[PATCH 00/15] bloom: changed-path Bloom filters v2

[PATCH 00/15] bloom: changed-path Bloom filters v2

[Taylor Blau]

This series combines the efforts in [1] and [2] to culminate in v2 of
the changed-path Bloom filter format (which uses a bona-fide murmur3
implementation) as well as an efficient upgrade path for repositories
with few non-ASCII paths.

The first seven patches are from [1], and are unchanged from that
version. Most of the remaining patches are from [2], and they have been
modified based on the review therein.

The final patch is new, and avoids leaking memory when the Bloom
subsystem is initialized both in the read- and read/write-cases.

Thanks to Jonathan and Peff who have both helped a great deal in putting
these patches together. And, as always, thanks in advance for your
review!

[1]: https://lore.kernel.org/git/cover.1684790529.git.jonathantanmy@google.com/
[2]: https://lore.kernel.org/git/cover.1691426160.git.me@ttaylorr.com/

Jonathan Tan (4):
  gitformat-commit-graph: describe version 2 of BDAT
  t4216: test changed path filters with high bit paths
  repo-settings: introduce commitgraph.changedPathsVersion
  commit-graph: new filter ver. that fixes murmur3

Taylor Blau (11):
  t/helper/test-read-graph.c: extract `dump_graph_info()`
  bloom.h: make `load_bloom_filter_from_graph()` public
  t/helper/test-read-graph: implement `bloom-filters` mode
  bloom: annotate filters with hash version
  bloom: prepare to discard incompatible Bloom filters
  t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
  commit-graph.c: unconditionally load Bloom filters
  commit-graph: drop unnecessary `graph_read_bloom_data_context`
  object.h: fix mis-aligned flag bits table
  commit-graph: reuse existing Bloom filters where possible
  bloom: introduce `deinit_bloom_filters()`

 Documentation/config/commitgraph.txt     |  26 ++-
 Documentation/gitformat-commit-graph.txt |   9 +-
 bloom.c                                  | 208 +++++++++++++++++++++--
 bloom.h                                  |  38 ++++-
 commit-graph.c                           |  36 +++-
 object.h                                 |   3 +-
 oss-fuzz/fuzz-commit-graph.c             |   2 +-
 repo-settings.c                          |   6 +-
 repository.h                             |   2 +-
 t/helper/test-bloom.c                    |   9 +-
 t/helper/test-read-graph.c               |  67 ++++++--
 t/t0095-bloom.sh                         |   8 +
 t/t4216-log-bloom.sh                     | 184 +++++++++++++++++++-
 13 files changed, 548 insertions(+), 50 deletions(-)

-- 
2.42.0.4.g52b49bb434

[PATCH 01/15] gitformat-commit-graph: describe version 2 of BDAT

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

The code change to Git to support version 2 will be done in subsequent
commits.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 Documentation/gitformat-commit-graph.txt | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/Documentation/gitformat-commit-graph.txt b/Documentation/gitformat-commit-graph.txt
index 31cad585e2..3e906e8030 100644
--- a/Documentation/gitformat-commit-graph.txt
+++ b/Documentation/gitformat-commit-graph.txt
@@ -142,13 +142,16 @@ All multi-byte numbers are in network byte order.
 
 ==== Bloom Filter Data (ID: {'B', 'D', 'A', 'T'}) [Optional]
     * It starts with header consisting of three unsigned 32-bit integers:
-      - Version of the hash algorithm being used. We currently only support
-	value 1 which corresponds to the 32-bit version of the murmur3 hash
+      - Version of the hash algorithm being used. We currently support
+	value 2 which corresponds to the 32-bit version of the murmur3 hash
 	implemented exactly as described in
 	https://en.wikipedia.org/wiki/MurmurHash#Algorithm and the double
 	hashing technique using seed values 0x293ae76f and 0x7e646e2 as
 	described in https://doi.org/10.1007/978-3-540-30494-4_26 "Bloom Filters
-	in Probabilistic Verification"
+	in Probabilistic Verification". Version 1 Bloom filters have a bug that appears
+	when char is signed and the repository has path names that have characters >=
+	0x80; Git supports reading and writing them, but this ability will be removed
+	in a future version of Git.
       - The number of times a path is hashed and hence the number of bit positions
 	      that cumulatively determine whether a file is present in the commit.
       - The minimum number of bits 'b' per entry in the Bloom filter. If the filter
-- 
2.42.0.4.g52b49bb434

[PATCH 02/15] t/helper/test-read-graph.c: extract `dump_graph_info()`

[Taylor Blau]

Prepare for the 'read-graph' test helper to perform other tasks besides
dumping high-level information about the commit-graph by extracting its
main routine into a separate function.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/helper/test-read-graph.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)

diff --git a/t/helper/test-read-graph.c b/t/helper/test-read-graph.c
index 8c7a83f578..3375392f6c 100644
--- a/t/helper/test-read-graph.c
+++ b/t/helper/test-read-graph.c
@@ -5,20 +5,8 @@
 #include "bloom.h"
 #include "setup.h"
 
-int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+static void dump_graph_info(struct commit_graph *graph)
 {
-	struct commit_graph *graph = NULL;
-	struct object_directory *odb;
-
-	setup_git_directory();
-	odb = the_repository->objects->odb;
-
-	prepare_repo_settings(the_repository);
-
-	graph = read_commit_graph_one(the_repository, odb);
-	if (!graph)
-		return 1;
-
 	printf("header: %08x %d %d %d %d\n",
 		ntohl(*(uint32_t*)graph->data),
 		*(unsigned char*)(graph->data + 4),
@@ -57,6 +45,23 @@ int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
 	if (graph->topo_levels)
 		printf(" topo_levels");
 	printf("\n");
+}
+
+int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+{
+	struct commit_graph *graph = NULL;
+	struct object_directory *odb;
+
+	setup_git_directory();
+	odb = the_repository->objects->odb;
+
+	prepare_repo_settings(the_repository);
+
+	graph = read_commit_graph_one(the_repository, odb);
+	if (!graph)
+		return 1;
+
+	dump_graph_info(graph);
 
 	UNLEAK(graph);
 
-- 
2.42.0.4.g52b49bb434

[PATCH 03/15] bloom.h: make `load_bloom_filter_from_graph()` public

[Taylor Blau]

Prepare for a future commit to use the load_bloom_filter_from_graph()
function directly to load specific Bloom filters out of the commit-graph
for manual inspection (to be used during tests).

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c | 6 +++---
 bloom.h | 5 +++++
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index aef6b5fea2..3e78cfe79d 100644
--- a/bloom.c
+++ b/bloom.c
@@ -29,9 +29,9 @@ static inline unsigned char get_bitmask(uint32_t pos)
 	return ((unsigned char)1) << (pos & (BITS_PER_WORD - 1));
 }
 
-static int load_bloom_filter_from_graph(struct commit_graph *g,
-					struct bloom_filter *filter,
-					uint32_t graph_pos)
+int load_bloom_filter_from_graph(struct commit_graph *g,
+				 struct bloom_filter *filter,
+				 uint32_t graph_pos)
 {
 	uint32_t lex_pos, start_index, end_index;
 
diff --git a/bloom.h b/bloom.h
index adde6dfe21..1e4f612d2c 100644
--- a/bloom.h
+++ b/bloom.h
@@ -3,6 +3,7 @@
 
 struct commit;
 struct repository;
+struct commit_graph;
 
 struct bloom_filter_settings {
 	/*
@@ -68,6 +69,10 @@ struct bloom_key {
 	uint32_t *hashes;
 };
 
+int load_bloom_filter_from_graph(struct commit_graph *g,
+				 struct bloom_filter *filter,
+				 uint32_t graph_pos);
+
 /*
  * Calculate the murmur3 32-bit hash value for the given data
  * using the given seed.
-- 
2.42.0.4.g52b49bb434

[PATCH 04/15] t/helper/test-read-graph: implement `bloom-filters` mode

[Taylor Blau]

Implement a mode of the "read-graph" test helper to dump out the
hexadecimal contents of the Bloom filter(s) contained in a commit-graph.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/helper/test-read-graph.c | 44 +++++++++++++++++++++++++++++++++-----
 1 file changed, 39 insertions(+), 5 deletions(-)

diff --git a/t/helper/test-read-graph.c b/t/helper/test-read-graph.c
index 3375392f6c..da9ac8584d 100644
--- a/t/helper/test-read-graph.c
+++ b/t/helper/test-read-graph.c
@@ -47,10 +47,32 @@ static void dump_graph_info(struct commit_graph *graph)
 	printf("\n");
 }
 
-int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+static void dump_graph_bloom_filters(struct commit_graph *graph)
+{
+	uint32_t i;
+
+	for (i = 0; i < graph->num_commits + graph->num_commits_in_base; i++) {
+		struct bloom_filter filter = { 0 };
+		size_t j;
+
+		if (load_bloom_filter_from_graph(graph, &filter, i) < 0) {
+			fprintf(stderr, "missing Bloom filter for graph "
+				"position %"PRIu32"\n", i);
+			continue;
+		}
+
+		for (j = 0; j < filter.len; j++)
+			printf("%02x", filter.data[j]);
+		if (filter.len)
+			printf("\n");
+	}
+}
+
+int cmd__read_graph(int argc, const char **argv)
 {
 	struct commit_graph *graph = NULL;
 	struct object_directory *odb;
+	int ret = 0;
 
 	setup_git_directory();
 	odb = the_repository->objects->odb;
@@ -58,12 +80,24 @@ int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
 	prepare_repo_settings(the_repository);
 
 	graph = read_commit_graph_one(the_repository, odb);
-	if (!graph)
-		return 1;
+	if (!graph) {
+		ret = 1;
+		goto done;
+	}
 
-	dump_graph_info(graph);
+	if (argc <= 1)
+		dump_graph_info(graph);
+	else if (!strcmp(argv[1], "bloom-filters"))
+		dump_graph_bloom_filters(graph);
+	else {
+		fprintf(stderr, "unknown sub-command: '%s'\n", argv[1]);
+		ret = 1;
+	}
 
+done:
 	UNLEAK(graph);
 
-	return 0;
+	return ret;
 }
+
+
-- 
2.42.0.4.g52b49bb434

[PATCH 05/15] t4216: test changed path filters with high bit paths

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

Subsequent commits will teach Git another version of changed path
filter that has different behavior with paths that contain at least
one character with its high bit set, so test the existing behavior as
a baseline.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/t4216-log-bloom.sh | 43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)

diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index fa9d32facf..2d4a3fefee 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -404,4 +404,47 @@ test_expect_success 'Bloom generation backfills empty commits' '
 	)
 '
 
+get_first_changed_path_filter () {
+	test-tool read-graph bloom-filters >filters.dat &&
+	head -n 1 filters.dat
+}
+
+# chosen to be the same under all Unicode normalization forms
+CENT=$(printf "\302\242")
+
+test_expect_success 'set up repo with high bit path, version 1 changed-path' '
+	git init highbit1 &&
+	test_commit -C highbit1 c1 "$CENT" &&
+	git -C highbit1 commit-graph write --reachable --changed-paths
+'
+
+test_expect_success 'setup check value of version 1 changed-path' '
+	(
+		cd highbit1 &&
+		echo "52a9" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
+# expect will not match actual if char is unsigned by default. Write the test
+# in this way, so that a user running this test script can still see if the two
+# files match. (It will appear as an ordinary success if they match, and a skip
+# if not.)
+if test_cmp highbit1/expect highbit1/actual
+then
+	test_set_prereq SIGNED_CHAR_BY_DEFAULT
+fi
+test_expect_success SIGNED_CHAR_BY_DEFAULT 'check value of version 1 changed-path' '
+	# Only the prereq matters for this test.
+	true
+'
+
+test_expect_success 'version 1 changed-path used when version 1 requested' '
+	(
+		cd highbit1 &&
+		test_bloom_filters_used "-- $CENT"
+	)
+'
+
 test_done
-- 
2.42.0.4.g52b49bb434

[PATCH 06/15] repo-settings: introduce commitgraph.changedPathsVersion

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

A subsequent commit will introduce another version of the changed-path
filter in the commit graph file. In order to control which version to
write (and read), a config variable is needed.

Therefore, introduce this config variable. For forwards compatibility,
teach Git to not read commit graphs when the config variable
is set to an unsupported version. Because we teach Git this,
commitgraph.readChangedPaths is now redundant, so deprecate it and
define its behavior in terms of the config variable we introduce.

This commit does not change the behavior of writing (Git writes changed
path filters when explicitly instructed regardless of any config
variable), but a subsequent commit will restrict Git such that it will
only write when commitgraph.changedPathsVersion is a recognized value.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 Documentation/config/commitgraph.txt | 23 ++++++++++++++++++++---
 commit-graph.c                       |  2 +-
 oss-fuzz/fuzz-commit-graph.c         |  2 +-
 repo-settings.c                      |  6 +++++-
 repository.h                         |  2 +-
 5 files changed, 28 insertions(+), 7 deletions(-)

diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
index 30604e4a4c..2dc9170622 100644
--- a/Documentation/config/commitgraph.txt
+++ b/Documentation/config/commitgraph.txt
@@ -9,6 +9,23 @@ commitGraph.maxNewFilters::
 	commit-graph write` (c.f., linkgit:git-commit-graph[1]).
 
 commitGraph.readChangedPaths::
-	If true, then git will use the changed-path Bloom filters in the
-	commit-graph file (if it exists, and they are present). Defaults to
-	true. See linkgit:git-commit-graph[1] for more information.
+	Deprecated. Equivalent to commitGraph.changedPathsVersion=-1 if true, and
+	commitGraph.changedPathsVersion=0 if false. (If commitGraph.changedPathVersion
+	is also set, commitGraph.changedPathsVersion takes precedence.)
+
+commitGraph.changedPathsVersion::
+	Specifies the version of the changed-path Bloom filters that Git will read and
+	write. May be -1, 0 or 1.
++
+Defaults to -1.
++
+If -1, Git will use the version of the changed-path Bloom filters in the
+repository, defaulting to 1 if there are none.
++
+If 0, Git will not read any Bloom filters, and will write version 1 Bloom
+filters when instructed to write.
++
+If 1, Git will only read version 1 Bloom filters, and will write version 1
+Bloom filters.
++
+See linkgit:git-commit-graph[1] for more information.
diff --git a/commit-graph.c b/commit-graph.c
index 0aa1640d15..da99f15fdf 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -401,7 +401,7 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 			graph->read_generation_data = 1;
 	}
 
-	if (s->commit_graph_read_changed_paths) {
+	if (s->commit_graph_changed_paths_version) {
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
diff --git a/oss-fuzz/fuzz-commit-graph.c b/oss-fuzz/fuzz-commit-graph.c
index 2992079dd9..325c0b991a 100644
--- a/oss-fuzz/fuzz-commit-graph.c
+++ b/oss-fuzz/fuzz-commit-graph.c
@@ -19,7 +19,7 @@ int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
 	 * possible.
 	 */
 	the_repository->settings.commit_graph_generation_version = 2;
-	the_repository->settings.commit_graph_read_changed_paths = 1;
+	the_repository->settings.commit_graph_changed_paths_version = 1;
 	g = parse_commit_graph(&the_repository->settings, (void *)data, size);
 	repo_clear(the_repository);
 	free_commit_graph(g);
diff --git a/repo-settings.c b/repo-settings.c
index 525f69c0c7..db8fe817f3 100644
--- a/repo-settings.c
+++ b/repo-settings.c
@@ -24,6 +24,7 @@ void prepare_repo_settings(struct repository *r)
 	int value;
 	const char *strval;
 	int manyfiles;
+	int read_changed_paths;
 
 	if (!r->gitdir)
 		BUG("Cannot add settings for uninitialized repository");
@@ -54,7 +55,10 @@ void prepare_repo_settings(struct repository *r)
 	/* Commit graph config or default, does not cascade (simple) */
 	repo_cfg_bool(r, "core.commitgraph", &r->settings.core_commit_graph, 1);
 	repo_cfg_int(r, "commitgraph.generationversion", &r->settings.commit_graph_generation_version, 2);
-	repo_cfg_bool(r, "commitgraph.readchangedpaths", &r->settings.commit_graph_read_changed_paths, 1);
+	repo_cfg_bool(r, "commitgraph.readchangedpaths", &read_changed_paths, 1);
+	repo_cfg_int(r, "commitgraph.changedpathsversion",
+		     &r->settings.commit_graph_changed_paths_version,
+		     read_changed_paths ? -1 : 0);
 	repo_cfg_bool(r, "gc.writecommitgraph", &r->settings.gc_write_commit_graph, 1);
 	repo_cfg_bool(r, "fetch.writecommitgraph", &r->settings.fetch_write_commit_graph, 0);
 
diff --git a/repository.h b/repository.h
index 5f18486f64..f71154e12c 100644
--- a/repository.h
+++ b/repository.h
@@ -29,7 +29,7 @@ struct repo_settings {
 
 	int core_commit_graph;
 	int commit_graph_generation_version;
-	int commit_graph_read_changed_paths;
+	int commit_graph_changed_paths_version;
 	int gc_write_commit_graph;
 	int fetch_write_commit_graph;
 	int command_requires_full_index;
-- 
2.42.0.4.g52b49bb434

[PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

The murmur3 implementation in bloom.c has a bug when converting series
of 4 bytes into network-order integers when char is signed (which is
controllable by a compiler option, and the default signedness of char is
platform-specific). When a string contains characters with the high bit
set, this bug causes results that, although internally consistent within
Git, does not accord with other implementations of murmur3 (thus,
the changed path filters wouldn't be readable by other off-the-shelf
implementatios of murmur3) and even with Git binaries that were compiled
with different signedness of char. This bug affects both how Git writes
changed path filters to disk and how Git interprets changed path filters
on disk.

Therefore, introduce a new version (2) of changed path filters that
corrects this problem. The existing version (1) is still supported and
is still the default, but users should migrate away from it as soon
as possible.

Because this bug only manifests with characters that have the high bit
set, it may be possible that some (or all) commits in a given repo would
have the same changed path filter both before and after this fix is
applied. However, in order to determine whether this is the case, the
changed paths would first have to be computed, at which point it is not
much more expensive to just compute a new changed path filter.

So this patch does not include any mechanism to "salvage" changed path
filters from repositories. There is also no "mixed" mode - for each
invocation of Git, reading and writing changed path filters are done
with the same version number; this version number may be explicitly
stated (typically if the user knows which version they need) or
automatically determined from the version of the existing changed path
filters in the repository.

There is a change in write_commit_graph(). graph_read_bloom_data()
makes it possible for chunk_bloom_data to be non-NULL but
bloom_filter_settings to be NULL, which causes a segfault later on. I
produced such a segfault while developing this patch, but couldn't find
a way to reproduce it neither after this complete patch (or before),
but in any case it seemed like a good thing to include that might help
future patch authors.

The value in t0095 was obtained from another murmur3 implementation
using the following Go source code:

  package main

  import "fmt"
  import "github.com/spaolacci/murmur3"

  func main() {
          fmt.Printf("%x\n", murmur3.Sum32([]byte("Hello world!")))
          fmt.Printf("%x\n", murmur3.Sum32([]byte{0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}))
  }

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 Documentation/config/commitgraph.txt |  5 +-
 bloom.c                              | 69 +++++++++++++++++++-
 bloom.h                              |  8 ++-
 commit-graph.c                       | 32 ++++++++--
 t/helper/test-bloom.c                |  9 ++-
 t/t0095-bloom.sh                     |  8 +++
 t/t4216-log-bloom.sh                 | 96 ++++++++++++++++++++++++++++
 7 files changed, 214 insertions(+), 13 deletions(-)

diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
index 2dc9170622..acc74a2f27 100644
--- a/Documentation/config/commitgraph.txt
+++ b/Documentation/config/commitgraph.txt
@@ -15,7 +15,7 @@ commitGraph.readChangedPaths::
 
 commitGraph.changedPathsVersion::
 	Specifies the version of the changed-path Bloom filters that Git will read and
-	write. May be -1, 0 or 1.
+	write. May be -1, 0, 1, or 2.
 +
 Defaults to -1.
 +
@@ -28,4 +28,7 @@ filters when instructed to write.
 If 1, Git will only read version 1 Bloom filters, and will write version 1
 Bloom filters.
 +
+If 2, Git will only read version 2 Bloom filters, and will write version 2
+Bloom filters.
++
 See linkgit:git-commit-graph[1] for more information.
diff --git a/bloom.c b/bloom.c
index 3e78cfe79d..ebef5cfd2f 100644
--- a/bloom.c
+++ b/bloom.c
@@ -66,7 +66,64 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
  * Not considered to be cryptographically secure.
  * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
  */
-uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
+uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len)
+{
+	const uint32_t c1 = 0xcc9e2d51;
+	const uint32_t c2 = 0x1b873593;
+	const uint32_t r1 = 15;
+	const uint32_t r2 = 13;
+	const uint32_t m = 5;
+	const uint32_t n = 0xe6546b64;
+	int i;
+	uint32_t k1 = 0;
+	const char *tail;
+
+	int len4 = len / sizeof(uint32_t);
+
+	uint32_t k;
+	for (i = 0; i < len4; i++) {
+		uint32_t byte1 = (uint32_t)(unsigned char)data[4*i];
+		uint32_t byte2 = ((uint32_t)(unsigned char)data[4*i + 1]) << 8;
+		uint32_t byte3 = ((uint32_t)(unsigned char)data[4*i + 2]) << 16;
+		uint32_t byte4 = ((uint32_t)(unsigned char)data[4*i + 3]) << 24;
+		k = byte1 | byte2 | byte3 | byte4;
+		k *= c1;
+		k = rotate_left(k, r1);
+		k *= c2;
+
+		seed ^= k;
+		seed = rotate_left(seed, r2) * m + n;
+	}
+
+	tail = (data + len4 * sizeof(uint32_t));
+
+	switch (len & (sizeof(uint32_t) - 1)) {
+	case 3:
+		k1 ^= ((uint32_t)(unsigned char)tail[2]) << 16;
+		/*-fallthrough*/
+	case 2:
+		k1 ^= ((uint32_t)(unsigned char)tail[1]) << 8;
+		/*-fallthrough*/
+	case 1:
+		k1 ^= ((uint32_t)(unsigned char)tail[0]) << 0;
+		k1 *= c1;
+		k1 = rotate_left(k1, r1);
+		k1 *= c2;
+		seed ^= k1;
+		break;
+	}
+
+	seed ^= (uint32_t)len;
+	seed ^= (seed >> 16);
+	seed *= 0x85ebca6b;
+	seed ^= (seed >> 13);
+	seed *= 0xc2b2ae35;
+	seed ^= (seed >> 16);
+
+	return seed;
+}
+
+static uint32_t murmur3_seeded_v1(uint32_t seed, const char *data, size_t len)
 {
 	const uint32_t c1 = 0xcc9e2d51;
 	const uint32_t c2 = 0x1b873593;
@@ -131,8 +188,14 @@ void fill_bloom_key(const char *data,
 	int i;
 	const uint32_t seed0 = 0x293ae76f;
 	const uint32_t seed1 = 0x7e646e2c;
-	const uint32_t hash0 = murmur3_seeded(seed0, data, len);
-	const uint32_t hash1 = murmur3_seeded(seed1, data, len);
+	uint32_t hash0, hash1;
+	if (settings->hash_version == 2) {
+		hash0 = murmur3_seeded_v2(seed0, data, len);
+		hash1 = murmur3_seeded_v2(seed1, data, len);
+	} else {
+		hash0 = murmur3_seeded_v1(seed0, data, len);
+		hash1 = murmur3_seeded_v1(seed1, data, len);
+	}
 
 	key->hashes = (uint32_t *)xcalloc(settings->num_hashes, sizeof(uint32_t));
 	for (i = 0; i < settings->num_hashes; i++)
diff --git a/bloom.h b/bloom.h
index 1e4f612d2c..138d57a86b 100644
--- a/bloom.h
+++ b/bloom.h
@@ -8,9 +8,11 @@ struct commit_graph;
 struct bloom_filter_settings {
 	/*
 	 * The version of the hashing technique being used.
-	 * We currently only support version = 1 which is
+	 * The newest version is 2, which is
 	 * the seeded murmur3 hashing technique implemented
-	 * in bloom.c.
+	 * in bloom.c. Bloom filters of version 1 were created
+	 * with prior versions of Git, which had a bug in the
+	 * implementation of the hash function.
 	 */
 	uint32_t hash_version;
 
@@ -80,7 +82,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
  * Not considered to be cryptographically secure.
  * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
  */
-uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len);
+uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len);
 
 void fill_bloom_key(const char *data,
 		    size_t len,
diff --git a/commit-graph.c b/commit-graph.c
index da99f15fdf..f7322c4fff 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -304,17 +304,26 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
 	return 0;
 }
 
+struct graph_read_bloom_data_context {
+	struct commit_graph *g;
+	int *commit_graph_changed_paths_version;
+};
+
 static int graph_read_bloom_data(const unsigned char *chunk_start,
 				  size_t chunk_size, void *data)
 {
-	struct commit_graph *g = data;
+	struct graph_read_bloom_data_context *c = data;
+	struct commit_graph *g = c->g;
 	uint32_t hash_version;
-	g->chunk_bloom_data = chunk_start;
 	hash_version = get_be32(chunk_start);
 
-	if (hash_version != 1)
+	if (*c->commit_graph_changed_paths_version == -1) {
+		*c->commit_graph_changed_paths_version = hash_version;
+	} else if (hash_version != *c->commit_graph_changed_paths_version) {
 		return 0;
+	}
 
+	g->chunk_bloom_data = chunk_start;
 	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
 	g->bloom_filter_settings->hash_version = hash_version;
 	g->bloom_filter_settings->num_hashes = get_be32(chunk_start + 4);
@@ -402,10 +411,14 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 	}
 
 	if (s->commit_graph_changed_paths_version) {
+		struct graph_read_bloom_data_context context = {
+			.g = graph,
+			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
+		};
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
-			   graph_read_bloom_data, graph);
+			   graph_read_bloom_data, &context);
 	}
 
 	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
@@ -2376,6 +2389,13 @@ int write_commit_graph(struct object_directory *odb,
 	}
 	if (!commit_graph_compatible(r))
 		return 0;
+	if (r->settings.commit_graph_changed_paths_version < -1
+	    || r->settings.commit_graph_changed_paths_version > 2) {
+		warning(_("attempting to write a commit-graph, but "
+			  "'commitgraph.changedPathsVersion' (%d) is not supported"),
+			r->settings.commit_graph_changed_paths_version);
+		return 0;
+	}
 
 	CALLOC_ARRAY(ctx, 1);
 	ctx->r = r;
@@ -2388,6 +2408,8 @@ int write_commit_graph(struct object_directory *odb,
 	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
 	ctx->num_generation_data_overflows = 0;
 
+	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
+		? 2 : 1;
 	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
 						      bloom_settings.bits_per_entry);
 	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
@@ -2417,7 +2439,7 @@ int write_commit_graph(struct object_directory *odb,
 		g = ctx->r->objects->commit_graph;
 
 		/* We have changed-paths already. Keep them in the next graph */
-		if (g && g->chunk_bloom_data) {
+		if (g && g->bloom_filter_settings) {
 			ctx->changed_paths = 1;
 			ctx->bloom_settings = g->bloom_filter_settings;
 		}
diff --git a/t/helper/test-bloom.c b/t/helper/test-bloom.c
index aabe31d724..3cbc0a5b50 100644
--- a/t/helper/test-bloom.c
+++ b/t/helper/test-bloom.c
@@ -50,6 +50,7 @@ static void get_bloom_filter_for_commit(const struct object_id *commit_oid)
 
 static const char *bloom_usage = "\n"
 "  test-tool bloom get_murmur3 <string>\n"
+"  test-tool bloom get_murmur3_seven_highbit\n"
 "  test-tool bloom generate_filter <string> [<string>...]\n"
 "  test-tool bloom get_filter_for_commit <commit-hex>\n";
 
@@ -64,7 +65,13 @@ int cmd__bloom(int argc, const char **argv)
 		uint32_t hashed;
 		if (argc < 3)
 			usage(bloom_usage);
-		hashed = murmur3_seeded(0, argv[2], strlen(argv[2]));
+		hashed = murmur3_seeded_v2(0, argv[2], strlen(argv[2]));
+		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
+	}
+
+	if (!strcmp(argv[1], "get_murmur3_seven_highbit")) {
+		uint32_t hashed;
+		hashed = murmur3_seeded_v2(0, "\x99\xaa\xbb\xcc\xdd\xee\xff", 7);
 		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
 	}
 
diff --git a/t/t0095-bloom.sh b/t/t0095-bloom.sh
index b567383eb8..c8d84ab606 100755
--- a/t/t0095-bloom.sh
+++ b/t/t0095-bloom.sh
@@ -29,6 +29,14 @@ test_expect_success 'compute unseeded murmur3 hash for test string 2' '
 	test_cmp expect actual
 '
 
+test_expect_success 'compute unseeded murmur3 hash for test string 3' '
+	cat >expect <<-\EOF &&
+	Murmur3 Hash with seed=0:0xa183ccfd
+	EOF
+	test-tool bloom get_murmur3_seven_highbit >actual &&
+	test_cmp expect actual
+'
+
 test_expect_success 'compute bloom key for empty string' '
 	cat >expect <<-\EOF &&
 	Hashes:0x5615800c|0x5b966560|0x61174ab4|0x66983008|0x6c19155c|0x7199fab0|0x771ae004|
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 2d4a3fefee..775e59d864 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -447,4 +447,100 @@ test_expect_success 'version 1 changed-path used when version 1 requested' '
 	)
 '
 
+test_expect_success 'version 1 changed-path not used when version 2 requested' '
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion 2 &&
+		test_bloom_filters_not_used "-- $CENT"
+	)
+'
+
+test_expect_success 'version 1 changed-path used when autodetect requested' '
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		test_bloom_filters_used "-- $CENT"
+	)
+'
+
+test_expect_success 'when writing another commit graph, preserve existing version 1 of changed-path' '
+	test_commit -C highbit1 c1double "$CENT$CENT" &&
+	git -C highbit1 commit-graph write --reachable --changed-paths &&
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		echo "options: bloom(1,10,7) read_generation_data" >expect &&
+		test-tool read-graph >full &&
+		grep options full >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'set up repo with high bit path, version 2 changed-path' '
+	git init highbit2 &&
+	git -C highbit2 config --add commitgraph.changedPathsVersion 2 &&
+	test_commit -C highbit2 c2 "$CENT" &&
+	git -C highbit2 commit-graph write --reachable --changed-paths
+'
+
+test_expect_success 'check value of version 2 changed-path' '
+	(
+		cd highbit2 &&
+		echo "c01f" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'version 2 changed-path used when version 2 requested' '
+	(
+		cd highbit2 &&
+		test_bloom_filters_used "-- $CENT"
+	)
+'
+
+test_expect_success 'version 2 changed-path not used when version 1 requested' '
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion 1 &&
+		test_bloom_filters_not_used "-- $CENT"
+	)
+'
+
+test_expect_success 'version 2 changed-path used when autodetect requested' '
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		test_bloom_filters_used "-- $CENT"
+	)
+'
+
+test_expect_success 'when writing another commit graph, preserve existing version 2 of changed-path' '
+	test_commit -C highbit2 c2double "$CENT$CENT" &&
+	git -C highbit2 commit-graph write --reachable --changed-paths &&
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		echo "options: bloom(2,10,7) read_generation_data" >expect &&
+		test-tool read-graph >full &&
+		grep options full >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
+	git init doublewrite &&
+	test_commit -C doublewrite c "$CENT" &&
+	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
+	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
+	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	(
+		cd doublewrite &&
+		echo "c01f" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
 test_done
-- 
2.42.0.4.g52b49bb434

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[SZEDER Gábor]

On Mon, Aug 21, 2023 at 05:44:15PM -0400, Taylor Blau wrote:
> From: Jonathan Tan <jonathantanmy@google.com>
> 
> The murmur3 implementation in bloom.c has a bug when converting series
> of 4 bytes into network-order integers when char is signed (which is
> controllable by a compiler option, and the default signedness of char is
> platform-specific). When a string contains characters with the high bit
> set, this bug causes results that, although internally consistent within
> Git, does not accord with other implementations of murmur3 (thus,
> the changed path filters wouldn't be readable by other off-the-shelf
> implementatios of murmur3) and even with Git binaries that were compiled
> with different signedness of char. This bug affects both how Git writes
> changed path filters to disk and how Git interprets changed path filters
> on disk.
> 
> Therefore, introduce a new version (2) of changed path filters that
> corrects this problem. The existing version (1) is still supported and
> is still the default, but users should migrate away from it as soon
> as possible.
> 
> Because this bug only manifests with characters that have the high bit
> set, it may be possible that some (or all) commits in a given repo would
> have the same changed path filter both before and after this fix is
> applied. However, in order to determine whether this is the case, the
> changed paths would first have to be computed, at which point it is not
> much more expensive to just compute a new changed path filter.
> 
> So this patch does not include any mechanism to "salvage" changed path
> filters from repositories. There is also no "mixed" mode - for each
> invocation of Git, reading and writing changed path filters are done
> with the same version number; this version number may be explicitly
> stated (typically if the user knows which version they need) or
> automatically determined from the version of the existing changed path
> filters in the repository.
> 
> There is a change in write_commit_graph(). graph_read_bloom_data()
> makes it possible for chunk_bloom_data to be non-NULL but
> bloom_filter_settings to be NULL, which causes a segfault later on. I
> produced such a segfault while developing this patch, but couldn't find
> a way to reproduce it neither after this complete patch (or before),
> but in any case it seemed like a good thing to include that might help
> future patch authors.
> 
> The value in t0095 was obtained from another murmur3 implementation
> using the following Go source code:
> 
>   package main
> 
>   import "fmt"
>   import "github.com/spaolacci/murmur3"
> 
>   func main() {
>           fmt.Printf("%x\n", murmur3.Sum32([]byte("Hello world!")))
>           fmt.Printf("%x\n", murmur3.Sum32([]byte{0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}))
>   }
> 
> Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> ---
>  Documentation/config/commitgraph.txt |  5 +-
>  bloom.c                              | 69 +++++++++++++++++++-
>  bloom.h                              |  8 ++-
>  commit-graph.c                       | 32 ++++++++--
>  t/helper/test-bloom.c                |  9 ++-
>  t/t0095-bloom.sh                     |  8 +++
>  t/t4216-log-bloom.sh                 | 96 ++++++++++++++++++++++++++++
>  7 files changed, 214 insertions(+), 13 deletions(-)
> 
> diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
> index 2dc9170622..acc74a2f27 100644
> --- a/Documentation/config/commitgraph.txt
> +++ b/Documentation/config/commitgraph.txt
> @@ -15,7 +15,7 @@ commitGraph.readChangedPaths::
>  
>  commitGraph.changedPathsVersion::
>  	Specifies the version of the changed-path Bloom filters that Git will read and
> -	write. May be -1, 0 or 1.
> +	write. May be -1, 0, 1, or 2.
>  +
>  Defaults to -1.
>  +
> @@ -28,4 +28,7 @@ filters when instructed to write.
>  If 1, Git will only read version 1 Bloom filters, and will write version 1
>  Bloom filters.
>  +
> +If 2, Git will only read version 2 Bloom filters, and will write version 2
> +Bloom filters.
> ++
>  See linkgit:git-commit-graph[1] for more information.
> diff --git a/bloom.c b/bloom.c
> index 3e78cfe79d..ebef5cfd2f 100644
> --- a/bloom.c
> +++ b/bloom.c
> @@ -66,7 +66,64 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
>   * Not considered to be cryptographically secure.
>   * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
>   */
> -uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
> +uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len)

Nit: MurmurHash3 implementations in C/C++ (apart from the sample
implementation on Wikipedia), and other hash functions taking data
pointer and buffer size parameters in general, have a 'const void
*data' parameter, not 'const char*'.

> +{
> +	const uint32_t c1 = 0xcc9e2d51;
> +	const uint32_t c2 = 0x1b873593;
> +	const uint32_t r1 = 15;
> +	const uint32_t r2 = 13;
> +	const uint32_t m = 5;
> +	const uint32_t n = 0xe6546b64;
> +	int i;
> +	uint32_t k1 = 0;
> +	const char *tail;
> +
> +	int len4 = len / sizeof(uint32_t);
> +
> +	uint32_t k;
> +	for (i = 0; i < len4; i++) {
> +		uint32_t byte1 = (uint32_t)(unsigned char)data[4*i];
> +		uint32_t byte2 = ((uint32_t)(unsigned char)data[4*i + 1]) << 8;
> +		uint32_t byte3 = ((uint32_t)(unsigned char)data[4*i + 2]) << 16;
> +		uint32_t byte4 = ((uint32_t)(unsigned char)data[4*i + 3]) << 24;
> +		k = byte1 | byte2 | byte3 | byte4;
> +		k *= c1;
> +		k = rotate_left(k, r1);
> +		k *= c2;
> +
> +		seed ^= k;
> +		seed = rotate_left(seed, r2) * m + n;
> +	}
> +
> +	tail = (data + len4 * sizeof(uint32_t));
> +
> +	switch (len & (sizeof(uint32_t) - 1)) {
> +	case 3:
> +		k1 ^= ((uint32_t)(unsigned char)tail[2]) << 16;
> +		/*-fallthrough*/
> +	case 2:
> +		k1 ^= ((uint32_t)(unsigned char)tail[1]) << 8;
> +		/*-fallthrough*/
> +	case 1:
> +		k1 ^= ((uint32_t)(unsigned char)tail[0]) << 0;
> +		k1 *= c1;
> +		k1 = rotate_left(k1, r1);
> +		k1 *= c2;
> +		seed ^= k1;
> +		break;
> +	}
> +
> +	seed ^= (uint32_t)len;
> +	seed ^= (seed >> 16);
> +	seed *= 0x85ebca6b;
> +	seed ^= (seed >> 13);
> +	seed *= 0xc2b2ae35;
> +	seed ^= (seed >> 16);
> +
> +	return seed;
> +}
> +
> +static uint32_t murmur3_seeded_v1(uint32_t seed, const char *data, size_t len)
>  {
>  	const uint32_t c1 = 0xcc9e2d51;
>  	const uint32_t c2 = 0x1b873593;
> @@ -131,8 +188,14 @@ void fill_bloom_key(const char *data,
>  	int i;
>  	const uint32_t seed0 = 0x293ae76f;
>  	const uint32_t seed1 = 0x7e646e2c;
> -	const uint32_t hash0 = murmur3_seeded(seed0, data, len);
> -	const uint32_t hash1 = murmur3_seeded(seed1, data, len);
> +	uint32_t hash0, hash1;
> +	if (settings->hash_version == 2) {
> +		hash0 = murmur3_seeded_v2(seed0, data, len);
> +		hash1 = murmur3_seeded_v2(seed1, data, len);
> +	} else {
> +		hash0 = murmur3_seeded_v1(seed0, data, len);
> +		hash1 = murmur3_seeded_v1(seed1, data, len);
> +	}
>  
>  	key->hashes = (uint32_t *)xcalloc(settings->num_hashes, sizeof(uint32_t));
>  	for (i = 0; i < settings->num_hashes; i++)
> diff --git a/bloom.h b/bloom.h
> index 1e4f612d2c..138d57a86b 100644
> --- a/bloom.h
> +++ b/bloom.h
> @@ -8,9 +8,11 @@ struct commit_graph;
>  struct bloom_filter_settings {
>  	/*
>  	 * The version of the hashing technique being used.
> -	 * We currently only support version = 1 which is
> +	 * The newest version is 2, which is
>  	 * the seeded murmur3 hashing technique implemented
> -	 * in bloom.c.
> +	 * in bloom.c. Bloom filters of version 1 were created
> +	 * with prior versions of Git, which had a bug in the
> +	 * implementation of the hash function.
>  	 */
>  	uint32_t hash_version;
>  
> @@ -80,7 +82,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
>   * Not considered to be cryptographically secure.
>   * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
>   */
> -uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len);
> +uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len);
>  
>  void fill_bloom_key(const char *data,
>  		    size_t len,
> diff --git a/commit-graph.c b/commit-graph.c
> index da99f15fdf..f7322c4fff 100644
> --- a/commit-graph.c
> +++ b/commit-graph.c
> @@ -304,17 +304,26 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
>  	return 0;
>  }
>  
> +struct graph_read_bloom_data_context {
> +	struct commit_graph *g;
> +	int *commit_graph_changed_paths_version;
> +};
> +
>  static int graph_read_bloom_data(const unsigned char *chunk_start,
>  				  size_t chunk_size, void *data)
>  {
> -	struct commit_graph *g = data;
> +	struct graph_read_bloom_data_context *c = data;
> +	struct commit_graph *g = c->g;
>  	uint32_t hash_version;
> -	g->chunk_bloom_data = chunk_start;
>  	hash_version = get_be32(chunk_start);
>  
> -	if (hash_version != 1)
> +	if (*c->commit_graph_changed_paths_version == -1) {
> +		*c->commit_graph_changed_paths_version = hash_version;
> +	} else if (hash_version != *c->commit_graph_changed_paths_version) {
>  		return 0;
> +	}
>  
> +	g->chunk_bloom_data = chunk_start;
>  	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
>  	g->bloom_filter_settings->hash_version = hash_version;
>  	g->bloom_filter_settings->num_hashes = get_be32(chunk_start + 4);
> @@ -402,10 +411,14 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
>  	}
>  
>  	if (s->commit_graph_changed_paths_version) {
> +		struct graph_read_bloom_data_context context = {
> +			.g = graph,
> +			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
> +		};
>  		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
>  			   &graph->chunk_bloom_indexes);
>  		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
> -			   graph_read_bloom_data, graph);
> +			   graph_read_bloom_data, &context);
>  	}
>  
>  	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
> @@ -2376,6 +2389,13 @@ int write_commit_graph(struct object_directory *odb,
>  	}
>  	if (!commit_graph_compatible(r))
>  		return 0;
> +	if (r->settings.commit_graph_changed_paths_version < -1
> +	    || r->settings.commit_graph_changed_paths_version > 2) {
> +		warning(_("attempting to write a commit-graph, but "
> +			  "'commitgraph.changedPathsVersion' (%d) is not supported"),
> +			r->settings.commit_graph_changed_paths_version);
> +		return 0;
> +	}
>  
>  	CALLOC_ARRAY(ctx, 1);
>  	ctx->r = r;
> @@ -2388,6 +2408,8 @@ int write_commit_graph(struct object_directory *odb,
>  	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
>  	ctx->num_generation_data_overflows = 0;
>  
> +	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
> +		? 2 : 1;
>  	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
>  						      bloom_settings.bits_per_entry);
>  	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
> @@ -2417,7 +2439,7 @@ int write_commit_graph(struct object_directory *odb,
>  		g = ctx->r->objects->commit_graph;
>  
>  		/* We have changed-paths already. Keep them in the next graph */
> -		if (g && g->chunk_bloom_data) {
> +		if (g && g->bloom_filter_settings) {
>  			ctx->changed_paths = 1;
>  			ctx->bloom_settings = g->bloom_filter_settings;
>  		}
> diff --git a/t/helper/test-bloom.c b/t/helper/test-bloom.c
> index aabe31d724..3cbc0a5b50 100644
> --- a/t/helper/test-bloom.c
> +++ b/t/helper/test-bloom.c
> @@ -50,6 +50,7 @@ static void get_bloom_filter_for_commit(const struct object_id *commit_oid)
>  
>  static const char *bloom_usage = "\n"
>  "  test-tool bloom get_murmur3 <string>\n"
> +"  test-tool bloom get_murmur3_seven_highbit\n"
>  "  test-tool bloom generate_filter <string> [<string>...]\n"
>  "  test-tool bloom get_filter_for_commit <commit-hex>\n";
>  
> @@ -64,7 +65,13 @@ int cmd__bloom(int argc, const char **argv)
>  		uint32_t hashed;
>  		if (argc < 3)
>  			usage(bloom_usage);
> -		hashed = murmur3_seeded(0, argv[2], strlen(argv[2]));
> +		hashed = murmur3_seeded_v2(0, argv[2], strlen(argv[2]));
> +		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
> +	}
> +
> +	if (!strcmp(argv[1], "get_murmur3_seven_highbit")) {
> +		uint32_t hashed;
> +		hashed = murmur3_seeded_v2(0, "\x99\xaa\xbb\xcc\xdd\xee\xff", 7);
>  		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
>  	}
>  
> diff --git a/t/t0095-bloom.sh b/t/t0095-bloom.sh
> index b567383eb8..c8d84ab606 100755
> --- a/t/t0095-bloom.sh
> +++ b/t/t0095-bloom.sh
> @@ -29,6 +29,14 @@ test_expect_success 'compute unseeded murmur3 hash for test string 2' '
>  	test_cmp expect actual
>  '
>  
> +test_expect_success 'compute unseeded murmur3 hash for test string 3' '
> +	cat >expect <<-\EOF &&
> +	Murmur3 Hash with seed=0:0xa183ccfd
> +	EOF
> +	test-tool bloom get_murmur3_seven_highbit >actual &&
> +	test_cmp expect actual
> +'
> +
>  test_expect_success 'compute bloom key for empty string' '
>  	cat >expect <<-\EOF &&
>  	Hashes:0x5615800c|0x5b966560|0x61174ab4|0x66983008|0x6c19155c|0x7199fab0|0x771ae004|
> diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
> index 2d4a3fefee..775e59d864 100755
> --- a/t/t4216-log-bloom.sh
> +++ b/t/t4216-log-bloom.sh
> @@ -447,4 +447,100 @@ test_expect_success 'version 1 changed-path used when version 1 requested' '
>  	)
>  '
>  
> +test_expect_success 'version 1 changed-path not used when version 2 requested' '
> +	(
> +		cd highbit1 &&
> +		git config --add commitgraph.changedPathsVersion 2 &&
> +		test_bloom_filters_not_used "-- $CENT"
> +	)
> +'
> +
> +test_expect_success 'version 1 changed-path used when autodetect requested' '
> +	(
> +		cd highbit1 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		test_bloom_filters_used "-- $CENT"
> +	)
> +'
> +
> +test_expect_success 'when writing another commit graph, preserve existing version 1 of changed-path' '
> +	test_commit -C highbit1 c1double "$CENT$CENT" &&
> +	git -C highbit1 commit-graph write --reachable --changed-paths &&

Nit: Since there is a subshell cd-ing into the 'highbit1' directory
anyway, it would look clearer to put these two commands into that
subshell as well.
This applies to some of the later test cases as well.

> +	(
> +		cd highbit1 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		echo "options: bloom(1,10,7) read_generation_data" >expect &&
> +		test-tool read-graph >full &&
> +		grep options full >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +
> +test_expect_success 'set up repo with high bit path, version 2 changed-path' '
> +	git init highbit2 &&
> +	git -C highbit2 config --add commitgraph.changedPathsVersion 2 &&
> +	test_commit -C highbit2 c2 "$CENT" &&
> +	git -C highbit2 commit-graph write --reachable --changed-paths
> +'
> +
> +test_expect_success 'check value of version 2 changed-path' '
> +	(
> +		cd highbit2 &&
> +		echo "c01f" >expect &&
> +		get_first_changed_path_filter >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +
> +test_expect_success 'version 2 changed-path used when version 2 requested' '
> +	(
> +		cd highbit2 &&
> +		test_bloom_filters_used "-- $CENT"

This test_bloom_filter_used helper runs two pathspec-limited 'git log'
invocations, one with disabled and the other with enabled
commit-graph, and thus with disabled and enabled modified path Bloom
filters, and compares their output.

One of the flaws of the current modified path Bloom filters
implementation is that it doesn't check Bloom filters for root
commits.

In several of the above test cases test_bloom_filters_used is invoked
in a repository with only a root commit, so they don't check that
the output is the same with and without Bloom filters.

> +	)
> +'
> +
> +test_expect_success 'version 2 changed-path not used when version 1 requested' '
> +	(
> +		cd highbit2 &&
> +		git config --add commitgraph.changedPathsVersion 1 &&
> +		test_bloom_filters_not_used "-- $CENT"
> +	)
> +'
> +
> +test_expect_success 'version 2 changed-path used when autodetect requested' '
> +	(
> +		cd highbit2 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		test_bloom_filters_used "-- $CENT"
> +	)
> +'
> +
> +test_expect_success 'when writing another commit graph, preserve existing version 2 of changed-path' '
> +	test_commit -C highbit2 c2double "$CENT$CENT" &&
> +	git -C highbit2 commit-graph write --reachable --changed-paths &&
> +	(
> +		cd highbit2 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		echo "options: bloom(2,10,7) read_generation_data" >expect &&
> +		test-tool read-graph >full &&
> +		grep options full >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +
> +test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
> +	git init doublewrite &&
> +	test_commit -C doublewrite c "$CENT" &&
> +	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
> +	git -C doublewrite commit-graph write --reachable --changed-paths &&
> +	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
> +	git -C doublewrite commit-graph write --reachable --changed-paths &&
> +	(
> +		cd doublewrite &&
> +		echo "c01f" >expect &&
> +		get_first_changed_path_filter >actual &&
> +		test_cmp expect actual
> +	)
> +'

The string "split" occurs twice in this patch series, but only in
patch hunk contexts, and it doesn't occur at all in the previous long
thread about the original patch series.

Unfortunately, split commit-graphs weren't really considered in the
design of the modified path Bloom filters feature, and layers with
different Bloom filter settings weren't considered at all.  I've
reported it back then, but the fixes so far were incomplete, and e.g.
the test cases shown in

  https://public-inbox.org/git/20201015132147.GB24954@szeder.dev/

still fail.

Since the interaction of different versions and split commit-graphs
was neither mentioned in any of the commit messages nor discussed
during the previous rounds, and there isn't any test case excercising
it, and since the Bloom filter version information is stored in the
same 'g->bloom_filter_settings' structure as the number of hashes, I'm
afraid (though haven't actually checked) that handling commit-graph
layers with different Bloom filter versions is prone to the same
issues as well.

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Jonathan Tan]

SZEDER Gábor <szeder.dev@gmail.com> writes:
> > @@ -66,7 +66,64 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
> >   * Not considered to be cryptographically secure.
> >   * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
> >   */
> > -uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
> > +uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len)
> 
> Nit: MurmurHash3 implementations in C/C++ (apart from the sample
> implementation on Wikipedia), and other hash functions taking data
> pointer and buffer size parameters in general, have a 'const void
> *data' parameter, not 'const char*'.

I think either works, so I'll stick with what the existing code uses.
(Well, we probably should have used "unsigned char" in the first place.)

> > +test_expect_success 'when writing another commit graph, preserve existing version 1 of changed-path' '
> > +	test_commit -C highbit1 c1double "$CENT$CENT" &&
> > +	git -C highbit1 commit-graph write --reachable --changed-paths &&
> 
> Nit: Since there is a subshell cd-ing into the 'highbit1' directory
> anyway, it would look clearer to put these two commands into that
> subshell as well.
> This applies to some of the later test cases as well.

Makes sense, but this patch series has already been reviewed a few times
so I don't think it's worth making this change at this point in the
review process for a small increase in readability.

> > +test_expect_success 'version 2 changed-path used when version 2 requested' '
> > +	(
> > +		cd highbit2 &&
> > +		test_bloom_filters_used "-- $CENT"
> 
> This test_bloom_filter_used helper runs two pathspec-limited 'git log'
> invocations, one with disabled and the other with enabled
> commit-graph, and thus with disabled and enabled modified path Bloom
> filters, and compares their output.
> 
> One of the flaws of the current modified path Bloom filters
> implementation is that it doesn't check Bloom filters for root
> commits.
> 
> In several of the above test cases test_bloom_filters_used is invoked
> in a repository with only a root commit, so they don't check that
> the output is the same with and without Bloom filters.

Ah...you are right. Indeed, when I flip one of the tests from
test_bloom_filters_used to _not_, the test still passes. I'll change
the tests.

> The string "split" occurs twice in this patch series, but only in
> patch hunk contexts, and it doesn't occur at all in the previous long
> thread about the original patch series.
> 
> Unfortunately, split commit-graphs weren't really considered in the
> design of the modified path Bloom filters feature, and layers with
> different Bloom filter settings weren't considered at all.  I've
> reported it back then, but the fixes so far were incomplete, and e.g.
> the test cases shown in
> 
>   https://public-inbox.org/git/20201015132147.GB24954@szeder.dev/
> 
> still fail.
> 
> Since the interaction of different versions and split commit-graphs
> was neither mentioned in any of the commit messages nor discussed
> during the previous rounds, and there isn't any test case excercising
> it, and since the Bloom filter version information is stored in the
> same 'g->bloom_filter_settings' structure as the number of hashes, I'm
> afraid (though haven't actually checked) that handling commit-graph
> layers with different Bloom filter versions is prone to the same
> issues as well.

My original design (up to patch 7 in this patch set) defends against
this by taking the very first version detected and rejecting every
other version, and Taylor's subsequent design reads every version, but
annotates filters with its version. So I think we're covered.

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[SZEDER Gábor]

On Tue, Aug 29, 2023 at 09:31:23AM -0700, Jonathan Tan wrote:
> SZEDER Gábor <szeder.dev@gmail.com> writes:
 
> > > +test_expect_success 'version 2 changed-path used when version 2 requested' '
> > > +	(
> > > +		cd highbit2 &&
> > > +		test_bloom_filters_used "-- $CENT"
> > 
> > This test_bloom_filter_used helper runs two pathspec-limited 'git log'
> > invocations, one with disabled and the other with enabled
> > commit-graph, and thus with disabled and enabled modified path Bloom
> > filters, and compares their output.
> > 
> > One of the flaws of the current modified path Bloom filters
> > implementation is that it doesn't check Bloom filters for root
> > commits.
> > 
> > In several of the above test cases test_bloom_filters_used is invoked
> > in a repository with only a root commit, so they don't check that
> > the output is the same with and without Bloom filters.
> 
> Ah...you are right. Indeed, when I flip one of the tests from
> test_bloom_filters_used to _not_, the test still passes. I'll change
> the tests.

I'd prefer to leave the test cases unchanged, and make the revision
walking machinery look at Bloom filters even for root commits, because
this is an annoying and recurring testing issue.  I remember it
annoyed me back then, when I wanted to reproduce a couple of bugs that
I knew were there, but my initial test cases didn't fail because the
Bloom filter of the root commit was ignored; Derrick overlooked this
in b16a8277644, you overlooked it now, and none of the reviewers then
and now caught it, either.

> > The string "split" occurs twice in this patch series, but only in
> > patch hunk contexts, and it doesn't occur at all in the previous long
> > thread about the original patch series.
> > 
> > Unfortunately, split commit-graphs weren't really considered in the
> > design of the modified path Bloom filters feature, and layers with
> > different Bloom filter settings weren't considered at all.  I've
> > reported it back then, but the fixes so far were incomplete, and e.g.
> > the test cases shown in
> > 
> >   https://public-inbox.org/git/20201015132147.GB24954@szeder.dev/
> > 
> > still fail.
> > 
> > Since the interaction of different versions and split commit-graphs
> > was neither mentioned in any of the commit messages nor discussed
> > during the previous rounds, and there isn't any test case excercising
> > it, and since the Bloom filter version information is stored in the
> > same 'g->bloom_filter_settings' structure as the number of hashes, I'm
> > afraid (though haven't actually checked) that handling commit-graph
> > layers with different Bloom filter versions is prone to the same
> > issues as well.
> 
> My original design (up to patch 7 in this patch set) defends against
> this by taking the very first version detected and rejecting every
> other version, and Taylor's subsequent design reads every version, but
> annotates filters with its version. So I think we're covered.

In the meantime I adapted the test cases from the above linked message
to write commit-graph layers with different Bloom filter versions, and
it does fail, because commits from the bottom commit-graph layer are
omitted from the revision walk's output.  And the test case doesn't
even need a middle layer without modified path Bloom filters to "hide"
the different version in the bottom layer.  Merging the layers seems
to work, though.

Besides fixing this issue, I think that the interaction of different
Bloom filter versions and split commit-graphs needs to be thoroughly
covered with test cases and discussed in the commit messages before
this series could be considered good for merging.


diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 48f8109a66..55f67e5110 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -586,4 +586,40 @@ test_expect_success 'when writing commit graph, reuse changed-path of another ve
 	test_filter_upgraded 1 trace2.txt
 '
 
+test_expect_success 'split commit graph vs changed paths breakage - setup' '
+	git init split-breakage &&
+	(
+		cd split-breakage &&
+		git commit --allow-empty -m "Bloom filters are written but still ignored for root commits :(" &&
+		for i in 1 2 3
+		do
+			echo $i >$CENT &&
+			git add $CENT &&
+			git commit -m "$i" || return 1
+		done &&
+		git log --oneline -- $CENT >expect
+	)
+'
+
+test_expect_failure 'split commit graph vs changed paths breakage - split' '
+	(
+		cd split-breakage &&
+
+		# Compute v1 Bloom filters for the commits at the bottom.
+		git rev-parse HEAD^ | git commit-graph write --stdin-commits --changed-paths --split &&
+		# Compute v2 Bloom filters for the rest of the commits at the top.
+		git rev-parse HEAD | git -c commitgraph.changedPathsVersion=2 commit-graph write --stdin-commits --changed-paths --split=no-merge &&
+
+		# Just to make sure that there are as many graph layers as I
+		# think there should be.
+		test_line_count = 2 .git/objects/info/commit-graphs/commit-graph-chain &&
+
+		# This checks Bloom filters using version information in the
+		# top layer, thus misses commits modifying the file in the
+		# bottom commit-graph layer.
+		git log --oneline -- $CENT >actual &&
+		test_cmp expect actual
+	)
+'
+
 test_done

It fails with:

  + cd split-breakage
  + git rev-parse HEAD^
  + git commit-graph write --stdin-commits --changed-paths --split
  + git rev-parse HEAD
  + git -c commitgraph.changedPathsVersion=2 commit-graph write --stdin-commits --changed-paths --split=no-merge
  + test_line_count = 2 .git/objects/info/commit-graphs/commit-graph-chain
  + git log --oneline -- ¢
  + test_cmp expect actual
  --- expect      2023-08-30 19:07:59.882066827 +0000
  +++ actual      2023-08-30 19:07:59.894067148 +0000
  @@ -1,3 +1 @@
   1db2248 3
  -cfcc97f 2
  -bd9c2c8 1
  error: last command exited with $?=1

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Jonathan Tan]

SZEDER Gábor <szeder.dev@gmail.com> writes:
> I'd prefer to leave the test cases unchanged, and make the revision
> walking machinery look at Bloom filters even for root commits, because
> this is an annoying and recurring testing issue.  I remember it
> annoyed me back then, when I wanted to reproduce a couple of bugs that
> I knew were there, but my initial test cases didn't fail because the
> Bloom filter of the root commit was ignored; Derrick overlooked this
> in b16a8277644, you overlooked it now, and none of the reviewers then
> and now caught it, either.

I agree that making the revwalk look at Bloom filters of root commits
is an urgent matter for the reasons you describe (it's something
easily missed that slows down future development and/or makes future
development error-prone), but so is moving to a correct murmur3
implementation, I think, and one shouldn't stop the other. There could
be an argument that because the revwalk doesn't look at root commit
Bloom filters, any development on a new Bloom filter hash version is
suspect, but I don't think that has to be completely true.

> > My original design (up to patch 7 in this patch set) defends against
> > this by taking the very first version detected and rejecting every
> > other version, and Taylor's subsequent design reads every version, but
> > annotates filters with its version. So I think we're covered.
> 
> In the meantime I adapted the test cases from the above linked message
> to write commit-graph layers with different Bloom filter versions, and
> it does fail, because commits from the bottom commit-graph layer are
> omitted from the revision walk's output.  And the test case doesn't
> even need a middle layer without modified path Bloom filters to "hide"
> the different version in the bottom layer.  Merging the layers seems
> to work, though.

For what it's worth, your test case below (with test_expect_success
instead of test_expect_failure) passes with my original design. With the
full patch set, it does fail, but for what it's worth, I did spot this,
providing an incomplete solution [1] and then a complete one [2]. Your
test case passes if I also include the following:

  diff --git a/bloom.c b/bloom.c
  index ff131893cd..1bafd62a4e 100644
  --- a/bloom.c
  +++ b/bloom.c
  @@ -344,6 +344,10 @@ struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
   
          prepare_repo_settings(r);
          hash_version = r->settings.commit_graph_changed_paths_version;
  +       if (hash_version == -1) {
  +               struct bloom_filter_settings *s = get_bloom_filter_settings(r);
  +               hash_version = (s && s->hash_version == 2) ? 2 : 1;
  +       }
   
          if (!(hash_version == -1 || hash_version == filter->version))
                  return NULL; /* unusable filter */

[1] https://lore.kernel.org/git/20230824222051.2320003-1-jonathantanmy@google.com/
[2] https://lore.kernel.org/git/20230829220432.558674-1-jonathantanmy@google.com/

> Besides fixing this issue, I think that the interaction of different
> Bloom filter versions and split commit-graphs needs to be thoroughly
> covered with test cases and discussed in the commit messages before
> this series could be considered good for merging.

Regarding commit messages, I can see that in "commit-graph: new filter
ver. that fixes murmur3", I could add "(the first version read if there
are multiple versions of changed path filters in the repository)" after
"automatically determined from the version of the existing changed path
filters in the repository". Taylor's patches already inherently cover
multiple versions, I think, since Bloom filters are annotated with their
versions, individually.

Regarding tests, yes, we could add the one you provided.

If you (or anyone else) can spot anything else that should be added,
please let us know.

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Taylor Blau]

On Fri, Sep 01, 2023 at 01:56:15PM -0700, Jonathan Tan wrote:
> > > My original design (up to patch 7 in this patch set) defends against
> > > this by taking the very first version detected and rejecting every
> > > other version, and Taylor's subsequent design reads every version, but
> > > annotates filters with its version. So I think we're covered.
> >
> > In the meantime I adapted the test cases from the above linked message
> > to write commit-graph layers with different Bloom filter versions, and
> > it does fail, because commits from the bottom commit-graph layer are
> > omitted from the revision walk's output.  And the test case doesn't
> > even need a middle layer without modified path Bloom filters to "hide"
> > the different version in the bottom layer.  Merging the layers seems
> > to work, though.
>
> For what it's worth, your test case below (with test_expect_success
> instead of test_expect_failure) passes with my original design. With the
> full patch set, it does fail, but for what it's worth, I did spot this,
> providing an incomplete solution [1] and then a complete one [2]. Your
> test case passes if I also include the following:
>
>   diff --git a/bloom.c b/bloom.c
>   index ff131893cd..1bafd62a4e 100644
>   --- a/bloom.c
>   +++ b/bloom.c
>   @@ -344,6 +344,10 @@ struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
>
>           prepare_repo_settings(r);
>           hash_version = r->settings.commit_graph_changed_paths_version;
>   +       if (hash_version == -1) {
>   +               struct bloom_filter_settings *s = get_bloom_filter_settings(r);
>   +               hash_version = (s && s->hash_version == 2) ? 2 : 1;
>   +       }
>
>           if (!(hash_version == -1 || hash_version == filter->version))
>                   return NULL; /* unusable filter */
>
> [1] https://lore.kernel.org/git/20230824222051.2320003-1-jonathantanmy@google.com/
> [2] https://lore.kernel.org/git/20230829220432.558674-1-jonathantanmy@google.com/

Hmm. I am confused -- are you saying that this series breaks existing
functionality, or merely does not patch an existing breakage? I *think*
that it's the latter, since this test case fails identically on master,
but I am not sure.

If my understanding is correct, I think that patching this would involve
annotating each Bloom filter with a pointer to the bloom_filter_settings
it was written with, and then using those settings when querying it.

Thanks,
Taylor

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[SZEDER Gábor]

On Mon, Sep 25, 2023 at 07:03:18PM -0400, Taylor Blau wrote:
> On Fri, Sep 01, 2023 at 01:56:15PM -0700, Jonathan Tan wrote:
> > > > My original design (up to patch 7 in this patch set) defends against
> > > > this by taking the very first version detected and rejecting every
> > > > other version, and Taylor's subsequent design reads every version, but
> > > > annotates filters with its version. So I think we're covered.
> > >
> > > In the meantime I adapted the test cases from the above linked message
> > > to write commit-graph layers with different Bloom filter versions, and
> > > it does fail, because commits from the bottom commit-graph layer are
> > > omitted from the revision walk's output.  And the test case doesn't
> > > even need a middle layer without modified path Bloom filters to "hide"
> > > the different version in the bottom layer.  Merging the layers seems
> > > to work, though.
> >
> > For what it's worth, your test case below (with test_expect_success
> > instead of test_expect_failure) passes with my original design. With the
> > full patch set, it does fail, but for what it's worth, I did spot this,
> > providing an incomplete solution [1] and then a complete one [2]. Your
> > test case passes if I also include the following:
> >
> >   diff --git a/bloom.c b/bloom.c
> >   index ff131893cd..1bafd62a4e 100644
> >   --- a/bloom.c
> >   +++ b/bloom.c
> >   @@ -344,6 +344,10 @@ struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
> >
> >           prepare_repo_settings(r);
> >           hash_version = r->settings.commit_graph_changed_paths_version;
> >   +       if (hash_version == -1) {
> >   +               struct bloom_filter_settings *s = get_bloom_filter_settings(r);
> >   +               hash_version = (s && s->hash_version == 2) ? 2 : 1;
> >   +       }
> >
> >           if (!(hash_version == -1 || hash_version == filter->version))
> >                   return NULL; /* unusable filter */
> >
> > [1] https://lore.kernel.org/git/20230824222051.2320003-1-jonathantanmy@google.com/
> > [2] https://lore.kernel.org/git/20230829220432.558674-1-jonathantanmy@google.com/
> 
> Hmm. I am confused -- are you saying that this series breaks existing
> functionality, or merely does not patch an existing breakage? I *think*
> that it's the latter,

It's neither: the new functionality added in this series is broken.

> since this test case fails identically on master,
> but I am not sure.

Not sure what test you are referring to.  My test demonstrating the
breakage succeeds when adaped to master, because master doesn't
understand the commitgraph.changedPathsVersion=2 setting, and keeps
writing v1 Bloom filter chunks instead, so all commit-graphs layers
contain the same version.

> If my understanding is correct, I think that patching this would involve
> annotating each Bloom filter with a pointer to the bloom_filter_settings
> it was written with, and then using those settings when querying it.

In general we are better off when we don't write Bloom filter chunks
with different settings in the first place.

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Taylor Blau]

On Sun, Oct 08, 2023 at 04:35:23PM +0200, SZEDER Gábor wrote:
> > Hmm. I am confused -- are you saying that this series breaks existing
> > functionality, or merely does not patch an existing breakage? I *think*
> > that it's the latter,
>
> It's neither: the new functionality added in this series is broken.
>
> > since this test case fails identically on master,
> > but I am not sure.
>
> Not sure what test you are referring to.  My test demonstrating the
> breakage succeeds when adaped to master, because master doesn't
> understand the commitgraph.changedPathsVersion=2 setting, and keeps
> writing v1 Bloom filter chunks instead, so all commit-graphs layers
> contain the same version.

I was referring to the test you sent back in:

    https://public-inbox.org/git/20201015132147.GB24954@szeder.dev/

but I think that I should have been looking at the one you sent more
recently in:

    https://lore.kernel.org/git/20230830200218.GA5147@szeder.dev/

Thanks,
Taylor

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Taylor Blau]

On Mon, Oct 09, 2023 at 02:17:54PM -0400, Taylor Blau wrote:
> On Sun, Oct 08, 2023 at 04:35:23PM +0200, SZEDER Gábor wrote:
> > > Hmm. I am confused -- are you saying that this series breaks existing
> > > functionality, or merely does not patch an existing breakage? I *think*
> > > that it's the latter,
> >
> > It's neither: the new functionality added in this series is broken.
> >
> > > since this test case fails identically on master,
> > > but I am not sure.
> >
> > Not sure what test you are referring to.  My test demonstrating the
> > breakage succeeds when adaped to master, because master doesn't
> > understand the commitgraph.changedPathsVersion=2 setting, and keeps
> > writing v1 Bloom filter chunks instead, so all commit-graphs layers
> > contain the same version.
>
> I was referring to the test you sent back in:
>
>     https://public-inbox.org/git/20201015132147.GB24954@szeder.dev/
>
> but I think that I should have been looking at the one you sent more
> recently in:
>
>     https://lore.kernel.org/git/20230830200218.GA5147@szeder.dev/

OK, I think that I now have my head wrapped around what you're saying.

However, I am not entirely sure I agree with you that this is a "new"
issue. At least in the sense that Git (on 'master') does not currently
know how to deal with Bloom filters that have different settings across
commit-graph layers.

IOW, you could produce this problem today using the test you wrote in
<20201015132147.GB24954@szeder.dev> using different values of the
GIT_BLOOM_SETTINGS environment variables as a proxy for different values
of the commitGraph.changedPathsVersion configuration variable introduced
in this series.

So I think that this series makes it easier to fall into that trap, but
the trap itself is not new. I think a reasonable stopgap (which IIUC you
have suggested earlier) is to prevent writing a new commit-graph layer
with a different hash version than the previous layer.

How does that sound?

Thanks,
Taylor

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Junio C Hamano]

Taylor Blau <me@ttaylorr.com> writes:

> However, I am not entirely sure I agree with you that this is a "new"
> issue. At least in the sense that Git (on 'master') does not currently
> know how to deal with Bloom filters that have different settings across
> commit-graph layers.
>
> IOW, you could produce this problem today using the test you wrote in
> <20201015132147.GB24954@szeder.dev> using different values of the
> GIT_BLOOM_SETTINGS environment variables as a proxy for different values
> of the commitGraph.changedPathsVersion configuration variable introduced
> in this series.
>
> So I think that this series makes it easier to fall into that trap, but
> the trap itself is not new. I think a reasonable stopgap (which IIUC you
> have suggested earlier) is to prevent writing a new commit-graph layer
> with a different hash version than the previous layer.

What we probably want more urgently than that stopgap is to perhaps
teach the code pretend as if commit-graph did not exist when we
detect multiple layers use different hash versions (or perhaps only
use the base layer and ignore the rest as an anti-pessimization), to
protect correctness first, no?

Re: [PATCH 07/15] commit-graph: new filter ver. that fixes murmur3

[Taylor Blau]

On Mon, Oct 09, 2023 at 12:52:13PM -0700, Junio C Hamano wrote:
> Taylor Blau <me@ttaylorr.com> writes:
>
> > However, I am not entirely sure I agree with you that this is a "new"
> > issue. At least in the sense that Git (on 'master') does not currently
> > know how to deal with Bloom filters that have different settings across
> > commit-graph layers.
> >
> > IOW, you could produce this problem today using the test you wrote in
> > <20201015132147.GB24954@szeder.dev> using different values of the
> > GIT_BLOOM_SETTINGS environment variables as a proxy for different values
> > of the commitGraph.changedPathsVersion configuration variable introduced
> > in this series.
> >
> > So I think that this series makes it easier to fall into that trap, but
> > the trap itself is not new. I think a reasonable stopgap (which IIUC you
> > have suggested earlier) is to prevent writing a new commit-graph layer
> > with a different hash version than the previous layer.
>
> What we probably want more urgently than that stopgap is to perhaps
> teach the code pretend as if commit-graph did not exist when we
> detect multiple layers use different hash versions (or perhaps only
> use the base layer and ignore the rest as an anti-pessimization), to
> protect correctness first, no?

Very good suggestion, thanks.

Thanks,
Taylor

[PATCH 08/15] bloom: annotate filters with hash version

[Taylor Blau]

In subsequent commits, we will want to load existing Bloom filters out
of a commit-graph, even when the hash version they were computed with
does not match the value of `commitGraph.changedPathVersion`.

In order to differentiate between the two, add a "version" field to each
Bloom filter.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c | 11 ++++++++---
 bloom.h |  1 +
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index ebef5cfd2f..9b6a30f6f6 100644
--- a/bloom.c
+++ b/bloom.c
@@ -55,6 +55,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
 	filter->data = (unsigned char *)(g->chunk_bloom_data +
 					sizeof(unsigned char) * start_index +
 					BLOOMDATA_CHUNK_HEADER_SIZE);
+	filter->version = g->bloom_filter_settings->hash_version;
 
 	return 1;
 }
@@ -240,11 +241,13 @@ static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 	return strcmp(e1->path, e2->path);
 }
 
-static void init_truncated_large_filter(struct bloom_filter *filter)
+static void init_truncated_large_filter(struct bloom_filter *filter,
+					int version)
 {
 	filter->data = xmalloc(1);
 	filter->data[0] = 0xFF;
 	filter->len = 1;
+	filter->version = version;
 }
 
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
@@ -329,13 +332,15 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 		}
 
 		if (hashmap_get_size(&pathmap) > settings->max_changed_paths) {
-			init_truncated_large_filter(filter);
+			init_truncated_large_filter(filter,
+						    settings->hash_version);
 			if (computed)
 				*computed |= BLOOM_TRUNC_LARGE;
 			goto cleanup;
 		}
 
 		filter->len = (hashmap_get_size(&pathmap) * settings->bits_per_entry + BITS_PER_WORD - 1) / BITS_PER_WORD;
+		filter->version = settings->hash_version;
 		if (!filter->len) {
 			if (computed)
 				*computed |= BLOOM_TRUNC_EMPTY;
@@ -355,7 +360,7 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 	} else {
 		for (i = 0; i < diff_queued_diff.nr; i++)
 			diff_free_filepair(diff_queued_diff.queue[i]);
-		init_truncated_large_filter(filter);
+		init_truncated_large_filter(filter, settings->hash_version);
 
 		if (computed)
 			*computed |= BLOOM_TRUNC_LARGE;
diff --git a/bloom.h b/bloom.h
index 138d57a86b..330a140520 100644
--- a/bloom.h
+++ b/bloom.h
@@ -55,6 +55,7 @@ struct bloom_filter_settings {
 struct bloom_filter {
 	unsigned char *data;
 	size_t len;
+	int version;
 };
 
 /*
-- 
2.42.0.4.g52b49bb434

[PATCH 09/15] bloom: prepare to discard incompatible Bloom filters

[Taylor Blau]

Callers use the inline `get_bloom_filter()` implementation as a thin
wrapper around `get_or_compute_bloom_filter()`. The former calls the
latter with a value of "0" for `compute_if_not_present`, making
`get_bloom_filter()` the default read-only path for fetching an existing
Bloom filter.

Callers expect the value returned from `get_bloom_filter()` is usable,
that is that it's compatible with the configured value corresponding to
`commitGraph.changedPathsVersion`.

This is OK, since the commit-graph machinery only initializes its BDAT
chunk (thereby enabling it to service Bloom filter queries) when the
Bloom filter hash_version is compatible with our settings. So any value
returned by `get_bloom_filter()` is trivially useable.

However, subsequent commits will load the BDAT chunk even when the Bloom
filters are built with incompatible hash versions. Prepare to handle
this by teaching `get_bloom_filter()` to discard filters that are
incompatible with the configured hash version.

Callers who wish to read incompatible filters (e.g., for upgrading
filters from v1 to v2) may use the lower level routine,
`get_or_compute_bloom_filter()`.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c | 20 +++++++++++++++++++-
 bloom.h | 20 ++++++++++++++++++--
 2 files changed, 37 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index 9b6a30f6f6..739fa093ba 100644
--- a/bloom.c
+++ b/bloom.c
@@ -250,6 +250,23 @@ static void init_truncated_large_filter(struct bloom_filter *filter,
 	filter->version = version;
 }
 
+struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
+{
+	struct bloom_filter *filter;
+	int hash_version;
+
+	filter = get_or_compute_bloom_filter(r, c, 0, NULL, NULL);
+	if (!filter)
+		return NULL;
+
+	prepare_repo_settings(r);
+	hash_version = r->settings.commit_graph_changed_paths_version;
+
+	if (!(hash_version == -1 || hash_version == filter->version))
+		return NULL; /* unusable filter */
+	return filter;
+}
+
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						 struct commit *c,
 						 int compute_if_not_present,
@@ -275,7 +292,8 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						     filter, graph_pos);
 	}
 
-	if (filter->data && filter->len)
+	if ((filter->data && filter->len) &&
+	    (!settings || settings->hash_version == filter->version))
 		return filter;
 	if (!compute_if_not_present)
 		return NULL;
diff --git a/bloom.h b/bloom.h
index 330a140520..bfe389e29c 100644
--- a/bloom.h
+++ b/bloom.h
@@ -110,8 +110,24 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						 const struct bloom_filter_settings *settings,
 						 enum bloom_filter_computed *computed);
 
-#define get_bloom_filter(r, c) get_or_compute_bloom_filter( \
-	(r), (c), 0, NULL, NULL)
+/*
+ * Find the Bloom filter associated with the given commit "c".
+ *
+ * If any of the following are true
+ *
+ *   - the repository does not have a commit-graph, or
+ *   - the repository disables reading from the commit-graph, or
+ *   - the given commit does not have a Bloom filter computed, or
+ *   - there is a Bloom filter for commit "c", but it cannot be read
+ *     because the filter uses an incompatible version of murmur3
+ *
+ * , then `get_bloom_filter()` will return NULL. Otherwise, the corresponding
+ * Bloom filter will be returned.
+ *
+ * For callers who wish to inspect Bloom filters with incompatible hash
+ * versions, use get_or_compute_bloom_filter().
+ */
+struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c);
 
 int bloom_filter_contains(const struct bloom_filter *filter,
 			  const struct bloom_key *key,
-- 
2.42.0.4.g52b49bb434

[PATCH 10/15] t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`

[Taylor Blau]

The existing implementation of test_bloom_filters_not_used() asserts
that the Bloom filter sub-system has not been initialized at all, by
checking for the absence of any data from it from trace2.

In the following commit, it will become possible to load Bloom filters
without using them (e.g., because `commitGraph.changedPathVersion` is
incompatible with the hash version with which the commit-graph's Bloom
filters were written).

When this is the case, it's possible to initialize the Bloom filter
sub-system, while still not using any Bloom filters. When this is the
case, check that the data dump from the Bloom sub-system is all zeros,
indicating that no filters were used.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/t4216-log-bloom.sh | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 775e59d864..a77caca789 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -81,7 +81,19 @@ test_bloom_filters_used () {
 test_bloom_filters_not_used () {
 	log_args=$1
 	setup "$log_args" &&
-	! grep -q "statistics:{\"filter_not_present\":" "$TRASH_DIRECTORY/trace.perf" &&
+
+	if grep -q "statistics:{\"filter_not_present\":" "$TRASH_DIRECTORY/trace.perf"
+	then
+		# if the Bloom filter system is initialized, ensure that no
+		# filters were used
+		data="statistics:{"
+		data="$data\"filter_not_present\":0,"
+		data="$data\"maybe\":0,"
+		data="$data\"definitely_not\":0,"
+		data="$data\"false_positive\":0}"
+
+		grep -q "$data" "$TRASH_DIRECTORY/trace.perf"
+	fi &&
 	test_cmp log_wo_bloom log_w_bloom
 }
 
-- 
2.42.0.4.g52b49bb434

[PATCH 11/15] commit-graph.c: unconditionally load Bloom filters

[Taylor Blau]

In 9e4df4da07 (commit-graph: new filter ver. that fixes murmur3,
2023-08-01), we began ignoring the Bloom data ("BDAT") chunk for
commit-graphs whose Bloom filters were computed using a hash version
incompatible with the value of `commitGraph.changedPathVersion`.

Now that the Bloom API has been hardened to discard these incompatible
filters (with the exception of low-level APIs), we can safely load these
Bloom filters unconditionally.

We no longer want to return early from `graph_read_bloom_data()`, and
similarly do not want to set the bloom_settings' `hash_version` field as
a side-effect. The latter is because we want to wait until we know which
Bloom settings we're using (either the defaults, from the GIT_TEST
variables, or from the previous commit-graph layer) before deciding what
hash_version to use.

If we detect an existing BDAT chunk, we'll infer the rest of the
settings (e.g., number of hashes, bits per entry, and maximum number of
changed paths) from the earlier graph layer. The hash_version will be
inferred from the previous layer as well, unless one has already been
specified via configuration.

Once all of that is done, we normalize the value of the hash_version to
either "1" or "2".

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 commit-graph.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/commit-graph.c b/commit-graph.c
index f7322c4fff..665a3edf78 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -317,12 +317,6 @@ static int graph_read_bloom_data(const unsigned char *chunk_start,
 	uint32_t hash_version;
 	hash_version = get_be32(chunk_start);
 
-	if (*c->commit_graph_changed_paths_version == -1) {
-		*c->commit_graph_changed_paths_version = hash_version;
-	} else if (hash_version != *c->commit_graph_changed_paths_version) {
-		return 0;
-	}
-
 	g->chunk_bloom_data = chunk_start;
 	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
 	g->bloom_filter_settings->hash_version = hash_version;
@@ -2408,8 +2402,7 @@ int write_commit_graph(struct object_directory *odb,
 	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
 	ctx->num_generation_data_overflows = 0;
 
-	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
-		? 2 : 1;
+	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version;
 	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
 						      bloom_settings.bits_per_entry);
 	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
@@ -2441,10 +2434,18 @@ int write_commit_graph(struct object_directory *odb,
 		/* We have changed-paths already. Keep them in the next graph */
 		if (g && g->bloom_filter_settings) {
 			ctx->changed_paths = 1;
-			ctx->bloom_settings = g->bloom_filter_settings;
+
+			/* don't propagate the hash_version unless unspecified */
+			if (bloom_settings.hash_version == -1)
+				bloom_settings.hash_version = g->bloom_filter_settings->hash_version;
+			bloom_settings.bits_per_entry = g->bloom_filter_settings->bits_per_entry;
+			bloom_settings.num_hashes = g->bloom_filter_settings->num_hashes;
+			bloom_settings.max_changed_paths = g->bloom_filter_settings->max_changed_paths;
 		}
 	}
 
+	bloom_settings.hash_version = bloom_settings.hash_version == 2 ? 2 : 1;
+
 	if (ctx->split) {
 		struct commit_graph *g = ctx->r->objects->commit_graph;
 
-- 
2.42.0.4.g52b49bb434

[PATCH 12/15] commit-graph: drop unnecessary `graph_read_bloom_data_context`

[Taylor Blau]

The `graph_read_bloom_data_context` struct was introduced in an earlier
commit in order to pass pointers to the commit-graph and changed-path
Bloom filter version when reading the BDAT chunk.

The previous commit no longer writes through the changed_paths_version
pointer, making the surrounding context structure unnecessary. Drop it
and pass a pointer to the commit-graph directly when reading the BDAT
chunk.

Noticed-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 commit-graph.c | 14 ++------------
 1 file changed, 2 insertions(+), 12 deletions(-)

diff --git a/commit-graph.c b/commit-graph.c
index 665a3edf78..a8e33c0739 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -304,16 +304,10 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
 	return 0;
 }
 
-struct graph_read_bloom_data_context {
-	struct commit_graph *g;
-	int *commit_graph_changed_paths_version;
-};
-
 static int graph_read_bloom_data(const unsigned char *chunk_start,
 				  size_t chunk_size, void *data)
 {
-	struct graph_read_bloom_data_context *c = data;
-	struct commit_graph *g = c->g;
+	struct commit_graph *g = data;
 	uint32_t hash_version;
 	hash_version = get_be32(chunk_start);
 
@@ -405,14 +399,10 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 	}
 
 	if (s->commit_graph_changed_paths_version) {
-		struct graph_read_bloom_data_context context = {
-			.g = graph,
-			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
-		};
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
-			   graph_read_bloom_data, &context);
+			   graph_read_bloom_data, graph);
 	}
 
 	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
-- 
2.42.0.4.g52b49bb434

[PATCH 13/15] object.h: fix mis-aligned flag bits table

[Taylor Blau]

Bit position 23 is one column too far to the left.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 object.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/object.h b/object.h
index 114d45954d..db25714b4e 100644
--- a/object.h
+++ b/object.h
@@ -62,7 +62,7 @@ void object_array_init(struct object_array *array);
 
 /*
  * object flag allocation:
- * revision.h:               0---------10         15             23------27
+ * revision.h:               0---------10         15               23------27
  * fetch-pack.c:             01    67
  * negotiator/default.c:       2--5
  * walker.c:                 0-2
-- 
2.42.0.4.g52b49bb434

[PATCH 14/15] commit-graph: reuse existing Bloom filters where possible

[Taylor Blau]

In 9e4df4da07 (commit-graph: new filter ver. that fixes murmur3,
2023-08-01), a bug was described where it's possible for Git to produce
non-murmur3 hashes when the platform's "char" type is signed, and there
are paths with characters whose highest bit is set (i.e. all characters
>= 0x80).

That patch allows the caller to control which version of Bloom filters
are read and written. However, even on platforms with a signed "char"
type, it is possible to reuse existing Bloom filters if and only if
there are no changed paths in any commit's first parent tree-diff whose
characters have their highest bit set.

When this is the case, we can reuse the existing filter without having
to compute a new one. This is done by marking trees which are known to
have (or not have) any such paths. When a commit's root tree is verified
to not have any such paths, we mark it as such and declare that the
commit's Bloom filter is reusable.

Note that this heuristic only goes in one direction. If neither a commit
nor its first parent have any paths in their trees with non-ASCII
characters, then we know for certain that a path with non-ASCII
characters will not appear in a tree-diff against that commit's first
parent. The reverse isn't necessarily true: just because the tree-diff
doesn't contain any such paths does not imply that no such paths exist
in either tree.

So we end up recomputing some Bloom filters that we don't strictly have
to (i.e. their bits are the same no matter which version of murmur3 we
use). But culling these out is impossible, since we'd have to perform
the full tree-diff, which is the same effort as computing the Bloom
filter from scratch.

But because we can cache our results in each tree's flag bits, we can
often avoid recomputing many filters, thereby reducing the time it takes
to run

    $ git commit-graph write --changed-paths --reachable

when upgrading from v1 to v2 Bloom filters.

To benchmark this, let's generate a commit-graph in linux.git with v1
changed-paths in generation order[^1]:

    $ git clone git@github.com:torvalds/linux.git
    $ cd linux
    $ git commit-graph write --reachable --changed-paths
    $ graph=".git/objects/info/commit-graph"
    $ mv $graph{,.bak}

Then let's time how long it takes to go from v1 to v2 filters (with and
without the upgrade path enabled), resetting the state of the
commit-graph each time:

    $ git config commitGraph.changedPathsVersion 2
    $ hyperfine -p 'cp -f $graph.bak $graph' -L v 0,1 \
        'GIT_TEST_UPGRADE_BLOOM_FILTERS={v} git.compile commit-graph write --reachable --changed-paths'

On linux.git (where there aren't any non-ASCII paths), the timings
indicate that this patch represents a speed-up over recomputing all
Bloom filters from scratch:

    Benchmark 1: GIT_TEST_UPGRADE_BLOOM_FILTERS=0 git.compile commit-graph write --reachable --changed-paths
      Time (mean ± σ):     124.873 s ±  0.316 s    [User: 124.081 s, System: 0.643 s]
      Range (min … max):   124.621 s … 125.227 s    3 runs

    Benchmark 2: GIT_TEST_UPGRADE_BLOOM_FILTERS=1 git.compile commit-graph write --reachable --changed-paths
      Time (mean ± σ):     79.271 s ±  0.163 s    [User: 74.611 s, System: 4.521 s]
      Range (min … max):   79.112 s … 79.437 s    3 runs

    Summary
      'GIT_TEST_UPGRADE_BLOOM_FILTERS=1 git.compile commit-graph write --reachable --changed-paths' ran
        1.58 ± 0.01 times faster than 'GIT_TEST_UPGRADE_BLOOM_FILTERS=0 git.compile commit-graph write --reachable --changed-paths'

On git.git, we do have some non-ASCII paths, giving us a more modest
improvement from 4.163 seconds to 3.348 seconds, for a 1.24x speed-up.
On my machine, the stats for git.git are:

  - 8,285 Bloom filters computed from scratch
  - 10 Bloom filters generated as empty
  - 4 Bloom filters generated as truncated due to too many changed paths
  - 65,114 Bloom filters were reused when transitioning from v1 to v2.

[^1]: Note that this is is important, since `--stdin-packs` or
  `--stdin-commits` orders commits in the commit-graph by their pack
  position (with `--stdin-packs`) or in the raw input (with
  `--stdin-commits`).

  Since we compute Bloom filters in the same order that commits appear
  in the graph, we must see a commit's (first) parent before we process
  the commit itself. This is only guaranteed to happen when sorting
  commits by their generation number.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c              | 90 ++++++++++++++++++++++++++++++++++++++++++--
 bloom.h              |  1 +
 commit-graph.c       |  5 +++
 object.h             |  1 +
 t/t4216-log-bloom.sh | 35 ++++++++++++++++-
 5 files changed, 127 insertions(+), 5 deletions(-)

diff --git a/bloom.c b/bloom.c
index 739fa093ba..24dd874e46 100644
--- a/bloom.c
+++ b/bloom.c
@@ -7,6 +7,9 @@
 #include "commit-graph.h"
 #include "commit.h"
 #include "commit-slab.h"
+#include "tree.h"
+#include "tree-walk.h"
+#include "config.h"
 
 define_commit_slab(bloom_filter_slab, struct bloom_filter);
 
@@ -250,6 +253,73 @@ static void init_truncated_large_filter(struct bloom_filter *filter,
 	filter->version = version;
 }
 
+#define VISITED   (1u<<21)
+#define HIGH_BITS (1u<<22)
+
+static int has_entries_with_high_bit(struct repository *r, struct tree *t)
+{
+	if (parse_tree(t))
+		return 1;
+
+	if (!(t->object.flags & VISITED)) {
+		struct tree_desc desc;
+		struct name_entry entry;
+
+		init_tree_desc(&desc, t->buffer, t->size);
+		while (tree_entry(&desc, &entry)) {
+			size_t i;
+			for (i = 0; i < entry.pathlen; i++) {
+				if (entry.path[i] & 0x80) {
+					t->object.flags |= HIGH_BITS;
+					goto done;
+				}
+			}
+
+			if (S_ISDIR(entry.mode)) {
+				struct tree *sub = lookup_tree(r, &entry.oid);
+				if (sub && has_entries_with_high_bit(r, sub)) {
+					t->object.flags |= HIGH_BITS;
+					goto done;
+				}
+			}
+
+		}
+
+done:
+		t->object.flags |= VISITED;
+	}
+
+	return !!(t->object.flags & HIGH_BITS);
+}
+
+static int commit_tree_has_high_bit_paths(struct repository *r,
+					  struct commit *c)
+{
+	struct tree *t;
+	if (repo_parse_commit(r, c))
+		return 1;
+	t = repo_get_commit_tree(r, c);
+	if (!t)
+		return 1;
+	return has_entries_with_high_bit(r, t);
+}
+
+static struct bloom_filter *upgrade_filter(struct repository *r, struct commit *c,
+					   struct bloom_filter *filter,
+					   int hash_version)
+{
+	struct commit_list *p = c->parents;
+	if (commit_tree_has_high_bit_paths(r, c))
+		return NULL;
+
+	if (p && commit_tree_has_high_bit_paths(r, p->item))
+		return NULL;
+
+	filter->version = hash_version;
+
+	return filter;
+}
+
 struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
 {
 	struct bloom_filter *filter;
@@ -292,9 +362,23 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						     filter, graph_pos);
 	}
 
-	if ((filter->data && filter->len) &&
-	    (!settings || settings->hash_version == filter->version))
-		return filter;
+	if (filter->data && filter->len) {
+		struct bloom_filter *upgrade;
+		if (!settings || settings->hash_version == filter->version)
+			return filter;
+
+		/* version mismatch, see if we can upgrade */
+		if (compute_if_not_present &&
+		    git_env_bool("GIT_TEST_UPGRADE_BLOOM_FILTERS", 1)) {
+			upgrade = upgrade_filter(r, c, filter,
+						 settings->hash_version);
+			if (upgrade) {
+				if (computed)
+					*computed |= BLOOM_UPGRADED;
+				return upgrade;
+			}
+		}
+	}
 	if (!compute_if_not_present)
 		return NULL;
 
diff --git a/bloom.h b/bloom.h
index bfe389e29c..e3a9b68905 100644
--- a/bloom.h
+++ b/bloom.h
@@ -102,6 +102,7 @@ enum bloom_filter_computed {
 	BLOOM_COMPUTED     = (1 << 1),
 	BLOOM_TRUNC_LARGE  = (1 << 2),
 	BLOOM_TRUNC_EMPTY  = (1 << 3),
+	BLOOM_UPGRADED     = (1 << 4),
 };
 
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
diff --git a/commit-graph.c b/commit-graph.c
index a8e33c0739..a3473df515 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -1045,6 +1045,7 @@ struct write_commit_graph_context {
 	int count_bloom_filter_not_computed;
 	int count_bloom_filter_trunc_empty;
 	int count_bloom_filter_trunc_large;
+	int count_bloom_filter_upgraded;
 };
 
 static int write_graph_chunk_fanout(struct hashfile *f,
@@ -1651,6 +1652,8 @@ static void trace2_bloom_filter_write_statistics(struct write_commit_graph_conte
 			   ctx->count_bloom_filter_trunc_empty);
 	trace2_data_intmax("commit-graph", ctx->r, "filter-trunc-large",
 			   ctx->count_bloom_filter_trunc_large);
+	trace2_data_intmax("commit-graph", ctx->r, "filter-upgraded",
+			   ctx->count_bloom_filter_upgraded);
 }
 
 static void compute_bloom_filters(struct write_commit_graph_context *ctx)
@@ -1692,6 +1695,8 @@ static void compute_bloom_filters(struct write_commit_graph_context *ctx)
 				ctx->count_bloom_filter_trunc_empty++;
 			if (computed & BLOOM_TRUNC_LARGE)
 				ctx->count_bloom_filter_trunc_large++;
+		} else if (computed & BLOOM_UPGRADED) {
+			ctx->count_bloom_filter_upgraded++;
 		} else if (computed & BLOOM_NOT_COMPUTED)
 			ctx->count_bloom_filter_not_computed++;
 		ctx->total_bloom_filter_data_size += filter
diff --git a/object.h b/object.h
index db25714b4e..2e5e08725f 100644
--- a/object.h
+++ b/object.h
@@ -75,6 +75,7 @@ void object_array_init(struct object_array *array);
  * commit-reach.c:                                  16-----19
  * sha1-name.c:                                              20
  * list-objects-filter.c:                                      21
+ * bloom.c:                                                    2122
  * builtin/fsck.c:           0--3
  * builtin/gc.c:             0
  * builtin/index-pack.c:                                     2021
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index a77caca789..48f8109a66 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -217,6 +217,10 @@ test_filter_trunc_large () {
 	grep "\"key\":\"filter-trunc-large\",\"value\":\"$1\"" $2
 }
 
+test_filter_upgraded () {
+	grep "\"key\":\"filter-upgraded\",\"value\":\"$1\"" $2
+}
+
 test_expect_success 'correctly report changes over limit' '
 	git init limits &&
 	(
@@ -543,10 +547,19 @@ test_expect_success 'when writing another commit graph, preserve existing versio
 test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
 	git init doublewrite &&
 	test_commit -C doublewrite c "$CENT" &&
+
 	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
-	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C doublewrite commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
 	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
-	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C doublewrite commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
 	(
 		cd doublewrite &&
 		echo "c01f" >expect &&
@@ -555,4 +568,22 @@ test_expect_success 'when writing commit graph, do not reuse changed-path of ano
 	)
 '
 
+test_expect_success 'when writing commit graph, reuse changed-path of another version where possible' '
+	git init upgrade &&
+
+	test_commit -C upgrade base no-high-bits &&
+
+	git -C upgrade config --add commitgraph.changedPathsVersion 1 &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C upgrade commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
+	git -C upgrade config --add commitgraph.changedPathsVersion 2 &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C upgrade commit-graph write --reachable --changed-paths &&
+	test_filter_computed 0 trace2.txt &&
+	test_filter_upgraded 1 trace2.txt
+'
+
 test_done
-- 
2.42.0.4.g52b49bb434

[PATCH 15/15] bloom: introduce `deinit_bloom_filters()`

[Taylor Blau]

After we are done using Bloom filters, we do not currently clean up any
memory allocated by the commit slab used to store those filters in the
first place.

Besides the bloom_filter structures themselves, there is mostly nothing
to free() in the first place, since in the read-only path all Bloom
filter's `data` members point to a memory mapped region in the
commit-graph file itself.

But when generating Bloom filters from scratch (or initializing
truncated filters) we allocate additional memory to store the filter's
data.

Keep track of when we need to free() this additional chunk of memory by
using an extra pointer `to_free`. Most of the time this will be NULL
(indicating that we are representing an existing Bloom filter stored in
a memory mapped region). When it is non-NULL, free it before discarding
the Bloom filters slab.

Suggested-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c        | 16 +++++++++++++++-
 bloom.h        |  3 +++
 commit-graph.c |  4 ++++
 3 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/bloom.c b/bloom.c
index 24dd874e46..ff131893cd 100644
--- a/bloom.c
+++ b/bloom.c
@@ -59,6 +59,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
 					sizeof(unsigned char) * start_index +
 					BLOOMDATA_CHUNK_HEADER_SIZE);
 	filter->version = g->bloom_filter_settings->hash_version;
+	filter->to_free = NULL;
 
 	return 1;
 }
@@ -231,6 +232,18 @@ void init_bloom_filters(void)
 	init_bloom_filter_slab(&bloom_filters);
 }
 
+static void free_one_bloom_filter(struct bloom_filter *filter)
+{
+	if (!filter)
+		return;
+	free(filter->to_free);
+}
+
+void deinit_bloom_filters(void)
+{
+	deep_clear_bloom_filter_slab(&bloom_filters, free_one_bloom_filter);
+}
+
 static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 		       const struct hashmap_entry *eptr,
 		       const struct hashmap_entry *entry_or_key,
@@ -247,7 +260,7 @@ static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 static void init_truncated_large_filter(struct bloom_filter *filter,
 					int version)
 {
-	filter->data = xmalloc(1);
+	filter->data = filter->to_free = xmalloc(1);
 	filter->data[0] = 0xFF;
 	filter->len = 1;
 	filter->version = version;
@@ -449,6 +462,7 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 			filter->len = 1;
 		}
 		CALLOC_ARRAY(filter->data, filter->len);
+		filter->to_free = filter->data;
 
 		hashmap_for_each_entry(&pathmap, &iter, e, entry) {
 			struct bloom_key key;
diff --git a/bloom.h b/bloom.h
index e3a9b68905..d20e64bfbb 100644
--- a/bloom.h
+++ b/bloom.h
@@ -56,6 +56,8 @@ struct bloom_filter {
 	unsigned char *data;
 	size_t len;
 	int version;
+
+	void *to_free;
 };
 
 /*
@@ -96,6 +98,7 @@ void add_key_to_filter(const struct bloom_key *key,
 		       const struct bloom_filter_settings *settings);
 
 void init_bloom_filters(void);
+void deinit_bloom_filters(void);
 
 enum bloom_filter_computed {
 	BLOOM_NOT_COMPUTED = (1 << 0),
diff --git a/commit-graph.c b/commit-graph.c
index a3473df515..585539da2f 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -723,6 +723,7 @@ static void close_commit_graph_one(struct commit_graph *g)
 void close_commit_graph(struct raw_object_store *o)
 {
 	close_commit_graph_one(o->commit_graph);
+	deinit_bloom_filters();
 	o->commit_graph = NULL;
 }
 
@@ -2523,6 +2524,9 @@ int write_commit_graph(struct object_directory *odb,
 
 	res = write_commit_graph_file(ctx);
 
+	if (ctx->changed_paths)
+		deinit_bloom_filters();
+
 	if (ctx->split)
 		mark_commit_graphs(ctx);
 
-- 
2.42.0.4.g52b49bb434

Re: [PATCH 00/15] bloom: changed-path Bloom filters v2

[Jonathan Tan]

Taylor Blau <me@ttaylorr.com> writes:
> Jonathan Tan (4):
>   gitformat-commit-graph: describe version 2 of BDAT
>   t4216: test changed path filters with high bit paths
>   repo-settings: introduce commitgraph.changedPathsVersion
>   commit-graph: new filter ver. that fixes murmur3
> 
> Taylor Blau (11):
>   t/helper/test-read-graph.c: extract `dump_graph_info()`
>   bloom.h: make `load_bloom_filter_from_graph()` public
>   t/helper/test-read-graph: implement `bloom-filters` mode
>   bloom: annotate filters with hash version
>   bloom: prepare to discard incompatible Bloom filters
>   t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
>   commit-graph.c: unconditionally load Bloom filters
>   commit-graph: drop unnecessary `graph_read_bloom_data_context`
>   object.h: fix mis-aligned flag bits table
>   commit-graph: reuse existing Bloom filters where possible
>   bloom: introduce `deinit_bloom_filters()`

Thanks. I had one small comment (sent as an email reply to one of the
patches), but everything else looks good.
 

Re: [PATCH 00/15] bloom: changed-path Bloom filters v2

[Jonathan Tan]

Jonathan Tan <jonathantanmy@google.com> writes:
> Taylor Blau <me@ttaylorr.com> writes:
> > Jonathan Tan (4):
> >   gitformat-commit-graph: describe version 2 of BDAT
> >   t4216: test changed path filters with high bit paths
> >   repo-settings: introduce commitgraph.changedPathsVersion
> >   commit-graph: new filter ver. that fixes murmur3
> > 
> > Taylor Blau (11):
> >   t/helper/test-read-graph.c: extract `dump_graph_info()`
> >   bloom.h: make `load_bloom_filter_from_graph()` public
> >   t/helper/test-read-graph: implement `bloom-filters` mode
> >   bloom: annotate filters with hash version
> >   bloom: prepare to discard incompatible Bloom filters
> >   t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
> >   commit-graph.c: unconditionally load Bloom filters
> >   commit-graph: drop unnecessary `graph_read_bloom_data_context`
> >   object.h: fix mis-aligned flag bits table
> >   commit-graph: reuse existing Bloom filters where possible
> >   bloom: introduce `deinit_bloom_filters()`
> 
> Thanks. I had one small comment (sent as an email reply to one of the
> patches), but everything else looks good.

I mistakenly sent my reply to an earlier version [1]. (Taylor has seen
it, so this note is more for future readers who might be curious about
what that email reply contains.)

[1] https://lore.kernel.org/git/20230824222051.2320003-1-jonathantanmy@google.com/

Re: [PATCH 00/15] bloom: changed-path Bloom filters v2

[Jonathan Tan]

Jonathan Tan <jonathantanmy@google.com> writes:
> I mistakenly sent my reply to an earlier version [1]. (Taylor has seen
> it, so this note is more for future readers who might be curious about
> what that email reply contains.)
> 
> [1] https://lore.kernel.org/git/20230824222051.2320003-1-jonathantanmy@google.com/

I think that this patch set is good for merging. There is a discussion
between Taylor and me at the link above in which Taylor thinks something
is OK and I think that that thing is OK except in a very specific
situation (there exists a commit-graph chain in which the different
layers have different Bloom filter versions). (Hopefully that's a good
summary of the discussion.) I think that the situation is narrow enough
and have seen that Taylor (whose capabilities I have a high opinion of)
has looked into it. I think the only way for me to be fully convinced
is to write a test that includes these different layers and see how the
code functions.

So I think that we can merge this patch set. An alternative would be
to wait for me (or someone else) to write such a test, but if it's just
me that's worried about this situation, it's probably not worth waiting
just for this.

The other outstanding thing is that Szeder Gabor pointed out that Bloom
filters are not applied to root commits so some of the tests don't
test what you would expect [2]. I've updated the tests and pushed the
results to GitHub [3]. I'm OK with using that version, or the current
version (in which case I'll resend the updated tests once this version
is merged).

[2] https://lore.kernel.org/git/20230826150610.GA1928@szeder.dev/
[3] https://github.com/jonathantanmy/git/tree/changedpath

Re: [PATCH 00/15] bloom: changed-path Bloom filters v2

[Junio C Hamano]

Jonathan Tan <jonathantanmy@google.com> writes:

> So I think that we can merge this patch set. An alternative would be
> to wait for me (or someone else) to write such a test, but if it's just
> me that's worried about this situation, it's probably not worth waiting
> just for this.
>
> The other outstanding thing is that Szeder Gabor pointed out that Bloom
> filters are not applied to root commits so some of the tests don't
> test what you would expect [2]. I've updated the tests and pushed the
> results to GitHub [3]. I'm OK with using that version, or the current
> version (in which case I'll resend the updated tests once this version
> is merged).
>
> [2] https://lore.kernel.org/git/20230826150610.GA1928@szeder.dev/
> [3] https://github.com/jonathantanmy/git/tree/changedpath

Thanks for a concise and well thought out summary.

If we know that some tests in the current round is not doing the
right thing, and we already have an updated set of tests to fix
them, I doubt that the topic is so urgent that merging a known to be
incomplete version is preferrable than seeing the hopefully final
version on the list that everybody can agree with.

Thanks.

[PATCH v2 00/15] bloom: changed-path Bloom filters v2

[Jonathan Tan]

Here's an updated patch set containing updates to the tests, making them
run on non-root commits to ensure that we exercise the Bloom filters.
Thanks to SZEDER Gábor for spotting this.

Jonathan Tan (4):
  gitformat-commit-graph: describe version 2 of BDAT
  t4216: test changed path filters with high bit paths
  repo-settings: introduce commitgraph.changedPathsVersion
  commit-graph: new filter ver. that fixes murmur3

Taylor Blau (11):
  t/helper/test-read-graph.c: extract `dump_graph_info()`
  bloom.h: make `load_bloom_filter_from_graph()` public
  t/helper/test-read-graph: implement `bloom-filters` mode
  bloom: annotate filters with hash version
  bloom: prepare to discard incompatible Bloom filters
  t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
  commit-graph.c: unconditionally load Bloom filters
  commit-graph: drop unnecessary `graph_read_bloom_data_context`
  object.h: fix mis-aligned flag bits table
  commit-graph: reuse existing Bloom filters where possible
  bloom: introduce `deinit_bloom_filters()`

 Documentation/config/commitgraph.txt     |  26 ++-
 Documentation/gitformat-commit-graph.txt |   9 +-
 bloom.c                                  | 208 +++++++++++++++++++++--
 bloom.h                                  |  38 ++++-
 commit-graph.c                           |  36 +++-
 object.h                                 |   3 +-
 oss-fuzz/fuzz-commit-graph.c             |   2 +-
 repo-settings.c                          |   6 +-
 repository.h                             |   2 +-
 t/helper/test-bloom.c                    |   9 +-
 t/helper/test-read-graph.c               |  65 +++++--
 t/t0095-bloom.sh                         |   8 +
 t/t4216-log-bloom.sh                     | 202 +++++++++++++++++++++-
 13 files changed, 564 insertions(+), 50 deletions(-)

Range-diff against v1:
 1:  dcfc987741 =  1:  dcfc987741 gitformat-commit-graph: describe version 2 of BDAT
 2:  0c56f2a9e9 =  2:  0c56f2a9e9 t/helper/test-read-graph.c: extract `dump_graph_info()`
 3:  8405b845e5 =  3:  8405b845e5 bloom.h: make `load_bloom_filter_from_graph()` public
 4:  3a25f90c15 =  4:  3a25f90c15 t/helper/test-read-graph: implement `bloom-filters` mode
 5:  e300d338e1 !  5:  7858010665 t4216: test changed path filters with high bit paths
    @@ t/t4216-log-bloom.sh: test_expect_success 'Bloom generation backfills empty comm
     +	true
     +'
     +
    ++test_expect_success 'setup make another commit' '
    ++	# "git log" does not use Bloom filters for root commits - see how, in
    ++	# revision.c, rev_compare_tree() (the only code path that eventually calls
    ++	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
    ++	# has one parent. Therefore, make another commit so that we perform the tests on
    ++	# a non-root commit.
    ++	test_commit -C highbit1 anotherc1 "another$CENT"
    ++'
    ++
     +test_expect_success 'version 1 changed-path used when version 1 requested' '
     +	(
     +		cd highbit1 &&
    -+		test_bloom_filters_used "-- $CENT"
    ++		test_bloom_filters_used "-- another$CENT"
     +	)
     +'
     +
 6:  6bc665e1d3 =  6:  94ad289dbb repo-settings: introduce commitgraph.changedPathsVersion
 7:  7ef3b2bbbd !  7:  44d3163125 commit-graph: new filter ver. that fixes murmur3
    @@ t/t4216-log-bloom.sh: test_expect_success 'version 1 changed-path used when vers
     +	(
     +		cd highbit1 &&
     +		git config --add commitgraph.changedPathsVersion 2 &&
    -+		test_bloom_filters_not_used "-- $CENT"
    ++		test_bloom_filters_not_used "-- another$CENT"
     +	)
     +'
     +
    @@ t/t4216-log-bloom.sh: test_expect_success 'version 1 changed-path used when vers
     +	(
     +		cd highbit1 &&
     +		git config --add commitgraph.changedPathsVersion -1 &&
    -+		test_bloom_filters_used "-- $CENT"
    ++		test_bloom_filters_used "-- another$CENT"
     +	)
     +'
     +
    @@ t/t4216-log-bloom.sh: test_expect_success 'version 1 changed-path used when vers
     +	)
     +'
     +
    ++test_expect_success 'setup make another commit' '
    ++	# "git log" does not use Bloom filters for root commits - see how, in
    ++	# revision.c, rev_compare_tree() (the only code path that eventually calls
    ++	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
    ++	# has one parent. Therefore, make another commit so that we perform the tests on
    ++	# a non-root commit.
    ++	test_commit -C highbit2 anotherc2 "another$CENT"
    ++'
    ++
     +test_expect_success 'version 2 changed-path used when version 2 requested' '
     +	(
     +		cd highbit2 &&
    -+		test_bloom_filters_used "-- $CENT"
    ++		test_bloom_filters_used "-- another$CENT"
     +	)
     +'
     +
    @@ t/t4216-log-bloom.sh: test_expect_success 'version 1 changed-path used when vers
     +	(
     +		cd highbit2 &&
     +		git config --add commitgraph.changedPathsVersion 1 &&
    -+		test_bloom_filters_not_used "-- $CENT"
    ++		test_bloom_filters_not_used "-- another$CENT"
     +	)
     +'
     +
    @@ t/t4216-log-bloom.sh: test_expect_success 'version 1 changed-path used when vers
     +	(
     +		cd highbit2 &&
     +		git config --add commitgraph.changedPathsVersion -1 &&
    -+		test_bloom_filters_used "-- $CENT"
    ++		test_bloom_filters_used "-- another$CENT"
     +	)
     +'
     +
 8:  302caee39d =  8:  a5bf23a7d3 bloom: annotate filters with hash version
 9:  d2b0726266 =  9:  3de6cd8460 bloom: prepare to discard incompatible Bloom filters
10:  9a9992220f = 10:  ef04389a0e t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
11:  607945ab05 = 11:  e0c1c1ccec commit-graph.c: unconditionally load Bloom filters
12:  e397d83895 = 12:  4d57f51854 commit-graph: drop unnecessary `graph_read_bloom_data_context`
13:  bd3ad6b6c0 = 13:  a3b4e7ef59 object.h: fix mis-aligned flag bits table
14:  2996f0fdb6 = 14:  05357f9533 commit-graph: reuse existing Bloom filters where possible
15:  a3b5b22db0 = 15:  58a1d90e6d bloom: introduce `deinit_bloom_filters()`
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 01/15] gitformat-commit-graph: describe version 2 of BDAT

[Jonathan Tan]

The code change to Git to support version 2 will be done in subsequent
commits.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 Documentation/gitformat-commit-graph.txt | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/Documentation/gitformat-commit-graph.txt b/Documentation/gitformat-commit-graph.txt
index 31cad585e2..3e906e8030 100644
--- a/Documentation/gitformat-commit-graph.txt
+++ b/Documentation/gitformat-commit-graph.txt
@@ -142,13 +142,16 @@ All multi-byte numbers are in network byte order.
 
 ==== Bloom Filter Data (ID: {'B', 'D', 'A', 'T'}) [Optional]
     * It starts with header consisting of three unsigned 32-bit integers:
-      - Version of the hash algorithm being used. We currently only support
-	value 1 which corresponds to the 32-bit version of the murmur3 hash
+      - Version of the hash algorithm being used. We currently support
+	value 2 which corresponds to the 32-bit version of the murmur3 hash
 	implemented exactly as described in
 	https://en.wikipedia.org/wiki/MurmurHash#Algorithm and the double
 	hashing technique using seed values 0x293ae76f and 0x7e646e2 as
 	described in https://doi.org/10.1007/978-3-540-30494-4_26 "Bloom Filters
-	in Probabilistic Verification"
+	in Probabilistic Verification". Version 1 Bloom filters have a bug that appears
+	when char is signed and the repository has path names that have characters >=
+	0x80; Git supports reading and writing them, but this ability will be removed
+	in a future version of Git.
       - The number of times a path is hashed and hence the number of bit positions
 	      that cumulatively determine whether a file is present in the commit.
       - The minimum number of bits 'b' per entry in the Bloom filter. If the filter
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 02/15] t/helper/test-read-graph.c: extract `dump_graph_info()`

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

Prepare for the 'read-graph' test helper to perform other tasks besides
dumping high-level information about the commit-graph by extracting its
main routine into a separate function.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 t/helper/test-read-graph.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)

diff --git a/t/helper/test-read-graph.c b/t/helper/test-read-graph.c
index 8c7a83f578..3375392f6c 100644
--- a/t/helper/test-read-graph.c
+++ b/t/helper/test-read-graph.c
@@ -5,20 +5,8 @@
 #include "bloom.h"
 #include "setup.h"
 
-int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+static void dump_graph_info(struct commit_graph *graph)
 {
-	struct commit_graph *graph = NULL;
-	struct object_directory *odb;
-
-	setup_git_directory();
-	odb = the_repository->objects->odb;
-
-	prepare_repo_settings(the_repository);
-
-	graph = read_commit_graph_one(the_repository, odb);
-	if (!graph)
-		return 1;
-
 	printf("header: %08x %d %d %d %d\n",
 		ntohl(*(uint32_t*)graph->data),
 		*(unsigned char*)(graph->data + 4),
@@ -57,6 +45,23 @@ int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
 	if (graph->topo_levels)
 		printf(" topo_levels");
 	printf("\n");
+}
+
+int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+{
+	struct commit_graph *graph = NULL;
+	struct object_directory *odb;
+
+	setup_git_directory();
+	odb = the_repository->objects->odb;
+
+	prepare_repo_settings(the_repository);
+
+	graph = read_commit_graph_one(the_repository, odb);
+	if (!graph)
+		return 1;
+
+	dump_graph_info(graph);
 
 	UNLEAK(graph);
 
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 03/15] bloom.h: make `load_bloom_filter_from_graph()` public

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

Prepare for a future commit to use the load_bloom_filter_from_graph()
function directly to load specific Bloom filters out of the commit-graph
for manual inspection (to be used during tests).

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 bloom.c | 6 +++---
 bloom.h | 5 +++++
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index aef6b5fea2..3e78cfe79d 100644
--- a/bloom.c
+++ b/bloom.c
@@ -29,9 +29,9 @@ static inline unsigned char get_bitmask(uint32_t pos)
 	return ((unsigned char)1) << (pos & (BITS_PER_WORD - 1));
 }
 
-static int load_bloom_filter_from_graph(struct commit_graph *g,
-					struct bloom_filter *filter,
-					uint32_t graph_pos)
+int load_bloom_filter_from_graph(struct commit_graph *g,
+				 struct bloom_filter *filter,
+				 uint32_t graph_pos)
 {
 	uint32_t lex_pos, start_index, end_index;
 
diff --git a/bloom.h b/bloom.h
index adde6dfe21..1e4f612d2c 100644
--- a/bloom.h
+++ b/bloom.h
@@ -3,6 +3,7 @@
 
 struct commit;
 struct repository;
+struct commit_graph;
 
 struct bloom_filter_settings {
 	/*
@@ -68,6 +69,10 @@ struct bloom_key {
 	uint32_t *hashes;
 };
 
+int load_bloom_filter_from_graph(struct commit_graph *g,
+				 struct bloom_filter *filter,
+				 uint32_t graph_pos);
+
 /*
  * Calculate the murmur3 32-bit hash value for the given data
  * using the given seed.
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 04/15] t/helper/test-read-graph: implement `bloom-filters` mode

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

Implement a mode of the "read-graph" test helper to dump out the
hexadecimal contents of the Bloom filter(s) contained in a commit-graph.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 t/helper/test-read-graph.c | 42 +++++++++++++++++++++++++++++++++-----
 1 file changed, 37 insertions(+), 5 deletions(-)

diff --git a/t/helper/test-read-graph.c b/t/helper/test-read-graph.c
index 3375392f6c..899b5f41cc 100644
--- a/t/helper/test-read-graph.c
+++ b/t/helper/test-read-graph.c
@@ -47,10 +47,32 @@ static void dump_graph_info(struct commit_graph *graph)
 	printf("\n");
 }
 
-int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+static void dump_graph_bloom_filters(struct commit_graph *graph)
+{
+	uint32_t i;
+
+	for (i = 0; i < graph->num_commits + graph->num_commits_in_base; i++) {
+		struct bloom_filter filter = { 0 };
+		size_t j;
+
+		if (load_bloom_filter_from_graph(graph, &filter, i) < 0) {
+			fprintf(stderr, "missing Bloom filter for graph "
+				"position %"PRIu32"\n", i);
+			continue;
+		}
+
+		for (j = 0; j < filter.len; j++)
+			printf("%02x", filter.data[j]);
+		if (filter.len)
+			printf("\n");
+	}
+}
+
+int cmd__read_graph(int argc, const char **argv)
 {
 	struct commit_graph *graph = NULL;
 	struct object_directory *odb;
+	int ret = 0;
 
 	setup_git_directory();
 	odb = the_repository->objects->odb;
@@ -58,12 +80,22 @@ int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
 	prepare_repo_settings(the_repository);
 
 	graph = read_commit_graph_one(the_repository, odb);
-	if (!graph)
-		return 1;
+	if (!graph) {
+		ret = 1;
+		goto done;
+	}
 
-	dump_graph_info(graph);
+	if (argc <= 1)
+		dump_graph_info(graph);
+	else if (!strcmp(argv[1], "bloom-filters"))
+		dump_graph_bloom_filters(graph);
+	else {
+		fprintf(stderr, "unknown sub-command: '%s'\n", argv[1]);
+		ret = 1;
+	}
 
+done:
 	UNLEAK(graph);
 
-	return 0;
+	return ret;
 }
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 05/15] t4216: test changed path filters with high bit paths

[Jonathan Tan]

Subsequent commits will teach Git another version of changed path
filter that has different behavior with paths that contain at least
one character with its high bit set, so test the existing behavior as
a baseline.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 t/t4216-log-bloom.sh | 52 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 52 insertions(+)

diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index fa9d32facf..0aa9719934 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -404,4 +404,56 @@ test_expect_success 'Bloom generation backfills empty commits' '
 	)
 '
 
+get_first_changed_path_filter () {
+	test-tool read-graph bloom-filters >filters.dat &&
+	head -n 1 filters.dat
+}
+
+# chosen to be the same under all Unicode normalization forms
+CENT=$(printf "\302\242")
+
+test_expect_success 'set up repo with high bit path, version 1 changed-path' '
+	git init highbit1 &&
+	test_commit -C highbit1 c1 "$CENT" &&
+	git -C highbit1 commit-graph write --reachable --changed-paths
+'
+
+test_expect_success 'setup check value of version 1 changed-path' '
+	(
+		cd highbit1 &&
+		echo "52a9" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
+# expect will not match actual if char is unsigned by default. Write the test
+# in this way, so that a user running this test script can still see if the two
+# files match. (It will appear as an ordinary success if they match, and a skip
+# if not.)
+if test_cmp highbit1/expect highbit1/actual
+then
+	test_set_prereq SIGNED_CHAR_BY_DEFAULT
+fi
+test_expect_success SIGNED_CHAR_BY_DEFAULT 'check value of version 1 changed-path' '
+	# Only the prereq matters for this test.
+	true
+'
+
+test_expect_success 'setup make another commit' '
+	# "git log" does not use Bloom filters for root commits - see how, in
+	# revision.c, rev_compare_tree() (the only code path that eventually calls
+	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
+	# has one parent. Therefore, make another commit so that we perform the tests on
+	# a non-root commit.
+	test_commit -C highbit1 anotherc1 "another$CENT"
+'
+
+test_expect_success 'version 1 changed-path used when version 1 requested' '
+	(
+		cd highbit1 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
 test_done
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 06/15] repo-settings: introduce commitgraph.changedPathsVersion

[Jonathan Tan]

A subsequent commit will introduce another version of the changed-path
filter in the commit graph file. In order to control which version to
write (and read), a config variable is needed.

Therefore, introduce this config variable. For forwards compatibility,
teach Git to not read commit graphs when the config variable
is set to an unsupported version. Because we teach Git this,
commitgraph.readChangedPaths is now redundant, so deprecate it and
define its behavior in terms of the config variable we introduce.

This commit does not change the behavior of writing (Git writes changed
path filters when explicitly instructed regardless of any config
variable), but a subsequent commit will restrict Git such that it will
only write when commitgraph.changedPathsVersion is a recognized value.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 Documentation/config/commitgraph.txt | 23 ++++++++++++++++++++---
 commit-graph.c                       |  2 +-
 oss-fuzz/fuzz-commit-graph.c         |  2 +-
 repo-settings.c                      |  6 +++++-
 repository.h                         |  2 +-
 5 files changed, 28 insertions(+), 7 deletions(-)

diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
index 30604e4a4c..2dc9170622 100644
--- a/Documentation/config/commitgraph.txt
+++ b/Documentation/config/commitgraph.txt
@@ -9,6 +9,23 @@ commitGraph.maxNewFilters::
 	commit-graph write` (c.f., linkgit:git-commit-graph[1]).
 
 commitGraph.readChangedPaths::
-	If true, then git will use the changed-path Bloom filters in the
-	commit-graph file (if it exists, and they are present). Defaults to
-	true. See linkgit:git-commit-graph[1] for more information.
+	Deprecated. Equivalent to commitGraph.changedPathsVersion=-1 if true, and
+	commitGraph.changedPathsVersion=0 if false. (If commitGraph.changedPathVersion
+	is also set, commitGraph.changedPathsVersion takes precedence.)
+
+commitGraph.changedPathsVersion::
+	Specifies the version of the changed-path Bloom filters that Git will read and
+	write. May be -1, 0 or 1.
++
+Defaults to -1.
++
+If -1, Git will use the version of the changed-path Bloom filters in the
+repository, defaulting to 1 if there are none.
++
+If 0, Git will not read any Bloom filters, and will write version 1 Bloom
+filters when instructed to write.
++
+If 1, Git will only read version 1 Bloom filters, and will write version 1
+Bloom filters.
++
+See linkgit:git-commit-graph[1] for more information.
diff --git a/commit-graph.c b/commit-graph.c
index 0aa1640d15..da99f15fdf 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -401,7 +401,7 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 			graph->read_generation_data = 1;
 	}
 
-	if (s->commit_graph_read_changed_paths) {
+	if (s->commit_graph_changed_paths_version) {
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
diff --git a/oss-fuzz/fuzz-commit-graph.c b/oss-fuzz/fuzz-commit-graph.c
index 2992079dd9..325c0b991a 100644
--- a/oss-fuzz/fuzz-commit-graph.c
+++ b/oss-fuzz/fuzz-commit-graph.c
@@ -19,7 +19,7 @@ int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
 	 * possible.
 	 */
 	the_repository->settings.commit_graph_generation_version = 2;
-	the_repository->settings.commit_graph_read_changed_paths = 1;
+	the_repository->settings.commit_graph_changed_paths_version = 1;
 	g = parse_commit_graph(&the_repository->settings, (void *)data, size);
 	repo_clear(the_repository);
 	free_commit_graph(g);
diff --git a/repo-settings.c b/repo-settings.c
index 525f69c0c7..db8fe817f3 100644
--- a/repo-settings.c
+++ b/repo-settings.c
@@ -24,6 +24,7 @@ void prepare_repo_settings(struct repository *r)
 	int value;
 	const char *strval;
 	int manyfiles;
+	int read_changed_paths;
 
 	if (!r->gitdir)
 		BUG("Cannot add settings for uninitialized repository");
@@ -54,7 +55,10 @@ void prepare_repo_settings(struct repository *r)
 	/* Commit graph config or default, does not cascade (simple) */
 	repo_cfg_bool(r, "core.commitgraph", &r->settings.core_commit_graph, 1);
 	repo_cfg_int(r, "commitgraph.generationversion", &r->settings.commit_graph_generation_version, 2);
-	repo_cfg_bool(r, "commitgraph.readchangedpaths", &r->settings.commit_graph_read_changed_paths, 1);
+	repo_cfg_bool(r, "commitgraph.readchangedpaths", &read_changed_paths, 1);
+	repo_cfg_int(r, "commitgraph.changedpathsversion",
+		     &r->settings.commit_graph_changed_paths_version,
+		     read_changed_paths ? -1 : 0);
 	repo_cfg_bool(r, "gc.writecommitgraph", &r->settings.gc_write_commit_graph, 1);
 	repo_cfg_bool(r, "fetch.writecommitgraph", &r->settings.fetch_write_commit_graph, 0);
 
diff --git a/repository.h b/repository.h
index 5f18486f64..f71154e12c 100644
--- a/repository.h
+++ b/repository.h
@@ -29,7 +29,7 @@ struct repo_settings {
 
 	int core_commit_graph;
 	int commit_graph_generation_version;
-	int commit_graph_read_changed_paths;
+	int commit_graph_changed_paths_version;
 	int gc_write_commit_graph;
 	int fetch_write_commit_graph;
 	int command_requires_full_index;
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 07/15] commit-graph: new filter ver. that fixes murmur3

[Jonathan Tan]

The murmur3 implementation in bloom.c has a bug when converting series
of 4 bytes into network-order integers when char is signed (which is
controllable by a compiler option, and the default signedness of char is
platform-specific). When a string contains characters with the high bit
set, this bug causes results that, although internally consistent within
Git, does not accord with other implementations of murmur3 (thus,
the changed path filters wouldn't be readable by other off-the-shelf
implementatios of murmur3) and even with Git binaries that were compiled
with different signedness of char. This bug affects both how Git writes
changed path filters to disk and how Git interprets changed path filters
on disk.

Therefore, introduce a new version (2) of changed path filters that
corrects this problem. The existing version (1) is still supported and
is still the default, but users should migrate away from it as soon
as possible.

Because this bug only manifests with characters that have the high bit
set, it may be possible that some (or all) commits in a given repo would
have the same changed path filter both before and after this fix is
applied. However, in order to determine whether this is the case, the
changed paths would first have to be computed, at which point it is not
much more expensive to just compute a new changed path filter.

So this patch does not include any mechanism to "salvage" changed path
filters from repositories. There is also no "mixed" mode - for each
invocation of Git, reading and writing changed path filters are done
with the same version number; this version number may be explicitly
stated (typically if the user knows which version they need) or
automatically determined from the version of the existing changed path
filters in the repository.

There is a change in write_commit_graph(). graph_read_bloom_data()
makes it possible for chunk_bloom_data to be non-NULL but
bloom_filter_settings to be NULL, which causes a segfault later on. I
produced such a segfault while developing this patch, but couldn't find
a way to reproduce it neither after this complete patch (or before),
but in any case it seemed like a good thing to include that might help
future patch authors.

The value in t0095 was obtained from another murmur3 implementation
using the following Go source code:

  package main

  import "fmt"
  import "github.com/spaolacci/murmur3"

  func main() {
          fmt.Printf("%x\n", murmur3.Sum32([]byte("Hello world!")))
          fmt.Printf("%x\n", murmur3.Sum32([]byte{0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}))
  }

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 Documentation/config/commitgraph.txt |   5 +-
 bloom.c                              |  69 +++++++++++++++++-
 bloom.h                              |   8 +-
 commit-graph.c                       |  32 ++++++--
 t/helper/test-bloom.c                |   9 ++-
 t/t0095-bloom.sh                     |   8 ++
 t/t4216-log-bloom.sh                 | 105 +++++++++++++++++++++++++++
 7 files changed, 223 insertions(+), 13 deletions(-)

diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
index 2dc9170622..acc74a2f27 100644
--- a/Documentation/config/commitgraph.txt
+++ b/Documentation/config/commitgraph.txt
@@ -15,7 +15,7 @@ commitGraph.readChangedPaths::
 
 commitGraph.changedPathsVersion::
 	Specifies the version of the changed-path Bloom filters that Git will read and
-	write. May be -1, 0 or 1.
+	write. May be -1, 0, 1, or 2.
 +
 Defaults to -1.
 +
@@ -28,4 +28,7 @@ filters when instructed to write.
 If 1, Git will only read version 1 Bloom filters, and will write version 1
 Bloom filters.
 +
+If 2, Git will only read version 2 Bloom filters, and will write version 2
+Bloom filters.
++
 See linkgit:git-commit-graph[1] for more information.
diff --git a/bloom.c b/bloom.c
index 3e78cfe79d..ebef5cfd2f 100644
--- a/bloom.c
+++ b/bloom.c
@@ -66,7 +66,64 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
  * Not considered to be cryptographically secure.
  * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
  */
-uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
+uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len)
+{
+	const uint32_t c1 = 0xcc9e2d51;
+	const uint32_t c2 = 0x1b873593;
+	const uint32_t r1 = 15;
+	const uint32_t r2 = 13;
+	const uint32_t m = 5;
+	const uint32_t n = 0xe6546b64;
+	int i;
+	uint32_t k1 = 0;
+	const char *tail;
+
+	int len4 = len / sizeof(uint32_t);
+
+	uint32_t k;
+	for (i = 0; i < len4; i++) {
+		uint32_t byte1 = (uint32_t)(unsigned char)data[4*i];
+		uint32_t byte2 = ((uint32_t)(unsigned char)data[4*i + 1]) << 8;
+		uint32_t byte3 = ((uint32_t)(unsigned char)data[4*i + 2]) << 16;
+		uint32_t byte4 = ((uint32_t)(unsigned char)data[4*i + 3]) << 24;
+		k = byte1 | byte2 | byte3 | byte4;
+		k *= c1;
+		k = rotate_left(k, r1);
+		k *= c2;
+
+		seed ^= k;
+		seed = rotate_left(seed, r2) * m + n;
+	}
+
+	tail = (data + len4 * sizeof(uint32_t));
+
+	switch (len & (sizeof(uint32_t) - 1)) {
+	case 3:
+		k1 ^= ((uint32_t)(unsigned char)tail[2]) << 16;
+		/*-fallthrough*/
+	case 2:
+		k1 ^= ((uint32_t)(unsigned char)tail[1]) << 8;
+		/*-fallthrough*/
+	case 1:
+		k1 ^= ((uint32_t)(unsigned char)tail[0]) << 0;
+		k1 *= c1;
+		k1 = rotate_left(k1, r1);
+		k1 *= c2;
+		seed ^= k1;
+		break;
+	}
+
+	seed ^= (uint32_t)len;
+	seed ^= (seed >> 16);
+	seed *= 0x85ebca6b;
+	seed ^= (seed >> 13);
+	seed *= 0xc2b2ae35;
+	seed ^= (seed >> 16);
+
+	return seed;
+}
+
+static uint32_t murmur3_seeded_v1(uint32_t seed, const char *data, size_t len)
 {
 	const uint32_t c1 = 0xcc9e2d51;
 	const uint32_t c2 = 0x1b873593;
@@ -131,8 +188,14 @@ void fill_bloom_key(const char *data,
 	int i;
 	const uint32_t seed0 = 0x293ae76f;
 	const uint32_t seed1 = 0x7e646e2c;
-	const uint32_t hash0 = murmur3_seeded(seed0, data, len);
-	const uint32_t hash1 = murmur3_seeded(seed1, data, len);
+	uint32_t hash0, hash1;
+	if (settings->hash_version == 2) {
+		hash0 = murmur3_seeded_v2(seed0, data, len);
+		hash1 = murmur3_seeded_v2(seed1, data, len);
+	} else {
+		hash0 = murmur3_seeded_v1(seed0, data, len);
+		hash1 = murmur3_seeded_v1(seed1, data, len);
+	}
 
 	key->hashes = (uint32_t *)xcalloc(settings->num_hashes, sizeof(uint32_t));
 	for (i = 0; i < settings->num_hashes; i++)
diff --git a/bloom.h b/bloom.h
index 1e4f612d2c..138d57a86b 100644
--- a/bloom.h
+++ b/bloom.h
@@ -8,9 +8,11 @@ struct commit_graph;
 struct bloom_filter_settings {
 	/*
 	 * The version of the hashing technique being used.
-	 * We currently only support version = 1 which is
+	 * The newest version is 2, which is
 	 * the seeded murmur3 hashing technique implemented
-	 * in bloom.c.
+	 * in bloom.c. Bloom filters of version 1 were created
+	 * with prior versions of Git, which had a bug in the
+	 * implementation of the hash function.
 	 */
 	uint32_t hash_version;
 
@@ -80,7 +82,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
  * Not considered to be cryptographically secure.
  * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
  */
-uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len);
+uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len);
 
 void fill_bloom_key(const char *data,
 		    size_t len,
diff --git a/commit-graph.c b/commit-graph.c
index da99f15fdf..f7322c4fff 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -304,17 +304,26 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
 	return 0;
 }
 
+struct graph_read_bloom_data_context {
+	struct commit_graph *g;
+	int *commit_graph_changed_paths_version;
+};
+
 static int graph_read_bloom_data(const unsigned char *chunk_start,
 				  size_t chunk_size, void *data)
 {
-	struct commit_graph *g = data;
+	struct graph_read_bloom_data_context *c = data;
+	struct commit_graph *g = c->g;
 	uint32_t hash_version;
-	g->chunk_bloom_data = chunk_start;
 	hash_version = get_be32(chunk_start);
 
-	if (hash_version != 1)
+	if (*c->commit_graph_changed_paths_version == -1) {
+		*c->commit_graph_changed_paths_version = hash_version;
+	} else if (hash_version != *c->commit_graph_changed_paths_version) {
 		return 0;
+	}
 
+	g->chunk_bloom_data = chunk_start;
 	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
 	g->bloom_filter_settings->hash_version = hash_version;
 	g->bloom_filter_settings->num_hashes = get_be32(chunk_start + 4);
@@ -402,10 +411,14 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 	}
 
 	if (s->commit_graph_changed_paths_version) {
+		struct graph_read_bloom_data_context context = {
+			.g = graph,
+			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
+		};
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
-			   graph_read_bloom_data, graph);
+			   graph_read_bloom_data, &context);
 	}
 
 	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
@@ -2376,6 +2389,13 @@ int write_commit_graph(struct object_directory *odb,
 	}
 	if (!commit_graph_compatible(r))
 		return 0;
+	if (r->settings.commit_graph_changed_paths_version < -1
+	    || r->settings.commit_graph_changed_paths_version > 2) {
+		warning(_("attempting to write a commit-graph, but "
+			  "'commitgraph.changedPathsVersion' (%d) is not supported"),
+			r->settings.commit_graph_changed_paths_version);
+		return 0;
+	}
 
 	CALLOC_ARRAY(ctx, 1);
 	ctx->r = r;
@@ -2388,6 +2408,8 @@ int write_commit_graph(struct object_directory *odb,
 	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
 	ctx->num_generation_data_overflows = 0;
 
+	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
+		? 2 : 1;
 	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
 						      bloom_settings.bits_per_entry);
 	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
@@ -2417,7 +2439,7 @@ int write_commit_graph(struct object_directory *odb,
 		g = ctx->r->objects->commit_graph;
 
 		/* We have changed-paths already. Keep them in the next graph */
-		if (g && g->chunk_bloom_data) {
+		if (g && g->bloom_filter_settings) {
 			ctx->changed_paths = 1;
 			ctx->bloom_settings = g->bloom_filter_settings;
 		}
diff --git a/t/helper/test-bloom.c b/t/helper/test-bloom.c
index aabe31d724..3cbc0a5b50 100644
--- a/t/helper/test-bloom.c
+++ b/t/helper/test-bloom.c
@@ -50,6 +50,7 @@ static void get_bloom_filter_for_commit(const struct object_id *commit_oid)
 
 static const char *bloom_usage = "\n"
 "  test-tool bloom get_murmur3 <string>\n"
+"  test-tool bloom get_murmur3_seven_highbit\n"
 "  test-tool bloom generate_filter <string> [<string>...]\n"
 "  test-tool bloom get_filter_for_commit <commit-hex>\n";
 
@@ -64,7 +65,13 @@ int cmd__bloom(int argc, const char **argv)
 		uint32_t hashed;
 		if (argc < 3)
 			usage(bloom_usage);
-		hashed = murmur3_seeded(0, argv[2], strlen(argv[2]));
+		hashed = murmur3_seeded_v2(0, argv[2], strlen(argv[2]));
+		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
+	}
+
+	if (!strcmp(argv[1], "get_murmur3_seven_highbit")) {
+		uint32_t hashed;
+		hashed = murmur3_seeded_v2(0, "\x99\xaa\xbb\xcc\xdd\xee\xff", 7);
 		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
 	}
 
diff --git a/t/t0095-bloom.sh b/t/t0095-bloom.sh
index b567383eb8..c8d84ab606 100755
--- a/t/t0095-bloom.sh
+++ b/t/t0095-bloom.sh
@@ -29,6 +29,14 @@ test_expect_success 'compute unseeded murmur3 hash for test string 2' '
 	test_cmp expect actual
 '
 
+test_expect_success 'compute unseeded murmur3 hash for test string 3' '
+	cat >expect <<-\EOF &&
+	Murmur3 Hash with seed=0:0xa183ccfd
+	EOF
+	test-tool bloom get_murmur3_seven_highbit >actual &&
+	test_cmp expect actual
+'
+
 test_expect_success 'compute bloom key for empty string' '
 	cat >expect <<-\EOF &&
 	Hashes:0x5615800c|0x5b966560|0x61174ab4|0x66983008|0x6c19155c|0x7199fab0|0x771ae004|
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 0aa9719934..1d0e11d7c1 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -456,4 +456,109 @@ test_expect_success 'version 1 changed-path used when version 1 requested' '
 	)
 '
 
+test_expect_success 'version 1 changed-path not used when version 2 requested' '
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion 2 &&
+		test_bloom_filters_not_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'version 1 changed-path used when autodetect requested' '
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'when writing another commit graph, preserve existing version 1 of changed-path' '
+	test_commit -C highbit1 c1double "$CENT$CENT" &&
+	git -C highbit1 commit-graph write --reachable --changed-paths &&
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		echo "options: bloom(1,10,7) read_generation_data" >expect &&
+		test-tool read-graph >full &&
+		grep options full >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'set up repo with high bit path, version 2 changed-path' '
+	git init highbit2 &&
+	git -C highbit2 config --add commitgraph.changedPathsVersion 2 &&
+	test_commit -C highbit2 c2 "$CENT" &&
+	git -C highbit2 commit-graph write --reachable --changed-paths
+'
+
+test_expect_success 'check value of version 2 changed-path' '
+	(
+		cd highbit2 &&
+		echo "c01f" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'setup make another commit' '
+	# "git log" does not use Bloom filters for root commits - see how, in
+	# revision.c, rev_compare_tree() (the only code path that eventually calls
+	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
+	# has one parent. Therefore, make another commit so that we perform the tests on
+	# a non-root commit.
+	test_commit -C highbit2 anotherc2 "another$CENT"
+'
+
+test_expect_success 'version 2 changed-path used when version 2 requested' '
+	(
+		cd highbit2 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'version 2 changed-path not used when version 1 requested' '
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion 1 &&
+		test_bloom_filters_not_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'version 2 changed-path used when autodetect requested' '
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'when writing another commit graph, preserve existing version 2 of changed-path' '
+	test_commit -C highbit2 c2double "$CENT$CENT" &&
+	git -C highbit2 commit-graph write --reachable --changed-paths &&
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		echo "options: bloom(2,10,7) read_generation_data" >expect &&
+		test-tool read-graph >full &&
+		grep options full >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
+	git init doublewrite &&
+	test_commit -C doublewrite c "$CENT" &&
+	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
+	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
+	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	(
+		cd doublewrite &&
+		echo "c01f" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
 test_done
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 08/15] bloom: annotate filters with hash version

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

In subsequent commits, we will want to load existing Bloom filters out
of a commit-graph, even when the hash version they were computed with
does not match the value of `commitGraph.changedPathVersion`.

In order to differentiate between the two, add a "version" field to each
Bloom filter.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 bloom.c | 11 ++++++++---
 bloom.h |  1 +
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index ebef5cfd2f..9b6a30f6f6 100644
--- a/bloom.c
+++ b/bloom.c
@@ -55,6 +55,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
 	filter->data = (unsigned char *)(g->chunk_bloom_data +
 					sizeof(unsigned char) * start_index +
 					BLOOMDATA_CHUNK_HEADER_SIZE);
+	filter->version = g->bloom_filter_settings->hash_version;
 
 	return 1;
 }
@@ -240,11 +241,13 @@ static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 	return strcmp(e1->path, e2->path);
 }
 
-static void init_truncated_large_filter(struct bloom_filter *filter)
+static void init_truncated_large_filter(struct bloom_filter *filter,
+					int version)
 {
 	filter->data = xmalloc(1);
 	filter->data[0] = 0xFF;
 	filter->len = 1;
+	filter->version = version;
 }
 
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
@@ -329,13 +332,15 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 		}
 
 		if (hashmap_get_size(&pathmap) > settings->max_changed_paths) {
-			init_truncated_large_filter(filter);
+			init_truncated_large_filter(filter,
+						    settings->hash_version);
 			if (computed)
 				*computed |= BLOOM_TRUNC_LARGE;
 			goto cleanup;
 		}
 
 		filter->len = (hashmap_get_size(&pathmap) * settings->bits_per_entry + BITS_PER_WORD - 1) / BITS_PER_WORD;
+		filter->version = settings->hash_version;
 		if (!filter->len) {
 			if (computed)
 				*computed |= BLOOM_TRUNC_EMPTY;
@@ -355,7 +360,7 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 	} else {
 		for (i = 0; i < diff_queued_diff.nr; i++)
 			diff_free_filepair(diff_queued_diff.queue[i]);
-		init_truncated_large_filter(filter);
+		init_truncated_large_filter(filter, settings->hash_version);
 
 		if (computed)
 			*computed |= BLOOM_TRUNC_LARGE;
diff --git a/bloom.h b/bloom.h
index 138d57a86b..330a140520 100644
--- a/bloom.h
+++ b/bloom.h
@@ -55,6 +55,7 @@ struct bloom_filter_settings {
 struct bloom_filter {
 	unsigned char *data;
 	size_t len;
+	int version;
 };
 
 /*
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 09/15] bloom: prepare to discard incompatible Bloom filters

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

Callers use the inline `get_bloom_filter()` implementation as a thin
wrapper around `get_or_compute_bloom_filter()`. The former calls the
latter with a value of "0" for `compute_if_not_present`, making
`get_bloom_filter()` the default read-only path for fetching an existing
Bloom filter.

Callers expect the value returned from `get_bloom_filter()` is usable,
that is that it's compatible with the configured value corresponding to
`commitGraph.changedPathsVersion`.

This is OK, since the commit-graph machinery only initializes its BDAT
chunk (thereby enabling it to service Bloom filter queries) when the
Bloom filter hash_version is compatible with our settings. So any value
returned by `get_bloom_filter()` is trivially useable.

However, subsequent commits will load the BDAT chunk even when the Bloom
filters are built with incompatible hash versions. Prepare to handle
this by teaching `get_bloom_filter()` to discard filters that are
incompatible with the configured hash version.

Callers who wish to read incompatible filters (e.g., for upgrading
filters from v1 to v2) may use the lower level routine,
`get_or_compute_bloom_filter()`.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 bloom.c | 20 +++++++++++++++++++-
 bloom.h | 20 ++++++++++++++++++--
 2 files changed, 37 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index 9b6a30f6f6..739fa093ba 100644
--- a/bloom.c
+++ b/bloom.c
@@ -250,6 +250,23 @@ static void init_truncated_large_filter(struct bloom_filter *filter,
 	filter->version = version;
 }
 
+struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
+{
+	struct bloom_filter *filter;
+	int hash_version;
+
+	filter = get_or_compute_bloom_filter(r, c, 0, NULL, NULL);
+	if (!filter)
+		return NULL;
+
+	prepare_repo_settings(r);
+	hash_version = r->settings.commit_graph_changed_paths_version;
+
+	if (!(hash_version == -1 || hash_version == filter->version))
+		return NULL; /* unusable filter */
+	return filter;
+}
+
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						 struct commit *c,
 						 int compute_if_not_present,
@@ -275,7 +292,8 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						     filter, graph_pos);
 	}
 
-	if (filter->data && filter->len)
+	if ((filter->data && filter->len) &&
+	    (!settings || settings->hash_version == filter->version))
 		return filter;
 	if (!compute_if_not_present)
 		return NULL;
diff --git a/bloom.h b/bloom.h
index 330a140520..bfe389e29c 100644
--- a/bloom.h
+++ b/bloom.h
@@ -110,8 +110,24 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						 const struct bloom_filter_settings *settings,
 						 enum bloom_filter_computed *computed);
 
-#define get_bloom_filter(r, c) get_or_compute_bloom_filter( \
-	(r), (c), 0, NULL, NULL)
+/*
+ * Find the Bloom filter associated with the given commit "c".
+ *
+ * If any of the following are true
+ *
+ *   - the repository does not have a commit-graph, or
+ *   - the repository disables reading from the commit-graph, or
+ *   - the given commit does not have a Bloom filter computed, or
+ *   - there is a Bloom filter for commit "c", but it cannot be read
+ *     because the filter uses an incompatible version of murmur3
+ *
+ * , then `get_bloom_filter()` will return NULL. Otherwise, the corresponding
+ * Bloom filter will be returned.
+ *
+ * For callers who wish to inspect Bloom filters with incompatible hash
+ * versions, use get_or_compute_bloom_filter().
+ */
+struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c);
 
 int bloom_filter_contains(const struct bloom_filter *filter,
 			  const struct bloom_key *key,
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 10/15] t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

The existing implementation of test_bloom_filters_not_used() asserts
that the Bloom filter sub-system has not been initialized at all, by
checking for the absence of any data from it from trace2.

In the following commit, it will become possible to load Bloom filters
without using them (e.g., because `commitGraph.changedPathVersion` is
incompatible with the hash version with which the commit-graph's Bloom
filters were written).

When this is the case, it's possible to initialize the Bloom filter
sub-system, while still not using any Bloom filters. When this is the
case, check that the data dump from the Bloom sub-system is all zeros,
indicating that no filters were used.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 t/t4216-log-bloom.sh | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 1d0e11d7c1..940a71d8b8 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -81,7 +81,19 @@ test_bloom_filters_used () {
 test_bloom_filters_not_used () {
 	log_args=$1
 	setup "$log_args" &&
-	! grep -q "statistics:{\"filter_not_present\":" "$TRASH_DIRECTORY/trace.perf" &&
+
+	if grep -q "statistics:{\"filter_not_present\":" "$TRASH_DIRECTORY/trace.perf"
+	then
+		# if the Bloom filter system is initialized, ensure that no
+		# filters were used
+		data="statistics:{"
+		data="$data\"filter_not_present\":0,"
+		data="$data\"maybe\":0,"
+		data="$data\"definitely_not\":0,"
+		data="$data\"false_positive\":0}"
+
+		grep -q "$data" "$TRASH_DIRECTORY/trace.perf"
+	fi &&
 	test_cmp log_wo_bloom log_w_bloom
 }
 
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 11/15] commit-graph.c: unconditionally load Bloom filters

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

In 9e4df4da07 (commit-graph: new filter ver. that fixes murmur3,
2023-08-01), we began ignoring the Bloom data ("BDAT") chunk for
commit-graphs whose Bloom filters were computed using a hash version
incompatible with the value of `commitGraph.changedPathVersion`.

Now that the Bloom API has been hardened to discard these incompatible
filters (with the exception of low-level APIs), we can safely load these
Bloom filters unconditionally.

We no longer want to return early from `graph_read_bloom_data()`, and
similarly do not want to set the bloom_settings' `hash_version` field as
a side-effect. The latter is because we want to wait until we know which
Bloom settings we're using (either the defaults, from the GIT_TEST
variables, or from the previous commit-graph layer) before deciding what
hash_version to use.

If we detect an existing BDAT chunk, we'll infer the rest of the
settings (e.g., number of hashes, bits per entry, and maximum number of
changed paths) from the earlier graph layer. The hash_version will be
inferred from the previous layer as well, unless one has already been
specified via configuration.

Once all of that is done, we normalize the value of the hash_version to
either "1" or "2".

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 commit-graph.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/commit-graph.c b/commit-graph.c
index f7322c4fff..665a3edf78 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -317,12 +317,6 @@ static int graph_read_bloom_data(const unsigned char *chunk_start,
 	uint32_t hash_version;
 	hash_version = get_be32(chunk_start);
 
-	if (*c->commit_graph_changed_paths_version == -1) {
-		*c->commit_graph_changed_paths_version = hash_version;
-	} else if (hash_version != *c->commit_graph_changed_paths_version) {
-		return 0;
-	}
-
 	g->chunk_bloom_data = chunk_start;
 	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
 	g->bloom_filter_settings->hash_version = hash_version;
@@ -2408,8 +2402,7 @@ int write_commit_graph(struct object_directory *odb,
 	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
 	ctx->num_generation_data_overflows = 0;
 
-	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
-		? 2 : 1;
+	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version;
 	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
 						      bloom_settings.bits_per_entry);
 	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
@@ -2441,10 +2434,18 @@ int write_commit_graph(struct object_directory *odb,
 		/* We have changed-paths already. Keep them in the next graph */
 		if (g && g->bloom_filter_settings) {
 			ctx->changed_paths = 1;
-			ctx->bloom_settings = g->bloom_filter_settings;
+
+			/* don't propagate the hash_version unless unspecified */
+			if (bloom_settings.hash_version == -1)
+				bloom_settings.hash_version = g->bloom_filter_settings->hash_version;
+			bloom_settings.bits_per_entry = g->bloom_filter_settings->bits_per_entry;
+			bloom_settings.num_hashes = g->bloom_filter_settings->num_hashes;
+			bloom_settings.max_changed_paths = g->bloom_filter_settings->max_changed_paths;
 		}
 	}
 
+	bloom_settings.hash_version = bloom_settings.hash_version == 2 ? 2 : 1;
+
 	if (ctx->split) {
 		struct commit_graph *g = ctx->r->objects->commit_graph;
 
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 12/15] commit-graph: drop unnecessary `graph_read_bloom_data_context`

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

The `graph_read_bloom_data_context` struct was introduced in an earlier
commit in order to pass pointers to the commit-graph and changed-path
Bloom filter version when reading the BDAT chunk.

The previous commit no longer writes through the changed_paths_version
pointer, making the surrounding context structure unnecessary. Drop it
and pass a pointer to the commit-graph directly when reading the BDAT
chunk.

Noticed-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 commit-graph.c | 14 ++------------
 1 file changed, 2 insertions(+), 12 deletions(-)

diff --git a/commit-graph.c b/commit-graph.c
index 665a3edf78..a8e33c0739 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -304,16 +304,10 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
 	return 0;
 }
 
-struct graph_read_bloom_data_context {
-	struct commit_graph *g;
-	int *commit_graph_changed_paths_version;
-};
-
 static int graph_read_bloom_data(const unsigned char *chunk_start,
 				  size_t chunk_size, void *data)
 {
-	struct graph_read_bloom_data_context *c = data;
-	struct commit_graph *g = c->g;
+	struct commit_graph *g = data;
 	uint32_t hash_version;
 	hash_version = get_be32(chunk_start);
 
@@ -405,14 +399,10 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 	}
 
 	if (s->commit_graph_changed_paths_version) {
-		struct graph_read_bloom_data_context context = {
-			.g = graph,
-			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
-		};
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
-			   graph_read_bloom_data, &context);
+			   graph_read_bloom_data, graph);
 	}
 
 	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 13/15] object.h: fix mis-aligned flag bits table

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

Bit position 23 is one column too far to the left.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 object.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/object.h b/object.h
index 114d45954d..db25714b4e 100644
--- a/object.h
+++ b/object.h
@@ -62,7 +62,7 @@ void object_array_init(struct object_array *array);
 
 /*
  * object flag allocation:
- * revision.h:               0---------10         15             23------27
+ * revision.h:               0---------10         15               23------27
  * fetch-pack.c:             01    67
  * negotiator/default.c:       2--5
  * walker.c:                 0-2
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 14/15] commit-graph: reuse existing Bloom filters where possible

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

In 9e4df4da07 (commit-graph: new filter ver. that fixes murmur3,
2023-08-01), a bug was described where it's possible for Git to produce
non-murmur3 hashes when the platform's "char" type is signed, and there
are paths with characters whose highest bit is set (i.e. all characters
>= 0x80).

That patch allows the caller to control which version of Bloom filters
are read and written. However, even on platforms with a signed "char"
type, it is possible to reuse existing Bloom filters if and only if
there are no changed paths in any commit's first parent tree-diff whose
characters have their highest bit set.

When this is the case, we can reuse the existing filter without having
to compute a new one. This is done by marking trees which are known to
have (or not have) any such paths. When a commit's root tree is verified
to not have any such paths, we mark it as such and declare that the
commit's Bloom filter is reusable.

Note that this heuristic only goes in one direction. If neither a commit
nor its first parent have any paths in their trees with non-ASCII
characters, then we know for certain that a path with non-ASCII
characters will not appear in a tree-diff against that commit's first
parent. The reverse isn't necessarily true: just because the tree-diff
doesn't contain any such paths does not imply that no such paths exist
in either tree.

So we end up recomputing some Bloom filters that we don't strictly have
to (i.e. their bits are the same no matter which version of murmur3 we
use). But culling these out is impossible, since we'd have to perform
the full tree-diff, which is the same effort as computing the Bloom
filter from scratch.

But because we can cache our results in each tree's flag bits, we can
often avoid recomputing many filters, thereby reducing the time it takes
to run

    $ git commit-graph write --changed-paths --reachable

when upgrading from v1 to v2 Bloom filters.

To benchmark this, let's generate a commit-graph in linux.git with v1
changed-paths in generation order[^1]:

    $ git clone git@github.com:torvalds/linux.git
    $ cd linux
    $ git commit-graph write --reachable --changed-paths
    $ graph=".git/objects/info/commit-graph"
    $ mv $graph{,.bak}

Then let's time how long it takes to go from v1 to v2 filters (with and
without the upgrade path enabled), resetting the state of the
commit-graph each time:

    $ git config commitGraph.changedPathsVersion 2
    $ hyperfine -p 'cp -f $graph.bak $graph' -L v 0,1 \
        'GIT_TEST_UPGRADE_BLOOM_FILTERS={v} git.compile commit-graph write --reachable --changed-paths'

On linux.git (where there aren't any non-ASCII paths), the timings
indicate that this patch represents a speed-up over recomputing all
Bloom filters from scratch:

    Benchmark 1: GIT_TEST_UPGRADE_BLOOM_FILTERS=0 git.compile commit-graph write --reachable --changed-paths
      Time (mean ± σ):     124.873 s ±  0.316 s    [User: 124.081 s, System: 0.643 s]
      Range (min … max):   124.621 s … 125.227 s    3 runs

    Benchmark 2: GIT_TEST_UPGRADE_BLOOM_FILTERS=1 git.compile commit-graph write --reachable --changed-paths
      Time (mean ± σ):     79.271 s ±  0.163 s    [User: 74.611 s, System: 4.521 s]
      Range (min … max):   79.112 s … 79.437 s    3 runs

    Summary
      'GIT_TEST_UPGRADE_BLOOM_FILTERS=1 git.compile commit-graph write --reachable --changed-paths' ran
        1.58 ± 0.01 times faster than 'GIT_TEST_UPGRADE_BLOOM_FILTERS=0 git.compile commit-graph write --reachable --changed-paths'

On git.git, we do have some non-ASCII paths, giving us a more modest
improvement from 4.163 seconds to 3.348 seconds, for a 1.24x speed-up.
On my machine, the stats for git.git are:

  - 8,285 Bloom filters computed from scratch
  - 10 Bloom filters generated as empty
  - 4 Bloom filters generated as truncated due to too many changed paths
  - 65,114 Bloom filters were reused when transitioning from v1 to v2.

[^1]: Note that this is is important, since `--stdin-packs` or
  `--stdin-commits` orders commits in the commit-graph by their pack
  position (with `--stdin-packs`) or in the raw input (with
  `--stdin-commits`).

  Since we compute Bloom filters in the same order that commits appear
  in the graph, we must see a commit's (first) parent before we process
  the commit itself. This is only guaranteed to happen when sorting
  commits by their generation number.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 bloom.c              | 90 ++++++++++++++++++++++++++++++++++++++++++--
 bloom.h              |  1 +
 commit-graph.c       |  5 +++
 object.h             |  1 +
 t/t4216-log-bloom.sh | 35 ++++++++++++++++-
 5 files changed, 127 insertions(+), 5 deletions(-)

diff --git a/bloom.c b/bloom.c
index 739fa093ba..24dd874e46 100644
--- a/bloom.c
+++ b/bloom.c
@@ -7,6 +7,9 @@
 #include "commit-graph.h"
 #include "commit.h"
 #include "commit-slab.h"
+#include "tree.h"
+#include "tree-walk.h"
+#include "config.h"
 
 define_commit_slab(bloom_filter_slab, struct bloom_filter);
 
@@ -250,6 +253,73 @@ static void init_truncated_large_filter(struct bloom_filter *filter,
 	filter->version = version;
 }
 
+#define VISITED   (1u<<21)
+#define HIGH_BITS (1u<<22)
+
+static int has_entries_with_high_bit(struct repository *r, struct tree *t)
+{
+	if (parse_tree(t))
+		return 1;
+
+	if (!(t->object.flags & VISITED)) {
+		struct tree_desc desc;
+		struct name_entry entry;
+
+		init_tree_desc(&desc, t->buffer, t->size);
+		while (tree_entry(&desc, &entry)) {
+			size_t i;
+			for (i = 0; i < entry.pathlen; i++) {
+				if (entry.path[i] & 0x80) {
+					t->object.flags |= HIGH_BITS;
+					goto done;
+				}
+			}
+
+			if (S_ISDIR(entry.mode)) {
+				struct tree *sub = lookup_tree(r, &entry.oid);
+				if (sub && has_entries_with_high_bit(r, sub)) {
+					t->object.flags |= HIGH_BITS;
+					goto done;
+				}
+			}
+
+		}
+
+done:
+		t->object.flags |= VISITED;
+	}
+
+	return !!(t->object.flags & HIGH_BITS);
+}
+
+static int commit_tree_has_high_bit_paths(struct repository *r,
+					  struct commit *c)
+{
+	struct tree *t;
+	if (repo_parse_commit(r, c))
+		return 1;
+	t = repo_get_commit_tree(r, c);
+	if (!t)
+		return 1;
+	return has_entries_with_high_bit(r, t);
+}
+
+static struct bloom_filter *upgrade_filter(struct repository *r, struct commit *c,
+					   struct bloom_filter *filter,
+					   int hash_version)
+{
+	struct commit_list *p = c->parents;
+	if (commit_tree_has_high_bit_paths(r, c))
+		return NULL;
+
+	if (p && commit_tree_has_high_bit_paths(r, p->item))
+		return NULL;
+
+	filter->version = hash_version;
+
+	return filter;
+}
+
 struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
 {
 	struct bloom_filter *filter;
@@ -292,9 +362,23 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						     filter, graph_pos);
 	}
 
-	if ((filter->data && filter->len) &&
-	    (!settings || settings->hash_version == filter->version))
-		return filter;
+	if (filter->data && filter->len) {
+		struct bloom_filter *upgrade;
+		if (!settings || settings->hash_version == filter->version)
+			return filter;
+
+		/* version mismatch, see if we can upgrade */
+		if (compute_if_not_present &&
+		    git_env_bool("GIT_TEST_UPGRADE_BLOOM_FILTERS", 1)) {
+			upgrade = upgrade_filter(r, c, filter,
+						 settings->hash_version);
+			if (upgrade) {
+				if (computed)
+					*computed |= BLOOM_UPGRADED;
+				return upgrade;
+			}
+		}
+	}
 	if (!compute_if_not_present)
 		return NULL;
 
diff --git a/bloom.h b/bloom.h
index bfe389e29c..e3a9b68905 100644
--- a/bloom.h
+++ b/bloom.h
@@ -102,6 +102,7 @@ enum bloom_filter_computed {
 	BLOOM_COMPUTED     = (1 << 1),
 	BLOOM_TRUNC_LARGE  = (1 << 2),
 	BLOOM_TRUNC_EMPTY  = (1 << 3),
+	BLOOM_UPGRADED     = (1 << 4),
 };
 
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
diff --git a/commit-graph.c b/commit-graph.c
index a8e33c0739..a3473df515 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -1045,6 +1045,7 @@ struct write_commit_graph_context {
 	int count_bloom_filter_not_computed;
 	int count_bloom_filter_trunc_empty;
 	int count_bloom_filter_trunc_large;
+	int count_bloom_filter_upgraded;
 };
 
 static int write_graph_chunk_fanout(struct hashfile *f,
@@ -1651,6 +1652,8 @@ static void trace2_bloom_filter_write_statistics(struct write_commit_graph_conte
 			   ctx->count_bloom_filter_trunc_empty);
 	trace2_data_intmax("commit-graph", ctx->r, "filter-trunc-large",
 			   ctx->count_bloom_filter_trunc_large);
+	trace2_data_intmax("commit-graph", ctx->r, "filter-upgraded",
+			   ctx->count_bloom_filter_upgraded);
 }
 
 static void compute_bloom_filters(struct write_commit_graph_context *ctx)
@@ -1692,6 +1695,8 @@ static void compute_bloom_filters(struct write_commit_graph_context *ctx)
 				ctx->count_bloom_filter_trunc_empty++;
 			if (computed & BLOOM_TRUNC_LARGE)
 				ctx->count_bloom_filter_trunc_large++;
+		} else if (computed & BLOOM_UPGRADED) {
+			ctx->count_bloom_filter_upgraded++;
 		} else if (computed & BLOOM_NOT_COMPUTED)
 			ctx->count_bloom_filter_not_computed++;
 		ctx->total_bloom_filter_data_size += filter
diff --git a/object.h b/object.h
index db25714b4e..2e5e08725f 100644
--- a/object.h
+++ b/object.h
@@ -75,6 +75,7 @@ void object_array_init(struct object_array *array);
  * commit-reach.c:                                  16-----19
  * sha1-name.c:                                              20
  * list-objects-filter.c:                                      21
+ * bloom.c:                                                    2122
  * builtin/fsck.c:           0--3
  * builtin/gc.c:             0
  * builtin/index-pack.c:                                     2021
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 940a71d8b8..502115abc3 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -217,6 +217,10 @@ test_filter_trunc_large () {
 	grep "\"key\":\"filter-trunc-large\",\"value\":\"$1\"" $2
 }
 
+test_filter_upgraded () {
+	grep "\"key\":\"filter-upgraded\",\"value\":\"$1\"" $2
+}
+
 test_expect_success 'correctly report changes over limit' '
 	git init limits &&
 	(
@@ -561,10 +565,19 @@ test_expect_success 'when writing another commit graph, preserve existing versio
 test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
 	git init doublewrite &&
 	test_commit -C doublewrite c "$CENT" &&
+
 	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
-	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C doublewrite commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
 	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
-	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C doublewrite commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
 	(
 		cd doublewrite &&
 		echo "c01f" >expect &&
@@ -573,4 +586,22 @@ test_expect_success 'when writing commit graph, do not reuse changed-path of ano
 	)
 '
 
+test_expect_success 'when writing commit graph, reuse changed-path of another version where possible' '
+	git init upgrade &&
+
+	test_commit -C upgrade base no-high-bits &&
+
+	git -C upgrade config --add commitgraph.changedPathsVersion 1 &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C upgrade commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
+	git -C upgrade config --add commitgraph.changedPathsVersion 2 &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C upgrade commit-graph write --reachable --changed-paths &&
+	test_filter_computed 0 trace2.txt &&
+	test_filter_upgraded 1 trace2.txt
+'
+
 test_done
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

[PATCH v2 15/15] bloom: introduce `deinit_bloom_filters()`

[Jonathan Tan]

From: Taylor Blau <me@ttaylorr.com>

After we are done using Bloom filters, we do not currently clean up any
memory allocated by the commit slab used to store those filters in the
first place.

Besides the bloom_filter structures themselves, there is mostly nothing
to free() in the first place, since in the read-only path all Bloom
filter's `data` members point to a memory mapped region in the
commit-graph file itself.

But when generating Bloom filters from scratch (or initializing
truncated filters) we allocate additional memory to store the filter's
data.

Keep track of when we need to free() this additional chunk of memory by
using an extra pointer `to_free`. Most of the time this will be NULL
(indicating that we are representing an existing Bloom filter stored in
a memory mapped region). When it is non-NULL, free it before discarding
the Bloom filters slab.

Suggested-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 bloom.c        | 16 +++++++++++++++-
 bloom.h        |  3 +++
 commit-graph.c |  4 ++++
 3 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/bloom.c b/bloom.c
index 24dd874e46..ff131893cd 100644
--- a/bloom.c
+++ b/bloom.c
@@ -59,6 +59,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
 					sizeof(unsigned char) * start_index +
 					BLOOMDATA_CHUNK_HEADER_SIZE);
 	filter->version = g->bloom_filter_settings->hash_version;
+	filter->to_free = NULL;
 
 	return 1;
 }
@@ -231,6 +232,18 @@ void init_bloom_filters(void)
 	init_bloom_filter_slab(&bloom_filters);
 }
 
+static void free_one_bloom_filter(struct bloom_filter *filter)
+{
+	if (!filter)
+		return;
+	free(filter->to_free);
+}
+
+void deinit_bloom_filters(void)
+{
+	deep_clear_bloom_filter_slab(&bloom_filters, free_one_bloom_filter);
+}
+
 static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 		       const struct hashmap_entry *eptr,
 		       const struct hashmap_entry *entry_or_key,
@@ -247,7 +260,7 @@ static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 static void init_truncated_large_filter(struct bloom_filter *filter,
 					int version)
 {
-	filter->data = xmalloc(1);
+	filter->data = filter->to_free = xmalloc(1);
 	filter->data[0] = 0xFF;
 	filter->len = 1;
 	filter->version = version;
@@ -449,6 +462,7 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 			filter->len = 1;
 		}
 		CALLOC_ARRAY(filter->data, filter->len);
+		filter->to_free = filter->data;
 
 		hashmap_for_each_entry(&pathmap, &iter, e, entry) {
 			struct bloom_key key;
diff --git a/bloom.h b/bloom.h
index e3a9b68905..d20e64bfbb 100644
--- a/bloom.h
+++ b/bloom.h
@@ -56,6 +56,8 @@ struct bloom_filter {
 	unsigned char *data;
 	size_t len;
 	int version;
+
+	void *to_free;
 };
 
 /*
@@ -96,6 +98,7 @@ void add_key_to_filter(const struct bloom_key *key,
 		       const struct bloom_filter_settings *settings);
 
 void init_bloom_filters(void);
+void deinit_bloom_filters(void);
 
 enum bloom_filter_computed {
 	BLOOM_NOT_COMPUTED = (1 << 0),
diff --git a/commit-graph.c b/commit-graph.c
index a3473df515..585539da2f 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -723,6 +723,7 @@ static void close_commit_graph_one(struct commit_graph *g)
 void close_commit_graph(struct raw_object_store *o)
 {
 	close_commit_graph_one(o->commit_graph);
+	deinit_bloom_filters();
 	o->commit_graph = NULL;
 }
 
@@ -2523,6 +2524,9 @@ int write_commit_graph(struct object_directory *odb,
 
 	res = write_commit_graph_file(ctx);
 
+	if (ctx->changed_paths)
+		deinit_bloom_filters();
+
 	if (ctx->split)
 		mark_commit_graphs(ctx);
 
-- 
2.42.0.rc2.253.gd59a3bf2b4-goog

Re: [PATCH v2 00/15] bloom: changed-path Bloom filters v2

[Junio C Hamano]

Jonathan Tan <jonathantanmy@google.com> writes:

> Here's an updated patch set containing updates to the tests, making them
> run on non-root commits to ensure that we exercise the Bloom filters.

Thanks for a quick turnaround.

> Thanks to SZEDER Gábor for spotting this.

With this everybody involved are happy, I presume.  Thank you,
everybody, who's worked on polishing and assembling these patches.

[PATCH v3 00/17] bloom: changed-path Bloom filters v2 (& sundries)

[Taylor Blau]

(Rebased onto the tip of 'master', which is 3a06386e31 (The fifteenth
batch, 2023-10-04), at the time of writing).

This series is a reroll of the combined efforts of [1] and [2] to
introduce the v2 changed-path Bloom filters, which fixes a bug in our
existing implementation of murmur3 paths with non-ASCII characters (when
the "char" type is signed).

In large part, this is the same as the previous round. But this round
includes some extra bits that address issues pointed out by SZEDER
Gábor, which are:

  - not reading Bloom filters for root commits
  - corrupting Bloom filter reads by tweaking the filter settings
    between layers.

These issues were discussed in (among other places) [3], and [4],
respectively.

Thanks to Jonathan, Peff, and SZEDER who have helped a great deal in
assembling these patches. As usual, a range-diff is included below.
Thanks in advance for your
review!

[1]: https://lore.kernel.org/git/cover.1684790529.git.jonathantanmy@google.com/
[2]: https://lore.kernel.org/git/cover.1691426160.git.me@ttaylorr.com/
[3]: https://public-inbox.org/git/20201015132147.GB24954@szeder.dev/
[4]: https://lore.kernel.org/git/20230830200218.GA5147@szeder.dev/

Jonathan Tan (4):
  gitformat-commit-graph: describe version 2 of BDAT
  t4216: test changed path filters with high bit paths
  repo-settings: introduce commitgraph.changedPathsVersion
  commit-graph: new filter ver. that fixes murmur3

Taylor Blau (13):
  t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
  revision.c: consult Bloom filters for root commits
  commit-graph: ensure Bloom filters are read with consistent settings
  t/helper/test-read-graph.c: extract `dump_graph_info()`
  bloom.h: make `load_bloom_filter_from_graph()` public
  t/helper/test-read-graph: implement `bloom-filters` mode
  bloom: annotate filters with hash version
  bloom: prepare to discard incompatible Bloom filters
  commit-graph.c: unconditionally load Bloom filters
  commit-graph: drop unnecessary `graph_read_bloom_data_context`
  object.h: fix mis-aligned flag bits table
  commit-graph: reuse existing Bloom filters where possible
  bloom: introduce `deinit_bloom_filters()`

 Documentation/config/commitgraph.txt     |  26 ++-
 Documentation/gitformat-commit-graph.txt |   9 +-
 bloom.c                                  | 208 ++++++++++++++++-
 bloom.h                                  |  38 +++-
 commit-graph.c                           |  61 ++++-
 object.h                                 |   3 +-
 oss-fuzz/fuzz-commit-graph.c             |   2 +-
 repo-settings.c                          |   6 +-
 repository.h                             |   2 +-
 revision.c                               |  26 ++-
 t/helper/test-bloom.c                    |   9 +-
 t/helper/test-read-graph.c               |  67 ++++--
 t/t0095-bloom.sh                         |   8 +
 t/t4216-log-bloom.sh                     | 272 ++++++++++++++++++++++-
 14 files changed, 682 insertions(+), 55 deletions(-)

Range-diff against v2:
10:  002a06d1e9 !  1:  fe671d616c t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
    @@ Commit message
         indicating that no filters were used.
     
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## t/t4216-log-bloom.sh ##
     @@ t/t4216-log-bloom.sh: test_bloom_filters_used () {
 -:  ---------- >  2:  7d0fa93543 revision.c: consult Bloom filters for root commits
 -:  ---------- >  3:  2ecc0a2d58 commit-graph: ensure Bloom filters are read with consistent settings
 1:  5fa681b58e !  4:  17703ed89a gitformat-commit-graph: describe version 2 of BDAT
    @@ Commit message
         Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
         Signed-off-by: Junio C Hamano <gitster@pobox.com>
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## Documentation/gitformat-commit-graph.txt ##
     @@ Documentation/gitformat-commit-graph.txt: All multi-byte numbers are in network byte order.
 2:  623d840575 !  5:  94552abf45 t/helper/test-read-graph.c: extract `dump_graph_info()`
    @@ Commit message
         Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
         Signed-off-by: Junio C Hamano <gitster@pobox.com>
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## t/helper/test-read-graph.c ##
     @@
 3:  bc9d77ae60 !  6:  3d81efa27b bloom.h: make `load_bloom_filter_from_graph()` public
    @@ Commit message
         Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
         Signed-off-by: Junio C Hamano <gitster@pobox.com>
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## bloom.c ##
     @@ bloom.c: static inline unsigned char get_bitmask(uint32_t pos)
 4:  ac7008aed3 !  7:  d23cd89037 t/helper/test-read-graph: implement `bloom-filters` mode
    @@ Commit message
         Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
         Signed-off-by: Junio C Hamano <gitster@pobox.com>
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## t/helper/test-read-graph.c ##
     @@ t/helper/test-read-graph.c: static void dump_graph_info(struct commit_graph *graph)
    @@ t/helper/test-read-graph.c: int cmd__read_graph(int argc UNUSED, const char **ar
     -	return 0;
     +	return ret;
      }
    ++
    ++
 5:  71755ba856 !  8:  cba766f224 t4216: test changed path filters with high bit paths
    @@ Commit message
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## t/t4216-log-bloom.sh ##
    -@@ t/t4216-log-bloom.sh: test_expect_success 'Bloom generation backfills empty commits' '
    - 	)
    +@@ t/t4216-log-bloom.sh: test_expect_success 'merge graph layers with incompatible Bloom settings' '
    + 	! grep "disabling Bloom filters" err
      '
      
     +get_first_changed_path_filter () {
 6:  9768d92c0f !  9:  a08a961f41 repo-settings: introduce commitgraph.changedPathsVersion
    @@ Commit message
         Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
         Signed-off-by: Junio C Hamano <gitster@pobox.com>
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## Documentation/config/commitgraph.txt ##
     @@ Documentation/config/commitgraph.txt: commitGraph.maxNewFilters::
 7:  f911b4bfab = 10:  61d44519a5 commit-graph: new filter ver. that fixes murmur3
 8:  35009900df ! 11:  a8c10f8de8 bloom: annotate filters with hash version
    @@ Commit message
         Bloom filter.
     
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## bloom.c ##
     @@ bloom.c: int load_bloom_filter_from_graph(struct commit_graph *g,
 9:  138bc16905 ! 12:  2ba10a4b4b bloom: prepare to discard incompatible Bloom filters
    @@ Commit message
         `get_or_compute_bloom_filter()`.
     
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## bloom.c ##
     @@ bloom.c: static void init_truncated_large_filter(struct bloom_filter *filter,
11:  2437e62813 ! 13:  09d8669c3a commit-graph.c: unconditionally load Bloom filters
    @@ Commit message
         either "1" or "2".
     
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## commit-graph.c ##
     @@ commit-graph.c: static int graph_read_bloom_data(const unsigned char *chunk_start,
12:  fe8fb2f5fe ! 14:  0d4f9dc4ee commit-graph: drop unnecessary `graph_read_bloom_data_context`
    @@ Commit message
     
         Noticed-by: Jonathan Tan <jonathantanmy@google.com>
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## commit-graph.c ##
     @@ commit-graph.c: static int graph_read_oid_lookup(const unsigned char *chunk_start,
13:  825af91e11 ! 15:  1f7f27bc47 object.h: fix mis-aligned flag bits table
    @@ Commit message
         Bit position 23 is one column too far to the left.
     
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## object.h ##
     @@ object.h: void object_array_init(struct object_array *array);
14:  593b317192 ! 16:  abbef95ae8 commit-graph: reuse existing Bloom filters where possible
    @@ Commit message
           commits by their generation number.
     
         Signed-off-by: Taylor Blau <me@ttaylorr.com>
    -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
    -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
     
      ## bloom.c ##
     @@
15:  8bf2c9cf98 = 17:  ca362408d5 bloom: introduce `deinit_bloom_filters()`
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 01/17] t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`

[Taylor Blau]

The existing implementation of test_bloom_filters_not_used() asserts
that the Bloom filter sub-system has not been initialized at all, by
checking for the absence of any data from it from trace2.

In the following commit, it will become possible to load Bloom filters
without using them (e.g., because `commitGraph.changedPathVersion` is
incompatible with the hash version with which the commit-graph's Bloom
filters were written).

When this is the case, it's possible to initialize the Bloom filter
sub-system, while still not using any Bloom filters. When this is the
case, check that the data dump from the Bloom sub-system is all zeros,
indicating that no filters were used.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/t4216-log-bloom.sh | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index fa9d32facf..487fc3d6b9 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -81,7 +81,19 @@ test_bloom_filters_used () {
 test_bloom_filters_not_used () {
 	log_args=$1
 	setup "$log_args" &&
-	! grep -q "statistics:{\"filter_not_present\":" "$TRASH_DIRECTORY/trace.perf" &&
+
+	if grep -q "statistics:{\"filter_not_present\":" "$TRASH_DIRECTORY/trace.perf"
+	then
+		# if the Bloom filter system is initialized, ensure that no
+		# filters were used
+		data="statistics:{"
+		data="$data\"filter_not_present\":0,"
+		data="$data\"maybe\":0,"
+		data="$data\"definitely_not\":0,"
+		data="$data\"false_positive\":0}"
+
+		grep -q "$data" "$TRASH_DIRECTORY/trace.perf"
+	fi &&
 	test_cmp log_wo_bloom log_w_bloom
 }
 
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 02/17] revision.c: consult Bloom filters for root commits

[Taylor Blau]

The commit-graph stores changed-path Bloom filters which represent the
set of paths included in a tree-level diff between a commit's root tree
and that of its parent.

When a commit has no parents, the tree-diff is computed against that
commit's root tree and the empty tree. In other words, every path in
that commit's tree is stored in the Bloom filter (since they all appear
in the diff).

Consult these filters during pathspec-limited traversals in the function
`rev_same_tree_as_empty()`. Doing so yields a performance improvement
where we can avoid enumerating the full set of paths in a parentless
commit's root tree when we know that the path(s) of interest were not
listed in that commit's changed-path Bloom filter.

Suggested-by: SZEDER Gábor <szeder.dev@gmail.com>
Original-patch-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 revision.c           | 26 ++++++++++++++++++++++----
 t/t4216-log-bloom.sh |  8 ++++++--
 2 files changed, 28 insertions(+), 6 deletions(-)

diff --git a/revision.c b/revision.c
index e789834dd1..dae569a547 100644
--- a/revision.c
+++ b/revision.c
@@ -834,17 +834,28 @@ static int rev_compare_tree(struct rev_info *revs,
 	return tree_difference;
 }
 
-static int rev_same_tree_as_empty(struct rev_info *revs, struct commit *commit)
+static int rev_same_tree_as_empty(struct rev_info *revs, struct commit *commit,
+				  int nth_parent)
 {
 	struct tree *t1 = repo_get_commit_tree(the_repository, commit);
+	int bloom_ret = 1;
 
 	if (!t1)
 		return 0;
 
+	if (nth_parent == 1 && revs->bloom_keys_nr) {
+		bloom_ret = check_maybe_different_in_bloom_filter(revs, commit);
+		if (!bloom_ret)
+			return 1;
+	}
+
 	tree_difference = REV_TREE_SAME;
 	revs->pruning.flags.has_changes = 0;
 	diff_tree_oid(NULL, &t1->object.oid, "", &revs->pruning);
 
+	if (bloom_ret == 1 && tree_difference == REV_TREE_SAME)
+		count_bloom_filter_false_positive++;
+
 	return tree_difference == REV_TREE_SAME;
 }
 
@@ -882,7 +893,7 @@ static int compact_treesame(struct rev_info *revs, struct commit *commit, unsign
 		if (nth_parent != 0)
 			die("compact_treesame %u", nth_parent);
 		old_same = !!(commit->object.flags & TREESAME);
-		if (rev_same_tree_as_empty(revs, commit))
+		if (rev_same_tree_as_empty(revs, commit, nth_parent))
 			commit->object.flags |= TREESAME;
 		else
 			commit->object.flags &= ~TREESAME;
@@ -978,7 +989,14 @@ static void try_to_simplify_commit(struct rev_info *revs, struct commit *commit)
 		return;
 
 	if (!commit->parents) {
-		if (rev_same_tree_as_empty(revs, commit))
+		/*
+		 * Pretend as if we are comparing ourselves to the
+		 * (non-existent) first parent of this commit object. Even
+		 * though no such parent exists, its changed-path Bloom filter
+		 * (if one exists) is relative to the empty tree, using Bloom
+		 * filters is allowed here.
+		 */
+		if (rev_same_tree_as_empty(revs, commit, 1))
 			commit->object.flags |= TREESAME;
 		return;
 	}
@@ -1059,7 +1077,7 @@ static void try_to_simplify_commit(struct rev_info *revs, struct commit *commit)
 
 		case REV_TREE_NEW:
 			if (revs->remove_empty_trees &&
-			    rev_same_tree_as_empty(revs, p)) {
+			    rev_same_tree_as_empty(revs, p, nth_parent)) {
 				/* We are adding all the specified
 				 * paths from this parent, so the
 				 * history beyond this parent is not
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 487fc3d6b9..322640feeb 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -87,7 +87,11 @@ test_bloom_filters_not_used () {
 		# if the Bloom filter system is initialized, ensure that no
 		# filters were used
 		data="statistics:{"
-		data="$data\"filter_not_present\":0,"
+		# unusable filters (e.g., those computed with a
+		# different value of commitGraph.changedPathsVersion)
+		# are counted in the filter_not_present bucket, so any
+		# value is OK there.
+		data="$data\"filter_not_present\":[0-9][0-9]*,"
 		data="$data\"maybe\":0,"
 		data="$data\"definitely_not\":0,"
 		data="$data\"false_positive\":0}"
@@ -174,7 +178,7 @@ test_expect_success 'setup - add commit-graph to the chain with Bloom filters' '
 
 test_bloom_filters_used_when_some_filters_are_missing () {
 	log_args=$1
-	bloom_trace_prefix="statistics:{\"filter_not_present\":3,\"maybe\":6,\"definitely_not\":9"
+	bloom_trace_prefix="statistics:{\"filter_not_present\":3,\"maybe\":6,\"definitely_not\":10"
 	setup "$log_args" &&
 	grep -q "$bloom_trace_prefix" "$TRASH_DIRECTORY/trace.perf" &&
 	test_cmp log_wo_bloom log_w_bloom
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 03/17] commit-graph: ensure Bloom filters are read with consistent settings

[Taylor Blau]

The changed-path Bloom filter mechanism is parameterized by a couple of
variables, notably the number of bits per hash (typically "m" in Bloom
filter literature) and the number of hashes themselves (typically "k").

It is critically important that filters are read with the Bloom filter
settings that they were written with. Failing to do so would mean that
each query is liable to compute different fingerprints, meaning that the
filter itself could return a false negative. This goes against a basic
assumption of using Bloom filters (that they may return false positives,
but never false negatives) and can lead to incorrect results.

We have some existing logic to carry forward existing Bloom filter
settings from one layer to the next. In `write_commit_graph()`, we have
something like:

    if (!(flags & COMMIT_GRAPH_NO_WRITE_BLOOM_FILTERS)) {
        struct commit_graph *g = ctx->r->objects->commit_graph;

        /* We have changed-paths already. Keep them in the next graph */
        if (g && g->chunk_bloom_data) {
            ctx->changed_paths = 1;
            ctx->bloom_settings = g->bloom_filter_settings;
        }
    }

, which drags forward Bloom filter settings across adjacent layers.

This doesn't quite address all cases, however, since it is possible for
intermediate layers to contain no Bloom filters at all. For example,
suppose we have two layers in a commit-graph chain, say, {G1, G2}. If G1
contains Bloom filters, but G2 doesn't, a new G3 (whose base graph is
G2) may be written with arbitrary Bloom filter settings, because we only
check the immediately adjacent layer's settings for compatibility.

This behavior has existed since the introduction of changed-path Bloom
filters. But in practice, this is not such a big deal, since the only
way up until this point to modify the Bloom filter settings at write
time is with the undocumented environment variables:

  - GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY
  - GIT_TEST_BLOOM_SETTINGS_NUM_HASHES
  - GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS

(it is still possible to tweak MAX_CHANGED_PATHS between layers, but
this does not affect reads, so is allowed to differ across multiple
graph layers).

But in future commits, we will introduce another parameter to change the
hash algorithm used to compute Bloom fingerprints itself. This will be
exposed via a configuration setting, making this foot-gun easier to use.

To prevent this potential issue, validate that all layers of a split
commit-graph have compatible settings with the newest layer which
contains Bloom filters.

Reported-by: SZEDER Gábor <szeder.dev@gmail.com>
Original-test-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 commit-graph.c       | 25 +++++++++++++++++
 t/t4216-log-bloom.sh | 64 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 89 insertions(+)

diff --git a/commit-graph.c b/commit-graph.c
index 1a56efcf69..ae0902f7f4 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -498,6 +498,30 @@ static int validate_mixed_generation_chain(struct commit_graph *g)
 	return 0;
 }
 
+static void validate_mixed_bloom_settings(struct commit_graph *g)
+{
+	struct bloom_filter_settings *settings = NULL;
+	for (; g; g = g->base_graph) {
+		if (!g->bloom_filter_settings)
+			continue;
+		if (!settings) {
+			settings = g->bloom_filter_settings;
+			continue;
+		}
+
+		if (g->bloom_filter_settings->bits_per_entry != settings->bits_per_entry ||
+		    g->bloom_filter_settings->num_hashes != settings->num_hashes) {
+			g->chunk_bloom_indexes = NULL;
+			g->chunk_bloom_data = NULL;
+			FREE_AND_NULL(g->bloom_filter_settings);
+
+			warning(_("disabling Bloom filters for commit-graph "
+				  "layer '%s' due to incompatible settings"),
+				oid_to_hex(&g->oid));
+		}
+	}
+}
+
 static int add_graph_to_chain(struct commit_graph *g,
 			      struct commit_graph *chain,
 			      struct object_id *oids,
@@ -614,6 +638,7 @@ struct commit_graph *load_commit_graph_chain_fd_st(struct repository *r,
 	}
 
 	validate_mixed_generation_chain(graph_chain);
+	validate_mixed_bloom_settings(graph_chain);
 
 	free(oids);
 	fclose(fp);
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 322640feeb..f49a8f2fbf 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -420,4 +420,68 @@ test_expect_success 'Bloom generation backfills empty commits' '
 	)
 '
 
+graph=.git/objects/info/commit-graph
+graphdir=.git/objects/info/commit-graphs
+chain=$graphdir/commit-graph-chain
+
+test_expect_success 'setup for mixed Bloom setting tests' '
+	repo=mixed-bloom-settings &&
+
+	git init $repo &&
+	for i in one two three
+	do
+		test_commit -C $repo $i file || return 1
+	done
+'
+
+test_expect_success 'split' '
+	# Compute Bloom filters with "unusual" settings.
+	git -C $repo rev-parse one >in &&
+	GIT_TEST_BLOOM_SETTINGS_NUM_HASHES=3 git -C $repo commit-graph write \
+		--stdin-commits --changed-paths --split <in &&
+	layer=$(head -n 1 $repo/$chain) &&
+
+	# A commit-graph layer without Bloom filters "hides" the layers
+	# below ...
+	git -C $repo rev-parse two >in &&
+	git -C $repo commit-graph write --stdin-commits --no-changed-paths \
+		--split=no-merge <in &&
+
+	# Another commit-graph layer that has Bloom filters, but with
+	# standard settings, and is thus incompatible with the base
+	# layer written above.
+	git -C $repo rev-parse HEAD >in &&
+	git -C $repo commit-graph write --stdin-commits --changed-paths \
+		--split=no-merge <in &&
+
+	test_line_count = 3 $repo/$chain &&
+
+	# Ensure that incompatible Bloom filters are ignored.
+	git -C $repo -c core.commitGraph=false log --oneline --no-decorate -- file \
+		>expect 2>err &&
+	git -C $repo log --oneline --no-decorate -- file >actual 2>err &&
+	test_cmp expect actual &&
+	grep "disabling Bloom filters for commit-graph layer .$layer." err
+'
+
+test_expect_success 'merge graph layers with incompatible Bloom settings' '
+	# Ensure that incompatible Bloom filters are ignored when
+	# generating new layers.
+	git -C $repo commit-graph write --reachable --changed-paths 2>err &&
+	grep "disabling Bloom filters for commit-graph layer .$layer." err &&
+
+	test_path_is_file $repo/$graph &&
+	test_dir_is_empty $repo/$graphdir &&
+
+	# ...and merging existing ones.
+	git -C $repo -c core.commitGraph=false log --oneline --no-decorate -- file \
+		>expect 2>err &&
+	GIT_TRACE2_PERF="$(pwd)/trace.perf" \
+		git -C $repo log --oneline --no-decorate -- file >actual 2>err &&
+
+	test_cmp expect actual && cat err &&
+	grep "statistics:{\"filter_not_present\":0" trace.perf &&
+	! grep "disabling Bloom filters" err
+'
+
 test_done
-- 
2.42.0.342.g8bb3a896ee

Re: [PATCH v3 03/17] commit-graph: ensure Bloom filters are read with consistent settings

[Patrick Steinhardt]

[-- Attachment #1: Type: text/plain, Size: 8299 bytes --]

On Tue, Oct 10, 2023 at 04:33:26PM -0400, Taylor Blau wrote:
> The changed-path Bloom filter mechanism is parameterized by a couple of
> variables, notably the number of bits per hash (typically "m" in Bloom
> filter literature) and the number of hashes themselves (typically "k").
> 
> It is critically important that filters are read with the Bloom filter
> settings that they were written with. Failing to do so would mean that
> each query is liable to compute different fingerprints, meaning that the
> filter itself could return a false negative. This goes against a basic
> assumption of using Bloom filters (that they may return false positives,
> but never false negatives) and can lead to incorrect results.
> 
> We have some existing logic to carry forward existing Bloom filter
> settings from one layer to the next. In `write_commit_graph()`, we have
> something like:
> 
>     if (!(flags & COMMIT_GRAPH_NO_WRITE_BLOOM_FILTERS)) {
>         struct commit_graph *g = ctx->r->objects->commit_graph;
> 
>         /* We have changed-paths already. Keep them in the next graph */
>         if (g && g->chunk_bloom_data) {
>             ctx->changed_paths = 1;
>             ctx->bloom_settings = g->bloom_filter_settings;
>         }
>     }
> 
> , which drags forward Bloom filter settings across adjacent layers.
> 
> This doesn't quite address all cases, however, since it is possible for
> intermediate layers to contain no Bloom filters at all. For example,
> suppose we have two layers in a commit-graph chain, say, {G1, G2}. If G1
> contains Bloom filters, but G2 doesn't, a new G3 (whose base graph is
> G2) may be written with arbitrary Bloom filter settings, because we only
> check the immediately adjacent layer's settings for compatibility.
> 
> This behavior has existed since the introduction of changed-path Bloom
> filters. But in practice, this is not such a big deal, since the only
> way up until this point to modify the Bloom filter settings at write
> time is with the undocumented environment variables:
> 
>   - GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY
>   - GIT_TEST_BLOOM_SETTINGS_NUM_HASHES
>   - GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS
> 
> (it is still possible to tweak MAX_CHANGED_PATHS between layers, but
> this does not affect reads, so is allowed to differ across multiple
> graph layers).
> 
> But in future commits, we will introduce another parameter to change the
> hash algorithm used to compute Bloom fingerprints itself. This will be
> exposed via a configuration setting, making this foot-gun easier to use.
> 
> To prevent this potential issue, validate that all layers of a split
> commit-graph have compatible settings with the newest layer which
> contains Bloom filters.
> 
> Reported-by: SZEDER Gábor <szeder.dev@gmail.com>
> Original-test-by: SZEDER Gábor <szeder.dev@gmail.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> ---
>  commit-graph.c       | 25 +++++++++++++++++
>  t/t4216-log-bloom.sh | 64 ++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 89 insertions(+)
> 
> diff --git a/commit-graph.c b/commit-graph.c
> index 1a56efcf69..ae0902f7f4 100644
> --- a/commit-graph.c
> +++ b/commit-graph.c
> @@ -498,6 +498,30 @@ static int validate_mixed_generation_chain(struct commit_graph *g)
>  	return 0;
>  }
>  
> +static void validate_mixed_bloom_settings(struct commit_graph *g)
> +{
> +	struct bloom_filter_settings *settings = NULL;
> +	for (; g; g = g->base_graph) {
> +		if (!g->bloom_filter_settings)
> +			continue;
> +		if (!settings) {
> +			settings = g->bloom_filter_settings;
> +			continue;
> +		}
> +
> +		if (g->bloom_filter_settings->bits_per_entry != settings->bits_per_entry ||
> +		    g->bloom_filter_settings->num_hashes != settings->num_hashes) {
> +			g->chunk_bloom_indexes = NULL;
> +			g->chunk_bloom_data = NULL;
> +			FREE_AND_NULL(g->bloom_filter_settings);
> +
> +			warning(_("disabling Bloom filters for commit-graph "
> +				  "layer '%s' due to incompatible settings"),
> +				oid_to_hex(&g->oid));
> +		}
> +	}
> +}
> +
>  static int add_graph_to_chain(struct commit_graph *g,
>  			      struct commit_graph *chain,
>  			      struct object_id *oids,
> @@ -614,6 +638,7 @@ struct commit_graph *load_commit_graph_chain_fd_st(struct repository *r,
>  	}
>  
>  	validate_mixed_generation_chain(graph_chain);
> +	validate_mixed_bloom_settings(graph_chain);
>  
>  	free(oids);
>  	fclose(fp);
> diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
> index 322640feeb..f49a8f2fbf 100755
> --- a/t/t4216-log-bloom.sh
> +++ b/t/t4216-log-bloom.sh
> @@ -420,4 +420,68 @@ test_expect_success 'Bloom generation backfills empty commits' '
>  	)
>  '
>  
> +graph=.git/objects/info/commit-graph
> +graphdir=.git/objects/info/commit-graphs
> +chain=$graphdir/commit-graph-chain
> +
> +test_expect_success 'setup for mixed Bloom setting tests' '
> +	repo=mixed-bloom-settings &&
> +
> +	git init $repo &&
> +	for i in one two three
> +	do
> +		test_commit -C $repo $i file || return 1
> +	done
> +'
> +
> +test_expect_success 'split' '
> +	# Compute Bloom filters with "unusual" settings.
> +	git -C $repo rev-parse one >in &&
> +	GIT_TEST_BLOOM_SETTINGS_NUM_HASHES=3 git -C $repo commit-graph write \
> +		--stdin-commits --changed-paths --split <in &&
> +	layer=$(head -n 1 $repo/$chain) &&
> +
> +	# A commit-graph layer without Bloom filters "hides" the layers
> +	# below ...
> +	git -C $repo rev-parse two >in &&
> +	git -C $repo commit-graph write --stdin-commits --no-changed-paths \
> +		--split=no-merge <in &&
> +
> +	# Another commit-graph layer that has Bloom filters, but with
> +	# standard settings, and is thus incompatible with the base
> +	# layer written above.
> +	git -C $repo rev-parse HEAD >in &&
> +	git -C $repo commit-graph write --stdin-commits --changed-paths \
> +		--split=no-merge <in &&
> +
> +	test_line_count = 3 $repo/$chain &&
> +
> +	# Ensure that incompatible Bloom filters are ignored.
> +	git -C $repo -c core.commitGraph=false log --oneline --no-decorate -- file \
> +		>expect 2>err &&
> +	git -C $repo log --oneline --no-decorate -- file >actual 2>err &&
> +	test_cmp expect actual &&
> +	grep "disabling Bloom filters for commit-graph layer .$layer." err
> +'

Up to this point everything looks sensible to me.

> +test_expect_success 'merge graph layers with incompatible Bloom settings' '
> +	# Ensure that incompatible Bloom filters are ignored when
> +	# generating new layers.
> +	git -C $repo commit-graph write --reachable --changed-paths 2>err &&
> +	grep "disabling Bloom filters for commit-graph layer .$layer." err &&
> +
> +	test_path_is_file $repo/$graph &&
> +	test_dir_is_empty $repo/$graphdir &&
> +
> +	# ...and merging existing ones.
> +	git -C $repo -c core.commitGraph=false log --oneline --no-decorate -- file \
> +		>expect 2>err &&
> +	GIT_TRACE2_PERF="$(pwd)/trace.perf" \
> +		git -C $repo log --oneline --no-decorate -- file >actual 2>err &&

But this test is a bit confusing to me, to be honest, also because the
comment for the second block here reads funny. We don't really merge
anything, do we? We only generate logs and compare that the log with and
without the resulting merged commit graph is the same. The actual logic
happened before.

> +	test_cmp expect actual && cat err &&

The `cat err` looks like a leftover from debugging.

> +	grep "statistics:{\"filter_not_present\":0" trace.perf &&

Also, why should the filter not be present here? If we merge the
commit-graphs with `--changed-paths` I'd have expected that we either
carry over bloom filters from preexisting commit graphs if compatible,
or otherwise generate them if they are either incompatible or don't
exist.

I feel like I'm missing something obvious, so this may be me just
missing the bigger picture.

> +	! grep "disabling Bloom filters" err

Can we make this assertion stricter and verify that `err` is empty? I
always think that `! grep` is quite a fragile pattern as it is quite
prone to becoming stale, e.g. when the error message itself would
change.

Patrick

> +'
> +
>  test_done
> -- 
> 2.42.0.342.g8bb3a896ee
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[PATCH v3 04/17] gitformat-commit-graph: describe version 2 of BDAT

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

The code change to Git to support version 2 will be done in subsequent
commits.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 Documentation/gitformat-commit-graph.txt | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/Documentation/gitformat-commit-graph.txt b/Documentation/gitformat-commit-graph.txt
index 31cad585e2..3e906e8030 100644
--- a/Documentation/gitformat-commit-graph.txt
+++ b/Documentation/gitformat-commit-graph.txt
@@ -142,13 +142,16 @@ All multi-byte numbers are in network byte order.
 
 ==== Bloom Filter Data (ID: {'B', 'D', 'A', 'T'}) [Optional]
     * It starts with header consisting of three unsigned 32-bit integers:
-      - Version of the hash algorithm being used. We currently only support
-	value 1 which corresponds to the 32-bit version of the murmur3 hash
+      - Version of the hash algorithm being used. We currently support
+	value 2 which corresponds to the 32-bit version of the murmur3 hash
 	implemented exactly as described in
 	https://en.wikipedia.org/wiki/MurmurHash#Algorithm and the double
 	hashing technique using seed values 0x293ae76f and 0x7e646e2 as
 	described in https://doi.org/10.1007/978-3-540-30494-4_26 "Bloom Filters
-	in Probabilistic Verification"
+	in Probabilistic Verification". Version 1 Bloom filters have a bug that appears
+	when char is signed and the repository has path names that have characters >=
+	0x80; Git supports reading and writing them, but this ability will be removed
+	in a future version of Git.
       - The number of times a path is hashed and hence the number of bit positions
 	      that cumulatively determine whether a file is present in the commit.
       - The minimum number of bits 'b' per entry in the Bloom filter. If the filter
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 05/17] t/helper/test-read-graph.c: extract `dump_graph_info()`

[Taylor Blau]

Prepare for the 'read-graph' test helper to perform other tasks besides
dumping high-level information about the commit-graph by extracting its
main routine into a separate function.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/helper/test-read-graph.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)

diff --git a/t/helper/test-read-graph.c b/t/helper/test-read-graph.c
index 8c7a83f578..3375392f6c 100644
--- a/t/helper/test-read-graph.c
+++ b/t/helper/test-read-graph.c
@@ -5,20 +5,8 @@
 #include "bloom.h"
 #include "setup.h"
 
-int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+static void dump_graph_info(struct commit_graph *graph)
 {
-	struct commit_graph *graph = NULL;
-	struct object_directory *odb;
-
-	setup_git_directory();
-	odb = the_repository->objects->odb;
-
-	prepare_repo_settings(the_repository);
-
-	graph = read_commit_graph_one(the_repository, odb);
-	if (!graph)
-		return 1;
-
 	printf("header: %08x %d %d %d %d\n",
 		ntohl(*(uint32_t*)graph->data),
 		*(unsigned char*)(graph->data + 4),
@@ -57,6 +45,23 @@ int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
 	if (graph->topo_levels)
 		printf(" topo_levels");
 	printf("\n");
+}
+
+int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+{
+	struct commit_graph *graph = NULL;
+	struct object_directory *odb;
+
+	setup_git_directory();
+	odb = the_repository->objects->odb;
+
+	prepare_repo_settings(the_repository);
+
+	graph = read_commit_graph_one(the_repository, odb);
+	if (!graph)
+		return 1;
+
+	dump_graph_info(graph);
 
 	UNLEAK(graph);
 
-- 
2.42.0.342.g8bb3a896ee

Re: [PATCH v3 05/17] t/helper/test-read-graph.c: extract `dump_graph_info()`

[Patrick Steinhardt]

[-- Attachment #1: Type: text/plain, Size: 2160 bytes --]

On Tue, Oct 10, 2023 at 04:33:33PM -0400, Taylor Blau wrote:
> Prepare for the 'read-graph' test helper to perform other tasks besides
> dumping high-level information about the commit-graph by extracting its
> main routine into a separate function.
> 
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>

Nit: your signoff is duplicated here. This is also still the case for
some of the other commits.

Patrick

> ---
>  t/helper/test-read-graph.c | 31 ++++++++++++++++++-------------
>  1 file changed, 18 insertions(+), 13 deletions(-)
> 
> diff --git a/t/helper/test-read-graph.c b/t/helper/test-read-graph.c
> index 8c7a83f578..3375392f6c 100644
> --- a/t/helper/test-read-graph.c
> +++ b/t/helper/test-read-graph.c
> @@ -5,20 +5,8 @@
>  #include "bloom.h"
>  #include "setup.h"
>  
> -int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
> +static void dump_graph_info(struct commit_graph *graph)
>  {
> -	struct commit_graph *graph = NULL;
> -	struct object_directory *odb;
> -
> -	setup_git_directory();
> -	odb = the_repository->objects->odb;
> -
> -	prepare_repo_settings(the_repository);
> -
> -	graph = read_commit_graph_one(the_repository, odb);
> -	if (!graph)
> -		return 1;
> -
>  	printf("header: %08x %d %d %d %d\n",
>  		ntohl(*(uint32_t*)graph->data),
>  		*(unsigned char*)(graph->data + 4),
> @@ -57,6 +45,23 @@ int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
>  	if (graph->topo_levels)
>  		printf(" topo_levels");
>  	printf("\n");
> +}
> +
> +int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
> +{
> +	struct commit_graph *graph = NULL;
> +	struct object_directory *odb;
> +
> +	setup_git_directory();
> +	odb = the_repository->objects->odb;
> +
> +	prepare_repo_settings(the_repository);
> +
> +	graph = read_commit_graph_one(the_repository, odb);
> +	if (!graph)
> +		return 1;
> +
> +	dump_graph_info(graph);
>  
>  	UNLEAK(graph);
>  
> -- 
> 2.42.0.342.g8bb3a896ee
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v3 05/17] t/helper/test-read-graph.c: extract `dump_graph_info()`

[Taylor Blau]

On Tue, Oct 17, 2023 at 10:45:18AM +0200, Patrick Steinhardt wrote:
> On Tue, Oct 10, 2023 at 04:33:33PM -0400, Taylor Blau wrote:
> > Prepare for the 'read-graph' test helper to perform other tasks besides
> > dumping high-level information about the commit-graph by extracting its
> > main routine into a separate function.
> >
> > Signed-off-by: Taylor Blau <me@ttaylorr.com>
> > Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
> > Signed-off-by: Junio C Hamano <gitster@pobox.com>
> > Signed-off-by: Taylor Blau <me@ttaylorr.com>
>
> Nit: your signoff is duplicated here. This is also still the case for
> some of the other commits.

Yeah, this is an artifact of having tossed these patches back and forth
(originally Jonathan sent some of these, then I sent another round, then
Jonathan, now me again). It's a little verbose, but accurately tracks
the DCO across multiple rounds.

Thanks,
Taylor

Re: [PATCH v3 05/17] t/helper/test-read-graph.c: extract `dump_graph_info()`

[Junio C Hamano]

Taylor Blau <me@ttaylorr.com> writes:

> On Tue, Oct 17, 2023 at 10:45:18AM +0200, Patrick Steinhardt wrote:
>> On Tue, Oct 10, 2023 at 04:33:33PM -0400, Taylor Blau wrote:
>> > Prepare for the 'read-graph' test helper to perform other tasks besides
>> > dumping high-level information about the commit-graph by extracting its
>> > main routine into a separate function.
>> >
>> > Signed-off-by: Taylor Blau <me@ttaylorr.com>
>> > Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
>> > Signed-off-by: Junio C Hamano <gitster@pobox.com>
>> > Signed-off-by: Taylor Blau <me@ttaylorr.com>
>>
>> Nit: your signoff is duplicated here. This is also still the case for
>> some of the other commits.
>
> Yeah, this is an artifact of having tossed these patches back and forth
> (originally Jonathan sent some of these, then I sent another round, then
> Jonathan, now me again). It's a little verbose, but accurately tracks
> the DCO across multiple rounds.

Well, between you and Jonathan Tan, that might be true, but my
involvement in the chain is only that I happened to have looked at
an earlier round and without doing anything else.  Even if you
refetched from my tree and based the new round on that version, it
would have been the same if you started from what you sent out
earlier without even looking at my tree.  So I do not think that
chain of DCO adds very little value by recording my sign-off there.

[PATCH v3 06/17] bloom.h: make `load_bloom_filter_from_graph()` public

[Taylor Blau]

Prepare for a future commit to use the load_bloom_filter_from_graph()
function directly to load specific Bloom filters out of the commit-graph
for manual inspection (to be used during tests).

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c | 6 +++---
 bloom.h | 5 +++++
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index aef6b5fea2..3e78cfe79d 100644
--- a/bloom.c
+++ b/bloom.c
@@ -29,9 +29,9 @@ static inline unsigned char get_bitmask(uint32_t pos)
 	return ((unsigned char)1) << (pos & (BITS_PER_WORD - 1));
 }
 
-static int load_bloom_filter_from_graph(struct commit_graph *g,
-					struct bloom_filter *filter,
-					uint32_t graph_pos)
+int load_bloom_filter_from_graph(struct commit_graph *g,
+				 struct bloom_filter *filter,
+				 uint32_t graph_pos)
 {
 	uint32_t lex_pos, start_index, end_index;
 
diff --git a/bloom.h b/bloom.h
index adde6dfe21..1e4f612d2c 100644
--- a/bloom.h
+++ b/bloom.h
@@ -3,6 +3,7 @@
 
 struct commit;
 struct repository;
+struct commit_graph;
 
 struct bloom_filter_settings {
 	/*
@@ -68,6 +69,10 @@ struct bloom_key {
 	uint32_t *hashes;
 };
 
+int load_bloom_filter_from_graph(struct commit_graph *g,
+				 struct bloom_filter *filter,
+				 uint32_t graph_pos);
+
 /*
  * Calculate the murmur3 32-bit hash value for the given data
  * using the given seed.
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 07/17] t/helper/test-read-graph: implement `bloom-filters` mode

[Taylor Blau]

Implement a mode of the "read-graph" test helper to dump out the
hexadecimal contents of the Bloom filter(s) contained in a commit-graph.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/helper/test-read-graph.c | 44 +++++++++++++++++++++++++++++++++-----
 1 file changed, 39 insertions(+), 5 deletions(-)

diff --git a/t/helper/test-read-graph.c b/t/helper/test-read-graph.c
index 3375392f6c..da9ac8584d 100644
--- a/t/helper/test-read-graph.c
+++ b/t/helper/test-read-graph.c
@@ -47,10 +47,32 @@ static void dump_graph_info(struct commit_graph *graph)
 	printf("\n");
 }
 
-int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
+static void dump_graph_bloom_filters(struct commit_graph *graph)
+{
+	uint32_t i;
+
+	for (i = 0; i < graph->num_commits + graph->num_commits_in_base; i++) {
+		struct bloom_filter filter = { 0 };
+		size_t j;
+
+		if (load_bloom_filter_from_graph(graph, &filter, i) < 0) {
+			fprintf(stderr, "missing Bloom filter for graph "
+				"position %"PRIu32"\n", i);
+			continue;
+		}
+
+		for (j = 0; j < filter.len; j++)
+			printf("%02x", filter.data[j]);
+		if (filter.len)
+			printf("\n");
+	}
+}
+
+int cmd__read_graph(int argc, const char **argv)
 {
 	struct commit_graph *graph = NULL;
 	struct object_directory *odb;
+	int ret = 0;
 
 	setup_git_directory();
 	odb = the_repository->objects->odb;
@@ -58,12 +80,24 @@ int cmd__read_graph(int argc UNUSED, const char **argv UNUSED)
 	prepare_repo_settings(the_repository);
 
 	graph = read_commit_graph_one(the_repository, odb);
-	if (!graph)
-		return 1;
+	if (!graph) {
+		ret = 1;
+		goto done;
+	}
 
-	dump_graph_info(graph);
+	if (argc <= 1)
+		dump_graph_info(graph);
+	else if (!strcmp(argv[1], "bloom-filters"))
+		dump_graph_bloom_filters(graph);
+	else {
+		fprintf(stderr, "unknown sub-command: '%s'\n", argv[1]);
+		ret = 1;
+	}
 
+done:
 	UNLEAK(graph);
 
-	return 0;
+	return ret;
 }
+
+
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 08/17] t4216: test changed path filters with high bit paths

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

Subsequent commits will teach Git another version of changed path
filter that has different behavior with paths that contain at least
one character with its high bit set, so test the existing behavior as
a baseline.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 t/t4216-log-bloom.sh | 52 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 52 insertions(+)

diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index f49a8f2fbf..da67c40134 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -484,4 +484,56 @@ test_expect_success 'merge graph layers with incompatible Bloom settings' '
 	! grep "disabling Bloom filters" err
 '
 
+get_first_changed_path_filter () {
+	test-tool read-graph bloom-filters >filters.dat &&
+	head -n 1 filters.dat
+}
+
+# chosen to be the same under all Unicode normalization forms
+CENT=$(printf "\302\242")
+
+test_expect_success 'set up repo with high bit path, version 1 changed-path' '
+	git init highbit1 &&
+	test_commit -C highbit1 c1 "$CENT" &&
+	git -C highbit1 commit-graph write --reachable --changed-paths
+'
+
+test_expect_success 'setup check value of version 1 changed-path' '
+	(
+		cd highbit1 &&
+		echo "52a9" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
+# expect will not match actual if char is unsigned by default. Write the test
+# in this way, so that a user running this test script can still see if the two
+# files match. (It will appear as an ordinary success if they match, and a skip
+# if not.)
+if test_cmp highbit1/expect highbit1/actual
+then
+	test_set_prereq SIGNED_CHAR_BY_DEFAULT
+fi
+test_expect_success SIGNED_CHAR_BY_DEFAULT 'check value of version 1 changed-path' '
+	# Only the prereq matters for this test.
+	true
+'
+
+test_expect_success 'setup make another commit' '
+	# "git log" does not use Bloom filters for root commits - see how, in
+	# revision.c, rev_compare_tree() (the only code path that eventually calls
+	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
+	# has one parent. Therefore, make another commit so that we perform the tests on
+	# a non-root commit.
+	test_commit -C highbit1 anotherc1 "another$CENT"
+'
+
+test_expect_success 'version 1 changed-path used when version 1 requested' '
+	(
+		cd highbit1 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
 test_done
-- 
2.42.0.342.g8bb3a896ee

Re: [PATCH v3 08/17] t4216: test changed path filters with high bit paths

[Patrick Steinhardt]

[-- Attachment #1: Type: text/plain, Size: 3157 bytes --]

On Tue, Oct 10, 2023 at 04:33:42PM -0400, Taylor Blau wrote:
> From: Jonathan Tan <jonathantanmy@google.com>
> 
> Subsequent commits will teach Git another version of changed path
> filter that has different behavior with paths that contain at least
> one character with its high bit set, so test the existing behavior as
> a baseline.
> 
> Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>

Nit: the signoffs are still funny here.

> ---
>  t/t4216-log-bloom.sh | 52 ++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 52 insertions(+)
> 
> diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
> index f49a8f2fbf..da67c40134 100755
> --- a/t/t4216-log-bloom.sh
> +++ b/t/t4216-log-bloom.sh
> @@ -484,4 +484,56 @@ test_expect_success 'merge graph layers with incompatible Bloom settings' '
>  	! grep "disabling Bloom filters" err
>  '
>  
> +get_first_changed_path_filter () {
> +	test-tool read-graph bloom-filters >filters.dat &&
> +	head -n 1 filters.dat
> +}
> +
> +# chosen to be the same under all Unicode normalization forms
> +CENT=$(printf "\302\242")
> +
> +test_expect_success 'set up repo with high bit path, version 1 changed-path' '
> +	git init highbit1 &&
> +	test_commit -C highbit1 c1 "$CENT" &&
> +	git -C highbit1 commit-graph write --reachable --changed-paths
> +'
> +
> +test_expect_success 'setup check value of version 1 changed-path' '
> +	(
> +		cd highbit1 &&
> +		echo "52a9" >expect &&
> +		get_first_changed_path_filter >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +
> +# expect will not match actual if char is unsigned by default. Write the test
> +# in this way, so that a user running this test script can still see if the two
> +# files match. (It will appear as an ordinary success if they match, and a skip
> +# if not.)
> +if test_cmp highbit1/expect highbit1/actual
> +then
> +	test_set_prereq SIGNED_CHAR_BY_DEFAULT
> +fi
> +test_expect_success SIGNED_CHAR_BY_DEFAULT 'check value of version 1 changed-path' '
> +	# Only the prereq matters for this test.
> +	true
> +'

Doesn't this mean that the preceding test where we `test_cmp expect
actual` can fail on some platforms depending on the signedness of
`char`?

Patrick

> +test_expect_success 'setup make another commit' '
> +	# "git log" does not use Bloom filters for root commits - see how, in
> +	# revision.c, rev_compare_tree() (the only code path that eventually calls
> +	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
> +	# has one parent. Therefore, make another commit so that we perform the tests on
> +	# a non-root commit.
> +	test_commit -C highbit1 anotherc1 "another$CENT"
> +'
> +
> +test_expect_success 'version 1 changed-path used when version 1 requested' '
> +	(
> +		cd highbit1 &&
> +		test_bloom_filters_used "-- another$CENT"
> +	)
> +'
> +
>  test_done
> -- 
> 2.42.0.342.g8bb3a896ee
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v3 08/17] t4216: test changed path filters with high bit paths

[Taylor Blau]

On Tue, Oct 17, 2023 at 10:45:13AM +0200, Patrick Steinhardt wrote:
> > +test_expect_success 'setup check value of version 1 changed-path' '
> > +	(
> > +		cd highbit1 &&
> > +		echo "52a9" >expect &&
> > +		get_first_changed_path_filter >actual &&
> > +		test_cmp expect actual
> > +	)
> > +'
> > +
> > +# expect will not match actual if char is unsigned by default. Write the test
> > +# in this way, so that a user running this test script can still see if the two
> > +# files match. (It will appear as an ordinary success if they match, and a skip
> > +# if not.)
> > +if test_cmp highbit1/expect highbit1/actual
> > +then
> > +	test_set_prereq SIGNED_CHAR_BY_DEFAULT
> > +fi
> > +test_expect_success SIGNED_CHAR_BY_DEFAULT 'check value of version 1 changed-path' '
> > +	# Only the prereq matters for this test.
> > +	true
> > +'
>
> Doesn't this mean that the preceding test where we `test_cmp expect
> actual` can fail on some platforms depending on the signedness of
> `char`?

Great catch, I am surprised this slipped by in earlier rounds. This
should do the trick, since we don't actually care about conditioning
that test's passing on test_cmp coming up clean. We check that in the if
statement you pointed out here, so:

--- 8< ---
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 114672e904..400dce2193 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -502,8 +502,7 @@ test_expect_success 'setup check value of version 1 changed-path' '
 	(
 		cd highbit1 &&
 		echo "52a9" >expect &&
-		get_first_changed_path_filter >actual &&
-		test_cmp expect actual
+		get_first_changed_path_filter >actual
 	)
 '
--- >8 ---

Thanks,
Taylor

[PATCH v3 09/17] repo-settings: introduce commitgraph.changedPathsVersion

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

A subsequent commit will introduce another version of the changed-path
filter in the commit graph file. In order to control which version to
write (and read), a config variable is needed.

Therefore, introduce this config variable. For forwards compatibility,
teach Git to not read commit graphs when the config variable
is set to an unsupported version. Because we teach Git this,
commitgraph.readChangedPaths is now redundant, so deprecate it and
define its behavior in terms of the config variable we introduce.

This commit does not change the behavior of writing (Git writes changed
path filters when explicitly instructed regardless of any config
variable), but a subsequent commit will restrict Git such that it will
only write when commitgraph.changedPathsVersion is a recognized value.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 Documentation/config/commitgraph.txt | 23 ++++++++++++++++++++---
 commit-graph.c                       |  2 +-
 oss-fuzz/fuzz-commit-graph.c         |  2 +-
 repo-settings.c                      |  6 +++++-
 repository.h                         |  2 +-
 5 files changed, 28 insertions(+), 7 deletions(-)

diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
index 30604e4a4c..2dc9170622 100644
--- a/Documentation/config/commitgraph.txt
+++ b/Documentation/config/commitgraph.txt
@@ -9,6 +9,23 @@ commitGraph.maxNewFilters::
 	commit-graph write` (c.f., linkgit:git-commit-graph[1]).
 
 commitGraph.readChangedPaths::
-	If true, then git will use the changed-path Bloom filters in the
-	commit-graph file (if it exists, and they are present). Defaults to
-	true. See linkgit:git-commit-graph[1] for more information.
+	Deprecated. Equivalent to commitGraph.changedPathsVersion=-1 if true, and
+	commitGraph.changedPathsVersion=0 if false. (If commitGraph.changedPathVersion
+	is also set, commitGraph.changedPathsVersion takes precedence.)
+
+commitGraph.changedPathsVersion::
+	Specifies the version of the changed-path Bloom filters that Git will read and
+	write. May be -1, 0 or 1.
++
+Defaults to -1.
++
+If -1, Git will use the version of the changed-path Bloom filters in the
+repository, defaulting to 1 if there are none.
++
+If 0, Git will not read any Bloom filters, and will write version 1 Bloom
+filters when instructed to write.
++
+If 1, Git will only read version 1 Bloom filters, and will write version 1
+Bloom filters.
++
+See linkgit:git-commit-graph[1] for more information.
diff --git a/commit-graph.c b/commit-graph.c
index ae0902f7f4..ea677c87fb 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -411,7 +411,7 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 			graph->read_generation_data = 1;
 	}
 
-	if (s->commit_graph_read_changed_paths) {
+	if (s->commit_graph_changed_paths_version) {
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
diff --git a/oss-fuzz/fuzz-commit-graph.c b/oss-fuzz/fuzz-commit-graph.c
index 2992079dd9..325c0b991a 100644
--- a/oss-fuzz/fuzz-commit-graph.c
+++ b/oss-fuzz/fuzz-commit-graph.c
@@ -19,7 +19,7 @@ int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
 	 * possible.
 	 */
 	the_repository->settings.commit_graph_generation_version = 2;
-	the_repository->settings.commit_graph_read_changed_paths = 1;
+	the_repository->settings.commit_graph_changed_paths_version = 1;
 	g = parse_commit_graph(&the_repository->settings, (void *)data, size);
 	repo_clear(the_repository);
 	free_commit_graph(g);
diff --git a/repo-settings.c b/repo-settings.c
index 525f69c0c7..db8fe817f3 100644
--- a/repo-settings.c
+++ b/repo-settings.c
@@ -24,6 +24,7 @@ void prepare_repo_settings(struct repository *r)
 	int value;
 	const char *strval;
 	int manyfiles;
+	int read_changed_paths;
 
 	if (!r->gitdir)
 		BUG("Cannot add settings for uninitialized repository");
@@ -54,7 +55,10 @@ void prepare_repo_settings(struct repository *r)
 	/* Commit graph config or default, does not cascade (simple) */
 	repo_cfg_bool(r, "core.commitgraph", &r->settings.core_commit_graph, 1);
 	repo_cfg_int(r, "commitgraph.generationversion", &r->settings.commit_graph_generation_version, 2);
-	repo_cfg_bool(r, "commitgraph.readchangedpaths", &r->settings.commit_graph_read_changed_paths, 1);
+	repo_cfg_bool(r, "commitgraph.readchangedpaths", &read_changed_paths, 1);
+	repo_cfg_int(r, "commitgraph.changedpathsversion",
+		     &r->settings.commit_graph_changed_paths_version,
+		     read_changed_paths ? -1 : 0);
 	repo_cfg_bool(r, "gc.writecommitgraph", &r->settings.gc_write_commit_graph, 1);
 	repo_cfg_bool(r, "fetch.writecommitgraph", &r->settings.fetch_write_commit_graph, 0);
 
diff --git a/repository.h b/repository.h
index 5f18486f64..f71154e12c 100644
--- a/repository.h
+++ b/repository.h
@@ -29,7 +29,7 @@ struct repo_settings {
 
 	int core_commit_graph;
 	int commit_graph_generation_version;
-	int commit_graph_read_changed_paths;
+	int commit_graph_changed_paths_version;
 	int gc_write_commit_graph;
 	int fetch_write_commit_graph;
 	int command_requires_full_index;
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 10/17] commit-graph: new filter ver. that fixes murmur3

[Taylor Blau]

From: Jonathan Tan <jonathantanmy@google.com>

The murmur3 implementation in bloom.c has a bug when converting series
of 4 bytes into network-order integers when char is signed (which is
controllable by a compiler option, and the default signedness of char is
platform-specific). When a string contains characters with the high bit
set, this bug causes results that, although internally consistent within
Git, does not accord with other implementations of murmur3 (thus,
the changed path filters wouldn't be readable by other off-the-shelf
implementatios of murmur3) and even with Git binaries that were compiled
with different signedness of char. This bug affects both how Git writes
changed path filters to disk and how Git interprets changed path filters
on disk.

Therefore, introduce a new version (2) of changed path filters that
corrects this problem. The existing version (1) is still supported and
is still the default, but users should migrate away from it as soon
as possible.

Because this bug only manifests with characters that have the high bit
set, it may be possible that some (or all) commits in a given repo would
have the same changed path filter both before and after this fix is
applied. However, in order to determine whether this is the case, the
changed paths would first have to be computed, at which point it is not
much more expensive to just compute a new changed path filter.

So this patch does not include any mechanism to "salvage" changed path
filters from repositories. There is also no "mixed" mode - for each
invocation of Git, reading and writing changed path filters are done
with the same version number; this version number may be explicitly
stated (typically if the user knows which version they need) or
automatically determined from the version of the existing changed path
filters in the repository.

There is a change in write_commit_graph(). graph_read_bloom_data()
makes it possible for chunk_bloom_data to be non-NULL but
bloom_filter_settings to be NULL, which causes a segfault later on. I
produced such a segfault while developing this patch, but couldn't find
a way to reproduce it neither after this complete patch (or before),
but in any case it seemed like a good thing to include that might help
future patch authors.

The value in t0095 was obtained from another murmur3 implementation
using the following Go source code:

  package main

  import "fmt"
  import "github.com/spaolacci/murmur3"

  func main() {
          fmt.Printf("%x\n", murmur3.Sum32([]byte("Hello world!")))
          fmt.Printf("%x\n", murmur3.Sum32([]byte{0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}))
  }

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 Documentation/config/commitgraph.txt |   5 +-
 bloom.c                              |  69 +++++++++++++++++-
 bloom.h                              |   8 +-
 commit-graph.c                       |  32 ++++++--
 t/helper/test-bloom.c                |   9 ++-
 t/t0095-bloom.sh                     |   8 ++
 t/t4216-log-bloom.sh                 | 105 +++++++++++++++++++++++++++
 7 files changed, 223 insertions(+), 13 deletions(-)

diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
index 2dc9170622..acc74a2f27 100644
--- a/Documentation/config/commitgraph.txt
+++ b/Documentation/config/commitgraph.txt
@@ -15,7 +15,7 @@ commitGraph.readChangedPaths::
 
 commitGraph.changedPathsVersion::
 	Specifies the version of the changed-path Bloom filters that Git will read and
-	write. May be -1, 0 or 1.
+	write. May be -1, 0, 1, or 2.
 +
 Defaults to -1.
 +
@@ -28,4 +28,7 @@ filters when instructed to write.
 If 1, Git will only read version 1 Bloom filters, and will write version 1
 Bloom filters.
 +
+If 2, Git will only read version 2 Bloom filters, and will write version 2
+Bloom filters.
++
 See linkgit:git-commit-graph[1] for more information.
diff --git a/bloom.c b/bloom.c
index 3e78cfe79d..ebef5cfd2f 100644
--- a/bloom.c
+++ b/bloom.c
@@ -66,7 +66,64 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
  * Not considered to be cryptographically secure.
  * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
  */
-uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
+uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len)
+{
+	const uint32_t c1 = 0xcc9e2d51;
+	const uint32_t c2 = 0x1b873593;
+	const uint32_t r1 = 15;
+	const uint32_t r2 = 13;
+	const uint32_t m = 5;
+	const uint32_t n = 0xe6546b64;
+	int i;
+	uint32_t k1 = 0;
+	const char *tail;
+
+	int len4 = len / sizeof(uint32_t);
+
+	uint32_t k;
+	for (i = 0; i < len4; i++) {
+		uint32_t byte1 = (uint32_t)(unsigned char)data[4*i];
+		uint32_t byte2 = ((uint32_t)(unsigned char)data[4*i + 1]) << 8;
+		uint32_t byte3 = ((uint32_t)(unsigned char)data[4*i + 2]) << 16;
+		uint32_t byte4 = ((uint32_t)(unsigned char)data[4*i + 3]) << 24;
+		k = byte1 | byte2 | byte3 | byte4;
+		k *= c1;
+		k = rotate_left(k, r1);
+		k *= c2;
+
+		seed ^= k;
+		seed = rotate_left(seed, r2) * m + n;
+	}
+
+	tail = (data + len4 * sizeof(uint32_t));
+
+	switch (len & (sizeof(uint32_t) - 1)) {
+	case 3:
+		k1 ^= ((uint32_t)(unsigned char)tail[2]) << 16;
+		/*-fallthrough*/
+	case 2:
+		k1 ^= ((uint32_t)(unsigned char)tail[1]) << 8;
+		/*-fallthrough*/
+	case 1:
+		k1 ^= ((uint32_t)(unsigned char)tail[0]) << 0;
+		k1 *= c1;
+		k1 = rotate_left(k1, r1);
+		k1 *= c2;
+		seed ^= k1;
+		break;
+	}
+
+	seed ^= (uint32_t)len;
+	seed ^= (seed >> 16);
+	seed *= 0x85ebca6b;
+	seed ^= (seed >> 13);
+	seed *= 0xc2b2ae35;
+	seed ^= (seed >> 16);
+
+	return seed;
+}
+
+static uint32_t murmur3_seeded_v1(uint32_t seed, const char *data, size_t len)
 {
 	const uint32_t c1 = 0xcc9e2d51;
 	const uint32_t c2 = 0x1b873593;
@@ -131,8 +188,14 @@ void fill_bloom_key(const char *data,
 	int i;
 	const uint32_t seed0 = 0x293ae76f;
 	const uint32_t seed1 = 0x7e646e2c;
-	const uint32_t hash0 = murmur3_seeded(seed0, data, len);
-	const uint32_t hash1 = murmur3_seeded(seed1, data, len);
+	uint32_t hash0, hash1;
+	if (settings->hash_version == 2) {
+		hash0 = murmur3_seeded_v2(seed0, data, len);
+		hash1 = murmur3_seeded_v2(seed1, data, len);
+	} else {
+		hash0 = murmur3_seeded_v1(seed0, data, len);
+		hash1 = murmur3_seeded_v1(seed1, data, len);
+	}
 
 	key->hashes = (uint32_t *)xcalloc(settings->num_hashes, sizeof(uint32_t));
 	for (i = 0; i < settings->num_hashes; i++)
diff --git a/bloom.h b/bloom.h
index 1e4f612d2c..138d57a86b 100644
--- a/bloom.h
+++ b/bloom.h
@@ -8,9 +8,11 @@ struct commit_graph;
 struct bloom_filter_settings {
 	/*
 	 * The version of the hashing technique being used.
-	 * We currently only support version = 1 which is
+	 * The newest version is 2, which is
 	 * the seeded murmur3 hashing technique implemented
-	 * in bloom.c.
+	 * in bloom.c. Bloom filters of version 1 were created
+	 * with prior versions of Git, which had a bug in the
+	 * implementation of the hash function.
 	 */
 	uint32_t hash_version;
 
@@ -80,7 +82,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
  * Not considered to be cryptographically secure.
  * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
  */
-uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len);
+uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len);
 
 void fill_bloom_key(const char *data,
 		    size_t len,
diff --git a/commit-graph.c b/commit-graph.c
index ea677c87fb..db623afd09 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -314,17 +314,26 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
 	return 0;
 }
 
+struct graph_read_bloom_data_context {
+	struct commit_graph *g;
+	int *commit_graph_changed_paths_version;
+};
+
 static int graph_read_bloom_data(const unsigned char *chunk_start,
 				  size_t chunk_size, void *data)
 {
-	struct commit_graph *g = data;
+	struct graph_read_bloom_data_context *c = data;
+	struct commit_graph *g = c->g;
 	uint32_t hash_version;
-	g->chunk_bloom_data = chunk_start;
 	hash_version = get_be32(chunk_start);
 
-	if (hash_version != 1)
+	if (*c->commit_graph_changed_paths_version == -1) {
+		*c->commit_graph_changed_paths_version = hash_version;
+	} else if (hash_version != *c->commit_graph_changed_paths_version) {
 		return 0;
+	}
 
+	g->chunk_bloom_data = chunk_start;
 	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
 	g->bloom_filter_settings->hash_version = hash_version;
 	g->bloom_filter_settings->num_hashes = get_be32(chunk_start + 4);
@@ -412,10 +421,14 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 	}
 
 	if (s->commit_graph_changed_paths_version) {
+		struct graph_read_bloom_data_context context = {
+			.g = graph,
+			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
+		};
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
-			   graph_read_bloom_data, graph);
+			   graph_read_bloom_data, &context);
 	}
 
 	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
@@ -2441,6 +2454,13 @@ int write_commit_graph(struct object_directory *odb,
 	}
 	if (!commit_graph_compatible(r))
 		return 0;
+	if (r->settings.commit_graph_changed_paths_version < -1
+	    || r->settings.commit_graph_changed_paths_version > 2) {
+		warning(_("attempting to write a commit-graph, but "
+			  "'commitgraph.changedPathsVersion' (%d) is not supported"),
+			r->settings.commit_graph_changed_paths_version);
+		return 0;
+	}
 
 	CALLOC_ARRAY(ctx, 1);
 	ctx->r = r;
@@ -2453,6 +2473,8 @@ int write_commit_graph(struct object_directory *odb,
 	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
 	ctx->num_generation_data_overflows = 0;
 
+	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
+		? 2 : 1;
 	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
 						      bloom_settings.bits_per_entry);
 	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
@@ -2482,7 +2504,7 @@ int write_commit_graph(struct object_directory *odb,
 		g = ctx->r->objects->commit_graph;
 
 		/* We have changed-paths already. Keep them in the next graph */
-		if (g && g->chunk_bloom_data) {
+		if (g && g->bloom_filter_settings) {
 			ctx->changed_paths = 1;
 			ctx->bloom_settings = g->bloom_filter_settings;
 		}
diff --git a/t/helper/test-bloom.c b/t/helper/test-bloom.c
index aabe31d724..3cbc0a5b50 100644
--- a/t/helper/test-bloom.c
+++ b/t/helper/test-bloom.c
@@ -50,6 +50,7 @@ static void get_bloom_filter_for_commit(const struct object_id *commit_oid)
 
 static const char *bloom_usage = "\n"
 "  test-tool bloom get_murmur3 <string>\n"
+"  test-tool bloom get_murmur3_seven_highbit\n"
 "  test-tool bloom generate_filter <string> [<string>...]\n"
 "  test-tool bloom get_filter_for_commit <commit-hex>\n";
 
@@ -64,7 +65,13 @@ int cmd__bloom(int argc, const char **argv)
 		uint32_t hashed;
 		if (argc < 3)
 			usage(bloom_usage);
-		hashed = murmur3_seeded(0, argv[2], strlen(argv[2]));
+		hashed = murmur3_seeded_v2(0, argv[2], strlen(argv[2]));
+		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
+	}
+
+	if (!strcmp(argv[1], "get_murmur3_seven_highbit")) {
+		uint32_t hashed;
+		hashed = murmur3_seeded_v2(0, "\x99\xaa\xbb\xcc\xdd\xee\xff", 7);
 		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
 	}
 
diff --git a/t/t0095-bloom.sh b/t/t0095-bloom.sh
index b567383eb8..c8d84ab606 100755
--- a/t/t0095-bloom.sh
+++ b/t/t0095-bloom.sh
@@ -29,6 +29,14 @@ test_expect_success 'compute unseeded murmur3 hash for test string 2' '
 	test_cmp expect actual
 '
 
+test_expect_success 'compute unseeded murmur3 hash for test string 3' '
+	cat >expect <<-\EOF &&
+	Murmur3 Hash with seed=0:0xa183ccfd
+	EOF
+	test-tool bloom get_murmur3_seven_highbit >actual &&
+	test_cmp expect actual
+'
+
 test_expect_success 'compute bloom key for empty string' '
 	cat >expect <<-\EOF &&
 	Hashes:0x5615800c|0x5b966560|0x61174ab4|0x66983008|0x6c19155c|0x7199fab0|0x771ae004|
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index da67c40134..8f8b5d4966 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -536,4 +536,109 @@ test_expect_success 'version 1 changed-path used when version 1 requested' '
 	)
 '
 
+test_expect_success 'version 1 changed-path not used when version 2 requested' '
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion 2 &&
+		test_bloom_filters_not_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'version 1 changed-path used when autodetect requested' '
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'when writing another commit graph, preserve existing version 1 of changed-path' '
+	test_commit -C highbit1 c1double "$CENT$CENT" &&
+	git -C highbit1 commit-graph write --reachable --changed-paths &&
+	(
+		cd highbit1 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		echo "options: bloom(1,10,7) read_generation_data" >expect &&
+		test-tool read-graph >full &&
+		grep options full >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'set up repo with high bit path, version 2 changed-path' '
+	git init highbit2 &&
+	git -C highbit2 config --add commitgraph.changedPathsVersion 2 &&
+	test_commit -C highbit2 c2 "$CENT" &&
+	git -C highbit2 commit-graph write --reachable --changed-paths
+'
+
+test_expect_success 'check value of version 2 changed-path' '
+	(
+		cd highbit2 &&
+		echo "c01f" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'setup make another commit' '
+	# "git log" does not use Bloom filters for root commits - see how, in
+	# revision.c, rev_compare_tree() (the only code path that eventually calls
+	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
+	# has one parent. Therefore, make another commit so that we perform the tests on
+	# a non-root commit.
+	test_commit -C highbit2 anotherc2 "another$CENT"
+'
+
+test_expect_success 'version 2 changed-path used when version 2 requested' '
+	(
+		cd highbit2 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'version 2 changed-path not used when version 1 requested' '
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion 1 &&
+		test_bloom_filters_not_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'version 2 changed-path used when autodetect requested' '
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		test_bloom_filters_used "-- another$CENT"
+	)
+'
+
+test_expect_success 'when writing another commit graph, preserve existing version 2 of changed-path' '
+	test_commit -C highbit2 c2double "$CENT$CENT" &&
+	git -C highbit2 commit-graph write --reachable --changed-paths &&
+	(
+		cd highbit2 &&
+		git config --add commitgraph.changedPathsVersion -1 &&
+		echo "options: bloom(2,10,7) read_generation_data" >expect &&
+		test-tool read-graph >full &&
+		grep options full >actual &&
+		test_cmp expect actual
+	)
+'
+
+test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
+	git init doublewrite &&
+	test_commit -C doublewrite c "$CENT" &&
+	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
+	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
+	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	(
+		cd doublewrite &&
+		echo "c01f" >expect &&
+		get_first_changed_path_filter >actual &&
+		test_cmp expect actual
+	)
+'
+
 test_done
-- 
2.42.0.342.g8bb3a896ee

Re: [PATCH v3 10/17] commit-graph: new filter ver. that fixes murmur3

[Patrick Steinhardt]

[-- Attachment #1: Type: text/plain, Size: 17868 bytes --]

On Tue, Oct 10, 2023 at 04:33:49PM -0400, Taylor Blau wrote:
> From: Jonathan Tan <jonathantanmy@google.com>
> 
> The murmur3 implementation in bloom.c has a bug when converting series
> of 4 bytes into network-order integers when char is signed (which is
> controllable by a compiler option, and the default signedness of char is
> platform-specific). When a string contains characters with the high bit
> set, this bug causes results that, although internally consistent within
> Git, does not accord with other implementations of murmur3 (thus,
> the changed path filters wouldn't be readable by other off-the-shelf
> implementatios of murmur3) and even with Git binaries that were compiled
> with different signedness of char. This bug affects both how Git writes
> changed path filters to disk and how Git interprets changed path filters
> on disk.
> 
> Therefore, introduce a new version (2) of changed path filters that
> corrects this problem. The existing version (1) is still supported and
> is still the default, but users should migrate away from it as soon
> as possible.
> 
> Because this bug only manifests with characters that have the high bit
> set, it may be possible that some (or all) commits in a given repo would
> have the same changed path filter both before and after this fix is
> applied. However, in order to determine whether this is the case, the
> changed paths would first have to be computed, at which point it is not
> much more expensive to just compute a new changed path filter.
> 
> So this patch does not include any mechanism to "salvage" changed path
> filters from repositories. There is also no "mixed" mode - for each
> invocation of Git, reading and writing changed path filters are done
> with the same version number; this version number may be explicitly
> stated (typically if the user knows which version they need) or
> automatically determined from the version of the existing changed path
> filters in the repository.
> 
> There is a change in write_commit_graph(). graph_read_bloom_data()
> makes it possible for chunk_bloom_data to be non-NULL but
> bloom_filter_settings to be NULL, which causes a segfault later on. I
> produced such a segfault while developing this patch, but couldn't find
> a way to reproduce it neither after this complete patch (or before),
> but in any case it seemed like a good thing to include that might help
> future patch authors.
> 
> The value in t0095 was obtained from another murmur3 implementation
> using the following Go source code:
> 
>   package main
> 
>   import "fmt"
>   import "github.com/spaolacci/murmur3"
> 
>   func main() {
>           fmt.Printf("%x\n", murmur3.Sum32([]byte("Hello world!")))
>           fmt.Printf("%x\n", murmur3.Sum32([]byte{0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}))
>   }
> 
> Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> ---
>  Documentation/config/commitgraph.txt |   5 +-
>  bloom.c                              |  69 +++++++++++++++++-
>  bloom.h                              |   8 +-
>  commit-graph.c                       |  32 ++++++--
>  t/helper/test-bloom.c                |   9 ++-
>  t/t0095-bloom.sh                     |   8 ++
>  t/t4216-log-bloom.sh                 | 105 +++++++++++++++++++++++++++
>  7 files changed, 223 insertions(+), 13 deletions(-)
> 
> diff --git a/Documentation/config/commitgraph.txt b/Documentation/config/commitgraph.txt
> index 2dc9170622..acc74a2f27 100644
> --- a/Documentation/config/commitgraph.txt
> +++ b/Documentation/config/commitgraph.txt
> @@ -15,7 +15,7 @@ commitGraph.readChangedPaths::
>  
>  commitGraph.changedPathsVersion::
>  	Specifies the version of the changed-path Bloom filters that Git will read and
> -	write. May be -1, 0 or 1.
> +	write. May be -1, 0, 1, or 2.
>  +
>  Defaults to -1.
>  +
> @@ -28,4 +28,7 @@ filters when instructed to write.
>  If 1, Git will only read version 1 Bloom filters, and will write version 1
>  Bloom filters.
>  +
> +If 2, Git will only read version 2 Bloom filters, and will write version 2
> +Bloom filters.
> ++
>  See linkgit:git-commit-graph[1] for more information.
> diff --git a/bloom.c b/bloom.c
> index 3e78cfe79d..ebef5cfd2f 100644
> --- a/bloom.c
> +++ b/bloom.c
> @@ -66,7 +66,64 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
>   * Not considered to be cryptographically secure.
>   * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
>   */
> -uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
> +uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len)
> +{
> +	const uint32_t c1 = 0xcc9e2d51;
> +	const uint32_t c2 = 0x1b873593;
> +	const uint32_t r1 = 15;
> +	const uint32_t r2 = 13;
> +	const uint32_t m = 5;
> +	const uint32_t n = 0xe6546b64;
> +	int i;
> +	uint32_t k1 = 0;
> +	const char *tail;
> +
> +	int len4 = len / sizeof(uint32_t);
> +
> +	uint32_t k;
> +	for (i = 0; i < len4; i++) {
> +		uint32_t byte1 = (uint32_t)(unsigned char)data[4*i];
> +		uint32_t byte2 = ((uint32_t)(unsigned char)data[4*i + 1]) << 8;
> +		uint32_t byte3 = ((uint32_t)(unsigned char)data[4*i + 2]) << 16;
> +		uint32_t byte4 = ((uint32_t)(unsigned char)data[4*i + 3]) << 24;
> +		k = byte1 | byte2 | byte3 | byte4;
> +		k *= c1;
> +		k = rotate_left(k, r1);
> +		k *= c2;
> +
> +		seed ^= k;
> +		seed = rotate_left(seed, r2) * m + n;
> +	}
> +
> +	tail = (data + len4 * sizeof(uint32_t));
> +
> +	switch (len & (sizeof(uint32_t) - 1)) {
> +	case 3:
> +		k1 ^= ((uint32_t)(unsigned char)tail[2]) << 16;
> +		/*-fallthrough*/
> +	case 2:
> +		k1 ^= ((uint32_t)(unsigned char)tail[1]) << 8;
> +		/*-fallthrough*/
> +	case 1:
> +		k1 ^= ((uint32_t)(unsigned char)tail[0]) << 0;
> +		k1 *= c1;
> +		k1 = rotate_left(k1, r1);
> +		k1 *= c2;
> +		seed ^= k1;
> +		break;
> +	}
> +
> +	seed ^= (uint32_t)len;
> +	seed ^= (seed >> 16);
> +	seed *= 0x85ebca6b;
> +	seed ^= (seed >> 13);
> +	seed *= 0xc2b2ae35;
> +	seed ^= (seed >> 16);
> +
> +	return seed;
> +}
> +
> +static uint32_t murmur3_seeded_v1(uint32_t seed, const char *data, size_t len)
>  {
>  	const uint32_t c1 = 0xcc9e2d51;
>  	const uint32_t c2 = 0x1b873593;
> @@ -131,8 +188,14 @@ void fill_bloom_key(const char *data,
>  	int i;
>  	const uint32_t seed0 = 0x293ae76f;
>  	const uint32_t seed1 = 0x7e646e2c;
> -	const uint32_t hash0 = murmur3_seeded(seed0, data, len);
> -	const uint32_t hash1 = murmur3_seeded(seed1, data, len);
> +	uint32_t hash0, hash1;
> +	if (settings->hash_version == 2) {
> +		hash0 = murmur3_seeded_v2(seed0, data, len);
> +		hash1 = murmur3_seeded_v2(seed1, data, len);
> +	} else {
> +		hash0 = murmur3_seeded_v1(seed0, data, len);
> +		hash1 = murmur3_seeded_v1(seed1, data, len);
> +	}
>  
>  	key->hashes = (uint32_t *)xcalloc(settings->num_hashes, sizeof(uint32_t));
>  	for (i = 0; i < settings->num_hashes; i++)
> diff --git a/bloom.h b/bloom.h
> index 1e4f612d2c..138d57a86b 100644
> --- a/bloom.h
> +++ b/bloom.h
> @@ -8,9 +8,11 @@ struct commit_graph;
>  struct bloom_filter_settings {
>  	/*
>  	 * The version of the hashing technique being used.
> -	 * We currently only support version = 1 which is
> +	 * The newest version is 2, which is
>  	 * the seeded murmur3 hashing technique implemented
> -	 * in bloom.c.
> +	 * in bloom.c. Bloom filters of version 1 were created
> +	 * with prior versions of Git, which had a bug in the
> +	 * implementation of the hash function.
>  	 */
>  	uint32_t hash_version;
>  
> @@ -80,7 +82,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
>   * Not considered to be cryptographically secure.
>   * Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
>   */
> -uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len);
> +uint32_t murmur3_seeded_v2(uint32_t seed, const char *data, size_t len);
>  
>  void fill_bloom_key(const char *data,
>  		    size_t len,
> diff --git a/commit-graph.c b/commit-graph.c
> index ea677c87fb..db623afd09 100644
> --- a/commit-graph.c
> +++ b/commit-graph.c
> @@ -314,17 +314,26 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
>  	return 0;
>  }
>  
> +struct graph_read_bloom_data_context {
> +	struct commit_graph *g;
> +	int *commit_graph_changed_paths_version;
> +};
> +
>  static int graph_read_bloom_data(const unsigned char *chunk_start,
>  				  size_t chunk_size, void *data)
>  {
> -	struct commit_graph *g = data;
> +	struct graph_read_bloom_data_context *c = data;
> +	struct commit_graph *g = c->g;
>  	uint32_t hash_version;
> -	g->chunk_bloom_data = chunk_start;
>  	hash_version = get_be32(chunk_start);
>  
> -	if (hash_version != 1)
> +	if (*c->commit_graph_changed_paths_version == -1) {
> +		*c->commit_graph_changed_paths_version = hash_version;
> +	} else if (hash_version != *c->commit_graph_changed_paths_version) {
>  		return 0;
> +	}

In case we have `c->commit_graph_changed_paths_version == -1` we lose
the check that the hash version is something that we know and support,
don't we? And while we do start to handle `-1` in the writing path, I
think we don't in the reading path unless I missed something.

> +	g->chunk_bloom_data = chunk_start;
>  	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
>  	g->bloom_filter_settings->hash_version = hash_version;
>  	g->bloom_filter_settings->num_hashes = get_be32(chunk_start + 4);
> @@ -412,10 +421,14 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
>  	}
>  
>  	if (s->commit_graph_changed_paths_version) {
> +		struct graph_read_bloom_data_context context = {
> +			.g = graph,
> +			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
> +		};
>  		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
>  			   &graph->chunk_bloom_indexes);
>  		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
> -			   graph_read_bloom_data, graph);
> +			   graph_read_bloom_data, &context);
>  	}
>  
>  	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
> @@ -2441,6 +2454,13 @@ int write_commit_graph(struct object_directory *odb,
>  	}
>  	if (!commit_graph_compatible(r))
>  		return 0;
> +	if (r->settings.commit_graph_changed_paths_version < -1
> +	    || r->settings.commit_graph_changed_paths_version > 2) {
> +		warning(_("attempting to write a commit-graph, but "
> +			  "'commitgraph.changedPathsVersion' (%d) is not supported"),
> +			r->settings.commit_graph_changed_paths_version);
> +		return 0;
> +	}
>  
>  	CALLOC_ARRAY(ctx, 1);
>  	ctx->r = r;
> @@ -2453,6 +2473,8 @@ int write_commit_graph(struct object_directory *odb,
>  	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
>  	ctx->num_generation_data_overflows = 0;
>  
> +	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
> +		? 2 : 1;
>  	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
>  						      bloom_settings.bits_per_entry);
>  	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
> @@ -2482,7 +2504,7 @@ int write_commit_graph(struct object_directory *odb,
>  		g = ctx->r->objects->commit_graph;
>  
>  		/* We have changed-paths already. Keep them in the next graph */
> -		if (g && g->chunk_bloom_data) {
> +		if (g && g->bloom_filter_settings) {
>  			ctx->changed_paths = 1;
>  			ctx->bloom_settings = g->bloom_filter_settings;
>  		}
> diff --git a/t/helper/test-bloom.c b/t/helper/test-bloom.c
> index aabe31d724..3cbc0a5b50 100644
> --- a/t/helper/test-bloom.c
> +++ b/t/helper/test-bloom.c
> @@ -50,6 +50,7 @@ static void get_bloom_filter_for_commit(const struct object_id *commit_oid)
>  
>  static const char *bloom_usage = "\n"
>  "  test-tool bloom get_murmur3 <string>\n"
> +"  test-tool bloom get_murmur3_seven_highbit\n"
>  "  test-tool bloom generate_filter <string> [<string>...]\n"
>  "  test-tool bloom get_filter_for_commit <commit-hex>\n";
>  
> @@ -64,7 +65,13 @@ int cmd__bloom(int argc, const char **argv)
>  		uint32_t hashed;
>  		if (argc < 3)
>  			usage(bloom_usage);
> -		hashed = murmur3_seeded(0, argv[2], strlen(argv[2]));
> +		hashed = murmur3_seeded_v2(0, argv[2], strlen(argv[2]));
> +		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
> +	}
> +
> +	if (!strcmp(argv[1], "get_murmur3_seven_highbit")) {
> +		uint32_t hashed;
> +		hashed = murmur3_seeded_v2(0, "\x99\xaa\xbb\xcc\xdd\xee\xff", 7);
>  		printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
>  	}
>  
> diff --git a/t/t0095-bloom.sh b/t/t0095-bloom.sh
> index b567383eb8..c8d84ab606 100755
> --- a/t/t0095-bloom.sh
> +++ b/t/t0095-bloom.sh
> @@ -29,6 +29,14 @@ test_expect_success 'compute unseeded murmur3 hash for test string 2' '
>  	test_cmp expect actual
>  '
>  
> +test_expect_success 'compute unseeded murmur3 hash for test string 3' '
> +	cat >expect <<-\EOF &&
> +	Murmur3 Hash with seed=0:0xa183ccfd
> +	EOF
> +	test-tool bloom get_murmur3_seven_highbit >actual &&
> +	test_cmp expect actual
> +'
> +
>  test_expect_success 'compute bloom key for empty string' '
>  	cat >expect <<-\EOF &&
>  	Hashes:0x5615800c|0x5b966560|0x61174ab4|0x66983008|0x6c19155c|0x7199fab0|0x771ae004|
> diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
> index da67c40134..8f8b5d4966 100755
> --- a/t/t4216-log-bloom.sh
> +++ b/t/t4216-log-bloom.sh
> @@ -536,4 +536,109 @@ test_expect_success 'version 1 changed-path used when version 1 requested' '
>  	)
>  '
>  
> +test_expect_success 'version 1 changed-path not used when version 2 requested' '
> +	(
> +		cd highbit1 &&
> +		git config --add commitgraph.changedPathsVersion 2 &&
> +		test_bloom_filters_not_used "-- another$CENT"
> +	)
> +'
> +
> +test_expect_success 'version 1 changed-path used when autodetect requested' '
> +	(
> +		cd highbit1 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		test_bloom_filters_used "-- another$CENT"
> +	)
> +'
> +
> +test_expect_success 'when writing another commit graph, preserve existing version 1 of changed-path' '
> +	test_commit -C highbit1 c1double "$CENT$CENT" &&
> +	git -C highbit1 commit-graph write --reachable --changed-paths &&
> +	(
> +		cd highbit1 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		echo "options: bloom(1,10,7) read_generation_data" >expect &&
> +		test-tool read-graph >full &&
> +		grep options full >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +
> +test_expect_success 'set up repo with high bit path, version 2 changed-path' '
> +	git init highbit2 &&
> +	git -C highbit2 config --add commitgraph.changedPathsVersion 2 &&
> +	test_commit -C highbit2 c2 "$CENT" &&
> +	git -C highbit2 commit-graph write --reachable --changed-paths
> +'
> +
> +test_expect_success 'check value of version 2 changed-path' '
> +	(
> +		cd highbit2 &&
> +		echo "c01f" >expect &&
> +		get_first_changed_path_filter >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +
> +test_expect_success 'setup make another commit' '
> +	# "git log" does not use Bloom filters for root commits - see how, in
> +	# revision.c, rev_compare_tree() (the only code path that eventually calls
> +	# get_bloom_filter()) is only called by try_to_simplify_commit() when the commit
> +	# has one parent. Therefore, make another commit so that we perform the tests on
> +	# a non-root commit.
> +	test_commit -C highbit2 anotherc2 "another$CENT"
> +'
> +
> +test_expect_success 'version 2 changed-path used when version 2 requested' '
> +	(
> +		cd highbit2 &&
> +		test_bloom_filters_used "-- another$CENT"
> +	)
> +'
> +
> +test_expect_success 'version 2 changed-path not used when version 1 requested' '
> +	(
> +		cd highbit2 &&
> +		git config --add commitgraph.changedPathsVersion 1 &&
> +		test_bloom_filters_not_used "-- another$CENT"
> +	)
> +'
> +
> +test_expect_success 'version 2 changed-path used when autodetect requested' '
> +	(
> +		cd highbit2 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		test_bloom_filters_used "-- another$CENT"
> +	)
> +'
> +
> +test_expect_success 'when writing another commit graph, preserve existing version 2 of changed-path' '
> +	test_commit -C highbit2 c2double "$CENT$CENT" &&
> +	git -C highbit2 commit-graph write --reachable --changed-paths &&
> +	(
> +		cd highbit2 &&
> +		git config --add commitgraph.changedPathsVersion -1 &&
> +		echo "options: bloom(2,10,7) read_generation_data" >expect &&
> +		test-tool read-graph >full &&
> +		grep options full >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +
> +test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
> +	git init doublewrite &&
> +	test_commit -C doublewrite c "$CENT" &&
> +	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
> +	git -C doublewrite commit-graph write --reachable --changed-paths &&
> +	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
> +	git -C doublewrite commit-graph write --reachable --changed-paths &&
> +	(
> +		cd doublewrite &&
> +		echo "c01f" >expect &&
> +		get_first_changed_path_filter >actual &&
> +		test_cmp expect actual
> +	)
> +'
> +

With the supposedly missing check in mind, should we also add tests for
currently unknown versions like 3 or -2?

Patrick

>  test_done
> -- 
> 2.42.0.342.g8bb3a896ee
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v3 10/17] commit-graph: new filter ver. that fixes murmur3

[Taylor Blau]

On Tue, Oct 17, 2023 at 10:45:24AM +0200, Patrick Steinhardt wrote:
> > @@ -314,17 +314,26 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
> >  	return 0;
> >  }
> >
> > +struct graph_read_bloom_data_context {
> > +	struct commit_graph *g;
> > +	int *commit_graph_changed_paths_version;
> > +};
> > +
> >  static int graph_read_bloom_data(const unsigned char *chunk_start,
> >  				  size_t chunk_size, void *data)
> >  {
> > -	struct commit_graph *g = data;
> > +	struct graph_read_bloom_data_context *c = data;
> > +	struct commit_graph *g = c->g;
> >  	uint32_t hash_version;
> > -	g->chunk_bloom_data = chunk_start;
> >  	hash_version = get_be32(chunk_start);
> >
> > -	if (hash_version != 1)
> > +	if (*c->commit_graph_changed_paths_version == -1) {
> > +		*c->commit_graph_changed_paths_version = hash_version;
> > +	} else if (hash_version != *c->commit_graph_changed_paths_version) {
> >  		return 0;
> > +	}
>
> In case we have `c->commit_graph_changed_paths_version == -1` we lose
> the check that the hash version is something that we know and support,
> don't we? And while we do start to handle `-1` in the writing path, I
> think we don't in the reading path unless I missed something.

We don't have to deal with c->commit_graph_changed_paths_version being
-1 here, since we normalize it when reading the BDAT chunk. See
commit-graph.c::graph_read_bloom_data(), particularly:

    if (*c->commit_graph_changed_paths_version == -1)
        *c->commit_graph_changed_paths_version = hash_version;
    else if (hash_version != *c->commit_graph_changed_paths_version)
        return 0;

> > +test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
> > +	git init doublewrite &&
> > +	test_commit -C doublewrite c "$CENT" &&
> > +	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
> > +	git -C doublewrite commit-graph write --reachable --changed-paths &&
> > +	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
> > +	git -C doublewrite commit-graph write --reachable --changed-paths &&
> > +	(
> > +		cd doublewrite &&
> > +		echo "c01f" >expect &&
> > +		get_first_changed_path_filter >actual &&
> > +		test_cmp expect actual
> > +	)
> > +'
> > +
>
> With the supposedly missing check in mind, should we also add tests for
> currently unknown versions like 3 or -2?

Good idea, I'll update the test to reflect.

Thanks,
Taylor

[PATCH v3 11/17] bloom: annotate filters with hash version

[Taylor Blau]

In subsequent commits, we will want to load existing Bloom filters out
of a commit-graph, even when the hash version they were computed with
does not match the value of `commitGraph.changedPathVersion`.

In order to differentiate between the two, add a "version" field to each
Bloom filter.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c | 11 ++++++++---
 bloom.h |  1 +
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index ebef5cfd2f..9b6a30f6f6 100644
--- a/bloom.c
+++ b/bloom.c
@@ -55,6 +55,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
 	filter->data = (unsigned char *)(g->chunk_bloom_data +
 					sizeof(unsigned char) * start_index +
 					BLOOMDATA_CHUNK_HEADER_SIZE);
+	filter->version = g->bloom_filter_settings->hash_version;
 
 	return 1;
 }
@@ -240,11 +241,13 @@ static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 	return strcmp(e1->path, e2->path);
 }
 
-static void init_truncated_large_filter(struct bloom_filter *filter)
+static void init_truncated_large_filter(struct bloom_filter *filter,
+					int version)
 {
 	filter->data = xmalloc(1);
 	filter->data[0] = 0xFF;
 	filter->len = 1;
+	filter->version = version;
 }
 
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
@@ -329,13 +332,15 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 		}
 
 		if (hashmap_get_size(&pathmap) > settings->max_changed_paths) {
-			init_truncated_large_filter(filter);
+			init_truncated_large_filter(filter,
+						    settings->hash_version);
 			if (computed)
 				*computed |= BLOOM_TRUNC_LARGE;
 			goto cleanup;
 		}
 
 		filter->len = (hashmap_get_size(&pathmap) * settings->bits_per_entry + BITS_PER_WORD - 1) / BITS_PER_WORD;
+		filter->version = settings->hash_version;
 		if (!filter->len) {
 			if (computed)
 				*computed |= BLOOM_TRUNC_EMPTY;
@@ -355,7 +360,7 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 	} else {
 		for (i = 0; i < diff_queued_diff.nr; i++)
 			diff_free_filepair(diff_queued_diff.queue[i]);
-		init_truncated_large_filter(filter);
+		init_truncated_large_filter(filter, settings->hash_version);
 
 		if (computed)
 			*computed |= BLOOM_TRUNC_LARGE;
diff --git a/bloom.h b/bloom.h
index 138d57a86b..330a140520 100644
--- a/bloom.h
+++ b/bloom.h
@@ -55,6 +55,7 @@ struct bloom_filter_settings {
 struct bloom_filter {
 	unsigned char *data;
 	size_t len;
+	int version;
 };
 
 /*
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 12/17] bloom: prepare to discard incompatible Bloom filters

[Taylor Blau]

Callers use the inline `get_bloom_filter()` implementation as a thin
wrapper around `get_or_compute_bloom_filter()`. The former calls the
latter with a value of "0" for `compute_if_not_present`, making
`get_bloom_filter()` the default read-only path for fetching an existing
Bloom filter.

Callers expect the value returned from `get_bloom_filter()` is usable,
that is that it's compatible with the configured value corresponding to
`commitGraph.changedPathsVersion`.

This is OK, since the commit-graph machinery only initializes its BDAT
chunk (thereby enabling it to service Bloom filter queries) when the
Bloom filter hash_version is compatible with our settings. So any value
returned by `get_bloom_filter()` is trivially useable.

However, subsequent commits will load the BDAT chunk even when the Bloom
filters are built with incompatible hash versions. Prepare to handle
this by teaching `get_bloom_filter()` to discard filters that are
incompatible with the configured hash version.

Callers who wish to read incompatible filters (e.g., for upgrading
filters from v1 to v2) may use the lower level routine,
`get_or_compute_bloom_filter()`.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c | 20 +++++++++++++++++++-
 bloom.h | 20 ++++++++++++++++++--
 2 files changed, 37 insertions(+), 3 deletions(-)

diff --git a/bloom.c b/bloom.c
index 9b6a30f6f6..739fa093ba 100644
--- a/bloom.c
+++ b/bloom.c
@@ -250,6 +250,23 @@ static void init_truncated_large_filter(struct bloom_filter *filter,
 	filter->version = version;
 }
 
+struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
+{
+	struct bloom_filter *filter;
+	int hash_version;
+
+	filter = get_or_compute_bloom_filter(r, c, 0, NULL, NULL);
+	if (!filter)
+		return NULL;
+
+	prepare_repo_settings(r);
+	hash_version = r->settings.commit_graph_changed_paths_version;
+
+	if (!(hash_version == -1 || hash_version == filter->version))
+		return NULL; /* unusable filter */
+	return filter;
+}
+
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						 struct commit *c,
 						 int compute_if_not_present,
@@ -275,7 +292,8 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						     filter, graph_pos);
 	}
 
-	if (filter->data && filter->len)
+	if ((filter->data && filter->len) &&
+	    (!settings || settings->hash_version == filter->version))
 		return filter;
 	if (!compute_if_not_present)
 		return NULL;
diff --git a/bloom.h b/bloom.h
index 330a140520..bfe389e29c 100644
--- a/bloom.h
+++ b/bloom.h
@@ -110,8 +110,24 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						 const struct bloom_filter_settings *settings,
 						 enum bloom_filter_computed *computed);
 
-#define get_bloom_filter(r, c) get_or_compute_bloom_filter( \
-	(r), (c), 0, NULL, NULL)
+/*
+ * Find the Bloom filter associated with the given commit "c".
+ *
+ * If any of the following are true
+ *
+ *   - the repository does not have a commit-graph, or
+ *   - the repository disables reading from the commit-graph, or
+ *   - the given commit does not have a Bloom filter computed, or
+ *   - there is a Bloom filter for commit "c", but it cannot be read
+ *     because the filter uses an incompatible version of murmur3
+ *
+ * , then `get_bloom_filter()` will return NULL. Otherwise, the corresponding
+ * Bloom filter will be returned.
+ *
+ * For callers who wish to inspect Bloom filters with incompatible hash
+ * versions, use get_or_compute_bloom_filter().
+ */
+struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c);
 
 int bloom_filter_contains(const struct bloom_filter *filter,
 			  const struct bloom_key *key,
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 13/17] commit-graph.c: unconditionally load Bloom filters

[Taylor Blau]

In 9e4df4da07 (commit-graph: new filter ver. that fixes murmur3,
2023-08-01), we began ignoring the Bloom data ("BDAT") chunk for
commit-graphs whose Bloom filters were computed using a hash version
incompatible with the value of `commitGraph.changedPathVersion`.

Now that the Bloom API has been hardened to discard these incompatible
filters (with the exception of low-level APIs), we can safely load these
Bloom filters unconditionally.

We no longer want to return early from `graph_read_bloom_data()`, and
similarly do not want to set the bloom_settings' `hash_version` field as
a side-effect. The latter is because we want to wait until we know which
Bloom settings we're using (either the defaults, from the GIT_TEST
variables, or from the previous commit-graph layer) before deciding what
hash_version to use.

If we detect an existing BDAT chunk, we'll infer the rest of the
settings (e.g., number of hashes, bits per entry, and maximum number of
changed paths) from the earlier graph layer. The hash_version will be
inferred from the previous layer as well, unless one has already been
specified via configuration.

Once all of that is done, we normalize the value of the hash_version to
either "1" or "2".

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 commit-graph.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/commit-graph.c b/commit-graph.c
index db623afd09..fa3b58e762 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -327,12 +327,6 @@ static int graph_read_bloom_data(const unsigned char *chunk_start,
 	uint32_t hash_version;
 	hash_version = get_be32(chunk_start);
 
-	if (*c->commit_graph_changed_paths_version == -1) {
-		*c->commit_graph_changed_paths_version = hash_version;
-	} else if (hash_version != *c->commit_graph_changed_paths_version) {
-		return 0;
-	}
-
 	g->chunk_bloom_data = chunk_start;
 	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
 	g->bloom_filter_settings->hash_version = hash_version;
@@ -2473,8 +2467,7 @@ int write_commit_graph(struct object_directory *odb,
 	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
 	ctx->num_generation_data_overflows = 0;
 
-	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
-		? 2 : 1;
+	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version;
 	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
 						      bloom_settings.bits_per_entry);
 	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
@@ -2506,10 +2499,18 @@ int write_commit_graph(struct object_directory *odb,
 		/* We have changed-paths already. Keep them in the next graph */
 		if (g && g->bloom_filter_settings) {
 			ctx->changed_paths = 1;
-			ctx->bloom_settings = g->bloom_filter_settings;
+
+			/* don't propagate the hash_version unless unspecified */
+			if (bloom_settings.hash_version == -1)
+				bloom_settings.hash_version = g->bloom_filter_settings->hash_version;
+			bloom_settings.bits_per_entry = g->bloom_filter_settings->bits_per_entry;
+			bloom_settings.num_hashes = g->bloom_filter_settings->num_hashes;
+			bloom_settings.max_changed_paths = g->bloom_filter_settings->max_changed_paths;
 		}
 	}
 
+	bloom_settings.hash_version = bloom_settings.hash_version == 2 ? 2 : 1;
+
 	if (ctx->split) {
 		struct commit_graph *g = ctx->r->objects->commit_graph;
 
-- 
2.42.0.342.g8bb3a896ee

Re: [PATCH v3 13/17] commit-graph.c: unconditionally load Bloom filters

[Patrick Steinhardt]

[-- Attachment #1: Type: text/plain, Size: 4073 bytes --]

On Tue, Oct 10, 2023 at 04:33:59PM -0400, Taylor Blau wrote:
> In 9e4df4da07 (commit-graph: new filter ver. that fixes murmur3,

Nit: It's a bit funny to read this reference to a commit ID when the
commit in question is part of the same series. Isn't it likely to grow
stale?

> 2023-08-01), we began ignoring the Bloom data ("BDAT") chunk for
> commit-graphs whose Bloom filters were computed using a hash version
> incompatible with the value of `commitGraph.changedPathVersion`.
> 
> Now that the Bloom API has been hardened to discard these incompatible
> filters (with the exception of low-level APIs), we can safely load these
> Bloom filters unconditionally.
> 
> We no longer want to return early from `graph_read_bloom_data()`, and
> similarly do not want to set the bloom_settings' `hash_version` field as
> a side-effect. The latter is because we want to wait until we know which
> Bloom settings we're using (either the defaults, from the GIT_TEST
> variables, or from the previous commit-graph layer) before deciding what
> hash_version to use.
> 
> If we detect an existing BDAT chunk, we'll infer the rest of the
> settings (e.g., number of hashes, bits per entry, and maximum number of
> changed paths) from the earlier graph layer. The hash_version will be
> inferred from the previous layer as well, unless one has already been
> specified via configuration.
> 
> Once all of that is done, we normalize the value of the hash_version to
> either "1" or "2".
> 
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> ---
>  commit-graph.c | 19 ++++++++++---------
>  1 file changed, 10 insertions(+), 9 deletions(-)
> 
> diff --git a/commit-graph.c b/commit-graph.c
> index db623afd09..fa3b58e762 100644
> --- a/commit-graph.c
> +++ b/commit-graph.c
> @@ -327,12 +327,6 @@ static int graph_read_bloom_data(const unsigned char *chunk_start,
>  	uint32_t hash_version;
>  	hash_version = get_be32(chunk_start);
>  
> -	if (*c->commit_graph_changed_paths_version == -1) {
> -		*c->commit_graph_changed_paths_version = hash_version;
> -	} else if (hash_version != *c->commit_graph_changed_paths_version) {
> -		return 0;
> -	}
> -
>  	g->chunk_bloom_data = chunk_start;
>  	g->bloom_filter_settings = xmalloc(sizeof(struct bloom_filter_settings));
>  	g->bloom_filter_settings->hash_version = hash_version;
> @@ -2473,8 +2467,7 @@ int write_commit_graph(struct object_directory *odb,
>  	ctx->write_generation_data = (get_configured_generation_version(r) == 2);
>  	ctx->num_generation_data_overflows = 0;
>  
> -	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version == 2
> -		? 2 : 1;
> +	bloom_settings.hash_version = r->settings.commit_graph_changed_paths_version;
>  	bloom_settings.bits_per_entry = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_BITS_PER_ENTRY",
>  						      bloom_settings.bits_per_entry);
>  	bloom_settings.num_hashes = git_env_ulong("GIT_TEST_BLOOM_SETTINGS_NUM_HASHES",
> @@ -2506,10 +2499,18 @@ int write_commit_graph(struct object_directory *odb,
>  		/* We have changed-paths already. Keep them in the next graph */
>  		if (g && g->bloom_filter_settings) {
>  			ctx->changed_paths = 1;
> -			ctx->bloom_settings = g->bloom_filter_settings;
> +
> +			/* don't propagate the hash_version unless unspecified */
> +			if (bloom_settings.hash_version == -1)
> +				bloom_settings.hash_version = g->bloom_filter_settings->hash_version;
> +			bloom_settings.bits_per_entry = g->bloom_filter_settings->bits_per_entry;
> +			bloom_settings.num_hashes = g->bloom_filter_settings->num_hashes;
> +			bloom_settings.max_changed_paths = g->bloom_filter_settings->max_changed_paths;
>  		}
>  	}
>  
> +	bloom_settings.hash_version = bloom_settings.hash_version == 2 ? 2 : 1;
> +

What if there is a future version of Git that writes Bloom filters with
hash version 3? Should we really normalize that to `1`?

Patrick

>  	if (ctx->split) {
>  		struct commit_graph *g = ctx->r->objects->commit_graph;
>  
> -- 
> 2.42.0.342.g8bb3a896ee
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[PATCH v3 14/17] commit-graph: drop unnecessary `graph_read_bloom_data_context`

[Taylor Blau]

The `graph_read_bloom_data_context` struct was introduced in an earlier
commit in order to pass pointers to the commit-graph and changed-path
Bloom filter version when reading the BDAT chunk.

The previous commit no longer writes through the changed_paths_version
pointer, making the surrounding context structure unnecessary. Drop it
and pass a pointer to the commit-graph directly when reading the BDAT
chunk.

Noticed-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 commit-graph.c | 14 ++------------
 1 file changed, 2 insertions(+), 12 deletions(-)

diff --git a/commit-graph.c b/commit-graph.c
index fa3b58e762..e0fc62e110 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -314,16 +314,10 @@ static int graph_read_oid_lookup(const unsigned char *chunk_start,
 	return 0;
 }
 
-struct graph_read_bloom_data_context {
-	struct commit_graph *g;
-	int *commit_graph_changed_paths_version;
-};
-
 static int graph_read_bloom_data(const unsigned char *chunk_start,
 				  size_t chunk_size, void *data)
 {
-	struct graph_read_bloom_data_context *c = data;
-	struct commit_graph *g = c->g;
+	struct commit_graph *g = data;
 	uint32_t hash_version;
 	hash_version = get_be32(chunk_start);
 
@@ -415,14 +409,10 @@ struct commit_graph *parse_commit_graph(struct repo_settings *s,
 	}
 
 	if (s->commit_graph_changed_paths_version) {
-		struct graph_read_bloom_data_context context = {
-			.g = graph,
-			.commit_graph_changed_paths_version = &s->commit_graph_changed_paths_version
-		};
 		pair_chunk(cf, GRAPH_CHUNKID_BLOOMINDEXES,
 			   &graph->chunk_bloom_indexes);
 		read_chunk(cf, GRAPH_CHUNKID_BLOOMDATA,
-			   graph_read_bloom_data, &context);
+			   graph_read_bloom_data, graph);
 	}
 
 	if (graph->chunk_bloom_indexes && graph->chunk_bloom_data) {
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 15/17] object.h: fix mis-aligned flag bits table

[Taylor Blau]

Bit position 23 is one column too far to the left.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 object.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/object.h b/object.h
index 114d45954d..db25714b4e 100644
--- a/object.h
+++ b/object.h
@@ -62,7 +62,7 @@ void object_array_init(struct object_array *array);
 
 /*
  * object flag allocation:
- * revision.h:               0---------10         15             23------27
+ * revision.h:               0---------10         15               23------27
  * fetch-pack.c:             01    67
  * negotiator/default.c:       2--5
  * walker.c:                 0-2
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 16/17] commit-graph: reuse existing Bloom filters where possible

[Taylor Blau]

In 9e4df4da07 (commit-graph: new filter ver. that fixes murmur3,
2023-08-01), a bug was described where it's possible for Git to produce
non-murmur3 hashes when the platform's "char" type is signed, and there
are paths with characters whose highest bit is set (i.e. all characters
>= 0x80).

That patch allows the caller to control which version of Bloom filters
are read and written. However, even on platforms with a signed "char"
type, it is possible to reuse existing Bloom filters if and only if
there are no changed paths in any commit's first parent tree-diff whose
characters have their highest bit set.

When this is the case, we can reuse the existing filter without having
to compute a new one. This is done by marking trees which are known to
have (or not have) any such paths. When a commit's root tree is verified
to not have any such paths, we mark it as such and declare that the
commit's Bloom filter is reusable.

Note that this heuristic only goes in one direction. If neither a commit
nor its first parent have any paths in their trees with non-ASCII
characters, then we know for certain that a path with non-ASCII
characters will not appear in a tree-diff against that commit's first
parent. The reverse isn't necessarily true: just because the tree-diff
doesn't contain any such paths does not imply that no such paths exist
in either tree.

So we end up recomputing some Bloom filters that we don't strictly have
to (i.e. their bits are the same no matter which version of murmur3 we
use). But culling these out is impossible, since we'd have to perform
the full tree-diff, which is the same effort as computing the Bloom
filter from scratch.

But because we can cache our results in each tree's flag bits, we can
often avoid recomputing many filters, thereby reducing the time it takes
to run

    $ git commit-graph write --changed-paths --reachable

when upgrading from v1 to v2 Bloom filters.

To benchmark this, let's generate a commit-graph in linux.git with v1
changed-paths in generation order[^1]:

    $ git clone git@github.com:torvalds/linux.git
    $ cd linux
    $ git commit-graph write --reachable --changed-paths
    $ graph=".git/objects/info/commit-graph"
    $ mv $graph{,.bak}

Then let's time how long it takes to go from v1 to v2 filters (with and
without the upgrade path enabled), resetting the state of the
commit-graph each time:

    $ git config commitGraph.changedPathsVersion 2
    $ hyperfine -p 'cp -f $graph.bak $graph' -L v 0,1 \
        'GIT_TEST_UPGRADE_BLOOM_FILTERS={v} git.compile commit-graph write --reachable --changed-paths'

On linux.git (where there aren't any non-ASCII paths), the timings
indicate that this patch represents a speed-up over recomputing all
Bloom filters from scratch:

    Benchmark 1: GIT_TEST_UPGRADE_BLOOM_FILTERS=0 git.compile commit-graph write --reachable --changed-paths
      Time (mean ± σ):     124.873 s ±  0.316 s    [User: 124.081 s, System: 0.643 s]
      Range (min … max):   124.621 s … 125.227 s    3 runs

    Benchmark 2: GIT_TEST_UPGRADE_BLOOM_FILTERS=1 git.compile commit-graph write --reachable --changed-paths
      Time (mean ± σ):     79.271 s ±  0.163 s    [User: 74.611 s, System: 4.521 s]
      Range (min … max):   79.112 s … 79.437 s    3 runs

    Summary
      'GIT_TEST_UPGRADE_BLOOM_FILTERS=1 git.compile commit-graph write --reachable --changed-paths' ran
        1.58 ± 0.01 times faster than 'GIT_TEST_UPGRADE_BLOOM_FILTERS=0 git.compile commit-graph write --reachable --changed-paths'

On git.git, we do have some non-ASCII paths, giving us a more modest
improvement from 4.163 seconds to 3.348 seconds, for a 1.24x speed-up.
On my machine, the stats for git.git are:

  - 8,285 Bloom filters computed from scratch
  - 10 Bloom filters generated as empty
  - 4 Bloom filters generated as truncated due to too many changed paths
  - 65,114 Bloom filters were reused when transitioning from v1 to v2.

[^1]: Note that this is is important, since `--stdin-packs` or
  `--stdin-commits` orders commits in the commit-graph by their pack
  position (with `--stdin-packs`) or in the raw input (with
  `--stdin-commits`).

  Since we compute Bloom filters in the same order that commits appear
  in the graph, we must see a commit's (first) parent before we process
  the commit itself. This is only guaranteed to happen when sorting
  commits by their generation number.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c              | 90 ++++++++++++++++++++++++++++++++++++++++++--
 bloom.h              |  1 +
 commit-graph.c       |  5 +++
 object.h             |  1 +
 t/t4216-log-bloom.sh | 35 ++++++++++++++++-
 5 files changed, 127 insertions(+), 5 deletions(-)

diff --git a/bloom.c b/bloom.c
index 739fa093ba..24dd874e46 100644
--- a/bloom.c
+++ b/bloom.c
@@ -7,6 +7,9 @@
 #include "commit-graph.h"
 #include "commit.h"
 #include "commit-slab.h"
+#include "tree.h"
+#include "tree-walk.h"
+#include "config.h"
 
 define_commit_slab(bloom_filter_slab, struct bloom_filter);
 
@@ -250,6 +253,73 @@ static void init_truncated_large_filter(struct bloom_filter *filter,
 	filter->version = version;
 }
 
+#define VISITED   (1u<<21)
+#define HIGH_BITS (1u<<22)
+
+static int has_entries_with_high_bit(struct repository *r, struct tree *t)
+{
+	if (parse_tree(t))
+		return 1;
+
+	if (!(t->object.flags & VISITED)) {
+		struct tree_desc desc;
+		struct name_entry entry;
+
+		init_tree_desc(&desc, t->buffer, t->size);
+		while (tree_entry(&desc, &entry)) {
+			size_t i;
+			for (i = 0; i < entry.pathlen; i++) {
+				if (entry.path[i] & 0x80) {
+					t->object.flags |= HIGH_BITS;
+					goto done;
+				}
+			}
+
+			if (S_ISDIR(entry.mode)) {
+				struct tree *sub = lookup_tree(r, &entry.oid);
+				if (sub && has_entries_with_high_bit(r, sub)) {
+					t->object.flags |= HIGH_BITS;
+					goto done;
+				}
+			}
+
+		}
+
+done:
+		t->object.flags |= VISITED;
+	}
+
+	return !!(t->object.flags & HIGH_BITS);
+}
+
+static int commit_tree_has_high_bit_paths(struct repository *r,
+					  struct commit *c)
+{
+	struct tree *t;
+	if (repo_parse_commit(r, c))
+		return 1;
+	t = repo_get_commit_tree(r, c);
+	if (!t)
+		return 1;
+	return has_entries_with_high_bit(r, t);
+}
+
+static struct bloom_filter *upgrade_filter(struct repository *r, struct commit *c,
+					   struct bloom_filter *filter,
+					   int hash_version)
+{
+	struct commit_list *p = c->parents;
+	if (commit_tree_has_high_bit_paths(r, c))
+		return NULL;
+
+	if (p && commit_tree_has_high_bit_paths(r, p->item))
+		return NULL;
+
+	filter->version = hash_version;
+
+	return filter;
+}
+
 struct bloom_filter *get_bloom_filter(struct repository *r, struct commit *c)
 {
 	struct bloom_filter *filter;
@@ -292,9 +362,23 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 						     filter, graph_pos);
 	}
 
-	if ((filter->data && filter->len) &&
-	    (!settings || settings->hash_version == filter->version))
-		return filter;
+	if (filter->data && filter->len) {
+		struct bloom_filter *upgrade;
+		if (!settings || settings->hash_version == filter->version)
+			return filter;
+
+		/* version mismatch, see if we can upgrade */
+		if (compute_if_not_present &&
+		    git_env_bool("GIT_TEST_UPGRADE_BLOOM_FILTERS", 1)) {
+			upgrade = upgrade_filter(r, c, filter,
+						 settings->hash_version);
+			if (upgrade) {
+				if (computed)
+					*computed |= BLOOM_UPGRADED;
+				return upgrade;
+			}
+		}
+	}
 	if (!compute_if_not_present)
 		return NULL;
 
diff --git a/bloom.h b/bloom.h
index bfe389e29c..e3a9b68905 100644
--- a/bloom.h
+++ b/bloom.h
@@ -102,6 +102,7 @@ enum bloom_filter_computed {
 	BLOOM_COMPUTED     = (1 << 1),
 	BLOOM_TRUNC_LARGE  = (1 << 2),
 	BLOOM_TRUNC_EMPTY  = (1 << 3),
+	BLOOM_UPGRADED     = (1 << 4),
 };
 
 struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
diff --git a/commit-graph.c b/commit-graph.c
index e0fc62e110..571f38335a 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -1109,6 +1109,7 @@ struct write_commit_graph_context {
 	int count_bloom_filter_not_computed;
 	int count_bloom_filter_trunc_empty;
 	int count_bloom_filter_trunc_large;
+	int count_bloom_filter_upgraded;
 };
 
 static int write_graph_chunk_fanout(struct hashfile *f,
@@ -1716,6 +1717,8 @@ static void trace2_bloom_filter_write_statistics(struct write_commit_graph_conte
 			   ctx->count_bloom_filter_trunc_empty);
 	trace2_data_intmax("commit-graph", ctx->r, "filter-trunc-large",
 			   ctx->count_bloom_filter_trunc_large);
+	trace2_data_intmax("commit-graph", ctx->r, "filter-upgraded",
+			   ctx->count_bloom_filter_upgraded);
 }
 
 static void compute_bloom_filters(struct write_commit_graph_context *ctx)
@@ -1757,6 +1760,8 @@ static void compute_bloom_filters(struct write_commit_graph_context *ctx)
 				ctx->count_bloom_filter_trunc_empty++;
 			if (computed & BLOOM_TRUNC_LARGE)
 				ctx->count_bloom_filter_trunc_large++;
+		} else if (computed & BLOOM_UPGRADED) {
+			ctx->count_bloom_filter_upgraded++;
 		} else if (computed & BLOOM_NOT_COMPUTED)
 			ctx->count_bloom_filter_not_computed++;
 		ctx->total_bloom_filter_data_size += filter
diff --git a/object.h b/object.h
index db25714b4e..2e5e08725f 100644
--- a/object.h
+++ b/object.h
@@ -75,6 +75,7 @@ void object_array_init(struct object_array *array);
  * commit-reach.c:                                  16-----19
  * sha1-name.c:                                              20
  * list-objects-filter.c:                                      21
+ * bloom.c:                                                    2122
  * builtin/fsck.c:           0--3
  * builtin/gc.c:             0
  * builtin/index-pack.c:                                     2021
diff --git a/t/t4216-log-bloom.sh b/t/t4216-log-bloom.sh
index 8f8b5d4966..a321d8d713 100755
--- a/t/t4216-log-bloom.sh
+++ b/t/t4216-log-bloom.sh
@@ -221,6 +221,10 @@ test_filter_trunc_large () {
 	grep "\"key\":\"filter-trunc-large\",\"value\":\"$1\"" $2
 }
 
+test_filter_upgraded () {
+	grep "\"key\":\"filter-upgraded\",\"value\":\"$1\"" $2
+}
+
 test_expect_success 'correctly report changes over limit' '
 	git init limits &&
 	(
@@ -629,10 +633,19 @@ test_expect_success 'when writing another commit graph, preserve existing versio
 test_expect_success 'when writing commit graph, do not reuse changed-path of another version' '
 	git init doublewrite &&
 	test_commit -C doublewrite c "$CENT" &&
+
 	git -C doublewrite config --add commitgraph.changedPathsVersion 1 &&
-	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C doublewrite commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
 	git -C doublewrite config --add commitgraph.changedPathsVersion 2 &&
-	git -C doublewrite commit-graph write --reachable --changed-paths &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C doublewrite commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
 	(
 		cd doublewrite &&
 		echo "c01f" >expect &&
@@ -641,4 +654,22 @@ test_expect_success 'when writing commit graph, do not reuse changed-path of ano
 	)
 '
 
+test_expect_success 'when writing commit graph, reuse changed-path of another version where possible' '
+	git init upgrade &&
+
+	test_commit -C upgrade base no-high-bits &&
+
+	git -C upgrade config --add commitgraph.changedPathsVersion 1 &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C upgrade commit-graph write --reachable --changed-paths &&
+	test_filter_computed 1 trace2.txt &&
+	test_filter_upgraded 0 trace2.txt &&
+
+	git -C upgrade config --add commitgraph.changedPathsVersion 2 &&
+	GIT_TRACE2_EVENT="$(pwd)/trace2.txt" \
+		git -C upgrade commit-graph write --reachable --changed-paths &&
+	test_filter_computed 0 trace2.txt &&
+	test_filter_upgraded 1 trace2.txt
+'
+
 test_done
-- 
2.42.0.342.g8bb3a896ee

[PATCH v3 17/17] bloom: introduce `deinit_bloom_filters()`

[Taylor Blau]

After we are done using Bloom filters, we do not currently clean up any
memory allocated by the commit slab used to store those filters in the
first place.

Besides the bloom_filter structures themselves, there is mostly nothing
to free() in the first place, since in the read-only path all Bloom
filter's `data` members point to a memory mapped region in the
commit-graph file itself.

But when generating Bloom filters from scratch (or initializing
truncated filters) we allocate additional memory to store the filter's
data.

Keep track of when we need to free() this additional chunk of memory by
using an extra pointer `to_free`. Most of the time this will be NULL
(indicating that we are representing an existing Bloom filter stored in
a memory mapped region). When it is non-NULL, free it before discarding
the Bloom filters slab.

Suggested-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
---
 bloom.c        | 16 +++++++++++++++-
 bloom.h        |  3 +++
 commit-graph.c |  4 ++++
 3 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/bloom.c b/bloom.c
index 24dd874e46..ff131893cd 100644
--- a/bloom.c
+++ b/bloom.c
@@ -59,6 +59,7 @@ int load_bloom_filter_from_graph(struct commit_graph *g,
 					sizeof(unsigned char) * start_index +
 					BLOOMDATA_CHUNK_HEADER_SIZE);
 	filter->version = g->bloom_filter_settings->hash_version;
+	filter->to_free = NULL;
 
 	return 1;
 }
@@ -231,6 +232,18 @@ void init_bloom_filters(void)
 	init_bloom_filter_slab(&bloom_filters);
 }
 
+static void free_one_bloom_filter(struct bloom_filter *filter)
+{
+	if (!filter)
+		return;
+	free(filter->to_free);
+}
+
+void deinit_bloom_filters(void)
+{
+	deep_clear_bloom_filter_slab(&bloom_filters, free_one_bloom_filter);
+}
+
 static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 		       const struct hashmap_entry *eptr,
 		       const struct hashmap_entry *entry_or_key,
@@ -247,7 +260,7 @@ static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 static void init_truncated_large_filter(struct bloom_filter *filter,
 					int version)
 {
-	filter->data = xmalloc(1);
+	filter->data = filter->to_free = xmalloc(1);
 	filter->data[0] = 0xFF;
 	filter->len = 1;
 	filter->version = version;
@@ -449,6 +462,7 @@ struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
 			filter->len = 1;
 		}
 		CALLOC_ARRAY(filter->data, filter->len);
+		filter->to_free = filter->data;
 
 		hashmap_for_each_entry(&pathmap, &iter, e, entry) {
 			struct bloom_key key;
diff --git a/bloom.h b/bloom.h
index e3a9b68905..d20e64bfbb 100644
--- a/bloom.h
+++ b/bloom.h
@@ -56,6 +56,8 @@ struct bloom_filter {
 	unsigned char *data;
 	size_t len;
 	int version;
+
+	void *to_free;
 };
 
 /*
@@ -96,6 +98,7 @@ void add_key_to_filter(const struct bloom_key *key,
 		       const struct bloom_filter_settings *settings);
 
 void init_bloom_filters(void);
+void deinit_bloom_filters(void);
 
 enum bloom_filter_computed {
 	BLOOM_NOT_COMPUTED = (1 << 0),
diff --git a/commit-graph.c b/commit-graph.c
index 571f38335a..7ccec429cb 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -787,6 +787,7 @@ static void close_commit_graph_one(struct commit_graph *g)
 void close_commit_graph(struct raw_object_store *o)
 {
 	close_commit_graph_one(o->commit_graph);
+	deinit_bloom_filters();
 	o->commit_graph = NULL;
 }
 
@@ -2588,6 +2589,9 @@ int write_commit_graph(struct object_directory *odb,
 
 	res = write_commit_graph_file(ctx);
 
+	if (ctx->changed_paths)
+		deinit_bloom_filters();
+
 	if (ctx->split)
 		mark_commit_graphs(ctx);
 
-- 
2.42.0.342.g8bb3a896ee

Re: [PATCH v3 00/17] bloom: changed-path Bloom filters v2 (& sundries)

[Patrick Steinhardt]

[-- Attachment #1: Type: text/plain, Size: 10543 bytes --]

On Tue, Oct 10, 2023 at 04:33:17PM -0400, Taylor Blau wrote:
> (Rebased onto the tip of 'master', which is 3a06386e31 (The fifteenth
> batch, 2023-10-04), at the time of writing).
> 
> This series is a reroll of the combined efforts of [1] and [2] to
> introduce the v2 changed-path Bloom filters, which fixes a bug in our
> existing implementation of murmur3 paths with non-ASCII characters (when
> the "char" type is signed).
> 
> In large part, this is the same as the previous round. But this round
> includes some extra bits that address issues pointed out by SZEDER
> Gábor, which are:
> 
>   - not reading Bloom filters for root commits
>   - corrupting Bloom filter reads by tweaking the filter settings
>     between layers.
> 
> These issues were discussed in (among other places) [3], and [4],
> respectively.
> 
> Thanks to Jonathan, Peff, and SZEDER who have helped a great deal in
> assembling these patches. As usual, a range-diff is included below.
> Thanks in advance for your
> review!

As this patch series has been sitting around without reviews for a week
I've tried my best to give it a go. Note though that this area is mostly
outside of my own comfort zone, so some of the questions and suggestions
might ultimately not apply.

Patrick

> [1]: https://lore.kernel.org/git/cover.1684790529.git.jonathantanmy@google.com/
> [2]: https://lore.kernel.org/git/cover.1691426160.git.me@ttaylorr.com/
> [3]: https://public-inbox.org/git/20201015132147.GB24954@szeder.dev/
> [4]: https://lore.kernel.org/git/20230830200218.GA5147@szeder.dev/
> 
> Jonathan Tan (4):
>   gitformat-commit-graph: describe version 2 of BDAT
>   t4216: test changed path filters with high bit paths
>   repo-settings: introduce commitgraph.changedPathsVersion
>   commit-graph: new filter ver. that fixes murmur3
> 
> Taylor Blau (13):
>   t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
>   revision.c: consult Bloom filters for root commits
>   commit-graph: ensure Bloom filters are read with consistent settings
>   t/helper/test-read-graph.c: extract `dump_graph_info()`
>   bloom.h: make `load_bloom_filter_from_graph()` public
>   t/helper/test-read-graph: implement `bloom-filters` mode
>   bloom: annotate filters with hash version
>   bloom: prepare to discard incompatible Bloom filters
>   commit-graph.c: unconditionally load Bloom filters
>   commit-graph: drop unnecessary `graph_read_bloom_data_context`
>   object.h: fix mis-aligned flag bits table
>   commit-graph: reuse existing Bloom filters where possible
>   bloom: introduce `deinit_bloom_filters()`
> 
>  Documentation/config/commitgraph.txt     |  26 ++-
>  Documentation/gitformat-commit-graph.txt |   9 +-
>  bloom.c                                  | 208 ++++++++++++++++-
>  bloom.h                                  |  38 +++-
>  commit-graph.c                           |  61 ++++-
>  object.h                                 |   3 +-
>  oss-fuzz/fuzz-commit-graph.c             |   2 +-
>  repo-settings.c                          |   6 +-
>  repository.h                             |   2 +-
>  revision.c                               |  26 ++-
>  t/helper/test-bloom.c                    |   9 +-
>  t/helper/test-read-graph.c               |  67 ++++--
>  t/t0095-bloom.sh                         |   8 +
>  t/t4216-log-bloom.sh                     | 272 ++++++++++++++++++++++-
>  14 files changed, 682 insertions(+), 55 deletions(-)
> 
> Range-diff against v2:
> 10:  002a06d1e9 !  1:  fe671d616c t/t4216-log-bloom.sh: harden `test_bloom_filters_not_used()`
>     @@ Commit message
>          indicating that no filters were used.
>      
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## t/t4216-log-bloom.sh ##
>      @@ t/t4216-log-bloom.sh: test_bloom_filters_used () {
>  -:  ---------- >  2:  7d0fa93543 revision.c: consult Bloom filters for root commits
>  -:  ---------- >  3:  2ecc0a2d58 commit-graph: ensure Bloom filters are read with consistent settings
>  1:  5fa681b58e !  4:  17703ed89a gitformat-commit-graph: describe version 2 of BDAT
>     @@ Commit message
>          Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
>          Signed-off-by: Junio C Hamano <gitster@pobox.com>
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## Documentation/gitformat-commit-graph.txt ##
>      @@ Documentation/gitformat-commit-graph.txt: All multi-byte numbers are in network byte order.
>  2:  623d840575 !  5:  94552abf45 t/helper/test-read-graph.c: extract `dump_graph_info()`
>     @@ Commit message
>          Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
>          Signed-off-by: Junio C Hamano <gitster@pobox.com>
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## t/helper/test-read-graph.c ##
>      @@
>  3:  bc9d77ae60 !  6:  3d81efa27b bloom.h: make `load_bloom_filter_from_graph()` public
>     @@ Commit message
>          Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
>          Signed-off-by: Junio C Hamano <gitster@pobox.com>
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## bloom.c ##
>      @@ bloom.c: static inline unsigned char get_bitmask(uint32_t pos)
>  4:  ac7008aed3 !  7:  d23cd89037 t/helper/test-read-graph: implement `bloom-filters` mode
>     @@ Commit message
>          Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
>          Signed-off-by: Junio C Hamano <gitster@pobox.com>
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## t/helper/test-read-graph.c ##
>      @@ t/helper/test-read-graph.c: static void dump_graph_info(struct commit_graph *graph)
>     @@ t/helper/test-read-graph.c: int cmd__read_graph(int argc UNUSED, const char **ar
>      -	return 0;
>      +	return ret;
>       }
>     ++
>     ++
>  5:  71755ba856 !  8:  cba766f224 t4216: test changed path filters with high bit paths
>     @@ Commit message
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## t/t4216-log-bloom.sh ##
>     -@@ t/t4216-log-bloom.sh: test_expect_success 'Bloom generation backfills empty commits' '
>     - 	)
>     +@@ t/t4216-log-bloom.sh: test_expect_success 'merge graph layers with incompatible Bloom settings' '
>     + 	! grep "disabling Bloom filters" err
>       '
>       
>      +get_first_changed_path_filter () {
>  6:  9768d92c0f !  9:  a08a961f41 repo-settings: introduce commitgraph.changedPathsVersion
>     @@ Commit message
>          Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
>          Signed-off-by: Junio C Hamano <gitster@pobox.com>
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## Documentation/config/commitgraph.txt ##
>      @@ Documentation/config/commitgraph.txt: commitGraph.maxNewFilters::
>  7:  f911b4bfab = 10:  61d44519a5 commit-graph: new filter ver. that fixes murmur3
>  8:  35009900df ! 11:  a8c10f8de8 bloom: annotate filters with hash version
>     @@ Commit message
>          Bloom filter.
>      
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## bloom.c ##
>      @@ bloom.c: int load_bloom_filter_from_graph(struct commit_graph *g,
>  9:  138bc16905 ! 12:  2ba10a4b4b bloom: prepare to discard incompatible Bloom filters
>     @@ Commit message
>          `get_or_compute_bloom_filter()`.
>      
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## bloom.c ##
>      @@ bloom.c: static void init_truncated_large_filter(struct bloom_filter *filter,
> 11:  2437e62813 ! 13:  09d8669c3a commit-graph.c: unconditionally load Bloom filters
>     @@ Commit message
>          either "1" or "2".
>      
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## commit-graph.c ##
>      @@ commit-graph.c: static int graph_read_bloom_data(const unsigned char *chunk_start,
> 12:  fe8fb2f5fe ! 14:  0d4f9dc4ee commit-graph: drop unnecessary `graph_read_bloom_data_context`
>     @@ Commit message
>      
>          Noticed-by: Jonathan Tan <jonathantanmy@google.com>
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## commit-graph.c ##
>      @@ commit-graph.c: static int graph_read_oid_lookup(const unsigned char *chunk_start,
> 13:  825af91e11 ! 15:  1f7f27bc47 object.h: fix mis-aligned flag bits table
>     @@ Commit message
>          Bit position 23 is one column too far to the left.
>      
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## object.h ##
>      @@ object.h: void object_array_init(struct object_array *array);
> 14:  593b317192 ! 16:  abbef95ae8 commit-graph: reuse existing Bloom filters where possible
>     @@ Commit message
>            commits by their generation number.
>      
>          Signed-off-by: Taylor Blau <me@ttaylorr.com>
>     -    Signed-off-by: Junio C Hamano <gitster@pobox.com>
>     -    Signed-off-by: Taylor Blau <me@ttaylorr.com>
>      
>       ## bloom.c ##
>      @@
> 15:  8bf2c9cf98 = 17:  ca362408d5 bloom: introduce `deinit_bloom_filters()`
> -- 
> 2.42.0.342.g8bb3a896ee

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v3 00/17] bloom: changed-path Bloom filters v2 (& sundries)

[Taylor Blau]

On Tue, Oct 17, 2023 at 10:45:36AM +0200, Patrick Steinhardt wrote:
> > Thanks to Jonathan, Peff, and SZEDER who have helped a great deal in
> > assembling these patches. As usual, a range-diff is included below.
> > Thanks in advance for your
> > review!
>
> As this patch series has been sitting around without reviews for a week
> I've tried my best to give it a go. Note though that this area is mostly
> outside of my own comfort zone, so some of the questions and suggestions
> might ultimately not apply.

Thanks for giving it a look! I generated a few small tweaks on top of
what I already had here based on your review, so I'll send a reroll
shortly.

Thanks,
Taylor