[PATCH 0/4] odb: eagerly load alternates

[PATCH 0/4] odb: eagerly load alternates

[Patrick Steinhardt]

Hi,

when initializing the object database we only eagerly initialize the
primary object database source. If the primary source has alternates,
those alternates are only initialized the first time we really access
the object database.

When introduced in ace1534d6f (Introduce SHA1_FILE_DIRECTORIES to
support multiple object databases., 2005-05-07), alternates were
originally only loaded when a given object wasn't found in the primary
object database. This was also reinforced by later optimization, for
example in 693d2bc625 (Attempt to delay prepare_alt_odb during get_sha1,
2007-05-26), where we tried to avoid loading alternates in even more
cases. But as Git has evolved, we eventually started to eagerly parse
alternates all over the codebase, including on every single object
lookup, and consequently deferring this operation does not really buy us
much anymore.

The result of this is that we have calls to `odb_prepare_alternates()`
cluttered all over the code base. This is somewhat awkward, and as
almost every Git command ends up reading objects at it doesn't even buy
us anything.

This patch series thus gets rid of the lazy-loading. Besides simplifying
the codebase a bit, it also prepares us for moving alternates into the
"files" backend as discussed in [1].

The series is built on top of 010afd3166 (The 12th batch, 2026-08-07)
with ps/odb-make-creation-pluggable at e927cfeb21 (odb: make creation of
on-disk structures pluggable, 2026-08-07) merged into it.

Thanks!

Patrick

[1]: <amLgMqkqxR8mKIbT@pks.im>

---
Patrick Steinhardt (4):
      odb: decouple source path comparisons from `the_repository`
      odb: eagerly initialize alternates
      odb: drop `loaded_alternates` field
      odb: drop `alternates_db` field

 builtin/fsck.c         |   3 --
 builtin/pack-objects.c |   3 --
 commit-graph.c         |   4 --
 loose.c                |   1 -
 object-name.c          |   1 -
 odb.c                  | 106 ++++++++++++++++++++++++-------------------------
 odb.h                  |  22 +++++-----
 odb/source.h           |   7 ++++
 odb/streaming.c        |   1 -
 pack-bitmap.c          |   2 -
 packfile.c             |   1 -
 packfile.h             |   2 -
 12 files changed, 68 insertions(+), 85 deletions(-)


---
base-commit: f6ad67a7977439ad8351d42e6ccfd11f714db765
change-id: 20260804-pks-odb-eagerly-prepare-alternates-3efb0a38e0dd

[PATCH 1/4] odb: decouple source path comparisons from `the_repository`

[Patrick Steinhardt]

When registering alternates we deduplicate object database sources by
their path so that the same source won't be added twice. Ever since
cf2dc1c238 (speed up alt_odb_usable() with many alternates, 2021-07-07)
this duplicate check is backed by a map keyed by the source's path,
using `fspathhash()` and `fspatheq()` as hash and equality functions,
respectively.

These functions are problematic in this context for two reasons:

  - They implicitly depend on `the_repository` instead of the
    repository that owns the object database.

  - They derive case-sensitivity from `repo_ignore_case()`, which
    returns a default value in case the repository's configuration has
    not been parsed yet. Object database sources may be registered
    before that is the case, so the answer may flip depending on when a
    source gets registered.

Fix this by making the comparison self-contained in the object
database. Instead of using `fspathhash()` and `fspatheq()` we resolve
"core.ignoreCase" manually and then use the correct comparison function
based on the result. This requires us to migrate to a `struct hashmap`,
as the khash interface does not give us the ability to change these
functions.

Note that we can unconditionally use `strihash()` to compute entry
hashes regardless of case sensitivity: a hash function only needs to
guarantee that equal keys have equal hashes, and a case-insensitive
hash satisfies this requirement for both case-sensitive and
case-insensitive equality.

Overall it's quite debatable whether all of this complexity really is
worth it, or whether we should just linearly search through all sources
to find duplicates. But the mentioned commit cares about cases with
thousands of alternates, and a linear search would of course regress
performance quite a bit. This doesn't really feel like a reasonable case
to care about though, but I don't feel comfortable regressing it anyway.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 odb.c        | 63 ++++++++++++++++++++++++++++++++++++++++--------------------
 odb.h        | 15 ++++++++++++++-
 odb/source.h |  7 +++++++
 3 files changed, 63 insertions(+), 22 deletions(-)

diff --git a/odb.c b/odb.c
index bd02d8ad54..51da386f22 100644
--- a/odb.c
+++ b/odb.c
@@ -2,11 +2,10 @@
 #include "abspath.h"
 #include "commit-graph.h"
 #include "config.h"
-#include "dir.h"
 #include "environment.h"
 #include "gettext.h"
+#include "hashmap.h"
 #include "hex.h"
-#include "khash.h"
 #include "lockfile.h"
 #include "loose.h"
 #include "midx.h"
@@ -29,8 +28,32 @@
 #include "trace2.h"
 #include "write-or-die.h"
 
-KHASH_INIT(odb_path_map, const char * /* key: odb_path */,
-	struct odb_source *, 1, fspathhash, fspatheq)
+static int odb_source_paths_cmp(struct object_database *o,
+				const char *a, const char *b)
+{
+	if (o->source_paths_icase < 0) {
+		int icase = 0;
+		repo_config_get_bool(o->repo, "core.ignorecase", &icase);
+		o->source_paths_icase = icase;
+	}
+
+	return o->source_paths_icase ? strcasecmp(a, b) : strcmp(a, b);
+}
+
+static int odb_source_by_path_cmp(const void *cb_data,
+				  const struct hashmap_entry *entry,
+				  const struct hashmap_entry *entry_or_key,
+				  const void *keydata)
+{
+	struct object_database *o = (struct object_database *)cb_data;
+	const struct odb_source *source = container_of(entry, const struct odb_source, by_path_entry);
+	const char *path = keydata;
+
+	if (!path)
+		path = container_of(entry_or_key, const struct odb_source, by_path_entry)->path;
+
+	return odb_source_paths_cmp(o, source->path, path);
+}
 
 int odb_mkstemp(struct object_database *odb,
 		struct strbuf *temp_filename, const char *pattern)
@@ -58,8 +81,8 @@ int odb_mkstemp(struct object_database *odb,
  */
 static bool odb_is_source_usable(struct object_database *o, const char *path)
 {
-	int r;
 	struct strbuf normalized_objdir = STRBUF_INIT;
+	struct hashmap_entry key;
 	bool usable = false;
 
 	strbuf_realpath(&normalized_objdir, o->sources->path, 1);
@@ -76,20 +99,18 @@ static bool odb_is_source_usable(struct object_database *o, const char *path)
 	 * Prevent the common mistake of listing the same
 	 * thing twice, or object directory itself.
 	 */
-	if (!o->source_by_path) {
-		khiter_t p;
-
-		o->source_by_path = kh_init_odb_path_map();
+	if (!hashmap_get_size(&o->source_by_path)) {
 		assert(!o->sources->next);
-		p = kh_put_odb_path_map(o->source_by_path, o->sources->path, &r);
-		assert(r == 1); /* never used */
-		kh_value(o->source_by_path, p) = o->sources;
+		hashmap_entry_init(&o->sources->by_path_entry,
+				   strihash(o->sources->path));
+		hashmap_add(&o->source_by_path, &o->sources->by_path_entry);
 	}
 
-	if (fspatheq(path, normalized_objdir.buf))
+	if (!odb_source_paths_cmp(o, path, normalized_objdir.buf))
 		goto out;
 
-	if (kh_get_odb_path_map(o->source_by_path, path) < kh_end(o->source_by_path))
+	hashmap_entry_init(&key, strihash(path));
+	if (hashmap_get(&o->source_by_path, &key, path))
 		goto out;
 
 	usable = true;
@@ -172,8 +193,6 @@ static struct odb_source *odb_add_alternate_recursively(struct object_database *
 {
 	struct odb_source *alternate = NULL;
 	struct strvec sources = STRVEC_INIT;
-	khiter_t pos;
-	int ret;
 
 	if (!odb_is_source_usable(odb, source))
 		goto error;
@@ -184,10 +203,11 @@ static struct odb_source *odb_add_alternate_recursively(struct object_database *
 	*odb->sources_tail = alternate;
 	odb->sources_tail = &(alternate->next);
 
-	pos = kh_put_odb_path_map(odb->source_by_path, alternate->path, &ret);
-	if (!ret)
+	hashmap_entry_init(&alternate->by_path_entry, strihash(alternate->path));
+	if (hashmap_get(&odb->source_by_path, &alternate->by_path_entry,
+			alternate->path))
 		BUG("source must not yet exist");
-	kh_value(odb->source_by_path, pos) = alternate;
+	hashmap_add(&odb->source_by_path, &alternate->by_path_entry);
 
 	/* recursively add alternates */
 	odb_source_read_alternates(alternate, &sources);
@@ -1056,6 +1076,8 @@ struct object_database *odb_new(struct repository *repo,
 	o->repo = repo;
 	pthread_mutex_init(&o->replace_mutex, NULL);
 	string_list_init_dup(&o->submodule_source_paths);
+	hashmap_init(&o->source_by_path, odb_source_by_path_cmp, o, 0);
+	o->source_paths_icase = -1;
 
 	if (flags & ODB_NEW_HONOR_ENV) {
 		primary_source = xstrdup_or_null(getenv(DB_ENVIRONMENT));
@@ -1094,8 +1116,7 @@ static void odb_free_sources(struct object_database *o)
 	odb_source_free(o->inmemory_objects);
 	o->inmemory_objects = NULL;
 
-	kh_destroy_odb_path_map(o->source_by_path);
-	o->source_by_path = NULL;
+	hashmap_clear(&o->source_by_path);
 }
 
 void odb_free(struct object_database *o)
diff --git a/odb.h b/odb.h
index 8eb4e85d64..71af7450a9 100644
--- a/odb.h
+++ b/odb.h
@@ -1,6 +1,7 @@
 #ifndef ODB_H
 #define ODB_H
 
+#include "hashmap.h"
 #include "object.h"
 #include "oidset.h"
 #include "oidmap.h"
@@ -54,7 +55,19 @@ struct object_database {
 	 */
 	struct odb_source *sources;
 	struct odb_source **sources_tail;
-	struct kh_odb_path_map *source_by_path;
+
+	/*
+	 * Map of object database sources, keyed by their respective paths.
+	 * This map is used to detect the case where the same source is
+	 * registered multiple times.
+	 */
+	struct hashmap source_by_path;
+
+	/*
+	 * Whether source paths shall be compared case-insensitively, as
+	 * determined by "core.ignoreCase".
+	 */
+	int source_paths_icase;
 
 	int loaded_alternates;
 
diff --git a/odb/source.h b/odb/source.h
index 4bc037b8d6..82cda8ad75 100644
--- a/odb/source.h
+++ b/odb/source.h
@@ -1,6 +1,7 @@
 #ifndef ODB_SOURCE_H
 #define ODB_SOURCE_H
 
+#include "hashmap.h"
 #include "object.h"
 #include "odb.h"
 #include "odb/transaction.h"
@@ -50,6 +51,12 @@ struct strvec;
 struct odb_source {
 	struct odb_source *next;
 
+	/*
+	 * Entry in the object database's map of sources, keyed by this
+	 * source's path.
+	 */
+	struct hashmap_entry by_path_entry;
+
 	/* Object database that owns this object source. */
 	struct object_database *odb;
 

-- 
2.55.0.679.g6767b8d81c.dirty

[PATCH 2/4] odb: eagerly initialize alternates

[Patrick Steinhardt]

When creating the object database we initialize the main object database
source, but we don't yet initialize its alternates. Instead, we have
many calls to `odb_prepare_alternates()` cluttered around the code base
whenever we are about to iterate through the sources.

This lazy loading doesn't really add much value: the moment where read
any object we _have_ to load the alternates anyway. So given that most
of our commands would access the object database this optimization is
not really buying us much in the first place. Quite on the contrary, it
makes the code harder to understand and is a potential source of bugs in
case any callsite forgot to prepare alternates before we iterate through
the sources.

Historically though there was a reason why we deferred lazy-loading: it
may happen that the repository has "core.ignoreCase" configured, and we
use that to deduplicate the list of alternates in case we had the same
alternate configured multiple times, but with different casing. We used
to initialize the object database before we had fully configured the
owning repository though, and consequently we couldn't access that
configuration yet. This has changed in the preceding commit though where
we started to parse "core.ignoreCase" manually.

Eagerly prepare alternates both when creating the object database and
when flushing its caches. Drop the now-unneeded calls to prepare the
alternates that are scattered across the code base.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 builtin/fsck.c         |  3 ---
 builtin/pack-objects.c |  3 ---
 commit-graph.c         |  4 ----
 loose.c                |  1 -
 object-name.c          |  1 -
 odb.c                  | 26 ++++----------------------
 odb.h                  |  6 ------
 odb/streaming.c        |  1 -
 pack-bitmap.c          |  2 --
 packfile.c             |  1 -
 packfile.h             |  2 --
 11 files changed, 4 insertions(+), 46 deletions(-)

diff --git a/builtin/fsck.c b/builtin/fsck.c
index a6c054e45b..892c5661d9 100644
--- a/builtin/fsck.c
+++ b/builtin/fsck.c
@@ -1069,7 +1069,6 @@ int cmd_fsck(int argc,
 		odb_for_each_object(repo->objects, NULL,
 				    mark_object_for_connectivity, repo, 0);
 	} else {
-		odb_prepare_alternates(repo->objects);
 		for (source = repo->objects->sources; source; source = source->next)
 			fsck_source(repo, source);
 
@@ -1155,7 +1154,6 @@ int cmd_fsck(int argc,
 	if (repo->settings.core_commit_graph) {
 		struct child_process commit_graph_verify = CHILD_PROCESS_INIT;
 
-		odb_prepare_alternates(repo->objects);
 		for (source = repo->objects->sources; source; source = source->next) {
 			child_process_init(&commit_graph_verify);
 			commit_graph_verify.git_cmd = 1;
@@ -1173,7 +1171,6 @@ int cmd_fsck(int argc,
 	if (repo->settings.core_multi_pack_index) {
 		struct child_process midx_verify = CHILD_PROCESS_INIT;
 
-		odb_prepare_alternates(repo->objects);
 		for (source = repo->objects->sources; source; source = source->next) {
 			child_process_init(&midx_verify);
 			midx_verify.git_cmd = 1;
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index 1ec5b6f206..48d37e8e32 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -1779,8 +1779,6 @@ static int want_object_in_pack_mtime(const struct object_id *oid,
 		*found_offset = 0;
 	}
 
-	odb_prepare_alternates(the_repository->objects);
-
 	for (source = the_repository->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		struct multi_pack_index *m = get_multi_pack_index(files->packed);
@@ -4520,7 +4518,6 @@ static void add_objects_in_unpacked_packs(void)
 		.source_infop = &source_info,
 	};
 
-	odb_prepare_alternates(to_pack.repo->objects);
 	for (source = to_pack.repo->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 
diff --git a/commit-graph.c b/commit-graph.c
index 49e8f63930..983c11ce85 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -651,8 +651,6 @@ struct commit_graph *load_commit_graph_chain_fd_st(struct object_database *odb,
 	count = st->st_size / (odb->repo->hash_algo->hexsz + 1);
 	CALLOC_ARRAY(oids, count);
 
-	odb_prepare_alternates(odb);
-
 	for (i = 0; i < count; i++) {
 		struct odb_source *source;
 
@@ -768,7 +766,6 @@ static struct commit_graph *prepare_commit_graph(struct repository *r)
 	if (!commit_graph_compatible(r))
 		return NULL;
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		r->objects->commit_graph = read_commit_graph_one(source);
 		if (r->objects->commit_graph)
@@ -2018,7 +2015,6 @@ static void fill_oids_from_all_packs(struct write_commit_graph_context *ctx)
 			_("Finding commits for commit graph among packed objects"),
 			ctx->approx_nr_objects);
 
-	odb_prepare_alternates(ctx->r->objects);
 	for (source = ctx->r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		odb_source_for_each_object(&files->packed->base, &oi, add_packed_commits_oi,
diff --git a/loose.c b/loose.c
index aa3cb1b4fc..c159d29d2d 100644
--- a/loose.c
+++ b/loose.c
@@ -115,7 +115,6 @@ int repo_read_loose_object_map(struct repository *repo)
 {
 	struct odb_source *source;
 
-	odb_prepare_alternates(repo->objects);
 	for (source = repo->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		if (loose_object_map_load(files->loose) < 0)
diff --git a/object-name.c b/object-name.c
index 83efba0ba6..34a08d76dd 100644
--- a/object-name.c
+++ b/object-name.c
@@ -280,7 +280,6 @@ static int init_object_disambiguation(struct repository *r,
 
 	ds->len = len;
 	ds->repo = r;
-	odb_prepare_alternates(r->objects);
 	return 0;
 }
 
diff --git a/odb.c b/odb.c
index 51da386f22..2ae8228dd2 100644
--- a/odb.c
+++ b/odb.c
@@ -237,11 +237,6 @@ void odb_add_to_alternates_file(struct object_database *odb,
 struct odb_source *odb_add_to_alternates_memory(struct object_database *odb,
 						const char *dir)
 {
-	/*
-	 * Make sure alternates are initialized, or else our entry may be
-	 * overwritten when they are.
-	 */
-	odb_prepare_alternates(odb);
 	return odb_add_alternate_recursively(odb, dir, 0);
 }
 
@@ -250,12 +245,6 @@ struct odb_source *odb_set_temporary_primary_source(struct object_database *odb,
 {
 	struct odb_source *source;
 
-	/*
-	 * Make sure alternates are initialized, or else our entry may be
-	 * overwritten when they are.
-	 */
-	odb_prepare_alternates(odb);
-
 	/*
 	 * Make a new primary odb and link the old primary ODB in as an
 	 * alternate
@@ -361,7 +350,6 @@ struct odb_source *odb_find_source(struct object_database *odb, const char *obj_
 	char *obj_dir_real = real_pathdup(obj_dir, 1);
 	struct strbuf odb_path_real = STRBUF_INIT;
 
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next) {
 		strbuf_realpath(&odb_path_real, source->path, 1);
 		if (!strcmp(obj_dir_real, odb_path_real.buf))
@@ -495,7 +483,6 @@ int odb_for_each_alternate(struct object_database *odb,
 	struct odb_source *alternate;
 	int r = 0;
 
-	odb_prepare_alternates(odb);
 	for (alternate = odb->sources->next; alternate; alternate = alternate->next) {
 		r = cb(alternate, payload);
 		if (r)
@@ -504,7 +491,7 @@ int odb_for_each_alternate(struct object_database *odb,
 	return r;
 }
 
-void odb_prepare_alternates(struct object_database *odb)
+static void odb_prepare_alternates(struct object_database *odb)
 {
 	struct strvec sources = STRVEC_INIT;
 
@@ -523,7 +510,6 @@ void odb_prepare_alternates(struct object_database *odb)
 
 int odb_has_alternates(struct object_database *odb)
 {
-	odb_prepare_alternates(odb);
 	return !!odb->sources->next;
 }
 
@@ -583,8 +569,6 @@ static int do_oid_object_info_extended(struct object_database *odb,
 	if (!odb_source_read_object_info(odb->inmemory_objects, oid, oi, flags))
 		return 0;
 
-	odb_prepare_alternates(odb);
-
 	while (1) {
 		struct odb_source *source;
 
@@ -847,7 +831,6 @@ int odb_freshen_object(struct object_database *odb,
 		       const struct object_id *oid)
 {
 	struct odb_source *source;
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next)
 		if (odb_source_freshen_object(source, oid, NULL))
 			return 1;
@@ -862,7 +845,6 @@ int odb_for_each_object_ext(struct object_database *odb,
 {
 	int ret;
 
-	odb_prepare_alternates(odb);
 	for (struct odb_source *source = odb->sources; source; source = source->next) {
 		if (opts->flags & ODB_FOR_EACH_OBJECT_LOCAL_ONLY && !source->local)
 			continue;
@@ -900,7 +882,6 @@ int odb_count_objects(struct object_database *odb,
 		return 0;
 	}
 
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next) {
 		unsigned long c;
 
@@ -980,7 +961,6 @@ int odb_find_abbrev_len(struct object_database *odb,
 		goto out;
 	}
 
-	odb_prepare_alternates(odb);
 	for (struct odb_source *source = odb->sources; source; source = source->next) {
 		ret = odb_source_find_abbrev_len(source, oid, len, &len);
 		if (ret)
@@ -1091,6 +1071,8 @@ struct object_database *odb_new(struct repository *repo,
 	o->alternate_db = secondary_sources;
 	o->inmemory_objects = &odb_source_inmemory_new(o)->base;
 
+	odb_prepare_alternates(o);
+
 	free(primary_source);
 	return o;
 }
@@ -1151,10 +1133,10 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
 	 */
 	if (flags & ODB_PREPARE_FLUSH_CACHES) {
 		o->loaded_alternates = 0;
+		odb_prepare_alternates(o);
 		o->object_count_valid = 0;
 	}
 
-	odb_prepare_alternates(o);
 	for (source = o->sources; source; source = source->next)
 		odb_source_prepare(source, flags);
 
diff --git a/odb.h b/odb.h
index 71af7450a9..fbafee174b 100644
--- a/odb.h
+++ b/odb.h
@@ -273,12 +273,6 @@ void odb_for_each_alternate_ref(struct object_database *odb,
 int odb_mkstemp(struct object_database *odb,
 		struct strbuf *temp_filename, const char *pattern);
 
-/*
- * Prepare alternate object sources for the given database by reading
- * "objects/info/alternates" and opening the respective sources.
- */
-void odb_prepare_alternates(struct object_database *odb);
-
 /*
  * Check whether the object database has any alternates. The primary object
  * source does not count as alternate.
diff --git a/odb/streaming.c b/odb/streaming.c
index 20531e864c..37642768e9 100644
--- a/odb/streaming.c
+++ b/odb/streaming.c
@@ -184,7 +184,6 @@ static int istream_source(struct odb_read_stream **out,
 {
 	struct odb_source *source;
 
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next)
 		if (!odb_source_read_object_stream(out, source, oid))
 			return 0;
diff --git a/pack-bitmap.c b/pack-bitmap.c
index e85bd69ba4..e0fb57d332 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -717,7 +717,6 @@ static int open_bitmap(struct repository *r,
 
 	assert(!bitmap_git->map);
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 
@@ -3417,7 +3416,6 @@ int verify_bitmap_files(struct repository *r)
 	struct packed_git *p;
 	int res = 0;
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		struct multi_pack_index *m = get_multi_pack_index(files->packed);
diff --git a/packfile.c b/packfile.c
index 0eee45055f..d870de90ed 100644
--- a/packfile.c
+++ b/packfile.c
@@ -1938,7 +1938,6 @@ int has_object_pack(struct repository *r, const struct object_id *oid)
 {
 	struct odb_source *source;
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		if (!odb_source_read_object_info(&files->packed->base, oid, NULL, 0))
diff --git a/packfile.h b/packfile.h
index e1f77152b5..10de24f477 100644
--- a/packfile.h
+++ b/packfile.h
@@ -77,8 +77,6 @@ static inline struct repo_for_each_pack_data repo_for_eack_pack_data_init(struct
 {
 	struct repo_for_each_pack_data data = { 0 };
 
-	odb_prepare_alternates(repo->objects);
-
 	for (struct odb_source *source = repo->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		struct packfile_list_entry *entry = packfile_store_get_packs(files->packed);

-- 
2.55.0.679.g6767b8d81c.dirty

[PATCH 3/4] odb: drop `loaded_alternates` field

[Patrick Steinhardt]

The `struct object_database::loaded_alternates` field tells us whether
or not alternates have been loaded already. This field was useful before
the preceding commit as we were indeed lazy-loading alternates. But now
that we started to eagerly load them we can assume them to be loaded
after `odb_new()`, and hence the field does not serve any purpose
anymore.

Remove it.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 odb.c | 9 +--------
 odb.h | 2 --
 2 files changed, 1 insertion(+), 10 deletions(-)

diff --git a/odb.c b/odb.c
index 2ae8228dd2..2eb37a2f44 100644
--- a/odb.c
+++ b/odb.c
@@ -230,8 +230,7 @@ void odb_add_to_alternates_file(struct object_database *odb,
 	int ret = odb_source_write_alternate(odb->sources, dir);
 	if (ret < 0)
 		die(NULL);
-	if (odb->loaded_alternates)
-		odb_add_alternate_recursively(odb, dir, 0);
+	odb_add_alternate_recursively(odb, dir, 0);
 }
 
 struct odb_source *odb_add_to_alternates_memory(struct object_database *odb,
@@ -495,16 +494,11 @@ static void odb_prepare_alternates(struct object_database *odb)
 {
 	struct strvec sources = STRVEC_INIT;
 
-	if (odb->loaded_alternates)
-		return;
-
 	parse_alternates(odb->alternate_db, PATH_SEP, NULL, &sources);
 	odb_source_read_alternates(odb->sources, &sources);
 	for (size_t i = 0; i < sources.nr; i++)
 		odb_add_alternate_recursively(odb, sources.v[i], 0);
 
-	odb->loaded_alternates = 1;
-
 	strvec_clear(&sources);
 }
 
@@ -1132,7 +1126,6 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
 	 * the lifetime of the process.
 	 */
 	if (flags & ODB_PREPARE_FLUSH_CACHES) {
-		o->loaded_alternates = 0;
 		odb_prepare_alternates(o);
 		o->object_count_valid = 0;
 	}
diff --git a/odb.h b/odb.h
index fbafee174b..aefb34213f 100644
--- a/odb.h
+++ b/odb.h
@@ -69,8 +69,6 @@ struct object_database {
 	 */
 	int source_paths_icase;
 
-	int loaded_alternates;
-
 	/*
 	 * A list of alternate object directories loaded from the environment;
 	 * this should not generally need to be accessed directly, but will

-- 
2.55.0.679.g6767b8d81c.dirty

[PATCH 4/4] odb: drop `alternates_db` field

[Patrick Steinhardt]

The `struct object_database::alternates_db` field tracks the value of
the "GIT_ALTERNATE_OBJECT_DIRECTORIES" environment variable and is
used in `odb_prepare_alternates()`. It's not necessary to store it as a
separate field anymore though, as we stopped lazy-loading alternates.
Consequently, we can simply pass it to `odb_prepare_alternates()` via
`odb_new()` now.

Do so and remove the field.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 odb.c | 14 +++++++-------
 odb.h |  7 -------
 2 files changed, 7 insertions(+), 14 deletions(-)

diff --git a/odb.c b/odb.c
index 2eb37a2f44..fc21199f80 100644
--- a/odb.c
+++ b/odb.c
@@ -490,12 +490,14 @@ int odb_for_each_alternate(struct object_database *odb,
 	return r;
 }
 
-static void odb_prepare_alternates(struct object_database *odb)
+static void odb_prepare_alternates(struct object_database *odb,
+				   const char *alternate_db)
 {
 	struct strvec sources = STRVEC_INIT;
 
-	parse_alternates(odb->alternate_db, PATH_SEP, NULL, &sources);
+	parse_alternates(alternate_db, PATH_SEP, NULL, &sources);
 	odb_source_read_alternates(odb->sources, &sources);
+
 	for (size_t i = 0; i < sources.nr; i++)
 		odb_add_alternate_recursively(odb, sources.v[i], 0);
 
@@ -1062,11 +1064,11 @@ struct object_database *odb_new(struct repository *repo,
 
 	o->sources = odb_source_new(o, primary_source, true);
 	o->sources_tail = &o->sources->next;
-	o->alternate_db = secondary_sources;
 	o->inmemory_objects = &odb_source_inmemory_new(o)->base;
 
-	odb_prepare_alternates(o);
+	odb_prepare_alternates(o, secondary_sources);
 
+	free(secondary_sources);
 	free(primary_source);
 	return o;
 }
@@ -1100,8 +1102,6 @@ void odb_free(struct object_database *o)
 	if (!o)
 		return;
 
-	free(o->alternate_db);
-
 	oidmap_clear(&o->replace_map, 1);
 	pthread_mutex_destroy(&o->replace_mutex);
 
@@ -1126,7 +1126,7 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
 	 * the lifetime of the process.
 	 */
 	if (flags & ODB_PREPARE_FLUSH_CACHES) {
-		odb_prepare_alternates(o);
+		odb_prepare_alternates(o, NULL);
 		o->object_count_valid = 0;
 	}
 
diff --git a/odb.h b/odb.h
index aefb34213f..748366a610 100644
--- a/odb.h
+++ b/odb.h
@@ -69,13 +69,6 @@ struct object_database {
 	 */
 	int source_paths_icase;
 
-	/*
-	 * A list of alternate object directories loaded from the environment;
-	 * this should not generally need to be accessed directly, but will
-	 * populate the "sources" list when odb_prepare_alternates() is run.
-	 */
-	char *alternate_db;
-
 	/*
 	 * Objects that should be substituted by other objects
 	 * (see git-replace(1)).

-- 
2.55.0.679.g6767b8d81c.dirty

Re: [PATCH 1/4] odb: decouple source path comparisons from `the_repository`

[Justin Tobler]

On 26/08/10 03:33PM, Patrick Steinhardt wrote:
> When registering alternates we deduplicate object database sources by
> their path so that the same source won't be added twice. Ever since
> cf2dc1c238 (speed up alt_odb_usable() with many alternates, 2021-07-07)
> this duplicate check is backed by a map keyed by the source's path,
> using `fspathhash()` and `fspatheq()` as hash and equality functions,
> respectively.
> 
> These functions are problematic in this context for two reasons:
> 
>   - They implicitly depend on `the_repository` instead of the
>     repository that owns the object database.
> 
>   - They derive case-sensitivity from `repo_ignore_case()`, which
>     returns a default value in case the repository's configuration has
>     not been parsed yet. Object database sources may be registered
>     before that is the case, so the answer may flip depending on when a
>     source gets registered.

Are alternates currently always registered after repository
configuration has been parsed? Or is this an existing bug?

> Fix this by making the comparison self-contained in the object
> database. Instead of using `fspathhash()` and `fspatheq()` we resolve
> "core.ignoreCase" manually and then use the correct comparison function
> based on the result. This requires us to migrate to a `struct hashmap`,
> as the khash interface does not give us the ability to change these
> functions.
> 
> Note that we can unconditionally use `strihash()` to compute entry
> hashes regardless of case sensitivity: a hash function only needs to
> guarantee that equal keys have equal hashes, and a case-insensitive
> hash satisfies this requirement for both case-sensitive and
> case-insensitive equality.

Ok IIUC, even if we want to be case-sensitive, its ok to use
`strihash()` and have hash collisions because the compare function will
still properly distinguish between the cases. Makes sense.

> Overall it's quite debatable whether all of this complexity really is
> worth it, or whether we should just linearly search through all sources
> to find duplicates. But the mentioned commit cares about cases with
> thousands of alternates, and a linear search would of course regress
> performance quite a bit. This doesn't really feel like a reasonable case
> to care about though, but I don't feel comfortable regressing it anyway.

Ya, my first though here was also whether all of this song and dance is
really needed for alternates. There may be someone out there with tons
of alternates I guess though. Probably good to be on the safe side.

> Signed-off-by: Patrick Steinhardt <ps@pks.im>
> ---
>  odb.c        | 63 ++++++++++++++++++++++++++++++++++++++++--------------------
>  odb.h        | 15 ++++++++++++++-
>  odb/source.h |  7 +++++++
>  3 files changed, 63 insertions(+), 22 deletions(-)
> 
> diff --git a/odb.c b/odb.c
> index bd02d8ad54..51da386f22 100644
> --- a/odb.c
> +++ b/odb.c
> @@ -2,11 +2,10 @@
>  #include "abspath.h"
>  #include "commit-graph.h"
>  #include "config.h"
> -#include "dir.h"
>  #include "environment.h"
>  #include "gettext.h"
> +#include "hashmap.h"
>  #include "hex.h"
> -#include "khash.h"
>  #include "lockfile.h"
>  #include "loose.h"
>  #include "midx.h"
> @@ -29,8 +28,32 @@
>  #include "trace2.h"
>  #include "write-or-die.h"
>  
> -KHASH_INIT(odb_path_map, const char * /* key: odb_path */,
> -	struct odb_source *, 1, fspathhash, fspatheq)
> +static int odb_source_paths_cmp(struct object_database *o,
> +				const char *a, const char *b)
> +{
> +	if (o->source_paths_icase < 0) {
> +		int icase = 0;
> +		repo_config_get_bool(o->repo, "core.ignorecase", &icase);
> +		o->source_paths_icase = icase;
> +	}

We now parse ignorecase configuration here directly and store the result
in `source_paths_icase`. This ensures configuration is correctly applied
regardless of whether repository configuration has been fully read yet.

> +	return o->source_paths_icase ? strcasecmp(a, b) : strcmp(a, b);
> +}
> +
> +static int odb_source_by_path_cmp(const void *cb_data,
> +				  const struct hashmap_entry *entry,
> +				  const struct hashmap_entry *entry_or_key,
> +				  const void *keydata)
> +{
> +	struct object_database *o = (struct object_database *)cb_data;
> +	const struct odb_source *source = container_of(entry, const struct odb_source, by_path_entry);
> +	const char *path = keydata;
> +
> +	if (!path)
> +		path = container_of(entry_or_key, const struct odb_source, by_path_entry)->path;
> +
> +	return odb_source_paths_cmp(o, source->path, path);
> +}

Here is the comparison callback that is used for the hashmap.

>  int odb_mkstemp(struct object_database *odb,
>  		struct strbuf *temp_filename, const char *pattern)
> @@ -58,8 +81,8 @@ int odb_mkstemp(struct object_database *odb,
>   */
>  static bool odb_is_source_usable(struct object_database *o, const char *path)
>  {
> -	int r;
>  	struct strbuf normalized_objdir = STRBUF_INIT;
> +	struct hashmap_entry key;
>  	bool usable = false;
>  
>  	strbuf_realpath(&normalized_objdir, o->sources->path, 1);
> @@ -76,20 +99,18 @@ static bool odb_is_source_usable(struct object_database *o, const char *path)
>  	 * Prevent the common mistake of listing the same
>  	 * thing twice, or object directory itself.
>  	 */
> -	if (!o->source_by_path) {
> -		khiter_t p;
> -
> -		o->source_by_path = kh_init_odb_path_map();
> +	if (!hashmap_get_size(&o->source_by_path)) {
>  		assert(!o->sources->next);
> -		p = kh_put_odb_path_map(o->source_by_path, o->sources->path, &r);
> -		assert(r == 1); /* never used */
> -		kh_value(o->source_by_path, p) = o->sources;
> +		hashmap_entry_init(&o->sources->by_path_entry,
> +				   strihash(o->sources->path));
> +		hashmap_add(&o->source_by_path, &o->sources->by_path_entry);

The hashmap is lazily set up with the primary source. I do find some of
the variable names like "source_by_path" a bit vague, but that isn't
really anything new here.

>  	}
>  
> -	if (fspatheq(path, normalized_objdir.buf))
> +	if (!odb_source_paths_cmp(o, path, normalized_objdir.buf))
>  		goto out;

If the path matches the first entry in the sources list then we know it
is not an alternate.

>  
> -	if (kh_get_odb_path_map(o->source_by_path, path) < kh_end(o->source_by_path))
> +	hashmap_entry_init(&key, strihash(path));
> +	if (hashmap_get(&o->source_by_path, &key, path))
>  		goto out;

If the alternates source cannot be found for the given path, then we
also know it is not a usuable alternate.

>  	usable = true;
> @@ -172,8 +193,6 @@ static struct odb_source *odb_add_alternate_recursively(struct object_database *
>  {
>  	struct odb_source *alternate = NULL;
>  	struct strvec sources = STRVEC_INIT;
> -	khiter_t pos;
> -	int ret;
>  
>  	if (!odb_is_source_usable(odb, source))
>  		goto error;
> @@ -184,10 +203,11 @@ static struct odb_source *odb_add_alternate_recursively(struct object_database *
>  	*odb->sources_tail = alternate;
>  	odb->sources_tail = &(alternate->next);
>  
> -	pos = kh_put_odb_path_map(odb->source_by_path, alternate->path, &ret);
> -	if (!ret)
> +	hashmap_entry_init(&alternate->by_path_entry, strihash(alternate->path));
> +	if (hashmap_get(&odb->source_by_path, &alternate->by_path_entry,
> +			alternate->path))
>  		BUG("source must not yet exist");
> -	kh_value(odb->source_by_path, pos) = alternate;
> +	hashmap_add(&odb->source_by_path, &alternate->by_path_entry);

Here is where alternates get registered and added to the hashmap. Makes
sense.

Overall this patch looks good.

-Justin

Re: [PATCH 2/4] odb: eagerly initialize alternates

[Justin Tobler]

On 26/08/10 03:33PM, Patrick Steinhardt wrote:
> When creating the object database we initialize the main object database
> source, but we don't yet initialize its alternates. Instead, we have
> many calls to `odb_prepare_alternates()` cluttered around the code base
> whenever we are about to iterate through the sources.
> 
> This lazy loading doesn't really add much value: the moment where read

Should this say "where we read" instead?

> any object we _have_ to load the alternates anyway. So given that most
> of our commands would access the object database this optimization is
> not really buying us much in the first place. Quite on the contrary, it
> makes the code harder to understand and is a potential source of bugs in
> case any callsite forgot to prepare alternates before we iterate through
> the sources.
> 
> Historically though there was a reason why we deferred lazy-loading: it
> may happen that the repository has "core.ignoreCase" configured, and we
> use that to deduplicate the list of alternates in case we had the same
> alternate configured multiple times, but with different casing. We used
> to initialize the object database before we had fully configured the
> owning repository though, and consequently we couldn't access that
> configuration yet. This has changed in the preceding commit though where
> we started to parse "core.ignoreCase" manually.
> 
> Eagerly prepare alternates both when creating the object database and
> when flushing its caches. Drop the now-unneeded calls to prepare the
> alternates that are scattered across the code base.

This sounds like the right direction and overall much simpler. Nice.

> Signed-off-by: Patrick Steinhardt <ps@pks.im>
> ---
[snip]
> @@ -1091,6 +1071,8 @@ struct object_database *odb_new(struct repository *repo,
>  	o->alternate_db = secondary_sources;
>  	o->inmemory_objects = &odb_source_inmemory_new(o)->base;
>  
> +	odb_prepare_alternates(o);

Now we eagerly prepare alternates at time of ODB creation.

> +
>  	free(primary_source);
>  	return o;
>  }
> @@ -1151,10 +1133,10 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
>  	 */
>  	if (flags & ODB_PREPARE_FLUSH_CACHES) {
>  		o->loaded_alternates = 0;
> +		odb_prepare_alternates(o);
>  		o->object_count_valid = 0;
>  	}

Ok we also invoke `odb_prepare_alternates()` when we need to refresh all
alternate sources. Makes sense.

The rest of this patch just removes the now-unneeded
`odb_prepare_alternates()` call sites and looks good.

-Justin

Re: [PATCH 3/4] odb: drop `loaded_alternates` field

[Justin Tobler]

On 26/08/10 03:33PM, Patrick Steinhardt wrote:
> The `struct object_database::loaded_alternates` field tells us whether
> or not alternates have been loaded already. This field was useful before
> the preceding commit as we were indeed lazy-loading alternates. But now
> that we started to eagerly load them we can assume them to be loaded
> after `odb_new()`, and hence the field does not serve any purpose
> anymore.

Now that alternates are eagerly set up, it is safe to assume, if we have
an ODB, the alternates have been loaded. Makes sense.

> Remove it.
> 
> Signed-off-by: Patrick Steinhardt <ps@pks.im>
> ---
[snip]
> @@ -1132,7 +1126,6 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
>  	 * the lifetime of the process.
>  	 */
>  	if (flags & ODB_PREPARE_FLUSH_CACHES) {
> -		o->loaded_alternates = 0;
>  		odb_prepare_alternates(o);

Also nice to see this go away as I thought it was little bit awkward to
unset it just to allow the us to reprepare the alternates.

-Justin

Re: [PATCH 4/4] odb: drop `alternates_db` field

[Justin Tobler]

On 26/08/10 03:33PM, Patrick Steinhardt wrote:
> The `struct object_database::alternates_db` field tracks the value of
> the "GIT_ALTERNATE_OBJECT_DIRECTORIES" environment variable and is
> used in `odb_prepare_alternates()`. It's not necessary to store it as a
> separate field anymore though, as we stopped lazy-loading alternates.
> Consequently, we can simply pass it to `odb_prepare_alternates()` via
> `odb_new()` now.
> 
> Do so and remove the field.
> 
> Signed-off-by: Patrick Steinhardt <ps@pks.im>
> ---
[snip]
> @@ -1126,7 +1126,7 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
>  	 * the lifetime of the process.
>  	 */
>  	if (flags & ODB_PREPARE_FLUSH_CACHES) {
> -		odb_prepare_alternates(o);
> +		odb_prepare_alternates(o, NULL);
>  		o->object_count_valid = 0;
>  	}

Naive question: is the reason we don't need to wire the
`GIT_ALTERNATE_OBJECT_DIRECTORIES` environment variable here because
they have already been added as sources? IOW, when we invoke
`odb_prepare_alternates()` after the initial set up, we only really care
about re-reading the alternates file.

-Justin

Re: [PATCH 1/4] odb: decouple source path comparisons from `the_repository`

[Patrick Steinhardt]

On Tue, Aug 11, 2026 at 05:04:58PM -0500, Justin Tobler wrote:
> On 26/08/10 03:33PM, Patrick Steinhardt wrote:
> > When registering alternates we deduplicate object database sources by
> > their path so that the same source won't be added twice. Ever since
> > cf2dc1c238 (speed up alt_odb_usable() with many alternates, 2021-07-07)
> > this duplicate check is backed by a map keyed by the source's path,
> > using `fspathhash()` and `fspatheq()` as hash and equality functions,
> > respectively.
> > 
> > These functions are problematic in this context for two reasons:
> > 
> >   - They implicitly depend on `the_repository` instead of the
> >     repository that owns the object database.
> > 
> >   - They derive case-sensitivity from `repo_ignore_case()`, which
> >     returns a default value in case the repository's configuration has
> >     not been parsed yet. Object database sources may be registered
> >     before that is the case, so the answer may flip depending on when a
> >     source gets registered.
> 
> Are alternates currently always registered after repository
> configuration has been parsed? Or is this an existing bug?

They are, because of the lazy-loading. So this is not a bug, we merely
have to ensure that we retain this behaviour.

> > Overall it's quite debatable whether all of this complexity really is
> > worth it, or whether we should just linearly search through all sources
> > to find duplicates. But the mentioned commit cares about cases with
> > thousands of alternates, and a linear search would of course regress
> > performance quite a bit. This doesn't really feel like a reasonable case
> > to care about though, but I don't feel comfortable regressing it anyway.
> 
> Ya, my first though here was also whether all of this song and dance is
> really needed for alternates. There may be someone out there with tons
> of alternates I guess though. Probably good to be on the safe side.

cf2dc1c238 (speed up alt_odb_usable() with many alternates, 2021-07-07)
mentions a repository with 100k alternates in total, but that's an
artificial testing setup. I doubt you can get any kind of reasonable
performance out of such a repository, regardless of whether on not
parsing the alternates is going to be fast.

For now though I didn't want to remove this infra. It feels overblown,
but it's not an unmaintainable mess, either.

Patrick

Re: [PATCH 2/4] odb: eagerly initialize alternates

[Patrick Steinhardt]

On Tue, Aug 11, 2026 at 05:15:33PM -0500, Justin Tobler wrote:
> On 26/08/10 03:33PM, Patrick Steinhardt wrote:
> > When creating the object database we initialize the main object database
> > source, but we don't yet initialize its alternates. Instead, we have
> > many calls to `odb_prepare_alternates()` cluttered around the code base
> > whenever we are about to iterate through the sources.
> > 
> > This lazy loading doesn't really add much value: the moment where read
> 
> Should this say "where we read" instead?

Yes, indeed.

Patrick

Re: [PATCH 4/4] odb: drop `alternates_db` field

[Patrick Steinhardt]

On Tue, Aug 11, 2026 at 05:31:17PM -0500, Justin Tobler wrote:
> On 26/08/10 03:33PM, Patrick Steinhardt wrote:
> > The `struct object_database::alternates_db` field tracks the value of
> > the "GIT_ALTERNATE_OBJECT_DIRECTORIES" environment variable and is
> > used in `odb_prepare_alternates()`. It's not necessary to store it as a
> > separate field anymore though, as we stopped lazy-loading alternates.
> > Consequently, we can simply pass it to `odb_prepare_alternates()` via
> > `odb_new()` now.
> > 
> > Do so and remove the field.
> > 
> > Signed-off-by: Patrick Steinhardt <ps@pks.im>
> > ---
> [snip]
> > @@ -1126,7 +1126,7 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
> >  	 * the lifetime of the process.
> >  	 */
> >  	if (flags & ODB_PREPARE_FLUSH_CACHES) {
> > -		odb_prepare_alternates(o);
> > +		odb_prepare_alternates(o, NULL);
> >  		o->object_count_valid = 0;
> >  	}
> 
> Naive question: is the reason we don't need to wire the
> `GIT_ALTERNATE_OBJECT_DIRECTORIES` environment variable here because
> they have already been added as sources? IOW, when we invoke
> `odb_prepare_alternates()` after the initial set up, we only really care
> about re-reading the alternates file.

Yes, exactly. We set up alternates exactly once in `odb_new()`, and we
don't expect the environment variable to ever change in a running
process. And as `odb_prepare_alternates()` only adds but never removes
any it's fine to ignore those here.

I'll add a comment.

Patrick

[PATCH v2 0/4] odb: eagerly load alternates

[Patrick Steinhardt]

Hi,

when initializing the object database we only eagerly initialize the
primary object database source. If the primary source has alternates,
those alternates are only initialized the first time we really access
the object database.

When introduced in ace1534d6f (Introduce SHA1_FILE_DIRECTORIES to
support multiple object databases., 2005-05-07), alternates were
originally only loaded when a given object wasn't found in the primary
object database. This was also reinforced by later optimization, for
example in 693d2bc625 (Attempt to delay prepare_alt_odb during get_sha1,
2007-05-26), where we tried to avoid loading alternates in even more
cases. But as Git has evolved, we eventually started to eagerly parse
alternates all over the codebase, including on every single object
lookup, and consequently deferring this operation does not really buy us
much anymore.

The result of this is that we have calls to `odb_prepare_alternates()`
cluttered all over the code base. This is somewhat awkward, and as
almost every Git command ends up reading objects at it doesn't even buy
us anything.

This patch series thus gets rid of the lazy-loading. Besides simplifying
the codebase a bit, it also prepares us for moving alternates into the
"files" backend as discussed in [1].

The series is built on top of 010afd3166 (The 12th batch, 2026-08-07)
with ps/odb-make-creation-pluggable at e927cfeb21 (odb: make creation of
on-disk structures pluggable, 2026-08-07) merged into it.

Changes in v2:
  - Add a missing word to a commit message.
  - Explain why we don't have to handle GIT_ALTERNATE_OBJECT_DIRECTORIES
    when re-preparing the object database.
  - Link to v1: https://patch.msgid.link/20260810-pks-odb-eagerly-prepare-alternates-v1-0-f0fa4a4004e1@pks.im

Thanks!

Patrick

[1]: <amLgMqkqxR8mKIbT@pks.im>

---
Patrick Steinhardt (4):
      odb: decouple source path comparisons from `the_repository`
      odb: eagerly initialize alternates
      odb: drop `loaded_alternates` field
      odb: drop `alternates_db` field

 builtin/fsck.c         |   3 --
 builtin/pack-objects.c |   3 --
 commit-graph.c         |   4 --
 loose.c                |   1 -
 object-name.c          |   1 -
 odb.c                  | 109 ++++++++++++++++++++++++-------------------------
 odb.h                  |  22 +++++-----
 odb/source.h           |   7 ++++
 odb/streaming.c        |   1 -
 pack-bitmap.c          |   2 -
 packfile.c             |   1 -
 packfile.h             |   2 -
 12 files changed, 70 insertions(+), 86 deletions(-)

Range-diff versus v1:

1:  25802adffa = 1:  721907c60d odb: decouple source path comparisons from `the_repository`
2:  1e73b730d8 ! 2:  3b2c23566c odb: eagerly initialize alternates
    @@ Commit message
         many calls to `odb_prepare_alternates()` cluttered around the code base
         whenever we are about to iterate through the sources.
     
    -    This lazy loading doesn't really add much value: the moment where read
    -    any object we _have_ to load the alternates anyway. So given that most
    -    of our commands would access the object database this optimization is
    -    not really buying us much in the first place. Quite on the contrary, it
    -    makes the code harder to understand and is a potential source of bugs in
    -    case any callsite forgot to prepare alternates before we iterate through
    -    the sources.
    +    This lazy loading doesn't really add much value: the moment where we
    +    read any object we _have_ to load the alternates anyway. So given that
    +    most of our commands would access the object database this optimization
    +    is not really buying us much in the first place. Quite on the contrary,
    +    it makes the code harder to understand and is a potential source of bugs
    +    in case any callsite forgot to prepare alternates before we iterate
    +    through the sources.
     
         Historically though there was a reason why we deferred lazy-loading: it
         may happen that the repository has "core.ignoreCase" configured, and we
3:  2ca1aa2a37 = 3:  df5d7df91d odb: drop `loaded_alternates` field
4:  1e97c93bdf ! 4:  50a37ef385 odb: drop `alternates_db` field
    @@ odb.c: void odb_free(struct object_database *o)
      	pthread_mutex_destroy(&o->replace_mutex);
      
     @@ odb.c: void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
    - 	 * the lifetime of the process.
    + 	 * Reprepare alt odbs, in case the alternates file was modified
    + 	 * during the course of this process. This only _adds_ odbs to
    + 	 * the linked list, so existing odbs will continue to exist for
    +-	 * the lifetime of the process.
    ++	 * the lifetime of the process. Consequently, we don't have to
    ++	 * reprocess GIT_ALTERNATE_OBJECT_DIRECTORIES here.
      	 */
      	if (flags & ODB_PREPARE_FLUSH_CACHES) {
     -		odb_prepare_alternates(o);

---
base-commit: f6ad67a7977439ad8351d42e6ccfd11f714db765
change-id: 20260804-pks-odb-eagerly-prepare-alternates-3efb0a38e0dd

[PATCH v2 1/4] odb: decouple source path comparisons from `the_repository`

[Patrick Steinhardt]

When registering alternates we deduplicate object database sources by
their path so that the same source won't be added twice. Ever since
cf2dc1c238 (speed up alt_odb_usable() with many alternates, 2021-07-07)
this duplicate check is backed by a map keyed by the source's path,
using `fspathhash()` and `fspatheq()` as hash and equality functions,
respectively.

These functions are problematic in this context for two reasons:

  - They implicitly depend on `the_repository` instead of the
    repository that owns the object database.

  - They derive case-sensitivity from `repo_ignore_case()`, which
    returns a default value in case the repository's configuration has
    not been parsed yet. Object database sources may be registered
    before that is the case, so the answer may flip depending on when a
    source gets registered.

Fix this by making the comparison self-contained in the object
database. Instead of using `fspathhash()` and `fspatheq()` we resolve
"core.ignoreCase" manually and then use the correct comparison function
based on the result. This requires us to migrate to a `struct hashmap`,
as the khash interface does not give us the ability to change these
functions.

Note that we can unconditionally use `strihash()` to compute entry
hashes regardless of case sensitivity: a hash function only needs to
guarantee that equal keys have equal hashes, and a case-insensitive
hash satisfies this requirement for both case-sensitive and
case-insensitive equality.

Overall it's quite debatable whether all of this complexity really is
worth it, or whether we should just linearly search through all sources
to find duplicates. But the mentioned commit cares about cases with
thousands of alternates, and a linear search would of course regress
performance quite a bit. This doesn't really feel like a reasonable case
to care about though, but I don't feel comfortable regressing it anyway.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 odb.c        | 63 ++++++++++++++++++++++++++++++++++++++++--------------------
 odb.h        | 15 ++++++++++++++-
 odb/source.h |  7 +++++++
 3 files changed, 63 insertions(+), 22 deletions(-)

diff --git a/odb.c b/odb.c
index bd02d8ad54..51da386f22 100644
--- a/odb.c
+++ b/odb.c
@@ -2,11 +2,10 @@
 #include "abspath.h"
 #include "commit-graph.h"
 #include "config.h"
-#include "dir.h"
 #include "environment.h"
 #include "gettext.h"
+#include "hashmap.h"
 #include "hex.h"
-#include "khash.h"
 #include "lockfile.h"
 #include "loose.h"
 #include "midx.h"
@@ -29,8 +28,32 @@
 #include "trace2.h"
 #include "write-or-die.h"
 
-KHASH_INIT(odb_path_map, const char * /* key: odb_path */,
-	struct odb_source *, 1, fspathhash, fspatheq)
+static int odb_source_paths_cmp(struct object_database *o,
+				const char *a, const char *b)
+{
+	if (o->source_paths_icase < 0) {
+		int icase = 0;
+		repo_config_get_bool(o->repo, "core.ignorecase", &icase);
+		o->source_paths_icase = icase;
+	}
+
+	return o->source_paths_icase ? strcasecmp(a, b) : strcmp(a, b);
+}
+
+static int odb_source_by_path_cmp(const void *cb_data,
+				  const struct hashmap_entry *entry,
+				  const struct hashmap_entry *entry_or_key,
+				  const void *keydata)
+{
+	struct object_database *o = (struct object_database *)cb_data;
+	const struct odb_source *source = container_of(entry, const struct odb_source, by_path_entry);
+	const char *path = keydata;
+
+	if (!path)
+		path = container_of(entry_or_key, const struct odb_source, by_path_entry)->path;
+
+	return odb_source_paths_cmp(o, source->path, path);
+}
 
 int odb_mkstemp(struct object_database *odb,
 		struct strbuf *temp_filename, const char *pattern)
@@ -58,8 +81,8 @@ int odb_mkstemp(struct object_database *odb,
  */
 static bool odb_is_source_usable(struct object_database *o, const char *path)
 {
-	int r;
 	struct strbuf normalized_objdir = STRBUF_INIT;
+	struct hashmap_entry key;
 	bool usable = false;
 
 	strbuf_realpath(&normalized_objdir, o->sources->path, 1);
@@ -76,20 +99,18 @@ static bool odb_is_source_usable(struct object_database *o, const char *path)
 	 * Prevent the common mistake of listing the same
 	 * thing twice, or object directory itself.
 	 */
-	if (!o->source_by_path) {
-		khiter_t p;
-
-		o->source_by_path = kh_init_odb_path_map();
+	if (!hashmap_get_size(&o->source_by_path)) {
 		assert(!o->sources->next);
-		p = kh_put_odb_path_map(o->source_by_path, o->sources->path, &r);
-		assert(r == 1); /* never used */
-		kh_value(o->source_by_path, p) = o->sources;
+		hashmap_entry_init(&o->sources->by_path_entry,
+				   strihash(o->sources->path));
+		hashmap_add(&o->source_by_path, &o->sources->by_path_entry);
 	}
 
-	if (fspatheq(path, normalized_objdir.buf))
+	if (!odb_source_paths_cmp(o, path, normalized_objdir.buf))
 		goto out;
 
-	if (kh_get_odb_path_map(o->source_by_path, path) < kh_end(o->source_by_path))
+	hashmap_entry_init(&key, strihash(path));
+	if (hashmap_get(&o->source_by_path, &key, path))
 		goto out;
 
 	usable = true;
@@ -172,8 +193,6 @@ static struct odb_source *odb_add_alternate_recursively(struct object_database *
 {
 	struct odb_source *alternate = NULL;
 	struct strvec sources = STRVEC_INIT;
-	khiter_t pos;
-	int ret;
 
 	if (!odb_is_source_usable(odb, source))
 		goto error;
@@ -184,10 +203,11 @@ static struct odb_source *odb_add_alternate_recursively(struct object_database *
 	*odb->sources_tail = alternate;
 	odb->sources_tail = &(alternate->next);
 
-	pos = kh_put_odb_path_map(odb->source_by_path, alternate->path, &ret);
-	if (!ret)
+	hashmap_entry_init(&alternate->by_path_entry, strihash(alternate->path));
+	if (hashmap_get(&odb->source_by_path, &alternate->by_path_entry,
+			alternate->path))
 		BUG("source must not yet exist");
-	kh_value(odb->source_by_path, pos) = alternate;
+	hashmap_add(&odb->source_by_path, &alternate->by_path_entry);
 
 	/* recursively add alternates */
 	odb_source_read_alternates(alternate, &sources);
@@ -1056,6 +1076,8 @@ struct object_database *odb_new(struct repository *repo,
 	o->repo = repo;
 	pthread_mutex_init(&o->replace_mutex, NULL);
 	string_list_init_dup(&o->submodule_source_paths);
+	hashmap_init(&o->source_by_path, odb_source_by_path_cmp, o, 0);
+	o->source_paths_icase = -1;
 
 	if (flags & ODB_NEW_HONOR_ENV) {
 		primary_source = xstrdup_or_null(getenv(DB_ENVIRONMENT));
@@ -1094,8 +1116,7 @@ static void odb_free_sources(struct object_database *o)
 	odb_source_free(o->inmemory_objects);
 	o->inmemory_objects = NULL;
 
-	kh_destroy_odb_path_map(o->source_by_path);
-	o->source_by_path = NULL;
+	hashmap_clear(&o->source_by_path);
 }
 
 void odb_free(struct object_database *o)
diff --git a/odb.h b/odb.h
index 8eb4e85d64..71af7450a9 100644
--- a/odb.h
+++ b/odb.h
@@ -1,6 +1,7 @@
 #ifndef ODB_H
 #define ODB_H
 
+#include "hashmap.h"
 #include "object.h"
 #include "oidset.h"
 #include "oidmap.h"
@@ -54,7 +55,19 @@ struct object_database {
 	 */
 	struct odb_source *sources;
 	struct odb_source **sources_tail;
-	struct kh_odb_path_map *source_by_path;
+
+	/*
+	 * Map of object database sources, keyed by their respective paths.
+	 * This map is used to detect the case where the same source is
+	 * registered multiple times.
+	 */
+	struct hashmap source_by_path;
+
+	/*
+	 * Whether source paths shall be compared case-insensitively, as
+	 * determined by "core.ignoreCase".
+	 */
+	int source_paths_icase;
 
 	int loaded_alternates;
 
diff --git a/odb/source.h b/odb/source.h
index 4bc037b8d6..82cda8ad75 100644
--- a/odb/source.h
+++ b/odb/source.h
@@ -1,6 +1,7 @@
 #ifndef ODB_SOURCE_H
 #define ODB_SOURCE_H
 
+#include "hashmap.h"
 #include "object.h"
 #include "odb.h"
 #include "odb/transaction.h"
@@ -50,6 +51,12 @@ struct strvec;
 struct odb_source {
 	struct odb_source *next;
 
+	/*
+	 * Entry in the object database's map of sources, keyed by this
+	 * source's path.
+	 */
+	struct hashmap_entry by_path_entry;
+
 	/* Object database that owns this object source. */
 	struct object_database *odb;
 

-- 
2.55.0.679.g6767b8d81c.dirty

[PATCH v2 2/4] odb: eagerly initialize alternates

[Patrick Steinhardt]

When creating the object database we initialize the main object database
source, but we don't yet initialize its alternates. Instead, we have
many calls to `odb_prepare_alternates()` cluttered around the code base
whenever we are about to iterate through the sources.

This lazy loading doesn't really add much value: the moment where we
read any object we _have_ to load the alternates anyway. So given that
most of our commands would access the object database this optimization
is not really buying us much in the first place. Quite on the contrary,
it makes the code harder to understand and is a potential source of bugs
in case any callsite forgot to prepare alternates before we iterate
through the sources.

Historically though there was a reason why we deferred lazy-loading: it
may happen that the repository has "core.ignoreCase" configured, and we
use that to deduplicate the list of alternates in case we had the same
alternate configured multiple times, but with different casing. We used
to initialize the object database before we had fully configured the
owning repository though, and consequently we couldn't access that
configuration yet. This has changed in the preceding commit though where
we started to parse "core.ignoreCase" manually.

Eagerly prepare alternates both when creating the object database and
when flushing its caches. Drop the now-unneeded calls to prepare the
alternates that are scattered across the code base.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 builtin/fsck.c         |  3 ---
 builtin/pack-objects.c |  3 ---
 commit-graph.c         |  4 ----
 loose.c                |  1 -
 object-name.c          |  1 -
 odb.c                  | 26 ++++----------------------
 odb.h                  |  6 ------
 odb/streaming.c        |  1 -
 pack-bitmap.c          |  2 --
 packfile.c             |  1 -
 packfile.h             |  2 --
 11 files changed, 4 insertions(+), 46 deletions(-)

diff --git a/builtin/fsck.c b/builtin/fsck.c
index a6c054e45b..892c5661d9 100644
--- a/builtin/fsck.c
+++ b/builtin/fsck.c
@@ -1069,7 +1069,6 @@ int cmd_fsck(int argc,
 		odb_for_each_object(repo->objects, NULL,
 				    mark_object_for_connectivity, repo, 0);
 	} else {
-		odb_prepare_alternates(repo->objects);
 		for (source = repo->objects->sources; source; source = source->next)
 			fsck_source(repo, source);
 
@@ -1155,7 +1154,6 @@ int cmd_fsck(int argc,
 	if (repo->settings.core_commit_graph) {
 		struct child_process commit_graph_verify = CHILD_PROCESS_INIT;
 
-		odb_prepare_alternates(repo->objects);
 		for (source = repo->objects->sources; source; source = source->next) {
 			child_process_init(&commit_graph_verify);
 			commit_graph_verify.git_cmd = 1;
@@ -1173,7 +1171,6 @@ int cmd_fsck(int argc,
 	if (repo->settings.core_multi_pack_index) {
 		struct child_process midx_verify = CHILD_PROCESS_INIT;
 
-		odb_prepare_alternates(repo->objects);
 		for (source = repo->objects->sources; source; source = source->next) {
 			child_process_init(&midx_verify);
 			midx_verify.git_cmd = 1;
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index 1ec5b6f206..48d37e8e32 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -1779,8 +1779,6 @@ static int want_object_in_pack_mtime(const struct object_id *oid,
 		*found_offset = 0;
 	}
 
-	odb_prepare_alternates(the_repository->objects);
-
 	for (source = the_repository->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		struct multi_pack_index *m = get_multi_pack_index(files->packed);
@@ -4520,7 +4518,6 @@ static void add_objects_in_unpacked_packs(void)
 		.source_infop = &source_info,
 	};
 
-	odb_prepare_alternates(to_pack.repo->objects);
 	for (source = to_pack.repo->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 
diff --git a/commit-graph.c b/commit-graph.c
index 49e8f63930..983c11ce85 100644
--- a/commit-graph.c
+++ b/commit-graph.c
@@ -651,8 +651,6 @@ struct commit_graph *load_commit_graph_chain_fd_st(struct object_database *odb,
 	count = st->st_size / (odb->repo->hash_algo->hexsz + 1);
 	CALLOC_ARRAY(oids, count);
 
-	odb_prepare_alternates(odb);
-
 	for (i = 0; i < count; i++) {
 		struct odb_source *source;
 
@@ -768,7 +766,6 @@ static struct commit_graph *prepare_commit_graph(struct repository *r)
 	if (!commit_graph_compatible(r))
 		return NULL;
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		r->objects->commit_graph = read_commit_graph_one(source);
 		if (r->objects->commit_graph)
@@ -2018,7 +2015,6 @@ static void fill_oids_from_all_packs(struct write_commit_graph_context *ctx)
 			_("Finding commits for commit graph among packed objects"),
 			ctx->approx_nr_objects);
 
-	odb_prepare_alternates(ctx->r->objects);
 	for (source = ctx->r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		odb_source_for_each_object(&files->packed->base, &oi, add_packed_commits_oi,
diff --git a/loose.c b/loose.c
index aa3cb1b4fc..c159d29d2d 100644
--- a/loose.c
+++ b/loose.c
@@ -115,7 +115,6 @@ int repo_read_loose_object_map(struct repository *repo)
 {
 	struct odb_source *source;
 
-	odb_prepare_alternates(repo->objects);
 	for (source = repo->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		if (loose_object_map_load(files->loose) < 0)
diff --git a/object-name.c b/object-name.c
index 83efba0ba6..34a08d76dd 100644
--- a/object-name.c
+++ b/object-name.c
@@ -280,7 +280,6 @@ static int init_object_disambiguation(struct repository *r,
 
 	ds->len = len;
 	ds->repo = r;
-	odb_prepare_alternates(r->objects);
 	return 0;
 }
 
diff --git a/odb.c b/odb.c
index 51da386f22..2ae8228dd2 100644
--- a/odb.c
+++ b/odb.c
@@ -237,11 +237,6 @@ void odb_add_to_alternates_file(struct object_database *odb,
 struct odb_source *odb_add_to_alternates_memory(struct object_database *odb,
 						const char *dir)
 {
-	/*
-	 * Make sure alternates are initialized, or else our entry may be
-	 * overwritten when they are.
-	 */
-	odb_prepare_alternates(odb);
 	return odb_add_alternate_recursively(odb, dir, 0);
 }
 
@@ -250,12 +245,6 @@ struct odb_source *odb_set_temporary_primary_source(struct object_database *odb,
 {
 	struct odb_source *source;
 
-	/*
-	 * Make sure alternates are initialized, or else our entry may be
-	 * overwritten when they are.
-	 */
-	odb_prepare_alternates(odb);
-
 	/*
 	 * Make a new primary odb and link the old primary ODB in as an
 	 * alternate
@@ -361,7 +350,6 @@ struct odb_source *odb_find_source(struct object_database *odb, const char *obj_
 	char *obj_dir_real = real_pathdup(obj_dir, 1);
 	struct strbuf odb_path_real = STRBUF_INIT;
 
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next) {
 		strbuf_realpath(&odb_path_real, source->path, 1);
 		if (!strcmp(obj_dir_real, odb_path_real.buf))
@@ -495,7 +483,6 @@ int odb_for_each_alternate(struct object_database *odb,
 	struct odb_source *alternate;
 	int r = 0;
 
-	odb_prepare_alternates(odb);
 	for (alternate = odb->sources->next; alternate; alternate = alternate->next) {
 		r = cb(alternate, payload);
 		if (r)
@@ -504,7 +491,7 @@ int odb_for_each_alternate(struct object_database *odb,
 	return r;
 }
 
-void odb_prepare_alternates(struct object_database *odb)
+static void odb_prepare_alternates(struct object_database *odb)
 {
 	struct strvec sources = STRVEC_INIT;
 
@@ -523,7 +510,6 @@ void odb_prepare_alternates(struct object_database *odb)
 
 int odb_has_alternates(struct object_database *odb)
 {
-	odb_prepare_alternates(odb);
 	return !!odb->sources->next;
 }
 
@@ -583,8 +569,6 @@ static int do_oid_object_info_extended(struct object_database *odb,
 	if (!odb_source_read_object_info(odb->inmemory_objects, oid, oi, flags))
 		return 0;
 
-	odb_prepare_alternates(odb);
-
 	while (1) {
 		struct odb_source *source;
 
@@ -847,7 +831,6 @@ int odb_freshen_object(struct object_database *odb,
 		       const struct object_id *oid)
 {
 	struct odb_source *source;
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next)
 		if (odb_source_freshen_object(source, oid, NULL))
 			return 1;
@@ -862,7 +845,6 @@ int odb_for_each_object_ext(struct object_database *odb,
 {
 	int ret;
 
-	odb_prepare_alternates(odb);
 	for (struct odb_source *source = odb->sources; source; source = source->next) {
 		if (opts->flags & ODB_FOR_EACH_OBJECT_LOCAL_ONLY && !source->local)
 			continue;
@@ -900,7 +882,6 @@ int odb_count_objects(struct object_database *odb,
 		return 0;
 	}
 
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next) {
 		unsigned long c;
 
@@ -980,7 +961,6 @@ int odb_find_abbrev_len(struct object_database *odb,
 		goto out;
 	}
 
-	odb_prepare_alternates(odb);
 	for (struct odb_source *source = odb->sources; source; source = source->next) {
 		ret = odb_source_find_abbrev_len(source, oid, len, &len);
 		if (ret)
@@ -1091,6 +1071,8 @@ struct object_database *odb_new(struct repository *repo,
 	o->alternate_db = secondary_sources;
 	o->inmemory_objects = &odb_source_inmemory_new(o)->base;
 
+	odb_prepare_alternates(o);
+
 	free(primary_source);
 	return o;
 }
@@ -1151,10 +1133,10 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
 	 */
 	if (flags & ODB_PREPARE_FLUSH_CACHES) {
 		o->loaded_alternates = 0;
+		odb_prepare_alternates(o);
 		o->object_count_valid = 0;
 	}
 
-	odb_prepare_alternates(o);
 	for (source = o->sources; source; source = source->next)
 		odb_source_prepare(source, flags);
 
diff --git a/odb.h b/odb.h
index 71af7450a9..fbafee174b 100644
--- a/odb.h
+++ b/odb.h
@@ -273,12 +273,6 @@ void odb_for_each_alternate_ref(struct object_database *odb,
 int odb_mkstemp(struct object_database *odb,
 		struct strbuf *temp_filename, const char *pattern);
 
-/*
- * Prepare alternate object sources for the given database by reading
- * "objects/info/alternates" and opening the respective sources.
- */
-void odb_prepare_alternates(struct object_database *odb);
-
 /*
  * Check whether the object database has any alternates. The primary object
  * source does not count as alternate.
diff --git a/odb/streaming.c b/odb/streaming.c
index 20531e864c..37642768e9 100644
--- a/odb/streaming.c
+++ b/odb/streaming.c
@@ -184,7 +184,6 @@ static int istream_source(struct odb_read_stream **out,
 {
 	struct odb_source *source;
 
-	odb_prepare_alternates(odb);
 	for (source = odb->sources; source; source = source->next)
 		if (!odb_source_read_object_stream(out, source, oid))
 			return 0;
diff --git a/pack-bitmap.c b/pack-bitmap.c
index e85bd69ba4..e0fb57d332 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -717,7 +717,6 @@ static int open_bitmap(struct repository *r,
 
 	assert(!bitmap_git->map);
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 
@@ -3417,7 +3416,6 @@ int verify_bitmap_files(struct repository *r)
 	struct packed_git *p;
 	int res = 0;
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		struct multi_pack_index *m = get_multi_pack_index(files->packed);
diff --git a/packfile.c b/packfile.c
index 0eee45055f..d870de90ed 100644
--- a/packfile.c
+++ b/packfile.c
@@ -1938,7 +1938,6 @@ int has_object_pack(struct repository *r, const struct object_id *oid)
 {
 	struct odb_source *source;
 
-	odb_prepare_alternates(r->objects);
 	for (source = r->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		if (!odb_source_read_object_info(&files->packed->base, oid, NULL, 0))
diff --git a/packfile.h b/packfile.h
index e1f77152b5..10de24f477 100644
--- a/packfile.h
+++ b/packfile.h
@@ -77,8 +77,6 @@ static inline struct repo_for_each_pack_data repo_for_eack_pack_data_init(struct
 {
 	struct repo_for_each_pack_data data = { 0 };
 
-	odb_prepare_alternates(repo->objects);
-
 	for (struct odb_source *source = repo->objects->sources; source; source = source->next) {
 		struct odb_source_files *files = odb_source_files_downcast(source);
 		struct packfile_list_entry *entry = packfile_store_get_packs(files->packed);

-- 
2.55.0.679.g6767b8d81c.dirty

[PATCH v2 3/4] odb: drop `loaded_alternates` field

[Patrick Steinhardt]

The `struct object_database::loaded_alternates` field tells us whether
or not alternates have been loaded already. This field was useful before
the preceding commit as we were indeed lazy-loading alternates. But now
that we started to eagerly load them we can assume them to be loaded
after `odb_new()`, and hence the field does not serve any purpose
anymore.

Remove it.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 odb.c | 9 +--------
 odb.h | 2 --
 2 files changed, 1 insertion(+), 10 deletions(-)

diff --git a/odb.c b/odb.c
index 2ae8228dd2..2eb37a2f44 100644
--- a/odb.c
+++ b/odb.c
@@ -230,8 +230,7 @@ void odb_add_to_alternates_file(struct object_database *odb,
 	int ret = odb_source_write_alternate(odb->sources, dir);
 	if (ret < 0)
 		die(NULL);
-	if (odb->loaded_alternates)
-		odb_add_alternate_recursively(odb, dir, 0);
+	odb_add_alternate_recursively(odb, dir, 0);
 }
 
 struct odb_source *odb_add_to_alternates_memory(struct object_database *odb,
@@ -495,16 +494,11 @@ static void odb_prepare_alternates(struct object_database *odb)
 {
 	struct strvec sources = STRVEC_INIT;
 
-	if (odb->loaded_alternates)
-		return;
-
 	parse_alternates(odb->alternate_db, PATH_SEP, NULL, &sources);
 	odb_source_read_alternates(odb->sources, &sources);
 	for (size_t i = 0; i < sources.nr; i++)
 		odb_add_alternate_recursively(odb, sources.v[i], 0);
 
-	odb->loaded_alternates = 1;
-
 	strvec_clear(&sources);
 }
 
@@ -1132,7 +1126,6 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
 	 * the lifetime of the process.
 	 */
 	if (flags & ODB_PREPARE_FLUSH_CACHES) {
-		o->loaded_alternates = 0;
 		odb_prepare_alternates(o);
 		o->object_count_valid = 0;
 	}
diff --git a/odb.h b/odb.h
index fbafee174b..aefb34213f 100644
--- a/odb.h
+++ b/odb.h
@@ -69,8 +69,6 @@ struct object_database {
 	 */
 	int source_paths_icase;
 
-	int loaded_alternates;
-
 	/*
 	 * A list of alternate object directories loaded from the environment;
 	 * this should not generally need to be accessed directly, but will

-- 
2.55.0.679.g6767b8d81c.dirty

[PATCH v2 4/4] odb: drop `alternates_db` field

[Patrick Steinhardt]

The `struct object_database::alternates_db` field tracks the value of
the "GIT_ALTERNATE_OBJECT_DIRECTORIES" environment variable and is
used in `odb_prepare_alternates()`. It's not necessary to store it as a
separate field anymore though, as we stopped lazy-loading alternates.
Consequently, we can simply pass it to `odb_prepare_alternates()` via
`odb_new()` now.

Do so and remove the field.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 odb.c | 17 +++++++++--------
 odb.h |  7 -------
 2 files changed, 9 insertions(+), 15 deletions(-)

diff --git a/odb.c b/odb.c
index 2eb37a2f44..0212eaa998 100644
--- a/odb.c
+++ b/odb.c
@@ -490,12 +490,14 @@ int odb_for_each_alternate(struct object_database *odb,
 	return r;
 }
 
-static void odb_prepare_alternates(struct object_database *odb)
+static void odb_prepare_alternates(struct object_database *odb,
+				   const char *alternate_db)
 {
 	struct strvec sources = STRVEC_INIT;
 
-	parse_alternates(odb->alternate_db, PATH_SEP, NULL, &sources);
+	parse_alternates(alternate_db, PATH_SEP, NULL, &sources);
 	odb_source_read_alternates(odb->sources, &sources);
+
 	for (size_t i = 0; i < sources.nr; i++)
 		odb_add_alternate_recursively(odb, sources.v[i], 0);
 
@@ -1062,11 +1064,11 @@ struct object_database *odb_new(struct repository *repo,
 
 	o->sources = odb_source_new(o, primary_source, true);
 	o->sources_tail = &o->sources->next;
-	o->alternate_db = secondary_sources;
 	o->inmemory_objects = &odb_source_inmemory_new(o)->base;
 
-	odb_prepare_alternates(o);
+	odb_prepare_alternates(o, secondary_sources);
 
+	free(secondary_sources);
 	free(primary_source);
 	return o;
 }
@@ -1100,8 +1102,6 @@ void odb_free(struct object_database *o)
 	if (!o)
 		return;
 
-	free(o->alternate_db);
-
 	oidmap_clear(&o->replace_map, 1);
 	pthread_mutex_destroy(&o->replace_mutex);
 
@@ -1123,10 +1123,11 @@ void odb_prepare(struct object_database *o, enum odb_prepare_flags flags)
 	 * Reprepare alt odbs, in case the alternates file was modified
 	 * during the course of this process. This only _adds_ odbs to
 	 * the linked list, so existing odbs will continue to exist for
-	 * the lifetime of the process.
+	 * the lifetime of the process. Consequently, we don't have to
+	 * reprocess GIT_ALTERNATE_OBJECT_DIRECTORIES here.
 	 */
 	if (flags & ODB_PREPARE_FLUSH_CACHES) {
-		odb_prepare_alternates(o);
+		odb_prepare_alternates(o, NULL);
 		o->object_count_valid = 0;
 	}
 
diff --git a/odb.h b/odb.h
index aefb34213f..748366a610 100644
--- a/odb.h
+++ b/odb.h
@@ -69,13 +69,6 @@ struct object_database {
 	 */
 	int source_paths_icase;
 
-	/*
-	 * A list of alternate object directories loaded from the environment;
-	 * this should not generally need to be accessed directly, but will
-	 * populate the "sources" list when odb_prepare_alternates() is run.
-	 */
-	char *alternate_db;
-
 	/*
 	 * Objects that should be substituted by other objects
 	 * (see git-replace(1)).

-- 
2.55.0.679.g6767b8d81c.dirty

Re: [PATCH v2 0/4] odb: eagerly load alternates

[Junio C Hamano]

Patrick Steinhardt <ps@pks.im> writes:

> The series is built on top of 010afd3166 (The 12th batch, 2026-08-07)
> with ps/odb-make-creation-pluggable at e927cfeb21 (odb: make creation of
> on-disk structures pluggable, 2026-08-07) merged into it.

It is not a clean merge, though.  Please double check the
synthesized base when I push the integration results out later
today.  d296c52baa (Merge branch 'ps/odb-make-creation-pluggable'
into ps/odb-eagerly-load-alternates, 2026-08-12) will be the merge,
unless I notice and fix a mismerge in it before I push it out.

Thanks.