git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jeff King <peff@peff.net>
To: Junio C Hamano <gitster@pobox.com>
Cc: David Kastrup <dak@gnu.org>, git@vger.kernel.org
Subject: [PATCH] branch: clean up commit flags after merge-filter walk
Date: Thu, 18 Sep 2014 06:49:43 -0400	[thread overview]
Message-ID: <20140918104943.GA13481@peff.net> (raw)
In-Reply-To: <xmqqwq944ov0.fsf@gitster.dls.corp.google.com>

On Mon, Sep 15, 2014 at 01:20:51PM -0700, Junio C Hamano wrote:

> David Kastrup <dak@gnu.org> writes:
> 
> > This gives the same result as
> >
> > git branch --verbose --merged
> >
> > namely _only_ listing the current branch verbosely.
> 
> Hmph.  Then that is a different issue.  As I never use --merged
> myself, even though I use "git branch [--verbose] --no-merged pu"
> quite often to check the list of topics that I received but not yet
> merged anywhere in my tree, I wouldn't be entirely surprised that
> the combination does not work, but with a quick glance of the code,
> especially how matches_merge_filter() uses "--[no-]merged", I cannot
> offhand imagine how --no-merged would work and --merged break.

Ugh. The bug is caused by a failure to clean up the global flag state. I
am really tempted to try converting the revision walker to keep its
flags in a commit_slab so that this sort of thing can never bite us
again.

But here is the minimal fix.

-- >8 --
Subject: branch: clean up commit flags after merge-filter walk

When we run `branch --merged`, we use prepare_revision_walk
with the merge-filter marked as UNINTERESTING. Any branch
tips that are marked UNINTERESTING after it returns must be
ancestors of that commit. As we iterate through the list of
refs to show, we check item->commit->object.flags to see
whether it was marked.

This interacts badly with --verbose, which will do a
separate walk to find the ahead/behind information for each
branch. There are two bad things that can happen:

  1. The ahead/behind walk may get the wrong results,
     because it can see a bogus UNINTERESTING flag leftover
     from the merge-filter walk.

  2. We may omit some branches if their tips are involved in
     the ahead/behind traversal of a branch shown earlier.
     The ahead/behind walk carefully cleans up its commit
     flags, meaning it may also erase the UNINTERESTING
     flag that we expect to check later.

We can solve this by moving the merge-filter state for each
ref into its "struct ref_item" as soon as we finish the
merge-filter walk. That fixes (2). Then we are free to clear
the commit flags we used in the walk, fixing (1).

Note that we actually do away with the matches_merge_filter
helper entirely here, and inline it between the revision
walk and the flag-clearing. This ensures that nobody
accidentally calls it at the wrong time (it is only safe to
check in that instant between the setting and clearing of
the global flag).

Signed-off-by: Jeff King <peff@peff.net>
---
 builtin/branch.c           | 33 +++++++++++++++++++--------------
 t/t3201-branch-contains.sh | 29 +++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 14 deletions(-)

diff --git a/builtin/branch.c b/builtin/branch.c
index ced422b..9e4666f 100644
--- a/builtin/branch.c
+++ b/builtin/branch.c
@@ -280,6 +280,7 @@ struct ref_item {
 	char *dest;
 	unsigned int kind, width;
 	struct commit *commit;
+	int ignore;
 };
 
 struct ref_list {
@@ -385,6 +386,7 @@ static int append_ref(const char *refname, const unsigned char *sha1, int flags,
 	newitem->commit = commit;
 	newitem->width = utf8_strwidth(refname);
 	newitem->dest = resolve_symref(orig_refname, prefix);
+	newitem->ignore = 0;
 	/* adjust for "remotes/" */
 	if (newitem->kind == REF_REMOTE_BRANCH &&
 	    ref_list->kinds != REF_REMOTE_BRANCH)
@@ -484,17 +486,6 @@ static void fill_tracking_info(struct strbuf *stat, const char *branch_name,
 	free(ref);
 }
 
-static int matches_merge_filter(struct commit *commit)
-{
-	int is_merged;
-
-	if (merge_filter == NO_FILTER)
-		return 1;
-
-	is_merged = !!(commit->object.flags & UNINTERESTING);
-	return (is_merged == (merge_filter == SHOW_MERGED));
-}
-
 static void add_verbose_info(struct strbuf *out, struct ref_item *item,
 			     int verbose, int abbrev)
 {
@@ -522,10 +513,9 @@ static void print_ref_item(struct ref_item *item, int maxwidth, int verbose,
 {
 	char c;
 	int color;
-	struct commit *commit = item->commit;
 	struct strbuf out = STRBUF_INIT, name = STRBUF_INIT;
 
-	if (!matches_merge_filter(commit))
+	if (item->ignore)
 		return;
 
 	switch (item->kind) {
@@ -575,7 +565,7 @@ static int calc_maxwidth(struct ref_list *refs)
 {
 	int i, w = 0;
 	for (i = 0; i < refs->index; i++) {
-		if (!matches_merge_filter(refs->list[i].commit))
+		if (refs->list[i].ignore)
 			continue;
 		if (refs->list[i].width > w)
 			w = refs->list[i].width;
@@ -618,6 +608,7 @@ static void show_detached(struct ref_list *ref_list)
 		item.kind = REF_LOCAL_BRANCH;
 		item.dest = NULL;
 		item.commit = head_commit;
+		item.ignore = 0;
 		if (item.width > ref_list->maxwidth)
 			ref_list->maxwidth = item.width;
 		print_ref_item(&item, ref_list->maxwidth, ref_list->verbose, ref_list->abbrev, 1, "");
@@ -656,6 +647,20 @@ static int print_ref_list(int kinds, int detached, int verbose, int abbrev, stru
 
 		if (prepare_revision_walk(&ref_list.revs))
 			die(_("revision walk setup failed"));
+
+		for (i = 0; i < ref_list.index; i++) {
+			struct ref_item *item = &ref_list.list[i];
+			struct commit *commit = item->commit;
+			int is_merged = !!(commit->object.flags & UNINTERESTING);
+			item->ignore = is_merged != (merge_filter == SHOW_MERGED);
+		}
+
+		for (i = 0; i < ref_list.index; i++) {
+			struct ref_item *item = &ref_list.list[i];
+			clear_commit_marks(item->commit, ALL_REV_FLAGS);
+		}
+		clear_commit_marks(filter, ALL_REV_FLAGS);
+
 		if (verbose)
 			ref_list.maxwidth = calc_maxwidth(&ref_list);
 	}
diff --git a/t/t3201-branch-contains.sh b/t/t3201-branch-contains.sh
index 141b061..912a663 100755
--- a/t/t3201-branch-contains.sh
+++ b/t/t3201-branch-contains.sh
@@ -130,4 +130,33 @@ test_expect_success 'implicit --list conflicts with modification options' '
 
 '
 
+# We want to set up a case where the walk for the tracking info
+# of one branch crosses the tip of another branch (and make sure
+# that the latter walk does not mess up our flag to see if it was
+# merged).
+#
+# Here "topic" tracks "master" with one extra commit, and "zzz" points to the
+# same tip as master The name "zzz" must come alphabetically after "topic"
+# as we process them in that order.
+test_expect_success 'branch --merged with --verbose' '
+	git branch --track topic master &&
+	git branch zzz topic &&
+	git checkout topic &&
+	test_commit foo &&
+	git branch --merged topic >actual &&
+	cat >expect <<-\EOF &&
+	  master
+	* topic
+	  zzz
+	EOF
+	test_cmp expect actual &&
+	git branch --verbose --merged topic >actual &&
+	cat >expect <<-\EOF &&
+	  master c77a0a9 second on master
+	* topic  2c939f4 [ahead 1] foo
+	  zzz    c77a0a9 second on master
+	EOF
+	test_cmp expect actual
+'
+
 test_done
-- 
2.1.0.486.g83bb229

      reply	other threads:[~2014-09-18 10:49 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-09-14  7:52 git branch --merged and git branch --verbose do not combine David Kastrup
2014-09-15 18:06 ` Junio C Hamano
2014-09-15 19:07   ` David Kastrup
2014-09-15 20:20     ` Junio C Hamano
2014-09-18 10:49       ` Jeff King [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20140918104943.GA13481@peff.net \
    --to=peff@peff.net \
    --cc=dak@gnu.org \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).