Skip to contentSkip to conversation

git bisect fails to handle annotated tags

PATCH In review Andreas Schwab 4 participants 5 messages Mar 16, 2021 – Mar 17, 2021

lore
Share & tools
lore
#0 lore raw
$ git --version
git version 2.31.0
$ git bisect start
$ git bisect good v2.30.0
$ git bisect bad v2.31.0
3e90d4b58f3819cfd58ac61cb8668e83d3ea0563 was both good and bad
$ git bisect log
git bisect start
# good: [71ca53e8125e36efbda17293c50027d31681a41f] Git 2.30
git bisect good 2d9685d47a7e516281aa093bf0cddc8aafa72448
# bad: [a5828ae6b52137b913b978e16cd2334482eb4c1f] Git 2.31
git bisect bad 3e90d4b58f3819cfd58ac61cb8668e83d3ea0563
$ git bisect start
Already on 'master'
Your branch is up to date with 'origin/master'.
$ git bisect good v2.30.0^{}
$ git bisect bad v2.31.0^{}
Bisecting: 453 revisions left to test after this (roughly 9 steps)
[41abfe15d95ede4c2a047180a6062eac23d8f7d6] maintenance: add pack-refs task
$ git bisect log
git bisect start
# good: [71ca53e8125e36efbda17293c50027d31681a41f] Git 2.30
git bisect good 71ca53e8125e36efbda17293c50027d31681a41f
# bad: [a5828ae6b52137b913b978e16cd2334482eb4c1f] Git 2.31
git bisect bad a5828ae6b52137b913b978e16cd2334482eb4c1f

Andreas.

Signature · 3 lines
Andreas Schwab, [email protected]
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."
#1 lore raw
Quoted 6 lines · Andreas Schwab
$ git --version
git version 2.31.0
$ git bisect start
$ git bisect good v2.30.0
$ git bisect bad v2.31.0
3e90d4b58f3819cfd58ac61cb8668e83d3ea0563 was both good and bad

Looks like it bisects to 27257bc466 (bisect--helper: reimplement bisect_state & bisect_head shell functions in C, 2020-10-15), which isn't too surprising. So it broke in v2.30, but nobody seems to have noticed during the last cycle.

I'd guess it's just missing a call to peel the input oid.

-Peff

#2 lore raw

PATCH 1/1 bisect: peel annotated tags to commits

Quoted 9 lines · Jeff King
Quoted 6 lines · Andreas Schwab
$ git --version
git version 2.31.0
$ git bisect start
$ git bisect good v2.30.0
$ git bisect bad v2.31.0
3e90d4b58f3819cfd58ac61cb8668e83d3ea0563 was both good and bad
Quoted 6 lines

Looks like it bisects to 27257bc466 (bisect--helper: reimplement bisect_state & bisect_head shell functions in C, 2020-10-15), which isn't too surprising. So it broke in v2.30, but nobody seems to have noticed during the last cycle.

I'd guess it's just missing a call to peel the input oid.

Yep. Here's a fix. Again, not new in v2.31, so we don't have to worry about a brown-bag fix for yesterday's release. But I do think it's worth trying to get onto a maint release. I prepared this patch on top of mr/bisect-in-c-3.

-- >8 --
Subject: [PATCH] bisect: peel annotated tags to commits

This patch fixes a bug where git-bisect doesn't handle receiving annotated tags as "git bisect good <tag>", etc. It's a regression in 27257bc466 (bisect--helper: reimplement bisect_state & bisect_head shell functions in C, 2020-10-15).

The original shell code called:

  sha=$(git rev-parse --verify "$rev^{commit}") ||
          die "$(eval_gettext "Bad rev input: \$rev")"

which will peel the input to a commit (or complain if that's not possible). But the C code just calls get_oid(), which will yield the oid of the tag.

The fix is to peel to a commit. The error message here is a little non-idiomatic for Git (since it starts with a capital). I've mostly left it, as it matches the other converted messages (like the "Bad rev input" we print when get_oid() fails), though I did add an indication that it was the peeling that was the problem. It might be worth taking a pass through this converted code to modernize some of the error messages.

Note also that the test does a bare "grep" (not i18ngrep) on the expected "X is the first bad commit" output message. This matches the rest of the test script.

---
 builtin/bisect--helper.c    |  9 ++++++++-
 t/t6030-bisect-porcelain.sh | 12 ++++++++++++
 2 files changed, 20 insertions(+), 1 deletion(-)
builtin/bisect--helper.c +8 -1
index fc6ca257a4..f0eeb4a2f0 100644
@@ -876,12 +876,19 @@ static enum bisect_error bisect_state(struct bisect_terms *terms, const char **a
 	 */
 
 	for (; argc; argc--, argv++) {
+		struct commit *commit;
+
 		if (get_oid(*argv, &oid)){
 			error(_("Bad rev input: %s"), *argv);
 			oid_array_clear(&revs);
 			return BISECT_FAILED;
 		}
-		oid_array_append(&revs, &oid);
+
+		commit = lookup_commit_reference(the_repository, &oid);
+		if (!commit)
+			die(_("Bad rev input (not a commit): %s"), *argv);
+
+		oid_array_append(&revs, &commit->object.oid);
 	}
 
 	if (strbuf_read_file(&buf, git_path_bisect_expected_rev(), 0) < the_hash_algo->hexsz ||
t/t6030-bisect-porcelain.sh +12 -0
index b886529e59..9c389553a7 100755
@@ -929,4 +929,16 @@ test_expect_success 'git bisect reset cleans bisection state properly' '
 	test_path_is_missing "$GIT_DIR/BISECT_START"
 '
 
+test_expect_success 'bisect handles annotated tags' '
+	test_commit commit-one &&
+	git tag -m foo tag-one &&
+	test_commit commit-two &&
+	git tag -m foo tag-two &&
+	git bisect start &&
+	git bisect good tag-one &&
+	git bisect bad tag-two >output &&
+	bad=$(git rev-parse --verify tag-two^{commit}) &&
+	grep "$bad is the first bad commit" output
+'
+
 test_done
Signature · 1 line
2.31.0.559.g509d4a088b
Tested-by on v1 1/1
#3 lore raw

Re: [PATCH] bisect: peel annotated tags to commits

I can reproduce this issue with v2.31.0 (as you mentioned).

Applying the patch, bisecting between annotated tags now worked just before git bisect is rewritten in C.

Thanks.

Quoted 104 lines · Jeff King
Quoted 9 lines · Jeff King
Quoted 6 lines · Andreas Schwab
$ git --version
git version 2.31.0
$ git bisect start
$ git bisect good v2.30.0
$ git bisect bad v2.31.0
3e90d4b58f3819cfd58ac61cb8668e83d3ea0563 was both good and bad
Quoted 6 lines

Looks like it bisects to 27257bc466 (bisect--helper: reimplement bisect_state & bisect_head shell functions in C, 2020-10-15), which isn't too surprising. So it broke in v2.30, but nobody seems to have noticed during the last cycle.

I'd guess it's just missing a call to peel the input oid.

Yep. Here's a fix. Again, not new in v2.31, so we don't have to worry about a brown-bag fix for yesterday's release. But I do think it's worth trying to get onto a maint release. I prepared this patch on top of mr/bisect-in-c-3.

-- >8 --
Subject: [PATCH] bisect: peel annotated tags to commits

This patch fixes a bug where git-bisect doesn't handle receiving annotated tags as "git bisect good <tag>", etc. It's a regression in 27257bc466 (bisect--helper: reimplement bisect_state & bisect_head shell functions in C, 2020-10-15).

The original shell code called:

   sha=$(git rev-parse --verify "$rev^{commit}") ||
           die "$(eval_gettext "Bad rev input: \$rev")"

which will peel the input to a commit (or complain if that's not possible). But the C code just calls get_oid(), which will yield the oid of the tag.

The fix is to peel to a commit. The error message here is a little non-idiomatic for Git (since it starts with a capital). I've mostly left it, as it matches the other converted messages (like the "Bad rev input" we print when get_oid() fails), though I did add an indication that it was the peeling that was the problem. It might be worth taking a pass through this converted code to modernize some of the error messages.

Note also that the test does a bare "grep" (not i18ngrep) on the expected "X is the first bad commit" output message. This matches the rest of the test script.

---
  builtin/bisect--helper.c    |  9 ++++++++-
  t/t6030-bisect-porcelain.sh | 12 ++++++++++++
  2 files changed, 20 insertions(+), 1 deletion(-)
builtin/bisect--helper.c +8 -1
index fc6ca257a4..f0eeb4a2f0 100644
@@ -876,12 +876,19 @@ static enum bisect_error bisect_state(struct bisect_terms *terms, const char **a
  	 */
 
  	for (; argc; argc--, argv++) {
+		struct commit *commit;
+
  		if (get_oid(*argv, &oid)){
  			error(_("Bad rev input: %s"), *argv);
  			oid_array_clear(&revs);
  			return BISECT_FAILED;
  		}
-		oid_array_append(&revs, &oid);
+
+		commit = lookup_commit_reference(the_repository, &oid);
+		if (!commit)
+			die(_("Bad rev input (not a commit): %s"), *argv);
+
+		oid_array_append(&revs, &commit->object.oid);
  	}
 
  	if (strbuf_read_file(&buf, git_path_bisect_expected_rev(), 0) < the_hash_algo->hexsz ||
t/t6030-bisect-porcelain.sh +12 -0
index b886529e59..9c389553a7 100755
@@ -929,4 +929,16 @@ test_expect_success 'git bisect reset cleans bisection state properly' '
  	test_path_is_missing "$GIT_DIR/BISECT_START"
  '
 
+test_expect_success 'bisect handles annotated tags' '
+	test_commit commit-one &&
+	git tag -m foo tag-one &&
+	test_commit commit-two &&
+	git tag -m foo tag-two &&
+	git bisect start &&
+	git bisect good tag-one &&
+	git bisect bad tag-two >output &&
+	bad=$(git rev-parse --verify tag-two^{commit}) &&
+	grep "$bad is the first bad commit" output
+'
+
  test_done
Signature · 1 line
An old man doll... just what I always wanted! - Clara
#4 lore raw

Re: [PATCH] bisect: peel annotated tags to commits

In reply to Jeff King
Quoted 11 lines · Jeff King
Quoted 6 lines

Looks like it bisects to 27257bc466 (bisect--helper: reimplement bisect_state & bisect_head shell functions in C, 2020-10-15), which isn't too surprising. So it broke in v2.30, but nobody seems to have noticed during the last cycle.

I'd guess it's just missing a call to peel the input oid.

Yep. Here's a fix. Again, not new in v2.31, so we don't have to worry about a brown-bag fix for yesterday's release. But I do think it's worth trying to get onto a maint release. I prepared this patch on top of mr/bisect-in-c-3.

Thanks. Yes, if we ever do another update to 2.30.x, this probably should be in it, as it is expected that people will start with tags and not with individual commits.