Web lists-archives.com

[PATCH 06/20] blame: fix a bug, core.abbrev should work like --abbrev

Since 84393bfd73 ("blame: add --abbrev command line option and make it
honor core.abbrev", 2011-04-06) first released with v1.7.6 "git blame"
has supported both core.abbrev and --abbrev to change the abbreviation

Initially output wouldn't alter the abbreviation length to account for
the boundary commit. This was changed in 91229834c2 ("blame: fix
alignment with --abbrev=40", 2017-01-05) first released with v2.11.1.

That change had a bug which I'm fixing here. It didn't account for the
abbreviation length also being set via core.abbrev, not just via the
--abbrev command-line option.

So let's handle that. The easiest way to do that is to check if the
global default_abbrev variable (-1 by default) has been set by
git_default_core_config(), and if so pretend we had the --abbrev
option is set, in lieu of making everything that uses the "abbrev"
variable now read that OR default_abbrev).

The reason I'm documenting these past behaviors is that whatever our
desires with --porcelain people do parse the human-readable "git
blame" output, and for any machine use are likely to use
core.abbrev=40 or --abbrev=40. Documenting how the format has changed
over time will help those users avoid nasty surprises.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx>
 Documentation/git-blame.txt | 8 +++++++-
 builtin/blame.c             | 2 ++
 t/t0014-abbrev.sh           | 7 ++++++-
 3 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/Documentation/git-blame.txt b/Documentation/git-blame.txt
index 7b562494ac..d6cddbcb2e 100644
--- a/Documentation/git-blame.txt
+++ b/Documentation/git-blame.txt
@@ -92,7 +92,13 @@ include::blame-options.txt[]
 Because of this UI design, the only way to get the full SHA-1 of the
 boundary commit is to use the `--porcelain` format. With `--abbrev=40`
 only 39 characters of the boundary SHA-1 will be emitted, since one
-will be used for the caret to mark the boundary.
+will be used for the caret to mark the boundary. This behavior was
+different before 2.11.1, git would then emit the 40 character if
+requested, resulting in unaligned output.
+Before 2.19, setting `core.abbrev=40` wouldn't apply the above rule
+and would cause blame to emit output that was unaligned. This bug has
+since been fixed.
diff --git a/builtin/blame.c b/builtin/blame.c
index 4202584f97..6ab08561d1 100644
--- a/builtin/blame.c
+++ b/builtin/blame.c
@@ -868,6 +868,8 @@ int cmd_blame(int argc, const char **argv, const char *prefix)
 	} else if (show_progress < 0)
 		show_progress = isatty(2);
+	if (default_abbrev >= 0)
+		abbrev = default_abbrev;
 	if (0 < abbrev && abbrev < GIT_SHA1_HEXSZ)
 		/* one more abbrev length is needed for the boundary commit */
diff --git a/t/t0014-abbrev.sh b/t/t0014-abbrev.sh
index 77f15d5b0b..934c54a96b 100755
--- a/t/t0014-abbrev.sh
+++ b/t/t0014-abbrev.sh
@@ -135,7 +135,12 @@ do
 	test_expect_success "blame core.abbrev=$i and --abbrev=$i with boundary" "
 		# See the blame documentation for why this is off-by-one
 		git -c core.abbrev=$i blame A.t | cut_tr_d_n_field_n 1 | nocaret >blame &&
-		test_byte_count = $i blame &&
+		if test $i -eq 40
+		then
+			test_byte_count = 39 blame
+		else
+			test_byte_count = $i blame
+		fi &&
 		git blame --abbrev=$i A.t | cut_tr_d_n_field_n 1 | nocaret >blame &&
 		if test $i -eq 40