]> git.ipfire.org Git - thirdparty/git.git/commit
object-file: fix a unpack_loose_header() regression in 3b6a8db3b03
authorÆvar Arnfjörð Bjarmason <avarab@gmail.com>
Thu, 12 May 2022 22:32:18 +0000 (15:32 -0700)
committerJunio C Hamano <gitster@pobox.com>
Thu, 12 May 2022 22:42:26 +0000 (15:42 -0700)
commit4627c67fa68d5669be511962a6437a11c0db3c99
treee770420ab4ae42fe090c42d1cd32ac65e8b13075
parent29d8e21d6e97d5363233a960e7730e0afc26a8b1
object-file: fix a unpack_loose_header() regression in 3b6a8db3b03

Fix a regression in my 3b6a8db3b03 (object-file.c: use "enum" return
type for unpack_loose_header(), 2021-10-01) revealed both by running
the test suite with --valgrind, and with the amended "git fsck" test.

In practice this regression in v2.34.0 caused us to claim that we
couldn't parse the header, as opposed to not being able to unpack
it. Before the change in the C code the test_cmp added here would emit:

-error: unable to unpack header of ./objects/e6/9de29bb2d1d6434b8b29ae775ad8c2e48c5391
+error: unable to parse header of ./objects/e6/9de29bb2d1d6434b8b29ae775ad8c2e48c5391

I.e. we'd proceed to call parse_loose_header() on the uninitialized
"hdr" value, and it would have been very unlikely for that
uninitialized memory to be a valid git object.

The other callers of unpack_loose_header() were already checking the
enum values exhaustively. See 3b6a8db3b03 and
5848fb11acd (object-file.c: return ULHR_TOO_LONG on "header too long",
2021-10-01).

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
object-file.c
t/t1006-cat-file.sh
t/t1450-fsck.sh