fsck: use oidset instead of oid_array for skipList
commit3b41fb0cb217f4b4491f2e67ce4183e5d2a5d873
authorRené Scharfe <l.s.r@web.de>
Mon, 3 Sep 2018 14:49:27 +0000 (3 14:49 +0000)
committerJunio C Hamano <gitster@pobox.com>
Wed, 12 Sep 2018 22:17:46 +0000 (12 15:17 -0700)
treeb229f85f43239388a68e21c09c556f7d71974d6a
parentfb8952077dfa11dc1534b5a9d965cbb39b043af4
fsck: use oidset instead of oid_array for skipList

Change the implementation of the skipList feature to use oidset
instead of oid_array to store SHA-1s for later lookup.

This list is parsed once on startup by fsck, fetch-pack or
receive-pack depending on the *.skipList config in use. I.e. only once
per invocation, but note that for "clone --recurse-submodules" each
submodule will re-parse the list, in addition to the main project, and
it will be re-parsed when checking .gitmodules blobs, see
fb16287719 ("fsck: check skiplist for object in fsck_blob()",
2018-06-27).

Memory usage is a bit higher, but we don't need to keep track of the
sort order anymore. Embed the oidset into struct fsck_options to make
its ownership clear (no hidden sharing) and avoid unnecessary pointer
indirection.

The cumulative impact on performance of this & the preceding change,
using the test setup described in the previous commit:

    Test                                             HEAD~2            HEAD~                   HEAD
    ----------------------------------------------------------------------------------------------------------------
    1450.3: fsck with 0 skipped bad commits          7.70(7.31+0.38)   7.72(7.33+0.38) +0.3%   7.70(7.30+0.40) +0.0%
    1450.5: fsck with 1 skipped bad commits          7.84(7.47+0.37)   7.69(7.32+0.36) -1.9%   7.71(7.29+0.41) -1.7%
    1450.7: fsck with 10 skipped bad commits         7.81(7.40+0.40)   7.94(7.57+0.36) +1.7%   7.92(7.55+0.37) +1.4%
    1450.9: fsck with 100 skipped bad commits        7.81(7.42+0.38)   7.95(7.53+0.41) +1.8%   7.83(7.42+0.41) +0.3%
    1450.11: fsck with 1000 skipped bad commits      7.99(7.62+0.36)   7.90(7.50+0.40) -1.1%   7.86(7.49+0.37) -1.6%
    1450.13: fsck with 10000 skipped bad commits     7.98(7.57+0.40)   7.94(7.53+0.40) -0.5%   7.90(7.45+0.44) -1.0%
    1450.15: fsck with 100000 skipped bad commits    7.97(7.57+0.39)   8.03(7.67+0.36) +0.8%   7.84(7.43+0.41) -1.6%
    1450.17: fsck with 1000000 skipped bad commits   7.72(7.22+0.50)   7.28(7.07+0.20) -5.7%   7.13(6.87+0.25) -7.6%

Helped-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/config.txt
fsck.c
fsck.h