Failing test_commit_iteration

[yarikoptic]

First failed while building 1.0.1 so decided to check current snapshot -- the same story:

hopa:~deb/build-area/python-git-1.0.1+git72-g039e265
$> GIT_PYTHON_TEST_GIT_REPO_BASE=$PWD/build/test/GitPython nosetests -s -v --pdb git/test/performance/test_commit.py 
test_commit_iteration (git.test.performance.test_commit.TestPerformance) ... Iterated 1250 Commits in 0.108746051788 [s] ( 11494.670192 commits/s )
ok
test_commit_serialization (git.test.performance.test_commit.TestPerformance) ... > /home/yoh/deb/gits/build-area/python-git-1.0.1+git72-g039e265/git/test/test_commit.py(57)assert_commit_serialization()
-> assert istream.hexsha == cm.hexsha.encode('ascii')
(Pdb) print istream.hexsha
66aa07976c27f9de05fd2da30f09627f365154ef
(Pdb) cm.hexsha.encode('ascii')
'e3068025b64bee24efc1063aba5138708737c158'
(Pdb) l
 52             ns += 1
 53             streamlen = stream.tell()
 54             stream.seek(0)
 55     
 56             istream = rwrepo.odb.store(IStream(Commit.type, streamlen, stream))
 57  ->         assert istream.hexsha == cm.hexsha.encode('ascii')
 58     
 59             nc = Commit(rwrepo, Commit.NULL_BIN_SHA, cm.tree,
 60                         cm.author, cm.authored_date, cm.author_tz_offset,
 61                         cm.committer, cm.committed_date, cm.committer_tz_offset,
 62                         cm.message, cm.parents, cm.encoding)

$> locale
LANG=en_US.UTF-8
LANGUAGE=
LC_CTYPE="en_US.UTF-8"
LC_NUMERIC="en_US.UTF-8"
LC_TIME="en_US.UTF-8"
LC_COLLATE="en_US.UTF-8"
LC_MONETARY="en_US.UTF-8"
LC_MESSAGES="en_US.UTF-8"
LC_PAPER="en_US.UTF-8"
LC_NAME="en_US.UTF-8"
LC_ADDRESS="en_US.UTF-8"
LC_TELEPHONE="en_US.UTF-8"
LC_MEASUREMENT="en_US.UTF-8"
LC_IDENTIFICATION="en_US.UTF-8"
LC_ALL=

$> apt-cache policy python-gitdb
python-gitdb:
  Installed: 0.6.4-1
  Candidate: 0.6.4-1
  Version table:
 *** 0.6.4-1 0
        300 http://http.debian.net/debian/ experimental/main amd64 Packages
        100 /var/lib/dpkg/status

[yarikoptic]

so any ideas or instructions on what/how to test to provide more information to pin point what is going on?

[Byron]

Sorry for the late reply, lot's of stuff is happening around here.

It seems I can nicely reproduce the issue, and must say that I absolutely don't like it. Right now I have no clue why it occours, but a bisect would certainly help in that regard.

No sooner said than done ! The commit 58c78e6 is the one introducing the issue - probably the commit message is now a different one.
It managed to sneak by CI as it doesn't run performance tests - it would take too long there.

Now a fix is most certainly possible, let's see if I will manage it tonight.

[Byron]

I found the issue - it's actually a critical flaw, which is now, finally fixed. This is a zero-day bug I suppose, and if not it's very old at least.

I'd be happy if you could verify the fix on your end and close the issue if it works for you.
Thanks

[yarikoptic]

Thank you @Byron -- this issue seems to be gone now! But I will file a fresh one in a sec ;)

[yarikoptic]

FWIW -- feedback was provided