#760844 [ppc64el] git FTBFS: t7004.136 fails intermittently

Package:
git
Source:
git
Description:
fast, scalable, distributed revision control system
Submitter:
Andreas Barth
Date:
2014-09-09 14:27:11 UTC
Severity:
important
#760844#5
Date:
2014-09-08 12:31:26 UTC
From:
To:
Hi,

git FTBFS on ppc64l due to the following testsuite failure.

not ok 136 - --contains works in a deep repo
#
#               >expect &&
#               i=1 &&
#               while test $i -lt 4000
#               do
#                       echo "commit refs/heads/master
#       committer A U Thor <author@example.com> $((1000000000 + $i * 100)) +0200
#       data <<EOF
#       commit #$i
#       EOF"
#                       test $i = 1 && echo "from refs/heads/master^0"
#                       i=$(($i + 1))
#               done | git fast-import &&
#               git checkout master &&
#               git tag far-far-away HEAD^ &&
#               run_with_limited_stack git tag --contains HEAD >actual &&
#               test_cmp expect actual
#
# failed 1 among 136 test(s)
1..136
make[3]: *** [t7004-tag.sh] Error 1
Makefile:44: recipe for target 't7004-tag.sh' failed



Andi

#760844#10
Date:
2014-09-08 19:31:20 UTC
From:
To:
Hi Andi,

Thanks for spotting this one.

Although that failure is indeed in the build log, the build finishes
successfully [0].  Do you think it's a more serious problem?

The packages was built and installed in the archive some time ago [1],
and the git binary is in the binary package:

	git_2.1.0-1_ppc64el.deb
	[...]
	-rwxr-xr-x root/root   1895560 2014-09-08 20:45 ./usr/lib/git-core/git
	[...]

Kind regards,


[0]
https://buildd.debian.org/status/fetch.php?pkg=git&arch=ppc64el&ver=1%3A2.1.0-1&stamp=1410201979
[1] https://buildd.debian.org/status/package.php?a=ppc64el&suite=sid&p=git

#760844#15
Date:
2014-09-08 19:54:20 UTC
From:
To:
retitle 760844 [ppc64el] git FTBFS: t7004.136 fails intermittently
tags 760844 + upstream
# FTBFS on non-release arch
severity 757402 important
tags 757402 + upstream
quit

Hi Mauricio,

Mauricio Faria de Oliveira wrote:

Please keep in mind that these appear as emails in a crowded inbox,
where the subject line can provide useful context.

It looks like the same bug as #757402, which is a bug in the testsuite
(not in git itself).  What's odd is that ppc64el the corresponding
problem doesn't seem to be 100% reproducible --- e.g. in the log you
mentioned, the packaging runs tests twice, and they failed the first
time and passed the second time.

Based on your knowledge of the arch, does that make sense?  Would
limiting a process to a 64k stack with "ulimit -s 64" be expected to
make it sometimes fail and sometimes succeed?

Curious,
Jonathan

#760844#24
Date:
2014-09-09 14:21:48 UTC
From:
To:
Hi Jonathan,

Apologies. I believe I provided such great example of a totally context-
free / useless subject line. :-)  Thanks for the respectful reminder.

I'll have to take a deeper look at that soon. Although the ELFv2 ABI
(introduced in/for ppc64el) has differences/reductions in stack frame
[1], that should be set/assembly written at compile-time - so it isn't
expected to vary at run-time.  (we should expect consistency there :)

I'm not familiar with the git/testsuite, so I may be missing something.
Maybe it's something in it or the environment. So I'll set-up some time
to go through it and analyze this issue soon.

Thanks for the pointers and context.

[1] https://sourceware.org/ml/binutils/2013-10/msg00371.html