bug#28966: git fails test suite

bug#28966: git fails test suite

[Ricardo Wurmus]

As of 62d48234e20c732ebefc1b4deb425f590b4241ca git fails at least one of
its tests on my servers:

--8<---------------cut here---------------start------------->8---
not ok 185 - --contains and --no-contains work in a deep repo
#
#               >expect &&
#               i=1 &&
#               while test $i -lt 8000
#               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 &&
#               run_with_limited_stack git tag --no-contains HEAD >actual &&
#               test_line_count "-gt" 10 actual
#
--8<---------------cut here---------------end--------------->8---

The server has 1.5TB RAM and 192 cores.  I’ve tried to build git about 5
times; some of these times I used “--cores=1” to force a sequential
build, but the tests would still fail.

-- Ricardo

bug#28966: git fails test suite

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 1684 bytes --]

Ricardo Wurmus <ricardo.wurmus@mdc-berlin.de> writes:

> As of 62d48234e20c732ebefc1b4deb425f590b4241ca git fails at least one of
> its tests on my servers:
>
> --8<---------------cut here---------------start------------->8---
> not ok 185 - --contains and --no-contains work in a deep repo
> #
> #               >expect &&
> #               i=1 &&
> #               while test $i -lt 8000
> #               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 &&
> #               run_with_limited_stack git tag --no-contains HEAD >actual &&
> #               test_line_count "-gt" 10 actual
> #
> --8<---------------cut here---------------end--------------->8---

"185" is very early in the test run, there could be similar cases
hiding as well.  Are you able to try and disable these tests?  If not,
I'll patch out at least this one in the next few days.

> The server has 1.5TB RAM and 192 cores.  I’ve tried to build git about 5
> times; some of these times I used “--cores=1” to force a sequential
> build, but the tests would still fail.

Odd that it fails sequentially as well.  It would be good to report this
upstream.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

bug#28966: git fails test suite

[Marius Bakke]

[-- Attachment #1.1: Type: text/plain, Size: 1788 bytes --]

Ricardo Wurmus <ricardo.wurmus@mdc-berlin.de> writes:

> As of 62d48234e20c732ebefc1b4deb425f590b4241ca git fails at least one of
> its tests on my servers:
>
> --8<---------------cut here---------------start------------->8---
> not ok 185 - --contains and --no-contains work in a deep repo
> #
> #               >expect &&
> #               i=1 &&
> #               while test $i -lt 8000
> #               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 &&
> #               run_with_limited_stack git tag --no-contains HEAD >actual &&
> #               test_line_count "-gt" 10 actual
> #
> --8<---------------cut here---------------end--------------->8---
>
> The server has 1.5TB RAM and 192 cores.  I’ve tried to build git about 5
> times; some of these times I used “--cores=1” to force a sequential
> build, but the tests would still fail.

I wonder if this has anything to do with the fact that this test (and
one other) limits the stack size to 128 kB (ulimit -s).  FWIW here is
the commit that added this test:

https://github.com/git/git/commit/cbc60b67201e083a4970c8731c5382a575357e36

It has since been increased from 64kB to 128kB, so I wonder if the
attached patch will make any difference.  Can you try it?


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: 0001-gnu-git-Increase-stack-size-for-tests.patch --]
[-- Type: text/x-patch, Size: 1373 bytes --]

From f95ae3d32e5eab9797cb1591cca07c4d31f187ab Mon Sep 17 00:00:00 2001
From: Marius Bakke <mbakke@fastmail.com>
Date: Wed, 1 Nov 2017 21:11:29 +0100
Subject: [PATCH] gnu: git: Increase stack size for tests.

* gnu/packages/version-control.scm (git)[arguments]: Add substitution in
'patch-tests' phase.
---
 gnu/packages/version-control.scm | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/gnu/packages/version-control.scm b/gnu/packages/version-control.scm
index af880c29f..9059284ed 100644
--- a/gnu/packages/version-control.scm
+++ b/gnu/packages/version-control.scm
@@ -216,6 +216,12 @@ as well as the classic centralized workflow.")
             ;; Many tests contain inline shell scripts (hooks etc).
             (substitute* (find-files "t" "\\.sh$")
               (("#!/bin/sh") (string-append "#!" (which "sh"))))
+            ;; XXX: Some tests run with a limited stack size, but it may
+            ;; be too small for very large machines (1TB+).  See
+            ;; <https://bugs.gnu.org/28966>.
+            (substitute* "t/test-lib.sh"
+              (("\\(ulimit -s 128")
+               "(ulimit -s 256"))
             ;; Un-do shebang patching here to prevent checksum mismatch.
             (substitute* '("t/t4034/perl/pre" "t/t4034/perl/post")
               (("^#!.*/bin/perl") "#!/usr/bin/perl"))
-- 
2.14.3


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

bug#28966: Hunting: git fails test suite

[zimoun]

Dear,

This bug is about failing tests of Git. Possible because the stack size.
It was opened against the commit 62d48234e2 (Git version 2.14.3).

The message [1] contains a patch. Does it still make sense?
Does the test suite is still failing with the current version of Git?
From [2], it seems passing.

Could we close this bug?
Or can we confirm that the issue comes from the stack size and the
ulimit patch should be enough to pass the test suite, right? Merge it?


[1] https://debbugs.gnu.org/cgi/bugreport.cgi?bug=28966#11
[2] http://data.guix.gnu.org/revision/75385105348066201ef898b934917eeb6ceab87a/package/git/2.25.0


Thank you in advance for any comments.


All the best,
simon

bug#28966: Hunting: git fails test suite

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 285 bytes --]

zimoun <zimon.toutoune@gmail.com> writes:

> This bug is about failing tests of Git. Possible because the stack size.
> It was opened against the commit 62d48234e2 (Git version 2.14.3).

Since we haven't heard back from Ricardo regarding this I'll assume the
issue is fixed.  Closing!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]