[Python-Dev] Build failure in test_cmd_line on OSX-x86

[Stephen Hansen]

I'm sure this has to be my configuration somehow, but I'm getting a build
failure that I don't quite know how to debug, because I can't reproduce it
when I run the test manually. Any advice would be appreciated-- I'm a
buildslave newbie :-)

I'm referring to
http://www.python.org/dev/buildbot/builders/x86%20Snow%20Leopard%203.x/builds/13/steps/test/logs/stdio

Now, when I saw this the first thing I assumed I should do is to try to
reproduce it, so I did:

  Top-2:~ pythonbuildbot$ cp -R buildarea/3.x.hansen-osx-x86/ ~/test
  Top-2:~ pythonbuildbot$ cd ~/test/build/
  Top-2:build pythonbuildbot$ ./configure --with-pydebug
--with-computed-gotos
  [snip]
  Top-2:build pythonbuildbot$ make all
  [snip]

I made sure to go check out the stdio for configure/compile so the compiling
I'd do would be with the same options as the buildslave did. Then:

  Top-2:build pythonbuildbot$ ./python.exe -m test.regrtest -uall
test_cmd_line.py
  [1/1] test_cmd_line
  1 test OK.
  [84022 refs]

Doh. So, next I try running the whole test suite:

  Top-2:build pythonbuildbot$ ./python.exe -m test.regrtest -uall
  [snip]
  [ 44/348] test_cmd_line

And it passes too: but I notice its run way earlier in the test process then
it did on the buildbot, and I remember reading awhile ago about a test
failure that happened only when a certain test ran in a certain order. So, I
go garb the randseed from the buildbot run and use it:

  Top-2:build pythonbuildbot$ ./python.exe -m test.regrtest -uall -r
--randseed=9634655
  == CPython 3.2a2+ (py3k:85321, Oct 8 2010, 08:54:05) [GCC 4.2.1 (Apple
Inc. build 5664)]
  ==   Darwin-10.4.0-i386-64bit little-endian
  ==   /Users/pythonbuildbot/test/build/build/test_python_86644
  Using random seed 9634655
  [snip]

And long story short, it gets to 201 and runs test_cmd_line in the same
order as the buildbot did, and it succeeds too, and I curse the gods of the
netherworld, and am stumped with how to proceed. Two separate buildbot runs
and this same failure happened, yet for me, no. Or I'm doing something
differently then the buildbot is, and I can't see what.

--Stephen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-dev/attachments/20101008/0780dbd1/attachment.html>