932373 - Windows checktest perma-orange on b2g26

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Ryan VanderMeulen [:RyanVM]]

https://tbpl-dev.allizom.org/php/getParsedLog.php?id=29840514&tree=Mozilla-B2g26-v1.2

b2g_mozilla-b2g26_v1_2_win32_gecko build on 2013-10-29 07:07:02 PDT for push 805d4143f5d5
slave: w64-ix-slave43

TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_android
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_arm
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_crashreporter
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_debug
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_linux
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_mac
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_mac_universal
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_missing
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_unknown
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_win
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_x86
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_basic
TEST-PASS | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_mozinfo.py | test_fileobj
e:\builds\moz2_slave\m-b2g26J-w32_g-000000000000000\build\config\rules.mk:233:0: command 'PYTHONDONTWRITEBYTECODE=1 e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/obj-firefox/_virtualenv/Scripts/python.exe e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_pythonutil.py' failed, return code 1
TEST-UNEXPECTED-FAIL | e:/builds/moz2_slave/m-b2g26J-w32_g-000000000000000/build/python/mozbuild/mozbuild/test/test_pythonutil.py | line 18, test_iter_modules_in_path: Lists differ: ['e:\\builds\\moz2_slave\\m-b2... != ['e:\\builds\\moz2_slave\\m-b2...

First differing element 0:
e:\builds\moz2_slave\m-b2g26j-w32_g-000000000000000\build\python\mozbuild\mozbuild\__init__.py
e:\builds\moz2_slave\m-b2g26J-w32_g-000000000000000\build\python\mozbuild\mozbuild\__init__.py

Diff is 912 characters long. Set self.maxDiff to None to see it.

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Glandium, I see that test_pythonutil.py is your doing. Any insights into what might be going wrong here?

Comment 2

[Ryan VanderMeulen [:RyanVM]]

Happens on regular desktop Firefox builds as well.
https://tbpl-dev.allizom.org/php/getParsedLog.php?id=29852959&tree=Mozilla-B2g26-v1.2

Comment 3

[Mike Hommey [:glandium]]

b2g26j vs. b2g26J.

Comment 4

[Ryan VanderMeulen [:RyanVM]]

(In reply to Mike Hommey [:glandium] from comment #3)
> b2g26j vs. b2g26J.

Aki/Hal, any idea where this is coming from?

Comment 5

[Aki Sasaki (not active)]

Seems like a test/code problem... possibly in moz.build ?

Comment 6

[Justin Wood (:Callek)]

So I am pretty sure the underlying problem is a broken test, that said the 'J' is comming in from nowhere, and the bug with the J also affects other trees...

>>> name
'm-b2g26J-w32_g'
>>> for word, replacement in mappings.iteritems():
...     if word in name:
...         regex = re.compile(
...     r'(-|_|\A)' +   # Are the start of the string or after a separator
... ('%s' % word) + # Contain the word we're looking
... r'(-|_|\Z)'     # And don't have anything other the end of the string
... )
...         name = regex.sub(r'\1%s\2' % replacement, name)
...
>>> name
'm-b2g26J-w32_g'
>>> str('\112')
'J'
>>> name = 'm-b2g26J-w32_g'
>>> for word, replacement in mappings.iteritems():
...     if word in name:
...         regex = re.compile(
...     r'(-|_|\A)' +   # Are the start of the string or after a separator
... ('%s' % word) + # Contain the word we're looking
... r'(-|_|\Z)'     # And don't have anything other the end of the string
... )
...         name = regex.sub(r'\g<1>%s\g<2>' % replacement, name)
...
>>> name
'm-b26_12-w32_g'

applied patch with r=aki over IRC https://hg.mozilla.org/build/buildbotcustom/rev/d1b91671f6cc

Comment 7

[Justin Wood (:Callek)]

This is indeed an issue on windows though with the test file (and possibly other code based on these assumptions), so needinfo to glandium since he wrote the test.

Comment 8

[Armen [:armenzg]]

In production (not added to the CC list).

Comment 9

[Mike Hommey [:glandium]]

The problem, from the test standpoint, is that it's invoked with b2g26J, so __file__ returns b2g26J (or so I think), while the modules as returned by python use b2g26j, thus the mismatch. The test /could/ be fixed to normalize case of the paths before comparing them, but we likely have the same problem in various other places.

Comment 10

[Ryan VanderMeulen [:RyanVM]]

Windows builds are green again, so I'm going to call this fixed based on glandium's reply. Thanks everyone!