Tkinter tests failed on Windows buildbots

[serhiy-storchaka]

BPO	25848
Nosy	@loewis, @pfmoore, @larryhastings, @tjguk, @bitdancer, @zware, @serhiy-storchaka, @zooba
Files	pep101_custom_buildbots.diff

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2016-02-09.16:44:08.597>
created_at = <Date 2015-12-12.17:01:11.980>
labels = ['tests', 'type-bug', 'expert-tkinter', 'OS-windows']
title = 'Tkinter tests failed on Windows buildbots'
updated_at = <Date 2016-02-09.16:44:08.596>
user = 'https://github.com/serhiy-storchaka'

bugs.python.org fields:

activity = <Date 2016-02-09.16:44:08.596>
actor = 'zach.ware'
assignee = 'none'
closed = True
closed_date = <Date 2016-02-09.16:44:08.597>
closer = 'zach.ware'
components = ['Tests', 'Tkinter', 'Windows']
creation = <Date 2015-12-12.17:01:11.980>
creator = 'serhiy.storchaka'
dependencies = []
files = ['41389']
hgrepos = []
issue_num = 25848
keywords = ['patch']
message_count = 11.0
messages = ['256287', '256294', '256748', '256803', '256826', '256844', '256845', '256846', '256857', '259945', '259946']
nosy_count = 9.0
nosy_names = ['loewis', 'paul.moore', 'larry', 'tim.golden', 'r.david.murray', 'python-dev', 'zach.ware', 'serhiy.storchaka', 'steve.dower']
pr_nums = []
priority = 'low'
resolution = 'out of date'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue25848'
versions = ['Python 3.4']

[serhiy-storchaka]

Tkinter related tests are failed on 4 of 6 Windows 3.4 buildbots.

http://buildbot.python.org/all/builders/AMD64%20Windows7%20SP1%203.4/builds/1418/steps/test/logs/stdio
http://buildbot.python.org/all/builders/AMD64%20Windows10%203.4/builds/220/steps/test/logs/stdio
http://buildbot.python.org/all/builders/AMD64%20Windows8%203.4/builds/779/steps/test/logs/stdio
http://buildbot.python.org/all/builders/AMD64%20Windows8.1%20Non-Debug%203.4/builds/333/steps/test/logs/stdio

======================================================================
ERROR: setUpModule (test.test_tcl)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "C:\buildbot.python.org\3.4.kloth-win64\build\lib\test\test_tcl.py", line 685, in setUpModule
    tcl = Tcl()
  File "C:\buildbot.python.org\3.4.kloth-win64\build\lib\tkinter\__init__.py", line 1955, in Tcl
    return Tk(screenName, baseName, className, useTk)
  File "C:\buildbot.python.org\3.4.kloth-win64\build\lib\tkinter\__init__.py", line 1856, in __init__
    self.tk = _tkinter.create(screenName, baseName, className, interactive, wantobjects, useTk, sync, use)
_tkinter.TclError: Can't find a usable init.tcl in the following directories: 
    C:/buildbot.python.org/3.4.kloth-win64/build/PCbuild/lib/tcl8.6 C:/buildbot.python.org/3.4.kloth-win64/build/PCbuild/lib/tcl8.6 C:/buildbot.python.org/3.4.kloth-win64/build/lib/tcl8.6 C:/buildbot.python.org/3.4.kloth-win64/build/PCbuild/library C:/buildbot.python.org/3.4.kloth-win64/build/library C:/buildbot.python.org/3.4.kloth-win64/build/tcl8.6.1/library C:/buildbot.python.org/3.4.kloth-win64/tcl8.6.1/library

This probably means that Tcl wasn't installed properly.

----------------------------------------------------------------------

Is this a test suite defect, or there is an issue with using Tkinter on Windows? In latter case this is a release blocker.

[zooba]

I was encountering this recently trying to set up a build machine for the release, and it seems to be a build issue.

Martin is building the release and apparently has it working.

[larryhastings]

I'm assuming this is a configuration issue on the buildbots.

[zware]

Larry: considering that 3.4 is now officially out of bugfix mode, how long do you want to keep the 3.4 buildbots around? From a selfish buildmaster configurator perspective, axing the 3.4 bots simplifies a few things :)

[larryhastings]

I hear your plea. But isn't it almost as easy to just leave it alone? 3.4 won't get many checkins from now on, so it won't actually kick off many tasks. But it would mean the "check the buildbots" step of making a release would be super easy, rather than the huge pain of requesting a manual build on the buildbots.

[bitdancer]

Well, it clutters up the buildbot web interface with builders that aren't being used. Our historical pattern has been to deactivate the builders after a release goes into security-fix-only mode.

Zach, would it be practical to just comment them out and then reactivate them when Larry does a release, or would that eliminate the benefit from your point of view?

[zware]

Would pretty much eliminate the benefit; there are currently a few special cases for 3.4 (particularly for the ICC builders, and the XP bot), which could go away.

For the record, the secret undocumented method for running a build on all custom builders is to fill out the third-to-last set of entry boxes (the one that includes a "Repo path" box) on http://buildbot.python.org/all/builders and click the associated 'Force Build' button. It's really not that hard to request such a build, but I will grant that it's rather inconvenient.

On the other hand, it wouldn't be too hard to just add the branch back without the special cases around release time for a little while, but there may be some expected failures.

[bitdancer]

How about just keeping the stable set configs, minus any from that set that have special cases? Although if it is just one form submission, maybe we just need to document that in the release manager's guide?

[zware]

I think only keeping some builders would be more complex than any other option. Here's a patch against PEP-101 to document the custom builder trick.