Eliminate 'ThemeChanged' warning when running IDLE tests

[terryjreedy]

BPO	27196
Nosy	@terryjreedy, @ned-deily, @serhiy-storchaka

^{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 = 'https://github.com/terryjreedy'
closed_at = None
created_at = <Date 2016-06-03.08:00:06.971>
labels = ['expert-IDLE', 'type-bug', 'expert-tkinter', '3.9']
title = "Eliminate 'ThemeChanged' warning when running IDLE tests"
updated_at = <Date 2020-01-07.22:42:55.465>
user = 'https://github.com/terryjreedy'

bugs.python.org fields:

activity = <Date 2020-01-07.22:42:55.465>
actor = 'terry.reedy'
assignee = 'terry.reedy'
closed = False
closed_date = None
closer = None
components = ['IDLE', 'Tkinter']
creation = <Date 2016-06-03.08:00:06.971>
creator = 'terry.reedy'
dependencies = []
files = []
hgrepos = []
issue_num = 27196
keywords = []
message_count = 7.0
messages = ['267078', '267081', '267226', '267228', '268861', '269089', '359559']
nosy_count = 4.0
nosy_names = ['terry.reedy', 'ned.deily', 'python-dev', 'serhiy.storchaka']
pr_nums = []
priority = 'normal'
resolution = None
stage = 'needs patch'
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue27196'
versions = ['Python 3.9']

Linked PRs

• GH-71383: IDLE - Document testing subsets of modules #104463
• [3.11] GH-71383: IDLE - Document testing subsets of modules (GH-104463) #104464
• gh-71383: add upstream fix to macOS installer #108806
• gh-71383: Update Tcl/Tk version in Windows to our patched build containing a targeted upstream fix #112973

[terryjreedy]

When running "python_d.exe -m test_idle from console or when running test file from IDLE after starting it in either console or by import, the following sometimes appears.

can't invoke "event" command: application has been destroyed
while executing
"event generate $w <<ThemeChanged>>"
(procedure "ttk::ThemeChanged" line 6)
invoked from within
"ttk::ThemeChanged"

Serhiy, this is pobably not critical, but it is annoying. Perhaps the 'kill-callbacks' patch would fix this. Still, I try to have IDLE tests cleanup properly.

Known repeatable occurrences:

• test_configdialog.ConfigDialogTest.test_dialog, when run from IDLE after import, but not when part of suite.
• test_replace.ReplaceDialogTest.test_default_command, 3 times, when run as part suite, but not when run from IDLE after import.

[serhiy-storchaka]

This looks as a duplicate of bpo-20567.

[python-dev]

New changeset 0207c6203f84 by Terry Jan Reedy in branch '3.5':
Issue bpo-27196: Stop 'application destroyed' warnings when running IDLE tests.
https://hg.python.org/cpython/rev/0207c6203f84

New changeset 2d864ac472f0 by Terry Jan Reedy in branch 'default':
Merge issue bpo-27196 from 3.5: Stop IDLE test 'application destroyed' warnings.
https://hg.python.org/cpython/rev/2d864ac472f0

[terryjreedy]

I closed bpo-20567 after finishing what I intended to do 3 years ago. I decided make this a separate issue for trying Serhiy's suggestion of updata_idletasks. Pursuing 'no default root' is bpo-24137.

3 other files also gave the warning, once, when run alone. 2 of the 4 'warned with the test suite in 3.5 and 3 of 4 in 3.6. Adding root.update or root.update_idletasks to all 4 fixed all warnings I know of for both versions. It appears to be a coincidence that the warnings consistently appeared where they did. I also added a note about this fix to README.txt.

There is no known problem in 2.7. I fixed 3.5 also because the warnings sometimes appear in buildbot reports, and they don't need the noise.

[ned-deily]

FWIW, I'm still seeing these errors on OS X (10.11.5 with ActiveTcl 8.5.18) using 2.7.12rc1, 3.5.2rc1, and top-of-trunk 3.6.0. They do not occur on the same platform when using 2.7.11 or 3.5.1.

For example:
$ /usr/local/bin/python2.7
Python 2.7.12rc1 (v2.7.12rc1:13912cd1e7e8, Jun 11 2016, 15:32:34)
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)] on darwin [...]
$ /usr/local/bin/python2.7 -m test.regrtest -ugui -j3 test_idle
can't invoke "event" command: application has been destroyed
while executing
"event generate $w <<ThemeChanged>>"
(procedure "ttk::ThemeChanged" line 6)
invoked from within
"ttk::ThemeChanged"
can't invoke "event" command: application has been destroyed
while executing
"event generate $w <<ThemeChanged>>"
(procedure "ttk::ThemeChanged" line 6)
invoked from within
"ttk::ThemeChanged"
[1/1] test_idle
1 test OK.

$ /usr/local/bin/python3.5
Python 3.5.2rc1 (v3.5.2rc1:68feec6488b2, Jun 11 2016, 21:59:53)
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)] on darwin [...]
$ /usr/local/bin/python3.5 -m test.regrtest -ugui test_idle
[1/1] test_idle
can't invoke "event" command:  application has been destroyed
    while executing
"event generate $w <<ThemeChanged>>"
    (procedure "ttk::ThemeChanged" line 6)
    invoked from within
"ttk::ThemeChanged"
1 test OK.

$ ./bin/python3.6
Python 3.6.0a2+ (default, Jun 19 2016, 13:19:06)
[GCC 4.2.1 Compatible Apple LLVM 7.3.0 (clang-703.0.29)] on darwin [...]
$ ./bin/python3.6 -m test -ugui -j3 test_idle
Run tests in parallel using 3 child processes
0:00:02 [1/1] test_idle passed
can't invoke "event" command:  application has been destroyed
    while executing
"event generate $w <<ThemeChanged>>"
    (procedure "ttk::ThemeChanged" line 6)
    invoked from within
"ttk::ThemeChanged"
1 test OK.
Total duration: 0:00:03

[terryjreedy]

The only thing I still see is a pair of ThemeChanged warnings from 2.7. Using binary search, I identified test_editmenu as the culprit. This in spite of all the cleanup I can think of

    @classmethod
    def tearDownClass(cls):
        del cls.text, cls.entry, cls.spin
        cls.root.clipboard_clear()
        cls.root.update_idletasks()
        cls.root.update()
        cls.root.destroy()
        del cls.root

When I tried to determine which test method(s) gave the warning, as disable all to make sure the warnings vanished, which then did. But after re-enabling, the warnings did not come back -- until they did after a few more runs. I suspect that the only permanent solution will be Serhiy's patch to have .destroy cancel pending callbacks.

[terryjreedy]

2.7 EOL.

Ned, do you still see Themed Changed? I don't on 3.8 on Macbook. with 10.14 mohave(?)

[chrstphrchvz]

Based on the error message, I believe this is a known upstream Tk issue (edit: which will be fixed in Tk 8.6.14): https://core.tcl-lang.org/tk/info/2958768, https://core.tcl-lang.org/tk/info/310c74ecf440

[terryjreedy]

The 2010-date tcl-tk ticket says that this is known when "an interpreter is created and then immediately destroyed". Last reproduced in Sept 2022 with 8.6 on Win 10. I believe I have seen without immediate close, but will see if possible.

I currently see 3 with python -m test.test_idle. Adding -v and the messages appear after
test_iomenu.IOBindingTest.test_init
test_searchbase.SearchDialogBaseTest.test_open_and_close
test_textview.ButtonClickTest.test_view_text_bind_with_button
These are each the last test in the file.

The above is not directly useful as the current glitches disappear when running 1 test file. I discovered that the first appears when test_history, test_hyperparser, and test_iomenu are run together, but not when iomenu or the other two are run separately. My next PR will be to update the 2 master test files and add a comment in idle_test.init on how to run subsets.

[chrstphrchvz]

Tcl/Tk has since merged a fix which will appear in 8.6.14. However, a patch for the fix is small and should apply cleanly to the past few Tk releases. Is Python opposed to backporting accepted fixes for its bundled Tcl/Tk?

[terryjreedy]

When would you expect 8.6.14?

[chrstphrchvz]

The past 6 Tcl/Tk 8.6.x releases were around November each year, when the release manager got around to doing so and after any critical fixes were made to Tk Aqua for the latest macOS. So I would guess 8.6.14 is still several months away.

[chrstphrchvz]

I have opened #108806 and python/cpython-source-deps#29 in case Python is interested in applying the upstream fix to the bundled Tcl/Tk 8.6.8 and 8.6.13 in the macOS and Windows installers.

[ned-deily]

Thanks for the PRs. The fix for Tk in the macOS installer has now been merged for release in 3.13.0a2, 3.12.1, and 3.11.7.

[zooba]

Windows fix is now committed for 3.13. As discussed elsewhere, I'm not comfortable backporting the patched build to 3.12 or earlier without someone present to promptly debug and fix any issues that arise.

[terryjreedy]

Unless Serhiy wants to do the Windows backports, I am happy to have this closed. I currently cannot repeat the problems on 3.11/12 Windows I reported last May in #71383 (comment). (This is not the first time the messages have appeared and disappeared.) And the patch should prevent any future appearances. If something reappears in 3.12, I now know not to waste more time trying to make it go away. If something reappears in main after this patch, it should be a new issue.