Roundup Robot added the comment:
New changeset 2176c1867b34 by Terry Jan Reedy in branch 'default':
Issue #18081: Back out temporary changeset, 2a9e1eb3719c, to merge new patch.
http://hg.python.org/cpython/rev/2176c1867b34
--
___
Python tracker
Terry J. Reedy added the comment:
This was sent to #18242 but not here. I expected here also.
New changeset c15d0baac3d6 by Terry Jan Reedy in branch '3.3':
Issue *18081, #18242: Change Idle warnings capture in PyShell and run
http://hg.python.org/cpython/rev/c15d0baac3d6
New changeset
Terry J. Reedy added the comment:
Improved 3.3 patch with tests incorporates Vinay's capture_warnings function to
uncapture at end of import and main exit. I plan to apply after checking other
versions and close this issue.
--
Added file:
Terry J. Reedy added the comment:
I should have changed the Versions sooner, as soon as it became obvious that
this was not just a 3.4 issue. The 'temporary' 3.4-only patch breaks forward
merging of a better patch. I will back it out just before I commit a 3.3 patch
to both eliminate the
Terry J. Reedy added the comment:
Here is a patch that looks much bigger than it really is. Idle was written
before the warnings module, so the warnings system was conditioned on a
successful warnings import. The test is no longer needed, and neither Vinay nor
Victor repeated the test at the
Brett Cannon added the comment:
I replied on python-checkins, I'll state here as well: the fix is still off
as warnings.formatwarning is still being replaced.
On Jun 24, 2013 6:20 PM, STINNER Victor rep...@bugs.python.org wrote:
STINNER Victor added the comment:
This issue is *really*
STINNER Victor added the comment:
I replied on python-checkins, I'll state here as well: the fix is still off
as warnings.formatwarning is still being replaced.
My changeset only fixes the unit test (test_idle), it's just to fix
buildbots. As I wrote, it's just a workaround.
--
Roundup Robot added the comment:
New changeset 2a9e1eb3719c by Victor Stinner in branch 'default':
Issue #18081: Workaround ./python -m test_idle test_logging failure
http://hg.python.org/cpython/rev/2a9e1eb3719c
--
nosy: +python-dev
___
Python
STINNER Victor added the comment:
This issue is *really* annoying: it makes buildbots almost useless (it is no
more possible to check if a commit introduces a regression). So until the best
fix is decided, I applied a temporary fix (based on changes.diff written by
Vinay Sajip):
New
Vinay Sajip added the comment:
With the attached patch,
python -m test_idle test_logging
passes (though test_idle still alters locale and environment).
Running python -m idelib.idle, and using IDLE itself for inspection, the
warnings.formatwarning is set to the implementation in idlelib.run.
R. David Murray added the comment:
Note that it is a known issue that tk (not python's bindings, tk itself) alters
the locale.
--
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue18081
___
Brett Cannon added the comment:
Another odd thing about this failure is that
http://hg.python.org/cpython/rev/59a11c81dd3c should have fixed it as it should
be setting warnings.showwarning() back to it's original value before the test
began executing. So I checked PyShell again and found out
Brett Cannon added the comment:
And Lib/idlelib/run.py also does the substitution. I'll file a separate bug for
all of this.
--
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue18081
___
Changes by Brett Cannon br...@python.org:
--
dependencies: +IDLE should not be replacing warnings.formatwarning
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue18081
___
Brett Cannon added the comment:
Seems to be fairly consistent on Windows but more random on Linux. I have also
triggered it on my OS X machine randomly.
Can't tell if it's a timing issue or some other test doing something bad. I'm
worried solving it is going to require taking one of the
Brett Cannon added the comment:
Found the test cases to cause this::
./python.exe -m test test_idle test_logging
Adding Roger and Terry to see if they happen to remember any of the IDLE tests
using logging or warnings in a way that could cause this.
P.S.: I thought we had a script somewhere
Terry J. Reedy added the comment:
test_idle, which runs the tests in idlelib/idle_test, currently comprises 9
test methods and perhaps 100 lines that do not, as far as I know, use warnings
or logging. However, perhaps this has an effect:
F:\Python\dev\cpython\PCbuildpython_d -m test test_idle
Brett Cannon added the comment:
I actually don't get the warnings when I run test_idle under regrtest. But what
I did was followed a hunch and added a check for warnings.showwarning() and
sure enough, it's been left modified by IDLE: PyShell replaces the function as
a side-effect of import.
Terry J. Reedy added the comment:
I discovered the same thing a different way. Grepping Lib/*.py for
'Warning (from warnings module)' hits
F:\Python\dev\cpython\Lib\idlelib\PyShell.py: 67:
F:\Python\dev\cpython\Lib\idlelib\run.py: 34:
Both monkey-patch warnings, when imported, to show warnings
Vinay Sajip added the comment:
In the meanwhile, I will try using support.import_fresh_module('warnings') in
test_logging, or better, saving and restoring warnings in test_idle.
I agree the second option (changing test_idle) is better. The failing logging
test is checking to ensure that the
Terry J. Reedy added the comment:
Vinay, please comment on this: on my Win7-64 system, as least, test_logging
seems broken without running test_idle first.
F:\Python\dev\cpython\PCbuildpython_d -m test test_logging
[1/1] test_logging
dummy.py:42: UserWarning: Explicit
Warning --
Terry J. Reedy added the comment:
I reverted the change to test_logging, reran, and it runs fine, at least twice.
Before the revert, I checked that the only two changes were the intentional
ones.
I tried adding idle_test/test_zdummy.py with a tearDownModule that deletes
Vinay Sajip added the comment:
I would change the reverted test_logging to add
print(logging._warnings_showwarning) before the failing assertions. If this
is bound to the original warnings.showwarning implementation, all should be
well. If it points to e.g. idle_showwarning, you would expect
R. David Murray added the comment:
As a side note, the various warnings about the execution environment being
modified do not appear when tests are run under unittest because those warnings
are generated by regrtest itself.
--
___
Python tracker
Terry J. Reedy added the comment:
The only locale setting is IOBinding.py: 21
# Try setting the locale, so that we can find out what encoding to use
locale.setlocale(locale.LC_CTYPE, )
This section, up to line 65, might be wrapped either by a function or if
statement.
There are only 2
Changes by Arfrever Frehtes Taifersar Arahesis arfrever@gmail.com:
--
nosy: +Arfrever
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue18081
___
New submission from R. David Murray:
The following failure is showing up frequently in the buildbots. I haven't
checked closely enough to be sure if it is a heisenberg, but I suspect it is,
since it doesn't seem to be related to any of the recent changesets that would
have been included in
27 matches
Mail list logo
