homepage

in py3k, the following simple code will throw an uncatched exception when executed with pythonw:
import warnings
warnings.warn('test')
the problem occurs in showarning function: in py3k's pythonw , stderr/stdout is set to None, so the file.write(...) statement will thorw AttributeError uncatched. I think a catch-all except(delete 'OSError') can solve this.
def showwarning(message, category, filename, lineno, file=None, line=None):
 """Hook to write a warning to a file; replace if you like."""
 if file is None:
 file = sys.stderr
 try:
 file.write(formatwarning(message, category, filename, lineno, line))
 except OSError:
 pass # the file (probably stderr) is invalid - this warning gets lost.

Here is a patch. 2.7 is affected too.

1. py2 is unaffected, because stderr/stdout is not None in py2's pythonw and the catch works correctly.
2. as to py3, I prefer this patch:
- execpt OSError:
+ execpt:

+ execpt:
Please never use that, but "except Exception:" instead to not catch SystemExit or KeyboardInterrupt.

I afraid this will silence unexpected errors.

warnings_stderr_none.patch looks good to me. It can be applied on Python 2.7, 3.4 and 3.5.

New changeset d04dab84388f by Serhiy Storchaka in branch '3.4':
Issue #23016: A warning no longer produces AttributeError when the program
https://hg.python.org/cpython/rev/d04dab84388f
New changeset 0050e770b34c by Serhiy Storchaka in branch 'default':
Issue #23016: A warning no longer produces an AttributeError when the program
https://hg.python.org/cpython/rev/0050e770b34c
New changeset aeeec8a4b9b8 by Serhiy Storchaka in branch '2.7':
Issue #23016: A warning no longer produces an AttributeError when sys.stderr
https://hg.python.org/cpython/rev/aeeec8a4b9b8 

> + # sys.stderr is None when ran with pythonw.exe - warnings get lost
s/ran/run/

New changeset 70b6fe58c425 by Serhiy Storchaka in branch '3.4':
Fixed a typo in a comment (issue #23016).
https://hg.python.org/cpython/rev/70b6fe58c425
New changeset da1ec8e0e068 by Serhiy Storchaka in branch 'default':
Fixed a typo in a comment (issue #23016).
https://hg.python.org/cpython/rev/da1ec8e0e068 

Thanks Arfrever.