Stderr and Pythonw

[vm_usenet]

Hi everyone,

I've seen this subject come up in another thread, but not getting
enough attention - so I'll just ask it:
When running python code via pythonw, what exactly happens to stderr?
I tried running code that, among else, writes to stderr - and at some
point I get an IOError claiming "Bad file descriptor".

Why is this happending?

Thanks in advance,

vm

[Manlio Perillo]

On Wed, 30 Jun 2004 04:22:16 GMT, Dennis Lee Bieber
<wlf...@ix.netcom.com> wrote:

>On 29 Jun 2004 12:25:58 -0700, vm_u...@yahoo.com (vm_usenet) declaimed
>the following in comp.lang.python:

>
>> When running python code via pythonw, what exactly happens to stderr?
>

> You aren't invoking programs by typing "pythonw program.py" in a
>command console window, are you?
>
> pythonw was intended to be used for programs activated by
>clicking on icons in Windows. As such, these programs are presumed to
>have NO CONSOLE I/O, all I/O must be handled by the program creating a
>window (GUI) or file and writing to that.
>
> Since there is no console expected, none of the std***
>connections are created.
>

This not really corrected.
Actually 'some' std*** connections are created but are usefulness
(and moreover they have a bug).

If you write more than 4096 an exception is raised...

Regards Manlio Perillo

[Peter Hansen]

Manlio Perillo wrote:

> On Wed, 30 Jun 2004 04:22:16 GMT, Dennis Lee Bieber
> <wlf...@ix.netcom.com> wrote:
>>On 29 Jun 2004 12:25:58 -0700, vm_u...@yahoo.com (vm_usenet) declaimed
>>the following in comp.lang.python:
>>>When running python code via pythonw, what exactly happens to stderr?
>>

>> Since there is no console expected, none of the std***
>>connections are created.
>
> This not really corrected.
> Actually 'some' std*** connections are created but are usefulness
> (and moreover they have a bug).
>
> If you write more than 4096 an exception is raised...

And, for future reference, here is the bug report Manlio filed
on SourceForge (for which there is no response yet):

http://sourceforge.net/tracker/index.php?func=detail&aid=973507&group_id=5470&atid=105470

I've also confirmed this, using a test script very similar to one
Tim Peters posted a couple of months ago in
http://groups.google.ca/groups?selm=mailman.980.1082749579.20120.python-list%40python.org

The problem is confirmed to still exist with Python 2.3.4.

-Peter