[Bf-committers] Python console IO encoding: PYTHONIOENCODING envvar

Campbell Barton ideasman42 at gmail.com
Mon Sep 16 12:46:46 CEST 2013

While its good you found a solution for this case my main concern is
that we still have many other areas that print files or show filepaths
in the UI, so you can close one bug but others will be reported.

Basically if you were to apply this fix to Blender all over, it would
be a much larger patch.

I had a fix that replaced sys.stdout with one that had different
encoding, it worked well *except* it crashed on exit for debug builds
in windows (some file resource management assert caused by overwriting

I was hoping Python devs would provide a solution by next release, but
theres no activity on the bug report to indicate that.

We could also just investigate a better hack for sys.stdout to force
it to be the encoding we want and avoid having to add workarounds all
over our python scripts.
For now this is my preference since at least the change is in one
place and we don't have to worry about adding safe_print's to addons

Note, its not good practice to modify kwargs, you can do this instead.
 print(*args, file=sys.stderr, **kwargs)

On Mon, Sep 16, 2013 at 5:27 PM, Bastien Montagne <montagne29 at wanadoo.fr> wrote:
> Well, in the bug report error comes from a print in addon_utils.py...
> Here is a patch (small wrapper around print that sets 'file' kwarg to
> sys.stderr) that fixes the issue, at least for me (tested with an ascii
> console): http://www.pasteall.org/45600/diff
> On 16/09/2013 08:40, Campbell Barton wrote:
>> On Mon, Sep 16, 2013 at 4:17 PM, Bastien Montagne<montagne29 at wanadoo.fr>  wrote:
>>> Hi Campbell,
>>> Could we consider using stderr in these cases, as a temp workaround?
>>>   From doc, "For stderr, the :errorhandler part is ignored; the handler
>>> will always be 'backslashreplace'.", and indeed a quick test (under
>>> linux) prints escapes codes in an ASCII console, instead of erroring...
>> Ah nice, I wasn't aware of this,
>> however I think using the stderr isn't going to be able to resolve all
>> these issues.
>> print() can be replaced with sys.stderr.write(), but in the bug report
>> the error is in %r ->  repr(). which makes a string to pass to Blenders
>> UI.
>> So not sure how stderr would be used to get around this.
>>> On 14/09/2013 02:40, Campbell Barton wrote:
>>>> On Fri, Sep 13, 2013 at 11:02 PM, Bastien Montagne
>>>> <montagne29 at wanadoo.fr>   wrote:
>>>>> Hey devs,
>>>>> Just ran into this bug report:
>>>>> https://projects.blender.org/tracker/?func=detail&atid=498&aid=36722&group_id=9
>>>>> Most likely, it is an issue of non-encodable char in its path (stupid
>>>>> windows...)
>>>>> Anyway, searching a bit around, I found PYTHONIOENCODING envvar
>>>>> (http://docs.python.org/3/using/cmdline.html#envvar-PYTHONIOENCODING),
>>>>> which is supposed to force encoding of stdXXX streams. Are we already
>>>>> using this, or wouldn't it work in our case? Else, setting it to
>>>>> something like "utf-8:replace" could save us some issues under windows...
>>>> Afraid its a known issue, see comment in BPY_python_start():
>>>> 'PYTHONIOENCODING' is ignored in MS-Windows
>>>> bug report:
>>>> https://projects.blender.org/tracker/?func=detail&aid=31555
>>>> upstream report to CPython:
>>>> http://bugs.python.org/issue16129
>>> _______________________________________________
>>> Bf-committers mailing list
>>> Bf-committers at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-committers
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers

- Campbell

More information about the Bf-committers mailing list