Re: [gentoo-python] Supporting non-UTF-8 locales

[IAN DELANEY <della5@iinet.com.au>]

On Sun, 17 Feb 2013 13:48:28 -0500
Mike Gilbert <floppym@gentoo.org> wrote:

> This issue seems to keep coming up in bug reports, especially those
> filed by Diego from his tinderbox. Basically, the way I/O is handled
> in python3 often causes build failures or test failures if you do not
> have a UTF-8 locale set. Ideally, upstream developers would write
> their code to deal with this, but that's a bit of a pipe dream.
> 
> Gentoo as a whole can't seem to make its mind up, but can the python
> team come to a rough consensus as to whether we want to support
> non-UTF-8 locales for python packages?
> 
> I'm sitting on the fence at the moment: I tend to fix issues if they
> cause build failures, but I ignore any test failures.
> 
Well, I tend to fix issues, be they builds of tests.

> If we decide to support non-UTF-8 locales, we need to stop leaving
> bugzilla comments like "please use a sane locale".
> 
> If we decide against supporting non-UTF-8 locales, I think we should
> begin to close such bugs as WONTFIX on a consistent basis.
> 
> My vote would be to not support non-UTF-8 locales -- it just isn't
> worth the effort.
> 
> Thoughts?
> 

Firstly, thumbs up for floating the issue. "please use a sane locale"
fundamentally undermines the merit of the tinderbox meister making such
bugs.  He deliberately makes such locales to flush out such packages
supporting only the utf8.  He's clearly convinced of the righteousness
of the bug submissions, but then again, so to for everything else he
does under the banner of working gentoo. 

The ignoring of such tests to me implies it's not really a bug, and
frankly this notion has obvious merit.  If there are utf and non utf
locales and the utf makes it work, then arguably the package is
buildable and installable.  The selection of a normative locale across
the board has obvious merit. If not utf8 simply einfo the user to set
one.

One statement arguing against I could cite from am upstream dev who,
lumbered with this exact task, agreed in principal they should not
assume a utf locale, and a fix for it has been made already by a few
maintainers of other packages, e.g. ipython.  Trouble is, he's one of a
very long queue of upstream devs.

to not support non-UTF-8 locales -- it just isn't
> worth the effort.

This is more or less the assumed and invisible policy, and
frankly on that basis I go along with Mike's selection.  Making this a
formal issue and declaring a python team policy is afaict the only
legitimate method of 'informing' the tinderbox meister these are not
bugs, these just further spam bugzilla. Then we can close such bugs
as WONTFIX  on a consistent basis without anticipation of the next dose
o' the long ago banned whip. After all, these bug submissions come
almost exclusively from him.

-- 
kind regards

Ian Delaney