[FFmpeg-devel] Getting rid of codec_get_{bmp,wav}_{id,tag}()

[Reimar Döffinger]

Hello,
On Mon, Jul 09, 2007 at 05:58:26PM +0200, Benoit Fouet wrote:
> Reimar D?ffinger wrote:
> > On Mon, Jul 09, 2007 at 04:07:14PM +0100, M?ns Rullg?rd wrote:
> >> Reimar D?ffinger wrote:
> >>> On Mon, Jul 09, 2007 at 11:26:57AM +0100, M?ns Rullg?rd wrote:
> >>> [...]
> >>>       
> >>>>>>> you mean that values are correct but reordered, that's it ?
> >>>>>>>               
> >>>>>> At a glance, that's what it looks like.
> >>>>>>             
> >>>>> i attach the result of the following command:
> >>>>> $ ls tests/data/a-* tests/data/b-* | sort
> >>>>>
> >>>>> if you want to compare with yours...
> >>>>>           
> >>>> I'm at work now, but the number and names of files are consistent
> >>>> with what I recall from running the tests.
> >>>>
> >>>> It's the output of seektest that for some reason gets reordered.
> >>>>         
> >>> I think the reference file was created with a broken sort, because in
> >>> that "a-mpeg1b.mpg" comes before "a-mpeg1.mpg" which I'd consider wrong
> >>> no matter what kind of sorting.
> >>>       
> >> Some locales ignore punctuation when sorting, which would explain this.
> >
> > Yes, they succeed with LC_ALL=de_DE. We should place a LC_ALL=C in the
> > script and regenerate the reference file IMO.
> > I still don't consider that proper sorting, but maybe some standard
> > actually says that's the way you're supposed to do it...
> 
> before playing with env variables, have you tried giving arguments to
> sort (like -d for instance) ?

While that gives the same order as the current references file using it
is broken, since the order of e.g. a file named "a.b.c" vs "abc" is not
specified according to the manpage (though gnu sort seems to fallback to
C sorting).
I also can't find anything that would indicate that -d is independant of
the locale. Keep in mind that e.g. for tr_TR even the ascii range
(specifically the 'i') has different behaviour, so even if
"alphanumeric" was supposed to mean ascii-alphanumeric range (and I
tested, it also includes charset-specific things like ? and ?) it would
still fail in some cases.

Greetings,
Reimar D?ffinger