[FFmpeg-devel] [PATCH] high precision mp3 decoding needs CONFIG_AUDIO_NONSHORT which is never set

[Diego Biurrun]

On Tue, Dec 09, 2008 at 10:56:40PM +0100, Benoit Fouet wrote:
> Diego Biurrun wrote :
> > On Tue, Dec 09, 2008 at 08:57:57AM +0100, Benoit Fouet wrote:
> >   
> >> On 12/09/2008 12:42 AM, Diego Biurrun wrote:
> >>     
> >>> On Mon, Dec 08, 2008 at 12:29:16PM +0100, Benoit Fouet wrote:
> >>>       
> >>>> On 12/08/2008 12:10 PM, Diego Biurrun wrote:
> >>>>         
> >>>>> On Mon, Dec 08, 2008 at 10:00:57AM +0100, Benoit Fouet wrote:
> >>>>>   
> >>>>>> reg tests fail here with this patch:
> >>>>>>         
> >>>>> Do they also fail if you make the change in both mpegaudiodec.c and
> >>>>> mpegaudio.h?
> >>>>>           
> >>>> yes, they also fail
> >>>>         
> >>> This does not look right.  Why are the libavformat regression tests
> >>> failing due to a change in libavcodec?  And especially a change to
> >>> preprocessor conditions that are not normally fulfilled?
> >>>       
> >> they are not fulfilled at the moment.
> >> #ifdef CONFIG_AUDIO_NONSHORT is an #if 0, no ?
> >
> > Correct.  But both options are never defined, so going from
> >
> > #if defined(USE_HIGHPRECISION) && defined(CONFIG_AUDIO_NONSHORT)
> >
> > to
> >
> > #if defined(USE_HIGHPRECISION)
> >
> > is like going from '#if 0' to '#if 0'.  So it should not affect
> > regression tests at all.
> 
> USE_HIGHPRECISION is defined if CONFIG_MPEGAUDIO_HP is defined... which
> can be true

I know, but only if you manually set it, configure has no option for it.
Did you manually add it to config.h?  My guess would be no.  Then there
should be no way for this to affect regression tests.

Diego