[FFmpeg-devel] Broken compile with latest libavutil/common.h

[Aurelien Jacobs]

On Sun, 18 Jan 2009 00:24:30 +0000
M?ns Rullg?rd <mans at mansr.com> wrote:

> Aurelien Jacobs <aurel at gnuage.org> writes:
> 
> > On Sat, 17 Jan 2009 21:52:16 +0100
> > Diego Biurrun <diego at biurrun.de> wrote:
> >
> >> On Sat, Jan 17, 2009 at 08:30:06PM +0100, Aurelien Jacobs wrote:
> >> > On Sat, 17 Jan 2009 16:19:17 +0000
> >> > M?ns Rullg?rd <mans at mansr.com> wrote:
> >> > > 
> >> > > If it's going to libavcodec, I'd prefer it in mathops.h.  That file
> >> > > has per-arch overrides already, and separating the assembler in that
> >> > > function as well would be nice.
> >> > 
> >> > That sounds like a good idea. Attached patch moves mid_pred() in mathops.h
> >> > and the x86 specific implementation in x86/mathops.h.
> >> > 
> >> > --- libavcodec/mathops.h	(revision 16652)
> >> > +++ libavcodec/mathops.h	(working copy)
> >> > @@ -83,5 +83,35 @@
> >> >  
> >> > +/* median of 3 */
> >> > +#ifndef mid_pred
> >> > +#define mid_pred mid_pred
> >> > +static inline av_const int mid_pred(int a, int b, int c)
> >> > +{
> >> > +#if 0
> >> 
> >> Maybe this would be a good point to get rid of the disabled code..
> >
> > I agree, but it would probably be better in a separate patch.
> >
> >> > --- libavcodec/x86/mathops.h	(revision 16652)
> >> > +++ libavcodec/x86/mathops.h	(working copy)
> >> > @@ -22,6 +22,8 @@
> >> >  #ifndef AVCODEC_X86_MATHOPS_H
> >> >  #define AVCODEC_X86_MATHOPS_H
> >> >  
> >> > +#include "libavutil/avutil.h"
> >> 
> >> What is this header needed for?  It looks rather as if you should
> >> #include config.h and common.h...
> >
> > AFAIK, avutil.h is granting inclusion of config.h and common.h.
> > So I think it is simpler, cleaner and more future proof to
> > include avutil.h. Not that it is important...
> 
> avutil.h also contains/includes other stuff.  In general, dependencies
> should be kept minimal.  If you need avutil.h for some other reason,
> you may omit the things it promises to include.  Do not use it merely
> as a shorthand for what you actually need.

Well, OK. Changed #include in attached patch.

Aurel
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mathops_mid_pred.diff
Type: text/x-diff
Size: 8640 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090118/946562a8/attachment.diff>