[MPlayer-cvslog] r27788 - in trunk: TOOLS/fastmemcpybench.c configure cpudetect.c cpuinfo.c drivers/radeon_vid.c liba52/downmix.c liba52/imdct.c liba52/imdct_3dnow.h liba52/liba52_changes.diff liba52/resample_mmx.c liba52/test.c libmpcodecs/dec_video.c libmpcodecs/pullup.c libmpcodecs/vf_decimate.c libmpcodecs/vf_divtc.c libmpcodecs/vf_eq.c libmpcodecs/vf_eq2.c libmpcodecs/vf_filmdint.c libmpcodecs/vf_fspp.c libmpcodecs/vf_halfpack.c libmpcodecs/vf_ilpack.c libmpcodecs/vf_ivtc.c libmpcodecs/vf_noise.c libmpcodecs/vf_pp7.c libmpcodecs/vf_sab.c libmpcodecs/vf_scale.c libmpcodecs/vf_spp.c libmpcodecs/vf_tfields.c libmpcodecs/vf_unsharp.c libmpcodecs/vf_uspp.c libmpcodecs/vf_yadif.c libvo/aclib_template.c libvo/osd_template.c loader/win32.c loader/wine/winnt.h mp3lib/dct64_3dnow.c mp3lib/dct64_k7.c mp3lib/dct64_mmx.c mp3lib/dct64_sse.c mp3lib/decode_i586.c mp3lib/decode_mmx.c vidix/AsmMacros.h vidix/sysdep/pci_os2.c

Diego Biurrun diego at biurrun.de
Thu Oct 16 22:15:53 CEST 2008


On Thu, Oct 16, 2008 at 09:57:14PM +0200, Reimar Döffinger wrote:
> On Thu, Oct 16, 2008 at 09:43:27PM +0200, Diego Biurrun wrote:
> > On Thu, Oct 16, 2008 at 09:37:12PM +0200, Reimar Döffinger wrote:
> > > On Thu, Oct 16, 2008 at 08:59:27PM +0200, diego wrote:
> > > > --- trunk/vidix/AsmMacros.h	(original)
> > > > +++ trunk/vidix/AsmMacros.h	Thu Oct 16 20:59:27 2008
> > > > @@ -99,11 +99,6 @@ void outl(U16_t, U32_t);
> > > >  
> > > >  #else /* not _MINIX and _ACK */
> > > >  
> > > > -# if defined(__STDC__) && (__STDC__ == 1)
> > > > -#  ifndef NCR
> > > > -#  define asm __asm
> > > > -#  endif
> > > > -# endif
> > > >  # ifdef SVR4
> > > >  #  include <sys/types.h>
> > > >  #  ifndef __USLC__
> > > 
> > > You really shouldn't be mixing such changes in.
> > 
> > It's related to the asm --> __asm__ change.  If I leave this
> > preprocessor hunk, the line
> > 
> >   #define intr_disable() __asm__("cli")
> > 
> > would be transformed to
> > 
> >   #define intr_disable() ____asm__("cli")
> > 
> > and probably silently break...
> 
> Yes, but changing the define to
> #  define __asm__ __asm
> would probably have kept the same behaviour (not that I object to removing it
> though).

Right, that just had not occurred to me.  So shall I put it back or
leave it as-is?

> But regardless of that I think it is better not to change such tricky,
> ancient hacks together with such huge "mechanical" changes, it just
> feels wrong to me.

Yes, you are probably right.

Diego



More information about the MPlayer-cvslog mailing list