[MPlayer-dev-eng] [PATCH] Crash in vf_spp.c and other things

[Diego Biurrun]

On Mon, Aug 28, 2006 at 03:36:34AM +0200, Gianluigi Tiesi wrote:
> On Mon, Aug 28, 2006 at 03:12:40AM +0200, Gianluigi Tiesi wrote:
> > On Mon, Aug 28, 2006 at 12:00:09AM +0200, Diego Biurrun wrote:
> > > On Sun, Aug 27, 2006 at 10:33:03AM +0200, Michael Niedermayer wrote:
> > > > 
> > > > On Sun, Aug 27, 2006 at 06:22:58AM +0200, Gianluigi Tiesi wrote:
> > > > > Attached patch removes an if-def that rip out some code
> > > > > if CONFIG_ENCODERS is not defined.
> > > > > 
> > > > > vf_spp.c at line 412:
> > > > > 
> > > > > p->dsp.fdct(block);
> > > > > 
> > > > > but dsp.fdct is never assigned if CONFIG_ENCODERS
> > > > > is not defined in libavcodec/i386/dsputil_mmx.c
> > > > 
> > > > rejected, dct belongs to the encoding stuff and should be under that ifdef
> > > > add a CONFIG_DCT if you want instead
> > > 
> > > The crash is fixed now, I've reenabled CONFIG_ENCODERS for MPlayer as
> > > well.
> > > 
> > So I cannot use vf=ssp if I use --disable-mencoder?
> > Also what about fast_memcpy crash? it's asm code I can debug
> > with ollydbg, but I known nothing about mmx asm
> > It maybe a win32 related problem? (I'll try on linux)
> > 
> Crash in p->dsp.fdct seams to be fixed, at least on win32
> I get p->dsp.fdct maped to ff_fdct_mmx2.
> I see CONFIG_ENCODER is defined, this would increase mplayer
> executable size also when using --disabled-mencoder.

Yes, but that's better than a crash.

> Michael's solution looks better, but why if dct belongs to encoding
> stuff there is a postprocessing filter that uses it?

I'd be happy if someone implemented Michael's solution.  But some
encoding functions are necessary for other filters like vf_lavc for
example, which needs mpeg1 encoding capabilities.

Diego