[Ffmpeg-devel] H264 decoding problem

[Michael Niedermayer]

Hi

On Thu, Oct 12, 2006 at 05:22:50PM +0200, Guillaume Poirier wrote:
> Hi,
> 
> Uoti Urpala wrote:
> > On Thu, 2006-10-12 at 16:34 +0200, Guillaume Poirier wrote:
> > 
> >>The bug is still there with latest snapshot (*4.2 20061007), I filed a
> >>bug report here:
> >>http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29440
> > 
> > 
> > That doesn't look like a good bug report. 

and you dont look like a polite and diplomatic person
and anyway sending good bugreports to the gcc devels is hard (just try and
you will see how many hours it can take to make a small testcase.c)


> > Is there any indication that
> > the problems are caused by a compiler bug? You don't even say what the
> > problem is except "results in a grey picture". What good does attaching
> > the "miscompiled" files do if that's all the problem description you
> > give - do you think the gcc developers can check whether the produced
> > code will "give a gray picture", except by building whole FFmpeg?
> 
> I'm sorry, but it's just that I'm so terribly stupid that I don't know
> what else I can do at _this_ time.

you could send me a compiled but not assembled h264.c ...
1. start with the case which generates a broken ffmpeg
2. rm h264.o ; make
3. copy and paste the line which compiles h264.o
4. replace -c -o h264.o by -S -o h264.s
5. send h264.s to me or upload it somewhere (its ~ 3mb)

ill bet gcc ignored some of the asm constraints, at least thats what it
does most of the time when it misscompiles some asm

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

In the past you could go to a library and read, borrow or copy any book
Today you'd get arrested for mere telling someone where the library is