[FFmpeg-devel] [PATCH] lavfi: Port fspp to FFmpeg

Michael Niedermayer michaelni at gmx.at
Fri Dec 19 12:21:07 CET 2014


On Fri, Dec 19, 2014 at 06:45:13AM +0000, Timothy Gu wrote:
> On Dec 18, 2014 10:31 PM, "arwa arif" <arwaarif1994 at gmail.com> wrote:
> >
> > On Fri, Dec 19, 2014 at 10:40 AM, Michael Niedermayer <michaelni at gmx.at>
> > wrote:
> > >
> > > it segfaults without -cpuflags 0 here
> > > ./ffplay   -i matrixbench_mpeg2.mpg -vf fspp
> > >
> > >
> > I tried this command on my system, and it is not giving segfault on my
> > system.
> >
> >
> > > #0  0x00000000004d0e7a in store_slice_mmx (dst=0x7fffc4612b60 "",
> > > src=0x7ffff7e07e50, dst_stride=-1440, src_stride=32, width=720,
> height=8,
> > > log2_scale=1) at libavfilter/x86/vf_fspp.c:49
> > >
> >
> > Here, the value of dst_stride is -1440, which shouldn't be negative. So,
> > when I looked into mplayer code, store_slice_mmx has input arguments of
> > type long. Maybe that is the problem, so I have updated the patch
> replacing
> > int by int32_t.
> 
> Actually ptrdiff_t is a better choice here.

yes, src_stride/dst_stride and the width argument need to be ptrdiff_t
thats a consequence of the asm accessing them as 64bit variables and
int being 32bit.
sorry i didnt had time yesterday to look at why it crashed

also gdb cannot always be trusted, especially with (non working)
asm code, the values it prints, can be wrong

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

What does censorship reveal? It reveals fear. -- Julian Assange
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20141219/c7563841/attachment.asc>


More information about the ffmpeg-devel mailing list