[FFmpeg-devel] swscale-test segfault with 64-bit icc 11.1

[Michael Niedermayer]

On Wed, Jul 21, 2010 at 12:52:26AM -0300, Ramiro Polla wrote:
> On Wed, Jul 21, 2010 at 12:19 AM, Michael Niedermayer <michaelni at gmx.at> wrote:
> > On Tue, Jul 20, 2010 at 07:05:54PM +0200, Reimar D?ffinger wrote:
> >> On Mon, Jul 19, 2010 at 06:48:03PM -0300, Ramiro Polla wrote:
> >> > I was able to reproduce the bug with gcc but they haven't been very
> >> > helpful nor have they acknowledged it:
> >> > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44975
> >> >
> >> > I faintly remember some intel compiler dev on this list or something
> >> > similar, maybe he could comment on the issue?
> >> >
> >> > Otherwise I think the best solution is to switch to yasm like mans
> >> > suggested, but that would involve work and I'm not volunteering =).
> >>
> >> As a hack, just calling a empty function (of course make sure it won't
> >> be inlined) at the start would probably force the compiler to do it
> >> in a way that works.
> >
> > something alongthe lines of
> > sub $1234,%%esp before the call and add afterwards could be used

> > one just must make sure that the call doesnt read via the stack pointer
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

> 
> The local variable is located in rsp-8. If we sub rsp, we lose the
> local variable (through "m"(...)).

you can put it in a register, put it in some place in memory that can
be addressed by another register, whatever

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The real ebay dictionary, page 2
"100% positive feedback" - "All either got their money back or didnt complain"
"Best seller ever, very honest" - "Seller refunded buyer after failed scam"
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 190 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100721/de0de903/attachment.pgp>