[MPlayer-dev-eng] [PATCH] gcc-4.0 support

Guillaume POIRIER guillaume.poirier at etudiant.univ-rennes1.fr
Sat Jan 22 11:39:04 CET 2005


Le vendredi 21 janvier 2005 à 20:56 -0500, D Richard Felker III a
écrit :
> On Fri, Jan 21, 2005 at 10:53:21PM +0100, Guillaume POIRIER wrote:
> > Hi,
> > Here's a patch that allows the compilation with gcc-4.0.0 of libaf/af.c.
> > Tested also with gcc-2.95.4, 3.3.5, 3.4.4 on i386. They all compile
> > fine.
> > 
> > Please comment on that patch.
> > 
> > It's very simple, and this is the reason why I'm worried I maybe
> > underestimated the problem.
> > The issue addressed is just that it's obviously syntacticly incorrect to
> > use the "&" operand to take the address of a variable that is declared
> > as "register". 
> > Now I guess there must be a good reason that the code was written that
> > way, but I don't see it...
> No, the reason it was written that way is that Anders is an idiot.
> Actually the register keyword should almost never be used, certainly
> not here. We're shipping 100000l of cola his way; in the mean time
> just..
> > No commit without any go-ahead.
> ..commit. :)

Thanks a lot. Committed.

I'd like to call for volunteers who'd have the brainpower and competence
(I unfortunately lack ASM knowledge here) to fix the following files:

* libmpcodecs/native/RTjpegN.c (lots of assembler error messages: suffix
or operands invalid for pretty much all calls of `psubw', `punpcklwd',
`movq', and `paddw'. All other gcc version compile with no warning at

* postproc/swscale_template.c (memory inputs not directly addressable.
Fixes for thoses files might look like those for libavcodec that have
been flying around ffmpeg's mailing list)

* liba52/imdct.c (memory inputs not direcly addressable. Same as above)

... and that's it!

You you need gcc's output, either refer to my previous message or I
guess I can re-post them if needed.

Patches are thus welcome. I'd gladly test them.


More information about the MPlayer-dev-eng mailing list