Bug#379497: [sms@2BSD.COM: Re: [Mjpeg-developer] jpeg-xmms build issues with gcc 4.1]
see attached message from upstreams devel list.
----- Forwarded message from "Steven M. Schultz" <sms@2BSD.COM> -----
From: "Steven M. Schultz" <sms@2BSD.COM>
Date: Fri, 28 Jul 2006 19:52:48 -0700 (PDT)
To: Developer mailing list <firstname.lastname@example.org>
Subject: Re: [Mjpeg-developer] jpeg-xmms build issues with gcc 4.1
Reply-To: Developer mailing list <email@example.com>
On Fri, 28 Jul 2006, Michael Ablassmeier wrote:
> jpeg-mmx fails to build from source using gcc 4.1:
That's not surprising - seems each new version of gcc breaks
> I'am no ASM/SSE Expert, so i think it would be best if you guys could
> have a look at the patches provided in the above link and could tell me
> if its safe to include them in the package.
jpeg-mmx was deprecated a while ago. It was problematic in many
ways: it does not build on AMD64 platforms, was causing programs
to crash, is not compatible with libquicktime (you can not use
jpeg-mmx and libquicktime in the same program due to symbol name
clash/conflict), is not usable on PPC systems, there are other
reasons but that should be enough for now.
Support for jpeg-mmx was removed from mjpegtools quite some time
while ago (January 2006 to be exact). Too many problems and the speed
improvement wasn't all that great even when the routines did work.
From the spec file:
" jpeg-mmx is not supported, it crashes on IA32 systems, will not build on
X86_64 or PPC systems and even when it did work didn't provide much of
a speedup (jpeg decoding is a small portion of the overall encoding process)"
My advice would be to ignore jpeg-mmx. if you are using the
cvs version of mjpegtools then ignoring jpeg-mmx will be done
automatically for you (nothing in configure.ac will look for jpeg-mmx).
Mjpeg-developer mailing list
----- End forwarded message -----