Fabian Greffrath <[EMAIL PROTECTED]> writes: > Reinhard Tartler schrieb: >> Hm, that would be a severe regression to what we currently have, >> IMO. I'd take that option only as last resort, but not as obvious way to >> go. However I'm open for arguments why this wasn't that bad... > > Well, we allready do this for ARM. Furthermore we build our packages for > i386 and cannot expect every machine out there to support MMX. > > In many cases these architecture-specific optimizations do not gain too > much performance if compared to the loss of portability they cause.
Okay, if you make this consideration, then I have to say this makes sense. I think we both agree that we don't know the code enough to sensibly support these optimisations, and I also agree that having a working ffmpeg with less optimisation is better than not having it at all. >> Let's better revert your last change for now and rethink our options. > > Why revert the change? The problem is not that we now have found a way > to actually add '-fPIC -DPIC' to CFLAGS, but that some > architecture-specific files fail to build with these flags. I mean, > reverting back to a broken measure to add these flags - which actually > results in the flags not beeing added - is not a fix for the very > problem (was that understandable?). ;) Good Point! :) -- Gruesse/greetings, Reinhard Tartler, KeyID 945348A4 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]