Hello! Emerging the latest unstable ebuild (1.0_pre1) went fine, no errors. Mplayer itself bails out with an illegal instruction. After checking the ebuild and the Makefiles I found the CFLAGS -march=athlon-4 -mcpu=athlon-4. I'm using an Athlon XP. In one of my former ebuilds I wrote a function that greps/cuts the -march and -mcpu settings out of /etc/make.conf. I used this to pach mplayers config.mak. It compiled with the settings I defined in /etc/make.conf (-march=i686 -mcpu=athlon-xp). That binaries does not produce illegal instructions. So I consider using athlon-4 a bug.
Created attachment 17792 [details] Tarball of maplayer compile logs, config.maks and the illegal instruction-error. This is just logs, if anyone is interested... I used them to grep athlon-4 and so on from the compile output.
Created attachment 17793 [details] Ebuild that fixes compile.mak (if athlon-4 was detected) Please test this one. It worked fine on my machine as well as on a pentium2 laptop.
marc did you send this upstream ?
gentoo doesn't use CFLAGS any more for mplayer
Ok, maybe I didn't make myself clear here. Marty, I know that we don't use ouzr CFLAGS from make.conf in mplayer ebuilds. That exactly is the problem (at least that's what I found out on my machine). Mplayer's ./configure will detect my athlon-xp and uses -march=athlon4 which is/was known to cause problems. Dunno if it still does or if athlon4 is supportet in gcc3.2. My changes to the ebuild just changed the mcpu/march settings. If athlon4 is not a problem for athlon-xp users that's just fine. I know why we don't use our CFLAGS settings and that's fine for me too. The only thing I want to know is if anyone has had the same problems with mplayer that I had. My machine was so borked (the famous "athlon-xp not knowing nothing about sse" - problem), that it is very likely the whole affair is my fault and has nothing to do with mplayer at all ;)