Bugzilla Bug 26239

the CFLAGS stip needs to be uncommented to build on a K6-2, possibly others,
build will fail if that -march isnt stripped

Whats the error message for this error?  I think its biting me.  Is it just
certain compilers?  -- ekl@crc.homelinux.org

Haven't had time to dissect this error completely, all I know so far is
stripping the -march lets the build succeed. Here's the error-

/bin/sh ../libtool --mode=compile gcc -DPACKAGE=\"libvorbis\" -DVERSION=      
\"1.0\" -DHAVE_DLFCN_H=1 -DHAVE_ALLOCA_H=1 -DHAVE_ALLOCA=1 -DHAVE_SQRTF=      
1 -DHAVE_LOGF=1 -DHAVE_EXPF=1 -DHAVE_ACOSF=1 -DHAVE_ATANF=1 -DHAVE_FREXP      
F=1 -DHAVE_RINTF=1 -DHAVE_FLOORF=1  -I. -I.  -I../include -I/usr/include       
   -O20 -ffast-math  -D_REENTRANT -fsigned-char -O2 -march=k6-2 -pipe -      
fomit-frame-pointer -DUSE_MEMORY_H -c envelope.c
make[2]: *** [smallft.lo] Error 1
make[2]: *** Waiting for unfinished jobs....
rm -f .libs/envelope.lo
gcc -DPACKAGE=\"libvorbis\" -DVERSION=\"1.0\" -DHAVE_DLFCN_H=1 -DHAVE_AL      
LOCA_H=1 -DHAVE_ALLOCA=1 -DHAVE_SQRTF=1 -DHAVE_LOGF=1 -DHAVE_EXPF=1 -DHA      
VE_ACOSF=1 -DHAVE_ATANF=1 -DHAVE_FREXPF=1 -DHAVE_RINTF=1 -DHAVE_FLOORF=1       
-I. -I. -I../include -I/usr/include -O20 -ffast-math -D_REENTRANT -fsig      
ned-char -O2 -march=k6-2 -pipe -fomit-frame-pointer -DUSE_MEMORY_H -c en      
velope.c  -fPIC -DPIC -o .libs/envelope.lo
gcc -DPACKAGE=\"libvorbis\" -DVERSION=\"1.0\" -DHAVE_DLFCN_H=1 -DHAVE_AL      
LOCA_H=1 -DHAVE_ALLOCA=1 -DHAVE_SQRTF=1 -DHAVE_LOGF=1 -DHAVE_EXPF=1 -DHA      
VE_ACOSF=1 -DHAVE_ATANF=1 -DHAVE_FREXPF=1 -DHAVE_RINTF=1 -DHAVE_FLOORF=1       
-I. -I. -I../include -I/usr/include -O20 -ffast-math -D_REENTRANT -fsig      
ned-char -O2 -march=k6-2 -pipe -fomit-frame-pointer -DUSE_MEMORY_H -c en      
velope.c -o envelope.o >/dev/null 2>&1
mv -f .libs/envelope.lo envelope.lo
make[2]: Leaving directory `/var/tmp/portage/libvorbis-1.0-r3/work/libvo      
rbis-1.0/lib'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/var/tmp/portage/libvorbis-1.0-r3/work/libvo      
rbis-1.0/lib'
make: *** [all-recursive] Error 1

!!! ERROR: media-libs/libvorbis-1.0-r3 failed.
!!! Function src_compile, Line 43, Exitcode 2
!!! (no error message)

I think its the -fpic -DPIC with the -march on mine, but like I said,havent
really looked into the files yet.

Usually k6, k6-2, k6-3 have issues with -fomit-frame-pointer. I'm going to
commit an r4 that filters -fo-f-p, could you please test and let me know?

Thanks

Please test -r4 and let me know. Thanks.

Well, it'll build stripping the -f-o-p also, but I can say that after a couple
years and 200-something currently installed packages using that combo, that's
the first time stripping -f-o-p helped anything. Like I mentioned previously,
the -fpic -DPIC clashes more with the -march, especially when coupled in with
the predetermined -ffast-math supplied with the source. The k6 doesn't like any
-f options really,but that's proved stable for me for a while. I suggested the
stripping the -march because I figured that no -march + no frame pointers would
likely be smaller and faster  than -march + frame pointers. But in the end I
guess I really dont care,as I dont need the support and have been too lazy to
take it out, I really only bothered cuz it choked my emerge run for that night.
Thanks for the fix, and nice job with the 3dnow patch.

The last post in forum thread http://forums.gentoo.org/viewtopic.php?t=72949
seems to indicate that -mcpu=k6-3, -mcpu=k6-2, -mcpu=k6 causes libvorbis-1.0-r4
not to compile. I have a P4 processor so can't test this myself.

Looks like it needs to strip -fomit-frame-pointer if mcpu=k6... too tired to
deal with this now. Back tomorrow when sober.

If you need any help with this, let me know, since I was the original poster
which added the SIMD patch.

Thanks Chris. Been running -r4 no probs a while now...suprised this wasnt
closed yet.

closing with -r4