Bugzilla Bug 41021

While trying to emerge KDE 3.2, which at the time was masked so I hade to use
ACCEPT_KEYWORDS="~x86", one of the dependencies "mpg123-0.59s-r1" failed to
compile.

The output told me to file a bug report so that's what I'm doing.

If I have 


Reproducible: Always
Steps to Reproduce:
1.ACCEPT_KEYWORDS="~x86" emerge mpg123 or ACCEPT_KEYWORDS="~x86" emerge kde
2.
3.

Actual Results:  
tabinit.c: In function `make_decode_tables':
tabinit.c:83: internal compiler error: Floating point exception
Please submit a full bug report,
[...]
with preprocessed source if appropriate.
!!! ERROR: media-sound/mpg123-0.59s-r1 failed.
!!! Function src_compile, Line 56, Exitcode 2
!!! (no error message)



Expected Results:  
It should've installed :)

kernel: development-sources 2.6.1
gcc version 3.3.2 20031218 (Gentoo Linux 3.3.2-r5, propolice-3.3-7)

Created an attachment (id=25296) [edit]
Preprocessed source

Can you please provide the output of 'emerge info'.  It worked fine for me...

Can you please try again?  I'm also using gcc-3.3.2-r5.

Created an attachment (id=25351) [edit]
Output from emerge info

Can you please try again.  Your compiler itself borked while compileing due to
a floating point exception... Are you overclocking your system perchance? 
Getting random lockups/crashes?

Tried to re-emerge mpg123-0.59s-r1 but it failed in the exact same way as
before.

>> Are you overclocking your system perchance?  Getting random lockups/crashes?

Yes, my system is overclocked (CPU 1.6@2.24 GHz /w 1:1 FSB:RAM), but I've never
run into any stabilityproblems with it.
I emerged KDE 3.2 last night and it installed just fine.
To be able to emerge KDE I just emerged the standard ebuild (non-~x86) of
mpg123, and thus didn't need mpg123-0.59s-r1.

I should mention that I'm running Gentoo as a guest on VMWARE 4.0 running WinXP
as a host. That might be the source of the problem even though I think it's
rather unlikely.

Try to clock your system back at its real speed, re-emerge gcc, then re-emerge
mpg123

I'm having the same problem with the same output.  I'm running on a P3-500 Dell
Latitude, and I'm not overclocked.  I also have ACCEPT_KEYWORDS="~x86" and my
version of gcc is 3.3.2-r7.

alright... bouncing to the gcc guys as this seems to be a problem with gcc...

it looks like r2 has the same problem i have to use an older version for it to
work 

-r2 was just a security fix

Works fine here with the same CFLAGS as Torbj

Works fine here with the same CFLAGS as Torbjörn (and same gcc of course).

Without "-fprefetch-loop-arrays" in the CFLAGS it works for me. With the option
it fails.

CFLAGS working: "-O3 -march=pentium3"

Acer Travelmate 800; 1,3 GHz Centrino; gcc 3.3.2-r5; not overclocked

Same with -r2 here. Usually I use:

CFLAGS="-march=athlon-xp -m3dnow -msse -mfpmath=sse -mmmx -O3 -pipe -fforce-addr -fomit-frame-pointer -funroll-loops -frerun-cse-after-loop -frerun-loop-opt -falign-functions=4 -maccumulate-outgoing-args -ffast-math -fprefetch-loop-arrays"

*** Bug 41773 has been marked as a duplicate of this bug. ***

It compiled for me without "-fprefetch-loop-arrays."  The command I used was:  
CFLAGS="-O3 -march=pentium3 -funroll-loops -pipe" emerge mpg123

Created an attachment (id=25807) [edit]
this patch removes the deprecated -m486 compile-options from Makefile for the
linux-486- and linux-mmx-targets

The internal compiler error is caused when having -fprefetch-loop-arrays in
your CFLAGS in combination with the (deprecated) -m486-compiler option in the
Makefile of mpg123. If your architecture is x86 or you have the mmx-USE-Flag
set, ebuild either issues a "make linux-486" or "make linux-mmx"... In both
cases, the Makefile adds -m486 which causes the compiler to break. After
removing this from the Makefile, everything builds fine.

I'm having a similar problem I think the problem is almost fixed by that patch
but goes a little farther... here's a snapshot.

gcc -march=pentium4 -O3 -pipe -fomit-frame-pointer -fexpensive-optimizations
-ffast-math -falign-functions=4 -fforce-addr -funroll-loops
-maccumulate-outgoing-args -frerun-loop-opt -fprefetch-loop-arrays
-frerun-cse-after-loop -mfpmath=sse -DI386_ASSEM -DREAL_IS_FLOAT -DI486_OPT
-DLINUX               -DOSS -DTERM_CONTROL            -Wall -O2 -m486          
      -fomit-frame-pointer -funroll-all-loops                
-finline-functions -ffast-math   -c -o decode_i486.o decode_i486.c
`-m486' is deprecated. Use `-march=i486' or `-mcpu=i486' instead.
decode_i486.c: In function `synth_1to1_486':
decode_i486.c:249: internal compiler error: Floating point exception


As you can see -O2 -m486 -fomit-frame-pointer -funroll-all-loops
-finline-functions -ffast-math are all included from the Makefile. However I
have those as well in my CFLAGS and they can be seen earlier. I however have
-O3 and the Makefile has -O2 which is causing an interesting conflict.

However the Makefile patch submited by Thomas Veith does fix my problem so
maybe that is enough.

It only adds it for USE=-mmx only.  Anyhow, fixed in CVS, thanks.

*** Bug 42229 has been marked as a duplicate of this bug. ***