Setting SAGE_FAT_BINARY causes internal compiler error building pari

[williamstein]

On all 32-bit linux systems, if I set SAGE_FAT_BINARY in sage-4.2.1.rc0, I get an internal compiler error when building PARI. Since PARI hasn't been upgraded in a while, I think this is the fault of MPIR:

gcc  -c -O1 -Wall -fno-strict-aliasing -fomit-frame-pointer    -I. -I../src/headers -I../src/graph -o
 plotport.o ../src/graph/plotport.c
../src/graph/plotport.c: In function 'rectticks':
../src/graph/plotport.c:469: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-4.3/README.Bugs> for instructions.
make[3]: *** [plotport.o] Error 1
make[3]: Leaving directory `/tmp/wstein/farm/sage-4.2.1.rc0/spkg/build/pari-2.3.3.p5/src/Olinux-i686'
make[2]: *** [gp] Error 2
make[2]: Leaving directory `/tmp/wstein/farm/sage-4.2.1.rc0/spkg/build/pari-2.3.3.p5/src'
Error building GP

real    0m9.951s

Recommendation: we revert to

http://wstein.org/home/wstein/patches/mpir-1.2.p8.spkg

so we can release sage-4.2.1. This is a subtle error and should be reported to the MPIR dev's.

Note though that I can build sage with SAGE_FAT_BINARY="no", then rebuild MPIR with SAGE_FAT_BINARY="yes", and that seems to get around the compiler error. But I haven't tried running the full test suite, and the fact that upgrading or rebuilding PARI would then result in an internal compiler error is not encouraging.

Another alternative would be to use GMP instead of MPIR for our general distribution version. It might work better in this regard....

Component: build

Issue created by migration from https://trac.sagemath.org/ticket/7451

[mwhansen]

comment:2

This should have been fixed by #7471