get_d_2exp test failures: clang -emit-llvm/-flto/-O4; icc -ipo/-fast

bathtubdev bathtubdev at gmail.com
Wed Aug 21 01:00:58 CEST 2013


> Somehow dropping all optimization flags might be a
> working approach. Ideally, one should keep ABI-changing flags in a
> variable different from CFLAGS (possibly in CC), and then one could drop
> all of CFLAGS for this test. Listing known optimization flags in a sed
> command seems a bit brittle.

It is brittle. But what you're suggesting is basically the same but
with a whitelist instead
of a blacklist. I can get pretty ugly: Darwin is probably the worst
case scenario, where one
might see something like CC='arch -i386 gcc' CFLAGS='-arch x86_64 -m32
-march=core2'
and have to parse that correctly.

G.


More information about the gmp-bugs mailing list