Comment # 8 on bug 1111657 from Richard Biener

Sorry, I do not have an account on the upstream cffi page and do not want to
create one just to answer.  The result is that on tumbleweed the test program
behaves as expected:

> gcc t.c -lffi -lm -m32 -O2
> ./a.out
expected  = 0.334238
ffiresult = 0.334238

this means this is either a cffi issue or a glibc one.  From the build log
it appears that the ffi call results in NaN instead of the expected number.

Andreas - can you check the ffi test program with the updated glibc on i586?