Comment # 3 on bug 1141190 from Martin Li��ka

It's caused by a wrong configure with -flto:

w/o LTO:

 checking format of `long double' floating point... possibly double-double,
little endian
 configure: WARNING: This format is known on GCC/PowerPC platforms,
 configure: WARNING: but due to GCC PR26374, we can't test further.
 configure: WARNING: You can safely ignore this warning, though.

with LTO:

 checking format of `long double' floating point... unknown