https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88343
--- Comment #13 from Iain Sandoe <iains at gcc dot gnu.org> --- (In reply to Joseph S. Myers from comment #11) > This change results in miscompilation of glibc for 32-bit soft-float powerpc > (symptoms: many libm tests as run by "make regen-ulps" either segfault, or > produce spurious errors that don't occur with compilers without this change, > or in the case of tests for float fail the "Value outside of 100 +/- 1ulp." > internal sanity check). > > Specifically, on GCC 7 branch r267476 works OK but the tip of the branch, > which has no subsequent changes other than this patch and changes to > DATESTAMP, miscompiles glibc. On GCC 8 branch r267386 (the revision before > this change) works and tip of the branch miscompiles glibc (but I haven't > bisected there to confirm it is indeed this change that's responsible). > Trunk also miscompiles glibc (I haven't tried previous revisions there). Unfortunately, there doesn't appear to be anything in the GCC test suite that catches this (all languages reg-strap was clean, on gcc110). Does it show for a 32b multilib on a 64b host, or only with a 32b host? (if you have a specific reproducer that would be most welcome).