I'd like to pull in Dan's upstream libffi commit into trunk to fix a wrong code bug/testsuite failure on powerpc64le-linux with long double defaulting to ieee128. This passed bootstrap and regtesting with no regressions. Ok for trunk?
This bug is also on the GCC 12 and GCC 11 release branches. Ok there too assuming testing is clean? I can wait to push the gcc12 backport until after the release. Peter If there is a homogeneous struct with float128 members, they should be copied to vector register save area. The current code incorrectly copies only the value of the first member, not increasing the pointer with each iteration. Fix this. Merged from upstream libffi commit: 464b4b66e3cf3b5489e730c1466ee1bf825560e0 2023-05-03 Dan HorĂ¡k <d...@danny.cz> libffi/ PR libffi/109447 * src/powerpc/ffi_linux64.c (ffi_prep_args64): Update arg.f128 pointer. diff --git a/libffi/src/powerpc/ffi_linux64.c b/libffi/src/powerpc/ffi_linux64.c index 4d50878e402..3454dacd3d6 100644 --- a/libffi/src/powerpc/ffi_linux64.c +++ b/libffi/src/powerpc/ffi_linux64.c @@ -680,7 +680,7 @@ ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack) { if (vecarg_count < NUM_VEC_ARG_REGISTERS64 && i < nfixedargs) - memcpy (vec_base.f128++, arg.f128, sizeof (float128)); + memcpy (vec_base.f128++, arg.f128++, sizeof (float128)); else memcpy (next_arg.f128, arg.f128++, sizeof (float128)); if (++next_arg.f128 == gpr_end.f128)