this is a critical wrong-codegen bug in gcc 9 and 10 that can have
security impact, and oddly the fix still does not seem to have been
backported, so just disable builtin memcmp expansion entirely until
there's a known-good fix.
the 60840 fpu and later, and coldfire fpus, have precision-specific
sqrt instructions that get used and do the right thing. only when
targeting baseline (or near-baseline) ISA does the ld80-only
instruction get used, and that's all that needs to be suppressed.
gcc pr 95921. the m68k sqrt rtl handles excess precision wrong. I
couldn't figure out how to make gcc generate the insn only for long
double, but since long double usage is rare anyway, just remove it.
this can be replaced with a proper fix later if there's ever one
upstream.
previously this was missed because there's target-specific spec magic
for a few targets to do it, but many including arm don't have it.
without this, use of -static-pie produces broken binaries which lack
an interpreter but link to shared libraries on the targets affected.
use of -static -pie (preferred form) is not affected. gcc versions
prior to 8.x did not have -static-pie (because there was no upstream
static pie support) so this patch is only needed for 8+, and only to
prevent breakage if someone uses the -static-pie option instead of
-static with -pie.
this is binutils issue 23825, but it's caused by gcc using local-exec
model rather than initial-exec model with the intent of making
binutils generate copy relocations. this is harmful, unnecessary, and
not presently supported by musl (and probably should never be). patch
taken from https://github.com/riscv/riscv-gcc/pull/118.
Richard van Dijk
Rich Felker