On 11/7/20 8:41 AM, Iain Sandoe wrote: > Hi > > ** Actually, this was originally posted during last stage-1, but I forgot to > keep > pinging it… > > ==== > > At present, the output of .cfi_personality and .cfi_lsda assumes > ELF semantics for indirections. This isn't suitable for all targets > and is one blocker to moving Darwin to use .cfi_xxxx. > > The patch adds a target hook that allows non-ELF targets to use > indirections appropriate to their needs. > > tested across the Darwin range and on x86_64-linux-gnu, > OK for master? > thanks > Iain > > -- > > gcc/ChangeLog: > > * config/darwin-protos.h (darwin_make_eh_symbol_indirect): New. > * config/darwin.c (darwin_make_eh_symbol_indirect): New. Use > Mach-O semantics for personality and ldsa indirections. > * config/darwin.h (TARGET_ASM_MAKE_EH_SYMBOL_INDIRECT): New. > * doc/tm.texi: Regenerate. > * doc/tm.texi.in: Add TARGET_ASM_MAKE_EH_SYMBOL_INDIRECT hook. > * dwarf2out.c (dwarf2out_do_cfi_startproc): If the target defines > a hook for indirecting personality and ldsa references, use that > otherwise default to ELF semantics. > * target.def (make_eh_symbol_indirect): New target hook.
OK jeff
