[Bug target/66215] [4.8/4.9/5/6 Regression] Wrong after label NOP emission for -mhotpatch

jakub at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Thu May 21 07:59:00 GMT 2015


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66215

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
No, IMHO you can have many debug insns after that and before first real insn.
I'd go for something like:
  rtx_insn *insn = get_insns ();
  if (!active_insn_p (insn))
    insn = next_active_insn (insn);
and insert before, rather than after (otherwise you don't handle the
hypothetical case of an active insn being the first one).
That would require rewriting the nop insertion code after it, because you want
to insert the 6 byte nops first.  Or just gcc_assert the first insn is not
active, or if the first insn is active, emit a NOTE_INSN_DELETED note before
that
first active insn, emit the nops after that note and perhaps kill the note at
the end.
Please test
void foo (void) { __builtin_unreachable (); }
actually generates any active insns.



More information about the Gcc-bugs mailing list