https://gcc.gnu.org/bugzilla/show_bug.cgi?id=74563
--- Comment #8 from Jeffrey A. Law <law at redhat dot com> ---
So AFAICT this pattern:
(define_insn "*<optab>"
[(any_return)]
""
{
operands[0] = gen_rtx_REG (Pmode, RETURN_ADDR_REGNUM);
return mips_output_jump (operands, 0, -1, false);
}
[(set_attr "type" "jump")
(set_attr "mode" "none")])
Is used for shrink-wrapping -- in particular it may be used while the return
pointer is still sitting in $31.
The _internal pattern:
(define_insn "<optab>_internal"
[(any_return)
(use (match_operand 0 "pmode_register_operand" ""))]
""
{
operands[0] = gen_rtx_REG (Pmode, RETURN_ADDR_REGNUM);
return mips_output_jump (operands, 0, -1, false);
}
[(set_attr "type" "jump")
(set_attr "mode" "none")])
Is used directly by mips.c which should be passing in the right
RETURN_ADDR_REGNUM.
So like you, I think the solution is to just remove the offending line. I'm
going to spin one more test. But I think we're converging on the solution.
