Changing a SET that has a REG_CFA_REGISTER note is wrong if we are changing the SET_DEST, or if the REG_CFA_REGISTER has nil as its argument, and maybe some other cases. It's never really useful to propagate into such an instruction, so let's just bail whenever we see such a note.
Bootstrapped and tested on powerpc64-linux {-m32,-m64}. Is this okay for trunk? Segher 2018-05-08 Segher Boessenkool <seg...@kernel.crashing.org> PR rtl-optimization/85645 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an insn that has a REG_CFA_REGISTER note. --- gcc/regcprop.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/gcc/regcprop.c b/gcc/regcprop.c index a664f76..1813242 100644 --- a/gcc/regcprop.c +++ b/gcc/regcprop.c @@ -848,6 +848,12 @@ copyprop_hardreg_forward_1 (basic_block bb, struct value_data *vd) && reg_overlap_mentioned_p (XEXP (link, 0), SET_SRC (set))) set = NULL; } + + /* We need to keep CFI info correct, and the same on all paths, + so we cannot normally replace the registers REG_CFA_REGISTER + refers to. Bail. */ + if (REG_NOTE_KIND (link) == REG_CFA_REGISTER) + goto did_replacement; } /* Special-case plain move instructions, since we may well -- 1.8.3.1