Bernd Edlinger <bernd.edlin...@hotmail.de> writes: > Hi,<br><br>this patch fixes a regression that was triggered by commit > r225260.<br>See > pr66747 for details.<br><br>Is it OK for > trunk?<br><br><br>Thanks<br>Bernd.<br>
Thanks Bernd. I was just reviewing this PR. I think it will probably be safer to move the fix up to the mips_find_gp_ref and consider each instruction inside a sequence individually at that point. I'm not sure that mips_small_data_pattern_p is safe when given a sequence. It looks like we have never had to compute frame information after delay slot filling Previously, hence a bug in old code. I am happy to rework this to move it into mips_find_gp_ref. Thanks, Matthew
