Re: [C++ Patch] PR 53491

2012-05-26 Thread Jason Merrill
On 05/26/2012 11:31 AM, Paolo Carlini wrote: Ok. Like this it works, if I understand your suggestion. Yep, that's what I had in mind. But let's put it after the !TREE_SIDE_EFFECTS case. OK with that change. Jason

Re: [C++ Patch] PR 53491

2012-05-26 Thread Paolo Carlini
On 05/26/2012 04:21 PM, Jason Merrill wrote: I think I would rather fix stabilize_expr to handle void arguments properly: basically just stick the whole argument in *initp and return void_zero_node. Ok. Like this it works, if I understand your suggestion. Thanks, Paolo. /

Re: [C++ Patch] PR 53491

2012-05-26 Thread Jason Merrill
I think I would rather fix stabilize_expr to handle void arguments properly: basically just stick the whole argument in *initp and return void_zero_node. Jason

[C++ Patch] PR 53491

2012-05-25 Thread Paolo Carlini
Hi, an ICE on invalid, regression in 4.7 and 4.8. In 4.7 (vs 4.6) we started using stabilize_expr in one more place in cp_build_modify_expr without also adding a preliminary check that the expr isn't of void type, thus we can easily end up calling build_type_target_expr_with_type on it and cr