Re: [C++ Patch] PR 58810

2013-11-23 Thread Jason Merrill
On 10/21/2013 02:05 PM, Paolo Carlini wrote: Thus the idea of resolving the inconsistency by removing completely one of the two checks: that in grokfndecl is more precise (separate messages for cv-qualifiers and ref-qualifiers) and seems the best candidate. The patch below passes testing modulo t

Re: [C++ Patch] PR 58810

2013-10-21 Thread Paolo Carlini
... this is an alternate, more conservative approach: a check remains in grokdeclarator and it handles friends too. At variance with the check in grokfndecl it's about function *types*. Thanks, Paolo. /// Index: cp/decl.c

[C++ Patch] PR 58810

2013-10-21 Thread Paolo Carlini
Hi, this is just an ICE on invalid, but I think it hints to a small clean-up. Essentially the issue is that we diagnose qualified free functions in two different places: in the grokdeclarator itself and in grokfndecl. The former handles non-friend declarations and does type = TYPE_MAIN_VARIAN