[C++ Patch] PR 58633 (Take 2)

[Paolo Carlini <paolo dot carlini at oracle dot com>]

Hi,

this is a completely different approach at fixing the bug, which overall 
I like better.

In this case most of the patch touches cp_parser_decltype_expr: instead 
of using cp_parser_postfix_expression only for member access 
expressions, we accept all its valid return values (still identifying 
the member access expressions, of course, because the semantics of 
decltype is special for those). Thus we can also immediately accept the 
pseudo destructor calls which are the matter of this bug. Otherwise, as 
a last resort, if the expression at issue isn't a postfix expression at 
all, we use cp_parser_expression.

Patch works better than my previous try from the diagnostic point of 
view (eg, for c++/47277 - type illegal expressions as arguments of the 
decltype) and seems less "hackish" to me. Also, it seems to me that it's 
a computational waste to redo the parsing of valid postfix-expressions 
!= member access expression.

Tested x86_64-linux.

Thanks!
Paolo.

/////////////////////////

Attachment: CL_58633_2
Description: Text document

Attachment: patch_58633_2
Description: Text document