Re: c++/9384: [3.4 Regression] ICE in validate_nonmember_using_decl, at cp/decl2.c:4152

[Volker Reichelt <reichelt at igpm dot rwth-aachen dot de>]

The ICEs in the 3.3 branch and mainline are completely unrelated
as the follwoing reduced testcases show.

The bug on mainline can be triggered with the following line of code:

----------------------snip here----------------------
namespace A { using ::B; }
----------------------snip here----------------------

Compiling this with mainline I get

PR9384A.cc:1: internal compiler error: in validate_nonmember_using_decl, at 
   cp/decl2.c:4152

This is a regression on mainline.


The bug on the 3.3 branch can be triggered with the following testcase:

----------------------snip here--------------------------------------
template <typename T> struct A
{
    A(int);
};

template <typename T> A<T>::A(int i __attribute__ ((__unused__)) ) {}
----------------------snip here--------------------------------------

The code causes a compiler crash since gcc 3.0 (which is changed to a
"confused by earlier errors, ..." message in release versions):

PR9384B.cc:6: error: parse error before `__attribute__'
PR9384B.cc:6: internal compiler error: in poplevel, at cp/decl.c:1283

(gcc 2.95.3 rejects the code and mainline accepts it).

I don't know where you have to place the attribute, but placing it in
front of "int i" makes the code compile, writing just "int" without "i"
and attribute "__unused__" should have the same effect, right?

Given that this is just an error recovery problem and that there's an
easy workaround, I don't think that this will be fixed on 3.3 branch.

Regards,
Volker

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9384