Analysis of PR c++/2015

[Roberto Bagnara <bagnara at cs dot unipr dot it>]

nathan@gcc.gnu.org wrote:
> 
> Synopsis: New ICE in snapshot of February 17
> 
> State-Changed-From-To: open->analyzed
> State-Changed-By: nathan
> State-Changed-When: Fri Apr 27 10:32:55 2001
> State-Changed-Why:
>     confirmed as a bug
> 
> http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2015&database=gcc


I think the following analysis and the much simplified testcase
should help fixing PR #2015 in GNATS.
Consider the following snippet:

//--------------------------------//
static int i = 10;

// namespace N {
  int* p = new int[i];
// };

void foo() {
}
//--------------------------------//

This causes GCC 3.0 to behave as follows

$ /usr/local/beta/bin/g++ gccbug.cc 
gccbug.cc:6: variable-size type declared outside of any function
gccbug.cc:6: variable-size type declared outside of any function

which is fine, apart from the useless repetition.
If we uncomment the namespace enclosure the the variable-size array,
we obtain an ICE.

//--------------------------------//
static int i = 10;

namespace N {
  int* p = new int[i];
};

void foo() {
}
//--------------------------------//

$ /usr/local/beta/bin/g++ gccbug.cc 
gccbug.cc: In function `void __static_initialization_and_destruction_0(int, 
   int)':
gccbug.cc:5: Internal compiler error in find_function_data, at function.c:328
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.

The problem seems to be that some part of the compiler treats
what follows "namespace N" as a function body, and this prevents
the recognition of the erroneous situation.
When later find_function_data() is invoked the compiler realizes that
there is no containing function and reacts to this by calling abort().

Here is some more info about the compiler used and the machine:

$ /usr/local/beta/bin/g++ -v       
Reading specs from /usr/local/beta/lib/gcc-lib/i686-pc-linux-gnu/3.0/specs
Configured with: ../gcc-3.0/configure --prefix=/usr/local/beta
Thread model: single
gcc version 3.0
$ uname -a
Linux zoltan.unisuv.it 2.4.6 #16 Fri Jul 6 18:25:36 CEST 2001 i686 unknown

All the best,

     Roberto Bagnara

-- 
Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara@cs.unipr.it