This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Cause of regression 6387
- From: Wolfgang Bangerth <bangerth at ticam dot utexas dot edu>
- To: gcc at gcc dot gnu dot org
- Cc: jakub at redhat dot com, <obrien at FreeBSD dot org>
- Date: Sat, 14 Dec 2002 13:30:23 -0600 (CST)
- Subject: Cause of regression 6387
Just for the record: this patch (which was commited at the same time)
diff -r1.12379 -r1.12380
0a1,11
> 2001-12-18 Jakub Jelinek <jakub@redhat.com>
>
> * config/linux.h (PREFERRED_DEBUGGING_TYPE): Remove.
> * config/sparc/linux.h (PREFERRED_DEBUGGING_TYPE): Remove.
> * config/sparc/linux64.h (PREFERRED_DEBUGGING_TYPE): Remove.
> (CC1_SPEC): Don't default to -gstabs+ if -m32.
>
> 2001-12-05 David O'Brien <obrien@FreeBSD.org>
>
> * elfos.h (PREFERRED_DEBUGGING_TYPE): Change to DWARF2_DEBUG.
is where high priority regression PR 6387 started, which has this
testcase:
-----------------------
void foo();
void bar() {
struct A { A() { foo(); } } a;
}
-----------------------
and yields the following result with -g1 -fPIC:
bash-2.05$ ../bin/gcc-2001-12-25/bin/c++ -fPIC -g1 -c PR6387.cc
PR6387.cc:5: Internal compiler error in dwarf2out_finish, at
dwarf2out.c:11959
Given the log message, it might well be that the problem originated
earlier and a change in meaning of -g1 now exposed the problem. Maybe the
authors of the patch can comment on how the new meaning of -g1 can be
reproduced with compilers that preceded this patch?
Regards
Wolfgang
-------------------------------------------------------------------------
Wolfgang Bangerth email: bangerth@ticam.utexas.edu
www: http://www.ticam.utexas.edu/~bangerth