------- Comment #11 from dominiq at lps dot ens dot fr 2009-11-16 14:29 -------
The following code triggers the problem on i686-apple-darwin9 at revision
154075 (+ patches from fortran-dev):
[ibook-dhum] f90/bug% cat complex-sign-add_red_1.c
extern void exit (int);
void
check_add_float (void)
{
_Complex float a1;
volatile _Complex float a2, b2, c2;
a1 = (+ 1 == 1 ? + 1 == 1 ? 0.0f + 0.0if : 0.0f - 0.0if : + 1 == 1 ? -(0.0f -
0.0if) : -(0.0f + 0.0if));
a2 = (+ 1 == 1 ? + 1 == 1 ? 0.0f + 0.0if : 0.0f - 0.0if : + 1 == 1 ? -(0.0f -
0.0if) : -(0.0f + 0.0if));
b2 = (+ 1 == 1 ? + 1 == 1 ? 0.0f + 0.0if : 0.0f - 0.0if : + 1 == 1 ? -(0.0f -
0.0if) : -(0.0f + 0.0if));
c2 = a2 + b2;
}
int
main (void)
{
check_add_float ();
exit (0);
}
[ibook-dhum] f90/bug% gcc45 complex-sign-add_red_1.c -O1 -g -c
[ibook-dhum] f90/bug% gcc45 complex-sign-add_red_1.o -O1 -g
[ibook-dhum] f90/bug% dsymutil a.out
Assertion failed: (orig_str), function FixReferences, file
/SourceCache/dwarf_utilities/dwarf_utilities-70/source/DWARFdSYM.cpp, line
3641.
Abort
The reduced case does not trigger the error on powerpc-apple-darwin9 while the
full test does.
QUESTIONS: what does gcc 4.5 put in the a.out files that triggers (or not) the
failure?
How am I supposed to find it?
Note that 'gcc45 complex-sign-add_red_1.c -O1 -g' triggers the same error, but
in both cases the executable is produced and can be run (I did not tried to
debug it). Also the dsymutil bug is not triggered at -O0. Finally 'gcc45
complex-sign-add_red_1.c -O1 -g -lm' does not gives any error due to the fact
that dsymutil is not called. The -v option gives respectively:
...
COLLECT_GCC_OPTIONS='-mmacosx-version-min=10.5.8' '-O1' '-g' '-v'
'-mtune=generic'
dsymutil a.out
Assertion failed: (orig_str), function FixReferences, file
/SourceCache/dwarf_utilities/dwarf_utilities-70/source/DWARFdSYM.cpp, line
3641.
gcc45: Internal error: Abort trap (program dsymutil)
Please submit a full bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.
and
...
COLLECT_GCC_OPTIONS='-mmacosx-version-min=10.5.8' '-O1' '-g' '-v'
'-mtune=generic'
Is this the intended behavior?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41473
