------- Additional Comments From uweigand at gcc dot gnu dot org 2005-04-09
18:26 -------
Here's a reduced test case:
extern void abort (void);
extern void **alloc (void);
void *test (void)
{
void **p = alloc ();
if (!p) abort ();
__builtin_set_thread_pointer (p);
return *p;
}
It would appear the __builtin_set_thread_pointer changes
on s390 for GCC 4.0 have uncovered a latent bug in regmove;
this pass now generates an invalid REG_DEAD note.
I don't want to change regmove for 4.0, but I've found a
back-end workaround which I'm currently testing ...
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20917
