https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101403
Roger Sayle <roger at nextmovesoftware dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |roger at
nextmovesoftware dot com
--- Comment #3 from Roger Sayle <roger at nextmovesoftware dot com> ---
Testing a fix. Sorry for the inconvenience. An even more reduced test case is:
unsigned int foo (unsigned int a)
{
unsigned int u;
/* b == 0x8000 */
unsigned short b = __builtin_bswap16 (a);
/* result = 0x0008 */
// This works: return b >> 12;
// This doesn't:
return b >> (u, 12);
}
int main (void)
{
unsigned int x = foo (0x80);
if (x != 0x0008)
__builtin_abort ();
return 0;
}
The change in signedness of the right shift (from logical to arithmetic)
triggers an oversight in my recent folding optimization.
