[tree-ssa] Fix gcc.dg/tree-ssa/20030530-2.c

[Hans-Peter Nilsson]

On Thu, 12 Jun 2003 law@redhat.com wrote:
> You're highly unlikely to catch this with ENABLE_CHECKING.  It is not
> a memory corruption problem (FWIW).  It's just memory layout sensitive
> because of how we hash certain objects (their address).

So don't hash on addresses, then!  (Do you really have to?)
Maybe this is gone, it's been a few days.  In my armchair
position ;-) I'd think it should be in codingconventions.html to
never base hash values on addresses in GCC.  Just like those
unstable sort functions (ahem!), you might get different output
on different hosts.

brgds, H-P