This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Solaris bootstrap failure
- To: rth at redhat dot com (Richard Henderson)
- Subject: Re: Solaris bootstrap failure
- From: Joern Rennecke <amylaar at cambridge dot redhat dot com>
- Date: Mon, 12 Mar 2001 21:49:10 +0000 (GMT)
- Cc: mrs at windriver dot com (Mike Stump), bernds at cambridge dot redhat dot com,chowell at redhat dot com, gcc at gcc dot gnu dot org
>
> On Thu, Mar 08, 2001 at 11:54:25AM -0800, Mike Stump wrote:
> > It should be beyond rare that this overflows to 0 exactly, but if it
> > does, blam.
>
> I wouldn't think it'd go "blam". As long as objects that ought to
> compare equal hash to the same value, everything works. I think
> what you're describing is unnecessary collisions.
>
> What we're looking for is either (1) two objects that compare equal
> but _don't_ hash to the same value, or (2) two objects that only
> sometimes compare equal, or (3) miscompiled hash table routines that
> somehow fail to locate a value despite correctness of the hash and
> compare functions.
Or some argorithm that is looking for one of several things, and uses
the first thing it finds in the hash table - and is thus dependent on the
order of things in the hash table.