[RFA] Patch PR debug/41065

Jakub Jelinek jakub@redhat.com
Mon Aug 31 19:16:00 GMT 2009


On Mon, Aug 31, 2009 at 09:56:06AM -0700, Richard Henderson wrote:
> Please don't use pointer sets for anything that you need
> to traverse and produce assembly output from.  That way
> leads to code generation differences depending on exact
> pointer values.  If address space randomization is enabled,
> that means that consecutive compiler invocations do not
> produce identical output, which is a Bad Thing.

The same is true for htab_t's which hash on htab_hash_pointer.
I think the traversal isn't used to output anything, just to make
the DIEs die_perennial_p.  Anyway, Dodji came up with a reason why htab_t
should be used, if we don't have a guarantee the pointer in the htab won't
be the only reference to a tree, we need a htab_t because of GC.

	Jakub



More information about the Gcc-patches mailing list