On Wednesday 17 November 2010 23:30:00 Chris Adams wrote: > Once upon a time, Matthew Garrett <mjg59@xxxxxxxxxxxxx> said: > > On Wed, Nov 17, 2010 at 09:03:02PM -0600, Chris Adams wrote: > > > However, I still think that changing memcpy away from years of "it just > > > works" is an ABI change that should not be taken lightly and IMHO > > > shouldn't be done in a "stable" release of glibc. Is memcpy called > > > often enough (and on large enough blocks) that this change makes a real > > > performance difference (not just on a synthetic memcpy benchmark)? > > > > It's a change in behaviour on x86. Have you verified that overlapping > > memcpy worked on all other glibc-supported architectures until this > > point? > > How is that relevant? If the behavior changes on only some > architectures, then it is okay? Yes, if the behavior was undefined to begin with: http://lists.apple.com/archives/Carbon-dev/2006/Aug/msg00784.html Undefined is undefined, there is no escaping that. Code should not be written to depend on undefined behavior producing some specific results, or even producing the same results on every platform. -- Ben -- Message sent on: Wed Nov 17 23:39:42 EST 2010
Attachment:
signature.asc
Description: This is a digitally signed message part.
-- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
