This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH RFA: PR 29286: Handle placement new aliasing issues
- From: Michael Matz <matz at suse dot de>
- To: Ian Lance Taylor <iant at google dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 11 Jun 2007 07:52:15 +0200 (CEST)
- Subject: Re: PATCH RFA: PR 29286: Handle placement new aliasing issues
- References: <m3y7itm484.fsf@localhost.localdomain>
Hi,
On Sat, 9 Jun 2007, Ian Lance Taylor wrote:
> After 175 comments on PR 29286, I think we finally have a patch which
> is ready.
Over the course of the bugreport I discussed this with Richard quite
extensively, and I still like the solution where we would introduce
certain rules (comment 105) about reordering stores and loads _much_
better than introducing this tree code and doing fancy but IMHO fragile
things with it, even though you put so much energy into it. Those rules
would also help us becoming more robust when presented with code where
different people have different opinion regardings its validity (e.g. if a
write to a POD-type pointer changes the type of the pointed to memory),
unlike the special casing of only placement new. IMHO they are easier to
understand, and they don't have much impact on runtime as Richard showed.
Ciao,
Michael.