This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: RFA: Fix c/57821
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Joern Rennecke <joern dot rennecke at embecosm dot com>
- Cc: <gcc-patches at gcc dot gnu dot org>, Richard Biener <rguenther at suse dot de>
- Date: Thu, 4 Jul 2013 20:33:15 +0000
- Subject: Re: RFA: Fix c/57821
- References: <20130704150020 dot lv606jwfogcckk0c-nzlynne at webmail dot spamcop dot net> <Pine dot LNX dot 4 dot 64 dot 1307041919360 dot 15783 at digraph dot polyomino dot org dot uk> <20130704162842 dot 3huedi1hk4c8o0cg-nzlynne at webmail dot spamcop dot net>
On Thu, 4 Jul 2013, Joern Rennecke wrote:
> Quoting "Joseph S. Myers" <joseph@codesourcery.com>:
>
> > On Thu, 4 Jul 2013, Joern Rennecke wrote:
> >
> > > For this fix I've assumed that it is now the job of the various places
> > > that
> > > do calculations with sizetypes to flag overflows where this is desired.
> >
> > Integer constants may be shared, I don't think setting TREE_OVERFLOW on
> > existing constants like you do is necessarily safe.
>
> Hmm, indeed. I now added a copy_node call in set_init_index to avoid this
> problem. For complete_array_type, we need this information just a bit
> further thon in the function, so I just made it a boolean flag.
>
> > Also note c/ and c-family/ have their own ChangeLog files.
>
> Oops. I though I checked that. Foiled by ls localisation.
>
> I'm currently testing the attached updated patch.
This is OK if it passes testing.
--
Joseph S. Myers
joseph@codesourcery.com