Re: [PATCH] Fix PR 43662 (was Re: [RFC] Fix PR 43562 -- 2nd try (was Re: [RFC] Fix PR 43562))

[Jie Zhang <jie at codesourcery dot com>]

On 04/19/2010 04:51 PM, Richard Guenther wrote:
> On Mon, Apr 19, 2010 at 9:18 AM, Jie Zhang<jie@codesourcery.com>  wrote:
> > On 04/10/2010 11:27 PM, Jie Zhang wrote:
> > > On 04/10/2010 03:40 AM, Jie Zhang wrote:
> > > > On 04/10/2010 12:33 AM, H.J. Lu wrote:
> > > > > > > > > As you suggested, this patch moves call of init_caller_save to IRA
> > > > > > > > > main
> > > > > > > > > function. A flag "caller_save_initialized_p" is used to avoid
> > > > > > > > > duplicate
> > > > > > > > > initializations. When a reinitialization is need, that flag is
> > > > > > > > > set to
> > > > > > > > > false again such that caller-save will be initialized in the next
> > > > > > > > > call
> > > > > > > > > of init_caller_save. Testing is going on. Is it OK if no
> > > > > > > > > regressions?
> > > > > This caused:
> > > > >
> > > > > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43662
> > > > Thanks for reporting this. I'm testing this patch.
> > > No regressions found on bootstrap and testsuite on native
> > > x86_64-pc-linux-gnu.
> > This is an updated patch, which addressed the comments in bugzilla. I added
> >
> > /* { dg-require-effective-target lp64 } */
> >
> > to the test case as suggested by Mikael Pettersson. I also cleaned up the
> > test case a little.
> >
> > Is it OK?
>
> Ok.
Thanks. Committed on trunk, 4.5 and 4.4 branches.


Jie