On Fri, Feb 08, 2019 at 04:18:52PM +, Iain Sandoe wrote:
>
> > On 8 Feb 2019, at 16:16, Segher Boessenkool
> > wrote:
> >
> > On Fri, Feb 08, 2019 at 10:19:40PM +1030, Alan Modra wrote:
> >> That one regressed gcc.dg/20020312-2.c, due to my "cleverness" in
> >> simplifying the ABI_V4 case.
> On 8 Feb 2019, at 16:16, Segher Boessenkool
> wrote:
>
> On Fri, Feb 08, 2019 at 10:19:40PM +1030, Alan Modra wrote:
>> That one regressed gcc.dg/20020312-2.c, due to my "cleverness" in
>> simplifying the ABI_V4 case. This one passes regression testing.
>> OK to apply?
>
> I think this is
On Fri, Feb 08, 2019 at 10:19:40PM +1030, Alan Modra wrote:
> That one regressed gcc.dg/20020312-2.c, due to my "cleverness" in
> simplifying the ABI_V4 case. This one passes regression testing.
> OK to apply?
I think this is correct. Thanks! Okay for trunk. Does it need backports?
Segher
On Fri, Feb 08, 2019 at 11:13:50AM +1030, Alan Modra wrote:
> PR target/88343
> * config/rs6000/rs6000.c (save_reg_p): Correct calls_eh_return
> case. Match logic in rs6000_emit_prologue emitting pic_offset_table
> setup, simplifying ABI_V4 case to df_regs_ever_live_p.
Fixes lack of r30 save/restore on
// -m32 -fpic -ftls-model=initial-exec
__thread char* p;
char** f1 (void) { return }
and
// -m32 -fpic -msecure-plt
extern int foo (int);
int f1 (int x) { return foo (x); }
These are both caused by save_reg_p returning false when the pic
offset table reg (r30
