Hi Folks.
> On 5 Nov 2024, at 19:23, Harald Anlauf <anl...@gmx.de> wrote:
>
> Hi Paul,
>
> Am 05.11.24 um 16:24 schrieb Paul Richard Thomas:
>> Hi All,
>>
>> There is not much to say about the attached patch other than it is minimal
>> :-) The testcases are probably a bit more than is strictly needed since the
>> interface tests (proc_ptr_55.f90) are already tested elsewhere. However, it
>> is as well to check in this context.
>
> the fix is fine, but testcase gfortran.dg/proc_ptr_56.f90 should be
>
> ! { dg-do compile }
>
> instead of
>
> ! { dg-do run }
>
> You'll notice that you get
>
> # of unresolved testcases 6
>
> when running it under the testsuite harness, because
>
> PASS: gfortran.dg/proc_ptr_56.f90 -O0 (test for excess errors)
> UNRESOLVED: gfortran.dg/proc_ptr_56.f90 -O0 compilation failed to
> produce executable
You can most likely add target-specific additional options like “-z execstack”
to
suppress the warning on those platforms affected (and therefore get a test at
O0).
Note that, in the test case in PR117434 the code was completely elided
for O > 0 and therefore that’s also a consideration as to whether execute
testing is useful (at least someone should verify the revised testcase actually
does something)
Iain
>
> the first line is expected, but not the second.
>
>> OK for mainline and 14-branch after a week or two?
>
> OK with the testcase fixed.
>
>> The issue with the executable stack on some platforms should have its own
>> PR to ensure that it has the required visibility. I can make
>> proc_ptr_64.f90 compile-only until it is fixed.
>
> I do not have any preference here. If any of the testers are affected,
> then having it compile-only might be the right choice.
>
>> Regards
>>
>> Paul
>>
>
> Thanks for the patch!
>
> Harald
>
