Re: [PATCH] Character length cleanup for Coarray Fortran library

[Janne Blomqvist <blomqvist.janne@gmail.com>]

On Sat, Feb 10, 2018 at 9:25 PM, Zaak Beekman <zbeekman@gmail.com> wrote:
>>
>> Janne, I've scanned the patch and it appears to a
>> fairly straight foward mechanical s/int/size_t
>> (and similar for nodes).  OK to commit (although
>> you may want to wait for a response from Damain
>> or Izaak Beekman).
>>
>> As this breaks an ABI and we've broken it for 8.0,
>> I think this should go into trunk.  You may need
>> to ping Jakub or Richard for final approval.
>>
>> --
>> Steve
>>
>
> Hi Janne & Steve,
>
> I'm still getting used to the GFortran mailing list, and trying to get up
> to speed with your dev practices, so please bear with me.
>
> Also, let me add the caveat that, to date, I have mostly dealt with the
> OpenCoarrays build system and various project
> management/documentation/QA/CI and other administrivia and devops related
> stuff---I am not *super* intimately familiar with the actual code base and
> ABI. I am even less familiar with the GFortran inner workings. Alessandro
> Fanfarillo and Andre are certainly more knowledgeable than I am in this
> area.
>
> Is my understanding correct that I can just look at changes to caf_single.c
> and the header files to determine where the type changes will impact
> OpenCoarrays? A concise list of ABIs/APIs that need changes on the
> OpenCoarrays side would go a long way in helping us/me make the
> corresponding changes on the OpenCoarrays side. Also, this patch will only
> be merged into trunk, right? I.e., it will impact GFortran 8+, and not
> future releases on the 6 and 7 branches?

AFAICT, the change to libgfortran/caf/libcaf.h tells you what would be
necessary to change on the OpenCoarrays side. And yes, the idea would
be to push this change only to the 8 branch.

My problem is a bit the opposite of yours, namely that apart from
skimming through the github repo I have zero experience of
opencoarrays.

A related question (, not directed to you specifically), if anybody
happens to know: Is libgfortran/caf/mpi.c actually used by anything,
or is it some pre-OpenCoarrays attempt at creating an MPI runtime for
coarrays? In other words, can we remove it?




-- 
Janne Blomqvist