Re: [Patch, fortran] PR83076 - [8 Regression] ICE in gfc_deallocate_scalar_with_status, at fortran/trans.c:1598

[Damian Rouson <damian at sourceryinstitute dot org>]

Hi Paul,

Does breaking binary compatibility simply mean that CAF codes will need to be recompiled (which is fine) or does it mean that there will need to be work done on OpenCoarrays to support the changes in this patch (which is unlikely to happen soon without new contributors to OpenCoarrays)? If it’s the latter, could you make the appropriate changes to OpenCoarrays before committing this? We’re still dealing with some regressions that Andre introduced a year ago through changes in gfortran that have never been reflected in OpenCoarrays. In fact, just within the past week, the number of such regressions was finally reduced from 3 to 2 for the first time in many months. I’m hopeful that we can continue vectoring toward 0 rather than increasing the number. It’s already the case that several projects won’t even beginning evaluating CAF for production use until our 2 current regressions get fixed and there’s one project that is still stuck using GCC 6 because of the current regressions. We actually had to publish results using GCC 6 for that reason (see https://github.com/sourceryinstitute/coarray-icar-paw17/blob/master/main.pdf).  

Damian  

On December 27, 2017 at 10:04:13 AM, Thomas Koenig (tkoenig@netcologne.de(mailto:tkoenig@netcologne.de)) wrote:

> Hi Paul,
>  
> by the way, the patch is OK for trunk. It is just gcc-7 that I am
> worried about.
>  
> Regards
>  
> Thomas