(Closes #483) Recognize generic directives

[LonelyCat124]

No description provided.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.19%. Comparing base (eee1f4f) to head (37db93d).
⚠️ Report is 9 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #486   +/-   ##
=======================================
  Coverage   92.19%   92.19%           
=======================================
  Files          86       86           
  Lines       13802    13802           
=======================================
  Hits        12725    12725           
  Misses       1077     1077           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[LonelyCat124]

@sergisiso This is ready for a look - small change to support generic directives as discussed yesterday.

[sergisiso]

Thanks @LonelyCat124 see inline mostly questions and request for test regarding this.

[LonelyCat124]

@sergisiso Back to you, should be fixed up.

[LonelyCat124]

We still get a bit weird locality here:

Block_Nonlabel_Do_Construct(Comment(''), Directive('!$acc comment3'), Comment('!comment3'), Nonlabel_Do_Stmt('DO', Loop_Control(None, (Name('i'), [Int_Literal_Constant('1', None), Name('n'), Int_Literal_Constant('1', None)]), None, None))

comes from

!$acc comment 3
!comment3
do i =1, n, 1

This makes life a little tricky in PSyclone. Also the first one after the declarations is hard to find.

[sergisiso]

@LonelyCat124 This is better but I still have some comments about the tests.

We still get a bit weird locality here

I think that's alright for now. I think we should focus first in that "keep_comments" and "keep_directives" work, and we can improve the specifics latter.

[LonelyCat124]

@sergisiso I think I've tidied up the tests now. The remaining question is whether we do any further work on multi-line directives, but I think that might be difficult for me to do without really diving deep into fparser (and maybe require some significant rewriting of how we handle these now, making comment/directives very different). Maybe its best to just add notes into the documentation for now?

[sergisiso]

@LonelyCat124 I am still confused about the behaviour, it would be good to complete stfc/PSyclone#3196 so we can see the whole thing.

[LonelyCat124]

@sergisiso I added the additional tests here. I didn't add a fixed-form acc test though I can add it if you think its worth testing separately.
I think the stfc/PSyclone#3196 PR probably tests the functionality changes that come from this with fparser. In fact the only remaining issue I remember with this is what to do with directives that appear within the declaration list in the Fortran source, which I have no good solution for (since the current behaviour leaves them as comments, and then PSyclone adds a space into ! dir$, but CodeBlocks can't go into the corresponding part of the tree. I'll try to find the xisting tests that test the ignore_directives=True behaviour from previous work and confirm that directives don't appear as comments with this branch added, but I'm fairly sure those already would have been in fparser2 frontend + comment tests.

[sergisiso]

@LonelyCat124 Ok, I am satisfied with the fparser part of directives, the remaining part is for psyclone to fix. This is approved for merging.