[flang] Add a semantics test for co_sum

[rouson]

Summary

This patch checks for valid and invalid forms of calls to the collective subroutine co_sum.

[kiranchandramohan]

flang/test/Semantics/collectives01.f90

1

Thanks @rouson for these tests. Recently we switched to using python (test_errors.py) instead of shell scripts (test_errors.sh) so that these tests run on Windows as well.

Please see some of the other tests for an example. (https://github.com/llvm/llvm-project/blob/main/flang/test/Semantics/critical01.f90)

Please switch to using that otherwise, the tests will fail.

[ktras]

I have locally implemented co_sum and was able to get this test to pass if these two error messages are changed, and if the python program is added, as suggested by @kiranchandramohan. Is there a way to make this test expectedly fail to get it into the repository? And then I can submit a patch with my implementation of co_sum in flang/lib/Evaluate/intrinsics.cpp?

flang/test/Semantics/collectives01.f90

36

The expected error will be too many actual arguments for intrinsic 'co_sum'

[ktras]

flang/test/Semantics/collectives01.f90

32

Please change the ’ marks to ' marks.

[kiranchandramohan]

In D113076#3107363, @ktras wrote:

I have locally implemented co_sum and was able to get this test to pass if these two error messages are changed, and if the python program is added, as suggested by @kiranchandramohan. Is there a way to make this test expectedly fail to get it into the repository? And then I can submit a patch with my implementation of co_sum in flang/lib/Evaluate/intrinsics.cpp?

You can use XFAIL to mark the test as expected to fail.
See https://github.com/llvm/llvm-project/blob/main/flang/test/Semantics/omp-sections03.f90 for an example.
https://llvm.org/docs/TestingGuide.html#constraining-test-execution

[ktras]

In D113076#3108667, @kiranchandramohan wrote:

You can use XFAIL to mark the test as expected to fail.
See https://github.com/llvm/llvm-project/blob/main/flang/test/Semantics/omp-sections03.f90 for an example.
https://llvm.org/docs/TestingGuide.html#constraining-test-execution

Thank you @kiranchandramohan. I was trying to find that exact info, thanks for helping point me in the right direction!

[rouson]

1. Add co_sum test to expectedly failed list.
2. Replaced non-ASCII characters.
3. Correct expected error message.
4. Update run directive to use test_errors.py.

[rouson]

Made update based on comments.

[ktras]

Thanks for making the changes. LGTM.

[kiranchandramohan]

LGTM. Thanks for this testcase.

[kiranchandramohan]

@rouson @ktras Do you need help to submit this patch?

[ktras]

In D113076#3121738, @kiranchandramohan wrote:

@rouson @ktras Do you need help to submit this patch?

Yes, I think we do. I don't have commit privileges and I don't think @rouson does either.

[PeteSteinfeld]

I think it's best for you to do the commit yourself. Here’s how.

The first step is to get committer access to the llvm-project repository. You can request commit access for the llvm-project here: https://llvm.org/docs/DeveloperPolicy.html#obtaining-commit-access. Once access is granted, an invitation should be visible here: https://github.com/llvm.

Once you have committer access (actually, you can do this now), you should incorporate your changes into the latest source code from llvm-project, and make sure that everything builds and tests correctly. Here’s how to do that:

Start in your private repository in the branch that contains your changes.
If you have multiple commits, run git rebase -i to squash them into a single commit.
Merge the latest changes from llvm-project into your branch:

• git checkout main
• git pull
• git checkout mybranch
• git rebase main

Rebuild and retest to verify that your changes still work.
Push your changes to the main branch in the llvm-repository: git push origin mybranch:main

flang/test/Semantics/collectives01.f90

+! RUN: %S/test_errors.sh %s %t %flang_fc1
+! REQUIRES: shell
+! Check for semantic errors in co_sum() function calls
+
+module test_co_sum
+  implicit none
+
+contains
+
+  subroutine test
+  
+    integer i, status
+    real array(1)
+    complex z(1,1,1, 1,1,1, 1,1,1, 1,1,1, 1,1,1) 
+    character(len=1) message
+  
+    ! correct calls, should produce no errors
+    call co_sum(i)
+    call co_sum(i,   1)
+    call co_sum(i,   1,              status)
+    call co_sum(i,   1,              stat=status)
+    call co_sum(i,   1,                           errmsg=message)
+    call co_sum(i,   1,              stat=status, errmsg=message)
+    call co_sum(i,   result_image=1, stat=status, errmsg=message)
+    call co_sum(a=i, result_image=1, stat=status, errmsg=message)
+    call co_sum(i,   result_image=1, stat=status, errmsg=message)
+  
+    call co_sum(array, result_image=1, stat=status, errmsg=message)
+    call co_sum(z, result_image=1, stat=status, errmsg=message)
+ 
+    ! the error is seen as an incorrect type for the stat= argument
+    !ERROR: Actual argument for ‘stat=’ has bad type ‘CHARACTER(KIND=1,LEN=1_8)’
+    call co_sum(i, 1, message)
+ 
+    ! the error is seen as too many arguments to the co_sum() call
+    !ERROR: too many actual arguments for collective subroutines 'co_sum'
+    call co_sum(i,   result_image=1, stat=status, errmsg=message, 3.4)
+  
+    ! keyword argument with incorrect type
+    !ERROR: unknown keyword argument to intrinsic 'co_sum'
+    call co_sum(fake=3.4)
+  
+  end subroutine
+
+end module test_co_sum