fix nvhpc 25.9 compilation

[JorgeG94]

This fixes compiling test-drive with nvfortran 25.9. This seems to be like a compiler bug, I will create an MRE and report

[JorgeG94]

See: https://forums.developer.nvidia.com/t/import-problem-in-nvfortran-25-9/347164

[awvwgk]

Thanks for sharing. Surprising that the interface for the procedure pointer inside the type needs to be visible in the abstract interface, it seems to break the encapsulation the derived type should provide. Fine with including this if it makes test-drive available with nvfortran.

[JorgeG94]

Thanks for sharing. Surprising that the interface for the procedure pointer inside the type needs to be visible in the abstract interface, it seems to break the encapsulation the derived type should provide. Fine with including this if it makes test-drive available with nvfortran.

It also took me by surprise. I've logged a ticket with nvidia since it might be a bug on their side and we'll need to revert this later. I can leave a comment in the code for this

[jvdp1]

Could it be encapsulated in a CPP directive:

  import ...
#ifdef __NVCOMPILER
  import ...
#endif
 ....

[JorgeG94]

Could it be encapsulated in a CPP directive:

  import ...
#ifdef __NVCOMPILER
  import ...
#endif
 ....

See: 81e0bb4

[JorgeG94]

Thanks for the report and nice reproducer. I was able to recreate the error here and have filed a problem report, TPR #37894. I’ve asked engineer to investigate especially since this is a regression.

-Mat

Bug report has been filed with nvidia :) they've picked it up.

[JorgeG94]

The CI appears to be dead