Update Function Arguments in ?gees, ?geesx, ?gges, ?ggesx, and ?gges3

[foxtran]

Instead of using external keyword for Function Arguments, this PR introduces explicit interfaces for Function Arguments. The code should be more safe now, and source code reflect documented behaviour. Later, it will help a lot to generate proper interfaces for these functions.

BIND(C) is used since code can be called from other than Fortran languages.

Checklist

• The documentation has been updated: no need to update documentation.
• Closes Provide explicit interfaces for external functions passed as arguments #1171

[foxtran]

Hmmm.. I used flang 21.1.6 on MacOS Ventura 13.7.8 to check what wrong in CI, and all tests were passed.

Configure command:

cmake -G "Ninja" -DCMAKE_Fortran_COMPILER=flang -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=ON -DCMAKE_Fortran_FLAGS="-g" ..

testing:

ctest . --parallel 4

Where I can find where LAPACK calls are located?

[foxtran]

Call tree looks valid. I have no idea why test failed.

[foxtran]

@langou, could you please add --output-on-failure flag to ctest command into CI?

[langou]

I added one for .github/workflows/cmake.yml (seems like other 'ctest' had it.)
Let me know if you want it at other places.
Note: I'll need a "maintainer" to approve my PR so as to be able to merge it.

[foxtran]

Let me know if you want it at other places.

What is about continuous-integration/appveyor/pr?

[foxtran]

Ok. Found where it is :-)

[foxtran]

No, I definitely has no idea why appveyor fails.

[martin-frbg]

Could be it's getting miscompiled, because the CI job running on Appveyor appears to be still using flang 11 from four years ago when the new flang was barely able to compile again.
(I had to clamp the version number around then to keep the CI running at all, because conda was acting up and tried to update both flang and itself with the dependencies all wrong)

[martin-frbg]

closing&reopening to trigger a new CI run