Use GNUInstallDirs to fix installation problems on mac

[pbrady]

PR Summary

A make install on my mac generated the following error:

     479    -- Installing: /parthenon/.
  >> 480    CMake Error at src/cmake_install.cmake:57 (file):
     481      file INSTALL cannot make directory "/parthenon/.": No such file or
     482      directory.
     483    Call Stack (most recent call first):
     484      cmake_install.cmake:42 (include)
     485
     486
  >> 487    make: *** [Makefile:103: install] Error 1

This PR uses GNUInstallDirs to ensure things are defined. The installed directory structure looks correct but I've never installed it before so...

PR Checklist

• Code passes cpplint
• New features are documented.
• Adds a test for any bugs fixed. Adds tests for new features.
• Code is formatted
• Changes are summarized in CHANGELOG.md
• CI has been triggered on Darwin for performance regression tests.
• Docs build
• (@lanl.gov employees) Update copyright on changed files

[Yurlungur]

Oh! This bit me in a previous project and I forgot about it. Good catch! I think this is a sufficiently trivial fix we don't need changelog, etc...

P.S., We often work on branches in parthenon, rather than forks. Would you like push access so you can make those?

[pbrady]

Thanks! Push access would be great.

[Yurlungur]

Done

[pbrady]

Is the CI failure due to the fork workflow or just some random error?

[Yurlungur]

Looks to me like the CI machine had some internet connection issue:

     for file: [/__w/parthenon/parthenon/tst/regression/gold_standard/parthenon_regression_gold_v19.tgz]
      expected hash: [e1d1a06b9cf9b761d42d0b6b241056ac75658db90138b6b867b1ca7ead4308af4f980285af092b40aee1dbbfb68b4e8cb15efcc9b83d7930c18bf992ae95c729]
        actual hash: [cf83e1357eefb8bdf1542850d66d8007d620e4050b5715dc83f4a921d36ce9ce47d0d13c5d85f2b0ff8318d2877eec2f63b931bd47417a81a538327af927da3e]
             status: [52;"Server returned nothing (no headers, no **data)"]**

If this persists, you might try moving from a fork to a branch. It's possible the CI machine doesn't like running forks. (I vaguely remember hitting that previously)

[pgrete]

The CI error might have been related to a network outage here (in addition to coming from a fork).

I had a look at the code and do not really understand what the issue was and why the proposed changes fix it.
Is there any more documentation (or similar).
I'm mostly worried about potential side effects on various platforms/software environments (that we naturally do not cover with our existing regression tests).

[pbrady]

The install part was relying on CMAKE_INSTALL_... variables that are not necessarily set by default. Adding the include(GNUInstallDirs) sets all the CMAKE_INSTALL_... variables to the "standard" definitions. I disabled the auto-merge that @Yurlungur had set due to my not understanding how the headers were consumed by other projects (riot) and encountering some errors.

I'll remove the draft status after I sort things out.

[pbrady]

This is ready. I've tested out the installed parthenon with riot and that works so this is probably doing the right thing.

[Yurlungur]

@pgrete @pdmullen @jdolence @BenWibking @lroberts36 @forrestglines can we get one more approval?

[pdmullen]

I would be happy to approve this (hence instantiating the auto-merge), but @pgrete raised some concerns above, so I think it best that he takes a look too.

[pgrete]

Looks good. I just tested that at least locally this works as expected and it does!
I'm approving (tough I think it'd be nice to also add it to the changelog so that people like @anku94 who use it as library know that there was a change related how the install works/behaves).

[pgrete]

Looks like I had to re-enable auto-merge to trigger the extended CI tests.

[pgrete]

Style check still doesn't seem to trigger for forks. As the extended test pass I'm manually pulling the trigger.