Python install modifications for compatibility with conda-forge

[bennibbelink]

I think our python installation script is more complex than is necessary and it is causing builds in conda-forge/cyclus to fail at pytests due to not being able to find the cyclus package (stemming from incorrect installation).

My understanding of the issue is as follows:

• Conda has environment variables to define the installation location (e.g. PREFIX, SP_DIR) to support the portability of our package. These variables are kept as placeholders throughout the build process (you can see this in the install location in the build output - Installing to PREFIX/bin).
• In the cyclus build we currently generate SetupPyInstall.cmake based on some vars in the build. I think because we configure this during the build (substitute the @@ variables) the placeholder used by conda was not being kept as a placeholder, rather it was expanded out to the absolute path instead of SP_DIR/cyclus (as I saw in build output)
  This PR eliminates the need for configuring a separate CMake script by replacing the SCRIPT argument with CODE.

We can then pass in the conda placeholder via -DPYTHON_SITE_PACKAGES

[github-actions]

Downstream Build Status Report - ad33d53 - 2024-04-20 13:50:49 -0500

Build FROM cyclus_20.04_apt/cyclus

• Cycamore: Success ✅
• Cymetric: Success ✅

Build FROM cyclus_20.04_conda/cyclus

• Cycamore: Success ✅
• Cymetric: Success ✅

Build FROM cyclus_22.04_apt/cyclus

• Cycamore: Success ✅
• Cymetric: Success ✅

Build FROM cyclus_22.04_conda/cyclus

• Cycamore: Success ✅
• Cymetric: Success ✅

[coveralls]

Pull Request Test Coverage Report for Build 8767088217

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• 2 unchanged lines in 1 file lost coverage.
• Overall coverage decreased (-0.002%) to 32.599%

---

Files with Coverage Reduction	New Missed Lines	%
build/cyclus/py_lib_.cxx	2	34.75%

Totals
Change from base Build 8765205180:	-0.002%
Covered Lines:	53079
Relevant Lines:	129079

---

💛 - Coveralls

[bennibbelink]

Ok I had to make one more change (which is really a change to cycamore, but I replicated it here for consistency).

FindCyclus.cmake finds libcyclus.so for cycamore to link against. In the conda cycamore builds the cython generated libraries (eventhooks.so, pyinfile.so, and pymodule.so) were not being linked agains because they were not being reported by FindCyclus.cmake. I will make a PR in cycamore to implement this - I have it working locally with the two respective branches, but in order to see CI pass we will need to merge these changes in cycamore first.

This was not a problem before because we always export LD_LIBRARY_PATH to /root/.local/lib which is the location of these three cython generated libraries. Most literature I can find considers the use of LD_LIBRARY_PATH to be relatively unsafe. With these build changes we no longer need to set LD_LIBRARY_PATH for the linker to find all the libraries it needs.

[bennibbelink]

Hmm it looks like making the changes to FindCyclus.cmake in this PR fixed the issue. I now see that Cycamore's CMakeLists.txt we set the CMAKE_MODULE_PATH to look in /root/.local/share/cyclus/cmake which has the Cyclus version of FindCyclus.cmake. Thus modifying this repo's version of the file fixes our issue, but we should make these same changes in cycamore (#596).

[gonuke]

Thanks @bennibbelink - this all looks reasonable. Just one question about installing FindCyclus.cmake "properly"?

[gonuke]

Thanks for the update @bennibbelink