Update libcantera-devel

[ischoegl]

This PR includes some minor updates:

• Include Fortran modules in libcantera-devel
• Create "clean" sample folders for cxx and f90
• Install man pages in cantera
• Ensure that libboost is a dependency for libcantera-devel
• Fix extra level in matlab folder structure
• Add SCons option package_build = True

Closes #37.

[bryanwweber]

• Ensure that libboost is a dependency for libcantera-devel

This shouldn't be needed, as we only rely on the header portions of the library? The concern I have is that this will then bring in libboost for the cantera package too (because cantera depends on libcantera-devel).

[ischoegl]

• Ensure that libboost is a dependency for libcantera-devel

This shouldn't be needed, as we only rely on the header portions of the library? The concern I have is that this will then bring in libboost for the cantera package too (because cantera depends on libcantera-devel).

Since libcantera-devel depends on libcantera, there shouldn’t be a concern? I can confirm once the packages are done building. (This is still WIP after all 😆). I may have to make sure it’s headers only, but libcantera-devel does indeed need them and they currently aren’t installed.

[bryanwweber]

Ah, if cantera depends on libcantera, then libcantera-devel is separate. I don't think I realized we'd made that split. All should be fine then!

[ischoegl]

@bryanwweber ... it is ready now. I also just confirmed that

conda install -c file:///opt/conda/envs/conda-bld/conda-bld/ libcantera-devel

and

conda install -c file:///opt/conda/envs/conda-bld/conda-bld/ cantera

have the expected dependencies.

PS: since the latest GH Action, the only change was to remove a ignore_run_exports_from from the libcantera-devel recipe (which negates the dependency). A local build passes, so I think it's good to go.

[bryanwweber]

Two small things ☺️ I wish I had more time to be productive with Cantera!

[ischoegl]

@bryanwweber ... thanks! Things are taken care of. I triggered a GH Action to build the packages just to make sure.

[ischoegl]

@bryanwweber ... I documented the issues with Makefile's in #37 and in comments, and verified that Fortran modules are packaged properly. So I think this is good to go ...

[bryanwweber]

Another couple small clarifications here 😊

[ischoegl]

@bryanwweber ... fixed things per comments. I also caught one extra-level introduced by copying of some folders of the cantera-matlab package on *nix, which is also fixed.

[ischoegl]

@bryanwweber ... I'd suggest to table the discussion on whether or not to package 'broken' Makefile's. The issue is now documented - see #37 - and without it, this PR still introduces some features that are worthwhile (see top).

[ischoegl]

@bryanwweber …. As things stands, this PR is independent of #37 and Cantera/cantera#1366. As such, this is converged on my side.

Now that Cantera/cantera#1366 is merged, the missing SCons option can be added here.

[ischoegl]

Using artifacts from a CI run on GitHub Actions, I have confirmed that #37 is resolved on Linux. Tests on windows revealed issues beyond #37, although some of the build routes are broken even for a regular install (see Cantera/cantera#1375 etc.) and thus need to be taken care of upstream first.

I haven't had a chance to test on macOS yet, but don't expect any surprises as the recipe is largely identical to that of Linux. Edit: On macOS, CMake works for C++ examples, although make and SCons appear to miss some linker flags (it does work for a local build, although there are some issues with Fortran samples even there, see Cantera/cantera#1378).

And of course, SCons/scons#3664 is again popping up on Windows when trying to use SCons to compile samples ... (although it does appear like a resolution of that issue may be on the horizon).

Edit: Summarizing, this PR resolves issues on Linux; I'd suggest to address remaining issues with macOS and Windows in follow-up PR's