MWFilters are only inclued in the source of MRCPP

[bjorgve]

The MWFilters used when running mrchem are located in mrcpp/share/mwfilters. So they are not available if mrcpp is removed from the system.

[stigrj]

Is it not installed with make install?

[stigrj]

No, indeed this does not seem to be the case. Also the SAD input files for the initial guess needs to be installed under share/

[robertodr]

Anything else that should be installed and isn't? @bjorgve and I can prepare a patch easily.

[robertodr]

Note that there will always be an inconsistency between install and source trees when it comes to MW_FILTER_DIR in config.h. I can't think of a better way to bake in this info into the library...

[stigrj]

I think mrcpp/share/mwfilters and mrchem/share/sad_basis are the two things missing

[stigrj]

So it's not possible to get something that works both for the build and install trees at the same time? Is it not possible to reconfigure config.h at make install?

[robertodr]

OK, yes, I can patch it up at install time. It gets trickier with the conda package, but that's a separate issue.

[robertodr]

Even patching config.h is problematic (just tried and failed) The compile-time value of MW_FILTER_DIR is baked into the library since it's used in MWFilter.cpp. How inconvenient would it be to pass that as an input to MRCPP?

[stigrj]

Is it possible to use a relative path? Should be the same in build and install

[robertodr]

I don't know 🤔 It might not work when changing compilers.

[stigrj]

Wait! There should be a INSTALL_FILTERS=on/off cmake option to mrcpp. Not sure if/how it works, though

[robertodr]

Yes, that's where we started on this rabbithole. If you set that, then the tests fail, 'cause MW_FILTER_DIR is hardcoded to the install tree.

[robertodr]

Look at #105 @bjorgve has a solution that, I think, will make everyone happy.

[robertodr]

This was fixed with #112