Circular dependencies issues

[MilitaoLucas]

Hello, circular dependency issues make it impossible to build shared libraries for OCC. I plan to fix this in the future if it isn't important. I guess building shared libs should be disabled for now as it doesn't work.

[peterspackman]

Yes, I should fix this. It shouldn't be too hard, but I'm in the middle of about 3 major features at the moment so I'll get onto it once those are merged/done

[MilitaoLucas]

Oh, I also am working in some changes so it is easier to use OCC as a library from NoSpherA2. As NospherA2 is still used from the commandline in olex2, it isn't a problem yet.

I will open a PR after I manage to fix the NoSpherA2 things... I also made a hack so it creates static objects for everything and just links it with libraries with empty C++ files... It kind of works, but it isn't elegant or good...

[peterspackman]

This is a bit more effort than I'd thought - your workaround for now is probably the best way to go. I'll do it, but it's a pretty large refactor of a few sections (mostly qm, gto and io) so it'll take a bit to verify I haven't broken anything after getting everything compiling again.

[MilitaoLucas]

This is a bit more effort than I'd thought - your workaround for now is probably the best way to go. I'll do it, but it's a pretty large refactor of a few sections (mostly qm, gto and io) so it'll take a bit to verify I haven't broken anything after getting everything compiling again.

I will do a PR this week solving this. I also plan to do the PR with the MSVC compatibility for NoSpherA2, but I will do it after (if) the PR is accepted. I didn't work much on the nix draft and made it more as a proposal. I plan to work after I finish upstreaming the changes that NoSpherA2 depends on.

[peterspackman]

Just to flag it, it's partly solved in some changes I haven't pushed yet as well. Unfortunately I made the decision to do it on a working branch with a lot of bigger new things I'm doing. I'd be happy to accept the PR I think, but just keep in mind it might end up getting clobbered a bit in the future anyway. Some things like shell data structures moving to gto, and so on. My own recommendation is that unless it's useful for you downstream to have your changes merged into main soon, I'd just save the effort. If it helps though, and is blocking you then it's fine, will just be a bit of duplicated effort

[MilitaoLucas]

Just to flag it, it's partly solved in some changes I haven't pushed yet as well. Unfortunately I made the decision to do it on a working branch with a lot of bigger new things I'm doing. I'd be happy to accept the PR I think, but just keep in mind it might end up getting clobbered a bit in the future anyway. Some things like shell data structures moving to gto, and so on. My own recommendation is that unless it's useful for you downstream to have your changes merged into main soon, I'd just save the effort. If it helps though, and is blocking you then it's fine, will just be a bit of duplicated effort

I can do the PR to the branch you're working on if it is better. I saw that some parts were already moved to their own IO subdirectories.
I am using a downstream version of libcint that supports MSVC. The objective was to switch from my repository to a pinned commit in yours in the git submodule. Basically the patches would be to better support MSVC and to make a shared/static lib target for OCC. But that can wait.

[MilitaoLucas]

Oh, I just realized you said that you didn't push the branch...

[peterspackman]

Yeah I haven't yet. It's a bit of a big mess (implementing full multipole electrostatic interactions + optimisation for crystal structures) so it's a fair undertaking. Whatever works for you in the interim!

[MilitaoLucas]

Yeah I haven't yet. It's a bit of a big mess (implementing full multipole electrostatic interactions + optimisation for crystal structures) so it's a fair undertaking. Whatever works for you in the interim!

Thanks. I plan to discuss what is better with Florian. Also, I wish to keep using toml to set up everything in a similar way to CLI11 from the C++ API. I added Toml++ in my branch. Is that fine? I couldn't get CLI11 to work in a similar way to the CLI input.