Extend compile.sh for libmathsat5j for macOS dylib

[lembergerth]

The produced library can not be distributed yet, because the script does not sign it.

[kfriedberger]

Looks valid. I can not test this on MacOS.

Would it be possible/useful to

• cross-compile the bindings on Linux,
• publish the unsigned artifact via Ivy, and
• delay the signing process for the user, e.g. manually done after installation?

We might think of requesting a developer account from Apple to sign such libraries. Seems to be free of costs for open-source and educational projects.

Perhaps we can even get some Gitlab runner / Github action for building and signing our artifacts automatically. That could be a nice student project.

[lembergerth]

Perhaps we can even get some Gitlab runner / Github action for building and signing our artifacts automatically. That could be a nice student project.

I think this would be the first necessary step.
I would like java-smt to support more solvers for macOS out-of-the-box, but manual testing each release would be a nightmare.

[lembergerth]

cross-compile the bindings on Linux,

I don't think this is possible.
Just building the library with a non-Apple clang on macOS already disallowed the signing process (but there was no nice error message, so I am not sure about the concrete issue).
So a cross-compiled library may not be signable.

[kfriedberger]

AppVeyor has MacOS as platform available: https://www.appveyor.com/docs/macos-images-software/
This could be enough for JUnit tests. As first step, we could extend our CI-pipeline for AppVeyor for MacOS, as already written here: #384

As Apple is quite special in building and signing software, I am not sure about a efficient and maintainable long term solution.