Jiri Slaby

Fixes the issue for me too. Can we have this merged?

> Hmm, so, just so I understand: on some platforms, the above check fails and then we `rpath` the STP libs to be `lib64` (like it is relative path), Right....

> So why does the check fail here? Surely `/usr/lib64` **is** in `CMAKE_PLATFORM_IMPLICIT_LINK_DIRECTORIES`? But `CMAKE_INSTALL_LIBDIR` is `lib64`. `CMAKE_INSTALL_PREFIX` is `/usr`. And `CMAKE_INSTALL_FULL_LIBDIR` would be `/usr/lib64`, with `GNUInstallDirs`. With this `CMakeLists.txt`:...

> You seem to be hesitant for STP to adopt `GNUInstallDirs`; what's the disadvantage (if you know)? I use it for all my projects. But they are all linux. I...

> Another question: could we just "emulate" `GNUInstallDirs` by joining `CMAKE_INSTALL_PREFIX` with `CMAKE_INSTALL_LIBDIR`? `CMAKE_INSTALL_LIBDIR` could be absolute, so the join would fail in that case.

> Would you be happy to make a PR that uses `GNUInstallDirs` only on Linux? Heh: https://github.com/stp/stp/blob/e8d153f196dcdf609add70e867f2188d58ffcc97/CMakeLists.txt#L49 So it should be matter of correct use of `CMAKE_INSTALL_FULL_LIBDIR` now.

> Just so we know, the snippet you highlighted came from the Kitware (the makers of CMake) wiki: > > * https://gitlab.kitware.com/cmake/community/-/wikis/doc/cmake/RPATH-handling#always-full-rpath > > > Given it is also used...

> Just noticed that these two links behave differently than stp+minisat. They have `${CMAKE_INSTALL_PREFIX}/lib`. That would actually work. (But only on 32 bit -- lib vs lib64.) This was changed...

@eumagga0x2a compiler warns about the `default` label. `ADM_assert` expands to `ADM_backTrack`, but the latter is not defined with `no_return` attribute. And it actually can return on some platforms (HAIKU or...

Based on the previous comment, it should `return false`, given there is `return true` hidden in the `END_PROCESS` macros used there.