cctbx debian package: new commit

[Luc Bourhis]

Hi Baptiste,

> a few quick remarks:
> 
> 1) the setup.py does not build the python extensions. It just installs
> them, but delegates the buiding to the SCons build system. So we need to
> interface with SCons anyway.

Any project complex enough must use a proper build system and distutils does clearly not fit the bill. Thus there must be plenty of Debian ports where distutils must interact with make, cmake, SCons, waf, you name it! Or am I not understanding you correctly here?

> 2) there are a few C++ applications which depend on cctbx, the one I
> know is FOX ( http://vincefn.net/Fox/ , objcryst-fox in Debian). For
> those use cases, should we still build a libcctbx-dev package with the
> headers and static libs (for those we don't need to care about ABI
> stability)?

Iirc Vincent's build system requires to put next to the ObjCryst directory an archive containing precompiled libs, among which cctbx. Thus I would recommend providing a libcctbx-dev package containing all the files on sourceforge, and build it with static libs, then put that build where objcryst expects it and let it build itself.

> 3) the tests in the "run_tests.py" scripts are either C++ or python.
> Some of the python tests are part of importable python packages, some
> are not. Some of the C++ tests seem to need data files, I didn't
> encounter this problem with python tests. Luc, is there anything I
> missed? Finding and installing all those files to the right places is
> n°1 on my current Debian task list.

You don't plan to move the tests away from their location on our sourceforge repo, do you? In which case, you would just need to set a few environment variables as part of installing that libcctbx-dev to get it all working as is. As for the need for input files, you need to keep in mind that most Phenix developers assume that the tests will be run in the context of a full Phenix development environment. That means a directory phenix_regression containing some tests structures and a directory chem_data containing thousands of them. Those two are only available from private subversion repos in Berkeley. Nat would happily give you a user account on their machine of course. That's for the theory. In practice, I really need to check how many tests cannot run outside of that Phenix development environment. As you can imagine, I have had it set up for ages, so I have no idea!

Best wishes,

Luc