RePrimAnd thorn requires GSL 2.0 but does not test for it

Issue #2610 new
Former user created an issue

The RePrimAnd thorn currently does not document its need for GSL 2.0 in a way Cactus can check.

The meson build file contains the check there is no check in configuration.ccl or build.sh.

Comments (6)

  1. Wolfgang Kastaun

    What would be the cactus way of requiring a minimum GSL version from the externallibraries GSL thorn?

    Note: The GSL dependence will probably be removed in the foreseeable future in favor of BOOST.

  2. Roland Haas

    There’s no real good way. There is a “version” attached with an ExternalLibrary but that does not give any guarantee about the version of the library actually used (in particular when using a system provided one and not building from source).

    I have a bad way of testing for GSL in: https://github.com/rhaas80/RePrimAnd/commit/a8f5a6664739b676e1fb44208cae5a790e34e36a but is is kludgy since I basically do a mini autoconf.

    Meson of course contains the correct version dependency but for the ET build I am not using Meson (too many dependencies on Python).

  3. Roland Haas

    Though if this is slated to be removed in favor of Boost anyway, I may just sit it out. Though boost of course is also not part of the ET and not particularly version stable 🙂

  4. Roland Haas

    @Wolfgang Kastaun any updates on GSL dependencies? Also on Boost I guess (is that going to be more or less present given that it will not translate well to GPUs)?

  5. Wolfgang Kastaun

    The only remaining GSL dependency is monotonic spline interpolation on irregular spaced data. I have already implemented the monotonic spline interpolation for regularly spaced data in another code part, generalizing that is no too much work and I am optimistic to get rid of GSL soon.

    The Boost dependency is another matter. I am using a root-finder and ODE integrator from Boost which I don’t want to re-implement. However, those are header-only dependencies, so I think I will try to isolate the required headers and ship them in the repo.

  6. Log in to comment