Magma is not currently compatible with CUDA-11. While this is reflected
in the package, it is done with a comment in a `depends_on` directive,
which has the effect of trying to install a version of CUDA that may be
different from the one in the current environment, without any message
to the end user. A `conflicts` is a better way to handle this.
* Disable bash completion by default.
* flake8
* Adding explicit dependence on libuuid
* Adding explicit dependence on cryptsetup
This way we don't pick up host crypto packages by mistake.
* Fixing the completion directory.
* Update var/spack/repos/builtin/packages/util-linux/package.py
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* flake8
* Removing libuuid linkage according to @michaelkuhn on #18696
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
This PR adds the current release version of mumax and tweaks the install
of the previous beta version.
- Set the url parameter to reflect the release version over the beta
version. Hopefully, this will be consistent going forward.
- Set an explicit url for the previous beta version.
- Accept values for `cuda_arch`. The previous version had its own list
but the release version does not.
- Replace the built in cuda compute capabilities list with the one
provided by Spack for the 3.10beta version.
This PR fixes a couple of things with the libbeagle package.
- libbeagle can only be built for one GPU type. Add a test for that.
- version 2 had the arch statement in
libhmsbeagle/GPU/kernels/Makefile.am but version 3 has it in
configure.ac. Put the variant specified value in configure.ac for
consistency.
Due to recent changes in the `netcdf-c` package, it is now necessary to explicitly request a non-mpi-enabled hdf5 build if building a non-mpi-enabled seacas.
* libvdwxc: unbreak concretization, request fftw-api
mixing both fftw and fftw-api in a dependency tree can trigger the
following:
```
$ spack spec cp2k@master +sirius
==> [2020-09-16-12:36:06.552981] sirius applying constraint gsl
==> [2020-09-16-12:36:06.554270] sirius applying constraint openblas@0.3.10%gcc@7.5.0~consistent_fpcsr~ilp64+pic+shared threads=none arch=linux-opensuse_leap15-sandybridge
Traceback (most recent call last):
File "./bin/spack", line 64, in <module>
sys.exit(spack.main.main())
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/main.py", line 762, in main
return _invoke_command(command, parser, args, unknown)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/main.py", line 490, in _invoke_command
return_val = command(parser, args)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/cmd/spec.py", line 103, in spec
spec.concretize()
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/spec.py", line 2228, in concretize
user_spec_deps=user_spec_deps),
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/spec.py", line 2716, in normalize
visited, all_spec_deps, provider_index, tests)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/spec.py", line 2654, in _normalize_helper
dep, visited, spec_deps, provider_index, tests)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/spec.py", line 2613, in _merge_dependency
visited, spec_deps, provider_index, tests)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/spec.py", line 2654, in _normalize_helper
dep, visited, spec_deps, provider_index, tests)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/spec.py", line 2554, in _merge_dependency
provider = self._find_provider(dep, provider_index)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/spec.py", line 2489, in _find_provider
providers = provider_index.providers_for(vdep)
File "/data/tiziano/debug-spack/spack2/lib/spack/spack/provider_index.py", line 80, in providers_for
return sorted(s.copy() for s in result)
File "/data/tiziano/debug-spack/spack2/lib/spack/llnl/util/lang.py", line 249, in <lambda>
lambda s, o: o is not None and s._cmp_key() < o._cmp_key())
TypeError: '<' not supported between instances of 'str' and 'NoneType'
```
while at the same point disallowing MKL as a fftw provider.
Solving both issues by depending on `fftw-api@3` instead and adding a
conflict on `^fftw~mpi` when using `+mpi` (thanks to alalazo).
* cp2k: use conflicts instead of runtime checks for fftw/openblas variants
