alalazo
27280ea8be
fix : added regression tests
2016-04-12 17:28:23 +02:00
alalazo
80678b2188
fix : proper update of config file (before it was discarding architectures that were not the current one) fixes #774
2016-04-12 16:54:51 +02:00
alalazo
41f365112c
modules : added provenance comment in tcl header
2016-04-12 09:13:50 +02:00
Adam J. Stewart
cb6c6fb374
Add elk package
2016-04-11 16:28:30 -05:00
Erik Schnetter
882e2e42cf
Merge branch 'develop' into eschnett/julia
2016-04-11 15:40:27 -04:00
alalazo
3959ca6270
modules : added possibility to blacklist or whitelist module files
2016-04-11 18:10:06 +02:00
alalazo
0e2b1359e3
modules : fixed bug preventing the creation of modulefiles autoloading only direct dependencies
2016-04-11 11:01:47 +02:00
Massimiliano Culpo
a5a5dbc408
tcl module file : added new-lines in autoload (per @glennpj bug report)
2016-04-10 13:21:41 +02:00
Erik Schnetter
df7e3f8635
Correct version dependency for OpenBLAS patch
2016-04-09 09:58:48 -04:00
Massimiliano Culpo
6cffac79bd
openmpi : added a few variants
2016-04-09 11:21:29 +02:00
Mario Melara
3a281b2399
Commented out the add_x_from_string methods since we don't need to worry about it. Also, changed the way architecture is set. Todd wanted to have users be able to input just frontend and/or backend in the architecture spec. This will be able to parse those inputs.
2016-04-08 12:46:44 -07:00
Mario Melara
01a36ab333
Changed the logic of concretize. We don't have to worry about whether the part of the spec is a string since any string parsed is made into an object via the Arch constructor. Any dependencies will take the root arch spec as well
2016-04-08 12:44:27 -07:00
Mario Melara
5bcd1e7ccd
If we already have the name and the platform, then we should probably go ahead and concretize it in the constructor. Else leave it as None and concretize it later.
2016-04-08 12:43:08 -07:00
Ben Boeckel
20b9f34b5c
paraview: use internal hdf5
...
Spack's HDF5 is too new. Rather than forcing everything in a ParaView
chain to use older HDF5, use the internal one until ParaView is patched
properly.
2016-04-08 15:09:02 -04:00
Ben Boeckel
6e56ba9f24
paraview: use internal netcdf
...
VTK needs to learn to cope with netcdf being split like this.
2016-04-08 15:08:29 -04:00
Ben Boeckel
3a4aac0213
paraview: use the right cmake variables
2016-04-08 15:08:09 -04:00
Erik Schnetter
f461b99722
Building OpenBLAS requires recent binutils
2016-04-08 14:31:40 -04:00
Mario Melara
618b3f5f2d
Changed default to default_target so the tests can pass
2016-04-08 11:18:30 -07:00
Mario Melara
277efc1dfb
Added test that works on every type of combination the user could possibly enter for arch spec
2016-04-08 11:17:56 -07:00
Matthew LeGendre
1af88be371
Spack was no longer using $TMPDIR for its stage area at LLNL. Spack's directory search was unnecessarily putting candidates that contained your username at the end of its search list.
2016-04-08 11:06:14 -07:00
Mario Melara
0d1a1b7526
Changed _set_architecture so if user inputs only the target, then os is None. This prevents the os being referenced before assignment error
2016-04-08 10:29:37 -07:00
Mario Melara
6ff6c805af
Assigned self.front_os, and back_os to self.default. Helps with testing. If we're on a darwin or linux machine, chances are there is no back end or front end operating system, but rather than have those as None, I just assign them the default which is detected via python platform.
2016-04-08 10:28:30 -07:00
Ben Boeckel
9eaf735bcd
py-setuptools: add 20.6.7
2016-04-08 12:03:45 -04:00
Ben Boeckel
dcd6b19348
py-setuptools: sort versions
2016-04-08 12:03:36 -04:00
Ben Boeckel
6c409d6b92
py-matplotlib: depend on pkg-config
...
On OS X, freetype isn't found by default, but pkg-config can.
2016-04-08 12:03:05 -04:00
Ben Boeckel
7d19154e18
pkg-config: use the internal glib
...
glib requires pkg-config itself, so on machines without pkg-config,
there's a bootstrapping problem.
2016-04-08 12:01:28 -04:00
Mario Melara
7b777568ed
Added new concretization of architecture tests, test each combination of user input to make sure the correct os and target are chosen
2016-04-07 14:38:43 -07:00
Mario Melara
0ad317213c
Changed architecture parser so that if user just enters in frontend or backend, then both the os and target will take those names. In the concretize method the frontend target/os and backend target/os will be picked to match each other
2016-04-07 14:38:04 -07:00
Mario Melara
07df403203
Cleaned up the concretize_architecture method by removing commented out code
2016-04-07 14:37:00 -07:00
Mario Melara
143a4d7f76
Changed default to default_target, also added frontend and backend to operating system
2016-04-07 14:36:26 -07:00
alalazo
1714f410f0
Merge branch 'develop' of https://github.com/LLNL/spack into features/custom_modulefile_from_config
2016-04-07 10:48:16 +02:00
alalazo
7dc1524c08
modules : added stub for dotkit (to be verified)
2016-04-07 10:48:05 +02:00
Todd Gamblin
e73caad0d7
Merge pull request #747 from KineticTheory/pkg-global
...
Package global depends on ncurses
2016-04-07 00:20:15 -07:00
Glenn Johnson
8a7f34665b
Make sure base bioconductor package can be installed.
2016-04-06 23:05:49 -05:00
Glenn Johnson
b19d6ab9c5
Check for non-numeric bits in the stem.
2016-04-06 19:08:01 -05:00
Todd Gamblin
6059da918e
Merge pull request #758 from goxberry/metis-5-url-fix
...
metis 5: fix download url
2016-04-06 17:07:40 -07:00
Geoffrey Oxberry
bbd328d307
metis 5: fix download url
...
Spack picks up the wrong download for metis 5.1.0 from parsing
the url field for metis 4.0.3. Add an explicit url field to fix
this bug.
2016-04-06 16:58:48 -07:00
Todd Gamblin
4437bae16d
Merge pull request #750 from goxberry/metis-4.0.3
...
metis 4.0.3
2016-04-06 16:05:02 -07:00
Todd Gamblin
feee268970
Merge pull request #754 from epfl-scitas/fixes/issue_216
...
fix : module files are deleted on uninstall (fixes #216 )
2016-04-06 16:03:11 -07:00
Todd Gamblin
cac819766e
Merge pull request #755 from adamjstewart/features/bash
...
Add bash package
2016-04-06 16:02:10 -07:00
Todd Gamblin
25d48f700c
Merge pull request #752 from KineticTheory/pkg-netlib-lapack
...
Second attempt at PR#741
2016-04-06 16:01:33 -07:00
Geoffrey Oxberry
ddaba07d82
metis 4.0.3
...
Add version 4.0.3 to metis package. Attempted to implement reasonable
versions of all variants declared for metis@5.1.0; some of these do
not have analogues in metis@4.0.3, and errors are raised accordingly.
Also updated dependencies of packages with depends_on('metis') to
depends_on('metis@5:') to ensure that these packages still build.
2016-04-06 15:49:59 -07:00
Todd Gamblin
d62766a94d
Merge pull request #756 from epfl-scitas/fixes/mpi_small_fixes
...
fix : mpich and mvapich2 will use spack compilers during installation
2016-04-06 15:35:34 -07:00
Glenn Johnson
32779ab1f6
Add r-filehash to test version naming.
2016-04-06 17:02:34 -05:00
Massimiliano Culpo
c113d390a7
mvapich2 : MPI compiler wrapper will use spack compilers during the installation phase
...
mpich : fixed wrong function signature
2016-04-06 23:51:47 +02:00
Glenn Johnson
fbabfc593d
Make R extensable and add a couple of packages for verification.
...
Added R as a build system so that the create template will have the
correct configure line.
Added a regex for version parsing as the R URLs are a little odd.
2016-04-06 16:44:22 -05:00
Mario Melara
99117cbc90
pulled from current repo the build_environment.py. Added module loading methods from our cray port
2016-04-06 14:07:22 -07:00
Mario Melara
918ad28a93
Changed all target values to architecture. Also added a sort_key method so that older architecture specs can be sorted with the newer ones
2016-04-06 13:58:29 -07:00
Adam J. Stewart
380a2b23e6
Add bash package
2016-04-06 15:57:03 -05:00
Massimiliano Culpo
4001663982
leftover from cherry-pick
2016-04-06 22:19:07 +02:00
