summaryrefslogtreecommitdiff
path: root/lib
AgeCommit message (Collapse)AuthorFilesLines
2017-06-23mv variants: packages are now needed only during normalization (#4129)Massimiliano Culpo4-158/+240
* mv variants: packages are now needed only during normalization The relationship among different types of variants have been weakened, in the sense that now it is permitted to compare MV, SV and BV among each other. The mechanism that permits this is an implicit conversion of the variant passed as argument to the type of the variant asking to execute a constrain, satisfies, etc. operation. * asbtract variant: added a new type of variant An abstract variant is like a multi valued variant, but behaves differently on "satisfies" requests, because it will reply "True" to requests that **it could** satisfy eventually. Tests have been modified to reflect the fact that abstract variants are now what get parsed from expressions like `foo=bar` given by users. * Removed 'concrete=' and 'normal=' kwargs from Spec.__init__ These two keyword arguments where only used in one test module to force a Spec to 'appear' concrete. I suspect they are just a leftover from another refactoring, as now there's the private method '_mark_concrete' that does essentially the same job. Removed them to reduce a bit the clutter in Spec. * Moved yaml related functions from MultiValuedVariant to AbstractVariant. This is to fix the erros that are occurring in epfl-scitas#73, and that I can't reproduce locally.
2017-06-21fix issue #4577 (#4579)becker331-1/+1
2017-06-21Module cmd fix (#3250)becker337-70/+346
* Parse modules in a way that works for both lmod and tcl * added test and made method more robust * refactoring for pythonic clarity * Improved detection of 'module' shell function + refactored module utilities into spack.util.module_cmd * Improved regex to reject nested parentheses we are not prepared to handle * make tests backwards compatible with python 2.6 * Improved regex to account for sh being aliased to bash and used in bash module definition on some systems * Improve test compatibility with lmod * Added error for None module_cmd * Add test for get_module_cmd_from_which() Add test for get_module_cmd_from_which(). Add -c argument to Popen call to typeset -f module in get_module_cmd_from_bash(). * Increased detection options Included BASH_FUNC_module() variable outside of typeset as a detection option This should work on bash even in restricted_shell mode Kept the typeset detection as an option in case the module function is not exported in bash Also added try statements to tests, with environment recreation in finally blocks. * More tests added; some hackiness * increased test coverage for util/module_cmd
2017-06-16System config (#4518)becker335-20/+40
* Code changes to enable system config scope in /etc Files will go in either /etc/spack or /etc/spack/<platform> Required minor changes to conftest. * Updated documentation to match new config scope
2017-06-16bugfix: support EDITOR values with spaces (#4523)Todd Gamblin1-8/+8
- previous code called `which` on $EDITOR, but that doesn't work for EDITORs like `emacs -nw` or `emacsclient -t -nw`. - This patch just trusts EDITOR if it is set (same as previous behavior), and only uses the defaults if it's not.
2017-06-16issue 4492: DependencyMap.concrete checks for unconditional dependencies (#4499)Massimiliano Culpo2-10/+96
* issue 4492: added xfailing test, added owner to DependencyMap * DependencyMap.concrete checks if we have unconditional dependencies This fixes #4492 and #4107 using some heuristics to avoid an infinite recursion among Spec.satisfies, Spec.concrete and DependencyMap.concrete The idea, as suggested by @becker33, is to check just for unconditional dependencies. This is not covering the whole space and a package with just conditional dependencies can still fail in the same way. It should cover though all the **real** packages we have in our repo so far.
2017-06-15Fix tests on cray (#4298)Mario Melara3-5/+4
* Check for CRAYPE_VERSION instead of path Architecture tests would fail on Cray since it would not find the expected path. To make the test correctly work on Cray search for the CRAYPE version instead. * Catch SystemExit error in case flake8 not in path On shared systems having flake8 can involve starting own virtual env. Skip the test if no flake8 is found to avoid failure reporting. * Add compatibility to 1.5 svnadmin create The flag added is needed to correctly create svn repos on NERSC systems. This could be unnecessary for other sites. I'd like to see others test before this change gets merged.
2017-06-15Add latest version of JDK (#4317)Adam J. Stewart1-10/+0
* Add latest version of JDK * Use a more specific URL to get curl to work * Remove failing unit test
2017-06-15Fix dashes in variant parsing (#4498)Todd Gamblin2-8/+8
- Skip spack flake8 test when flake8 is not installed. - Fix parsing of dashes in specs broken by new help parser. - use argparse.REMAINDER instead of narg='?' - don't interpret parts of specs like -mpi as arguments.
2017-06-15Prefer vim to vi for default editor (#4230)Adam J. Stewart4-82/+94
* vim > vi * Allow which to accept multiple args * Update __init__ to use which with multiple args * Fix doc tests
2017-06-14Adding package namd (#4321)Nicolas Richart1-11/+12
* Initial version of the namd package * Modified charm to consider compile against intel/intel-mpi * Correction of namd to compile with intel-mkl and intel compiler * Adding inclue64 in the prefix * adding property for the build directory * removing useless function build
2017-06-13Override partial installs by default - part three (#4331)scheibelp4-3/+172
* During install, remove prior unfinished installs If a user performs an installation which fails, in some cases the install prefix is still present, and the stage path may also be present. With this commit, unless the user specifies '--keep-prefix', installs are guaranteed to begin with a clean slate. The database is used to decide whether an install finished, since a database record is not added until the end of the install process. * test updates * repair_partial uses keep_prefix and keep_stage * use of mock stage object to ensure that stage is destroyed when it should be destroyed (and otherwise not) * add --restage option to 'install' command; when this option is not set, the default is to reuse a stage if it is found.
2017-06-12Add an installcheck phase to MakefilePackage (#4476)Adam J. Stewart1-0/+12
* Add an installcheck phase to MakefilePackage * Minor changes to ESMF
2017-06-09Move description to top of `spack info` (#4475)Todd Gamblin1-7/+8
2017-06-09Add default list_url for GitLab, BitBucket, and CRAN (#4439)Adam J. Stewart1-5/+38
* Add default list_url for GitLab, BitBucket, and CRAN * Fix flake and doc tests
2017-06-09Move gpg section of docs to Getting Started (#4446)Adam J. Stewart2-64/+67
2017-06-09Fix url parse offset for SourceForge downloads (#4458)Adam J. Stewart2-1/+4
2017-06-07Supress output from gpg --version during tests (#4441)Adam J. Stewart1-2/+2
2017-06-07Fixed duplicated spec: nag@6.1 line in getting started guide (#4445)Stas Sergienko1-0/+1
2017-05-30Fix spack info bug for Python 3 (#4391)Adam J. Stewart1-2/+2
2017-05-27Document known issue with R concretization (#4313)Adam J. Stewart1-0/+40
2017-05-26Fix typo in mod_to_class docstring (#4371)Adam J. Stewart1-1/+1
2017-05-26Add --configure-args/vars support to RPackage (#4289)Adam J. Stewart2-5/+37
* Add --configure-args/vars support to RPackage * Docstring formatting change
2017-05-26gpg: add 'spack gpg subcommand (#3845)Ben Boeckel6-0/+541
- Add a `spack gpg` subcommand in anticipation of signed binaries. - GPG keys are stored in var/spack/gpg, and the spack gpg command manages them. - Docs are included on the command.
2017-05-25Touch up string expansion. (#4344)George Hartzell1-1/+2
* Touch up string expansion. I'm chasing this: ``` $ (module purge; spack install perl %gcc/5.4.0) ==> Error: No installed spec matches the hash: '%s' ``` There's something deeper going on, but the error message isn't helpful. After this change it tells me this: ``` $ (module purge; spack install perl %gcc/5.4.0) ==> Error: No installed spec matches the hash: '5.4.0' ``` Which is weird because `5.4.0` is not a hash... Whatever is going on here, the error message needs to be fixed. * Flake8 whitespace
2017-05-24Fix issues parsing multiple anonymous specs (#4199)becker332-18/+24
* fix parser * Removed xfails * cleaned up debug print statements * make use of these changes in gcc * Added comment explaining unreachable line, line left for added protection
2017-05-24Fix PGI version detection: 17.4-0 -> 17.4 (#4251)Adam J. Stewart1-1/+1
2017-05-17Move doc dependencies to requirements.txt for readthedocs (#4280)Todd Gamblin1-0/+5
* Move doc dependencies to requirements.txt for readthedocs * Move sphinx to doc requirements.
2017-05-17fix bug with executables setting their own environment. (#4237)Todd Gamblin1-3/+4
2017-05-17Sphinx no longer supports Python 2.6 (#4266)Adam J. Stewart6-301/+3
* Sphinx no longer supports Python 2.6 * Update vendored sphinxcontrib.programoutput from 0.9.0 to 0.10.0 * Documentation cannot be built in parallel * Let Travis install programoutput for us * Remove vendored sphinxcontrib-programoutput Recent updates to the sphinx package prevent the vendored version from being found in sys.path. We don't vendor sphinx, so it doesn't make sense to vendor sphinxcontrib-programoutput either.
2017-05-16Sphinx no longer ignores first argument (#4243)Adam J. Stewart1-1/+1
* Sphinx no longer ignores first argument * Duplicate first argument for maximum compatibility
2017-05-12Add helpful error message for uncompressed downloads (#4205)Adam J. Stewart1-6/+26
2017-05-11Fix typo in PythonPackage documentation (#4221)Adam J. Stewart1-1/+1
2017-05-11filesystem.py: fixed bug introduced in #3367 (scrambled order in output) (#4156)Massimiliano Culpo19-7/+95
PR #3367 inadvertently changed the semantics of _find_recursive and _find_non_recursive so that the returned list are not ordered as the input search list. This commit restores the original semantic, and adds tests to verify it.
2017-05-09cp2k: fixed compilation issues for intel stackMassimiliano Culpo1-1/+1
Added DFLAGS to the `make.inc` file being written. These macros are also added to the language specific variables like CFLAGS, CXXFLAGS and FCFLAGS. Changed `spec.satisfies('foo')` with `'foo' in spec` in `intel-mkl`, see #4135. Added a basic build interface to `intel-mpi`.
2017-05-08spec_syntax: added xfailing tests for cases in #4144 (#4151)Massimiliano Culpo1-0/+26
It seems that parse_anonymous_spec may fail if more than one part (variant, version range, etc.) is given to the function. Added tests to code against to fix the problem in #4144.
2017-05-08rework spack help (#3033)Todd Gamblin55-17/+711
- Full help is now only generated lazily, when needed. - Executing specific commands doesn't require loading all of them. - All commands are only loaded if we need them for help. - There is now short and long help: - short help (spack help) shows only basic spack options - long help (spack help -a) shows all spack options - Both divide help on commands into high-level sections - Commands now specify attributes from which help is auto-generated: - description: used in help to describe the command. - section: help section - level: short or long - Clean up command descriptions - Add a `spack docs` command to open full documentation in the browser. - move `spack doc` command to `spack pydoc` for clarity - Add a `spack --spec` command to show documentation on the spec syntax.
2017-05-07BarrierTimeoutError must derive from Exception (#4157)Massimiliano Culpo1-1/+1
Seen in https://travis-ci.org/LLNL/spack/builds/229484526, very likely due to a problem in the Travis builder.
2017-05-07Fix PGI compiler detection on PowerPC (#4150)Adam J. Stewart1-2/+10
2017-05-07Spec.__init__: removed dead code (#4146)Massimiliano Culpo1-5/+2
The conditionals are repeated in the statement before _add_dependencies
2017-05-05Remind developers to update tab completion script (#4148)Adam J. Stewart1-0/+4
2017-05-05Fix typos in Basic Installation Tutorial (#4127)Adam J. Stewart1-4/+4
2017-05-04SV variants are evaluated correctly in "when=" (#4118)Massimiliano Culpo3-41/+48
* SV variants are evaluated correctly in `when=` statements fixes #4113 The problem here was tricky: ```python spec.satisfies(other) ``` changes already the MV variants in others into SV variants (where necessary) if spec is concrete. If it is not concrete it does nothing because we may be acting at a pure syntactical level. When evaluating a `when=` keyword spec is for sure not concrete as it is in the middle of the concretization process. In this case we have to trigger manually the substitution in other to not end up comparing a MV variant "foo=bar" to a SV variant "foo=bar" and having False in return. Which is wrong. * sv variants: improved error message for typos in "when=" statements
2017-05-04add transitive run dependencies of direct build dependenciesDenis Davydov1-3/+3
2017-05-03Merge pull request #4069 from davydden/fix_transitive_build_depscheibelp1-1/+6
only add direct build-only dependencies to PATH
2017-05-03Update cray compiler options (#4086)Matthew Scott Krafczyk1-0/+12
2017-05-03spack: no stacktrace if not in debug mode + fix emacs variant (#4098)Massimiliano Culpo1-2/+2
* spack: no stacktrace if not in debug mode + fix emacs variant * emacs: removed dead code
2017-05-02Add link to spack view docs in command index (#4082)Adam J. Stewart1-3/+4
2017-05-02add run-time dependencies of direct build-time dependencies to PATHDenis Davydov1-0/+5
2017-05-01only add direct build-only dependencies to PATHDenis Davydov1-1/+1