* zoltan: only add gussed MPI libs if no MPI wrappers are used
- if MPI-wrappers are used for compilation, we can assume that
linking works without manually specifying MPI libs
(guessing may result in wrong libs, cf. #8979)
- thus, only guess the NPI libs and add them explicitly if no
MPI-wrappers are used
- use llnl.util.filesystem.find_libraries instead of a locally
defined routine to guess the MPI libs if needed
(cf. #8979)
* zoltan: rely on MPI-wrappers to know the required MPI libs
* This fixes a number of bugs:
* Patches were not properly downloaded and added to mirrors.
* Mirror create didn't respect `list_url` in packages
* Update the `spack mirror` command to add all packages in the
concretized DAG (where originally it only added the package specified
by the user). This is required in order to collect patches that are specified
by dependents. Example:
* if X->Y and X requires a patch on Y called Pxy, then Pxy will only
be discovered if you create a mirror with X.
* replace confusing --one-version-per-spec option for `spack mirror create`
with --versions-per-spec; support retrieving multiple versions for
concrete specs
* Implementation details:
* `spack mirror create` now uses regular staging logic to download files
into a mirror, instead of reimplementing it in `add_single_spec`.
* use a separate resource caching object to keep track of new
resources and already-existing resources; also accepts storing
resources retrieved from a cache (unlike the local cache)
* mirror cache object now stores resources that are considered
non-cachable, like (e.g. the tip of a branch);
* the 'create' function of the mirror module no longer traverses
dependencies since this was already handled by the 'mirror' command;
* Change handling of `--no-checksum`:
* now that 'mirror create' uses stages, the mirror tests disable
checksums when creating the mirror
* remove `no_checksum` argument from library functions - this is now
handled at the Spack-command-level (like for 'spack install')
* Store ccache directory explicitly in Travis.
Despite we started using ccache on `develop`, it seems the cache itself
is not stored from one CI build to the next. This might be du to the
fact that our language on Travis is Python and not C nor C++.
Hence here we store the ccache directory explicitly.
Lately many CI runs for PRs are failing due to the `mpich` build that
times out on Travis (10 mins. without output). As the timeout seems to
happen consistently during the build phase, increasing the verbosity of
that test can help working around the issue.
The sys_ustat.h.patch to file sanitizer_platform_limits_posix.cc from
PR #10046 does not apply cleanly to gcc 4.8 or 4.9 (or earlier).
GCC up to 4.8.x either don't have libsanitizer or else don't include
ustat.h in sanitizer_platform_limits_posix.cc.
GCC 4.9.x includes ustat.h, but needs a slightly different patch.
The patch applies to GCC 5.x up to 6.4, and 7.x up to 7.3 and also
8.1.0.
The patch is already included in the tar files for gcc 6.5.0, 7.4.0
and 8.2.0.
* flux-sched: fix compilation errors
configure was being run twice becuase the staging symlinks were
confusing autotools
mirror flux-core in building in a subdirectory to avoid other common
compilation problems, like the one in PR #3543
* flux-core: update dependencies
munge is no longer a dependency
provide hints for lua version so that `spack install flux-core` works
"out-of-the-box". Currently, when no explicity version is provided, the
concretizer attempts to use lua@5.3, which is invalid. Closes#10000
* Added a package for the MDAnalysis toolkit.
* Added Ninja as the generator for LBANN, Hydrogen, and Aluminum. Also
fixed a bug in how the Conduit variant is included in LBANN. Added
new versions for Hydrogen and LBANN.
* Updated lbann package so that the CMAKE parameters are set when the
variant is on and off.
