PR #15212 added a new connect_timeout option that can be overridden
using fetch_options but had to specified per-version. This adds a new
per-package variable that can be used to override fetch_options for
all versions in the package. This includes connect_timeout as well
as 'cookie' (e.g. for the jdk package).
Packages can combine package-level fetch_options with per-version
fetch_options, in which case the version fetch_options completely
override the package-level fetch_options.
This commit includes tests for the added behavior.
* Change py-merlinwf to py-merlin to match PyPi.
Change py-merlin to py-merlin-info.
Move to py-merlin_info.
Add py-merlin-info back in.
* Update dependent packages for the new merlin name.
* Remove non-working pyre and the associated packages, exchanger,
py-pythia and py-mlerin-info from citcoms.
* Remove blank line.
* scr: add develop, legacy branches; version 2.0.0
squash! scr: add develop and legacy versions
* filo: package for SCR component
* spath: package for SCR component
* axl: update for versions 0.3 and 0.2
* scr: build with components
* spath: structure of +mpi if/else
* 👌 capitalization of ecp-veloc
* scr: branches are always greater than any version
* Parallel-netcdf: update package.
* Add a temporary patch for version 'develop'.
* Rename version 'develop' to 'master'.
* Drop the patch for 'master'.
includes fixes for likwid-mpirun, better support for ARM and POWER,
other bugfixes.
For full support of ARM and POWER, #14183 has to be merged, too.
Added TomTheBear as maintainer. He is the current main developer of
LIKWID.
* new package: GunRock
* Update var/spack/repos/builtin/packages/gunrock/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* improve
Co-authored-by: Sinan81 <sbulut@3vgeomatics.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Sinan81 <Sinan81@github>
* add --skip-unstable-versions option to 'spack mirror create' which skips sources/resource for packages if their version is not stable (i.e. if they are the head of a git branch rather than a fixed commit)
* '--skip-unstable-versions' should skip all VCS sources/resources, not just those which are not cachable
* New package: bat
Add package for bat:
A cat(1) clone with wings.
* Update copyright date
* Embiggen comment re build env settings
Provide a bit more explanatory text about why setup_build_environment
needs to set LLVM_CONFIG_PATH and LIBCLANG_PATH.
Co-authored-by: George Hartzell <ghartzell@audentestx.com>
* eigen: updated url to point to gitlab
fixes#13890
Eigen migrated from bitbucket to gitlab
* eigen: simplified package (no dependencies other than stdlib)
* Added TODO list for future improvements
* libunwind: remove version 2018.10.12, add stable branch
Finish cleaning up the libunwind version numbers. The 2018.10.12
snapshot number didn't fit well with spack's ordering (my bad), and
1.4-rc1 is a near identical replacement.
Add a version for the 1.4-stable branch.
Add a variant for zlib compressed symbol tables (develop branch only).
Adjust packages caliper and hpctoolkit to adapt to the changes.
Add myself as maintainer.
* Flake
* Settle on renaming 'develop' to 'master' (to match the branch name)
and name the 'v1.4-stable' branch as '1.4-head'. 'stable' or
'1.4-stable' is a better name, but '1.4-head' (an infinity version)
sorts better.
* explicitly link against libtinfo
* Update to v15.9
* fixed indentation
* fixed url definition
* added url vor current version again
* fixed indentation
* moving url_version to the bottom
* Add version 0.5.14
* Add variant to allow setting the localstatedir: See below
* Add bzip2 dependency
* Add myself to maintainers (I just think, I can care for
this package)
About localstatedir:
munge has a server and a client.
They communicate via unix domain sockets.
This socket is in PREFIX/var.
This package provides the client, the server, and
development part (headers, libraries).
Let's assume one has the server part installed as a system
package. This generally is a good idea, so that the server
gets started during boot. This means, that the socket is in
the system's /var.
If one now wants to use the client part (library!) via
spack, one has a problem: spack's munge looks in
SPACK-PACKAGE-PREFIX/var for the socket.
There needs to be a way to let the spack installed package
use the system's socket.
So add a variant to override the path during build:
localstatedir=/var.
* Add Checksum for 1.4.1.4 Release
Add checksum for 1.4.1.4 release. Mark myself as maintainer. List develop version.
* Update var/spack/repos/builtin/packages/libhio/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update package.py
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
It was reported that UnifyFS had a bug with the --enable-mpi-mount
config option, which corresponds to the auto-mount variant. This bug
was fixed in the UnifyFS dev branch, however remains broken for the
0.9.0 version.
This adds a patch to the unifyfs package to fix the auto-mount
variant when installing with version 0.9.0.
This also removes the openssl dependency as unifyfs does not directly
depend on it. This was said to be a non-explicit dependency in #15258.
However, if it is needed, it is likely a non-explicit dependency of
one of unifyfs's dependencies and should be added there.
Fixes: #15292
* lammps: add most recent stable release
* Update var/spack/repos/builtin/packages/lammps/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Add package for ripgrep:
ripgrep is a line-oriented search tool that recursively searches
your current directory for a regex pattern. ripgrep is similar to
other popular search tools like The Silver Searcher, ack and grep.
* [opensubdiv] created stub for opensubdiv
* [opensubdiv] added homepage and description
* [opensubdiv] removed boilerplate
* [opensubdiv] working on dependencies and variants
* [opensubdiv] fixed syntax error
* [opensubdiv] defined spec
* [opensubdiv] added dev version
* [opensubdiv] building on CudaPackage
* [opensubdiv] always build with open gl and depends_on('gl')
* [opensubdiv] applying cuda flags
* [opensubdiv] worked on doc variant
* [opensubdiv] added some x11 libraries
* [opensubdiv] depends on glfw
* [opensubdiv] locating glew
* [opensubdiv] added openmp variant
* [opensubdiv] flake8 fixing
* [opensubdiv] fixed develop version name
* [opensubdiv] fixed description to not need @
* ADIOS2: Version `master`
Rename branch version to supported, real development branch `master`.
The old name is legacy Spack when alternative development branch
names were not yet supported.
* ADIOS: Simplify via spec Variable
use the already defined local variable `spec` to shorten
lines
Since 1.9.0 is broken for Darwin, which impacts many developers, and
the fix is still a RC, let's keep the previous release as default.
This avoids distruption for OSX developers and CI.
* @develop needs the full git repo to use "git describe"
properly
* If not specifying the cxxstd variant, let cmake use its
default
* Improve fmt dependencies: fairlogger < 1.6.2 does not
work with fmt >= 6.
* Small other stuff
Just `depend_on('ncurses')`, pkgconfig seems to cause the right thing
to happen if the package is built +termlib or ~termlib (tested both
ways on a CentOS 7 system).
Additional details in: https://github.com/spack/spack/issues/15281
Tested by building all of the releases of tmux on a CentOS 7 box.
* add preliminary afqmc support in qmcpack
* afqmc updates
* fix spack typos
* edit AFQMC to only allow up 3.7 or above
* added NCCL library support for AFQMC build
* add CMAKE args for BUILD_AFQMC_WITH_NCCL
* update for just AFQMC support. No AFQMC+GPU support
* remove nccl for afqmc
* flake8 whitespace fix
* Update flag_handler for 'netcdf-fortran'.
* Refactoring.
* Enable old versions of netcdf-fortran for NAG.
* Disable parallel 'make check' for versions before 4.5.0.
* Fix shared libraries built with NAG instead of conflicting it.
* Add 'skosukhin' as a maintainer of 'netcdf-fortran'.
This PR adds some fixes for the tcsh package.
- Adds new version
- adds list_url so fetching works for current and old tarballs
- sets ncurses dependency explicitly to `ncurses+termlib`
If `+termlib` is not set then it will link against the system libtinfo.
* Quantum-Espresso: qe-6.5 fails to detect MKL for FFT
qe-6.5 fails to detect MKL for FFT if BLAS_LIBS is set due to
an unfortunate upsteam change in their autoconf/configure:
- qe-6.5/install/m4/x_ac_qe_blas.m4 only sets 'have_blas'
but no 'have_mkl' if BLAS_LIBS is set (which seems to be o.k.)
- however, qe-6.5/install/m4/x_ac_qe_fft.m4 in 6.5 unfortunately
relies on x_ac_qe_blas.m4 to detect MKL and set 'have_mkl'
- qe-5.4 up to 6.4.1 had a different logic and worked fine with
BLAS_LIBS being set
However, MKL is correctly picked up by qe-6.5 for BLAS and FFT if
MKLROOT is set (which SPACK does automatically for ^intel-mkl).
Thus, do not set BLAS_LIBS when compiling qe-6.5 with intel-mkl.
* replace all '^intel-mkl' by '^mkl' to match other packages which also provide MKL
e.g. intel-parallel-studio+mkl as mentioned by @adamjstewart in #15276
* Filter problematic flag from std_cmake_args
Including '-DCMAKE_VERBOSE_MAKEFILE:BOOL=ON' in the cmake args causes
bits of cmake's output to end up in the autoconf-generated configure
script. See https://github.com/spack/spack/issues/15274 and
* Don't build in parallel (sigh)
* Constrain hackery for newer versions...
... to newer versions.
* [WIP] new package: qgis
* add maintainer
* further improvements
* reflect improvements to qscintilla installation paths
* comment out qtkeychain dependency as the package is created
* uncomment qtkeychaing dependency, since this package is now created
* a comment on webkit
* specify versions of dependencies, add variant
* fix variant description
* fix proj dependency logic
* adjust conflicts and dependencies so that one can compile qgis@2 with qt4, python2.7
* minor improvement
* fix some build errors, improve dependency specs
* qsci python bindings will be build by py-pyqt
* cmake variable QSCINTILLA_LIBRARY should point to library itself not the parent folder
* turn grass off explicitly, fix typo, turn qspatialite off explicitly
* fix typo
* specify more cmake options that doesn"t seem to be set properly, and use spack provided pkg-config
* fix libzip
* fix build issue with sqlite variant, add runtime dependencies
* add more runtime python package dependencies
* reflect variant name change in sqlite
* add maintainer, correct typo
* add TODO's
* add more versions
* improve style
* add latest versions
* netcdf -> netcdf-c
* add variants as shown in cmake config
* add conflict: v3.8.1 won't build if qt@5.13:
* change preferred version to latest long term release, 3.10.3
* add a zillion of build options
* improve style
* add descriptions for variants
* remove and already implemented compilation tip
* add "when" statements for optional dependencies
* make flake8 happy
* Update var/spack/repos/builtin/packages/qgis/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/qgis/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/qgis/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/qgis/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/qgis/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* convert conflicts to depends_on
* undo str conversion for path objects
* fix flake8 E131
* fix flake8 E128, E124
* improve style
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Caliper depends on python3.
The package needs to be told where to find it.
* More flake8 formatting edits.
* Change explicit python3 to spec['python'].command.path
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Removing defunct import for flake8
* Flake8 trailing whitespace warning.
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
This PR adds 'http://' to the homepage setting of a few packages that do
not have it set. Not having that set can cause problems with some wiki
apps when embedding the homepage value into markdown syntax.
- bowtie2
- exuberant-ctags
- perl-want
- samtools
- add version info for v1.8.0
- v1.8.0 adds a new source file, `file.c`, which need to be included
in our hardcoded list of objects to link.
I discovered this while demonstrating "how easy it is to add a new
package to Spack", only to scratch my head when it failed `spack
install` but worked when I ran `make` in the stage dir. Finally
looked at tree/package.py and it All Became Clear.
Perhaps someone should rewrite this to use MakefilePackage, but the
Makefile starts off with a bunch of twisty turny "uncomment these
lines to run on this platform", so it might not be worth it.
* Added package py-versioneer
* Update python version
* Added python2 to versioneer
* Added python2 to versioneer
* Update var/spack/repos/builtin/packages/py-versioneer/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Cleaning comments
* Removed temporarily @warner as a maintainer, waiting for answer
* Removed line
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* gromacs: add v2019.6
* Update var/spack/repos/builtin/packages/gromacs/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
- Tell make about the source code path
- Install actual header files, not a wrapper with wrong paths
- Add a patch to prevent compiler warnings
- Improve description
* Fixups for jupyter
This PR fixes a few things for some jupyter related packages.
py-ipython:
- make the python depends_on statements reflect needs of different
versions
- remove an unneeded conflicts directive
py-ipywidgets:
- add new version
- set version constraints for py-widgetsnbextension
py-jupyter-console
- add new version
- set python dependencies for versions as needed
- set version constraint for py-ipython
- set version constraints for py-prompt-toolkit
py-pyqt5
- build with py-sip
py-qtconsole
- add dependency on py-pyqt5
* Update var/spack/repos/builtin/packages/py-jupyter-console/package.py
Tweak version ranges.
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-jupyter-console/package.py
Tweak version range.
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Make py-pyqt5 a run dependency
Also, make formatting more consistent.
* Fix site_packages_dir
Change reference of site_packages_dir to self.site_packages_dir. Oddly,
this did not show up as a problem until I regenerated the module.
* Restore py-pyqt5 to previous state
* Explicitly set path to site_packages_dir
This change prevents an error when regenerating the py-pyqt5 module
file.
* Fix flake8 errors
* Make sure prefix is in join_path
* Fix flake8 errors
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* New package: py-librosa
This PR adds the py-librosa package, along with new dependency packages
and some updates of existing dependency packages.
- new package: py-audioread
- new package: py-resampy
- new package: py-soundfile
- update package: py-numba
- update package: py-llvmlite
py-numba:
- add updated version
- adjust constraints
py-llvmlite:
- add updated version
- adjust constraints
- fix version specifications for llvm
- add environment function to set PIC
* Update var/spack/repos/builtin/packages/py-numba/package.py
Ah, yes, I see that `setuptools` is listed in the `install_requires` array. I missed that before.
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Fix dependency
- Add dependency of py-soundfile depends on libsndfile
- Add new libsndfile package
* Add py-pytest-runner build dep
* Make numpy a variant for py-soundfile
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
When building gcc7 and gcc8 on RHEL6 with Spack and installing it as
a spack-available compiler, OpenBLAS will fail to compile because GCC
generates newer instructions than rhel6's `as` assembler knows about
(e.g. "vpermpd").
Building gcc8 with binutils succeeds, and it generates a GCC that can
then successfully build OpenBLAS. This is also expected to work for
gcc7 on RHEL6.
* Added version 0.12.0 to fix issue #15218
* Added dependencies specs with compatible versions
* Switched py-scipy dependency to variant (default F)
* Removed variant py-scipy and didn't add py-dask
* Fixed typo: missing '
* Update var/spack/repos/builtin/packages/py-pyfftw/package.py
Fixing typos from version ranges in dependencies.
Co-Authored-By: Axel Huebl <axel.huebl@plasma.ninja>
* Update var/spack/repos/builtin/packages/py-pyfftw/package.py
Removed repeating dependency option.
Co-Authored-By: Axel Huebl <axel.huebl@plasma.ninja>
* Update var/spack/repos/builtin/packages/py-pyfftw/package.py
Limited version of py-numpy dependency to <2.0.
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Adding an option to build the C api for Umpire.
This is useful if you need to link to a C code and you're using
a compiler suite that doesn't support fortran.
* Also updating the versions while I'm here.
* Adding conflict: Fortran requires C.
To ease transition and confusion, default to C-bindings being
present. This shouldn't hurt anyone who is upgrading an existing
installation.
LLVM is the only package that explicitly sets the "termlib" variant
of ncurses and it specifies +termlib. ncurses defaults to ~termlib;
if a package depends on LLVM and ncurses, there is a concretizer bug
that incorrectly detects a constraint conflict (see #267). Setting
+termlib as the default is a stopgap measure to avoid this conflict.
If other packages were to explicitly request ~termlib in the future,
the same issue would come up again (and could not be resolved by
adjusting the default of "termlib").
Setting +termlib on ncurses moves some symbols into a separate
"libtinfo". Not all packages may be able to detect libtinfo properly
so may require an update; vim, samtools, and libedit have been
updated to use ncurses+termlib (in the case of libedit, the only
necessary action was to add a newer version where the build system
was updated to check libtinfo).
* Flake8 OK
* Update var/spack/repos/builtin/packages/py-basis-set-exchange/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-basis-set-exchange/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Added missing dependencies proposed by @adamjstewart
* Without py-versioneer
* Added py-versioneer
* Python2 for bse
* Python build error
* Update var/spack/repos/builtin/packages/py-basis-set-exchange/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Removed py-versioneer, according to @adamjstewart
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
- When compiling qrupdate with `FFLAGS=-fdefault-integer-8` it can be perfectly used for larger problem dimensions.
- Improved the readability of the file with the added rules.
* Update py-bx-python package
- add update to py-bx-python
- switch to pypi downloads
- set dependencies
* Update var/spack/repos/builtin/packages/py-bx-python/package.py
I had initially pulled version 0.8.6 and then updated that to 0.8.8 but missed the change in the python specs between those two versions.
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Fix version 0.7.4
- set dependency on python2
- add dependency on py-python-lzo
- add py-python-lzo package
- set py-numpy dependency to correspond to latest version that works
with python2
* Add constraint for py-six dependency
* Update var/spack/repos/builtin/packages/py-bx-python/package.py
Ah, I had that `when` clause in and then took it out as it did not seem to be needed. I guess it is always better to be more explicit.
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Remove py-numpy constraint
Let the concretizer catch the conflict with python2 and py-numpy
versions.
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Fixes & additional variant(s) for HepMC3
* Syntax
* Restore recipe for HepMC2
* Remove FIXME
* Update package.py
* Apply suggestions from code review
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add extra version of py-jsonschema
* Update dependencies
* Update dependencies + flake8
* Add py-pyrsistent package
* Update package.py
* Update var/spack/repos/builtin/packages/py-jsonschema/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update package.py
* Update package.py
* Apply suggestions from code review
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Apply suggestions from code review
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add a variant to QE that suppresses upstream patching. Need in order to do ddependency patching.
* QE variant fails to build often. Set default variant to False as a user friendly change.
* QMCPACK converter patch collides with internal QE patches. Deactivate internal patches when performing dependency patching.
* Clearer description of QE patch variant that is also flake8 compliant.
* Add extra version of py-sqlalchemy
* Update package.py
* Update package.py
* Update package.py
* Update package.py
* Update package.py
* Apply suggestions from code review
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update package.py
* Update package.py
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
The hashing logic looks for function calls that are Spack directives.
It expects that when a Spack directive is used that it is referenced
directly by name, and that the directive function is not itself
retrieved by calling another function. When the hashing logic
encountered a function call where the function was determined
dynamically, it would fail (attempting to access a name attribute
that does not happen to exist in this case).
This updates the hashing logic to filter out function calls where the
function is determined dynamically when looking for uses of Spack
directives.
* Add extra version of py-terminado
* Update package.py
* Update var/spack/repos/builtin/packages/py-terminado/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add extra version of py-widgetsnbextension
* Update dependency version
* Update package.py
* Update package.py
* Update var/spack/repos/builtin/packages/py-widgetsnbextension/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add extra version of py-urllib3
* Update package.py
* Update package.py
* Update package.py
* Update var/spack/repos/builtin/packages/py-urllib3/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add extra version of vc
* Update package.py
* Update package.py
* Update var/spack/repos/builtin/packages/vc/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update package.py
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Mark conflicts with binutils on darwin
* Explicitly require binutils bootstrapping and mark conflict with nvptx
* Disable gold variant by default on darwin
* igv: adding package igv
* removing some remaining initial boilerplate
* changing path construction to be more correct
* adding in type for java dep, also forgot about prefix.bin etc
* Added IRPF90 package
* PEP8
* SHA256
* Update var/spack/repos/builtin/packages/py-irpf90/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
libfabric used to install fabtests only when installed
using --test. fabtests has tools that are useful on a
running system, so they should be installed always.
* Rewrote the build/install part to always install
fabtests alongside libfabric.
* Updated a few fabtests resources.
* Updated the test related stuff. Works for most versions
now.
* Include tcp and udp fabrics so that the test suite works.
Change hpctoolkit's dependency on libunwind from 2018.10.12 to 1.4:.
In libunwind, 2018.10.12 is going away in favor of 1.4-rc1 (they're
nearly identical commits).
Remove the 'gpu' version. This was a temporary branch that is now
folded into master.
* py-notebook: make py-setuptools a run dependency
The py-setuptools dependency in py-notebook needs to be a run
dependency. The following message is received if it is not in the run
environment.
Traceback (most recent call last): File "/opt/ssoft/apps/2020.1/linu
x-centos7-sandybridge/gcc-9.2.0/py-notebook-6.0.1-6usbn4c/bin/jupyter-notebook",
line 6, in <module>
from pkg_resources import load_entry_point
Module NotFoundError: No module named 'pkg_resources'
* Remove extraneous whitespace
* Add extra version of py-jedi
* Update package.py
* Update package.py
Correct dependency types
* Update var/spack/repos/builtin/packages/py-jedi/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-jedi/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-jedi/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Add py-parso package
* Remove boilerplate from py-parso
* Flake-8
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add extra version of py-isort
* Update package.py
* Update package.py
* Update var/spack/repos/builtin/packages/py-isort/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-isort/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-isort/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-isort/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
* Add extra version of py-dask
* Update package.py
* Add extra dependencies for py-dask+distributed
* Update package.py
* Update var/spack/repos/builtin/packages/py-heapdict/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-heapdict/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-distributed/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-distributed/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update package.py
* Update package.py
* Update package.py
* Update package.py
* Update var/spack/repos/builtin/packages/py-distributed/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update var/spack/repos/builtin/packages/py-distributed/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update package.py
* Update var/spack/repos/builtin/packages/py-distributed/package.py
Co-Authored-By: Adam J. Stewart <ajstewart426@gmail.com>
* Update package.py
* Flake-8
* Add patch step for py-distributed
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>
Fixes#9394Closes#13217.
## Background
Spack provides the ability to enable/disable parallel builds through two options: package `parallel` and configuration `build_jobs`. This PR changes the algorithm to allow multiple, simultaneous processes to coordinate the installation of the same spec (and specs with overlapping dependencies.).
The `parallel` (boolean) property sets the default for its package though the value can be overridden in the `install` method.
Spack's current parallel builds are limited to build tools supporting `jobs` arguments (e.g., `Makefiles`). The number of jobs actually used is calculated as`min(config:build_jobs, # cores, 16)`, which can be overridden in the package or on the command line (i.e., `spack install -j <# jobs>`).
This PR adds support for distributed (single- and multi-node) parallel builds. The goals of this work include improving the efficiency of installing packages with many dependencies and reducing the repetition associated with concurrent installations of (dependency) packages.
## Approach
### File System Locks
Coordination between concurrent installs of overlapping packages to a Spack instance is accomplished through bottom-up dependency DAG processing and file system locks. The runs can be a combination of interactive and batch processes affecting the same file system. Exclusive prefix locks are required to install a package while shared prefix locks are required to check if the package is installed.
Failures are communicated through a separate exclusive prefix failure lock, for concurrent processes, combined with a persistent store, for separate, related build processes. The resulting file contains the failing spec to facilitate manual debugging.
### Priority Queue
Management of dependency builds changed from reliance on recursion to use of a priority queue where the priority of a spec is based on the number of its remaining uninstalled dependencies.
Using a queue required a change to dependency build exception handling with the most visible issue being that the `install` method *must* install something in the prefix. Consequently, packages can no longer get away with an install method consisting of `pass`, for example.
## Caveats
- This still only parallelizes a single-rooted build. Multi-rooted installs (e.g., for environments) are TBD in a future PR.
Tasks:
- [x] Adjust package lock timeout to correspond to value used in the demo
- [x] Adjust database lock timeout to reduce contention on startup of concurrent
`spack install <spec>` calls
- [x] Replace (test) package's `install: pass` methods with file creation since post-install
`sanity_check_prefix` will otherwise error out with `Install failed .. Nothing was installed!`
- [x] Resolve remaining existing test failures
- [x] Respond to alalazo's initial feedback
- [x] Remove `bin/demo-locks.py`
- [x] Add new tests to address new coverage issues
- [x] Replace built-in package's `def install(..): pass` to "install" something
(i.e., only `apple-libunwind`)
- [x] Increase code coverage
