When the user explicitly sets ~fortran, mpich builds without fortran
support. This will make building C/C++ libraries using clang easier,
since clang does not offer a fortran compiler by default (yet).
Since the user has to disable Fortran support explicitly, this change
is not breaking.
* Handle uninstalled rootspecs in buildcache
- Do not parse specs / find matching specs when in an environment and no
package string is provided
- Error only when a spec.yaml or spec string are not installed. In an
environment it is fine when the root spec does not exist.
- When iterating through the matched specs, simply skip uninstalled
packages
* Run Python2.6 unit tests on Github Actions
* Skip url tests on Python 2.6 to reduce waiting times
* Skip foreground background tests on Python 2.6 to reduce waiting times
* Removed references to Travis in the documentation
* Deleted install_patchelf.sh (can be installed from repo on CentOS 6)
* Buildcache:
* Try mocking an install of quux, corge and garply using prebuilt binaries
* Put patchelf install after ccache restore
* Add script to install patchelf from source so it can be used on Ubuntu:Trusty which does not have a patchelf pat package. The script will skip building on macOS
* Remove mirror at end of bindist test
* Add patchelf to Ubuntu build env
* Revert mock patchelf package to allow other tests to run.
* Remove depends_on('patchelf', type='build') relying instead on
* Test fixture to ensure patchelf is available.
* Call g++ command to build libraries directly during test build
* Flake8
* Install patchelf in before_install stage using apt unless on Trusty where a build is done.
* Add some symbolic links between packages
* Flake8
* Flake8:
* Update mock packages to write their own source files
* Create the stage because spec search does not create it any longer
* updates after change of list command arguments
* cleanup after merge
* flake8
Style and documentation tests take just a few minutes
to run. Since in Github actions one can't restart a single
job but needs to restart an entire workflow, here we group
tests with similar duration together.
- [x] Remove references to `master` branch
- [x] Document how release branches are structured
- [x] Document how to make a major release
- [x] Document how to make a point release
- [x] Document how to do work in our release projects
* Move flake8 tests on Github Actions
* Move shell test to Github Actions
* Moved documentation build to Github Action
* Don't run coverage on Python 2.6
Since we get connection errors consistently on Travis
when trying to upload coverage results for Python 2.6,
avoid computing coverage entirely to speed-up tests.
This is needed because libcuda is used by the driver,
whereas libcudart is used by the runtime. CMake searches
for cudart instead of cuda.
On LLNL LC systems, libcuda is only found in compat and
stubs directories, meaning that the lookup of libraries
fails.
`spack -V` stopped working when we added the `releases/latest` tag to
track the most recent release. It started just reporting the version,
even on a `develop` checkout. We need to tell it to *only* search for
tags that start with `v`, so that it will ignore `releases/latest`.
`spack -V` also would print out unwanted git eror output on a shallow
clone.
- [x] add `--match 'v*'` to `git describe` arguments
- [x] route error output to `os.devnull`
`spack buildcache list` was trying to construct an `Arch` object and
compare it to `arch_for_spec(<spec>)`. for each spec in the buildcache.
`Arch` objects are only intended to be constructed for the machine they
describe. The `ArchSpec` object (part of the `Spec`) is the descriptor
that lets us talk about architectures anywhere.
- [x] Modify `spack buildcache list` and `spack buildcache install` to
filter with `Spec` matching instead of using `Arch`.
- [x] Make it easier to get a `Spec` with a proper `ArchSpec` from an
`Arch` object via new `Arch.to_spec()` method.
- [x] Pull `spack.architecture.default_arch()` out of
`spack.architecture.sys_type()` so we can get an `Arch` instead of
a string.
* Loosen Axom's variants, add shared variant for axom, fix clang/xlf rpath'ing problem on blueos
* Fix flake8
* Add main branch to list of known git branches
fixes#18028
Since now external packages support multiple modules
the correct thing to do is to check if the name of the
*first* module to be loaded contains the string "cray"
`cmake @3.16.3` is the version provided by Ubuntu 20.04. Adding this version here avoids the warning
```
==> Warning: Missing a source id for cmake@3.16.3
```
when using the system `cmake`.
* Spack recipes for ROCm Stage 1 Build components
* fix flake8 errors
* fixes for flake8 errors
* Add a patch for cmake 3.x suport
* Fix rpath issue where hsa-rocr-dev does not allow it to be filled in by spack
* Remove inherited cmake args from comgr
* Make hsakmt-roct compile: no -Werror because with const cast in numa, and actually add the numa dependency
* Remove redundant cmake args which is inherited
* Fix some dependencies
* Fix some python 2.x compatibilities
* Add amd gpu targets to rocfft
* Make comgr a link dep of rocm-dbgapi and remove redundant cmake args
* Remove redundant cmake args
* Remove more redundant cmake args
* Final redundant args
* Use cmake 3.x instead of a fixed version
* Remove random variable
* Use installed rocclr instead of nonexisting directory
* Don't build outside the staging folder
* Deploy some missing cmake target file
* Formatting
* Fix target list
* Properly handle the rocclr dependency
* Formatting
* Fix vermin test
* Make all 3.5.0 package depend exactly on eachother
* Add a few missing link dependencies
* Fix flake8
* Remove some other redundant flags
* Add gcc install prefix for gcc builds of llvm-amdgpu
* review changes for the spack recipes
* Do not hard-code versions
* Fix atmi install
- no more relative rpaths outside of install directory (required patch)
- fix build -> link dependencies
- remove unused build dependency
* Fix flake8 errors
* Remove unused variable and make things python 2.x compatible
* Fix flake8
* Move compiler config from rocfft -> hipcc
* Remove redundant dependency on fftw-api
* Remove redundant import
* Avoid hitting the ROCM_PATH variable altogether with a patch; also just fill in all variables
* Add missing deps z3, zlib and ncurses+termlib to llvm-amdgpu
* Fix perl shebang and add dep
* Fix typo and patch HIP_CLANG_ROOT detection in hip's cmake files
* fixing build failure due z3 and adding zlib for rocgdb
* new changes to add z3,curses dependency for llvm-amdgpu
* fix flake8 error
Co-authored-by: root <root@localhost.localdomain>
Co-authored-by: Harmen Stoppels <harmenstoppels@gmail.com>
This is needed because libcuda is used by the driver,
whereas libcudart is used by the runtime. CMake searches
for cudart instead of cuda.
On LLNL LC systems, libcuda is only found in compat and
stubs directories, meaning that the lookup of libraries
fails.
* Libtool: add spack external find support
* Less specific regex
* match -> search
* Clarify that min returns first alphabetically, not shortest
* Simplify version determination
The modifications in 193e8333fa
introduced a bug in the loading of compiler modules, since a
function that was expecting a list of string was just getting
a string.
This commit fixes the bug and adds an assertion to verify the
prerequisite of the function.
