* TensorFlow: Clean up/simplify the installation, make sure the headers are
installed so that horovod can find them successfully. Fix the 2.0.* builds.
* Backport of 837c8b6b upstream
"Remove contrib cloud bigtable and storage ops/kernels."
Allows 2.0.* releases to build with '--config=nogcp'
* comment regarding tensorflow issue #31187
Co-authored-by: Andrew W Elble <aweits@skl-a-00.rc.rit.edu>
## Summary
This PR updates and improves the Spack package for [UPC++](https://upcxx.lbl.gov).
I'm an LBL employee and developer on the UPC++ team, as well as the maintainer of this Spack package.
### Key Improvements:
* Adding new 2020.3.0 release and support for use of develop/master branches
- Our build infrastructure underwent a major change in this release, switching from a hand-rolled Python2 script to a bash-based autoconf work-alike.
- The new build system is NOT using autotools (nor does it support some of the more esoteric autoconf options), but the user interface for common builds is similar.
* Add explicit support for an MPI optional dependency
- New `mpi` variant enables use of the MPI-based spawner (most relevant on loosely coupled clusters), and the (unofficial) mpi-conduit backend
- This variant is OFF by default, since UPC++ works fine without MPI on many systems, increasing the likelihood first-time Spack users get a working build without needing to correctly setup MPI
* Add support for post-install testing using the test support deployed in the new build infrastructure
* Fix or workaround a few bugs observed during testing
### Status
The new package has been validated with a variety of specs across over seven different systems, including: NERSC cori, ALCF Theta, OLCF Summit, an in-house Linux cluster, and macOS laptops (Mojave and Catalina).
Expose serial/parallel build (MPI), CUDA/OpenMP backends, Clang, and
Ascent bindings.
Interestingly, `warpx +ascent` currently leads to an infinite loop in
the Spack concretizer.
Removed provider_index use of 'import from' and refactored a few routines to a further subclassing of _IndexBase for implementing user defined bindings of provider specs.
Git-based conditions database for HEP and other experiments.
Use latest release version and current master to support Linux and
macOS. Add core known dependencies and conflicts related to C++17
support.
cxxstd variant added to help transitive dependencies, and for future
support for newer standards in future.
* relocate: removed import from statements
* relocate: renamed *Exception to *Error
This aims at consistency in naming with both
the standard library (ValueError, AttributeError,
etc.) and other errors in 'spack.error'.
Improved existing docstrings
* relocate: simplified search function by un-nesting conditionals
The search function that searches for patchelf has been
refactored to remove deeply nested conditionals.
Extended docstring.
* relocate: removed a condition specific to unit tests
* relocate: added test for _patchelf
Our unit tests run many times. Any unit test which actually installs
a package (which involves fetching code on the internet) is a severe
bug because it runs an installation many times (i.e. re-downloading
the same package for each version of Python that we run unit tests
for).
This reverts commit 25893f1, which added tests that install real
packages.
If the Python used by Spack does not include Setuptools, then
'spack test' will fail because Spack's vendored pytest dependency
imports and uses Setuptools in some of its functions. It turns out
that Spack doesn't use the functionality those methods enable, so
this PR removes those functions and thereby allows 'spack test' to
run without Setuptools.
For any Spack test using Spack's YAML configuration, avoid using real
Spack configuration that has been cached by other tests and Spack
startup logic. Previously this was only done for tests using
'mutable_config' (i.e. those which expected to *change* the
configuration of Spack), but in fact all tests that read Spack config
should use it.
This was an issue when running tests within an environment, because
compiler configuration ends up being queried earlier, and the user's
real config "leaks" into the cache. Outside an environment, the cache
is never set until tests touch it, so we weren't seeing this issue.
