* Add +shared variant, which builds shared library in addition to the
static library.
* Install libraries even when specifying the header-only option
(header-only is just about installing an additional folder).
* An additional make call is not required to build generator
executables (they are built by default).
* Streamlined help-line of each variant.
* Fix Mac platform check for dependency in py-ipython package: 'when'
constraints in Spack directives must be Specs (either a Spec
object or a Spec in string format)
* Fix Mac version check in py-numpy: platform.mac_ver() returns a
3-part string as its first tuple item so the check as written would
never pass; use Spack Version object to simplify check.
* Fix Mac version check in qt package (the check was incorrectly
comparing ints and strings) and use Spack version object to
simplify check.
* Add versions 4.3.0, 4.3.1, and 4.3.2
* The URL format changed after 4.1.4, so this adds a url_for_version
function to choose the URL format based on the version
With this PR, a user can designate older versions of OpenGL as an
external Spack package, and dependents can use that installation
as long as they do not have version requirements.
MacOS currently comes with OpenGL 2.1; there is currently no
'provides' directive in the OpenGL Spack package which specifies a
gl version provided for versions earlier than OpenGL 3.3, so earlier
versions of OpenGL are not considered to provide any version of gl.
This PR records that any version of OpenGL provides 'gl' (which is
sufficient for any package that does not require a specific version
of gl).
* Version 1.3.0 requires python 3.5 or later (no Python 2 support)
* Remove test of scipy.weave import (not available since 1.0)
* Add more-sensitive py-numpy constraints based on py-scipy version
* Replace py-nose dependency with py-pytest
* Add adamjstewart as maintainer
* Rename py-pytorch to py-torch
* Add versions 1.1.0 and 1.0.1
* Define modules to test import of after installation (import_modules)
* py-typing dependency is only needed for older versions of Python
(3.4 or before)
* Newer versions of py-torch depend on newer versions of CUDA
* Add adamjstewart as maintainer
