Age | Commit message (Collapse) | Author | Files | Lines |
|
strategy is set correctly
|
|
|
|
platform dictionary
|
|
|
|
|
|
are used my operating system subclasses
|
|
|
|
|
|
abstract.architecture.target.concrete
|
|
|
|
|
|
of spec.target
|
|
of the arch tuple
|
|
so any time we need access to target we do architecture.target.
|
|
|
|
|
|
checks whether each piece of string is a valid platform, operating system and target. If the operating system or target are none it will use the defaults. Updated the documentation for that method. One thing that bothers me is how to account for the multitude of operating systems when cross compiling. If someone wants to compile with another operating system not found on current platform. How can spack check to see if it is valid?
|
|
weren't going to be used anyway
|
|
architecture
|
|
way Platforms are compared. Created a mini inherited class named Arch that inherits from namedtuple. The reason for this is to override the namedtuple __str__ method to make arch tuples look prettier when printed out. Rather than Arch(platform= ... ) it prints to platform-os-target which makes directory paths to packages look cleaner.
|
|
|
|
namedtuple. Also added some TODO comments to later consider changing how architecture is handled
|
|
operating systems are set as class default_os
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
operating system
|
|
detected by linux and darwin and manually set by cray and bgq"
This reverts commit 70088be24b2ed34076f7f5292a2a465a8655a886.
|
|
by linux and darwin and manually set by cray and bgq
|
|
|
|
compilers and then spack was giving multiple compiler match errors
|
|
syntax errors
|
|
|
|
|
|
|
|
|
|
|
|
compilers returned.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
the proper strategy
|