You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Seen in #61, there are a couple of problems with the current installation strategy.
our installation of Chapel (a core dependency of this project), is an additional
step outside of the python setup.py install and pip install pyChapel steps followed
to install normal Python modules. Is there a way to modify setup.py so that it performs
this installation automatically, without the user needing to perform this action themselves?
the recommended path (and that followed by our test_python.bash script) to
install Chapel is not robust to when multiple configurations of Chapel are available to the
user. It will try to grab all files in the lib directory, instead of just the one for the current
configuration, leading to potentially overwriting the settings desired with other Chapel
configurations (including those that pyChapel does not currently support). Improving
the handling of this is likely related to Expand Chapel settings that work with this feature #7
Other, related issues:
Investigate and improve tie to Chapel dependency. Can we change the build of pyChapel
so that it can find the Chapel library based on an environment variable setting (preferably
one that Chapel already relies upon, such as CHPL_HOME) instead of forcing us to create
a copy or a link to the library file in a specific location?
The text was updated successfully, but these errors were encountered:
I am assuming this is now the place to discuss the issue of the copying of the chapel libraries?
(Sorry I have been absent, I shall restart efforts on the Python 3 stuff shortly.)
I guess the first question I had ages ago was "Why do the files have to be copied? Why is it not possible to use environment variables to achieve the access to the built libraries?" I believe I then used symbolic links instead of copying. Building and installing PyChapel, requires (or required) a rebuild of Chapel so an update to the "copied" files anyway.
I don't know the answer, but I agree with you that we shouldn't have to copy the files. I would imagine it was more chosen as one way to solve the problem of "Where is my Chapel install?" rather than because it was viewed as the only way to find the Chapel install. Updating the issue description to include that as an action item
Or, maybe,... a "deployed" PyChapel with the copied files makes a directory that can be tarred or zipped up and it makes a complete distribution. I had previously been thinking only in terms of development. Putting on a packager/DevOps mindset it could be this is the first stage of making a release of a self-contained thing with only libc as a dependency.
Seen in #61, there are a couple of problems with the current installation strategy.
step outside of the
python setup.py install
andpip install pyChapel
steps followedto install normal Python modules. Is there a way to modify setup.py so that it performs
this installation automatically, without the user needing to perform this action themselves?
install Chapel is not robust to when multiple configurations of Chapel are available to the
user. It will try to grab all files in the lib directory, instead of just the one for the current
configuration, leading to potentially overwriting the settings desired with other Chapel
configurations (including those that pyChapel does not currently support). Improving
the handling of this is likely related to Expand Chapel settings that work with this feature #7
Other, related issues:
so that it can find the Chapel library based on an environment variable setting (preferably
one that Chapel already relies upon, such as CHPL_HOME) instead of forcing us to create
a copy or a link to the library file in a specific location?
The text was updated successfully, but these errors were encountered: