On Mar 20, 2010, at 14:49, Bill Janssen <jans...@parc.com> wrote:
Andi Vajda <va...@apache.org> wrote:
Ah ok, I missed that. The import of config.py can be surrounded by
try/
catch since SHARED is not needed during jcc build time. I'll add that
and simulate the execution path to find other such holes.
Here's one I don't see a good workaround for:
if __name__ == '__main__':
import jcc.__main__
else:
from _jcc import initVM
The import of "_jcc" will fail, too.
How about we put the code to find JAVAHOME (which is really JDKHOME, I
think) in file at the same level as setup.py? It's not needed for
later execution. That way we can just import it, use the code, and
proceed with setup.py.
I was thinking of putting it into a tree sibling to jcc, called
helpers, where all this build stuff could move to - including a mini
fork of setuptools for patch.43 if it really had to come to this.
So helpers.windows has your stuff, helpers.setuptools, the patched
linux shared lib support, and so on.. as needed. That helpers tree
would be part of a sdist archive, via MANIFEST.in but not included
when jcc is installed.
Andi..
Bill
