[ZODB-Dev] ZEO2, sys.path, start.py and docs

[Jeremy Hylton]

>>>>> "AvdD" == Adrian van den Dries <adriand@flow.com.au> writes:

  AvdD> adriand@baldrick:~/Zope/ZSS$ python lib/python/ZEO/start.py -p
  AvdD> 9999 -D Traceback (most recent call last):
  AvdD>   File "lib/python/ZEO/start.py", line 235, in main
  AvdD> ImportError: No module named StorageServer

  AvdD> Upon inspection, I see that start.py does some sys.path
  AvdD> munging, inserting 'lib/python'.  But later on, it does a
  AvdD> os.chdir(env.var), so this path is now relative to the var
  AvdD> directory.  The problem goes away if I use an absolute path to
  AvdD> start.py or disable ZDaemon.

It looks like the code to munge the path is buggy, but it's not clear
to me when (if ever) it worked correctly.

In all the ZEO installs I've seen/run, the ZEO software is installed
in site-packages or in a directory that is manually added to
PYTHONPATH before the start script is run.

  AvdD> Have I done something wrong or is the README incorrect?

The README is probably a bit out of date.  Either it or the code
should be fixed.

  AvdD> I am using stock Debian/woody python2.1.3.

  AvdD> P.S.  It seems to me that this whole path-munging business is
  AvdD> about finding where Zope lives, which would be solved with a
  AvdD> (suprise, surprise) java-style ZOPE_HOME, set either at
  AvdD> installation time or by the administrator.  Surely this would
  AvdD> be easier and less error-prone than cryptic path-munging
  AvdD> functions?

I agree that modifying sys.path at runtime is tricky.  I wouldn't
be in favor of adding new environment variables.  PYTHONPATH should be
sufficient to fix your problem.

Jeremy