[medusa] Re: trouble with 19990902 snapshot: asyncore.loop() exists immediately

[Sam Rushing]

jam writes:
> I have been plugging away a little more at this problem, and when I change
> the 'from medusa import ...' line to include 'asyncore', and remove it from
> the top of the script, everything works *fine*. this is odd because I copied
> the 'asyncore.py' file from the distribution into /usr/lib/python1.5, so
> there really shouldn't be a problem, should there?

Ok, I figured it out... the package system is letting you introduce
two different versions of the 'asyncore' module. The one used for
'asyncore.loop()' is not the one that knows about listening sockets.

Change the code thus:

>	hs.install_handler(dh)
>	print asyncore, id(asyncore)
>	print http_server.asyncore, id(http_server.asyncore)
>	asyncore.loop()

And you get this:

| rushing@s...:/usr/src/nm/apps$ python strange.py
| info: adding channel <caching_resolver at 814adf0>
| info: adding channel <http_server at 8149b18>
| info: Medusa (V3.35) started at Thu Sep 2 21:48:29 1999
| Hostname: seattle
| Port:8081
| 
| <module 'asyncore' from '/usr/local/lib/python1.5/asyncore.py'> 135185384
| <module 'medusa.asyncore' from 'medusa/asyncore.pyc'> 135223912

sneaky, but not really the package system's fault: the whole idea of a
package system is that you shouldn't have to worry about name
collisions between unrelated packages.

-Sam