[omniORB] omniORBpy: __init__.py installed in site-packages

[Thomas Lockhart]

On 06/13/2011 10:48 AM, Floris Bruynooghe wrote:
> On 5 June 2011 22:16, Thomas Lockhart<lockhart at fourpalms.org>  wrote:
>> On 06/05/2011 01:19 PM, Floris Bruynooghe wrote:
>>> It seems that __init__.py gets installed in the toplevel
>>> site-packages. =C2=A0While generally not very harmful I'm sure it can
>>> hardly be deemed as correct...
>> Isn't this required to do a clean install into an alternate location?
>> Breaking that installation is probably not a step forward.
> I doubt it.  omniorbpy needs to make it's own directories packages but
> it can't ensure that the directory containing those packages ends up
> on sys.path anyway.  And turning the directory on sys.path into a
> package itself just seems wrong.
OK, I see that my top-level directories on sys.path do not need an 
__init__.py file to be recognized as the top-level of a python library 
installation. Sorry, I should have checked that before responding. So I 
don't see that is is necessary for omniORBpy (or omniORB; did you look 
there too? There is some python code to support omniidl).

Duncan, does this look reasonable to apply to the main tree? Or should 
we put this in as a patch for RPM building etc. for awhile first?

                                        - Tom