[omniORB] Re: omniORB 4.0.7 and omniORBpy 2.7 soon (Duncan Grisby)

[David Fugate]

Duncan Grisby wrote:
> On Tuesday 20 December, David Fugate wrote:
> 
>> Hello Duncan, the necessity for a patch to omniidl's Python backend is
>> discussed here:
>> 	http://www.omniorb-support.com/pipermail/omniorb-list/2005-October/027202.html
>>
>> while the actual patch can be found here:
>> 	http://www.ras.ucalgary.ca/~dfugate/python.py
> 
> Sorry, I don't think I can include your patch. Firstly, I think it is an
> unusual situation to have generated stub files in more than one place,
> especially when the IDL files open the same module names. Certainly
> nobody has mentioned the problem before.
It is pretty common in my opinion to install the stubs in more than one 
location (e.g., this is particularly useful in cleanly testing small 
patches to large systems), but I do agree with you completely that it's 
a bad practice to do this when reopening IDL modules across different 
IDL files. Regardless of my opinion on this, a decision has been made 
that ALMA software must support this practice.

> Second, in the common case that omniidl builds stubs in one place then
> they're installed somewhere else, your modification can lead to imports
> being added when they shouldn't be, especially if IDL changes. Finally,
> your code won't work on Windows because it assumes the path separator is
> ":" -- on Windows it's ";".
As for your first point, I argued this as well in the beginning. The 
counterargument  I was given was that if someone makes the IDL changes, 
they basically need to keep track and be very careful of where they're 
installing packages. Also, they may need to recreate areas of 
$PYTHONPATH. Other than that, I've fixed the path separator issue.

Dave