[pywin32-bugs] [ pywin32-Patches-1457673 ] dispids can go wrong when using typelibs

[SourceForge.net <no...@so...>]

Patches item #1457673, was opened at 2006-03-24 11:18
Message generated for change (Comment added) made by tonyroberts
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=551956&aid=1457673&group_id=78018

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
>Group: Unstable (example)
Status: Open
Resolution: None
Priority: 5
Submitted By: Tony Roberts (tonyroberts)
Assigned to: Mark Hammond (mhammond)
Summary: dispids can go wrong when using typelibs

Initial Comment:
Hi,

In DesignatedWrapPolicy._wrap_, if a typelib is 
available it's used to populate _name_to_dispid_ the 
other dispid maps. However, the case of the 
methods/properties in the type library don't always 
match the python code as the midl compiler may 
produce a typelib with differently cased names to the 
original idl file.

When iterating over _public_attrs_, if a dispid 
already exists for name.lower(), it should be used 
instead of a new dispid - otherwise there's the 
posibility of having two dispids for what should be 
the same attribute. When iterating over 
_public_methods_, it's not always enough to check 
that name.lower() already exists in _name_to_dispid_ 
as the corresponding name in _dispid_to_func_ may 
have the wrong case.

cheers,
Tony.

----------------------------------------------------------------------

>Comment By: Tony Roberts (tonyroberts)
Date: 2006-09-03 11:55

Message:
Logged In: YES 
user_id=1484825

Hi Mark,

it's a slightly strange one - took me a while to remember 
exactly how to reproduce it!

Anyway, if you take a look at the attached file test.idl 
you'll see ITest1 has the method 'test', and ITest2 has 
the method 'Test'. If you compile this idl file and look 
at the resultant typelib, you should see that both 
interfaces have the method 'test' - so ITest2.Test has 
been renamed ITest2.test. I've just tested this using 
tlbrowse.py, but I think any typelib browser would show 
the same thing.

I've only tried this with midl 5.01 (the one that ships 
with vc6), possibly the problem doesn't exist with later 
versions.

As you say, VB makes no case distinction, but that 
shouldn't matter the way the python code works at the 
moment, it's really only when the case in the typelib 
isn't what you'd expect that you get this particular 
problem.

Let me know if you need anything else.

cheers,
Tony.

----------------------------------------------------------------------

Comment By: Mark Hammond (mhammond)
Date: 2006-08-30 09:35

Message:
Logged In: YES 
user_id=14198

Sorry for the delay in looking at this.  Can you give me an
idea of when MIDL may change the case of elements?  I'm
vaguely worried about the case where an object may want to
implement 2 interfaces, each with an attribute that differs
only by case - the current behaviour would be correct in
that case.  But given VB makes no case distinction, I can't
defend that argument too far - but would like an idea of
when MIDL can screw up like that...

Thanks!

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=551956&aid=1457673&group_id=78018