Does anyone have any objections?
Thanks for that input Mick. If I don't hear of any issues from others
I'll make this change in 1.2.1 which should freeze soon. I'm waiting
for someone to get back to me on one of my commentary changes before I
ask Gregory to finalize it.
Pete
Michael Curran wrote:
> Hi,
>
> Currently the IAccessible2 IDL imports the types and constants from
> oleacc (IAccessible etc) using the IDL import statement. This is correct
> in that this allows further types in the IDL file (such as
> IAccessible2) access to the types so that they can inherit from them.
>
> However, This also currently means that if you generate a type library
> from the IAccessible2 IDL, it will then contain *all* of the oleacc
> library inside it.
>
> This so far really hasn't been too much of a problem. However the
> disadvantages are:
> *The IA2 type library may be larger than it needs to be (truthfully only
> by about 3 kb)
> *The IA2 type library statically uses a version of oleacc's type library
> at the time of compile, rather than the version on the user's system.
>
> The actual way I noticed this was while testing something in NVDA (using
> Python comtypes):
> For a long time now NVDA has done an extra queryInterface to IAccessible
> even if the pointer is an IAccessible in the first place. This of course
> is redundant, so I tried taking it out.
> However, then a line of code lower down kept failing causing NVDA to not
> work at all. This line was asking if the pointer was in deed an
> IAccessible instance.
> if not isinstance(pacc,IAccessible): raise ValueError("pacc %s is not an
> IAccessible"%pacc)
>
> The ValueError was always raised. Yet the message in the ValueError
> printed showed me it was an IAccessible: "<POINTER(IAccessible) at
> 0x38bcf6" for example.
>
> After further debugging I ended up finding out that the definition of
> IAccessible that my Python module was using was the one from its
> Pythonified IAccessible2 type library, not the one from the Pythonified
> oleacc type library (which the actual pointer was using).
>
> In reality I could fix this very easily in NVDA by importing the oleacc
> type library after the IAccessible2 type library, which would allow the
> oleacc definition of IAccessible to clobber the IAccessible2 one. But, I
> think the most logical way would be to fix the IDL.
>
> The fix is very simple:
> Inside the Library block, there is already a line that says:
> importlib ("stdole2.tlb")
> Under this line, just put:
> importlib ("oleacc.dll")
>
> This instructs midl to link the generated Ia2 tlb with oleacc.dll, which
> contains oleacc's type library.
>
> Note that you still need to keep the other import statement at the top
> of the idl, that imports oleacc.idl, as midl itself still needs these
> definitions.
>
> Thanks
> Mick
>
> --
>
> --
> Michael Curran
> Email: [email protected]
> website: http://blog.kulgan.net
>
> _______________________________________________
> Accessibility-ia2 mailing list
> [email protected]
> https://lists.linux-foundation.org/mailman/listinfo/accessibility-ia2
>
>
--
*Pete Brunet*
a11ysoft - Accessibility Architecture and Development
(512) 238-6967 (work), (512) 689-4155 (cell)
Skype: pete.brunet
IM: ptbrunet (AOL, Google), [email protected] (MSN)
http://www.a11ysoft.com/about/
Ionosphere: WS4G
_______________________________________________
Accessibility-ia2 mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/accessibility-ia2
