Jacek Caban wrote:
Changelog:
- Added creation of dispinterface header
- Added creation of coclass header
- Added messages when fopen failes
Cool! How far away are we from having widl generate the proxy/stub marshalling code? I started writing some COM test cases last night but it's really hard to test anything non-trivial without a test object that supports marshalling and I doubt Alexandre would want MIDL code in the test suite (or maybe he would, that is what we have to run after all ...)
thanks -mike
Mike Hearn wrote:
Jacek Caban wrote:
Changelog:
- Added creation of dispinterface header
- Added creation of coclass header
- Added messages when fopen failes
Cool! How far away are we from having widl generate the proxy/stub marshalling code? I started writing some COM test cases last night but it's really hard to test anything non-trivial without a test object that supports marshalling and I doubt Alexandre would want MIDL code in the test suite (or maybe he would, that is what we have to run after all ...)
This has nothing to do with proxy/stub marshalling. It is a step towards type library generation though.
Rob
Robert Shearman wrote:
This has nothing to do with proxy/stub marshalling. It is a step towards type library generation though.
Hmm, I don't follow. Widl needs to be able to generate the NDR calls for the proxy/stubs. I've seen some work done on this lately, and was wondering how far this work has come. Widl needs to be able to do typelibs and old-style C marshallers, as IIRC some of the core COM interfaces can't be represented by type libraries.
thanks -mike
Mike Hearn wrote:
Robert Shearman wrote:
This has nothing to do with proxy/stub marshalling. It is a step towards type library generation though.
Hmm, I don't follow. Widl needs to be able to generate the NDR calls for the proxy/stubs. I've seen some work done on this lately, and was wondering how far this work has come. Widl needs to be able to do typelibs and old-style C marshallers, as IIRC some of the core COM interfaces can't be represented by type libraries.
Sorry, I thought you were commenting on the patch. Widl still needs support for generating type format strings (and I remember promising to Mike that I'd have a go at this) to generate any non-trivial proxies/stubs.
Rob
Robert Shearman wrote:
Sorry, I thought you were commenting on the patch. Widl still needs support for generating type format strings (and I remember promising to Mike that I'd have a go at this) to generate any non-trivial proxies/stubs.
Really? I thought that C/NDR based marshallers could marshal anything that MOPs could. But yeah, it'd be great to get widl generating all of these. Or, are you referring to code size?
thanks -mike
-
Mike Hearn -
Robert Shearman