Re: widl: let widl understand typedef enum

[Rob Shearman]

Vincent Béron wrote:
> Since the typedef patch to widl, typedef enum foo {A, B} FOO; was
> rejected by get_type_vt (unknown-type: 13).
>
> Changelog:
> A RPC_FC_ENUM16 is a VT_I2.

Are you sure about this? Elsewhere RPC_FC_ENUM16 is equivalent to a 
VT_I4 or VT_UI4 (although it is actually only 16-bits on the wire).

Rob

widl: let widl understand typedef enum

[Vincent Béron]

Le sam 05/02/2005 à 12:53, Rob Shearman a écrit :
> Vincent Béron wrote:
> > Since the typedef patch to widl, typedef enum foo {A, B} FOO; was
> > rejected by get_type_vt (unknown-type: 13).
> >
> > Changelog:
> > A RPC_FC_ENUM16 is a VT_I2.
>
> Are you sure about this? Elsewhere RPC_FC_ENUM16 is equivalent to a 
> VT_I4 or VT_UI4 (although it is actually only 16-bits on the wire).

Where elsewhere? I only found (in widl) proxy.c:marshall_size_arg(), in
which it's size is 2. I haven't looked in the rest of Wine source.

So no, I'm not 100% sure (no hard proof to lean on), but it was lacking
a type and VT_I2 looked like the proper one.

A type16 which is actually 4 bytes would indeed surprise me, especially
when a type32 does exist.

Vincent

widl: let widl understand typedef enum

[Rob Shearman]

Vincent Béron wrote:
> Le sam 05/02/2005 à 12:53, Rob Shearman a écrit :
> > Vincent Béron wrote:
> > > Since the typedef patch to widl, typedef enum foo {A, B} FOO; was
> > > rejected by get_type_vt (unknown-type: 13).
> > >
> > > Changelog:
> > > A RPC_FC_ENUM16 is a VT_I2.
> >
> > Are you sure about this? Elsewhere RPC_FC_ENUM16 is equivalent to a 
> > VT_I4 or VT_UI4 (although it is actually only 16-bits on the wire).
>
> Where elsewhere? I only found (in widl) proxy.c:marshall_size_arg(), in
> which it's size is 2. I haven't looked in the rest of Wine source.

Like I said, on the wire it is 16-bits.
> So no, I'm not 100% sure (no hard proof to lean on), but it was lacking
> a type and VT_I2 looked like the proper one.
>
> A type16 which is actually 4 bytes would indeed surprise me, especially
> when a type32 does exist.

Yeah, it seems strange, but I have a test program here that shows it is 
32-bits application-side for proxies/stubs. If the upper 16-bits are 
non-zero then the proxy code throws a RPC_X_ENUM_VALUE_OUT_OF_RANGE 
exception. I don't know whether it is the same for type libraries, but 
it really needs testing.

Rob

widl: let widl understand typedef enum

[Mike Hearn]

On Sat, 05 Feb 2005 12:37:08 -0600, Rob Shearman wrote:
> Yeah, it seems strange, but I have a test program here that shows it is 
> 32-bits application-side for proxies/stubs.

There's all kinds of brokenness going on in their variant marshalling, eg
some types are mapped seemingly at random to others. I have a list with
some of the details around here somewhere, but I'd totally believe
whatever Rob says on this no matter how weird.