Re: WIDL: use __RPC_FAR attribute for pointers in proxies

[Alexandre Julliard]

Mike McCormack  writes:
> ChangeLog:
> - use __RPC_FAR attribute for pointers in proxies

Is this really useful?  __RPC_FAR will always be an empty define, even
when building on Windows, and I doubt we'll want to add 16-bit support
to widl...

WIDL: use __RPC_FAR attribute for pointers in proxies

[Mike McCormack]

Alexandre Julliard wrote:
> > - use __RPC_FAR attribute for pointers in proxies
>
> Is this really useful?  __RPC_FAR will always be an empty define, even
> when building on Windows, and I doubt we'll want to add 16-bit support
> to widl...

Yes, because it makes it easier to compare the output of midl and widl. 
  Additionally, I hope that projects other than Wine can use widl too 
one day...

Mike

WIDL: use __RPC_FAR attribute for pointers in proxies

[Alexandre Julliard]

Mike McCormack  writes:
> Yes, because it makes it easier to compare the output of midl and
> widl. Additionally, I hope that projects other than Wine can use widl
> too one day...

Frankly I can't think of any project that would need that. It's just
historical noise, far pointers have fortunately gone the way of the
dinosaur. I'd suggest doing a s/__RPC_FAR// on the midl output before
comparison and avoiding it in Wine. In fact we explicitly forbid use
of FAR in Wine, and __RPC_FAR should probably be treated the same
way.

WIDL: use __RPC_FAR attribute for pointers in proxies

[Mike Hearn]

> Yes, because it makes it easier to compare the output of midl and widl. 
>  Additionally, I hope that projects other than Wine can use widl too one 
> day...

Yeah, I'd vote for *not* following the exact output of MIDL in WIDL 
actually.

- MIDL produces code that could compete in the worlds ugliest code
   competition.

- The casts it uses make new versions of gcc cry

- There's a lot of junk in there we probably don't need (stubless proxy
   support etc)

I'd rather we produced clean, easily readable Wine style code rather 
than copying the output of MIDL exactly. The code isn't magic, it should 
not be difficult to simply compare the meaning of the code if we're not 
sure it matches up.

thanks -mike

WIDL: use __RPC_FAR attribute for pointers in proxies

[Robert Shearman]

Mike Hearn wrote:
> > Yes, because it makes it easier to compare the output of midl and 
> > widl.  Additionally, I hope that projects other than Wine can use 
> > widl too one day...
>
> Yeah, I'd vote for *not* following the exact output of MIDL in WIDL 
> actually.
> -
>
> MIDL produces code that could compete in the worlds ugliest code
> competition.
> -
>
> The casts it uses make new versions of gcc cry
> -
>
> There's a lot of junk in there we probably don't need (stubless proxy
> support etc)

Stubless proxies are the only ones that are secure.
> I'd rather we produced clean, easily readable Wine style code rather 
> than copying the output of MIDL exactly. The code isn't magic, it 
> should not be difficult to simply compare the meaning of the code if 
> we're not sure it matches up.

Agreed.

Rob

WIDL: use __RPC_FAR attribute for pointers in proxies

[Mike Hearn]

> Stubless proxies are the only ones that are secure.

Secure? This is news to me, how is interpretive marshalling more secure 
than generated code based marshalling? Or are we talking about different 
things ...

WIDL: use __RPC_FAR attribute for pointers in proxies

[Robert Shearman]

Mike Hearn wrote:
> > Stubless proxies are the only ones that are secure.
>
> Secure? This is news to me, how is interpretive marshalling more 
> secure than generated code based marshalling?

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/midl/midl/_...

Rob

WIDL: use __RPC_FAR attribute for pointers in proxies

[Mike Hearn]

> http://msdn.microsoft.com/library/default.asp?url=/library/en-us/midl/midl/_...

Interesting. I wonder how this affects the output.

I think we should worry about this sort of thing when our DCOM code is 
actually mature enough that we can use it, though.

WIDL: use __RPC_FAR attribute for pointers in proxies

[Robert Shearman]

Mike Hearn wrote:
> > http://msdn.microsoft.com/library/default.asp?url=/library/en-us/midl/midl/_...
>
> Interesting. I wonder how this affects the output.

It changes the conformance descriptor in the type format string from 4 
bytes to 6 bytes.
> I think we should worry about this sort of thing when our DCOM code is 
> actually mature enough that we can use it, though.

I should really clean up my NdrClientCall2 work at some point.

Rob

WIDL: use __RPC_FAR attribute for pointers in proxies

[Mike McCormack]

> Yeah, I'd vote for *not* following the exact output of MIDL in WIDL 
> actually.

I don't want to make WIDL produce exactly the same output as MIDL, but 
it's nice to have some of the lines line up so it can be compared easily.

Mike