Re: GsTcpPort (NSDefaultRunLoopMode / _handleWindowNeedsDispaly:)

[Richard Frith-Macdonald]

On Monday, July 8, 2002, at 08:44 PM, David Ayers wrote:

> Hi Richard,
>
> you probably have other things to do before you leave, but in case you 
> do
> find some spare time:
> I've been trying to write a simple App/Server pair to test the basics. 
> My
> first problem was a
> -(oneway void)doOneWay
> invocation. The signature (*type) of this method during
> NSConnection's -forwardInvocation:forProxy: is identical to:
> -(void)doVoid
>
> "v8@0:4"
>
> in other words, _F_ONEWAY isn't set and my App won't continue processing
> until -(oneway void)doOneWay returns. I could provide you with my two
> ProjectCenter projects (App and Server) if you need something for 
> testing.
> Or is this a known to do?
>
> I'll try to look into where this is set, yet I belive this done by the
> compiler? I'm using 3.0.4 right now.

I'm using a pre-release 3.1.1 compiler, and it's working fine in that, 
and
of course it has worked fine in earlier compilers for several years.

I don't recall hearing that it was broken in 3.0.4 (thoguh of course it 
might be)
so perhaps the problem is in your code?

One thing is - you have to be sure that all the code using the oneway 
method can
see the method declaration in order to know that it's supposed to be 
oneway -
normally this means putting a protocol which declares the oneway method 
in a
header file, so both the server and the client can see it.
That's really about the only thing (other than compiler bug) that I can 
think of.

PS. I recommend looking at nsconnection_server.[hm] and 
nsconnection_client.m
in base/Testing