Re: win32: state of distributed objects

[Alex Perez]

Richard Frith-Macdonald wrote:
> On 19 Apr 2006, at 19:12, Wolfgang Sourdeau wrote:
>
> > Hi,
> >
> >
> > What is the current state of support for distributed objects in Win32?
> > During one of my tests, it seems the reliability of NSMessagePort is
> > unstable while NSSocketPort does not work.
> > Besides I get all kinds of NSLog message about blocking mode not set on
> > certain filehandles.
> >
> > Are there specific issues for Win32 that should be known or does anybody
> > intend to fix things so that they work seemlessly across platforms?
>
> There are no known bugs in distributed objects in win32 other than ...
> 1. The spurious NSLog messages ... which should not effect functionality.

If they are spurious, then why not do something like #ifndef MINGW32 for 
the time being?

> 2. NSMessagePort cannot tell if the other end has disconnected until it 
> tries to send something to it.
>
> So for most purposes, both should work.
> It would be nice to have patches to fix the bogus NSLogs.
> The issue about noticing when a connection has gone away could probably 
> be resolved by implementing a keepalive mechanism to send harmless 
> requests at intervals to check that the remote end is still there (eg 
> rootObjectForProxy requests) when the connection is idle.  We should 
> probably also send an explicit shutdown message to tell the other 
> process that we are going away when a connection is closed cleanly.