wild pointers in current named pipe implementation?

[Mike McCormack]

Hi All,

I'll get back to the named pipes patch after the release of Crossover 
Office 2.0.

I'm a little worried that we won't be able to get unix pipes to do all 
the things that Windows pipes can do, but if I want my patch in... :)

Mike

Gregory M. Turner wrote:

> No cause for despair!  IIRC the patch was just rejected because it used
> wineserver calls for the actual pipe i/o, which was considered too
> inefficient, even for wine, by Alexandre ;)  Alexandre suggested using
> connectionless sockets (I'm not sure if he means unix domain or inet sockets?).
> Far be it from me to have an opinion on things I do not understand, but if I
> /were/ to do so, I'd say that I don't see why connectionful unix domain sockets
> ought to be insufficient ;)
> 
> Anyhow, I vaguely recall that Mike replied that it would take some time, but
> that perhaps Alexandre's idea was workable... am I getting this right?  In case
> it will help illuminate some idea-emotocon's above brighter heads than mine,
> here is my shot at converting Mike's patch (which was posted in an almost
> unmentionable format) into a unified diff against CVS wine, courtesy of
> winetreediff: