signaling unix threads

[Jukka Heinonen]

Alexandre Julliard:
> It would certainly be nice if we could convince dosvm to stop using
> SIGUSR2, there are lots of other things that could make better use of
> it.  I don't really have a good solution to offer though (except maybe
> getting rid of dosvm altogether ;-)

Well, I'm not going to comment about dosvm usefulness, but
I guess there are quite a few types of asynchronous requests 
Windows processes make to each other, so some kind of multiplexing
system is going to be needed, anyway, and dosvm notification
signal probably could use the same system without adding any
extra complexity.

I was myself thinking about using a single signal 
whose handler just pulls asynchronous requests from
wineserver as an implementation of the multiplexing system.
Locking issues could be fixed using worker threads,
signal safe critical section implementation or
explicit signal blocking in protected code paths.
But since there are quite a few ways for implementing
multiplexing and fixing the resulting locking issues
I probably just wait and see what the implementation
turns out to be.

-- 
Jukka Heinonen <http://www.iki.fi/jhei>