preliminary ev_async implementation

Tony Arcieri tony at
Fri Feb 8 08:55:02 CET 2008

On Feb 3, 2008 7:47 PM, Tony Arcieri <tony at> wrote:

> One thing: the docs show ev_async_send() taking one argument (the ev_async
> struct) which makes sense.  But in the CVS version, it takes both the loop
> and the ev_async struct.  Shouldn't the ev_async struct store the loop it
> was started on?

Well, after having problems with async events apparently being gobbled and
sometimes managing to put the loop into a state where it ceased to block (if
callbacks were firing, I couldn't tell what they were), I decided to do a
Ruby implementation using my existing wrappers for ev_io watchers and pipes
for the time being.

This will let me drop ev_async in whenever it's in a stable state while
still preserving API compatibility.  Let me know if you've spotted any bugs
in it and I can try it out again.  If there's a portable C version available
I'll certainly switch to it.

Tony Arcieri
ClickCaster, Inc.
tony at
-------------- next part --------------
An HTML attachment was scrubbed...

More information about the libev mailing list