[omniORB] omniORB and global servant activator queue

[Duncan Grisby]

On Wednesday 19 July, Teemu Torma wrote:

> I noticed from a deadlock that poa has global servant activator queue, 
> resulting to deadlock if servant activator would trigger another 
> servant activator in another poa.  
> 
> I can not find any documentation saying this should be prohibited.  Is 
> this behaviour intentional or just accident?

The POA specification doesn't say that servant activators in different
POAs should not run concurrently, but it doesn't say that they must do
either. I'm pretty certain that omniORB has one single queue for a good
reason, to avoid a race condition leading to a deadlock, but I can't
actually remember the circumstances right now.

Cheers,

Duncan.

-- 
 -- Duncan Grisby         --
  -- duncan at grisby.org     --
   -- http://www.grisby.org --