[omniORB] Backlog for UNIX domain sockets

[Duncan Grisby]

On Fri, 2024-10-18 at 15:41 +0000, Serguei Kolos via omniORB-list
wrote:


> I have been using omniORB with the UNIX domain sockets transport and
> periodically encountered connection failures when running many
> clients simultaneously. This appeared to be caused by the small
> backlog number (5) of the listen call that is used in unixEndpoint.cc
> file (line 205). Changing this value to 1024 has fixed the problem. 
> May I suggest to change this number to the standard
> orbParameters::listenBacklog value like this is done for TCP
> transport?

Yes, that's a sensible thing to do. I've done that.

Thanks,

Duncan.

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