[omniORB] clientCallTimeoutPeriod bug in multithreading

[Duncan Grisby]

On Thursday 28 August, Thierry DESCOMBES wrote:

>    The corba client "Daq" interrogate the "sensors" (which are corba 
> servers). Each time, the "Daq" can't reach a "sensor", it calls a method on 
> the "Manager" (which acts has a corba server too).
> If the "Daq" interrogates simultaneously 2 or more of the "sensor" servers 
> (using one thread by request), it notices the "Manager" of the 
> impossibility to reach the "sensor" for only the first four, or first seven 
> unreachable servers  (depending on the number of simultaneous requests I 
> made)... and the "Daq" falls in a deadlock.
> However... If the "Daq" doesn't inform the "Manager", all works fine...

Without more details, it's impossible to track down. Please get a
trace with -ORBtraceLevel 25 -ORBtraceInvocations 1 -ORBtraceThreadId
1. If possible, also attach to the frozen "Daq" with a debugger and
get a backtrace of its threads.

Cheers,

Duncan.

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