[omniORB] omni::giopRope::match() crash

[Michael Teske]

Hi Duncan,

Am 10.04.15 um 16:01 schrieb Duncan Grisby:
> On Thu, 2015-04-09 at 09:13 +0200, Michael Teske wrote:
> 
>> with this patch it doesn't crash anymore.
> 
> Thanks for your investigation and fix. The code was protecting itself
> against concurrent invocations of acquireClient, but as you say, a
> thread in match() could become a victim of the changed address list.
> 
> Rather than your fix, which causes an extra unlock/relock to occur, does
> this similar patch work for you?  It just pushes the unlocking down into
> filterAndSortAddressList. Now, in the common case that there is only one
> address to resolve, it does only one unlock/relock pair.

Of course. I checked it several times, your patch simply _must_ work. :)
I will use that now. In the improble case that it doesn't work, I'll
write here.

Greetings,
  Michael