Re: autosave V4.1 and caRepeater

[Andrew Johnson <anj@aps.anl.gov>]

Tim Mooney wrote:
> I have not been able to figure out why this problem apparently
> goes away when autosave's status PV's are loaded.  Autosave calls
> ca_context_create() regardless of whether or not status PV's exist,
> and it makes CA connections to lots of PV's in any case.
>
> It must be the save_restore thread that's issuing the fork()
> call, since this thread does all of save_restore's CA work.

No, it's the Channel Access library that's calling fork() to create the 
caRepeater when it finds that there isn't one running at the time.  The 
fork() has nothing to do with threads/tasks (which are the same thing) 
inside the IOC.

My guess to what's happening is that asking save/restore to load the 
status PVs is sufficient to change the startup of the various threads 
sufficiently that the video server hasn't got around to opening that 
device before CA gets as far as spawning off the caRepeater.

This problem is therefor nothing to do with autosave at all, it should 
be possible to tickle it by doing some other startup operation instead. 
 The current best solution is, as Jeff has explained, to load 
caRepeater before starting the IOC.

- Andrew
--
English probably arose from Normans trying to pick up Saxon girls.