Re: StreamDevice ioc dying

[Dirk Zimoch <dirk.zimoch@psi.ch>]

Hi Mike,

you can try to switch on debugging (var streamDebug 1) to see what's going on inside StreamDevice.
This may produce a lot debugging output. Better only switch it on after iocInit.

StreamDevice should be able to handle null characters. Of course there may be a bug inside
StreamDevice that crashes the ioc under certain conditions. I need more detailled information about
the error conditions.

Can you send me your records and protocol file? What is your operating system?

I guess the "Killed" message comes from the operating system. It does not come from StreamDevice or
asyn or EPICS base. You can try to switch on core dumps (ulimit -c unlimited) before starting your
IOC and analyse it with a debugger.

Dirk

Mike Forster wrote:
> Hi folks,
>
> We've got a new ioc that controls a power supply over RS-232 with 
> StreamDevice.  When the control power on the power supply is cycled, the 
> ioc dies with only the message "Killed".  (Specifically the ioc dies 
> when the PS is turned back on.)
>
> Turning on asyn tracing shows that upon turning the PS back on,  some 
> extra data like null characters are read in through StreamDevice.  
> (Perhaps data that was interrupted?)  On reading this, the ioc dies.  
> The mismatch exception handling is turned on, but I'm not quite sure 
> what command to put inside the handler to deal with this problem. I've 
> tried several ideas but none seem to work.
>
> Using epics 3.14.8.2 <http://3.14.8.2>, streamDevice 2-1, and asyn 4-6.
> (Also tried with 3.14.7, 2-1, and 4-3)
>
> Any ideas on what could be going on that the ioc can't handle?
>
> Thanks,
> Mike Forster
> Wilson Lab
> Cornell University
>
> --
> .

--
Dr. Dirk Zimoch
Swiss Light Source
Computing and Controls
Paul Scherrer Institut
phone +41 56 310 5182
fax   +41 56 310 4413