Experimental Physics and
Industrial Control System

Michael Davidsaver via Core-talk <[email protected]> · Tue, 13 Aug 2019 20:30:28 -0700

On 8/13/19 2:39 PM, Mark Rivers via Core-talk wrote:
> I had a problem with pvget today.
> 
>  
> 
> I am running an areaDetector IOC that runs qsrv and also the NDPluginPva.  The IOC is running 7.0.3, linux-x86_64 on Ubuntu 18.
> 
>  
> 
> On this IOC this is the output of pvasr:

If this happens again, it would be interesting to run "pvasr 5"
which will list open connections.  So this would give information
about whether the client is still connected, and how it identified
itself.

This is relevant because Base 7.0.3 includes the first backwards
compatible PVA protocol change.

In my testing, I've found that older and newer peers work in all
the cases I tested (I think all 6).  However, it is possible that
I missed something.

https://github.com/epics-base/pvAccessCPP/pull/144

Aside from that, I've found (and become quite frustrated with) the
race conditions when a PVA peer (client or server) is shutting down.

Apparently some recent changes I've made have aggravated a preexisting
crash on exit.  This crash has been possible for some time, with 7.0.3
it became more likely.

https://github.com/epics-base/pvAccessCPP/issues/152

...
> I tried running the pvget –p pva command on another Linux system (Centos 7, base 7.0.3), talking to the same IOC.  It also hung up the same way, I had to kill it with -9.

If this happens again.  Please capture a stack trace before killing the process.
One way to do this is with gdb.

> $ gdb
> (gdb) attach <pid>
> (gdb) thread apply all backtrace

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System