RE: caget segment fault

["Jeff Hill" <johill@lanl.gov>]

> I am trying to compile epics base 3.14.8.2 and 3.14.9 on redhat 8.0
> After compilation, caget <PV> (most of the time) returns a value and
> immediately segment fault.
> Compilation of 3.14.6 works great.
> Any idea what that could be?

This might be an issue with proper orderly shutdown - possibly something is
different with process rundown with redhat 8.0. 

To get the problem fixed may require debugging the process while it shuts
down (a weak area in the past with the Linux debugger) and or looking at the
crash dump file with the debugger. A common cause of such problems can be
threads that don't have an orderly shutdown, and that might continue to use
resources that are being destroyed, when the process exits. In theory, all
of the CA related threads are shutdown when one deletes the client context,
but there are other threads that might be started implicitly when using the
libraries in libCom. To make further progress may require some sessions in
the debugger. If you can capture the stack trace of the thread that crashes
and email it, I might see the cause of the trouble.

> Make sure the running caRepeater application is from the same epics  
> version as caget.  You could be running the 3.14.6 caRepeater when  
> using the other two releases.

I am at this time unaware of any issues where the version of the caRepeater
deamon could have any impact on caget that might cause it to crash. 

The only scenario I can think of that would cause such behavior would be
possibly where the R3.13 caRepeater has pulled into memory the R3.13 ca and
or Com shareable library, and somehow the R3.14 caget application is using
the wrong shareable library? Nevertheless, this seems unlikely because I
presume that the Linux shareable library version numbering would prevent
such catastrophes.

Jeff

-----Original Message-----
From: tech-talk-bounces@aps.anl.gov [mailto:tech-talk-bounces@aps.anl.gov]
On Behalf Of Geoff Savage
Sent: Monday, July 09, 2007 1:26 PM
To: Emmanuel Mayssat
Cc: epics
Subject: Re: caget segment fault

Make sure the running caRepeater application is from the same epics  
version as caget.  You could be running the 3.14.6 caRepeater when  
using the other two releases.

For more info on caRepeater see Ralph's email on "beacons" from this  
morning.

Geoff

On Jul 9, 2007, at 1:54 PM, Emmanuel Mayssat wrote:

> Hellom
>
> I am trying to compile epics base 3.14.8.2 and 3.14.9 on redhat 8.0
> After compilation, caget <PV> (most of the time) returns a value and
> immediately segment fault.
> Compilation of 3.14.6 works great.
> Any idea what that could be?
> --
> Emmanuel
>