Experimental Physics and
Industrial Control System

"Hill, Jeff" <[email protected]> · Mon, 13 May 2013 17:49:04 +0000

Dirk,

When the compiler version changes, then sometimes the C++ template codes need a complete rebuild. 

Jeff

> -----Original Message-----
> From: [email protected] [mailto:tech-talk-
> [email protected]] On Behalf Of Dirk Zimoch
> Sent: Monday, May 13, 2013 3:50 AM
> To: J. Lewis Muir
> Cc: EPICS
> Subject: Re: Timer Queue crash
> 
> The problem went away after I re-compiled base. Maybe I had a version
> mismatch somewhere. Or a lib was broken during copy. Anyway, it works
> now!
> 
> Dirk
> 
> 
> On 08.05.2013 17:15, J. Lewis Muir wrote:
> > On 5/8/13 4:37 AM, Dirk Zimoch wrote:
> >> Hello,
> >>
> >> I have a problem using epicsTimerQueueActive.
> >>
> >> When I create a time queue with epicsTimerQueueActive::allocate(true)
> >> then release it with epicsTimerQueueActive::release() and after that
> >> call epicsTimerQueueActive::allocate(true) again, I get a segmentation
> >> fault due to access to address 0.
> >>
> >> I see no crash if I do not call release().
> >>
> >> Interestingly, I see this effect only on my armv5teb (aka xscale_be)
> >> embedded Linux system, but not on x86 or on p2020 ppc (big endian, too).
> >>
> >> All accesses are in the same thread (_main_) during device initialization.
> >>
> >> This happens when I use StreamDevice with at least two records
> connected
> >> to a non-existing asynOctet port.
> >>
> >> Any ideas?
> >
> > Hi, Dirk.
> >
> > What version of EPICS Base is this?
> >
> > What compiler was used to build for armv5teb?
> >
> > Does the problem also occur when built with all compiler optimizations
> > disabled (i.e. -O0)?
> >
> > Lewis
> >

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System