Experimental Physics and
Industrial Control System

Kevin Peterson <[email protected]> · Thu, 03 Nov 2011 16:20:41 -0500

Ron,

Is this the same problem that you and Andrew were looking into last week?

Kevin

On 11/3/11 3:08 PM, Matt Newville wrote:
> Hi,
> 
> I'm seeing an issue with Channel Access getting "large data arrays"
> and hope someone can provide some insight.  First, this is not an
> issue with EPICS_CA_MAX_ARRAY_BYTES, which is set large enough.
> Rather the issue I'm seeing seems to raise the naive question:
> 
>     Once I do a ca_array_get(type, count, chid, pvalue), when can I
> successfully read the data in pvalue?
> 
> That is, how do I know how long to wait in ca_pend_event() and
> ca_pend_io()?    In most cases of a fast network and scalar values or
> small arrays, values on the order of
>    ca_pend_event(1.e-3);
>    ca_pend_io(1.0);
> 
> seem to work well.   But, if the array is "large" or the network
> "slow", I have a hard time predicting these values, and accessing the
> data may give incorrect values unless I've waited "long enough".
> Using either a slow network or pend_event/io times that is clearly
> "too short",  say,
>    ca_pend_event(1.e-5);
>    ca_pend_io(0.01);
> 
> leaves the data at pvalue incorrect (all zeros).  Trying to access
> this data from python, I can get python to segfault for large enough
> data (say, 4.2M ints).  I'm not sure I fully understand why python is
> crashing -- the segfault happens well after ca_array_get() returns,
> but can happen in ca_pend_event() a short time later.  If I save but
> don't try to access the data in pvalue, or wait "long enough", it
> seems I never have a crash.
> 
> Related questions are:  What is supposed to happen if ca_pend_event()
> and/or ca_pend_io() do time out?  and: Is there a way to tell if
> pend_event() or pend_io() have timed out or if there are events that
> are pending?
> 
> I was hoping to be able to do something like
>    ret_ev = ca_pend_event(1.e-3);
>    ret_io = ca_pend_io(1.);
> 
>    while (ret_io == ECA_TIMEOUT) {
>       ret_ev = ca_pend_event(1.e-3);
>       ret_io = ca_pend_io(1.);
>    }
> 
> Unfortunately, it seems that checking the return values are not so
> helpful, as pend_event() always(?) returns ECA_TIMEOUT and pend_io()
> seems to return ECA_TIMEOUT no more than one time before returning
> ECA_NORMAL.
> 
> For what it's worth, I'm using base 3.14-12.1.   Thanks in advance for
> any insight.
> 
> --Matt Newville <newville at cars.uchicago.edu> 630-252-0431
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System