Experimental Physics and
Industrial Control System

Hi Xuan,
If you run your program directly, without gdb, does it show
a CA.Client.Exception w/ Context of "ca_search"?

If so, that's the result of the SEVCHK macro that calls ca_pend_io
throwing an exception because one of the PV's didn't connect.
I'm not sure why gdb shows that location in the backtrace, but
I saw the same thing w/ both gdb and valgrind.    Possibly something
to do with that being the next line of code to execute if the exception
hadn't been thrown.

It you're not getting a Context of "ca_search", please send the full
output message.

Cheers,
- Bruce

On 10/30/2014 06:06 PM, 吴煊 wrote:

Hi All,

    I write a simple CA program to read waveform data. If I monitor less than 41 waveform PVs(100 elements of float type and 25Hz scan rate), it works fine. If I monitor 42 or more(modify "CH_NUM" in archive.c), then segmentation fault occured. I used gdb to debug it.
(gdb) bt
#0  ca_element_count (pChan=0x31346677) at ../oldChannelNotify.cpp:634
#1  0x0000000000400fe3 in main (argc=<value optimized out>, argv=<value optimized out>) at archive.c:138
(gdb) where
#0  ca_element_count (pChan=0x31346677) at ../oldChannelNotify.cpp:634
#1  0x0000000000400fe3 in main (argc=<value optimized out>, argv=<value optimized out>) at archive.c:138
(gdb) l
629	/*
630	 * ca_element_count ()
631	 */
632	arrayElementCount epicsShareAPI ca_element_count ( chid pChan )
633	{
634	    epicsGuard < epicsMutex > guard ( pChan->cacCtx.mutexRef () );
635	    return pChan->io.nativeElementCount ( guard );
636	}
637	
638	/*

It seems that there is something wrong with "ca_element_count" and "epicsGuard < epicsMutex > guard ( pChan->cacCtx.mutexRef () );"
Then I used fixed count number 100 instead of using "ca_element_count", segment fault as well.
(gdb) bt
#0  ca_create_subscription (type=16, count=100, pChan=0x31346677, mask=5, pCallBack=0x401250 <monitor>, pCallBackArg=0x7fffc71f9470, monixptr=0x0) at ../oldChannelNotify.cpp:560
#1  0x0000000000400fd9 in main (argc=<value optimized out>, argv=<value optimized out>) at archive.c:141
(gdb) l
555	    if ( mask & ~maskMask ) {
556	        return ECA_BADMASK;
557	    }
558	
559	    try {
560	        epicsGuard < epicsMutex > guard ( pChan->cacCtx.mutexRef () );
561	        try {
562	            // if this stalls out on a live circuit then an exception
563	            // can be forthcoming which we must ignore (this is a
564	            // special case preserving legacy ca_create_subscription

I guess there is something to do with "epicsGuard < epicsMutex > guard ( pChan->cacCtx.mutexRef () );"  However, I don't know whether my code having bug or something else wrong. Any help will be appreciate

I'm using RedHat 6.5_x64 and EPICS base 3.14.12.1

Cheers

Xuan Wu
IHEP

--
Bruce Hill
Member Technical Staff
SLAC National Accelerator Lab
2575 Sand Hill Road M/S 10
Menlo Park, CA  94025

References:

read waveform data using ca_create_subscription 吴煊

Navigate by Date:

Prev: RE: Compiling SDDS problem Anderson, Janet B.

Next: Re: Sequencer: are variables shared? Damiano Bortolato

Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  <2014>  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024 

Navigate by Thread:

Prev: read waveform data using ca_create_subscription 吴煊

Next: Re: Re: read waveform data using ca_create_subscription 吴煊

Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  <2014>  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024