Experimental Physics and
Industrial Control System

"Hill, Jeffrey O" <[email protected]> · Wed, 11 Apr 2012 14:37:18 +0000

Hi Dirk,

> ca_pend_io() waits for search and get without callbacks but probably not
> for monitors? Jeff, please comment.

The ca_pend_io function is for implementing simple, short life span, clients 
like caget which don't need to reconnect if the IOC is rebooted.

The ca_pend_io function will 
o wait for channels created, specifying a nill connection state callback 
	handler function pointer, to connect
o additionally wait for all get requests, not specifying a completion 
	callback handler, to complete
o only wait for above mentioned requests issued after the previous call 
	to ca_pend_io or ca_context_create whichever is later
o will return early specifying an error condition if the timeout expires
o does also issue an implicit flush request before waiting
o does deal with any outstanding ca background activity while waiting
o will return immediately, after initiating the implicit flush, if no 
	such requests are pending

The ca_pend_io request does not wait for subscription request updates 
to complete. That's consistent considering that this type of request 
specifies a callback which is called multiple times. Also, a subscription
request is typically used by long lifespan clients which must deal with 
reconnect issues. 

Jeff

> -----Original Message-----
> From: Dirk Zimoch [mailto:[email protected]]
> Sent: Wednesday, April 11, 2012 1:53 AM
> To: Hill, Jeffrey O
> Cc: Florian Feldbauer; EPICS tech talk
> Subject: Re: What is void* puser in ca_create_channel
> 
> Hi Jeff, Florian,
> 
> I checked my presentation and found:
> 
> * Single-threaded context
>    - Callbacks are only called from within a ca_* function.
>      The program must call ca_pend_io(), ca_flush_io(), ca_pend_event()
> or ca_poll() regularly to receive any data (e.g. in event/idle loop).
> 
> This is wrong:
> ca_flush_io() will not do the job because it does not handle input.
> ca_pend_event() is definitely ok here.
> ca_poll() is ca_pend_event(1e-12) and thus is ok.
> ca_pend_io() waits for search and get without callbacks but probably not
> for monitors? Jeff, please comment.
> 
> Dirk
> 
> 
> Hill, Jeffrey O wrote:
> > Hi,
> >
> > Correct, a single-threaded CA client code must periodically spend some
> time in ca_pend_event() in order that the channels will connect. The
> channels will not transition from a disconnected to a connected state if
> your main thread doesn't periodically call ca_pend_event. The
> ca_pend_event(), among other tasks, arranges for the incoming TCP messages
> to be processed, and calls the user callbacks.
> >
> > The ca_flush_io() call is forcing all of the en-queued put/get/subscribe
> requests to be delivered to the TCP circuits in aggregate, but does not
> necessarily process the incoming messages on the TCP circuits, or call the
> connection callbacks.
> >
> > Jeff
> >
> >> -----Original Message-----
> >> From: Florian Feldbauer [mailto:[email protected]]
> >> Sent: Monday, April 02, 2012 8:05 AM
> >> To: Hill, Jeffrey O
> >> Cc: EPICS tech talk
> >> Subject: Re: What is void* puser in ca_create_channel
> >>
> >> Hey,
> >>
> >> after creating all channels and subscriptions I used ca_flush_io() in
> my
> >> program as
> >> mentioned in the talk "Channel Access Client Programming" from Dirk
> >> Zimoch at the last
> >> EPICS meeting at PSI.
> >> When I use ca_pend_event() instead, the CA is working.
> >>
> >> Best regards,
> >> Florian
> >>
> >>
> >> On 03/30/2012 04:10 PM, Hill, Jeffrey O wrote:
> >>>> Just one other question: Can I use the CA methods ca_create_channel
> and
> >>>> ca_create_subscription within a c++ class?
> >>> Yes, the necessary 'extern"C"' bracketing is in the ca header files so
> >> that its C interface can be called from C++.
> >>>> Any idea based on this few information, or should I post my code?
> >>> If the CA client and its IOC are not on the same subnet then you may
> >> need to add the IOC's address to the EPICS_CA_ADDR_LIST environment
> >> variable for the client. Otherwise, if they are intended to be on the
> same
> >> subnet them make certain that both hosts are configured with the same
> >> subnet mask.
> >>> There are also some additional suggestions in the troubleshooting
> >> section of the CA reference manual.
> >>> Jeff
> >>>
> >>>> -----Original Message-----
> >>>> From: Florian Feldbauer [mailto:[email protected]]
> >>>> Sent: Friday, March 30, 2012 6:29 AM
> >>>> To: Hill, Jeffrey O
> >>>> Cc: EPICS tech talk
> >>>> Subject: Re: What is void* puser in ca_create_channel
> >>>>
> >>>> Thanks for your answers.
> >>>>
> >>>> Just one other question: Can I use the CA methods ca_create_channel
> and
> >>>> ca_create_subscription within a c++ class?
> >>>> The project we are working on should use OO techniques.
> >>>> We need not only an interface to EPICS, but also to a simulation.
> >>>> So we wanted to have an abstract base class, which is used by the
> rest
> >>>> of our project, and  derived classes for the implementation of
> >>>> the individual "control systems". One derived class would contain the
> >> CA
> >>>> for EPICS and another the interface to our simulation.
> >>>>
> >>>> I can compile my test program for this object-structure, but I don't
> >> get
> >>>> a conncetion to my EPICS ioc.
> >>>> The two examples from makeBaseApp.pl are working correctly.
> >>>>
> >>>> Any idea based on this few information, or should I post my code?
> >>>>
> >>>> Best regards,
> >>>> Florian
> >>>>
> >>>> On 03/29/2012 05:19 PM, Hill, Jeffrey O wrote:
> >>>>> Hi,
> >>>>>
> >>>>> The puser pointer you pass to ca_create_channel is saved by the
> >> library
> >>>> with the channel and is later supplied as a puser argument to your
> >> channel
> >>>> connectione state update callback function. There is also a ca
> function
> >>>> "void * ca_puser ( CHID )" that will return the puser pointer that
> was
> >>>> supplied when the channel was created.
> >>>>> Likewise, the puser pointer you pass to ca_create_subscription is
> >> saved
> >>>> by the library with the subscription and is later supplied as a puser
> >>>> argument to your subscription update callback function.
> >>>>>> How does this parameter have to look like?
> >>>>>> Can I use a C-Struct with any member I want to have (even members
> >> which
> >>>>>> are not for the CA, but for another part of my program)? Are there
> >> any
> >>>>>> members which are mandatory? (like value or chid)
> >>>>> A pointer to any data structure can be used. The CA library does not
> >> in
> >>>> any way access, or place any restrictions on the content of, this
> user
> >>>> application private data structure. The user application is of course
> >>>> responsible for allocation, initialization, maintenance, and de-
> >> allocation
> >>>> (proper cleanup) of this data structure.
> >>>>> Use of void pointers passed to callbacks is an old technique which
> is
> >>>> roughly equivalent to a virtual function in modern software
> interfaces,
> >>>> BTW.
> >>>>> HTH, and feel free to send any other questions that you might have,
> >> and
> >>>> I will do my best to answer them.
> >>>>> Jeff
> >>>>>
> >>>>>> -----Original Message-----
> >>>>>> From: [email protected] [mailto:tech-talk-
> >>>> [email protected]]
> >>>>>> On Behalf Of Florian Feldbauer
> >>>>>> Sent: Thursday, March 29, 2012 4:01 AM
> >>>>>> To: EPICS tech talk
> >>>>>> Subject: What is void* puser in ca_create_channel
> >>>>>>
> >>>>>> Dear all,
> >>>>>>
> >>>>>> I currently write a CA Client that should monitor all measured
> >>>>>> temperatures in my experiment
> >>>>>> to regulate the chillers.
> >>>>>> This is my first try to program a CA Client, and I have a question
> >>>>>> concerning this:
> >>>>>>
> >>>>>> The ca_create_channel and the ca_create_subscription methods both
> >> have
> >>>>>> a parameter void* puser. How does this parameter have to look like?
> >>>>>> Can I use a C-Struct with any member I want to have (even members
> >> which
> >>>>>> are not
> >>>>>> for the CA, but for another part of my program)? Are there any
> >> members
> >>>>>> which are mandatory? (like value or chid)
> >>>>>>
> >>>>>> In the camonitor.c example generated with makeBaseApp.pl this puser
> >> is
> >>>>>> just a pointer to a C-Struct containing
> >>>>>> a c-string value, the chid and evid of the channel.
> >>>>>>
> >>>>>> Best regards,
> >>>>>> Florian
> >>>>>>
> >>>>>> --
> >>>>>> -------------------------------------
> >>>>>> | Florian Feldbauer                 |
> >>>>>> | Ph.D. student                     |
> >>>>>> |                                   |
> >>>>>> | Institut für Experimentalphysik I |
> >>>>>> | Ruhr-Universität Bochum           |
> >>>>>> | Universitätsstr. 150              |
> >>>>>> | D-44780 Bochum                    |
> >>>>>> |                                   |
> >>>>>> | Office: NB 2/173                  |
> >>>>>> | Phone:  (+49)234 / 32-23538       |
> >>>>>> | Fax:    (+49)234 / 32-14170       |
> >>>>>> | Web:    http://www.ep1.rub.de     |
> >>>>>> -------------------------------------
> >>>> --
> >>>> -------------------------------------
> >>>> | Florian Feldbauer                 |
> >>>> | Ph.D. student                     |
> >>>> |                                   |
> >>>> | Institut für Experimentalphysik I |
> >>>> | Ruhr-Universität Bochum           |
> >>>> | Universitätsstr. 150              |
> >>>> | D-44780 Bochum                    |
> >>>> |                                   |
> >>>> | Office: NB 2/173                  |
> >>>> | Phone:  (+49)234 / 32-23538       |
> >>>> | Fax:    (+49)234 / 32-14170       |
> >>>> | Web:    http://www.ep1.rub.de     |
> >>>> -------------------------------------
> >>
> >> --
> >> -------------------------------------
> >> | Florian Feldbauer                 |
> >> | Ph.D. student                     |
> >> |                                   |
> >> | Institut für Experimentalphysik I |
> >> | Ruhr-Universität Bochum           |
> >> | Universitätsstr. 150              |
> >> | D-44780 Bochum                    |
> >> |                                   |
> >> | Office: NB 2/173                  |
> >> | Phone:  (+49)234 / 32-23538       |
> >> | Fax:    (+49)234 / 32-14170       |
> >> | Web:    http://www.ep1.rub.de     |
> >> -------------------------------------
> >
> >
> >

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System