Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Tue, 17 Mar 2009 11:36:30 -0500

Hi Oleg,

On Monday 16 March 2009 16:43:26 Oleg Makarov wrote:
>
> It is strange though that it does not work without explicit pvFlush()
> calls. If  "+a" option is not set for snc compiler, one would expect
> synchronous IO to be performed.

I suspect part of the problem might be that you're calling epicsThreadSleep() 
inside the action statements, which completely freezes the sequence program 
during that delay period.  This almost always turns out to be the wrong thing 
to do since the sequence program can't break out of that delay if something 
unexpected happens.  You should instead call the delay() function inside a 
when() statement, which also permits the sequencer to perform other actions 
while it's waiting.  I would recommend changing your program to look more 
like this instead (not tested), which might help:

ss main {
  state init {
    when (pvConnectCount() == pvChannelCount()) {
    } state connected
  }
  state connected {
    when (1)
      strcpy(str_cmd, "I522");
      pvPut(str_cmd);
      pvFlush();
    } state action522
  }
  state action522 {
    when (delay(1.0)) {
      pvGet(str_rsp);
    } state getting
  }
  state getting {
    when (delay(1.0)) {
      printf("str_rsp = %s \n", str_rsp);
      strcpy(str_cmd, "I521");
      pvPut(str_cmd);
      pvFlush();
    } state action521
  }
  state action521 {
    when (pvConnectCount() != pvChannelCount()) {
    } state init
    when (delay(1.0)) {
    } state connected
  }
}

Note that you can easily add the ability to react to channel disconnection in 
the action522 and getting states if you wish.

- Andrew
-- 
The best FOSS code is written to be read by other humans -- Harold Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System