That there is a failure when polling for incoming characters in a TCP message stream implies a rather fundamental error within the environment.
Is IP configured and started in this environment?
Since this is IP, do mechanisms such as ping show connectivity with the ABB PLC controller?
And as others have requested, what's the error (if any)? I might well assume the result here is a no-data condition; that there's nothing arrived from the ABB PLC.
I am assuming the polling loop used is to deal with the fact that TCP is not a datagram protocol and that it can segment arriving character data in interesting ways. I didn't look at the provided C code in detail to see if it included re-assembly, but that you can and will receive partial messages can potentially be a surprise -- the ABB PLC or any other IP data source might well write a single unit of a couple of hundred bytes into its TCP socket, and the data can conceivably arrive as one message, or as a couple of hundred one-byte "messages". Combine this with a read buffer that doesn't get cleared and you can see all manner of "torn" messages and degenerate behavior. (This is not specific to OpenVMS, this is how TCP works, and how it is supposed to work.)
The specification of the address of the address for the ioctl looks a bit weird in the polling routing. I'd expect ioctl( fd, FIONREAD, ninbuf); as it appears that ninbuf is already a pointer. I'd confirm this.
Polling itself must be handled quite cautiously, and that solution can be very sensitive to changes in performance. An AlphaServer DS10 executes its polling loops rather faster than an AlphaServer 1000A series box does, for instance. It might well finish its polling before the ABB PLC sends its data, where the older system was slow enough that the polling windows lined up.
Moving to AST-based code can avoid the need for polling, but that's likely going to involve a reorganization of the existing C code. The expedient solution is to increase the polling count to offset the faster system.
I'd probably enable tracing in the code, such that the trace2 routines are active. If conditionalized through a header file, you can avoid the need to re-comment these routines when you next need to disable tracing.
I spent some time some time ago providing C and PLC-related and communications work. I feel your pain. :-)
Stephen Hoffman
HoffmanLabs
