DAQmx: Improve IOC

[DominicOram]

As a developer I would like the DAQmx IOC to be maintainable. It currently has a number of major issues:

• It uses the old C API for asyn rather than C++ one
• It has a very confusing state machine
• It is ~4500 lines of sparsely documented C
• It uses an older version of the DAQmx API (@KathrynBaker knows more)

Note this could be starting the driver from scratch implementing only the things we require or could be a gradual rewrite.

Acceptance Criteria

• There are system tests that have a good coverage of as many of the DAQmx modes as possible
• The IOC uses the C++ asyn interface
• The state machine logic is simplified and documented
• The IOC uses the latest version of the DAQmx API

[FreddieAkeroyd]

Would we try and keep compatibility with DAQmxBase in case we ever went to Linux?

[DominicOram]

TBD, the other thing to take into account is if we only implement the features we want and throw the rest of the codebase away then the EPICS community has effectively lost support for the rest

[KathrynBaker]

NI-DAQmx Base hasn't been updated since 2015, there are versions there for both Linux and Windows using NI-DAQmx Base. Only NI-DAQmx is being actively updated from what is visible on the website and was updated in 2019.

There are some NI instructions at http://www.ni.com/tutorial/5409/en/ which might help if we do rewrite it.

NI-DAQmx Base was originally being used in relation to Linux, https://www.ni.com/en-gb/support/documentation/supplemental/18/daqmx-for-desktop-linux-tools.html might have more useful information.

There is a python wrapper for the C API for DAQmx, which might also be helpful if it provides a suitable encapsulation of the two functions: https://nidaqmx-python.readthedocs.io/en/latest/

If we want to maintain Linux compatibility it might be worth reaching out to our NI account manager to ask about non-production options that might be available, although there may be caveats in relation to that.