Can interruptAccept checks be removed?

[mdavidsaver]

I came across this comment while trying to understand why a plugin unittest I am trying to write hangs on exit in callbackThread::~callbackThread().

asyn/asyn/asynPortDriver/asynPortDriver.cpp

Lines 919 to 922 in db991c2

	/* I thought this would be a temporary fix until EPICS supported PINI after interruptAccept, which would then be used
	* for input records that need callbacks after output records that also have PINI and that could affect them. But this
	* does not work with asyn device support because of the ring buffer. Records with SCAN=I/O Intr must not processed
	* for any other reason, including PINI, or the ring buffer can get out of sync. */

In all likelihood, the hang I am seeing is due to me having missed some preparation. Still, I thought it worth asking. Is the special handling of interruptAccept is still needed? Can some nicer solution can be found? (maybe involving registrars and initHooks)

fyi.

The field(PINI, "RUNNING") was added by epics-base/epics-base@68dbf8a in epics-base 3.14.11 back in 2009.

The initHook API moved into libCom in 7.0.2 (epics-base/epics-base@dc310a4).