Skip any missed timer events updating the timer to now

[kiwichris]

This stops a timer calling the callback for every missed event if the time jumps forward when corrected or moved.

The PR stops the EPICS shell locking up with NTP if no NTP server is available for RTEMS init to use or that request fails. The jump from the RTEMS epoch to the current time resulted the timer making all the missed timer calls which was a lot. The timer and callback looped until finished.

[mdavidsaver]

@kiwichris Could you describe the situation you are seeing leading to a lockup? Which timer and timer queue are involved? From context, I guess the effect is a live lock?

[mdavidsaver]

With the earlier

... (waitfor = queue.front()->timeToRun - now) ...

It would make more sense to me to express this condition as:

if (waitfor < -work->period) {

[JJL772]

@mdavidsaver (per RTEMS Discord discussion) The callback being run due to this bug is epics::pvAccess::ChannelSearchManager::callback()

[mdavidsaver]

@kiwichris I have rebased this PR, and pushed an additional commit. Please have a look and see if you agree with this addition.

For clarify, I reuse the previously calculated waitfor difference. I also opted to re-fetch the time after execution, which I think might help both in the time jump case, and also in a situation where eg. a periodic callback takes longer than one period to execute. Re-fetching the timer afterwards will effectively add a one period "penalty" to allow other threads some time.

[AppVeyorBot]

❌ Build pvDataCPP 1.0.63 failed (commit f797fe82d8 by @kiwichris)

[AppVeyorBot]

❌ Build pvDataCPP 1.0.67 failed (commit 8e926f8386 by @mdavidsaver)

[AppVeyorBot]

❌ Build pvDataCPP 1.0.68 failed (commit e3a9ded505 by @mdavidsaver)

[AppVeyorBot]

❌ Build pvDataCPP 1.0.67 failed (commit e3a9ded505 by @mdavidsaver)

[mdavidsaver]

I think this is enough of an improvement for now.

[kiwichris]

@kiwichris I have rebased this PR, and pushed an additional commit. Please have a look and see if you agree with this addition.

Sorry about the delay responding, I was away last week. Thanks for review and changes. They looks good and I have tested it on my setup and I do not see the start up lock up.