2.7.15

They were not faulty when Pncconf built them originally.
PID behavior was changed (I guess nobody thought to check how the configuration programs used them)

cmorley wrote: They were not faulty when Pncconf built them originally.

The HAL shown earlier had a signal connected to pid.N.command-deriv, but there was no driver for that signal.

While command-deriv was broken and ignored, that didn't matter, but I don't think it was correct.
It's curious that the command-vel output from motion (which I think is the correct signal to drive command-deriv) lives in the "debugging pins, might be removed" section.

Yes that was a bad choice for the PID component. but when pncconf originally was built that was fine - in fact there are lots of pins that pncconf adds signals to but not driving pins. It does so so that added HAL code files could be more consistent.
Breaking a system because you added a signal to a pin is not user friendly.

The problem was that the PID component had a bug (it ignored the command-deriv pin)
The fix implemented simply made the PID component match its long standing manual page description, that is, make the PID component calculate the command derivative internally if the command-deriv pin was unconnected but use the command-deriv pin value if connected.

This matches the feedback-deriv pin behaviour, and any other fix (mode pins etc) would likely have much broader impact.

yes but breaking peoples systems on a released version is not helpful.
Especially since it's very difficult to find the problem, if you don't know about the special behavior.
Adding the fix to master would have been ok (Though I still disagree with the black magic)
I realize nobody did this on purpose - but telling people that adding a signal to a pin that is not connected to another driving pin is incorrect - is wrong - other then for the PID component.

Chris

I agree that this should have been left in 2.8 and later.

I do think that making command-deriv match the manual page and
feedback-deriv behaviour has the least impact on existing hal files
while fixing a even more mystifying bug (command-deriv having no effect)