Fix HOME_AFTER_DEACTIVATE behavior when steppers released

[swissnorp]

Description

There are many situations where HOME_AFTER_DEACTIVATE does not work as i understand it to work. Especially if NO_MOTION_BEFORE_HOMING is enabled.

In most cases just axis_homed is checked (by TEST(axis_homed... or by function all_axes_homed) which does not respect HOME_AFTER_DEACTIVATE. Instead homing_needed() should be used instead of all_axes_homed()...

The easier and more strict way would be to change the behavior of DISABLE_AXIS_X(), DISABLE_AXIS_Y() and DISABLE_AXIS_Z() in indirection.h

• If HOME_AFTER_DEACTIVATE is enabled the axis will be no more trusted if stepper has been disabled. set_axis_not_trusted(axis) instead of CBI(axis_known_position, axis).

Benefits

HOME_AFTER_DEACTIVATE works as intended, especially if NO_MOTION_BEFORE_HOMING is activated.
More security for users.

[thinkyhead]

I don't think this can be applied without negative side-effects. We can't just throw away the differentiation between an axis being "homed" and an axis being "known" as there may be valid reasons to retain this distinction for some cases, even when using HOME_AFTER_DEACTIVATE.

[swissnorp]

I have looked as far as i could to find parts of marlin, where this could have a negative side effect.. i could not find any.

If HOME_AFTER_DEACTIVATE is enabled the whole machine should be in a state as it was never homed before after stepper have ben released. Homing should be forced as described. thats my opinion/understanding as a cnc mechanic :)

[thinkyhead]

Anything that calls axes_need_homing function will get the right answer based on HOME_AFTER_DEACTIVATE. If some feature is misbehaving —as you alluded to— then it must be accessing the flags directly and not calling this function. In each case that you have seen we will need to check whether there is a sound reason for that behavior.