Proposed fix for #20382 (Unreliable circle detection)

[ghost]

Description

This is a proposed fix for #20382, whereby NEAR_ZERO macro was leading to problems. New logic essentially deletes the NEAR_ZERO requirement because target and position vector should suffice for determination of circle presence (right?) .

Need @matbhe to please test code to see if his problem resolves.
UPDATE: matbhe reports it is working for him. Will mark as ready for review, but others should test/review this before merge just to be sure I didn't make mistake.

Benefits

1.) Proposed fix for #20382
2.) only perform expensive atan2 operation if not circle

Configurations

Default Config will suffice

Related Issues

#20382

[ghost]

There are now two commits and each one is a different proposed change. The first one attempts to fix the relevant issue but keeps the default behavior the same whereby angular rotation of zero is still assumed to represent a full circle and set accordingly. The second one includes the same fix of removing near_zero requirement but goes further and handles the angular_travel == 0 case differently by keeping it zero. i.e. only way to perform a full circle is when target == current position and if atan2 somehow yields 0 then it is kept zero instead of reset to +-2pi.

[ghost]

addendum just adds a return statement to bail when target != position and angle == 0. This means that center point, initial position, and target position all lie on the same line but with center not between the other two points. Either bail or maybe add some logic to replace G2/G3 command with a single G1 command to go straight to target position.

[ghost]

@matbhe reports it is working for him with no problems. So looks like this is a viable alternative. Will mark as ready for review so devs and community can review the code and keep what they like.

[sjasonsmith]

@yysh12 it looked like you might be planning to make changes based on feedback from @marknn3? What is the status of this?

[DerAndere1]

Ideally a straight line to target point in those circumstances would be ideal. I do not yet know enough marlin to know best method to replace with G1 when that happens though.

I guess that is possible with the following function calls as can be found in file G0_G1.cpp:

Marlin/Marlin/src/gcode/motion/G0_G1.cpp

Line 68 in 7afd274

get_destination_from_command(); // Get X Y Z E F (and set cutter power)

and

Marlin/Marlin/src/gcode/motion/G0_G1.cpp

Lines 98 to 102 in 7afd274

	#if IS_SCARA
	fast_move ? prepare_fast_move_to_destination() : prepare_line_to_destination();
	#else
	prepare_line_to_destination();
	#endif