Pathpilot Reset Functionality

I'm working on creating a new Reset Function that will clear the machine's E-STOP state. I'd like to implement this as an API route so that I can control the mill's functionality remotely from another computer.Currently, my route is based on thefunction from. However, PathPilot crashes approximately 1 out of every 5 times when I call this endpoint.Does anyone have any ideas what could be causing this intermittent crashing issue?

Weird.

Does it print out an error or something?

Does it always crash in the same way?

You are connecting from another computer via SSH, correct?

I would consider a 20% chance of crash to be more than intermittent, but maybe that's just me.

I wish it did print something out. I added a bunch of error logging to print to the system logs and It seems to be crash at different lines (which I find to be odd).

I have the two computers on the same network and I'm hitting those endpoints via HTTP.

If it helps here's the code that I've added to the tormach_mill_ui.py:

def ss_reset_route(self):
"""
/reset
Reset the machine after an emergency stop was pressed.
Note: This function is the same as on_reset_button_clicked()
"""
self.hal["pp-estop-fault"] = 0 # clear any existing pp software estops
self.halt_world_flag = False # note issues with halt world. Do we lock it out during reset
# and then check or reset this at end of on_reset_button_clicked
self.clear_message_line_text()
self.hide_m01_image()
self.unlock_enclosure_door()

# turn off air blast off rack tool changer
if self.machineconfig.rack_tool_changer_supported() and self.rack_tool_changer:
self.rack_tool_changer.stop_air_blast()

if self.hal["mesa-watchdog-has-bit"]:
# since resetting the mesa card io_error parameter is more involved now with ethernet,
# only do this if we really did see a watchdog bite.

# clear Mesa IO errors (if any). this must be done PRIOR to setting the mesa-watchdog-has-bit pin low.
clear_hostmot2_board_io_error(self.inifile)

# clear Mesa card watchdog
self.hal["mesa-watchdog-has-bit"] = 0
self.mesa_watchdog_has_bit_seen = False

# give it a second to re-establish IO link before jamming commands at it.
time.sleep(1.0)
self.poll()

# did the watchdog re-bite already? If so, re-establishing the IO link didn't work.
# leave us in e-stop.
if self.hal["mesa-watchdog-has-bit"]:
self.mesa_watchdog_has_bit_seen = True
self.error_handler.write(
"Machine interface error. Check cabling and power to machine and then press RESET.",
ALARM_LEVEL_MEDIUM,
)
self.call_ui_hook("reset_button")
return Server_Response_Factory.create_server_response(400, "Failed IO Link")

# Configure the A axis hardware
# This happens in the reset handler because __init__ is too early, and the change to traj.max_angular_ vel is ignored
a_axis = self.machineconfig.a_axis.selected()
if a_axis:
self.configure_a_axis(a_axis)

# order of these is important. Answer queries first, then check for random stop/reset presses
if self.set_response_cancel():
return Server_Response_Factory.create_server_response(
400, "Outstanding Prompts"
) # check for outstanding prompts and cancel,True is message answered

if self.atc.worker_thread_busy.is_set():
self.atc.stop_reset.set() # only if atc thread in progress
self.atc.feed_hold_clear.set() # signal that any feed holds are cleared

# RACK_QUESTION what about router pdb on and rack disabled?
if self.rack_tool_changer:
if self.rack_tool_changer.worker_thread_busy.is_set():
self.rack_tool_changer.stop_reset.set() # only if rack atc thread in progress
self.rack_tool_changer.feed_hold_clear.set() # signal that any feed holds are cleared
self.set_image("feedhold_image", "Feedhold-Black.jpg")

# clear feedhold
self.clear_feedhold()

# reset e-stop
if self.status.task_state != linuxcnc.STATE_ESTOP_RESET:
# this actually ends up doing a linuxcnc command abort internally
# and that will run any on_abort ngc code.
self.command.state(linuxcnc.STATE_ESTOP_RESET)
self.command.wait_complete()
self.poll()
if self.status.task_state not in [linuxcnc.STATE_ESTOP_RESET, linuxcnc.STATE_ON]:
self.error_handler.write(
"Failed to bring machine out of E-stop. Please check machine power, limit switches, and communication cable from the controller to the machine."
)
self.call_ui_hook("reset_button")
return Server_Response_Factory.create_server_response(
400, "Failed to bring machine out of E-stop"
)

# clear alarm
self.estop_alarm = False
self.display_estop_msg = True

# Prevent coming out of Reset if a limit switch is active.
if self.status.limit[0] != 0:
error_msg = "X limit switch active."
self.error_handler.write(error_msg, ALARM_LEVEL_MEDIUM)
if self.status.limit[1] != 0:
error_msg = "Y limit switch active."
self.error_handler.write(error_msg, ALARM_LEVEL_MEDIUM)
if self.status.limit[2] != 0:
error_msg = "Z limit switch active."
self.error_handler.write(error_msg, ALARM_LEVEL_MEDIUM)
if (self.status.limit[0] != 0) or (self.status.limit[1] != 0) or (self.status.limit[2] != 0):
error_msg = "Disable limit switches in Settings, then push Reset, then carefully jog off limit switch, then re-enable limit switches in Settings."
self.error_handler.write(error_msg, ALARM_LEVEL_MEDIUM)
self.call_ui_hook("reset_button")
return Server_Response_Factory.create_server_response(400, "Disable limit switches in Settings")

time.sleep(0.25) # TODO: Added time.sleep()
# must be turned on again after being reset from estop
if self.status.task_state != linuxcnc.STATE_ON:
# this actually ends up doing a linuxcnc command abort internally
# and that will run any on_abort ngc code.
self.command.state(linuxcnc.STATE_ON)
self.command.wait_complete()
self.poll()
if self.status.task_state != linuxcnc.STATE_ON:
err_msg = "Failed to bring machine out of E-stop. Please check machine power, limit switches, and communication cable from the controller to the machine."
self.error_handler.write(err_msg)
return Server_Response_Factory.create_server_response(400, err_msg)

# saw a rare case where the ATC stuff above times out after taking 5 long seconds and during that
# time, the operator presses the e-stop button. So just check again to be sure before we start
# running commands. If it is e-stopped, the periodics will take appropriate action.
if self.hal["machine-ok"] == False:
return Server_Response_Factory.create_server_response(400, "Failed to bring machine out of E-stop")

# stop motion
self.command.abort()
self.command.wait_complete()

# reset/rewind program
if (self.status.limit[0] == 0) and (self.status.limit[1] == 0) and (self.status.limit[2] == 0):
# HACK manually decrement the part counters by one since M30 increments them
# if this ever becomes optional, then make sure not to do this if M30 incrementing is disabled
self.issue_mdi("#5650 = [#5650 - 1] #5651 = [#5651 - 1]")
self.command.wait_complete()
self.issue_mdi("M30")
self.command.wait_complete()

# clear SB status
self.single_block_active = False
self.single_block_button.set_led_state(self.single_block_active)

# clear live plotter
self.gremlin.clear_live_plotter()

# refresh work offsets
self.refresh_work_offset_liststore()

# rewind program listing and set starting line
if self.is_gcode_program_loaded:
self.set_start_settings_and_mark(1)

# some folks got confused because their program ended, the M30 reset current line to 0 and
# the 50ms periodic auto-scrolled back up to the start line. But then they managed to scroll
# around in the view and then press the Reset button and they expect it to auto-scroll to the
# top again. The 50ms periodic doesn't do anything because the current line hasn't 'changed'
# from 0 so we need this here to always smack the display back to the start line.
self.scroll_sourceview_to_mark("start")

self.call_ui_hook("reset_button")

self.do_first_run_setup()

self.axis_motor_command(0, MOTOR_CMD_NORMAL)
self.axis_motor_command(1, MOTOR_CMD_NORMAL)
self.axis_motor_command(2, MOTOR_CMD_NORMAL)

# Make sure the override sliders are enabled. The ATC code disables them and tries to restore them, but
# in certain aborted situations, they can get stuck off.
self.command.set_feed_override(True)
self.command.set_spindle_override(True)

# g21 and machineconfig need to be accurate before setting scaled jog increment
# self.update_gui_for_units(self.g21) # TODO: this seems to breaking some things, need to figure why

# self.start_loop_recording() # TODO: this seems to breaking some things, need to figure why

return Server_Response_Factory.create_server_response(200, "Mill has been reset")

Perhaps the indentation is wrong?

Your code has a lot of errors and is not nearly as streamlined as it could be, though I have not worked with Python for about 2 years.

Bugs hide within overly long code.

Short, efficient code is much easier to debug.

Try to use methods to guarantee unifority of funtionality across the program!

Don't return a 400 "Bad Request" error if the limit switch is active.

Use 500 or 503 or something.

developer.mozilla.org/en-US/docs/Web/HTTP/Reference/Status

Thanks for the reply!

Yes, it's inside the mill(TormachUIBase) class. The indentation got messed up when I copied it over, but I can assure you it's correct in the actual code because the function does work 'sometimes'.

The suggestion to keep the code concise is a good idea, but I'm hesitant to trim it down too much as this mirrors the PathPilot reset code. The last thing I'd want is to initialize one of the HAL pins correctly and that would cause even more issues. I'm really not experienced in the CNCs and have been only tinkering on and on off for about 3 years.

You need to write some helper methods so this code makes more sense.

The "one huge method" approach does not work well, especially in Python.

Make sure the method can never have more than con simultaneous execution.

Have a variable that causes the method to immediately return if it is already running.

This method checks every limit switch and adds a corresponding value if it is tripped.
I really recommend you make useful little methods like this one instead of jamming everything into one method in one file.

If no limits are active, it returns 0, if all limits are tripped, it returns 2^1-1.

I recommend you make a separate .py file for little methods like this and import it so the main file is empty and easy to read/understand/debug.

I also strongly recommend you put the entire method in a try/catch and have different error types return different HTTP error codes and the error message.

Not a Python expert, but this is an idea:
Reference Java code: github.com/pacman-admin/Repeat/blob/mast...mmon/HTTPLogger.java

Instead of returning error codes, raise Exceptions, it's cleaner and makes more sense IMO.

Also, it makes the whole program more flexible.