using M70/72 can unexpectaly start the spindle on abort or error.

[c-morley]

If a user starts the spindle in MDI then stops the spindle with manual buttons, then uses a subprogram (such as probing routines) that record/reset codes with M70/M72 The spindle will start on abort or error.

I assume manual control doesn't reset the M5 in the internal settings. Explicitly setting M5 just before M70 will work around the problem.

https://forum.linuxcnc.org/qtvcp/50729-qtdragonhd-versa-probe-failed-probe-turns-on-spindle?start=10#286614

complaint was on 2.9. master is the same - never tested 2.8.

[andypugh]

Yikes! The one time you really don't want the spindle to start is when probing.

Is this an issue with the probe routine, or do we think that M70/M72 should never re-start the spindle?

[rmu75]

"spindle stop with manual buttons" just sends a stop command to emctask, right? state in interpreter is not changed? so maybe the button should instead send "M5"?

[rmu75]

btw, we really need a "max allowed rpm" field in the tool table, shouldn't be too hard to implement

[c-morley]

The problem is that manual stop and m5 are not equivalent in internal settings.

A really great feature would the ability to see what is being recorded and or ability to record different code with different options.

But for now just deciding how to make manual stop equivalent to m5 would be great.

To be frank I'm not sure why we enforce manual/mdi modes in the controller.

[c-morley]

In reality we mix them in the GUI

[c-morley]

Sorry missed a message....

The problem is manual stop and m5 are not equivalent in the internal settings.

I'm not sure why we still have manual/MDI modes in linuxcncs motion controller. We mix the modes in the GUI all the time.

[rmu75]

Main difference seems to be teleop/free mode in manual vs. coordinated in mdi.

[rmu75]

The problem with M70/M72 seems to be that interp internal "struct setup" is not in sync with status in emcmot. Don't know how to solve that -- why not make "external buttons" send "M5"?

[c-morley]

Because you can't send M5 in Manual mode, only in MDI. Also the fix needs to be deeper then the GUI code (if that is what you mean) otherwise the problem is still lurking.

[rmu75]

the problem is that emctask receives a NML command from the GUI to stop the spindle and instructs emcmot to do so. the interpreter has a separate status and doesn't see that command, so if it is told to restore state, in on_abort, it will restore to it's view of the world which has the spindle running.

either the gui button has to go through the interpreter or it has to reset the interpreter.

[rmu75]

maybe make the button send M5 if in MDI mode and NML message if in manual? but that sounds like a hack and would need consistent impl. in every GUI. not a good solution.

[andypugh]

I wonder if there is a way to make a spindle-stop NML command set the _setup[2] to stopped(5)? (and whether that would work)

https://github.com/LinuxCNC/linuxcnc/blob/master/src/emc/rs274ngc/interp_write.cc#L156C1-L157C63

  emz[2] = (settings->spindle_turning[0] == CANON_STOPPED) ? 5 :   /* 2 spindle     */
    (settings->spindle_turning[0] == CANON_CLOCKWISE) ? 3 : 4;

The workaround in the probe scripts is fairly easy, and should be done anyway. I doubt that many probe with a spinning spindle.

[c-morley]

Some probes require spinning spindles. M70 is supposed to record the current state and it does not. Maybe a better way to say this is the interpreter is not up to date with linuxcnc's actual state.

is there not a sync state command that MDI commands should call first or manual commands call after?

[samcoinc]

wouldn't any que buster cause that? Like a G4?

On Fri, Dec 8, 2023 at 2:05 PM c-morley @.***> wrote:

Some probes require spinning spindles. M70 is supposed to record the current state and it does not. Maybe a better way to say this is the interpreter is not up to date with linuxcnc's actual state.

is there not a sync state command that MDI commands should call first or manual commands call after?

— Reply to this email directly, view it on GitHub https://github.com/LinuxCNC/linuxcnc/issues/2764#issuecomment-1847778172, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEHRGQW54VJOABN2VQFHIMTYINXILAVCNFSM6AAAAAA75AYTESVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNBXG43TQMJXGI . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[c-morley]

in this case it's not a que problem. the que is empty after the initial (user typed in) mdi command to start the spindle. It's the manual button to stop the spindle that changes the actual state of the machine but the interpreter doesn't see it. When the MDI command (m70) comes, it records it as the spindle is not stopped.

[andypugh]

This seems like a difficult thing to solve in the general case. Even my idea above of writing to the interpreter state from motion won't always work if you consider pluggable interpreters.

I think that this should be documented in the M70 documentation as a limitation of the code.

I noticed yesterday that Chris has added an M5 to the script, and I hope to push out a 2.9 update in the next few days to distribute both this and the run-from-line fix. However that is dependent on working out how to fix the buildbot builds.

[new952036]

@aman952036

[new952036]

@aman952036