Mesa 7I96 + Renishaw LP2 probe

That routine seems to use G43.1 and some calculations. That seems like the wrong way to do it, to me.

Though, version that is actually distributed with LinuxCNC uses G43.1 too:
github.com/LinuxCNC/linuxcnc/blob/master...es/manual_change.ngc

(Note that that one also relies on a number of INI file entries to work out which pins to use, and a custom GUI panel, so might not be the ideal example to copy)

There is certainly a lot to be said for having a proper spindle with pre-set tools.

But both these routines store the newly-measured offset in the tool table. But only for Z (I can't remember if you are working with a lathe or a mill)

I can only guess how it is all meant to work by looking at the G-code.

I am wondering if perhaps you are using a machine that doesn't home repeatably? Do you by any chance use homing to set the workpiece zero, rather than home switches?

OK, then storing offsets in the tool table should be correct.

It looks like G10 L11 is intended for this use-case.
linuxcnc.org/docs/2.8/html/gcode/g-code.html#gcode:g10-l1

Though I was actually looking for a G10 variant which set the tool length and position in the machine coordinate system, as I think that generally tool-length probing is done with G93 moves.