gmoccapy/qtdragon: preview not respecting WCS other than G54 on file load/reload

[Sigma1912]

This is running 2.9.1 and master.

1. start configs/sim/gmoccapy/gmoccpy
2. load file without WCS definition, for example 'arcspiral.ngc'
3. Note how the spiral is centered on the origin in the preview:
   

4: run mdi-command 'g55'
5: run mdi-command 'g10 l20 p0 x100'
6: reload the ngc file.
What I expect it that the preview shows the center of the spiral in the new origin of g55 (ie moved 100mm to the right) however the preview only shows the new origin but does not move the preview:

[edit] same in qtdragon

This used to work in previous versions (2.9pre) and still works in 'axis'.

[Sigma1912]

Not sure what is going on here.
I copied and built the rip install folder from the machine where this works to one of the ones where it does not work and I see the same bug.

Maybe this is dependent on the installed opengl version or something?

[hansu]

But if you run the program, the spiral is machined around the new origin?

[Sigma1912]

Yes, the machine moves correctly. It's just the preview that is off.

[Sigma1912]

I've done a bit of digging in '/lib/python/rs274/glcanon.py' and 'interpret.py'
This, I think, is here is where the program preview is constructed in 'interpret.py' (with my print statements):

I then ran AXIS gui changed the offset from G54 to G55 and clicked on reload:

Then I did the same in gmoccpy:

Note how the run with axis correctly sets the offset to 2 (ie G55) with it's corresponding x value runs through the gcode and resets to G54 (persumebly because of the 'M2')

The run with Gmoccapy resets the offset to G54 before running the gcode hence the preview is always drawn in the origin of G54.

Does anybody have an idea as to why this happens?

I also think that this maybe dependent on timings as I have a machine where I have observed correct and incorrect behavior using gmoccpy (axis work correctly on all 3 systems I have tried)

[Sigma1912]

for gmoccpy this seems to get called from gremlin.py (if I interpret my print statment right )

interp.py line 60 +++++++++++++++ self:  <gremlin.StatCanon object at 0x7f178742d300>
interp.py line 62 +++++++++++++++ index:  2
interp.py line 64 +++++++++++++++ g5x-x:  2.0
interp.py line 60 +++++++++++++++ self:  <gremlin.StatCanon object at 0x7f178742d300>
interp.py line 62 +++++++++++++++ index:  1
interp.py line 64 +++++++++++++++ g5x-x:  0.0

[hansu]

Maybe @gmoccapy has an idea...

[PeterMue]

I've done some pretty weird debugging on this topic that eventually lead me to this log output:

emc/rs274ngc/gcodemodule.cc:839	PyArg_ParseTuple: try ParseTuple with format: sOO!|s:new-parse
emc/rs274ngc/gcodemodule.cc:843	PyArg_ParseTuple: set initcodes=nullptr (reason: PyArg_ParseTuple(args, "sOO!|s:new-parse",...) failed
emc/rs274ngc/gcodemodule.cc:846	PyArg_ParseTuple: fallback ParseTuple with format: sO|sss:parse
emc/rs274ngc/gcodemodule.cc:855	delete pinterp
emc/rs274ngc/gcodemodule.cc:864	pinterp = new Interp;
emc/rs274ngc/gcodemodule.cc:868	USER_DEFINED_FUNCTION_NUM=100
emc/rs274ngc/gcodemodule.cc:882	pinterp->init();
++++
  File "/home/peter/Develop/linuxcnc-dev/bin/gmoccapy", line 5787, in <module>
    Gtk.main()
  File "/usr/lib/python3/dist-packages/gi/overrides/Gtk.py", line 1641, in main
    return _Gtk_main(*args, **kwargs)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_actions.py", line 105, in _f
    return f(self, *a, **kw)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_filechooser.py", line 225, in on_activate
    self._load_file(self.gstat.stat.file)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_filechooser.py", line 129, in _load_file
    self.gstat.emit('file-loaded', filename)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_gremlin.py", line 149, in fileloaded
    self._load(f)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/rs274/glcanon.py", line 408, in inner
    return f(self, *args, **kw)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_gremlin.py", line 284, in _load
    return self.load(filename)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gremlin.py", line 331, in load
    result, seq = self.load_preview(filename, canon, unitcode, initcode)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/rs274/glcanon.py", line 1940, in load_preview
    result, seq = gcode.parse(f, canon, *args)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/rs274/interpret.py", line 64, in set_g5x_offset
    traceback.print_stack(limit=25)
++++ set_g5x_offset(index=2, x=-3.937007874015748, y=0.0, z=0.0, a=0.0, b=0.0, c=0.0, u=0.0, v=0.0, w=0.0
emc/rs274ngc/gcodemodule.cc:884	pinterp->open(f);
emc/rs274ngc/gcodemodule.cc:908	unitcode: unitcode=G21 && RESULT_OK
emc/rs274ngc/gcodemodule.cc:910	unitcode: result=0, RESULT_OK=1: check !RESULT_OK
emc/rs274ngc/gcodemodule.cc:912	unitcode: pinterp->execute()
emc/rs274ngc/gcodemodule.cc:916	initcode: initcode=G17 G21 G40 G43H0 G64P0.005 G80 G90 G94 G97 M5 M9 && RESULT_OK
emc/rs274ngc/gcodemodule.cc:918	initcode: result=0, RESULT_OK=1: check !RESULT_OK
emc/rs274ngc/gcodemodule.cc:920	initcode: pinterp->execute()
emc/rs274ngc/gcodemodule.cc:924	while: !interp_error && RESULT_OK
emc/rs274ngc/gcodemodule.cc:933	while: result=0, RESULT_OK=1: check !RESULT_OK
...truncated...
++++
  File "/home/peter/Develop/linuxcnc-dev/bin/gmoccapy", line 5787, in <module>
    Gtk.main()
  File "/usr/lib/python3/dist-packages/gi/overrides/Gtk.py", line 1641, in main
    return _Gtk_main(*args, **kwargs)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_actions.py", line 105, in _f
    return f(self, *a, **kw)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_filechooser.py", line 225, in on_activate
    self._load_file(self.gstat.stat.file)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_filechooser.py", line 129, in _load_file
    self.gstat.emit('file-loaded', filename)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_gremlin.py", line 149, in fileloaded
    self._load(f)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/rs274/glcanon.py", line 408, in inner
    return f(self, *args, **kw)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gladevcp/hal_gremlin.py", line 284, in _load
    return self.load(filename)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/gremlin.py", line 331, in load
    result, seq = self.load_preview(filename, canon, unitcode, initcode)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/rs274/glcanon.py", line 1940, in load_preview
    result, seq = gcode.parse(f, canon, *args)
  File "/home/peter/Develop/linuxcnc-dev/lib/python/rs274/interpret.py", line 64, in set_g5x_offset
    traceback.print_stack(limit=25)
++++ set_g5x_offset(index=1, x=0.0, y=0.0, z=0.0, a=0.0, b=0.0, c=0.0, u=0.0, v=0.0, w=0.0

So basically when loading or reloading the interpreter is initialized via pinterp->init().
This uses the current context and thus calls the set_g5x_offset method correctly (g55: index=2).
Then, the unitcode and initcodes are executed. By default, the gmoccapy sim machines define the following initcode:
G17 G21 G40 G43H0 G54 G64P0.005 G80 G90 G94 G97 M5 M9
(see section RS274NGC key RS274NGC_STARTUP_CODE in the ini file: gmoccapy.ini#L51)

Therefore it's drawn in G54.

So basically, it's doing what it should - isn't it?

[gmoccapy]

IMHO not,
The INI Default code should only be executed on start of the GUI or if you make a new file, or if an M2 is executed, but not if a file reload has been executed.

Loading a file should never been followed by the execution of the default INI code.

Norbert

[c-morley]

If I had to guess.. I wonder if there is a race between setting the initcode and loading the file?
I'm curious why the unit code is separate too.
We have noticed similar things in qtdragon - but it is not consistent.

[Sigma1912]

So basically, it's doing what it should - isn't it?

Interesting, thanks for looking into this.
I would disagree with the conclusion though:

1. It is not what it used to do in earlier versions
2. It is not what the Axis GUI still does
3. It does not do it consistently
4. This would basically make it impossible to properly load/reload any code that does not contain the full preamble.

Loading a file should never been followed by the execution of the default INI code.

I agree. I would expect 'startup code' to be executed on starting linuxcnc

I wonder if there is a race between setting the initcode and loading the file?

It certainly looks like a race condition to me but why would we rerun the initcode on a file reload in the first place?

[PeterMue]

You're probably right, but when looking at the code for loading a file both in axis and in gmoccapy, it always loading at least the configured default initcode:

https://github.com/LinuxCNC/linuxcnc/blame/master/src/emc/usr_intf/gremlin/gremlin.py#L330
https://github.com/LinuxCNC/linuxcnc/blob/master/src/emc/usr_intf/axis/scripts/axis.py#L1241-L1243

Both sources don't seem to have changed lately in the relevant sections.

For axis, there are some more gcodes appended based on the current linuxcnc stat: (s.gcodes)
https://github.com/LinuxCNC/linuxcnc/blob/master/src/emc/usr_intf/axis/scripts/axis.py#L1265

This entire part from axis is missing as far as i understand in gremlin.