RH post to Ian

[Ray Henry <rehenry-at-up.net>]

List - copy of my post to Ian
-----------------------------
Ian

Hey thanks for the complement and the bug report.  You have found program
code that creates NaN.  I didn't even know NaN existed until a post from
Matt a couple weeks ago.  I had no idea that it causes an almost
unstoppable runaway.

Here are the relevant lines from your file.
-------------------------------------------
G01 Z-0.1000 F 1.00 
G01 X-5.0000 Y 0.0000 F 5.00 
G01 X-5.0000 Y 0.0000 
G01 X 5.0000 Y 0.0000 
G02 X 5.0000 Y 0.0000 I-5.0000 J 0.0000         ########## Here Backplot
flips and takes off at high speed! #### 
G01 X 2.7000 Y 0.0000 
G01 X 2.7000 Y 0.0000 
G01 X 2.0000 Y 0.0000 
G02 X 2.0000 Y 0.0000 I-2.0000 J 0.0000 (also here)
G01 X-2.0000 Y 0.0000 
-------------------------------------------

Notice that the x and y locations for the offending lines and the one prior
to it are the same.  I don't know why flashcut is asking for the g02 radius
move to the same place it is.  There must be some assumption that they make
at that point that emc does not.

I revised those lines as follows to create an enc runnable program.
-------------------------------------------------------------------
G01 X-5.0000 Y 0.0000 F 5.00
G01 X-5.0000 Y 0.0000
G01 X 5.0000 Y 0.0000
G02 X-5.0000 Y 0.0000 I-5.0000 J 0.0000(Here Backplot flips and takes off
at high speed!)
G01 X 2.7000 Y 0.0000
G01 X 2.7000 Y 0.0000
G01 X 2.0000 Y 0.0000
G02 X-2.0000 Y 0.0000 I-2.0000 J 0.0000 (also here)
G01 X-2.0000 Y 0.0000
G00 Z 0.2500
-------------------------------------------------------------------
But this makes the assumption that what you wanted was an arc from x5 to
x-5 for the first offending line and an arc from x2 to x-2 for the second.
If you have a flashcut routed out copy of the result of this nc program you
could compare that with the attached plate1.gif which I just ran with
Tkbackplot from the revised lines here. (you can view it with your browser)

Any differences between would be instructive as to the different
assumptions made by flashcut and emc.  A general post on your findings
about those differences would be helpful.  Perhaps you could prepare some
stuff for the handbook about this problem.  (shameless hustle)

I'll post a brief bug report to the emc list about the cause of the NaN and
we'll see if we can get the awesome runaway fixed.  My steppers went wild
when they hit that line.  My <quit> or the <x> at the top of the window was
the only thing that stopped 'em short of the hard limits.

I'm going to be out for a couple days but I'm anxious to hear what you find.

Ray

Plate.gif