Re: PPMC progress

[Jon Elson <jmelson-at-artsci.wustl.edu>]

JohnDRoc-at-aol.com wrote:

> I can't offer much towards a solution, but I will definitely vouch for the existence of a Tcl/Tk "degradation" problem.  While working on the "machine from he--" I witnessed it firsthand, and it seemed more a result of starting and stopping EMC (which I did alot of).  It would get to the point that it was so slow that it would take minutes instead of seconds to start up.  That was in the May, June and July versions.  At first I thought it was a PPO (piss-poor operator) problem, but now I am inclined to think otherwise.  Wish I could offer more...

Ohhhh, this is all starting to tie together! I had observed at times
when using the TkEMC gui that the display would not update all
the time.  I wasn't sure it wasn't my hardware, but I didn't see this
on the same software and hardware with xemc.  What I was doing
was moving one hand-operated encoder from channel to channel
on my encoder counter card.  I sort of thought it might have been
from plugging in an encoder with the power on, but I also saw
it sometimes when I wasn't swapping the encoder to another channel.
What I saw was that the display didn't show any changes, while
I was moving the encoder back and forth.  Then, suddenly, the
display would jump a large distance.  That indicates the hardware
had been incrementing the position count all the time, but the
GUI was not updating the display.  This would happen perhaps
once a minute or so, and seemed to last from 1 to 5 seconds
(this is on a 333 MHz Pentium-II).

I really didn't know what to think about this particular observation,
but it obviously bothered me.  I didn't think of it being related to
the crashes/hangups, but this new info seems to indicate it
very well could be!  So, maybe there is some problem with the
dispatching of the Tk/tcl task, that sometimes resolves itself and
sometimes doesn't.  Or, possibly, there is a haywire pointer in
the c-coded resources for the TkEMC task, and sometimes the
pointer gets set back to a usable value before something critical
gets damaged, and sometimes not.   Hmmm, that might give one
some ideas where to look.  I've never even looked into the C-coded
support stuff for TkEMC, I just remeber that there is some.
I don't suppose you can have bad pointers in tcl, or can you?

A couple more observations like this, and we may have it
just about nailed down!  At least, user mode code under C you can
debug!

Jon