I really think we should. With Kivy apparently abandoning python 2, that will force changes. Even if we run Webcontrol on linux/windows/mac and access it via a local browser, I see a lot of value in eliminating the kivy development requirement. I think that there are going to be quite a few more people who can tweak a python based web server and links than who can get up to speed on kivy development
YES
the new holey version has some code changes to handle sending very small values to the arduino (things that python defaults to sending as e-notation)
the new holey version does a check at startup for differences between config and firmware values (this may already be in WC)
the new holey version adds a new parameter, which conflicts with a parameter that WC uses for optical calibration.
the screens (and calculations) for the holey calibration need to be added to WC
Exactly. I’ve been bothered by all the warnings I see while trying to set up the holey install and make it a “standard” Maslow CNC release…
The two already have diverged. I like the idea of having them as options but I think it gets complicated and confusing to go forward with two different software packages that don’t work the same way and do not have the same features.
When Bar started this thing, no else knew the code either.
There’s very little that’s in GC that’s not in WC… basically I think it’s just the simulator and that’s because I didn’t want to deal with it and not sure its of much value these days as is. The Holey Calibration model could “easily” be incorporated into WC (will take time but there’s nothing challenging in it). The new method for syncing values would take some time and thought… I’d like to rewrite the whole setting thing anyway (for the third time). I’d also like to incorporate observers (like Kivy) to handle unit switching better, but that will take time as well.
I’d like to help somehow. For the moment I’m committed to building a custom installer for Holey Calibration GC (EDIT: @madgrizzle: do you know who created the “official” GC package? There is a bat file that executes an exe etc which starts GC?). Manual python/kivy etc install is not working for several Windows folks and I believe it is due to the dependencies and requirement to use PIP via cmd prompt.
Although, I’ve never touched a R Pi, I am worried about the technical bar required to get started using WebControl. To continue with the Maslow CNC mission, it needs to be easy enough for novice technical people to use. Reading through this entire thread, and seeing the technical knowledge required to get WebControl going worries me in the same way that Holey Calibration GC worries me about requiring users to set up the Python dev environment, with Kivy, numpy etc.
I’m not a software engineer but I think docker and virtualization is the way to go. Just polishing and making bullet proof is the tedious part.
Thanks for all you do for this community.
I just posted a thought on the github ‘restart releases’ issue
with triangular calibration and WebControl hitting, does it make sense to say
that WebControl replaces GroundControl (running on linux/windows/mac) accessed
via a browser?
If we make a big step like that, this feels like a justification for a version
2.0 jump.
The value in the simulator is to help people understand the different types of
errors and how we can’t just give them a knob to fix an incorrect Y
from a developer point of view it’s not that useful, but I think it does need to
exist somewhere. I could be a web page that someone (including me) hosts instead
of having to be in WC. It initially got merged into GC in the hope that there
would be less code duplication, but I don’t know if anything else in GC/WC
actually does the kinematics calculations (as opposed to just working in the XY
space)
there is a pi image available that you just load on an SD card, set your wifi
settings, and everything ‘just works’
it’s not as bad as you think
but I am thinking that we need to make it into a package that we can run on
linux/windows/mac desktops like we do GC (accessing it via a browser), so this
would require some changes and us figuring out the packaging issues.
not needing kivy and it’s dependencies should drastically simplify the
packaging, but we still want to make it an .exe for windows.
There were plenty of issues commented above in trying to get it to work.
Also surprising to me is the issues with getting holey GC to work on Windows. I also don’t think its that bad, but recognize it needs to be packaged.
No idea… I wasn’t around back then, but I suspect @bar likely.
Someone could probably start selling a pre-packaged RPI with the sd card burnt with the docker image, but getting the RPI onto a wireless network will still be tricky. Someone could likely write a phone app to talk to find the RPI and talk to it though… like setting up any IOT device (lights, smart plugs, etc.)… i wonder if there’s an opensource method for that.
I don’t know that its not doable, but will take time to figure out. I looked into it once, but you have to also set up directories to contain the html/scripts, etc. So it’d need some kind of package installer.
This is what I’m learning at the moment. Its all doable (cool stuff)! See my comments here. The thing is, you have to have separate installer packages (using different packaging software) to support Windows, Mac and Linux. That is why I’m trying to contact Bar about how he did it with GC.
One of our goals is to sell identical complete kits that include everything except the plywood, but that is a few months off.
I think if one adds up all the time needed to calibrate and build a stock frame, then modify the frame and re calibrate and then trouble shoot stuff like counter weights, z axis issues, etc. and then compares that to a kit that has already been troubleshot with quirks worked out people will see value in a complete kit, because people value their time.
Umm, you boot to the pi desktop and click a WiFi icon in the task bar like a regular desktop. Select an said and enter a password like any WiFi. Done. No custom work needed these days.
