Welcome to LinuxCNC.org

Heres my take on this:


Its easy to interrogate the system for available pins and parameters (a 2 line HAL file is all thats required). Getting the HostMot2 pinout in a useable form is a little harder but doable (dmesg has it already).

Well for standard Hostmot2 components scale and units are known, on sserial, things are easier as scale is known and available from parameters, units are not currently exported from the driver but are available to it.

Thats what I have been saying, rather than 100s of XML files, the user only needs one, the one that matches his configuration. IMHO this is best generated dynamically from the current users hardware setup.

The philosopy of HostMot2 firmware and driver has always been that _any_ combination of features should _just work_. This is how the driver works, it does not care about the specific pinout or set of available modules, it does not need auxillary desciption files, it just discovers whats there and works with what it has.
There has never been a bias towards "common" bitfiles as all bitfiles should work the same
the only real limit being what the user needs, FPGA space, and daughterboard pinout constraints.

I think pncconf has been a great boon for people overwhelmed by LinuxCNC configuration complexity and I thank you for the huge amout of work you have put into it.

I do think it can be structured so that new boards and firmware are more automatically accomodated so there is not endless tweaking of XML files

There may be some driver/ firmare changes possible to make this easier as well

(one thing I am adding to HostMot2 is daughterboard hints so the firmware knows the appropriate daughterboard type(s))

well when you say hostmot2 'just works' your kinda missing a step. While the hardware always has 'just worked',
Working hardware and configured to do something with linuxcnc is two different things.

for the hardware to work (depends what you mean by work) it doesn't need to know scale, units, parameters etc but to do useful stuff in linuxcnc you sure do.
And with sserial cards there is now no way to predict how to use the components beside getting a unit and testing.
What I am getting at is that even if PNCconf can get the pin names and parameters unless it knows how to set the parameters you still can't configure the system. live interrogating is not the be all end all

But.. I am certainly not against trying it and seeing. Maybe my reservations are not as bad as I think.
Does anyone have code for pulling info from the boards?

I tried, but the recommended procedure does eliminate the second 7i77, but the 7i64 doesn't show up anywhere in PNCconf that I can see. Here are the various tab screens I get when I set encoders to 6 and sserial channels to 3:

Oops! I can't get this forum to upload images. It worked last night, but tonight all it does is place a blank IMG tag and nothing uploads. I don't have an FTP client on this CNC PC, and don't feel like installing one, so the best I can do is take the screen shot images home and see if I can upload them from there, or put them on my server and provide embedded links. Anyway, trust me, the 7i77 tabs are there (I/O Connector 2, I/O Connector 3, Smart Serial 0, and Smart Serial 1) but there are no tabs for the 7i64.

I thought I had the latest LinuxCNC build, but Update Manager said otherwise. Apparently there was a new build in the last week or so. I used Update Manager to fetch linuxcnc and linuxcnc-doc from buildbot and install it. No good. Same results.

I even pretended like it was a Windows computer and rebooted it.

I tried renaming the PNCconf config file with a .bak extension so it would force it to create a new configuration, in case there was something in the existing configuration that was confusing PNCconf, but it still showed the tabs for one 7i77 and no 7i76.

I can Launch Test Panel and it looks like it sees all of the 7i64 and 7i77 pins. Example: hm2_5i25.0.7i62.0.2.adcin.01.in shows up in the HAL Meter as 2.949xxx. The two analog inputs aren't connected to anything and the ADC is sampling noise, so the last three digits are fluctuating. The 00 analog input is 2.944xxx.

I don't know if it's significant, but before and after today's LinuxCNC update, PNCconf produces the following errors in the terminal about halfway through its startup.

/usr/bin/pncconf:3999: Warning: g_path_get_dirname: assertion `file_name != NULL' failed
self.xml = gtk.glade.xml_new_from_buffer(doc, len(doc), domain="linuxcnc")
/usr/bin/pncconf:3999: GtkWarning: gdk_pixbuf_new_from_file: assertion `filename != NULL' failed
self.xml = gtk.glade.xml_new_from_buffer(doc, len(doc), domain="linuxcnc")
/usr/bin/pncconf:3999: Warning: g_object_unref: assertion `G_IS_OBJECT (object)' failed
self.xml = gtk.glade.xml_new_from_buffer(doc, len(doc), domain="linuxcnc")
/home/beast/Desktop/custom_firmware/firmware.py

Still trying to get the 7i64 to configure in PNCconf. Based on the pins that are available in the Test Panel, it looks like the hardware is being seen, and it's active, but PNCconf won't provide any tabs for the 7i64.

Help me Obi Wan. You're my only hope.


PS - I tried to exit from the HAL Test Pad and send this post, and everything locked up solid. I couldn't even pull up another Linux command line window. The PC's reset button wouldn't work. I was forced to pull the plug and crash it. I guess misbehaving realtime operating systems can crash hard. When I rebooted, my post was still in Firefox.
Sweet.

Looks like the first screen shot went through. Here's a second.

I can only post one image per post?

My post count is loving this.