New LinuxCNC-Ethercat git tree

[edited]

The main error message here is the only difference between 1.13.1 and 1.13.0 -- it now actually reports SDO setting failures back instead of silently ignoring them.  You can try rolling back to 1.13.0 and see if that fixes things.

It'd be helpful to have more data.  This looks like a generic CiA 402 device?  Which hardware are you using, and can you share a copy of your XML file?  Also, what was logged in `dmesg` when this happened?

the device is a Delta ASDA A2-E
I attach the xml file


 


by giving power to the CNC everything works, if I can I'll try to go back to 1.13.0
Thank you

power activated, sees all OP slaves
maybe so...


 

Well I'm a total EtherCAT newbie, but isn't this just the lcec code trying to download settings to your Delta drive? If your CNC is powered off is the Delta drive is powered off too and so if (of course) can't accept LinuxCNC commands over EtherCAT.

I have parts including Delta B3 shipping "soon" and already have four Beckhoff I/O modules working just fine so anything you discover about LinuxCNC 2.9.2 and Delta drives is great.

As is often the case in Linux, if you happen to dig into a project just when it is maturing, life can be wonderful! I think LinuxCNC + EtherCAT "works" for years but is now "works+EASY"

Entertainingly, prior to 1.13.1, it would queue SDOs for the drive asynchronously, so if they failed then nothing actually noticed. This meant that when bad settings were sent *no one could tell* until the drive failed to work. There'd be a record in 'dmesg', but very few people notice unless there are problems.

But, the flip side of this is that the way SDO settings were queued up they'd be re-set when the device powered on. So, weirdly, talking to a powered-off drive should actually have worked. That wasn't a case that I'd really considered.

We now have better error reporting, *and* we still keep SDO settings around so we can re-set them if the slave reboots. The downside is that we require that devices are powered on up front.

Is supporting powered-down devices something that we should be paying attention to?

I'd think that a one-shot information message right after:

"Failed to execute SDO download"

something like:

"Are any of your EtherCAT devices powered off? Did you add/remove a device w/out editing your XML?

I suspect LinuxCNC isn't happy if an ethernet attached Mesa 7i76e is powered off either.

A simple hint could reassure the user that nothing may be broken, just forgot to power something up.

I had written something on this yesterday, but apparently failed to post ist.

If the devices are off, the EC master cannot connect and accordingly linuxcnc can't start. That to me is entirely logical behaviour and fine. If a mesa card is offline, linuxcnc will crash as well.

Scottlaird:
Is supporting powered-down devices something that we should be paying attention to?

hi, I don't know
during the changes of recent weeks there was a moment in which LCNC started when the devices were turned off,
now no longer.
I thought there was a problem, but everything is ok
Thank you

The problem is that there are probably hundreds of things that could go wrong at any point, and we don't always know which ones are the most likely. Right now, the main point of most of our error messages is simply to provide debugging data, along with enough information to figure out where in the process things broke. That's fine for development but (obviously) not a great user experience. It's part of the process of improving things, though.

Once things stabilize a bit more we can work on improving the error experience.

Unfortunately, I'm still in the "break things more often" phase while doing cleanups and backfilling missing hardware support. Hopefully that'll slow soon.

@scottlaird

Awesome work…  I know you and Rod Webster have gotten this all working much easier, but I still like to do the source builds just as a backup.  This morning, I did a quick build of LCNC 2.9 using Bullseye with LXQT and your branch of /linuxcnc-ethercat/linuxcnc-ethercat.git    Everything is working correctly without any build issues.  I like that you have the “realtime.mk” file fixed, that totally streamlines the build.  I’m overloaded with work, but when I get time I will amend the “source build sticky” and replace Sascha Ittner’s link as you are definitely are making things happen.  @tommylight (I will get rid of the CAPS, I know you hate them)

Also, if you need more Beckhoff terminals for testing, I am more than willing to send you some.  I have all types and can generate a list of what I have.   I have several test setups and can send you a complete setup if needed with everything mounted on a DIN rail and ready to plug into 110V wall outlet.  

Once again, thanks for taking your time to make all of this happen it is greatly appreciated by all of us!!!