[3.4.5] CAN connection drops mid-print (OoM error)

[dc42]

@Diamondback thanks. Are any of the endstops or the Z probe connected to the Duet 3 Mini?

[Diamondback]

@dc42 no, I made sure that all axis endstops and the Z probe are connected to the same board as the relevant steppers (ie all on the 6HC master)

[dc42]

@Diamondback thanks again. I am looking for data structures that consume 24 bytes of memory, and endstops and Z probes were one candidate.

Is it possible that the reduction in memory is correlated with operation of the tool-at-dock switches? Bear in mind that there may be contact bounce when they operate.

[Diamondback]

@dc42 no, these switches are not operated during the print at all (and I don't have any evidence that some sort of vibration would constantly trigger them by accident).
Maybe it's 3*8 bytes instead or something like this...

[dc42]

@Diamondback please install the firmware at https://www.dropbox.com/sh/wmrfidzkej24mcm/AAB7AFGrfMoYqssjbVKM5rwxa?dl=0 on the Mini5+ and monitor part of a print with M122 as before. I have added new fields heap top and heap limit to the report. When it has settled down, I expect either heap top to increase by 24 byes at a time, or heap limit to decrease by 24 bytes at a time. I need to know which.

[dc42]

@Diamondback further to my previous message, I have found and fixed a possible cause of the memory leak. Please try the firmware at https://www.dropbox.com/sh/ihw972tlfq3gmj3/AABxTz_F0RYRC3i124PL4Wvma?dl=0. Cancel that, I need to make a change to the fix.

[dc42]

@Diamondback second attempt: new firmware with possible memory leak fix at https://www.dropbox.com/sh/ihw972tlfq3gmj3/AABxTz_F0RYRC3i124PL4Wvma?dl=0.

[Diamondback]

@dc42 Which particular 6HC firmware should I use in combination with this? Is your other custom build from the reset/update via CAN issue ok?

[Diamondback]

Trying the fixed one first before the one with more debug output. Is the never used RAM supposed to be perfectly stable? It still does go down by 24bytes at a time, however a LOT slower than before.
Only lost 72 bytes in the last 6 minutes now and only 480 since I started the print 11 minutes ago.

[Diamondback]

Looks like it stopped decreasing now at 28076, the print finished without issues. Now trying to print a second time without any reset.

[dc42]

@Diamondback yes when you first print, expect the free memory to go down and then stabilise. It may go down in two sizes of step, one will be 24 (max. 120 steps) and the other will be greater (max. 40 steps). If you do a second print, it should not go down any more.

[Diamondback]

@dc42 After a few consecutive prints without resets, it still sits at 28076. So I guess this is fixed! Awesome work, thanks a lot!