7i73 LCD shows rubisch

newbynobi wrote:

I mentioned in an ealier post, the push that destroid the display communication.

That push did not have any effect on the lcd display for me. And as it did not touch the process data part of the driver code, I am at a loss to explain the apparent correlation.

I also found that the display worked just as well for me before and after the push, so I have no way at all to debug the problem.

I _did_ find that for very short display strings ("A|B|C|D") the display was jumpy and bad with _both_ 2.7 and the latest 2.8, so I fixed that hoping to fix the underlying issue.

Can you please explain, why you do change lcd.comp? It might not be clean coded, but the communication problem must be related to something else.

We have also found a long-standing race condition in the sserial driver (also present since at least 2.6. That has been fixed and _also_ has no affect on the LCD on my test system.

I have just reseted the computer from my machine to the working branch and since that step, my displayworks fine!

Which working branch?

newbynobi wrote:

Hallo Andi,

reseted master hard to be 6f0f2cafb89a914ea392c65d2e3dcbe754b1638d
(control.c: honor fractional accels for wheel jogging)
working correct!

The diff is hard to make sense of as I moved a lot of code to different places, and also accidentally introduced a lot of white-space changes.

github.com/LinuxCNC/linuxcnc/compare/6f0...1656abb536fd4c550fcd

But the critical code is hm2_prepare_tram_write:
github.com/LinuxCNC/linuxcnc/blob/2.7/sr...mot2/sserial.c#L1173
And specifically the part where it writes the stream data in line 1307

The above is as it was in 2.7.
This is the same code in master:
github.com/LinuxCNC/linuxcnc/blob/master...mot2/sserial.c#L1771

Which is, admittedly, now factored out into a function.

The parameter setting code should not even run unless there are changes to parameters in the HAL file. (and even then only if the HAL values are different to those stored in the remote)

Maybe we need to compare ethernet traffic logs to see if there are differences?

Norbert: One other thing, can you test the latest Master with the LCD and sserial updates from Sunday?

One other experiment to try, can you put the 7i73 into mode 2?
(sserial_port_0=22xxxxxx)
That is "fast display" mode. It makes no difference here (it works perfectly in either mode), but it's an easy experiment.

Arggh!! Dewey suggested this and made a fix, and my testing seemed to show that it wasn't the problem...


Try commenting out TWOPASS = true

I finally managed to make a config misbehave, and turning off twopass fixes it.

The problem seems to be that TWOPASS strips \n characters from argument strings.

Try Dewey's new branch, I think it might fix the problem (fixes it here).
github.com/LinuxCNC/linuxcnc/commit/185f...6757e77d9bc983b8fef7

Dewey's TWOPASS fix is now in Master. Can you test again?