Mesa UART doesnt work after update to linuxcnc2.8

i was using linuxcnc 2.8 about 1 year ago,recently i decided to update it to release version 2.8.
Before updating, i can use multi UART channels but now only 1 channel work. example: i use 4 UART channels, only the channel 0 work, then i disable channel 0, channel 1 will work.
I tried look at uart.c in mesa-hostmot driver but nothing difference between them.

did you update the OS also to debian stretch
and the master is at 2.9

I did a quick look for an obvious commit - but found nothing obvious to me.
Can you use git bisect to find the commit that broke the behavior?

Also Andy P probably can make more intelligent suggestions -I'm sure he will chime in.

Chris M

Thats odd, I dont think there have been any UART related changes in 2.8 at all

The individual enable/disable logic seems to be in the mesa_uart component
so thats a likely suspect

1. i'm using linuxcnc 2.8 on Ubuntu 16.04 with Preempt-rt, i dont think it's a reason
2. i tried git bisect but i dont see any bad revision show up
3. Here is what happen after i make a deeper text:
+ I used 4 mesa_uart(modified), everything works well in my old version
+ When i use newer version 2.8.
- If i load 4 uart channels and use all, only channel 1 work but data store in channel 1 is come from channel 4.
- If i disable channel 1 (still load it) then channel 2 will work but data store in it also come from channel 4.
- if i disable channel 4( keep 1,2,3 running), not anychannel work, no data is stored

Is this on identical hardware/firmware as 2.7?

PCW wrote:

Is this on identical hardware/firmware as 2.7?

No i didnt test it on 2.7 version. Both version are 2.8 and built in rip mode.

I'm not clear on what has changed, have you changed OS, Firmware, or hardware?

As far as I can by the file history tell nothing major changed in the UART code recently

There is not much change, i just switch from an old 2.8 version that i download about a year ago to recent 2.8 version. Both are built in same OS, Firmware and hardware even config file.
My bug look like all data is overwrited because it's fine if i use only1 channel but if i use 2 channels, the data is stored in channel 1 come from channel 2 and channel 2 dont have any data

So I would have to agree with cmorely
It probably will require git bisect to find the
commit that broke things