LIRC not working after October update

I guess that I have an unusual configuration for the Vero 4k. I don’t use the RF remote. I use an IR remote with 2 device code sets. I don’t like eventlircd because it ignores the remote name and sends everything on as “linux-input-layer” which doesn’t work for me. I would like to use lircd, but in the 2020-October build, a couple of things are broken.

  1. Any button press is repeated over a hundred times in linux-input-layer. This does not happen on the RPI version of 2020-October.

  2. For the lircd path (socket /var/run/lirc/lircd-lirc0) the first keypress is ignored. After the first press, I can hit keys and they show up until I stop hitting buttons for a while (about 5 seconds). This is roughly the amount of time that the autorepeat goes on for, but that may just be coincidence.

I would like to figure out a way to get back to a simple lircd configuration that I used to have long ago. I have tried disabling eventlircd, but that seems to also disable lircd. I am kind of a novice at this just, poking around until something works, but that is not very efficient. Any pointers would be appreciated.

I think my missing keypress issue is in the kernel. I found this bug report:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=a8f29e89f2b54fbf2c52be341f149bc195b63a8b

Does this mean anything to you? I don’t know what to do with it.

I finally got to a working system. Here is what I had to do. I had to abandon the USB IR receiver and go back to the internal (or extended) IR receiver. I think that that by itself my have gotten me back to the functionality that I had in the June 2020 build. However, I decided to go a step further…
I disabled eventlircd (sudo systemctl mask eventlircd) since that combines all remotes into one set. I need two remotes to be differentiated. I dealt with this in the past by giving the buttons on the second remote different names. By using lircd instead of eventlircd, I can use any names for the buttons (I am not restricted to a finite set of standard names) and it will attach the remote name to the code returned.
I only hope that future releases won’t break this (again). It is the bane of having niche requirements.