Pi - Key release not recognised under CPU load.

I have seen numerous instances of key release not being seen when the CPU is carrying high load.

In particular, I have FTPs running which, disappointly, is very CPU hungry especially when largish files are being transfered.

Every so often a key release is missed with the effect of generating a string of repeat characters (only stopped by tapping another key). I have seen this on a number of applications including !Edit.

Also, the key release missed might be the shift key which causes subsequent characters to appear in upper case when they should not. (only fixed by tapping the shift key again).

So what version of the ROM image are you running on the Pi. Is it the one in RC5 or RC6 or a later development version from the ROOL software downloads.

I ask this as there have been some improvements according to the CVS logs.

Also might be good to state what type of keyboard you have and how it is connected as the report is not clear and doesn’t give enough information to help narrow things down.

OK, Sorry. Good points.

I am running a downloaded copy of RC6, with no updates.

The keyboard is a Genious USB keyboard, connected via a Belkin powered USB hub, which provides the power for the Pi. The mouse is also connected to the hub and the other USB socket on the Pi is currently empty.

There are also other Apps running including !NetSurf (on which I am typing with difficultyyyyyyyyyyyyyyyyyyyyyyyyy.), !Webjames and Wget. Could this just be another instance of the well known Pi USB problem? (Whatever operating system).

I don’t see any problems when FTPs is not running.

It would be straightforward to put in an auto-cancelling timer, so that missing key-ups get simulated after a short delay.

Leaning on the down-arrow won’t work quite so well (release and repress) but on the upside it doesn’t get stuck.

As I routinely have my machines configured like this, I’m badly affected by stuck keys:

Delay: 28
Repeat: 2

I have now come to the conclusion that this problem is best solved by not using !FTPs. Everything works fine without it. I do not think that it is just the high CPU load alone which is causing the problems.

It’s not the high CPU load that’s causing the problems, it’s the high USB load. (Pi’s ethernet is on USB.)

Raspberry Pi-specific bug, and it’s not a RISC OS-specific bug, either, it affects Linux just as much. The USB bus has trouble at high load.

Agreed. I did mention “Whatever operating system” in my post two hours ago.

Incidently, I can transfer the same 126K of data with no problems if I use !WebJames instead of !FTPs. (and Wget instead of FTP on the remote Pi running Raspbian Linux).

Incidently, I can transfer the same 126K of data with no problems if I use !WebJames instead of !FTPs. (and Wget instead of FTP on the remote Pi running Raspbian Linux).

I suspect the transmitted packet size may be the source of the better transfer of wget. You would need to do a bit of packet capture to prove it.

This may be fixed in the latest Alpha build, the problem certainly stopped for me a few builds back.