Ion not responding to numeric keys

Hi Alister, 

haven't seen it before. Do you have a USB keyboard connected? Could it be that someone lays something on it pressing Numlock key? Could you try pressing Numlock when that happens again?

Yes - but if Eos registered a BUTTON Down Press - meaning something was laid on top of the Keyboard, and then the console rebooted, and something was still lying on the Keyboard, it won't send another button down press until its picked up and placed down on the Keyboard again, so the Problem would seem to have gone away with a reboot. But as Nick suggested - keeping an active thought about whats on the Keyboard and also swapping out the possibly defective Keyboard may possibly keep the Problem from reoccurring.

I did a bit of digging in the logs myself. I found the point where when my colleague was using the console it suddenly stopped responding to numeric keys, where, as was noted, a Tab -> Tab_Down occurs without a matching Tab_Up. What then follows is that most of the keys appear to be acknowledge, but the numeric keys don't. Any operation then appears to refeence Cue 1, pallete 1 etc. At various points the numeric keys can be seen being dequeued as a number of long numeric sequences which of course makes no sense to the console (the non numeric keys having been processed but of course in the wrong context. so it discards them and raises a syntax error.

The thing is, the identical problem had happened to me on the previous show. So I did some detective work and found the same Tab->Tab_Down without matching Tab_Up, followed by the out of sequence dequeuing  of the numerics. So far so good.

Spurred on by this, I've just tried an experiment in the offline editor. Let's suppose you are recording a cue and you're typing the label on the little keyboard, and let's suppose you accidentally clip the tab key with your fat fingers while you're typing. So you think you've typed

Record Cue 3 Label With Tabs

But actually you've typed

Record Cue 3 Label With Ta<Tab>bs

You get no visual indication of this, as tabs aren't valid in labels so just get discarded, except they don't completely, because when you press enter to complete the label the console still thinks the tab key is down, and the next numeric entry does indeed appear with the tiny, tell tale red "tab 34" (or whatever) legend showing that the console"thinks" it's processing a tab selection command. This is reproducible, and I think it's a bug, as it has the whiff of an error in the keyboard processing's state machine.

Pressing tab does clear the condition, however, so now we know about it it's not catastrophic but it is something I think needs looking at as it is confusing.

I really don't know where to start, but I'll try to address the issues in order.

First, until I started analyzing the logs, I hadn't made the connection. The two lockups we experienced occurred several weeks apart. Hindsight is a wonderful thing.

Second, I hadn't realised (sorry, my clairvoyance is sometimes weak) that when I'm asked to send the logs to a specific email address at etc that I should also 9off my own bat) post them into the open forum.

Third, you may have known a priori that external devices supplied with the console are known to cause problems, but not everyone has the benefit of this experience, so we sort of assumed that what was supplied with the console was supposed to work with it.

Fourth, we still aren't absolutely certain that it was Tab being pressed during a labeling operation, although that is indeed looking more and more likely. If you are now saying that, in your opinion, code will be added to fix this problem then (in my view)  that legitimises it as a problem, and gives the lie to PICNIC, books on the keyboard, misuse et.c. as previously suggested.

Fifth, I'm also very protective of my development team, being very used to fielding complaints from people who mud sling when they wish to deflect blame. Someone wants to blame my team, bring me the evidence.

In this case I don't believe I did mud sling - I offered a possibility of events being lost, based on my experience developing real time high through put systems. Next time, I'll say nothing.

Sixth, what comment did I make regarding OSC that was in any way other than enthusiastic?

I want to get this problem solved  (and I think we've now probably shown that there possibly is a state-machine based problem somewhere) and if I have offended the developers along the way then I apologise TO THE DEVELOPERS. As a developer myself it was never my intention to to put any developer noses out of joint. I know myself how tricky software development can be and how it can sometimes seem a thankless task.

I do have to ask though, Nick - are you an ETC employee? Up until now my experince with ETC has been nothing less than positive. This thread leaves a bad taste in my mouth