[BRLTTY] missing key release events, was Re: brltty5.2 alpine2.11 cursor tracking

[Stéphane Doyon]

On Sun, 4 Oct 2015, Dave Mielke wrote:

> [quoted lines by Nicolas Pitre on 2015/10/04 at 16:17 -0400]
>
>> And here's the culprit: commit 42bebdf197
>
> Thanks. That was done to make AtSpi2 work better. I've undone it for now.
> Please test the latest code.

Yes! Yes! That seems much much better!
OK it's only been 5 minutes, but I'm pretty sure.

I have indeed been experiencing the issue exactly as Nicolas described. 
Was trying to catch a good segfault backtrace but somehow could not manage 
it. I once caught a haep corruption warning from the memory sanitizer 
thingy but couldn't really make sense of it and had difficulty reproducing 
it...

Nicolas' idea of doing lots of cursor routing does seem like a good repro 
recipe. I reproduced the speech hang in a few seconds before doing the git 
pull, and now can't reproduce the issue. It's not just that though. Before 
it felt like some kind of sluggishness had been introduced, not entirely 
well defined but it seemed like speech echo at least was definitely 
sluggish, and that seems much better after your revert.

This all sounds a little subjective and not very actionable, which is why 
I hadn't spoken up yet and was still looking for some more definite info. 
But this is a huge relief, the sluggishness and speech hangs / forced 
restarts were frustrating. Huge thanks Nicolas for the bisection!

Now, how can we figure out what the problem was?