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

[Nicolas Pitre]

On Thu, 1 Oct 2015, Nicolas Pitre wrote:

> Now to complicate things further, I just replaced my workstation for a 
> much more powerful one.  The lost key issue doesn't appear to manifest 
> itself anymore but I now have other kinds of brltty problems resulting 
> in complete loss of braille output on random occasions (smells like 
> brltty segfaulting) I have yet to investigate properly.

Okay... In fact the lost key event is still there with my new PC.  And 
much more too!  Now I lose speech entirely on random occasions.  When 
that happens, doing a RESTARTSPEECH simply kills braille output as well!  
And I'm able to reproduce this in less than 2 minutes which makes it 
much easier to bisect.  Oh, and when braille is dead, I must issue a 
"killall -9 brltty" before restarting it or it doesn't work.  And then 
only speech works.  I also have to yank the USB cable from my display 
and plug it back before braille is available too.

Apparently, this issue is affecting Stéphane Doyon as well.

And here's the culprit: commit 42bebdf197

A build of brltty right before that commit is rock solid and stable.  
Going back to commit 42bebdf197, I can confirm that trouble occurs after 
a cursor routing action.  One such action out of 20 or so would fail to 
bring the cursor to the desired position half way, and from that point 
speech becomes dead and RESTARTSPEECH kills braille too.  I've also 
experienced weirder behaviors with slugish braille output on rare 
occasions, but if memory corruption is at play then anything can happen.


Nicolas