[BRLTTY] [Nvda-dev] Braille support?

James Teh jamie at jantrid.net
Sun Sep 14 19:21:47 EDT 2008


Hi all,

Resending this private mail to Samuel to the list now that I'm 
subscribed, as there may be others who can provide input.

Samuel Thibault wrote:
>> If I attach my display, brltty picks it up and starts using it. However,
>> if I power it down, it powers down due to idle timeout or I disconnect
>> it, brltty will never detect it again. Worse, it seems to cause USB to
>> stop functioning correctly. I suspect this relates to libusb-win32.
> I'm afraid that's libusb-win32's fault then indeed.
I've done some further testing on this. Here's the story:
* libusb-win32 has some serious problems, at least on my system. I 
haven't yet had a chance to test this on another system. Basically, if I 
ever disconnect the display while brltty is running, USB stops 
functioning altogether to the point where plugging other USB devices 
into any of my USB ports has no effect. It is as if USB just froze in 
the state it was in before I disconnected the display. I'm not sure 
whether there is anything brltty can do about this (some sort of release 
that isn't happening when it detects the device is disappearing?), but I 
suspect not.
* My display comes with a driver which emulates a serial port (com4 on 
my system), so I am now using brltty with serial:com4 as the device. 
This works much more nicely. Disconnecting and reconnecting the display 
works as it should.
* Even with this serial:com4 configuration, powering down the display 
causes problems. I see
Baum Powerdown: power switch
Subsequently, brltty starts chewing massive amounts of CPU. If I use -l 
debug -e, I see it trying to probe:
probing with baum protocol
but it seems to freeze at this point. I suspect there is some oddity in 
Windows with writing to a device which is not ready, even though it is 
present. If I disconnect the display at this point, brltty comes back to 
life such that connecting it again brings everything back to normal once 
the display is redetected. Starting brltty with the display plugged in 
but not switched on yields the same result, except that I don't see the 
probing message.

> Did you try
> libbraille, just to check whether it has the same problem?
I seem to remember having major problems with libbraille last time I 
tried. I'll try it again when I get a chance and see what happens.

>> brltty seems to behave rather badly after a while if no device is
>> attached. It chews up a lot of CPU as if it is spinning at a rather
>> insanely high rate.
I did see this once, but this may relate either to libusb-win32's bad 
behaviour or the issue mentioned above. I suggest we try to figure those 
out before tackling this one, as they may be related. :)

>> The service often seems to fail to start on my system, even though
>> brltty's greeting messages appear on my display. After the brltty
>> version, it says "1 startup problem". If I don't hit scroll forward
>> within the next few seconds, the service will fail to start.
I figured out that this one was caused by my configuration file having 
the wrong permissions. (In Windows land, the system account doesn't 
necessarily have permission to read files owned by another 
administrator. Grrr!) It does seem, though, that the service doesn't 
"detach" soon enough sometimes. For example, fi there is a startup 
problem, it should still continue as is by reporting that startup was 
successful to the service manager.

>> Is there a
>> way to get debug output from the service? The Windows event log doesn't
>> seem to show the messages correctly - has this been implemented?
> Odd, it should show up there indeed. I guess you can add -e and redirect
> to a file to get the debugging output.
You can't do this as a service. However, interestingly enough, I noticed 
that brltty seems to log (partial) messages to the event log when 
running normally, but it doesn't when running as a service. Any ideas?

I've also found some bugs in the brlapi Python bindings which I'll 
report (or perhaps patch) in due course.

Thanks!

Jamie


--
James Teh
Email: jamie at jantrid.net
WWW: http://www.jantrid.net/
MSN Messenger: jamie at jantrid.net
Jabber: jteh at jabber.org
Yahoo: jcs_teh


More information about the BRLTTY mailing list