[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#782743: unblock: brltty/5.2~20141018-5



Control: tag -1 d-i confirmed

Samuel Thibault <sthibault@debian.org> (2015-04-17):
> We have just received a report from a blind user saying that braille
> auto-detection in the debian-installer does not work any more for his
> device. This is reported as bug report #782732. I could reproduce the
> issue, and fix it. This is pretty bad because it matters for 4 brands
> (Albatross, Cebra, HumanWare and HandyTech) of braille devices of the
> 30 supported brands, among which HumanWare and HandyTech are AIUI very
> commonly found. The price of such device makes it unaffordable for their
> user to use a replacement device, so they really have no way install
> debian 8.0 by themselves.
> 
> What happens is that these braille devices use the same USB ID,
> 0403:6001, because their manufacturers were too lazy to change the
> USB-to-serial chip ID.  So when one of these is connected, brltty has
> to try the 4 different drivers for these various devices, which happens
> to work fine because the protocols are different enough for this to be
> safe.

How do we love it when that happens…

> However, the probing function of one of those drivers, Albatross,
> actually uses an infinite loop, keeping trying between 19200 bauds and
> 9600 bauds. As a result, when brltty tries it, it gets stuck inside
> that probing loop. It happens that in the past the driver would get
> an open error at some point, thus breaking its loop and letting other
> drivers try their protocol. This open error has apparently been fixed
> in the latest version of brltty, and thus the probing gets stuck inside
> the Albatross driver. In the attached change which I have uploaded to
> unstable, I change this infinite loop into a simple loop over the two
> 19200 and 9600 bauds values, i.e. the driver will just try them both,
> and if they both fail, notice the end of available baud possibilities,
> thus abort, and thus let the brltty core give other drivers a chance to
> do their probing, before the brltty core tries Albatross again. This
> infinite loop inside the Albatross driver was indeed spurious since the
> brltty core already performs it.
> 
> I'm Cc-ing debian-boot for udeb confirmation.

Don't worry, I had that on my radar already. Looks good for u/u-u/u
(unblock, unblock-udeb, urgent).

Mraw,
KiBi.

Attachment: signature.asc
Description: Digital signature


Reply to: