speakup todo?

[Samuel Thibault]

Hello,

Littlefield, Tyler, le Sun 16 Sep 2012 12:17:15 -0600, a écrit :
> I'm interested in learning more about kernel programming, and I figured I'd
> start by working on something I use almost daily.

That's usually the best way: it brings both motivation and good testing.

> I'm curious then if there's some sort of todo or improvements speakup
> could have to it.

See the TODO file in the speakup repository, and the TODO file in
linux/drivers/staging/speakup

> I'd also be curious if someone has thought about moving it to userspace

There are actually already some userspace readers: yasr, brltty.

> --as far as I know, the only thing that we really need the kernel for
> would be hardware speech

Actually only for a couple of them. Others can be driven from /dev/ttyS0
from userland.

> and accessing the console directly.

That, however is an important point.

> We could use sequence files and access the console through /proc. It could
> return a file of 2-byte chars, which I believe is how it works now--one byte
> is the color, and the other byte is the ascii value.

This is already available as /dev/vcsa*, and what brltty uses. But
that's not very convenient, because it only provides a snapshot, and
not the whole flow. If you cat a file, for instance, you want it spoken
completely, and not only the last screen. Without pausing the cat
process from the kernel, you won't get it.

The kernel could expose the flow to userland like it currently does to
speakup, but the pause feature would probably be frowned up. It'd
easily get stuck for good.

> If there are people willing to answer questions from time to time in
> terms of the kernel programming, since that's something I've not done
> before, I'm game to start coding.

See the kernelnewbies website and IRC channel, it's meant for that.

> Another question is then, how do people catch panics?

Through a serial console, for instance (console=ttyS1)

Samuel