Skip to content

Commit

Permalink
staging/speakup: Update TODO list
Browse files Browse the repository at this point in the history
Thanks to Okash's latest work, the TODO list is essentially empty, so
the way out from staging now seems open.

The status of the remaining issue mentioned in TODO is not clear, we
asked the speakup user mailing list for reproducer cases, but didn't get
a really specific scenario. One serious bug was fixed by 9d32c0c
("staging/speakup: fix get_word non-space look-ahead"), which does not
seem to really be related to the bug mentioned in TODO. Possibly the bug
mentioned in TODO has been fixed long ago and people have been thinking
that it was not because they can not distinguish the symptoms mentioned
in TODO from the symptoms of the bug fixed by 9d32c0c.

Signed-off-by: Samuel Thibault <[email protected]>

Link: https://lore.kernel.org/r/20200726165452.qhos3wrjcm6jmcmx@function
Signed-off-by: Greg Kroah-Hartman <[email protected]>
  • Loading branch information
sthibaul authored and gregkh committed Jul 28, 2020
1 parent 9038c6a commit 53754e4
Showing 1 changed file with 6 additions and 31 deletions.
37 changes: 6 additions & 31 deletions drivers/staging/speakup/TODO
Original file line number Diff line number Diff line change
Expand Up @@ -6,42 +6,17 @@ Speakup is a kernel based screen review package for the linux operating
system. It allows blind users to interact with applications on the
linux console by means of synthetic speech.

Currently, speakup has several issues we know of.
Currently, speakup has one issue we know of.

The first issue has to do with the way speakup communicates with serial
ports. Currently, we communicate directly with the hardware
ports. This however conflicts with the standard serial port drivers,
which poses various problems. This is also not working for modern hardware
such as PCI-based serial ports. Also, there is not a way we can
communicate with USB devices. The current serial port handling code is
in serialio.c in this directory.

Some places are currently using in_atomic() because speakup functions
are called in various contexts, and a couple of things can't happen
in these cases. Pushing work to some worker thread would probably help,
as was already done for the serial port driving part.

There is a duplication of the selection functions in selections.c. These
functions should get exported from drivers/char/selection.c (clear_selection
notably) and used from there instead.

The kobjects may have to move to a more proper place in /sys. The
discussion on lkml resulted to putting speech synthesizers in the
"speech" class, and the speakup screen reader itself into
/sys/class/vtconsole/vtcon0/speakup, the nasty path being handled by
userland tools.

Another issue seems to only happen on SMP systems. It seems
that text in the output buffer gets garbled because a lock is not set.
This bug happens regularly, but no one has been able to find a situation
which produces it consistently.
It seems to only happen on SMP systems. It seems that text in the output buffer
gets garbled because a lock is not set. This bug happens regularly, but no one
has been able to find a situation which produces it consistently.

Patches, suggestions, corrections, etc, are definitely welcome.

We prefer that you contact us on the mailing list; however, if you do
not want to subscribe to a mailing list, send your email to all of the
following:

[email protected], [email protected], [email protected] and
[email protected].

[email protected], [email protected], [email protected],
[email protected] and [email protected].

0 comments on commit 53754e4

Please sign in to comment.