support for HW_SKIP_CONSOLE breaks use by blind people

Michal Suchanek hramrach at gmail.com
Thu Jan 10 06:20:21 PST 2013


On 10 January 2013 13:50, Samuel Thibault <samuel.thibault at ens-lyon.org> wrote:
> Michal Suchanek, le Thu 10 Jan 2013 12:38:13 +0100, a écrit :

>> > That said, it's of course way simpler, instead of having to modify all
>> > drivers, to just make video-dummy and input-void set the SKIP flag,
>>
>> That's not going to work. If you do not set the flag in dummy the
>> console will open whatever you set in void. The console is open on
>> every driver probe starting with dummy. The HW_SKIP_CONSOLE hack
>> abuses the internal knowledge of driver probing order to set the flag
>> in the driver which is probed first but which is the wrong one for
>> setting the flag.
>
> Re-read the source code. The console will be opened as soon as either of
> the drivers do not set the SKIP flag.

Yes, that's why the flag has to be inverse. Alternatively the flag
handling could be improved so that it's checked for every driver and
not saved in the X server based on the first driver.

Dummy driver sets the flag because it does not need console and then
the console is not open although evedev which requires console and
does not set the flag is loaded later. X server saves the flag when
it's set on the first driver and opens the console when it is not set
on the first driver so the flag state on any latter driver is
irelevant.

Possibly handling the SKIP flag in hotplug would suffice on Linux
because then when evedev is plugged it will not set SKIP and the
console is opened in hotplug. This would not work on platforms that do
not use hotplug. however.

Thanks

Michal


More information about the xorg-devel mailing list