[PATCH libinput] touchpad: serial synaptics need to fake new touches on TRIPLETAP

[Hans de Goede]

Hi,

On 24-07-15 04:54, Peter Hutterer wrote:
> On the 4.1 kernels synaptics pretends to have 3 slots (the serial fw only does
> 2). This was added to avoid cursor jumps but has since been reverted.
> In some cases a TRIPLETAP may be triggered without slot 2 ever activating.
>
> While there are still those kernels out there, work around this bug by opening
> a new touch point where none exists if the fake finger count exceeds the slot
> count.
>
> Reported-by: Jan Alexander Steffens <jan.steffens at gmail.com>
> Signed-off-by: Peter Hutterer <peter.hutterer at who-t.net>
> Tested-by: Jan Alexander Steffens <jan.steffens at gmail.com>

I'm not sure if we really want to work around kernel bugs like this, hasn't the
revert of the broken commit been scheduled for a 4.1.x update ?

IMHO we should just urge people ti upgrade to 4.1.x then.

> ---
>   src/evdev-mt-touchpad.c | 19 ++++++++++++++-----
>   1 file changed, 14 insertions(+), 5 deletions(-)
>
> diff --git a/src/evdev-mt-touchpad.c b/src/evdev-mt-touchpad.c
> index 6718b61..376f394 100644
> --- a/src/evdev-mt-touchpad.c
> +++ b/src/evdev-mt-touchpad.c
> @@ -368,13 +368,22 @@ tp_restore_synaptics_touches(struct tp_dispatch *tp,
>   	for (i = 0; i < tp->num_slots; i++) {
>   		struct tp_touch *t = tp_get_touch(tp, i);
>
> -		if (t->state != TOUCH_END)
> +		switch(t->state) {
> +		case TOUCH_HOVERING:
> +		case TOUCH_BEGIN:
> +		case TOUCH_UPDATE:
>   			continue;
> +		/* new touch, move it through to begin immediately */
> +		case TOUCH_NONE:
> +		case TOUCH_END:
> +			tp_new_touch(tp, t, time);
> +			tp_begin_touch(tp, t, time);
> +			break;
> +		}
>
> -		/* new touch, move it through begin to update immediately */
> -		tp_new_touch(tp, t, time);
> -		tp_begin_touch(tp, t, time);
> -		t->state = TOUCH_UPDATE;
> +		/* touch just ended ,we need need to restore it to update */
> +		if (t->state == TOUCH_END)
> +			t->state = TOUCH_UPDATE;

This will now never trigger, if the touch was in TOUCH_END state it has gone
through tp_begin_touch in the switch/case and will now be in TOUCH_BEGIN.

Regards,

Hans