On Sun, 2009-01-11 at 23:36 -0800, Dmitry Torokhov wrote:

Hi Johannes,

On Tue, Jan 06, 2009 at 12:26:24PM +0100, Johannes Berg wrote:

quoted

Whenever you press and then release a key, the CPU wakes up
three times:
 * press
 * release
 * autorepeat timer exactly 250ms after press

The autorepeat timer has nothing to do, obviously, since you already
have released the key, so stop it on key release.

This introduces a slight change in behaviour (the key that is released
may not be one that is being autorepeated) but I think it still makes
sense to do it.

Hmm, good point, I thought it didn't, but it does when you press a, b
wait, release a then b doesn't continue autorepeating. Seems like a
corner case though, so I agree. We could fix that by checking which key
is being autorepeated though, I'd think, see below (untested as of now)

johannes

--- wireless-testing.orig/drivers/input/input.c	2009-01-10 11:24:06.000000000 +0100
+++ wireless-testing/drivers/input/input.c	2009-01-12 09:49:58.000000000 +0100

@@ -132,6 +132,14 @@ static void input_start_autorepeat(struc
 	}
 }
 
+static void input_stop_autorepeat(struct input_dev *dev, int code)
+{
+	if (dev->repeat_key != code)
+		return;
+
+	del_timer(&dev->timer);
+}
+
 #define INPUT_IGNORE_EVENT	0
 #define INPUT_PASS_TO_HANDLERS	1
 #define INPUT_PASS_TO_DEVICE	2

@@ -167,6 +175,8 @@ static void input_handle_event(struct in
 				__change_bit(code, dev->key);
 				if (value)
 					input_start_autorepeat(dev, code);
+				else
+					input_stop_autorepeat(dev, code);
 			}
 
 			disposition = INPUT_PASS_TO_HANDLERS;