On Tue, Oct 25, 2016 at 11:59:22AM +0200, Arnd Bergmann wrote:

On Monday, October 24, 2016 4:45:13 PM CEST Dmitry Torokhov wrote:

quoted

quoted

diff --git a/drivers/input/keyboard/tca8418_keypad.c b/drivers/input/keyboard/tca8418_keypad.c
index 9002298698fc..3048ef3e3e16 100644
--- a/drivers/input/keyboard/tca8418_keypad.c
+++ b/drivers/input/keyboard/tca8418_keypad.c

@@ -164,11 +164,18 @@ static void tca8418_read_keypad(struct tca8418_keypad *keypad_data)
      int error, col, row;
      u8 reg, state, code;
 
-     /* Initial read of the key event FIFO */
-     error = tca8418_read_byte(keypad_data, REG_KEY_EVENT_A, &reg);
+     do {
+             error = tca8418_read_byte(keypad_data, REG_KEY_EVENT_A, &reg);
+             if (error < 0) {
+                     dev_err(&keypad_data->client->dev,
+                             "unable to read REG_KEY_EVENT_A\n");
+                     break;
+             }
+
+             /* Assume that key code 0 signifies empty FIFO */
+             if (reg <= 0)
+                     break;

I am unconvinced that this rearrangement fixes the issue for all older
GCCs. Can we simply do:

        u8 uninitialized_var(reg);

and be done with it?

Yes, that would work. However:

- avoiding the fake intialization tends to produce better object
  code, as gcc actually knows what's going on
- Linus doesn't like uninitialized_var() and would rather see it
  removed from the kernel
- llvm produces warnings for uninitialized_var()

I have checked gcc-4.6/4.7/4.8/4.9/5.x/6.x, and only gcc-4.9
produces the warning. 4.9 changed the detection for uninitialized
variables significantly, which generally caused fewer false
positives but unfortunately introduced a couple of new ones
like this.

OK, I'll apply it then.

Thanks.

-- 
Dmitry