=?UTF-8?Q?lirc=5Fserial=20bugfix?=

[Andrei T. <an...@un...>]

The attached patch fixes two issues in lirc_serial
1. potential active level autodetection error (if there happens to be an IR signal present at the moment when the driver is loading). The new code takes 9 samples with 0.1s intervals and selects active high/low based on the maximum number of "votes".
2. if 2 consecutive "pulse" interrupts occur with more than 15 seconds interval, the driver would decide that "this MUST be a space" and flip the "sense" bit (even if it wasn't in autodetection mode!), becoming "deaf". Under normal conditions this (almost) never happened, with suspend/resume working the problem became very obvious. The patch just deletes this piece of code.

Regards,
Andrei.