Bug in 'daemons/hw_ftdi.c'

[Benjamin K. <be...@kr...>]

I believe I have found a bug in 'daemons/hw_ftdi.c'.

Line 440 should read:
pulsewidth = f_sample * ((__u64) (pulse & PULSE_MASK)) / 1000000ul;

Instead of what it currently reads:
pulsewidth = f_sample * ((__u32) (pulse & PULSE_MASK)) / 1000000ul;

I've attached a patch file with this change.

I discovered this bug while using this configuration file:
http://lirc.sourceforge.net/remotes/motorola/DCT2000
(included in the Ubuntu distribution of LIRC as:
/usr/share/lirc/extras/transmitters/motorola/dctxxxx.conf)

In that file, the header is specified as:
header       9036  4424

The relevant pieces of code in 'daemons/hw_ftdi.c' are:

static int tx_baud_rate = 65536; /* Default transmit baud rate */
__u32 f_sample = tx_baud_rate * 8; /* Default sampling rate 524288 Hz */
int pulsewidth = f_sample * ((__u32) (pulse & PULSE_MASK)) / 1000000ul;

The largest pulse from the configuration file has a duration of 9036us.
pulse = 9036;

Given that pulse:
9036 * 524288 = 4737466368
However, 2^32 = 4294967296
Difference    =  442499072

The result (verified using xmode2) was that the first header pulse was
lasting ~844us instead of 9036us.

Changing '__u32' to '__u64' resolved the issue. This was tested on a
machine running the 64bit version of Ubuntu 12.10.

This bug impacts the Motorola set-top-boxes, which from my experience
are the majority of the set-top-boxes distributed by cable providers in
my area (Boston, MA). I recently switched from Comcast to RCN as my
cable provider, and both use Motorola cable boxes which use this same
configuration file. This bug prevents the home-brew FTDI based
transmitters (using Albert's design from http://www.huitsing.nl/irftdi/)
from communicating with these set-top-boxes.

- Ben

-- 
Benjamin Kraus
be...@kr...