fix the bloody longterm vtimer bug

[LudwigKnuepfer]

You know who you are, I'm looking at you!

[LudwigKnuepfer]

PS:
If you want to test this, I suggest reducing SECONDS_PER_TICK and MICROSECONDS_PER_TICK to 10 seconds or so... (https://github.com/RIOT-OS/RIOT/blob/master/sys/vtimer/vtimer.c#L22)

[LudwigKnuepfer]

@kaspar030 This will need your ACK for the license again.

While you're at it: If you can still explain the purpose of https://github.com/RIOT-OS/RIOT/blob/master/sys/vtimer/vtimer.c#L70 I'll forgive you for not commenting it in the first place ;-)

    uint32_t next = hwtimer_next_absolute + longterm_tick_start;
    uint32_t now = HWTIMER_TICKS_TO_US(hwtimer_now());

    /* <YOUR COMMENT HERE> */
    if((next - HWTIMER_TICKS_TO_US(VTIMER_THRESHOLD) - now) > MICROSECONDS_PER_TICK ) {
        next = now + HWTIMER_TICKS_TO_US(VTIMER_BACKOFF);
    }

    hwtimer_id = hwtimer_set_absolute(HWTIMER_TICKS(next), vtimer_callback, NULL);

[OlegHahm]

The parts I understood are fine and the rest seems at least not break anything more than before: ACK.

[LudwigKnuepfer]

@kaspar030 ACK?

[mehlis]

tested on native with ccn, I see no problems!

[mehlis]

you can cherry pick https://github.com/mehlis/RIOT/commit/e0cdd99f6ec7d0fd7db9c9790605bf759f7b6c59 to fix all warnings in the vtimer (yay!)

[LudwigKnuepfer]

@kaspar030 ping

[kaspar030]

@LudwigOrtmann compiling now...

[LudwigKnuepfer]

@kaspar030 you should acquire a faster computer... ;-)

[kaspar030]

@LudwigOrtmann Actually it compiles and sems to work, but then I didn't really know how to test longtime timers, then I wanted to write an automated test, then there was other stuff to do. But as basic tests worked fine, I'll go with oleg. Can only be better, so
ACK. And good job, btw. ;)