sys: xtimer: WIP: xtimer refactoring

[malosek]

Hello,

addressing issue reported in #4902. The idea is to get rid of timer_set_absolute with 32bits target.
Now all the magic happens in _xtimer_set64 where there is irq disabled between getting the current absolute time and target calculations.
have had look at xtimer_sleep_until as well - not to duplicate code from _xtimer_sleep and use 64bits last_wakeup arg.

It it still WIP (need to adapt the rest of the sources for new xtimer_sleep_until arg), but would like to get general feedback.

wbr
malo

[kaspar030]

We have to solve #5338 first, otherwise, disableIRQ(), xtimer_now64() doesn't work as the overflow interrupt might get lost.

[kaspar030]

@malosek Can you take a look at #5428? It contains some ideas from this PR and should also fix #4902.

[malosek]

Hello @kaspar030 ,
quite a rework - finally:), looking forward, will try to give it play in "near" future.
wbr
malo

[kYc0o]

@kaspar030 @gebart are these fixes addressed by one of your PRs? Can you take a look?

[jnohlgard]

I think #5428 will fix this, @kaspar030 what do you think?

[kYc0o]

As #5428 was postponed, I think this one needs that too, however we could also look if that PR is addressing the same issue as this one.

[miri64]

#5428 was postponed so this one needs to go too (+ Kaspar is on vacation)

[OlegHahm]

What's the state of this PR?

[kYc0o]

Maybe #5607 #5608 and #5612 solved (partially) this?

[kYc0o]

I'd like to know if we can close this PR if the fixes it provides are already addressed by already merged PRs...

[smlng]

IMHO this PR is quite outdated, it doesn't use the xtimer_ticks32/64 types and is way before the major rework done in #5608.

+1 for closing

[malosek]

closing...

wbr
malo