cpu: native: fix thread_yield_higher isr_is_in() case

[kaspar030]

Previously, native created a new "setcontext" context whenever thread_yield_higher() was called. That's unnecessary and broken. At the end of interrupt our interrupt handlers, sched_context_switch_request is checked and triggers an eventually needed context switch.

Fixes #6642.

[OlegHahm]

I came to a similar solution. What I don't quite understand:
at the end of native's IRQ handler (native_irq_handler()) the function cpu_switch_context_exit() gets called. However, the documentation of this function says: "Call context switching at thread exit". Apparently, this function serves more than this purpose on native.

[OlegHahm]

The change seems valid to me, but I'd like to run a few more tests before I ACK.

[LudwigKnuepfer]

I think isr_thread_yield is not called from anywhere else - please remove if this is correct.

[OlegHahm]

It's not correct, see https://github.com/RIOT-OS/RIOT/pull/6660/files#diff-ae383327c6bd21463d3010729a9284b3R221

[kaspar030]

What I don't quite understand:
at the end of native's IRQ handler (native_irq_handler()) the function cpu_switch_context_exit() gets called.

Yeah, that's weird. IMHO unrelated, though.

[OlegHahm]

True

[OlegHahm]

ACK

[kaspar030]

@OlegHahm Does this patch also solve the stability issues you had when testing yout thesis?

[OlegHahm]

Trying hard to remember which issues you're talking about...

[OlegHahm]

I thought they were already fixed with #2071.

[OlegHahm]

But I found 9930e17 in my working branch that may be related.

[kaspar030]

But I found 9930e17 in my working branch that may be related.

Yes, that's the one I meant, also in #6135.