sys: xtimer: add xtimer_set_timeout_flag()

[kaspar030]

This PR adds an xtimer convenience function to set the timeout flag after a certain period.

This intentionally does not allow to set any flag (only THREAD_FLAG_TIMEOUT), or set a flag on any thread (not the calling one), for simplicity and performance reasons.

The thread_flags test is updated with a simple test of the new function.

[jnohlgard]

The changes to xtimer look sane.
I think that the unittests dependencies need updating. However, is it OK for a core feature test to depend on a sys module?

[kaspar030]

I think that the unittests dependencies need updating. However, is it OK for a core feature test to
depend on a sys module?

Which unittests do you mean? Thread flags don't have unittests (yet). The deps of tests/thread_flags are updated with the xtimer dependency.

[jnohlgard]

Re unit tests: I must have misread when I looked at the diff before.
Tested working on samr21-xpro.

There is another thread flags test thread_flags_xtimer which seem somewhat redundant now. Could it be removed? @haukepetersen @kaspar030
https://github.com/RIOT-OS/RIOT/blob/master/tests/thread_flags_xtimer/main.c

[kaspar030]

There is another thread flags test thread_flags_xtimer which seem somewhat redundant now. Could it be removed? @haukepetersen

Hm, that test was showing a bug in thread_flags, IIRC the thread flag was only triggered the first time. @haukepetersen what do you think, remove the test?

[jnohlgard]

Can we modify this test to catch that kind of error as well?

[kaspar030]

Can we modify this test to catch that kind of error as well?

I think the best way would be to do proper unit tests. How about we merge this and go from there?

There are some follow-ups anyway:

• xtimer_set_timeout_flag64()
• make xtimer_sleep* use them
• make xtimer_msg_timeout use the thread flag
• ...

[jnohlgard]

I'm fine with some follow ups