note: you need to use gcc 7.2 on master and this PR to verify the fix

error messages, as follows:

usleep:

/RIOT/examples/posix_sockets/udp.c:118:9: error: implicit declaration of function 'usleep'; did you mean 'sleep'? [-Werror=implicit-function-declaration]
         usleep(delay);
         ^~~~~~
         sleep
cc1: all warnings being treated as errors

and

/RIOT/sys/posix/pthread/pthread.c:239:13: error: this statement may fall through [-Werror=implicit-fallthrough=]
             thread_sleep();
             ^~~~~~~~~~~~~~
/RIOT/sys/posix/pthread/pthread.c:241:9: note: here
         case (PTS_ZOMBIE):

Do we provide a posix usleep function implementation? Do we need it for anything or is it just bloat? Do we provide a nanosleep function implementation that can be used instead?

At least for msp430-libc and avr-libc we do (the implementation is in sys/posix/time/posix_time.c. It exists to be able to run without explicitly calling xtimer_usleep() (but it just wraps this function) so POSIX libraries can be run on RIOT. I did not know about the deprecation, but if it is, we should indeed find an alternative (but this is unrelated to this PR IMHO).

I did not know about the deprecation,

(at least the Linux manpages do not mention that).

first and foremost this RP is about getting things to work with GCC 7.x. I didn't want to start kind of a philosophical discussion or question POSIX compatibility and so on.

getting things to work with GCC 7.x

POSIX stuff here, there are more fixes needed to actually compile all of RIOT with GCC 7.x

usleep:

Why not fix to use xtimer_usleep?

this is about POSIX and portability (I would say), i.e., using standard headers without need for RIOT specific ones. Look at the examples, they don't see xtimer but just use standard header functions. Fixing this by using xtimer, renders this tests redundant.

renders this tests redundant.

Maybe not redundant, but one wouldn't be able to run it on e.g. Linux anymore, which should be possible at the moment with this example.

ping @gebart?

ping, this bugfix could also go into the release (I'll provide a backport). It would be nice to have GCC 7 support, or not?

Sorry for the delay