tests/xtimer_now32_overflow: simplify

[kaspar030]

Contribution description

Vastly simplifies and speeds up the regression test added in #13850.

Testing procedure

Test should succeed on master. To see that it would fail, revert 212fe15.

Issues/PRs references

#13850

[kaspar030]

ping @miri64 @JulianHolzwarth

[fjmolinas]

Testing:

• pr:

RIOT native interrupts/signals initialized.
LED_RED_OFF
LED_GREEN_ON
RIOT native board initialized.
RIOT native hardware initialization complete.

Help: Press s to start test, r to print it is ready
READY
s
START
main(): This is RIOT! (Version: 2020.07-devel-5-gf0ca79-pr-13853)
[SUCCESS]

• reverted case:

RIOT native interrupts/signals initialized.
LED_RED_OFF
LED_GREEN_ON
RIOT native board initialized.
RIOT native hardware initialization complete.

Help: Press s to start test, r to print it is ready
READY
s
START
main(): This is RIOT! (Version: 2020.07-devel-6-g88a790-pr-13853)
[FAILED]

[fjmolinas]

The test looks good, and is a simpler test for the regression pointed out in the test README. This test doesn't check the timing of the offsets, but I do not see why that is needed for the regression test intended here, it seems more like a test for xtimer_set_msg64(), tests for the 64bit case seem to generally be missing, but I'm under the impression that the only differ from the 32bit version on the overflow that is tested here:

void _xtimer_set_msg(xtimer_t *timer, uint32_t offset, msg_t *msg, kernel_pid_t target_pid)
{
    _setup_msg(timer, msg, target_pid);
    _xtimer_set64(timer, offset, 0);
}

void _xtimer_set_msg64(xtimer_t *timer, uint64_t offset, msg_t *msg, kernel_pid_t target_pid)
{
    _setup_msg(timer, msg, target_pid);
    _xtimer_set64(timer, offset, offset >> 32);
}

Anyway since I'm not that acquainted with xtimer code, maybe @JulianHolzwarth can chime in and mention if the new test falls short somehow/

[kaspar030]

@miri64 @JulianHolzwarth could you take a closer look?

The test that's currently in master is actually buggy:

RIOT/tests/xtimer_now32_overflow/tests/01-run.py

Lines 34 to 39 in 5de5cea

	child.expect(r"now=(\d+)")
	offset = int(child.match.group(1))
	# get largest possible timeout for expects below
	timeout = _get_largest_timeout_difference(timers.values()) / (10**6) + 1
	for i in range(timers_numof):
	child.expect(r"#(\d):now=(\d+)", timeout=timeout)

(The \d+ are not terminated, leading to partial matches).

[fjmolinas]

Please squash @kaspar030 !

[kaspar030]

• squashed

[fjmolinas]

The change seems OK to me and the regression is still tested in a simpler way.

[leandrolanzieri]

Backport provided in #13905

[fjmolinas]

I have some asserts with this PR depending on the platform:

BOARD=iotlab-m3 CFLAGS=-DDEBUG_ASSERT_VERBOSE make -C tests/xtimer_now32_overflow/ flash term -j3

START
main(): This is RIOT! (Version: 2020.07-devel-5-g4ac5e-HEAD)
[SUCCESS]
sys/xtimer/xtimer_core.c:278 => 0x8000d49
*** RIOT kernel panic:
FAILED ASSERTION.

*** halted.

iotlab-m3, openbmote-b, usb-kw41z. No issues with samr21-pxro, nrf52dk, nucleo-l14

Seems I should have tested this PR on more platforms, any ideas?