tests/kernel: fifo_timeout: Remove wake-up order checking

[rgundi]

There is no guarantee of wake-up order when multiple threads
are woken up on the same tick. Hence, modified the tests
accordingly.

Fixes #8159.

Signed-off-by: Rajavardhan Gundi [email protected]

[spoorthik]

@rgundi: Is there any reason for introducing boot delay?

[rgundi]

Yes. The issue #8159 cannot be reproduced without this delay. It's a very timing dependent issue.

[nashif]

you should remove it though, we do not set this for tests by default.

[pizi-nordic]

I think that this PR throws out the baby with the bathwater, as it removes order check completely.

IMHO order check should be still there, but it has to take under account presence of tick granulation. For example we can record start time of each thread and check if the timeouts are in specified bounds (not earlier than _ms_to_ticks() and not later than _ms_to_ticks()+1).

[rgundi]

I can't see a better method as the timeouts get affected by system loading. If the interrupts get locked for whatever reason, the timeouts may go out of bounds (later than _ms_to_ticks()+1). So, all we can do is to ensure the timeouts do not happen earlier than _ms_to_ticks().

[rgundi]

@pizi-nordic .. can you please comment?

[pizi-nordic]

I agree that there is no upper limit on loaded system.
However during the test we are controlling system state (configuration, number of threads and so on).
So we can assume upper bound. What about the following checks:

1. Make sure that thread does not waited shorter than expected,
2. Make sure that thread was woken up not later than expected_time + X (X = 1-2 ticks should be sufficient).

[rgundi]

@pizi-nordic Agree. Have updated the test. Please check.

[codecov-io]

Codecov Report

Merging #10047 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master   #10047   +/-   ##
=======================================
  Coverage   53.05%   53.05%           
=======================================
  Files         214      214           
  Lines       26203    26203           
  Branches     5646     5646           
=======================================
  Hits        13903    13903           
  Misses      10028    10028           
  Partials     2272     2272

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4dfd311...1733655. Read the comment docs.

[pizi-nordic]

IMHO this check is not the one we want.
Whit this code, kernel might wake-up thread earlier than expected and the test will not notice that.

[rgundi]

That is already taken care in "is_timeout_in_range()" function. That function checks for the lower limit while the new code above checks for the upper limit.

[pizi-nordic]

Ok

[nashif]

you should remove it though, we do not set this for tests by default.

[rgundi]

@nashif : Done.