cross-browser-tests-runner / cross-browser-tests-runner / 170

93%

push

travis-ci

Fixes #25

There is a race condition between test.stop() and native runner
monitor.

Our stop implementation ensures that after a DELETE call, the status
returned for a given job does say 'stopped', which happens with retries.
The native runner monior is sneaking through this period to see twice
that status is stopped but native runner has not marked it as stopped
yet.

The solution used here is that as soon as the test results arrive and
we find a valid non-stopped test to stop, a state 'nativeRunnerStopping'
is added to the test. When native runner monitor sees this state, it
just won't process this test any more to figure out if it has been not
reported yet etc. So for actually failing tests that do not respond back,
the earlier logic (now encapsulated under the if condition checking for
this state) comes into play, and would determine correctly, free of the
race conditions, that the test did not stop correctly.

630 of 747 branches covered (84.34%)

Branch coverage included in aggregate %.

1963 of 2199 relevant lines covered (89.27%)

1798.97 hits per line