Make timing tests compare to upper / lower bounds #2068
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description of Change(s)
When running tests in parallel, we get intermittent failures of
timing related tests due to the elapsed time being outstide of the
allowed range.
However, this could just be due to thread scheduling issues, as
an indeterminate amount of time could pass between the recorded trace
and the standard library call which was used as "ground truth".
Now, I bracket the trace records on both sides with standard library
timing calls, to provide expected upper and lower bounds, which should
be more forgiving of randomness from thread scheduling.
A follow up to changes in:
cadce6f - Avoid hard-coded comparisons for timing tests
Fixes Issue(s)
Intermittent failures of testTrace and TfStopwatch tests when running tests with many processes
I have submitted a signed Contributor License Agreement