Summary is incorrect when there are fewer percentiles than expected
nicktindall opened this issue · comments
When a probe has fewer percentiles than expected for the number of benchmark iterations (perhaps it is not invoked once for every iteration) it renders like:
-------------------------------- SUMMARY (B) us ----------------------------------------------------
Percentile run1 run2 run3 % Variation
50.0: 0.10 0.10 0.10 0.00
90.0: 0.10 0.10 0.10 0.00
99.0: 0.10 0.10 0.10 0.00
99.7: 0.10 0.10 0.10 0.00
worst: Infinity Infinity Infinity -Infinity
----------------------------------------------------------------------------------------------------
When in fact it should render like
-------------------------------- SUMMARY (B) us ----------------------------------------------------
Percentile run1 run2 run3 % Variation
50.0: 0.10 0.10 0.10 0.00
90.0: 0.10 0.10 0.10 0.00
99.0: 0.10 0.10 0.10 0.00
worst: 0.10 0.10 0.10 0.00
----------------------------------------------------------------------------------------------------
i.e. it is the 99.7th percentile missing due to small sample size, not the "worst" measurement
Released in JLBH-1.23ea0, BOM-2.22.129