I think there's a bit of a bug in 0746a31
Before
The aggregate metric geomean seems to be the sum of the sub-subtest results. Eg, if I look at http://localhost:8081/?developerMode=true&suites=Perf-Dashboard&startAutomatically=true#details from a locally served copy on 0746a31's parent I get
Perf-Dashboard/Render 8.02 ± 1.03 (12.9%) ms
Perf-Dashboard/SelectingPoints 12.78 ± 0.35 (2.7%) ms
Perf-Dashboard/SelectingRange 8.42 ± 1.04 (12.3%) ms
and
Geomean 29.22 ± 2.08 (7.1%) ms
I get
Perf-Dashboard/Render 7.50 ± 0.37 (4.9%) ms
Perf-Dashboard/SelectingPoints 13.98 ± 1.57 (11.2%) ms
Perf-Dashboard/SelectingRange 7.91 ± 0.10 (1.2%) ms
and
Geomean 7.91 ± 0.10 (1.2%) ms
Which is exactly the result of SelectingRange -- it seems to overwrite the total.
I think there's a bit of a bug in 0746a31
Before
The aggregate metric geomean seems to be the sum of the sub-subtest results. Eg, if I look at http://localhost:8081/?developerMode=true&suites=Perf-Dashboard&startAutomatically=true#details from a locally served copy on 0746a31's parent I get
and
After 0746a31
I get
and
Which is exactly the result of SelectingRange -- it seems to overwrite the total.