SIDE-NOTE: no benches should be run in MQ
SIDE-NOTE: all pushes to try should be done using mach try perf
.
Ways of triggering bencher runs:
publish results using --branch=main
on bencher (and github checks or maybe job summary; need aditional work in bencher to support it using adtitonal flag)
publish results using --branch=try
on bencher and should not be parent of old try run (try runs are independent), base should be last common commit from main
report as github check
because not token is available here we cannot send results to bencher but we should still generate results like in servo/servo case
same as try push to servo/servo, but bencher should also generate comment on PR
(while user could also create try-labels personal fork, but this is not supported workflow so we will not cover it)