testing: panic in AllocsPerRun during parallel test

[rsc]

I propose that testing.AllocsPerRun panic if called during a parallel test,
because it cannot give accurate results in that context. Tests that use
the result in that context will be flaky.

For example, flake #70327 was caused by a test that called t.Parallel
and then used AllocsPerRun. The other tests running in parallel
had allocations that were observed by the AllocsPerRun, resulting
in inconsistent results and a flaky failure.

I've implemented this change in CL 630137.

[gabyhelp]

Related Code Changes

• testing: panic in AllocsPerRun if parallel tests are running

_{(Emoji vote if this was helpful or unhelpful; more detailed feedback welcome in this discussion.)}

[ianlancetaylor]

Personally I think this is a bug fix. I don't think it needs to go through the proposal process. Since it can't work correctly, it's hard for me to believe that anybody depends on the current behavior.

[rsc]

I was on the fence about proposal or not. If the committee says not, that's fine with me! 😃

[rsc]

Discussed at proposal review and decided on quick accept.

[rsc]

No change in consensus, so accepted. 🎉
This issue now tracks the work of implementing the proposal.
— rsc for the proposal review group

The proposal is for testing.AllocsPerRun to panic if called during a parallel test.