runtime: 'failed to trigger concurrent GC' in TestSelectStackAdjust

[bcmills]

On the linux-ppc64le-power9osu builder (https://build.golang.org/log/585e960d0f000d42817bf1b361c8ecbf51d1294d):

--- FAIL: TestSelectStackAdjust (0.02s)
    chan_test.go:723: failed to trigger concurrent GC
FAIL
FAIL	runtime	35.499s

A similar failure consistently occurs on the openbsd-arm builder, but I'm not confident in its results because it still seems to be misconfigured (#25498 (comment)).

See previously #14852.

CC @mknyszek @aclements

[laboger]

There's a discussion in #34391 that might be related to this failure. Some intermittent failures on ppc64 and ppc64le appear to be happening due to a problem with signals, and also affects linux/arm.

[bcmills]

linux-ppc64le-power9osu again: https://build.golang.org/log/25161e684426bdf04053d57fd93551dc53ee6234

[bcmills]

Also a recurring problem on certain BSD builders, it seems: there are instances on netbsd and dragonfly too.

2019-11-08T18:55:44-b2b0992/netbsd-arm-bsiegert
2019-11-07T14:51:06-543c6d2/openbsd-arm
2019-11-07T05:59:33-a8f57f4/openbsd-arm
2019-11-06T09:09:59-0c5d545/openbsd-arm
2019-11-06T09:09:21-0ea7440/openbsd-arm
2019-11-05T15:40:02-8550a58/openbsd-arm
2019-11-05T00:26:30-b3bd7ab/openbsd-arm
2019-11-04T17:19:36-1da575a/openbsd-arm
2019-11-04T05:27:25-fb29e22/openbsd-arm
2019-11-02T19:53:12-bdb5e9d/openbsd-arm
2019-11-02T05:52:33-8de0bb7/openbsd-arm
2019-11-02T05:36:43-971ec87/openbsd-arm
2019-11-02T00:36:59-48a0b97/openbsd-arm
2019-11-01T22:07:56-dc39be8/dragonfly-amd64
2019-11-01T21:38:07-4ff4582/plan9-386-0intro
2019-11-01T03:40:20-4a43a50/openbsd-arm
2019-11-01T01:51:26-efd395f/openbsd-arm
2019-10-31T15:14:45-02a5502/openbsd-arm
2019-10-31T01:09:58-5a210b5/openbsd-arm
2019-10-31T01:09:34-43d25a6/openbsd-arm
2019-10-31T01:09:14-851b1f4/openbsd-arm
2019-10-31T01:04:59-e90e6e7/openbsd-arm
2019-10-30T16:37:41-cf6e6ab/openbsd-arm
2019-10-30T08:17:29-f4e32ae/openbsd-arm
2019-10-30T03:48:03-9e094ea/openbsd-arm
2019-10-29T16:45:39-1a058cd/openbsd-arm

[odeke-em]

Just seen in https://storage.googleapis.com/go-build-log/026635dc/android-amd64-emu_6806dba8.log from https://go-review.googlesource.com/c/go/+/180919/5#message-4c33c31b5e6b04b4e02ec706d238879b733712f7

[bcmills]

2019-11-10T20:36:44-47bc240/netbsd-arm-bsiegert

[ianlancetaylor]

2019-11-27T02:18:51-95c9a65/netbsd-arm-bsiegert
2019-11-22T19:17:28-8324aca/netbsd-arm-bsiegert
2019-11-15T15:12:42-dd8bbc7/netbsd-arm-bsiegert
2019-11-12T01:09:23-e2cac31/netbsd-arm-bsiegert
2019-11-12T01:09:06-5b17b65/linux-arm
2019-11-11T22:13:42-0c703b3/netbsd-arm-bsiegert
2019-11-11T17:53:51-70be481/linux-ppc64le-power9osu

[ianlancetaylor]

Almost all the failures seem to happen when running the runtime tests with -test.cpu=1,2,4.

[gopherbot]

Change https://golang.org/cl/210217 mentions this issue: runtime: add Gosched to TestSelectStackAdjust loop