greplogs --dashboard -md -l -e 'FAIL: TestDialCancel .*(?:\n .*)*unexpected successful connection' --since=2021-01-01

2022-04-11T02:55:52-a6f6932/linux-amd64-wsl
2022-04-11T02:55:52-a6f6932/linux-mips64le-mengzhuo
2022-04-11T01:24:31-b6fb3af/linux-mips64le-mengzhuo
2022-04-11T01:24:31-b6fb3af/linux-riscv64-unmatched
2022-04-10T16:04:13-a10f158/linux-amd64-wsl
2022-04-10T16:04:13-a10f158/linux-mips64le-mengzhuo
2022-04-10T16:04:13-a10f158/linux-riscv64-unmatched
2022-04-10T15:52:08-8cd6aac/linux-mips64le-mengzhuo
2022-04-10T15:52:08-8cd6aac/linux-riscv64-unmatched
2022-04-07T23:33:12-5a90270/linux-riscv64-unmatched
2022-04-07T23:06:24-c451a02/linux-riscv64-unmatched
2022-04-07T20:56:02-8d581f5/linux-riscv64-unmatched
2022-04-07T20:39:32-6f6942e/linux-riscv64-unmatched
2022-04-07T19:16:46-79619c3/linux-riscv64-unmatched
2022-04-07T17:19:55-c0bbeb0/linux-riscv64-unmatched
2022-04-07T14:42:13-063f403/linux-riscv64-unmatched
2022-04-06T20:46:47-81ae993/linux-mips64le-mengzhuo
2022-04-06T03:13:34-5bb2628/linux-mips64le-mengzhuo
2022-04-06T01:11:21-6c17529/linux-mips64le-mengzhuo
2022-04-05T23:32:55-2de2f6d/linux-mips64le-mengzhuo
2022-04-05T22:09:11-dbb52cc/linux-mips64le-mengzhuo
2022-04-05T21:45:19-69756b3/linux-mips64le-mengzhuo
2022-04-05T18:58:52-5a6a830/linux-mips64le-mengzhuo
2022-04-05T18:01:26-9e16cc1/linux-mips64le-mengzhuo
2022-04-05T17:52:29-81431c7/linux-mips64le-mengzhuo

Curiously, these failures are all on builders owned by @mengzhuo.

At first I thought this might be a bug in the test — the addresses in question are reserved for “benchmarking”, but the test does not ensure that no benchmark is currently running. However, looking at the list of ports in RFC 2544 and RFC 5180, none of the prescribed benchmarks actually seem to use the addresses in the 198.18.0.* range. A successful dial on such an address seems to indicate an out-of-spec network configuration.