test: double test timeouts on arm machines

[bnoordhuis]

The ARM buildbots are notoriously slow. Update the test runner to
double the per-test time limit when it's running on one of them.

R=@iojs/build

CI: https://jenkins-iojs.nodesource.com/view/iojs/job/iojs+any-pr+multi/448/

[Fishrock123]

cc @silverwind

[brendanashworth]

Build 444 on armv8 looks like it had the same amount of timeouts as this build's, fwiw.

[silverwind]

Is this the 30s test time limit? On which boards is it an issue?

[bnoordhuis]

Is this the 30s test time limit?

Yes. Specifically, the limit that makes tests fail with --- TIMEOUT ---.

On which boards is it an issue?

The ARMv8 buildbot is the main offender and I noticed that's the one buildbot that seems unaffected by this change. I speculate that platform.machine() returns something different on that one. Will investigate.

[bnoordhuis]

I speculate that platform.machine() returns something different

It does: 'aarch64'. New run with fix-up: https://jenkins-iojs.nodesource.com/view/iojs/job/iojs+any-pr+multi/455/

[bnoordhuis]

The latest run suggests that the timeouts are persistent, doubling doesn't make a difference. I'm going to close this PR.

[silverwind]

Yeah, I wondered myself if that was really needed, as ARMv7 (RPi2) is almost fast enough to pass tests without any modifications. I expect ARMv8 to pass with our regular timeouts.

[bnoordhuis]

I did some digging and I expect that libuv is to blame here, see libuv/libuv#308.

[silverwind]

@bnoordhuis ARMv6 seems to hit the 30s timeout on the following tests:

test/sequential/test-pipe
test/parallel/test-stringbytes-external

That'd be two tests I can strike off my ARMv6 list if this gets merged. Would you reconsider this PR for these?

[bnoordhuis]

Oh, sure. The branch still exists. If you LGTM it, I'll land it.

[silverwind]

LGTM. Just ran the ARMv6 tests with this patch, and no more --- TIMEOUT --- 👍