Fix flaky tests

[shazow]

• [x] --- FAIL: TestHostNameCollision (0.01s) host_test.go:74: First client failed to get 'foo' name: " \* foo joined. (Connected: 1)" https://github.com/shazow/ssh-chat/commit/c57a5acca7096c056414dfb837e65a2bb750daef
• [ ] --- FAIL: TestIgnore (0.00s) room_test.go:148: should have messages
• [x] --- FAIL: TestSetExpiring (0.00s) set_test.go:78: b-prefix incorrect: ["barbar" "2017-05-11 11:54:27.161591982 -0400 EDT"]
• [ ] --- FAIL: TestHostGetPrompt (0.00s) host_test.go:44: Got: "[\x1b[38;05;172mfoo\x1b[0m] "; Expected: "[\x1b[38;05;88mfoo\x1b[0m] "

[tacerus]

Hello!

Is there an update on this?

I am working on a RPM package for ssh-chat, and encounter FAIL: TestHostGetPrompt on the architectures armv7l and i586 (which are not super important, as most people are likely using it on x86_64, but it would still be nice to widen the target audience - and I don't feel like disabling tests for those two architectures, if they work with x86_64 just fine). If you are interested in checking it out, you can find the build logs in https://build.opensuse.org/package/show/home:crameleon/ssh-chat (so far only SLES and openSUSE are relevant, still working on adapting the spec to Fedora and CentOS, so those are expected to fail).

Cheers!

[shazow]

@tacerus Hiya, thanks for following up! I appreciate you maintaining this distro package downstream. :)

We did make progress on x86 as you noticed, go test -test.timeout 5s -count 100 ./... works reliably.

I'm not certain why arm7l and i586 would still have issues. One suggestion: Could try running it with/without -race, I recall that have a side effect on whether some tests were flakey before or not.

For what it's worth, it's not a super important test, the flake is simply messages coming in out of order. If you'd like to disable it for those builds, I think it's reasonable.

[tacerus]

Thanks for the input and headsup!

[ 47s] go test: -race is only supported on linux/amd64, linux/ppc64le, linux/arm64, freebsd/amd64, netbsd/amd64, darwin/amd64, darwin/arm64, and windows/amd64

I found that the tests pass if I run them only against /cmd/ssh-chat and will hence go ahead and implement a clause that will do that only for the problematic architectures.

[tacerus]

I got excited too early, after some testing this workaround showed only reliable on i586, on armv7l it would still fail. I hence implemented the cmd/ssh-chat workaround for i586 and skipped tests completely on armv7l, adding a note referring this issue.