ISystemOperations.AuthenticateAsync now returns AuthResponse.

[Emdot]

Fixes #493

[jterry75]

LGTM

[jterry75]

@galvesribeiro - PTAL

[galvesribeiro]

LGTM. Thanks!

[galvesribeiro]

Just one thing before we merge. Why is the tests failing? I'm out of my desk now and I can't check. Can you please fix it and push the update? Thanks!

[Emdot]

It's failing MonitorEventsAsync_Succeeds. Running locally, it appears to be an intermittent failure.

[dgvives]

@galvesribeiro @Emdot errors on StreamUtils.MonitorStreamAsync methods are not related to this changeset but to previously existing changes.

Previous ContainerOperations.GetContainerLogsAsync task cancellation by disposing stream had some pitfalls, as it was disposing a stream while a streamReader was consuming from it.

Should be working fine now, pushed to #487 .
@Emdot feel free to review/fix/improve/criticise this PR. There is integrity within your criteria

[Emdot]

@galvesribeiro, can you retrigger the build?

[galvesribeiro]

Just did. Same error as it seems... ☹️

[Emdot]

I see that this is a known bug in the test case. "// On CI/CD Pipeline exception is thrown, not always"

I think there's more of a problem here than I'm willing to fix:

1. It looks like the event streams are delayed. Some of the events from CreateImageAsync don't arrive until after MonitorEventsAsync starts, so they show up in _onJSONMessageCalled. Some of the events from TagImageAsync don't arrive until after the assertion.
2. I can't be sure that that's what's happening, since stdio might be buffered too. Adding some Console.Out.Flush and Console.Error.Flush calls between commands can help check for that.
3. This probably isn't what's going wrong in this case, but it's a bug anyway: Using MonitorEventsAsync this way has a race condition. Unless you await MonitorEventsAsync's Task, you can't guarantee that the monitoring has started, so you can miss messages. The fix would is to replace MonitorEventsAsync with something like StartEventMonitorAsync that returns a subscription.

Would you accept a PR to disable the broken test case, until someone fixes the underlying problems?

[Emdot]

I wrote that up as #502.

[dgvives]

@Emdot move the call to include a call to remove the image after tagging it, that should be enough in this case

Try to include some delay before cancelling task so that I/O operation finishes
Task.Delay(TimeSpan.FromSeconds(1))

[Emdot]

I'm unwilling to fudge the tests to make them pass. If you fix the problematic code (as noted above) or disable the test, then this PR can proceed.