TestRunCommand fails on windows

[Martin Packman]

UniterSuite.TestRunCommand fails consistently on windows in CI.

<http://reports.vapour.ws/releases/issue/5576e8f1749a560d68f74bdb>

Note the issue also has old failures, see bug 1463408 for that.

<http://reports.vapour.ws/releases/3051/job/run-unit-tests-win2012-amd64/attempt/1193>

FAIL: util_windows_test.go:92: UniterSuite.TestRunCommand
...
test 4: run commands: async using rpc client
...
step 2:

uniter_test.verifyFile{filename:"c:\\...\\run.output", content:"juju run u/0\r\n"}
[LOG] 0:13.542 INFO juju.worker.uniter unit "u/0" shutting down: tomb: dying
[LOG] 0:13.542 DEBUG juju.worker.uniter juju-run listener stopping
[LOG] 0:13.542 DEBUG juju.worker.uniter RunCommands: {Commands:Set-Content c:\...\run.output "juju run $env:JUJU_UNIT_NAME" RelationId:-1 RemoteUnitName: ForceRemoteUnit:false}
...
util_windows_test.go:160:
    s.runUniterTests(c, []uniterTest{
...
    })
util_test.go:133:
    c.Assert(err, jc.ErrorIsNil)
... value syscall.Errno = 0x490 ("Element not found.")

Looks to be the last windows test regression from the maltese-falcon branch landing.

[Casey Marshall (cmars)]

I'm at eod and not getting anywhere. I did reproduce this on hyperv and found that the run listener seems to be restarting rather than executing the command:

----------------------------------------------------------------------
FAIL: util_windows_test.go:92: UniterSuite.TestRunCommand

...

test 4: run commands: async using rpc client

step 0:

...

uniter_test.waitHooks{"install", "leader-elected", "config-changed", "start"}
waiting for hooks: []string{"install", "leader-elected", "config-changed", "start"}
ctx.hooksCompleted: []string{"install", "leader-elected", "config-changed", "start"}
ctx.hooks: []string{"install", "leader-elected", "config-changed", "start"}
uniter_test.verifyCharm{revision:0, attemptedRevision:0, checkFiles:filetesting.Entries(nil)}
step 1:

uniter_test.asyncRunCommands{"Set-Content C:\\Users\\ADMINI~1\\AppData\\Local\\Temp\\2\\check-5577006791947779410\\8\\run.output \"juju run $env:JUJU_U
NIT_NAME\""}
step 2:

uniter_test.verifyFile{filename:"C:\\Users\\ADMINI~1\\AppData\\Local\\Temp\\2\\check-5577006791947779410\\8\\run.output", content:"juju run u/0\r\n"}
[LOG] 0:09.432 INFO juju.worker.uniter unit "u/0" shutting down: tomb: dying
[LOG] 0:09.432 ERROR juju.worker.uniter juju-run listener got: [{invalid argument}]
[LOG] 0:09.432 DEBUG juju.worker.uniter juju-run listener stopping

[Casey Marshall (cmars)]

http://reviews.vapour.ws/r/2668/

[Curtis Hovey (sinzui)]

Per the duplicate bug that just affects 1.24

As seen in
    http://reports.vapour.ws/releases/3092/job/run-unit-tests-win2012-amd64/attempt/1265
and the 2 subsequent attempt, UniterSuite.TestRunCommand fails.

util_test.go:1399:
    c.Assert(err, jc.ErrorIsNil)
... value rpc.ServerError = "r.runner.RunCommands: tomb: dying"

The only commit since the last pass is
    https://github.com/juju/juju/commit/832f636ce846d3dea63d4a3367da327de5bc2498

This issue is actually a duplicate of bug 1495591. The last commit merged some badness.

[Canonical Juju QA Bot (juju-qa-bot)]

Juju-CI verified that this issue is Fix Released in juju-core 1.24:
    http://reports.vapour.ws/releases/3096