environs/jujutest: fix test

environs/jujutest: fix test

LiveTests.TestStartStop did not work when called
with an already-bootstrapped environment.
Coincidentally, LiveTests.TestDestroy was
always called after the only other test to bootstrap,
so running all the tests worked.

https://code.launchpad.net/~rogpeppe/juju-core/175-ec2-fix-live-tests/+merge/136940

(do not edit description out of merge proposal)

[rog, 2012-11-29 14:24:27]

Please take a look.

[TheMue, 2012-11-29 14:29:33]

LGTM

[gz, 2012-11-29 14:56:27]

Tests pass (well, more so than before) for me now. Some nitpicking on specifics
but otherwise lgtm.

https://codereview.appspot.com/6850121/diff/2001/environs/jujutest/livetests.go
File environs/jujutest/livetests.go (right):

https://codereview.appspot.com/6850121/diff/2001/environs/jujutest/livetests....
environs/jujutest/livetests.go:119: // new instance.
This doesn't actually assert that insts is a set without duplicates. The
instance appearing in the list multiple times would pass the test. There's no
set builtin to just do difference and assert on that?

Also, asserting len+1 seems slightly fragile still, as this test doesn't control
the lifetime of other instances.

(Expecting to be able to cleanly run tests while doing other work on the same
account would be silly, but this code is already quite complex anyway, but all
these checks should really care about is the non-existence before and singual
existence after).

[rog, 2012-11-29 15:32:35]

*** Submitted:

environs/jujutest: fix test

LiveTests.TestStartStop did not work when called
with an already-bootstrapped environment.
Coincidentally, LiveTests.TestDestroy was
always called after the only other test to bootstrap,
so running all the tests worked.

R=TheMue, gz
CC=
https://codereview.appspot.com/6850121

https://codereview.appspot.com/6850121/diff/2001/environs/jujutest/livetests.go
File environs/jujutest/livetests.go (right):

https://codereview.appspot.com/6850121/diff/2001/environs/jujutest/livetests....
environs/jujutest/livetests.go:119: // new instance.
On 2012/11/29 14:56:27, gz wrote:
> This doesn't actually assert that insts is a set without duplicates. The
> instance appearing in the list multiple times would pass the test. There's no
> set builtin to just do difference and assert on that?

you're right that it doesn't test that the set has no dupes.
i'm not too concerned about that (we don't actually document that that's
necessarily true), so i'm just deleting the comment for the time being.

> Also, asserting len+1 seems slightly fragile still, as this test doesn't
control
> the lifetime of other instances.

the environment is unique for the tests, so this should be fine.