Code review comment for lp:~clint-fewbar/pyjuju/add-testrunner

Revision history for this message
Kapil Thangavelu (hazmat) wrote :

eat my data should be unesc. the data store for tests is a effectively a
ram store.

On Thu, Oct 25, 2012 at 6:28 PM, Jim Baker <email address hidden> wrote:

> Some minor issues I observed in trying this useful utility out:
> 1) eatmydata and iwatch need to be installed in advance; this issue is
> more pronounced for eatmydata since it will fail in the test phase. I'm
> not certain what the best approach is for this juju package. Perhaps
> just test installation of both tools before running, as opposed to
> making this a requirement of juju itself.
> 2) At least on my desktop, the test is always run twice, because iwatch
> seems to pick up the file change twice.

> 3) Users need to remember they have this utility running and not attempt
> to run more tests simultaneously, otherwise both test runs will now
> fail.
perhaps as part of a parellizations effort, or minimally the iwatch test
addr,the get zk test address could return chroot address per test.

i'm curious though what error were encountered as a result of running tests
in this manner, afaics each run should have a separate zk anyways (test zk
enclosure/context manager).


> --
> You are subscribed to branch lp:juju.

« Back to merge proposal