launchpad-dev team mailing list archive

[Stuart Bishop <stuart.bishop@xxxxxxxxxxxxx>]

On Thu, Oct 7, 2010 at 9:24 PM, Jonathan Lange <jml@xxxxxxxxxxxxx> wrote:
> On Thu, Oct 7, 2010 at 11:28 AM, Jonathan Lange <jml@xxxxxxxxxxxxx> wrote:
>> On Wed, Oct 6, 2010 at 11:18 PM, Maris Fogels
>> <maris.fogels@xxxxxxxxxxxxx> wrote:
>>> On 10/06/2010 05:46 PM, Maris Fogels wrote:
>>>> On 10/06/2010 05:28 PM, Jonathan Lange wrote:
>>>>> Hello all,
>>>>>
>>>>> The following tests fail in stable:
>>>>>   $ ./bin/test -cvv devscripts.ec2test.tests.test_remote
>>>>>
>>>>> 7 tests fail. As a reminder, stable is our known-good
>>>>> all-tests-pass-guaranteed branch.  This is not supposed to be
>>>>> possible.
>>>>>
>>>>> I don't know how many more tests are failing.  At this point, I
>>>>> suggest that we can no longer consider stable or db-stable to be
>>>>> deployable code.
>>>>>
>>>>> Could someone please:
>>>>>   * find out which tests are failing and fix them
>>>>>   * find out how this happened
>>>>>
>>>>> jml
>>>>>
>>>>
>>>> I can reproduce this in devel.
>>>>
>>>> My patch fixes four of them:
>>>> https://code.edge.launchpad.net/~mars/launchpad/fix-ec2test-utf-in-devel/+merge/37760
>>>>
>>>> The errors look to be the same.  I'm working on a quick patch based on the same
>>>> code, but if that does not work I will hand the baton off to someone else.
>>>>
>>>
>>> Here is the merge proposal that fixes all of the errors seen:
>>>
>>>  https://code.edge.launchpad.net/~mars/launchpad/fix-ec2test-utf-in-devel/+merge/37799
>>>
>>> I've sent it off to PQM, but I /really/ have to go now, and can't see this
>>> branch through.  If someone could shepherd it, it would be much appreciated!
>>
>> Thanks. There were a couple of glitches, but StevenK is landing a patch now.
>>
>> However:
>>  a) we still need to confirm that stable's tests are all passing. I'm
>> running a full test run on ec2 now.
>>  b) ditto for db-stable, but I can't seem to get ec2 test working
>> because of http://paste.ubuntu.com/507903/ (paramiko error)
>>  c) we still need to figure out how this happened in the first place
>
> After doing a full run of tests against stable, I can also confirm
> some more failures:
>
> lp.codehosting.puller.tests.test_worker.TestWorkerProgressReporting.test_network
> (leaking errors)
> lib/lp/registry/stories/person/xx-person-subscriptions.txt
>
> http://paste.ubuntu.com/508043/ has the errors in detail.

I think there is code relying on database ordering rather than
requesting explicit ordering.

For instance with the first failure, the user is subscribe to a bug
with two bugtasks (Firefox the product and Firefox the ubuntu
package), and +subscriptions is displaying one of them at random.


-- 
Stuart Bishop <stuart@xxxxxxxxxxxxxxxx>
http://www.stuartbishop.net/