couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dave Cottlehuber <...@jsonified.com>
Subject Re: git commit: Increase test timeout. Take spinning disks into account.
Date Tue, 26 Mar 2013 18:34:54 GMT
On 26 March 2013 18:46, Jan Lehnardt <jan@apache.org> wrote:
>
> On Mar 26, 2013, at 18:43 , Jan Lehnardt <jan@apache.org> wrote:
>
>>
>> On Mar 26, 2013, at 18:39 , Wendall Cada <wendallc@apache.org> wrote:
>>
>>> On 03/26/2013 06:45 AM, Jan Lehnardt wrote:
>>>> I’m not sure why this didn’t go into 1.3.x, but it’s probably my
>>>> fault, sorry about that.
>>>>
>>>> This should go into 1.3.x and we should recast the 1.3.0 vote as rc.3.
>>>>
>>>> Could we get a consensus from the people where this failed on what the
>>>> timeout should be? For one person on IRC it was 2000, but we can make
>>>> this bigger if required. Please send in the minimal value for your system
>>>> where this doesn’t fail for you.
>>> I agree with this. I have consistent failures with the javascript cli_runner.
I've been playing around with it to get everything passing. The test failures are all timing
based from what I can tell. The setup/teardown happens too fast between several of the tests
and it leaves a hanging couchdb process. The tests eventually time out and fail. Adding more
time is just masking the problem. I still strongly feel these issues are more closely related
to having couchdb stop/start properly. As a stop gap measure, adding more time to specific
tests to ensure that the tear down of the previous step happens is just fine, but this should
get a proper fix at some point in the near future.
>>
>> This is unrelated to the JS tests, but test/etap/250-*.t
>
> And yes, there is a lot we need to fix about this tests generally, but this is about
this one particular test on 1.3.x. Let’s focus on that :)
>
> If I remember correctly the correlation is the Erlang version (<=R15 fails and >=R15B02
succeeds). So in this case it isn’t really masking an actual issue, but working around version
discrepancies.
>
> Cheers
> Jan

I agree with redoing the release based on general feedback, but I
won't have a chance to vote/test before Tuesday next week.

For me on R15B03-1 OSX with spinning disk, I got repeated failures,
5000 works for me. This was on a heavily loaded system (for a mac). My
other box is a veritable ferrari so I didn't see it prior.

A+
Dave

Mime
View raw message