[Gluster-infra] Requesting separate labels in Gerrit for better testing results

[Kaushal M]

On Mon, Jan 18, 2016 at 1:03 PM, Raghavendra Talur <rtalur at redhat.com> wrote:
>
>
> On Fri, Jan 15, 2016 at 4:22 PM, Niels de Vos <ndevos at redhat.com> wrote:
>>
>> On Thu, Jan 14, 2016 at 10:26:46PM +0530, Kaushal M wrote:
>> > I'd pushed the config to a new branch instead of updating the
>> > `refs/meta/config` branch. I've corrected this now.
>> >
>> > The 3 new labels are,
>> > - Smoke
>> > - CentOS-regression
>> > - NetBSD-regression
>> >
>> > The new labels are active now. Changes cannot be merged without all of
>> > them being +1. Only the bot accounts (Gluster Build System and NetBSD
>> > Build System) can set them.
>
>
> Thanks Kaushal !
>
>>
>>
>> It seems that Verified is also a label that is required. Because this is
>> now the label for manual testing by reviewers/qa, I do not think it
>> should be a requirement anymore.
>>
>> Could the labels that are needed for merging be setup like this?
>>
>>   Code-Review=+2 && (Verified=+1 || (Smoke=+1 && CentOS-regression=+1 &&
>> NetBSD-regression=+1))
>
>
> I would prefer not having Verified=+1 here. A dev should not be allowed to
> override the restrictions.

I've made the Verified flag a `NoBlock` flag. No changes are
merge-able only with (Code-Review+2 && Smoke+1 && CentOS-regression+1
&& NetBSD-regression+1).

>
>>
>>
>> I managed to get http://review.gluster.org/13208 merged now, please
>> check if the added tags in the commit message are ok, or need to get
>> modified.
>>
>> Thanks,
>> Niels
>>
>>
>> >
>> > On Thu, Jan 14, 2016 at 9:22 PM, Kaushal M <kshlmster at gmail.com> wrote:
>> > > On Thu, Jan 14, 2016 at 5:12 PM, Niels de Vos <ndevos at redhat.com>
>> > > wrote:
>> > >> On Thu, Jan 14, 2016 at 03:46:02PM +0530, Kaushal M wrote:
>> > >>> On Thu, Jan 14, 2016 at 2:43 PM, Niels de Vos <ndevos at redhat.com>
>> > >>> wrote:
>> > >>> > On Thu, Jan 14, 2016 at 11:51:15AM +0530, Raghavendra Talur wrote:
>> > >>> >> On Tue, Jan 12, 2016 at 7:59 PM, Atin Mukherjee
>> > >>> >> <atin.mukherjee83 at gmail.com>
>> > >>> >> wrote:
>> > >>> >>
>> > >>> >> > -Atin
>> > >>> >> > Sent from one plus one
>> > >>> >> > On Jan 12, 2016 7:41 PM, "Niels de Vos" <ndevos at redhat.com>
>> > >>> >> > wrote:
>> > >>> >> > >
>> > >>> >> > > On Tue, Jan 12, 2016 at 07:21:37PM +0530, Raghavendra Talur
>> > >>> >> > > wrote:
>> > >>> >> > > > We have now changed the gerrit-jenkins workflow as follows:
>> > >>> >> > > >
>> > >>> >> > > > 1. Developer works on a new feature/bug fix and tests it
>> > >>> >> > > > locally(run
>> > >>> >> > > > run-tests.sh completely).
>> > >>> >> > > > 2. Developer sends the patch to gerrit using rfc.sh.
>> > >>> >> > > >
>> > >>> >> > > > +++Note that no regression runs have started automatically
>> > >>> >> > > > for this
>> > >>> >> > patch
>> > >>> >> > > > at this point.+++
>> > >>> >> > > >
>> > >>> >> > > > 3. Developer marks the patch as +1 verified on gerrit as a
>> > >>> >> > > > promise of
>> > >>> >> > > > having tested the patch completely. For cases where patches
>> > >>> >> > > > don't have
>> > >>> >> > a +1
>> > >>> >> > > > verified from the developer, maintainer has the following
>> > >>> >> > > > options
>> > >>> >> > > > a. just do the code-review and award a +2 code review.
>> > >>> >> > > > b. pull the patch locally and test completely and award a
>> > >>> >> > > > +1 verified.
>> > >>> >> > > > Both the above actions would result in triggering of
>> > >>> >> > > > regression runs
>> > >>> >> > for
>> > >>> >> > > > the patch.
>> > >>> >> > >
>> > >>> >> > > Would it not help if anyone giving +1 code-review starts the
>> > >>> >> > > regression
>> > >>> >> > > tests too? When developers ask me to review, I prefer to see
>> > >>> >> > > reviews
>> > >>> >> > > done by others first, and any regression failures should have
>> > >>> >> > > been fixed
>> > >>> >> > > by the time I look at the change.
>> > >>> >> > When this idea was originated (long back) I was in favour of
>> > >>> >> > having
>> > >>> >> > regression triggered on a +1, however verified flag set by the
>> > >>> >> > developer
>> > >>> >> > would still trigger the regression. Being a maintainer I would
>> > >>> >> > always
>> > >>> >> > prefer to look at a patch when its verified  flag is +1 which
>> > >>> >> > means the
>> > >>> >> > regression result would also be available.
>> > >>> >> >
>> > >>> >>
>> > >>> >>
>> > >>> >> Niels requested in IRC that it is good have a mechanism of
>> > >>> >> getting all
>> > >>> >> patches that have already passed all regressions before starting
>> > >>> >> review.
>> > >>> >> Here is what I found
>> > >>> >> a. You can use the search string
>> > >>> >> status:open label:Verified+1,user=build AND
>> > >>> >> label:Verified+1,user=nb7build
>> > >>> >> b. You can bookmark this link and it will take you directly to
>> > >>> >> the page
>> > >>> >> with list of such patches.
>> > >>> >>
>> > >>> >>
>> > >>> >> http://review.gluster.org/#/q/status:open+label:Verified%252B1%252Cuser%253Dbuild+AND+label:Verified%252B1%252Cuser%253Dnb7build
>> > >>> >
>> > >>> > Hmm, copy/pasting this URL does not work for me, I get an error:
>> > >>> >
>> > >>> >     Code Review - Error
>> > >>> >     line 1:26 no viable alternative at character '%'
>> > >>> >     [Continue]
>> > >>> >
>> > >>> >
>> > >>> > Kaushal, could you add the following labels to gerrit, so that we
>> > >>> > can
>> > >>> > update the Jenkins jobs and they can start setting their own
>> > >>> > labels?
>> > >>> >
>> > >>> >
>> > >>> > http://review.gluster.org/Documentation/config-labels.html#label_custom
>> > >>> >
>> > >>> > - Smoke: misc smoke testing, compile, bug check, posix, ..
>> > >>> > - NetBSD: NetBSD-7 regression
>> > >>> > - Linux: Linux regression on CentOS-6
>> > >>>
>> > >>> I added these labels to the gluster projects' project.config, but
>> > >>> they
>> > >>> don't seem to be showing up. I'll check once more when I get back
>> > >>> home.
>> > >>
>> > >> Might need a restart/reload of Gerrit? It seems required for the main
>> > >> gerrit.config file too:
>> > >>
>> > >>
>> > >> http://review.gluster.org/Documentation/config-gerrit.html#_file_code_etc_gerrit_config_code
>> > >
>> > > I was using Chromium and did a restart. Both hadn't helped. I'll try
>> > > again.
>> > >>
>> > >> Niels
>
>