Playing devil's advocate for a bit - are there more non-checkin-needed
backouts? That is, people who, err, feel it is unnecessary to push to
try to land something with checkin-needed, and therefore then land it
themselves and burn the tree? :-)
And also: has the throughput in checkin-needed landings gone down, ie
are we trading this off against making it a teeny bit "harder" for
people to contribute? (not intending a value judgment there - it is
likely still the right thing to do, but it would be good to know)
~ Gijs
On 30/05/2014 16:17, Ryan VanderMeulen wrote:
Just as a quick follow-up to this - we're already seeing much lower
checkin-needed backout rates since this change went into affect, so thank you
all for your help!
-Ryan
----- Original Message -----
From: "Ryan VanderMeulen" <rvandermeu...@mozilla.com>
To: "dev-b2g" <dev-...@lists.mozilla.org>, "dev.platform"
<dev-platform@lists.mozilla.org>, dev-g...@lists.mozilla.org
Cc: "Sheriffs" <sheri...@mozilla.com>
Sent: Friday, May 16, 2014 4:54:29 PM
Subject: Update on sheriff-assisted checkin-needed bugs
As many of you are aware, the sheriff team has been assisting with landing
checkin-needed bugs for some time now. However, we've also had to deal with the
fallout of a higher than average bustage frequency from them. As much as we
enjoy shooting ourselves in the foot, our team has decided that we needed to
tweak our process a bit to avoid tree closures and wasted time and energy.
Therefore, our team has decided that we will now require that a link to a
recent Try run be provided when requesting checkin before we will land the
patch. To be clear, this *ONLY* affects checkin-needed bugs where we're
assisting with the landing. We have no desire to police what other developers
do before pushing. As has always been the case, developers are expected to
ensure that their patches have received adequate testing prior to pushing
whether they are receiving our assistance or not.
Our team is also not going to dictate which specific builds/tests are required.
We're not experts in your code and we'll defer to your judgment as to what
counts as sufficient testing. As mentioned earlier today in another post, if in
doubt, we do have a set of general best practices for Try that can be used as a
guide [1]. We just want to ensure that patches have at least received some
baseline level of testing before being pushed to production. We've been testing
the water with this policy for the past couple weeks and have already seen a
reduction in the number of backouts needed.
For those of you mentoring bugs for new contributors, please also keep this in
mind in order to keep patches from being held up in landing. And consider
vouching for Level 1 commit access to further empower those contributors!
Thanks!
-Ryan
[1] https://wiki.mozilla.org/Sheriffing/How:To:Recommended_Try_Practices
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform
