Open Stack

On Mon, Dec 7, 2015 at 3:32 AM Roman Prykhodchenko <me at romcheg.me> wrote:

> Maciej, thanks for bringing this topic up for the discussion!
>
> I absolutely agree with the idea that at this point we have a problem with
> patch size. Patches that are too big usually have two major issues: they
> either don’t get reviewed for a very long time or get random +1s from
> people who don’t even bother reading all the changes there.
>
> Splitting a patch when it’s already published may be pretty problematic,
> so we should think about granularity of every feature on the design stage.
> Only in that way we will be able to submit small patches w/o making
> compromises on test coverage, stability or any other important aspect of
> the code quality.
>
> Folks proposed two different solutions here:
>   1. Set a CI job that puts a -1 to every patch that’s bigger than a
> certain limit.
>   2. Make reviewers responsible for -1ing a patch.
>
> Most of the patches that are bigger than 500 LOC can be split. However,
> there is a lot of corner cases we have to carry about, if we are going to
> put a -1 to all of them automatically. In my opinion we should not do that.
> Instead, we can send warning emails to core teams if this kind of patch set
> is submitted. Then a core team may easily analyze whether this patch can be
> split and propose a meaningful way to do that.
>
> Huge patches cause a different story — they are incredibly hard for making
> a good review, merging or reverting them can lead to a major regression in
> a lot of places, they cause conflicts for a lot of other patches and so on.
> A huge patch ALWAYS means exclusively one of the following:
>   1. It can be split
>   2. The design of the feature is completely wrong.
>

> We need to set a treshold for a huge patch and block all patches that
> exceed it by putting -2 scores.
>
> Summary:
>
> - Let’s set two thresholds — one for to big and one for huge patches.
> - Let’s create a job that will warn an appropriate core team if a too big
> patch is submitted and block huge patches.
>

On it's own, LOC is invalid as the metric for determining if a patch it too
large (its a good indicator that it might be too large but it does require
more datapoints). It can be easily obscured just with moving code around
and tests. While lare reviews can often be split This can, and should be
directly up to the reviewers to determine if it's necessary to split a
patch. At the same time it's not a unilateral decision on the reviewer's
part if it needs to be split. Like the code review in general, it needs to
be worked out between the quorum of the submitter and the reviewers. This
is the entire point of peer review and creating an auto-mated response to
something that is the responsibility of the review process seem silly. At
the same time, I'm not opposed to a code quality guideline that points out
that if your review is over X large its probably too big, and should be
split. Pointing out that this is an area of contention will help the
reviewer better prepare

For example our NOOP test yamls are huge [1] is 1291 lines. Simply adding
one set of data could easily become 4000+ lines (for example neutron vlan
l3 ha with ceph and ceilometer are 4604 lines.

[1]
https://github.com/openstack/fuel-library/blob/master/tests/noop/astute.yaml/neut_vlan.ceph.controller-ephemeral-ceph.yaml


> - romcheg
>
>
>
> 7 груд. 2015 р. о 11:23 Stanislaw Bogatkin <sbogatkin at mirantis.com>
> написав(ла):
>
>
> > What if you're not sure how the improved code should look like, but
> > you definitely sure that it shouldn't look like proposed one? :)
>
> I believe you should ask other people if you are right, as set '-1' to
> code that you
> cannot improve is not the best option, so
>
>
> > If you are not sure how the improved code would look like, just let it
> go!
> is right
>
>
> Also I think that set some strict boundaries, like 400 LOC per patch is
> wrong. For example, if you
> introduce new test fixture for fuel-library, it usually about 800+ LOC and
> you can't split it
> out, so we should either move such fixtures out of code or make an
> exeption for such type of code.
>
> On Mon, Dec 7, 2015 at 1:03 PM, Igor Kalnitsky <ikalnitsky at mirantis.com>
> wrote:
>
>> Hey Andrii,
>>
>> I'm totally agree with you about contribution guides, except one thing
>> - we need and should force users to split huge patches into set of
>> small ones. Mostly because it will simplify support and review of
>> patches. Previously we ignored this rule pretty often, and get a lot
>> of troubles due to buggy code.
>>
>> Also, see some comments below.
>>
>> > So, when an author of a patch gets -1 with the statement «split this
>> > code», I believe it is not constructive. At least you should roughly
>> > describe how you see it, how the patch could be split, you should be
>> > helpful to the author of a patch.
>>
>> No one is suggesting to set -1 without leaving a comment how the patch
>> could be divided.
>>
>>
>> > If you are not sure how the improved code would look like, just let it
>> go!
>>
>> What if you're not sure how the improved code should look like, but
>> you definitely sure that it shouldn't look like proposed one? :)
>>
>> I'd say it's a good reason to set -1 and start discussion. Not only
>> with author, but with other folks, since everyone in community is
>> responsible of code quality of the project.
>>
>> - I
>>
>> On Mon, Dec 7, 2015 at 3:28 AM, Andrey Tykhonov <atykhonov at mirantis.com>
>> wrote:
>> > I believe this is against the code review guidelines.
>> >
>> > «Comments must be meaningful and should help an author to change the
>> > code the right way.» [1]
>> >
>> > If you get a comment that says «split this change into the smaller
>> > commit» I'm sorry, but it doesn't help at all.
>> >
>> > «Leave constructive comments
>> >
>> > Not everyone in the community is a native English speaker, so make
>> > sure your remarks are meaningful and helpful for the patch author to
>> > change his code, but also polite and respectful.
>> >
>> > The review is not really about the score. It's all about the
>> > comments. When you are reviewing code, always make sure that your
>> > comments are useful and helpful to the author of the patch. Try to
>> > avoid leaving comments just to show that you reviewed something if
>> > they don't really add anything meaningful» [2]
>> >
>> > So, when an author of a patch gets -1 with the statement «split this
>> > code», I believe it is not constructive. At least you should roughly
>> > describe how you see it, how the patch could be split, you should be
>> > helpful to the author of a patch. So, first of all, you need to review
>> > the patch! :)
>> >
>> > I want to emphasize this: «The review is not really about the
>> > score. It's all about the comments.»
>> >
>> > «In almost all cases, a negative review should be accompanied by clear
>> > instructions for the submitter how they might fix the patch.» [4]
>> >
>> > I believe that the statement "split this change into the smaller
>> > commit" is too generic, it is mostly the same as the "this patch needs
>> > further work". It doesn't bring any additional instructions how
>> > exactly a patch could be fixed.
>> >
>> > Please also take a loot at the following conversation from mailing
>> > list: [3].
>> >
>> > «It's not so easy to guess what you mean, and in case of a clumsy
>> > piece of code, not even that certain that better code can be used
>> > instead. So always provide an example of what you would rather want to
>> > see. So instead of pointing to indentation rules, just show properly
>> > indented code. Instead of talking about grammar or spelling, just type
>> > the corrected comment or docstring. Finally, instead of saying "use
>> > list comprehension here" or "don't use has_key", just type your
>> > proposal of how the code should look like. Then we have something
>> > concrete to talk about. Of course, you can also say why you think this
>> > is better, but an example is very important. If you are not sure how
>> > the improved code would look like, just let it go, chances are it
>> > would look even worse.» [3]
>> >
>> > So, please, bring something concrete to talk about. If you are not
>> > sure how the improved code would look like, just let it go!
>> >
>> > «The simplest way to talk about code is to just show the code. When you
>> > want the author to fix something, rewrite it in a different way,
>> > format the code differently, etc. -- it's best to just write in the
>> > comment how you want that code to look like. It's much faster than
>> > having the author guess what you meant in your descriptions, and also
>> > lets us learn much faster by seeing examples.» [2]
>> >
>> >
>> > [1]
>> >
>> https://docs.google.com/document/d/1tyKhHQRQqTEW6tS7_LCajEpzqn55f-f5nDmtzIeJ2uY/edit
>> > [2] https://wiki.openstack.org/wiki/CodeReviewGuidelines
>> > [3]
>> >
>> http://www.mail-archive.com/openstack-dev@lists.openstack.org/msg07831.html
>> > [4] http://docs.openstack.org/infra/manual/developers.html#peer-review
>> >
>> >
>> > Best regards,
>> > Andrey Tykhonov (tkhno)
>> >
>> >
>> >
>> __________________________________________________________________________
>> > OpenStack Development Mailing List (not for usage questions)
>> > Unsubscribe:
>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> <http://OpenStack-dev-request@lists.openstack.org/?subject:unsubscribe>
>> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>> >
>>
>> __________________________________________________________________________
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe:
>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> <http://OpenStack-dev-request@lists.openstack.org/?subject:unsubscribe>
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
-- 

--

Andrew Woodward

Mirantis

Fuel Community Ambassador

Ceph Community
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20151207/b320496f/attachment.html>