[ovirt-devel] patch granularity, extracting patches

[Tomas Jelinek]

On Thu, Jun 29, 2017 at 6:16 PM, Greg Sheremeta <gshereme at redhat.com> wrote:

> Do we have or follow a standard on patch granularity?
>

I don't think we have or should have any strict standard about this. Every
reviewer is different and prefers different kind of patch granularity.

For example I really dislike if someone break one logical work (like
implement one small feature) into 10 smaller patches (all in the same area)
to easy review. To me reviewing 10 smaller patches is much more work than
review one bigger because I keep loosing the context and the bigger
picture. In such cases I typically end up downloading the series locally
and review in IDE. But than I loose the gerrit features like compare
patchsets.

Others, on the other hand, prefer to see tiny patches and that helps them
to review.

Ultimately, if you want to get a patch in, you need to know the culture and
the people to know how to shape it.
A rule of thumb: in VDSM break up the patches to as small as possible, in
engine, break up into logical areas of expertise (e.g. DB, core, frontend).


> When I started, I picked up on the culture that we do small, logical
> commits -- as much as possible, each commit should be focused on a specific
> purpose. I've perceived some reviewers prefer to have all orthogonal
> changes (fix a random spelling error, remove a duplicate semicolon)
> extracted to other patches for clarity.
>

>From my perspective this is a complete waste of time. It does not bring any
more clarity to the review nor to the git log. It may be aesthetically more
appealing but brings much more work while babysitting the patches and does
not bring anything of value. Also I feel it discourages people to fix
issues like typo in comment when crossing it because they will have to
create a separate patch, than either depend on it or have conflicts with
it.

I personally prefer to follow the boy scout rule. On the other hand, some
find it disturbing. That is fine, we are all humans, there is no one good
option and this is not a technical question...


> Others don't seem to mind. I feel like I always want to ask, but I feel
> bad because it's a hassle.
>


Contributing code is a way of communication with people. I don't think
asking anything is a problem...


>
> Also, when you are asked to extract something, do you have a trick to make
> it as easy as possible?
>
> Best wishes,
> Greg
>
> --
> Greg Sheremeta, MBA
> Sr. Software Engineer
> Red Hat, Inc.
> gshereme at redhat.com
>
> _______________________________________________
> Devel mailing list
> Devel at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/devel/attachments/20170630/4b307674/attachment.html>