writing Was: [PATCH weston 1/5] gl-renderer: Rename gl_renderer_create to gl_renderer_display_create

Bryce Harrington bryce at osg.samsung.com
Mon May 23 21:39:46 UTC 2016

On Wed, May 18, 2016 at 06:33:15PM -0400, Yong Bakos wrote:
> Hi,
> I'm partly to blame for the bikeshedding on writing, yet believe me,
> I'm always asking "is this worth it?", prioritizing corrections/clarity
> over mere style judgements, and am definitely sensitive to
> everyone's cognitive bandwidth. My goal is high quality documentation
> for Wayland, and I am open to suggestions on how to best deliver that.
> (Maybe we can chat about this over IRC rather than further hijack this thread.)
> On May 18, 2016, at 10:31 AM, Daniel Stone <daniel at fooishbar.org> wrote:
> > 
> > Hi,
> > 
> > On 18 May 2016 at 15:25, Derek Foreman <derekf at osg.samsung.com> wrote:
> >> On 18/05/16 08:41 AM, Mike Blumenkrantz wrote:
> >>> In fairness, we'd likely be less short on review bandwidth if the
> >>> majority of that bandwidth was not in use to make/revise trivial
> >>> criticisms such as whitespace usage and comment grammar which are
> >>> guaranteed to be cleaned up in future patches; this leads to burnout on
> >>> both the code-writing side as well as the reviewing side since everyone
> >>> has become hyper attuned to the insignificant and non-functional
> >>> minutiae of patches rather than focusing more on expediting the
> >>> technical development of the protocol.
> I agree with the crux of Mike's point. Yet, should trivial correctness
> be revised before merging or should post-merge corrections clutter the
> blame history? (Rhetorical question.)
> >> Fair points, though I'm not certain "will certainly get fixed up later"
> >> is a given.  Certainly indenting and basic style is a mechanical problem
> >> that could be tested pre-commit hooks, and there should be no reason to
> >> bike shed that on the list at all.
> Yes!
> >> Grammar probably needs more serious consideration for protocol doc than
> >> elsewhere due to its potential impact on compositor implementors - but
> >> ever there probably not to the degree we're seeing lately.
> >> 
> >> Follow up commits that do nothing but change style and grammar can make
> >> "git blame" less useful (when trying to figure out who would best review
> >> a piece of code - not just "arrrrgh who wrote this stupid bug") and
> >> provide churn for very little benefit, imho.
> > 
> > Yeah, I agree. I get that the bikeshedding can be annoying; I do (for
> > that reason, if no other) like tagging commits as 'RFC' or similar,
> > which is effectively, 'please just check out the technical concept and
> > don't worry about memory leaks or spelling mistakes right now'. But
> > given that it's pretty trivial to fix up, and you're likely to have to
> > rebase _anyway_, I don't see the harm in doing one round of review for
> > clarity.
> > 
> > Generally, there's no need to send out a subsequent revision round
> > just because someone has noted some typos - send it again if you need
> > a resend anyway to get people to pick it up after a rebase, or if
> > there have been notable changes, but you shouldn't be arriving at v17
> > just because you have difficulty spelling.
> Perhaps non-technical suggestions should be sent only directly to the
> author, for consideration in incorporating in the event of subsequent
> revisions, and we not send these to the list? This gives the author
> the benefit of the information, and does not clutter the list with
> non-technical reviews.

No, I don't think we need special rules for such things.  Just send them
to the list like anything else.  If you've done a spelling/grammar check
then posting it publically means the next reviewer knows they don't need
to comment on the same stuff and can focus on something else.

In general, we don't need to bikeshed the bikeshedding.  ;-)

> > Similarly, 'no, I disagree' is a reasonable response to someone
> > bikeshedding your exact choice of variables or naming. Review is meant
> > to be a discussion, not something you just have to unilaterally
> > acqiuesce to.
> > 
> >> While we're drifting just slightly off topic here, I'm also concerned
> >> about the basic usage of some of our tags:
> >> 
> >> Reviewed-by:  indicates rigorous technical review *AND* a firm
> >> conviction that the feature is important and should be included.
> >> 
> >> Acked-by: Indicates a firm conviction that the feature is important and
> >> should be included, but no rigorous review has taken place.
> >> 
> >> Signed-off-by: Indicates an assumption of responsibility for the code.
> The rigor is important. We should add this to:
> https://wayland.freedesktop.org/reviewing.html.

That's all fine but frankly I don't see this as a serious problem in
practice.  The real problem we have is more that patches aren't getting
*enough* R-b's.

When landing patches I do take into account who gave the R-b, and am
aware that each of us has areas we're stronger in than others.  I'm
guessing here that pq takes this into account as well.  I'd rather have
excessive R-b's than none.

A good technique I've seen (and tried to emulate myself) is to state
what aspects of the patch you've reviewed.  E.g. "Looks good from a
technical perspective, no idea on overall appropriateness."  But, from a
maintainer perspective it's usually clear what aspects the reviewer was
looking at, based on their comments (or lack thereof).


More information about the wayland-devel mailing list