[LLVMdev] [cfe-dev] Commit message policy?

[Sean Silva]

On Fri, Mar 6, 2015 at 4:15 PM, Mehdi Amini <mehdi.amini at apple.com> wrote:

>
> > On Mar 6, 2015, at 4:08 PM, Robinson, Paul <
> Paul_Robinson at playstation.sony.com> wrote:
> >
> >> -----Original Message-----
> >> From: cfe-dev-bounces at cs.uiuc.edu [mailto:cfe-dev-bounces at cs.uiuc.edu]
> On
> >> Behalf Of Mehdi Amini
> >> Sent: Friday, March 06, 2015 1:49 PM
> >> To: Davide Italiano
> >> Cc: Clang Dev; LLVM Dev
> >> Subject: Re: [cfe-dev] [LLVMdev] Commit message policy?
> >>
> >>
> >>> On Mar 6, 2015, at 1:36 PM, Davide Italiano <davide at freebsd.org>
> wrote:
> >>>
> >>> On Fri, Mar 6, 2015 at 1:12 PM, Renato Golin <renato.golin at linaro.org>
> >> wrote:
> >>>> On 6 March 2015 at 20:59, Reid Kleckner <rnk at google.com> wrote:
> >>>>> I think the only guideline we should have is that the first line
> >> should be
> >>>>> written as though it is an email subject, because it gets used for
> >> that. If
> >>>>> you write a long first line, then you get a long subject, and it
> looks
> >>>>> silly. If people want to embarrass themselves with strangely
> formatted
> >>>>> email, they it's on them. We don't need a specific hard or soft
> >> number.
> >>>>
> >>>> Not many people care about the email subject already, that's why they
> >>>> keep using ridiculously long first lines.
> >>>>
> >>>> IMO, "suggesting" to write short first lines is the same as not doing
> >>>> anything. Either we add a cap (say, 80 chars), or we don't do
> >>>> anything.
> >>>>
> >>>> Chandler's other suggestion, tough, is interesting: to write up a bit
> >>>> about what a *good* message would be, so the people that were really
> >>>> interested, could do it "right" (tm).
> >>>>
> >>>
> >>> Another guideline I would like to propose for commit messages is that
> >>> of attaching to the commit a link to the code review, if any.
> >>
> >> I believe it is documented here:
> >> http://llvm.org/docs/Phabricator.html#committing-a-change
> >>
> >> Mehdi
> >
> > That would be the norm for people doing reviews in Phabricator.
> > I think the suggestion is to do something similar for non-Phab reviews?
>
> Oh, like a link to the email thread, for instance:
> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20110314/117996.html
> ?
>

I really wish our mail web ui was better.... personally, I only use it as a
last resort. I'll turn first to my inbox, and then to gmane or marc.info or
any of the other mail archives.

As such, the most useful thing is for people to reply to review threads
with the revision at which the code landed. That way, I can find the commit
from the review thread (just the named revisions), or the review thread for
the commit (searching for the revision number turns up the review thread,
and also any relevant post-commit review or reference to it). The community
is actually pretty consistent about providing this and it's an awesome
system IMO.

I would vastly prefer we try to encourage this existing convention more
consistently, rather than trying to enforce putting links in commit
messages. The existing convention provides a bidirectional link.

-- Sean Silva


>
> As a Phabricator user, I haven’t thought about that, it makes sense I
> guess.
>
> Thanks for the clarification.
>
> —
> Mehdi
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150307/6e2ef21b/attachment.html>