[Development] [Qt-creator] gerrit-speak

[Oswald Buddenhagen]

On Thu, Dec 13, 2012 at 08:50:16PM +0000, Jenssen Tim wrote:
> > (Rethoric: should this go to a more generic qt list?)
> > 
> > Hello
> >
> > (First a side question, is there a more consequent documentation about the use
> > of gerrit?)
> > 
> > Is it only me that finds the "-1" review default message in gerrit rather
> > irritating?
> > 
> > "I would prefer that you didn't submit this"
> > 
> > reads to my brain much like "go f*k yourself" without the raw words.
> >
yes, it's sad ... that we have a huge wiki/howto, but nobody cares to
RTFM, and chooses to be insulted instead. never mind all the technical
screwups resulting from failure to read/understand that text.

On Fri, Dec 14, 2012 at 08:30:50AM +0000, Koehne Kai wrote:
> There's a point though that a lot of people (including me) are using
> -1 for 'this requires more work', and -2 for 'the patch, or target
> branch is just wrong, please abandon'.
> 
this is actually problematic, because people tend to overlook the -1
when somebody else gives +2.
ideally, we'd have the following:
* -1 => "needs some work"
* -2 => "needs serious work. this will break things - don't submit under
         any circumstances."
* -3 => "this is inherently rubbish or mistargeted. please abandon."
the thing is that (at least in the gerrit version we use) it is not
possible to make the two bottom levels blocking, so -2 has a double
meaning. the ambiguity is a minor problem (we all can read the comments,
right?), but -2's stick (to cover the -3 case above), so it needs
explicit unblocking, which is often annoying.

in the opposite direction, we actually also should have three levels:
* +1 => "yeah, whatever"
* +2 => "this is good ... but i'm no approver yet (or i'm a coward) ^^"
* +3 => "approved. i take the blame."

and to top it off, it should be possible to give explicit zero scores:
"others already scored this. i'm waiting for the next revision and want
the boldness to disappear from my dashboard". these should not appear as
comments, obviously.

how good that digia will put an internal resource on gerrit development
soon ...

On Fri, Dec 14, 2012 at 08:59:25AM +0000, Bache-Wiig Jens wrote:
> The css they provide for us was most likely never touched by a
> designer and I almost suspect they did not intend for it to be used
> as-is.
> 
gerrit is built with GWT. style *that*.

> I still have a hard time finding the clone link to checkout an
> existing project and I am sure others hit this problem as well. For
> some reason clicking on "project/qtdeclarative" does not actually tell
> me how to clone it. It really should be the first thing on that page.
>
that's because gerrit is (currently) not a project hosting environment,
but a review tool (which happens to come with an own git server, but
that's just typical java NIH).
so the point is that if you are looking for the clone url on gerrit, the
surrounding infrastructure has already failed you.

On Fri, Dec 14, 2012 at 10:46:02AM +0100, André Somers wrote:
> Op 14-12-2012 9:30, Koehne Kai schreef:
> > "you shouldn't submit this" - I don't think this is ready to go in as it is.
> > "I would prefer" -  ..but at the same time this isn't a veto, so if another approver thinks differently (with reasons), please go ahead.
> >
> +1. I really don't see the offending nature of the message, and it is 
> quite consise in what it says. "This requires more work" does not imply 
> "Please don't commit this (yet)" to me.
> 
"This needs more work before submission."
but any formulation won't fix the problem that people don't bother to
RTFM.

On Fri, Dec 14, 2012 at 11:20:11AM +0000, Koehne Kai wrote:
> Just out of interest, is the topic feature actually used a lot?
>
not really.
partly because it's still so bug-ridden that nobody dares to use it. it
doesn't behave quite intuitively, either.
another reason is that any sane commit policy prescribes atomicity on a
commit level, so forced grouping of commits isn't much of an advantage.