Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 30 Mar 2018 09:02:57 -0700 (PDT)
From:      "Rodney W. Grimes" <freebsd@pdx.rh.CN85.dnsmgr.net>
To:        Mark Linimon <linimon@lonesome.com>
Cc:        rgrimes@freebsd.org, Ian Lepore <ian@freebsd.org>, src-committers@freebsd.org, svn-src-stable@freebsd.org, svn-src-all@freebsd.org, svn-src-stable-11@freebsd.org, Marcelo Araujo <araujo@freebsd.org>
Subject:   Re: bug triaging (was: Re: svn commit: r331728 - in stable/11/etc: . rc.d)
Message-ID:  <201803301602.w2UG2vBr068679@pdx.rh.CN85.dnsmgr.net>
In-Reply-To: <20180330030443.GA15201@lonesome.com>

next in thread | previous in thread | raw e-mail | index | archive | help
> This is addressed to developers in general, not just rgrimes, but he
> made the comments, so ...

:-)

> 
> On Thu, Mar 29, 2018 at 09:33:44AM -0700, Rodney W. Grimes wrote:
> > It seems that the Phabricator review system is somewhat dysfunctional
> > in that actual review is only happening in some cases.  Some people
> > have even stated they flat out hate it.
> 
> I will have to state as someone who has spent a great deal of time on
> classifying/triaging bug reports in this project, that the attitude
> that some developers have that "I am not going to use tool xyz" is both
> disheartening and demotivating.  I find it difficult to remember when I
> triage: who it is that will or will not use which tool?

Agreed that this is difficult if not impossible to "remeber",
and imho there should be no reason to have to remeber it.

However we are also faced with the problem of volunteerism,
which means we can not just shove tools down peoples throats.

> Here: the plain facts are that our clearance rate for Phabriactor reviews,
> for both src and doc, are far better than for Bugzilla.  For ports, the
> opposite is true.  These are just facts.

There is probably a statistical relation between externally opened
(non committer) and internally opened (committer) that would answer
the above.  Most, but not all, Phabricator reviews are opened by
committers, where as most Bugzilla reports are opened by others.

The ports bugzillas are usually opened and handed directly to
a port maintainer.  That does not happen for src bugs, which
are generally assigned to a group which ends up with mail to
a list and hopefully someone is interested in fixing or helping
with that bug.  There is also a periodic summary/reminder of these
list assigned bugs, that was off for months (6?), probably causing
some accumulation.

Improvement:  We need to get better and the "take bug" task,
The default action of bugzilla when you take a bug is to remove
the current assigned value, this leads to the bugs dropping
out of view of the mailling list reminder nontices.  It is
documented in the procedures that you should "keep the current
assignee" when taking a bug.   This is not happening.   When
I notice this I was emailing the taker about it, that got some
fixing done, but now I just go in and fix it when I see it
happen, which also lead to some more peopole doing the correct
thing in this area.   What would REALLY fix it would be if
we could either have a "group asignee" and a "personal assignee",
or perhaps fix the "take bug" action to just move the group
address to the CC: list.

> 
> By and large IMHO phab is a plus.  (Disclaimer: I personally hate the
> web interface, it makes me want to pull out my few remaining hairs.)
> But I do not see it going away.  Nor, do I see bugzilla going away.
> Some people like the workflow of the one, some like the other.

Phabricator is a neat tool, I like it and hate it.  Its not the
tool that I tried to take issue with, but the lack of communications
that occurs with the tool as it stands today.  I should probably
learn to use arc, and then I might hate it less.

> > The problem is that most people are not notified that a review
> > of a change is even in process until the commit lands, this is
> > not a functional communications system.
> 
> But many developers also ignore bug reports coming through Bugzilla,
> echoed on the mailing lists.  What is your constructive suggestion
> here?  Do we make subscribing to Phab reviews per src bit mandatory?
> I would support it but imagine I would get a lot of pushback.

Improvement:  Add a "submitter selected" item that says this
Phabricator review belongs to one of the same mailling list
clasifications that you already usually triage the bugs into,
on phabricator SUBMISSION, a one or two line summary of the
review needs to go to a file, at the end of the day these
files get mailled to the appropriate lists.   That would
only create 1 mail per day per clasification.  Or it could
be sent as a single mail, which would be more, but as imp@ says
it is not a huge amount of traffic.  Phabricator has hearald
rules, and I have been trying for a month to get my custom
#bhyvce rule that was requested by the bhyve group promoted
to a global rule.  I have sent mail to phab admin, no response.
I manually add the #bhvye group when I get notice.  There
are a *few* groups, but by and far this just does not create
enough interest in reviews of code, I suspect because people
do not know there is a review up.

> > Requring us all to go sign up like imp@ did to receive all
> > submitted reviews, imho, is also a non functional situation.
> 
> So what is a constructive suggestion?

See above.

> 
> (Fair warning, folks: I won't consider "get rid of Phabricator" or
> "get rid of Bugzilla" as constructive.)

Nor would I.  But we can make improvements.  And I feel the place
to start with improvements is in the area of information flow.
I still assert, most interested parties are not aware of a review
until the commit lands, and that is a communictions problem.

-- 
Rod Grimes                                                 rgrimes@freebsd.org



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201803301602.w2UG2vBr068679>