Re: PR backlog (was: [RFC] Deprecation and removal of the drm2 driver)

On Thu, May 31, 2018 at 10:25:10PM -0500, Matthew D. Fuller wrote:
> because the incentives are rigged.  A bad outside contribution brought
> into ports more often yields "hey, you should have noticed" to the
> committer and more opprobrium back to the submitter.

I believe you're missing two important points:

 - there is feedback to the ports committers that goes on behind the
   scenes (e.g. not on public mailing lists);

 - everyone who uses FreeBSD needs src to work.  Not everyone who uses
   FreeBSD needs even a large fraction of ports to work.

Let me reframe this debate.  To me, the correct comparison is:

  "compare commits to src"

     vs. 

  "compare commits to key ports pieces such as Mk/, perl, apache24, &c"

Commits to the latter are thoroughly tested* in external staging areas
and/or "exp-runs" done by portmgr, _before_ they hit the tree.

For src committers that aren't aware: since adopting the exp-runs, there
have been far, far, fewer large-scale regressions of the ports tree.
Check the monthly portmgr reports to see how much work is going into
this -- and that doesn't count projects like gnome and kde that do their
own external precommit work.

Also, IMHO talking about whether this process is, or should be, automated
misses the point.  The distinguishing feature is the buy-in by the people
who are making changes to the tree to have done sufficient testing _first_.
Without that buy-in, the tools are irrelvant.

Next, let me assure you that anyone who breaks those key pieces of ports
hears about it _immediately_.

tl:dr; at least for the key pieces, FreeBSD ports has moved away from the
"throw it at the wall to see if it will stick" paradigm.  That's the part
of the codebase that ought to be the comparison point.

mcl

* almost always.  They are supposed to be, in any case.  Humans are involved.