[darcs-users] using screened for adventure

Eric Kow kowey at darcs.net
Fri Sep 24 10:12:38 UTC 2010

On Thu, Sep 23, 2010 at 15:24:51 +0100, Ganesh Sittampalam wrote:
> Petr would advertise in advance on the list any substantial changes,
> so that they can be generally discussed and objected to if
> necessary. We should probably maintain a wiki page with the current
> list.

We can use http://wiki.darcs.net/Development/Adventure for that
(or maybe a page in the Review/ hierarchy too)

> If any tests need to be broken temporarily, then they will get
> renamed to failing-adventure-xxx, so that we can check that they are
> all fixed later on.
> To keep the screened backlog down, we'll also need to be able to
> review the work in pieces. It therefore needs to be clear what sets
> of adventure patches can be pulled together. I'd suggest that the
> general rule should be that any set of patches that passes the tests
> is ok to pull, and that Petr should clearly flag if this is not the
> case.

Sounds good: special marking for tests that adventure regresses, and OK
to pull any patches with no test failures.

Does this mean no failing-adventure-xxx either? I imagine it does, but
perhaps this would lead to too much a backlog? We could also cross that
bridge when we come to it.

> The upside of doing this is that both Petr and other developers will
> be able to minimise conflicts and thus wasted work.
> The downside is that since we can't just leave things in screened
> indefinitely without stalling all review work, we will effectively
> be committing to accepting the adventure work into darcs in some
> form.

If it helps, I think we all want the library cleanup to happen, and we
have made effort to engage darcs-users on sacrificing old-fashioned.  So
unless I'm mistaken, we're all on board about the work being a good idea
in principle.

Recalling previous discussion [1], what we're after is to resolve the
two questions about how to adventure, Q1:logistics and Q2:correctness.
This thread started from a strong objection about the work being done
without some prior agreement on evidence of correctness.  In order to
make progress, I think we need to focus on one question at a time (*).

So let's settle Q1:logistics now.

Adapting our consensus on the screened branch [1] to

- Ganesh proposed it in the first place
- Petr seems willing to try it out (unless I misunderstand the IRC discussion)
- Eric is up for an experiment 
- Florent
- Reinier probably yes (since his willingness to try the screened branch was
  as an alternative to his preferred adventure branch approach)

And to avoid misunderstandings that we tripped over:

 1. This is consensus, not voting.
    The important part is the ideas and discussion.

 2. Consensus means finding an arrangement that all major participants
    can live with. This also means we all accept/understand the
    motivations for our choices even if we aren't too happy on them.

 3. We're building consensus on the broad principles rather than
    trying to settle the finer details up front (eg. must be possible to
    review in trickles, rather than which wiki page the review goes on)

 4. Discussions are open to whole community (we need broad set of
    perspectives).  Push comes to shove, it's the people who are the
    most implicated (ie. doing the hacking and the reviews) whose
    concerns/opinions weigh the most.

Just trying to learn from the adventure experience...


PS: Personally, I think we can tackle Q2:correctness only when we start
    to merge large chunks of adventure into mainline.

[1] http://lists.osuosl.org/pipermail/darcs-users/2010-August/025098.html
[2] http://lists.osuosl.org/pipermail/darcs-users/2010-September/025298.html

Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
For a faster response, try +44 (0)1273 64 2905 or
xmpp:kowey at jabber.fr (Jabber or Google Talk only)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.osuosl.org/pipermail/darcs-users/attachments/20100924/45f4d0b8/attachment.pgp>

More information about the darcs-users mailing list