I'm accepting these; notes below.

> Sat Dec  9 14:39:16 PST 2006  David Roundy <[EMAIL PROTECTED]>
>   * fix pending bug that broke several_commands.sh.
> 
> Sun Dec 10 13:18:46 PST 2006  David Roundy <[EMAIL PROTECTED]>
>   * resolve conflict in white space.

Ok, I don't entirely understand this, but I'll trust that David actually
understands the pending stuff really well, in any case much better than
me and I do see that tests do now pass where they didn't before.

Notes
-----
1) Remove.lhs - I smell a refactor.  Maybe?

2) What exactly does moving add_to_pending before the moving/replacing/etc
   accomplish?  Is this about correctness, robustness or something else?

3) add_to_pending now uses get_unrecorded, which includes read_pending that
   we used to invoke.  I'm guess this is the bit that fixes the bugs, but is
   there any chance that this introduces a new category of bugs?  Is there
   such a thing as unrecorded changes that shouldn't go on pending being
   dumped there by accident?

And now a general remark
------------------------
One thing that seems to be missing in our comments is an explanation how
pending really works.  Maybe there is none, because there is nothing to
explain.  I would suspect that other developers would appreciate some sort of
hand-holding on the subtleties behind pending vs.  regular old unrecord
changes, even a basic list of gotchas and sanity checks we should make when
mucking around in there.

I claim also that with David having recently cleaned up all this stuff and
pushed it into the Repository module, now would be a good time to work on said
documentation, while it's still fresh in our collective head.

> Sun Dec 10 13:35:36 PST 2006  David Roundy <[EMAIL PROTECTED]>
>   * change Maybe Patch to Hopefully Patch.
> 
> Sun Dec 10 15:16:23 PST 2006  David Roundy <[EMAIL PROTECTED]>
>   * don't use HopefullyPrivate outside of Hopefully.
> 
> Sun Dec 10 15:44:53 PST 2006  David Roundy <[EMAIL PROTECTED]>
>   * fix bug in haskell_policy check for HopefullyPrivate.

This version of the Hopefully stuff answers some of my previous
questions, so I've copied over the ones I still find relevant and
added some more

Notes 2
-------
4) should hopefully be expressed in terms of conscientiously?

5) There is one disadvantage to this new code, and that is that we lose
   the source code line numbers that tell use where the fromJust/
   hopefully error was triggered.  I would argue that this is worth
   getting back, because hopefully still gets called in many different
   places, so it might not be entirely clear why/when patch stuff
   becomes Unavailable.

   Would it be possible for hopefully to have some of the same
   impossible.h line number magic?  That would cover the top of the
   chain, just like fromJust did.  What would also be nice is if we
   could cover the bottom of the chain by putting the same magic in our
   calls to the Unavailable constructors.

6) Perhaps we should have the policy checker "enforce" that Unavailable
   should never be passed the empty string.   We now have much more
   capacity for error tracking, so it seems like we should adopt a
   discipline of actually using it.

7) Maybe refactor hopefullyNoParseError?
   I see it in two places; maybe there'll be more?

-- 
Eric Kow                     http://www.loria.fr/~kow
PGP Key ID: 08AC04F9         Merci de corriger mon français.

Attachment: pgpDWTLAISxqQ.pgp
Description: PGP signature

_______________________________________________
darcs-devel mailing list
[email protected]
http://www.abridgegame.org/cgi-bin/mailman/listinfo/darcs-devel

Reply via email to