[darcs-users] Fwd: Releasing 2.8 : UI questions

Florent Becker florent.becker at ens-lyon.org
Thu Jun 23 10:25:40 UTC 2011


Hi all,

some of the changes in 2.8 impact the UI of darcs, which we should
discuss to determine if/how we announce new and deprecated parts of
darcs' UI. We need to have these questions settled before 2.8 is
released in order to make changes to defaults in 2.10.

* First, there is the question of darcs put. I agree that having it is
important for the completeness of the command set. On the other hand,
the current implementation is really sub-par, and having users being
exposed to that as "normal darcs slugginess" sucks. We should at least
warn that it's currently advised against. I'd also like not to keep
bit-rotten parts in darcs, so I still think it should be removed from
2.10 if it has not been reimplemented in a viable way.

* Hunk-splitting in revert: the UI for this is still a bit crude, but I
don't know how to improve it. On the other hand, getting the UI into a
beta is the way to get more feedback about it and converge to a good
solution. I think we should enable this, with a warning, for 2.8.

* rollback --no-record. The consensus seems to be that this flag should
eventually become the default behaviour of rollback, but this may break
uses of rollback in script. Is 2.10 the right shade of "eventually" for
this change? Is a warning in the long-message part of recording within
rollback enough?

Florent


Le 19/06/2011 20:27, Florent Becker a écrit :
> Hi fellow darcs users and developpers,
> 
> it seems the time for 2.8 is now approaching. You can find all the
> details about it at http://wiki.darcs.net
> 
> I will release a first beta real soon now, just by branching off from
> the reviewed repository; see for this
> http://darcs.net/releases/branch-2.8. After this beta is released, there
> will be a three week period for testing ("elastic deep freeze"), with
> new betas and rcs as needed. Only bugfixes will go into these betas and
> rcs, via the branch http://darcs.net/releases/branch-2.8. During this
> period, patches that are intended for the release should be sent using:
> 'darcs send http://darcs.net/releases/branch-2.8'. Please state that the
> patch is for the release branch, especially if you use darcs send -o.
> 
> The end of the release cycle is uncertain: when new bugs are found in a
> release candidate, there will be another RC a week later. Only when bugs
> are not found in a release candidate for a week, it will be tagged and
> released as darcs 2.8.0.
> 
> This freeze is elastic: if you have code that corresponds to one of the
> unmet release goals and send it in next week, there'll be an extra week
> for integrating it, after which, we go back to the (translated) schedule.
> 
> Thus, the schedule is as follows:
> 
> - June 19, release-2.8 is branched and beta1 is uploaded.
> - On June 27th, beta2 is released
> - On July 7th, rc1 is released (aka 2.7.99)
> - Every thursday from there on, as long as bugs are found and fixed, a
> new rc is released. After that, the last rc becomes darcs2.8.0, and is
> tagged 2.8.0; That tag is backported to reviewed.
> 
> What's in 2.8
> =============
> 
> Since darcs 2.5 has been tagged, there have been 826 patches applied to
> darcs. Here is a (subjective) list of the most user-visible changes,
> report any omission:
> 
> 
> Thu Apr  7 02:04:36 CEST 2011  Owen Stephens <darcs at owenstephens.co.uk>
>   * resolve issue2052: Implicitly use unified diff, unless told not to.
> 
> Sat Apr  2 11:57:25 CEST 2011  Guillaume Hoffmann <guillaumh at gmail.com>
>   * default darcs get to use packs
> 
> Fri Apr  1 17:21:06 CEST 2011  Owen Stephens <darcs at owenstephens.co.uk>
>   * Ignore thisrepo source.
>   Do not read or write the 'thisrepo' source to the prefs/sources file.
> 
> Thu Feb 10 21:08:31 CET 2011  dixiecko at gmail.com
>   * Add inferred XML Schema for outputs.
> 
> Wed Feb  2 17:03:36 CET 2011  Guillaume Hoffmann <guillaumh at gmail.com>
>   * remove OF write support
>   Make low level pristine handling functions fail on OF repository
>   by printing an error message.
>   Adapted from work by Petr Rockai
> 
> Wed Feb  2 17:37:52 CET 2011  Guillaume Hoffmann <guillaumh at gmail.com>
>   * make OF get warning message point to wiki
> 
> Fri Jan 14 16:44:49 CET 2011  Florent Becker <florent.becker at ens-lyon.org>
>   * add a --no-record option to rollback
> 
> Sun Dec 19 22:42:19 CET 2010  Gabriel Kerneis <kerneis at pps.jussieu.fr>
>   * resolve issue1350: reliable patch transport
> 
>   Note that this patch does not compress attached patches, for the sake
>   of backward-compatibility.  But it does resolve the points of issue1350:
>   hide context when reviewing a patch and ensure a reliable transport with
>   application/x-darcs-patch instead of text/x-darcs-patch.
> 
> Sun Oct 17 17:41:31 CEST 2010  Florent Becker <florent.becker at ens-lyon.org>
>   * resolve issue114: allow hunk-splitting in revert
>   The UI is stolen from record-hunk-splitting and therefore subpar. It
> should be checked by native english-speakers and non-programmers.
> 
> Mon Nov  1 17:14:45 CET 2010  Florent Becker <florent.becker at ens-lyon.org>
>   * resolve issue1970: allow ssh:// urls
> 
> Sun Oct 17 10:52:26 CEST 2010  Florent Becker <florent.becker at ens-lyon.org>
>   * make progress message in pull less silly looking
> 
> Thu Sep  2 12:46:17 CEST 2010  Eric Kow <kowey at darcs.net>
>   * darcs show tags --repo (instead of --repodir).
> 
> Fri Sep 17 23:06:53 CEST 2010  Adolfo Builes <builes.adolfo at googlemail.com>
>   * Solve issue1923: cache mechanism warns about sources outside your
> control
> 
> Wed Aug 25 19:30:45 CEST 2010  Dmitry Astapov <dastapov at gmail.com>
>   * Fix for issue1932
>   Do _not_ check for colons in "isRelative" - everything is covered
>   inside Add.lhs and governed by use of "--reserved-ok" option.
> 
>   Replace isRelative and isAbsolute with functions from System.FilePath
>   for correct handling of platform specifics.
> 
> Wed Aug 25 17:21:35 CEST 2010  Eric Kow <kowey at darcs.net>
>   * Raise base package dependency to >= 4 (keeping it < 5).
>   We now require GHC 6.10.x or higher.
> 
> Fri Aug 13 07:54:54 CEST 2010  Jason Dagit <dagit at codersbase.com>
>   * dramatically improve the parser performance
>   This puts the parser's performance back to where it was before the
>   move to a Parsec-like API.  The cruxt of the fix was to implement
>   linesStartingWith as a primitive instead of implementing it using
>   other parser primitives.  Several other tricks are included, such as
>   using strict tuples that are partially specialized.
> 
> Sun Aug 15 23:13:08 CEST 2010  Eric Kow <kowey at darcs.net>
>   * Restore obliterate output options.
> 
>   They were removed for Darcs 2.5, but we can keep them in unstable
>   and resolve the underlying bug instead.
> 
>   rolling back:
> 
>   Sun Aug 15 20:20:10 BST 2010  Reinier Lamers <tux_rocker at reinier.de>
>     * resolve issue1924: remove output options of obliterate
> 
>        ./tests/issue1909-unrecord-O-misses-tag.sh ->
> ./tests/failing-issue1909-unrecord-O-misses-tag.sh
>       M ./src/Darcs/Commands/Unrecord.lhs -3 +1
> 
> Sun Jul 18 13:33:53 CEST 2010  Eric Kow <kowey at darcs.net>
>   * Resolve issue1530: allow user to pick from multiple email address.
>   If your global authors pref has more than one line in it, Darcs will
>   prompt you something like:
> 
>     You have saved the following email addresses to your global settings:
>     1. Fred Bloggs <fred at example.net>
>     2. Fred Bloggs <fred at home.example.tld>
>     3. BLOGGS Fred <fbloggs at example.com>
>     4. Other
>     Please select an email address for this repository.
> 
> Sat Aug  7 02:33:39 CEST 2010  builes.adolfo at googlemail.com
>   * Resolve issue 1599: automatically expire unused caches
> 
> Wed Jul 21 07:37:31 CEST 2010  Ganesh Sittampalam <ganesh at earth.li>
>   * resolve issue1896: enable witnesses for library
> 
> Fri Jul 23 15:18:47 CEST 2010  Eric Kow <kowey at darcs.net>
>   * Fix deprecation warning on GHC.Handle in Exec module.
>   It looks like it was deprecated in GHC 6.12 along with the
>   Unicode-oriented rewrite.
> 
> Wed Aug  4 20:53:57 CEST 2010  Petr Rockai <me at mornfall.net>
>   * Re-enable packs and optimize --http.
> 
> Thu Jul 15 12:16:08 CEST 2010  Eric Kow <kowey at darcs.net>
>   * Resolve issue1883: rename --patch-name option to --name.
>   This is usually used in darcs record in shorthand (-m) form.
>   This rename is aimed at eliminating the confusion with the --patch
>   matcher, which bites amend-record and rollback users.
> 
> What we need to make the release happen:
> ========================================
> 
> -Benchmarks, please use darcs-benchmark on HEAD, and report; currently,
> darcs-benchmark is broken for two reasons:
>    * it does not compile with modern ghcs. For this, please get it from
>      darcs get  http://tausenblum.lautre.net/darcsystuff/darcs-benchmark
>      for a modernized versions
>    * The benchmarks repositories are not available anymore. If you have
>      them somewhere, then you can use them (skip darcs-benchmark get);
>      otherwise, we'll make them available as soon as possible.
> 
> -Exotic platform support: we need someone to check that installing on
> windows and mac works. We'll also need a windows installer and something
> (what?) on mac.
> 
> -Check for release blocking bugs and regressions, please use the betas
> and report issues.
> 
> -Marketing: where do we announce and how?
> 
> Where we are with respect to the roadmap:
> =========================================
> In comparison to what http://wiki.darcs.net/Roadmap has for 2.8, a lot
> of things are still marked TODO. On the other hand, we are already
> behind schedule. For this reason, the freeze will be an "elastic
> freeze", as explained above. The unmet release goals are below, if you
> have patches for these points, please shout now!
> 
>    * issue1938: this seems to be definitely better than 2.4, but may be
> a regression with respect to 2.5. Needs investigation.
>    * issue 1934/305: Petr, can this garbage collecting be done in
> hashed-storage for the _darcs directory? Additionally we may or may not
> want to garbage collect the cache or limit the size of the cache. Unless
> a someone comes up real soon with a proposal, this is going to be
> deferred to 2.10
>    * break up cache: we had an implementation of this in progress, but
> no news since. Probably deferred too.
>    * filecache: idem
>    * darcs test: likewise
> 
> I don't want to add anything to this message, and that's fine with darcs.
> 
> Florent
> 



More information about the darcs-users mailing list