mirror of http://darcs.net/screened (fork of darcs's darcs-reviewed) (http://darcs.net/Development/GettingStarted)

remove unused color flag in darcs.cabal

It should have been removed by this patch from 2009: "Resolve issue1366: remove unreachable code for direct ncurses use".

ghTue Mar 24 18:48:40 UTC 2015

fast-export: rewriting single dots in tags was unneeded

ghFri Mar 20 20:38:32 UTC 2015

resolve issue2446: in fast-export delete dirs before dumping contents

ghFri Mar 20 20:27:02 UTC 2015

resolve issue2433: fix help strings of send and fetch

ghTue Mar 17 20:16:49 UTC 2015

clarify rollback help string

ghTue Mar 17 19:55:10 UTC 2015

resolve issue2420: commandGetArgPossibilities for apply should not list all files in the repo

bfrTue Mar 10 20:03:32 UTC 2015

use die instead of fail in Darcs.UI.RunCommand

bfrTue Mar 10 11:58:16 UTC 2015

added die action to Darcs.Util.Exception as a replacement for fail

It is commonly agreed upon that fail is an ugly wart in the Monad class and should not be used for serious programs. Its only sensible justification is to support de-sugaring of do blocks with failing pattern matches.

bfrTue Mar 10 11:54:23 UTC 2015

Darcs.UI.Defaults: upper case initial letter in error messages

This brings it in line with how messages in Darcs are formatted elsewhere.

bfrSat Mar 7 12:25:46 UTC 2015

purify Darcs.UI.Defaults, fixing two problems on the way

The IO that was in Darcs.UI.Defaults has been moved to the caller (Darcs.UI.RunCommand.runCommand). This allows us to fix a regression: relative paths given as option arguments in defaults files were not resolved with respect to the same working directory as when giving the option on the command line. The ordering constraints of commandPrereq vs. getGlobal/getPreflist are now concentrated in one place and documented.

Some changes related to errors and how to report them are also included in this patch since it would have been difficult to record them separately:

  • applyDefaults no longer stops on the first error, instead accumulates all errors and returns them together with the result in a pair. This makes it easy to postpone the actual failure to a point after the standard command actions (--help, --list-options, --disable) have been handled. Which in turn fixes a problem with the generalized contrib/_darcs.zsh.

  • replacing use of unlines with (intercalate "") when concatenating error messages avoids many uses of the ugly chompTrailingNewline

  • error messages returned by getOpt and those generated by applyDefaults are now in the same format

bfrSun Feb 22 21:14:01 UTC 2015

replace old command names in network tests (get, put, changes)

ghSun Mar 8 20:13:53 UTC 2015

reintroduce put command as deprecated

ghSun Mar 8 20:07:08 UTC 2015

allow flag to prevent patch index creation with log and annotate

ename --patch-index to --with-patch-index to avoid conflict with existing --patch flag used with log and annotate

ghSun Mar 8 19:50:19 UTC 2015

export a couple of things needed by darcsden

ganeshFri Mar 6 18:40:37 UTC 2015

add bounds to all dependencies

I've checked that new lower bounds (on binary, dataenc) work.

ganeshThu Mar 5 06:45:27 UTC 2015

bump utf8-string upper bound

ganeshWed Mar 4 22:55:58 UTC 2015

replace pointless import of Control.Monad.Except

This fixes the build with mtl<2.2.1

ganeshWed Mar 4 22:41:39 UTC 2015

silence warning (ghc<7.8) about the MINIMAL pragma

ganeshTue Mar 3 07:56:53 UTC 2015

remove a few unneeded -fno-warn-incomplete-patterns

ganeshTue Mar 3 07:52:39 UTC 2015

Turn off deprecation warning for use of 'parseTime' for now

ganeshWed Mar 4 06:29:06 UTC 2015

fix GHC 7.10 warning: explicitly quantify variables in type alias

ganeshWed Mar 4 06:10:30 UTC 2015

ErrorT is deprecated, switch to ExceptT

The two have different 'fail' implementations, but as far as I can see we don't have any explicit calls to that or any inexhaustive pattern matches in the relevant code.

ganeshTue Mar 3 19:06:41 UTC 2015

resolve issue2432: use the merged version of the local patches

The previous code gave a false sense of security in its use of witnesses on the Repository type, because those can be implicitly mutated by operations on the repository but the Repository value with its old witnesses is still available for use.

The bug was that tentativelyReplacePatches should be passed the replacement patches, as it just adds them to the repository at the end after commuting out the previous versions. So when doing pull/apply --reorder, we need to pass in the local patches merged with the remote patches, rather than in their original form.

I reworked the code so that the witnesses are correct, which provokes a type error, and fixed the error by changing merge2FL to return both the merged remote patches and the merged local patches so the latter can be passed in to tentativelyReplacePatches.

ganeshMon Mar 2 19:46:31 UTC 2015

fix warning

ganeshMon Mar 2 20:04:39 UTC 2015

bump version to 2.11.0

ghSat Feb 28 15:08:11 UTC 2015

bump version to 2.10.0

ghSat Feb 28 15:07:50 UTC 2015

fix patch-index tests after defaults change

ghThu Feb 26 17:52:40 UTC 2015

resolve issue2438: no longer build patch index by default on cloning

  • build automatically only with annotate or non-interactive log file
  • patch index creation only occurs if the lock can be taken
  • PI creation only happens on init, clone and convert if --patch-index is passed
  • PI creation no longer done when finalizing a repo job
  • the only way to have the file _darcs/no_patch_index created, is to run optimize disable-patch-index, or to ctrl-c PI creation (during annotate, log, init, clone or convert)
ghThu Feb 26 17:50:31 UTC 2015

in annotate command, push call to withRepository deeper

ghWed Feb 25 14:39:30 UTC 2015

license DateTime module as BSD3, with original author approval

ghFri Feb 27 17:29:04 UTC 2015