gmane.comp.version-control.darcs.user

Conventions on teams using darcs - naming, patch size,etc.

Matt Lamari — 2012-05-23T00:36:59

Are there any conventions that users tend to gravitate towards?

1.  Naming - with patches from a lot of people, do short names like : 
"Fixed the serialization of Foo" work?  Do long names wrap a line
telling exactly what you did work better?  Or do you fall back on bug
numbers?  What if you can't fall back on issue numbers?  Does
pinpointing a patch by name or in some context get tricky?
2.  Change granularity - do people tend to try to split a change up into
parts, e.g. if a library and end-code are in the same repo, maybe patch
the lib first and then the rest adding it as a dependency?
3.  Merging patches. . .  Does pollution of lots of tiny little changes
ever become a problem, or something you address in practice?  For
example, say people all work on some related issue, put in lots of
little fixes, then you're done - is it ever something where you want to
just merge them all and have everyone work in terms of the amended
repository?  Or does the UI/history/etc. make it easy to find, or not
find if you

Re: optimization in HEAD: darcs diff

Guillaume Hoffmann — 2012-05-21T21:29:00

A few memory allocation figures obtained with "+RTS -s" for two cases
(recent patch and old patch):

* patch "kill footnotes" (latest):

2.8
     287,535,128 bytes allocated in the heap
      59,356,640 bytes copied during GC
      14,969,680 bytes maximum residency (9 sample(s))
         342,752 bytes maximum slop
              32 MB total memory in use (0 MB lost due to fragmentation)
[..]
  %GC time      17.5%  (16.5% elapsed)
  Alloc rate    871,318,569 bytes per MUT second
  Productivity  82.5% of total user, 59.8% of total elapsed

HEAD
      62,756,136 bytes allocated in the heap
      37,237,544 bytes copied during GC
      12,593,440 bytes maximum residency (6 sample(s))
         332,928 bytes maximum slop
              24 MB total memory in use (0 MB lost due to fragmentation)
[..]
  %GC time       0.0%  (38.1% elapsed)
  Alloc rate    697,290,400 bytes per MUT second
  Productivity 100.0% of total user, 64.1% of total elapsed

* patch "Import bracketOnError from Workaound" from 2006:

2.8
  17,413

optimization in HEAD: darcs diff

Guillaume Hoffmann — 2012-05-21T19:59:56

I'd like to share to the darcs-users list another significant change
in HEAD that happened after the release of Darcs 2.8, this time an
optimization.

In 2010 Petr Rockai optimized the "darcs diff" command, but a few
things blocked the inclusion of this piece of code into darcs.
Notably, the optimization did not support old-fashioned repositories.
However, since version 2.8 Darcs no longer fully support these
repositories, so the new implementation of "diff" has been ported to
HEAD.

As in Petr's words (http://bugs.darcs.net/patch351):

----8<----
I have changed the diff code to only write those files that actually changed in
the temporary locations. On my other project (less than 1000 working copy
files):

(with cold cache)
head: darcs diff  1,09s user 0,58s system 9% cpu 16,752 total
now:  darcs diff  0,14s user 0,04s system 6% cpu 2,978 total

(with hot cache)
head: darcs diff  0,36s user 0,18s system 98% cpu 0,548 total
now:  darcs diff  0,06s user 0,01s system 86% cpu 0,078 total
---->8----

I've added

Re: Darcs really needs a book

Ash Moran — 2012-05-20T20:43:41

On 20 May 2012, at 21:17, Matt Lamari wrote:


+1 - Pragmatic Version Control Using Darcs? :)

Re: Darcs really needs a book

raichoo — 2012-05-20T20:42:53

On May 20, 2012, at 10:17 PM, Matt Lamari wrote:


+1. I'd buy one for sure.

Darcs really needs a book

Matt Lamari — 2012-05-20T20:17:54

I think that the subject matter really warrants a handy paperback.

Re: mission statement

Ash Moran — 2012-05-20T18:17:51

On 30 Mar 2012, at 17:16, Eric Kow wrote:


Hi Eric

Very delayed reply to this old post, but I found something that may help. I recently read Made to Stick[1] which is about the psychology of memorable ideas. It has examples of a variety of types of idea (urban legends, company slogans, advertising campaigns), but one that seems especially relevant is about the Murray Dranoff Duo Piano Foundation. "Most attendees had never heard of duo piano and asked about it. Duo Piano, shocked that nobody understood, explained the magnificence of the piano. " I found a snippet online[2], but it might be worth a flick through the book.

HTH
Ash

[1] http://www.amazon.co.uk/Made-Stick-ideas-others-unstuck/dp/009950569X/
[2] http://www.bookjive.com/wiki/Book:Made_to_Stick:_Why_Some_Ideas_Survive_and_Others_Die

Re: new feature in HEAD: exponential backoff teststrategy

Guillaume Hoffmann — 2012-05-20T15:39:17

I've updated the wiki page which url is now
http://wiki.darcs.net/Using/Test .

g.

Re: How to separate some files from a repository

Rob Moss — 2012-05-17T16:56:06

Thanks for the suggestion. I've successfully used darcs-fastconvert to
create a git repository, then I used git filter-branch to rewrite the
history, and then used darcs-fastconvert again to create a new darcs
repository. It was certainly much easier than doing things by hand!

If anyone else is interested, here are the steps I took:

# Create a git repository with identical contents
cd original_repo/..
mkdir git_repo && cd git_repo && git init
(cd ../original_repo; darcs-fastconvert export) | git fast-import
git checkout

# Remove all other files from every commit
git filter-branch --tree-filter "rm -rf $SKIP_FILES"

# Create a new darcs repository with the modified patches
git fast-export -M -C --progress=1 HEAD -- | darcs-fastconvert import
../new_repo

# Keep only the relevant (non-empty) patches
cd ../new_repo
darcs unrecord --matches 'not touch file1 && ... && not touch fileN' --all


Yes, certainly. My intention is to split out a one or two modules into
a separate library, which I can then use across

Re: How to separate some files from a repository

Lele Gaifax — 2012-05-17T14:56:34

Last time I did something like that, for exactly the same reasons, I
used tailor, using its pre-commit hook to filter out unwanted entries
from the changesets.

Nowadays I'd probably try to exercise the new fast-import/export way,
at least that's what I successfully used to do similar tweaks exporting
from a bzr repo to git.

Either way, you must be of course aware that the resulting repository
won't be “compatible” with the original, because as you says it will
contain *modified version* of the patches.

Hope this helps,
ciao, lele.

How to separate some files from a repository

Rob Moss — 2012-05-17T10:53:20

Hi,

I began a project about six months ago as a single repository. I've
now reached the point where I'd like to split out several files into a
separate library, for use across multiple projects. Of course, I also
want to preserve the history of these files in the new library
repository. However, several of these patches include changes to other
files (eg, in response to changes in module interfaces) and in order
to pull all of the library-specific patches, I end up having to pull
in a large number of additional patches. What I'd like to do is
identify the relevant patches ("darcs changes file1 ... fileN") and
pull modified versions of these patches that do not include any hunks
that modify other files in the repository.

I can't see a more efficient way of doing this than manually applying
and recording individual hunks (using "darcs changes -v file1 ...
fileN", or "darcs annotate --match XXX" and "darcs unrecord --last 1
--matches XXX"). It sounds as though the "amend-record on steroids"
feature would be u

Re: welcome Simon Michael to the Darcs team

Simon Michael — 2012-05-16T08:55:53

Thanks Eric!

Re: [Haskell-cafe] darcs patch dependencies in dotformat

Joachim Breitner — 2012-05-15T09:26:07

Hi,

Am Montag, den 14.05.2012, 07:21 -0700 schrieb Simon Michael:

I think you’d want to collapse [a]→[b] to [a,b], if no other edges leave
a or reach b. This way you only get arrows at branching and merging
points.

Greetings,
Joachim

welcome Simon Michael to the Darcs team

Eric Kow — 2012-05-14T18:38:44

Hi all,

Just announcing another newcomer to the Darcs team.

Simon Michael recently put a lot of work into revamping the Darcs homepage.
As part of the team, he will be working primarily on the page and documentation, and hopefully over time, expand his areas of interest.

Thanks,

darcs weekly news #96

Guillaume Hoffmann — 2012-05-14T18:49:47

# News and discussions

1.  Darcs 2.8.1 was released, the only change is a build dependency that
    will make it buildable with the next Haskell Platform:

    -   http://lists.osuosl.org/pipermail/darcs-users/2012-May/026506.html
    -   http://hackage.haskell.org/package/darcs-2.8.1

2.  Meanwhile in HEAD, a new test strategy has been implemented:

    -   http://lists.osuosl.org/pipermail/darcs-users/2012-May/026503.html

3.  On the developers' mailing list, we are discussing how to make the
    darcs development process more friendly:

    -    http://lists.osuosl.org/pipermail/darcs-devel/2012-May/013820.html

# Issues resolved in the last week (3)

issue1921 Owen Stephens
  ~ -   return Nothing in splitOnTag if the tag isn't in the patchset
    -   http://bugs.darcs.net/issue1921

issue2095 Ganesh Sittampalam
  ~ -   avoid new GHC encoding behaviour using a global setting
    -   http://bugs.darcs.net/issue2095

issue2161 Owen Stephens
  ~ -   - don't use a separate type-variable for recorded/tentativ

darcs 2.8.1 release

Ganesh Sittampalam — 2012-05-14T18:17:21

Hi,

I've just released darcs 2.8.1 to hackage. The only change it
incorporates is a bump to the dependency on mtl to allow it to build
against the upcoming Haskell Platform.

Therefore there's no need for any users to upgrade, nor is there any
particular need for updated binaries except perhaps to avoid confusion
about binaries being out of date.

Cheers,

Ganesh

Re: Please upgrade your hackage package to mtl-2.1

Ganesh Sittampalam — 2012-05-14T06:30:15

Hi,

On 13/05/2012 12:02, Joachim Breitner wrote:


I'll upload darcs 2.8.1 with this fix in the next day or two.

I've also changed the contact details in darcs.cabal to the darcs-devel
list now we've got that split out again.

Cheers,

Ganesh

Please upgrade your hackage package to mtl-2.1

Joachim Breitner — 2012-05-13T11:02:24

Dear Hackage package maintainer,

the release of the next haskell-platform is close, and it will most
likely ship with mtl-2.1.1. Your package on Hackage does not allow
building against this version, which makes it for your users and
distribution packagers difficult to use the package. Please consider
uploading a new version of your package with an extended mtl (and
possibly transformers) dependency range. In most cases, no version
changes are required.

The list of affected packages¹ are:
darcs
diagrams-cairo
ghc-events 
haskeline
leksah
ltk
template
vty
haskelldb
rss2irc

The current plan for package versions in the platform can be seen at
https://github.com/haskell/haskell-platform/blob/pre-release/haskell-platform.cabal


Thanks,
Joachim

¹ Only listing packages that are also shipped by Debian, not all
affected packages on Hackage.

new feature in HEAD: exponential backoff test strategy

Guillaume Hoffmann — 2012-05-12T23:03:17

Hi everyone,

First, for those who have missed it, since darcs 2.8 all code testing
features of darcs belong to the command "darcs test". Without flags,
"darcs test" runs a test on the latest recorded state of the
repository. However, flags --trackdown and --bisect look for the most
recent recorded version passing a given test.

We already know how --trackdown  and --bisect work: --trackdown runs a
test agains the last recorded version, then the last version minus one
patch, then minus two patches... until finding a passing test.
--bisect sort of does the same by jumping back and forth in the
history by dichotomy.  See http://wiki.darcs.net/Using/Trackdown for
more information and unusually friendly drawings :-)

So, the new interesting feature in HEAD is the exponential backoff
strategy, implemented by Michael Hendricks.  As explained by Michael
himself on the bug tracker:

----8<----
This patch series defines a new "darcs test" search strategy named
--backoff.  It's aimed at finding regressions which are c

Re: [Haskell-cafe] Darcs home page updated

Simon Michael — 2012-05-11T14:07:10

Hi Ben, thanks for the feedback. The body width is specified to keep the page looking coherent and readable when the 
window is very large, as is more common these days (a lot of sites do this). I specified the column widths to fit within 
that with the right proportions, and to avoid scrollbars at default font sizes on all the browsers/platforms I tried. 
Can you send me a screenshot (eg via http://pasteshack.net , it's super-easy) and tell me your browser, platform, and 
special font settings if you have any ?

Best - Simon

Re: darcs 2.8 release

Ben Franksen — 2012-05-10T01:56:36

It does not work. I think the correct path is $HOME/.darcs/author, at least 
that is what darcs says in teh comment in the _darcs/prefs/author it 
generated.

Cheers
Ben