Camelia, the Perl 6 bug

IRC log for #darcs, 2013-01-09

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:23 Igloo joined #darcs
00:31 tranma joined #darcs
00:53 mizu_no_oto joined #darcs
01:06 tranma joined #darcs
01:19 mizu_no_oto joined #darcs
01:22 gienah joined #darcs
01:40 tranma joined #darcs
02:15 tranma joined #darcs
02:23 intripoon joined #darcs
02:45 Igloo joined #darcs
02:49 tranma joined #darcs
02:51 favonia joined #darcs
03:23 tranma joined #darcs
03:58 tranma joined #darcs
05:17 * M30W has a correction for the darcs-wiki. So I just darcs get.. patch it, record and send ?
05:31 Heffalump sure, or change it online
05:32 M30W Online? O.o
05:32 Heffalump http://wiki.darcs.net
05:33 M30W Heh
05:33 M30W [Edit]
05:33 M30W Yea, I was online when I saw it.
05:33 M30W Neat
05:33 M30W I'm gonna get my darcs configured to use mutt (and therego setup mutt) ;)
06:35 jochu joined #darcs
07:17 jochu joined #darcs
08:45 schlaftier joined #darcs
08:45 donri joined #darcs
09:07 jeltsch joined #darcs
09:08 jeltsch joined #darcs
09:50 favonia joined #darcs
09:52 donri joined #darcs
10:19 raichoo joined #darcs
11:12 owst joined #darcs
12:03 mornfall Btw., in darcs 2.8.3, I can darcs mv files that are not in the repository.
12:04 mornfall The move gets recorded in pending.
12:04 owst Excellent
12:04 mornfall So doing a darcs mv *.x dir/ tends to corrupt the repository.
12:07 owst To the BTS-mobile!
12:07 mornfall :D
12:07 mornfall Can you reproduce?
12:09 owst You're running the mv from inside the repository directory?
12:09 owst % darcs mv ../file.txt foodir
12:09 owst Ignoring non-repository paths: ../file.txt
12:09 owst darcs failed:  Both source and destination must be valid.
12:09 owst (trying to move a file from above the repo into a dir in the repo)
12:09 owst That was on 2.9.5 (+ 34 patches)
13:14 mizu_no_oto joined #darcs
14:09 jyyou_ joined #darcs
14:32 mizu_no_oto joined #darcs
15:14 gh___ joined #darcs
15:39 mizu_no_oto joined #darcs
16:35 mizu_no_oto joined #darcs
17:30 MasseR joined #darcs
17:31 MasseR Hi, how can I install darcs-bridge? http://darcs.net/DarcsBridg​eUsage#how-can-i-install-it says to get **TBC** darcs-bridge, but what's the dbc?
17:33 owst MasseR: the TBC is "wait until it's not broken"
17:33 owst MasseR: what do you want to use it for?
17:34 MasseR git -> darcs
17:34 MasseR I was thinking of trying darcs for a year and see how I like it
17:35 MasseR (and later on darcs -> git to keep them still in github)
17:37 owst The problem I'm having is in the darcs->git direction, when keeping the two in sync, so if you could live without it, things would be ok (we have hub.darcs.net as a shared repo hosting solution, too)
17:39 MasseR What kind of problems?
17:40 owst Do you know how git and darcs are implemented?
17:40 owst I encode git's commit DAG into a linear sequence of patches and tags, unwravelling that linear sequence, so I can communicate changes back to git is proving to be hard
17:41 MasseR Ah
17:41 gh___ joined #darcs
17:43 owst So anyway, importing darcs->git is fine, nothing is lost (you get a bunch of extra tags, but that shouldn't cause too many issues), but trying to go the other way isn't possible right now
17:43 MasseR Hmm... darcs-to-git apparently is more or less working? So as long as I get git->darcs, I could still try
17:44 owst Does the git repo have any branches/merges in the history?
17:44 MasseR Yes
17:44 owst darcs-to-git will not do the right thing with the darcs repo you'll get from importing with darcs-bridge
17:45 owst none of the tools available will
17:46 MasseR Oh damn. Sorry to hear that :/
17:46 owst It's a tricky problem. Unfortunately
17:46 MasseR Heh I believe it is.
17:49 MasseR So no darcs for me then. Patch theory is interesting though. I've been thinking how it would fare as a todo medium (imagine taskwarrior or todo.sh). Taskwarrior merges are brittle beasts and todo.sh doesn't have them at all :P
17:50 owst Heffalump: btw, the tag-per-commit and clean-taggeds approach can't work, unless we know the hash of each patch, which isn't recorded in the tag meta-data :-(
17:52 * owst wonders if it's actually reasonable to only support re-exporting to git in a sane way if a bridge is set up and used, rather than manually performing a import and later an export
17:52 owst That way I can keep whatever metadata I like
17:54 owst and I don't have to add tags in the imported repo for each commit - I can just store the required inventories somewhere
17:55 markstos left #darcs
18:02 sm owst: is it simply that a darcs repo stores only a subset of what a git repo can store ? and really a git repo should convert to multiple darcs repos ?
18:03 owst Yeah, sort of. If you import a "branching" git repo, you do end up with multiple darcs repos
18:03 owst Figuring out what happened in git merges is the tricky part
18:10 * sm ponders merges
18:10 sm I would expect the darcs repo corresponding to that git branch to just have a change saying "stuff merged from XXX"
18:12 owst Alas, that's not how darcs works
18:12 owst To merge, you have to merge the patches
18:13 owst i.e. transfer them into the target repo (and obviously change their representations to make sense)
18:14 owst git can do merges by taking N parent pointers, figuring out what the conflicts are and then committing a single resolution commit
18:14 owst Well, the merge commit and the resolution are one and the same
18:14 owst If no conflicts, there's just a commit with N parent pointers
18:15 owst So anyway, the trickyness is, given a darcs repo with a bunch of merged in patches, how do we figure out what the original representations were for all those patches
18:15 sm I guess I'm not understanding what you just said - it sounds like you're trying to solve a harder problem than I'd really need as a user
18:17 owst Well, it depends on what you mean by a "darcs repo corresponding to a git branch"
18:17 sm I think maybe you're wanting to have the darcs patches corresponding to git merges have the same identity across all the generated darcs repos
18:17 owst Up to randomly generated hashes, yes
18:18 owst otherwise the whole thing would be insane :-) "This bridge keeps your changes in sync, but may give different representations for changes on each side"
18:19 owst Change-ordering doesn't matter to darcs "in theory", it tracks the changes neccessary to re-order things. In git, it absolutely matters
18:19 owst sm: The real issue is: given a git->darcs import, how can I export any subsequently added darcs patches, such that git can make sense of where they came from?
18:20 sm__ joined #darcs
18:22 sm so ongoing sync of a multi-branch git repo to darcs repos is hard
18:22 owst yes
18:23 sm but one-time sync of same to darcs repos is quite doable ? maybe working already ?
18:23 owst yes, it works in darcs-bridge
18:23 sm even that, maintained, would be very useful
18:24 owst my GSOC easily got that far, and then got stuck, very stuck, on the harder syncing problem
18:25 owst (I mean one-time multi-branch git -> multiple darcs repo import)
18:26 sm owst: and does single darcs repo -> git repo work ?
18:26 sm I think you said it did
18:27 owst In theory yes, but the code's pretty tangled up with the general case of unwravelling imported git merges
18:28 owst Heffalump: have you ever come across Petri nets? In particular, Petri net unfoldings? They very much remind me of mergers, and the unwravelling transformation on a merger
18:28 sm it would be really great to have a stable darcs-bridge 1.0 that makes darcs-to-git and one-time-git-to-darcses a mundane everyday activity
18:29 sm it would take away any fear of using darcs for a project
18:29 sm no lock-in
18:29 owst what, the darcs->git "escape hatch" being present?
18:29 sm yes
18:30 owst I'm pretty sure darcs-fastconvert on hackage would do fine at darcs->git
18:31 sm I have my doubts about that, I'll test it
18:31 owst Cool
18:31 sm I expect it to be uninstallable
18:32 owst Yeah, likely
18:32 owst it'll be against 2.5.2 darcs I guess
18:32 sm yup
18:33 sm so darcs is still a risky island of vcs-dom, as yet
18:33 owst Maybe I'll try and extract the working-core of darcs-bridge
18:33 sm unless you're an enthusiast, you'd be unwise to put an important project in it
18:34 * sm is wearing the skeptical project leader hat
18:34 owst A useful hat to wear
18:39 Heffalump owst: no, never really learnt about Petri nets, bit of a gap in my education :-)
18:39 owst I don't really understand unfoldings, but they are apparently a way of encoding a history of net transitions, as a net itself
18:40 owst Which you need in the situation when there are choices that the net can take - you need to know which choice was made
18:51 kerneis joined #darcs
19:12 kerneis joined #darcs
21:12 mizu_no_oto joined #darcs
21:56 sm__ joined #darcs
22:25 owst joined #darcs
22:29 jochu joined #darcs
23:38 schlaftier joined #darcs

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary