Camelia, the Perl 6 bug

IRC log for #darcs, 2013-05-17

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

All times shown according to UTC.

Time Nick Message
00:00 sm indeed.. you mean darcs, right
00:00 sm I think that's where the change is needed
00:00 sm for now if you have a better text for the ControlMaster answer in the FAQ, I'll gladly use it
00:00 mtp i'm not sure; the failure stopped happening when i turned ControlMaster off, so i would naturally blame the hub's sshd
00:02 mtp I'd say something about how if pushing doesn't seem to work, or it if appears to work and says  "Apply Failed!" anyway, then make sure you have ControlMaster off
00:02 sm well yes I guess implementing ControlMaster would also avoid the problem. That'd be in http://hackage.haskell.org/package/ssh
00:03 sm thanks, I'll add that. The whole docs need a rewrite
00:10 sm done
00:10 mtp sm: thanks :)
00:52 dleverton_ joined #darcs
00:52 byorgey joined #darcs
00:52 MasseR joined #darcs
00:52 tomejaguar joined #darcs
00:52 marmoute joined #darcs
00:52 kaol joined #darcs
00:52 dixie joined #darcs
00:52 Heffalump joined #darcs
00:52 kerneis joined #darcs
00:52 vikraman joined #darcs
00:52 idnar joined #darcs
00:52 mtp joined #darcs
00:52 delamonpansie joined #darcs
00:52 mal`` joined #darcs
00:52 lambdabot joined #darcs
00:52 konundra joined #darcs
00:52 dolio joined #darcs
00:52 xymox joined #darcs
00:52 gbeshers joined #darcs
00:52 edwardk joined #darcs
00:52 jyyou joined #darcs
00:52 lpsmith joined #darcs
00:52 thorkilnaur_ joined #darcs
00:52 mndrix joined #darcs
00:52 dcoutts_ joined #darcs
00:52 Igloo_ joined #darcs
00:52 markstos joined #darcs
00:52 whaletechno joined #darcs
00:52 iago joined #darcs
00:52 preflex joined #darcs
00:52 sm joined #darcs
00:52 stepcut joined #darcs
00:52 mulander joined #darcs
00:52 epta joined #darcs
00:52 teratorn joined #darcs
00:52 haasn joined #darcs
00:52 burp joined #darcs
00:52 alexsuraci_ joined #darcs
00:52 pr joined #darcs
00:52 alexsuraci joined #darcs
00:52 mornfall joined #darcs
01:18 mizu_no_oto joined #darcs
01:24 mizu_no_oto joined #darcs
01:24 favonia joined #darcs
01:51 edwardk joined #darcs
01:56 intripoon joined #darcs
02:18 mizu_no_oto joined #darcs
02:46 mizu_no_oto joined #darcs
03:05 preflex_ joined #darcs
04:14 mizu_no_oto joined #darcs
04:46 carter joined #darcs
05:37 slack1256 joined #darcs
05:37 slack1256 Let's say I am grabbing a copy of the linux kernel and I want to vcs with darcs.
05:37 slack1256 I just "darcs add -r *", wouldn't that create a giant patch?
05:38 mtp yes
05:39 mtp and that is what you want to create
05:39 slack1256 Really? wouldn't that being difficult to handle?
05:41 mtp what, specifically
05:41 mtp not all of your commits are going to be like the initial commit
05:43 slack1256 even so, if for example the linux kernel i do a "darcs add -r *" will freak out because there are so many files (no just some big files)
05:43 mtp i don't know, i've never tried to vcs the linux kernel with darcs
05:48 Heffalump yes, I think it would freak out
05:48 slack1256 http://darcs.net/HintsAndTips#initial-import​-of-existing-project-without-version-control
05:49 slack1256 Found it. well seems quite sane
05:49 slack1256 which vcs wouldn't freak out?
06:00 dolio Why would it freak out? Trying to commute giant patches past things?
06:31 konundra joined #darcs
06:38 lelit joined #darcs
06:42 stepcut joined #darcs
06:47 stepcut joined #darcs
07:38 felipe joined #darcs
07:47 povman joined #darcs
07:52 donri joined #darcs
08:02 amgarchIn9 joined #darcs
08:14 edwardk joined #darcs
09:28 raichoo joined #darcs
10:55 kowey joined #darcs
10:57 owst joined #darcs
11:43 lelit joined #darcs
12:16 kowey been using a lot of git lately
12:16 kowey and still think we get quite a lot right in subjective feel
12:16 kowey i mean sure, we don't have the niceness of rebase -i (but we also rebase less)
12:17 kowey but as nice as git can be in places, things which I still find uncomfortable coming from darcs:
12:17 kowey 1. operations that need a clean working dir (tend to stash, grumble, roll my eyes etc)
12:18 kowey 2. having to anticipate branching
12:18 kowey in Darcs I can use VCS with a lot less discipline, sort of decide after the fact that I'm working on separate things
12:18 kowey all through implicitness offered by interactive transparent cherry picking
12:19 kowey maybe the discipline mandated by Git usage is a good thing (moar planning)
12:19 kowey but I missed the joyful sloppiness of working with darcs
12:19 kowey 3. conflicts
12:19 kowey hey conflicts hurt bad in darcs too (but for different reasons)
12:19 kowey and maybe my confusion is on par with confusion felt by darcs newbie
12:20 kowey (probably is)
12:20 kowey Git doesn't suffer from our conflict marking woes
12:20 kowey but the workflow can feel convoluted/confusing unless you take the time to learn the Git mental model
12:21 kowey if there's one thing I hope we (eventually?) get right, is that having a VCS that people don't have to sit down and learn a mental model for
12:21 kowey that it just comes for free, implicitly through experience, but that without one you can still get by
12:21 kowey and we get this right in many places (the interactive everything, the small used-everywhere set of idioms)
12:21 kowey and of course fall down in other places (conflicts)
12:22 kowey But in an Ideal World, a Darcs for Computer Scientists and Darcs for Dummies would both by superfluous (the former only for fans who want to know)
12:23 kowey because nothing needs to be said (sorry, I had no idea I was going to blather to this extent, was expecting a couple lines or so).
12:42 konundra joined #darcs
12:43 raichoo joined #darcs
12:57 mizu_no_oto joined #darcs
13:14 iago joined #darcs
13:24 idnar joined #darcs
13:26 mizu_no_oto joined #darcs
13:31 uniquenick joined #darcs
14:02 rarasd joined #darcs
14:13 lelit kowey: almost agree on all points
14:15 lelit my most appreciated git feature is "inplace branch switching", followed very closely by the emacs magit package
14:15 markstos left #darcs
14:18 lelit git conflicts markers are better than darcs' for two reason: a. more "standard", i.e. most tools (editors or graphical viewers) deal with them by default; b. they carry better information about which patch is "responsible" of each "conflict branch"
14:18 lelit I know the latter point has been discussed here and has no "simple" solution...
14:39 kerneis hasn't this point been improved in darcs recently?
14:48 sm_ joined #darcs
14:48 gh_ joined #darcs
14:49 lelit it may be, it's been a while since my last conflict :)
14:54 kowey I think Heffalump has actually started some work on the marking issue (apologies if mistaken)
14:59 gh_ kowey, you mean patch name marking in conflicts?
15:00 gh_ he said it was vastly inefficient
15:01 gh_ with the current patch and repository formats
15:01 gh_ but maybe we want an existing inefficient feature over a non-existent feature?
15:10 kowey one worry is that even if we do fix marking, and of course we should find a way (even if would entail a format switch or if we have to find an answer to it being inherently inefficient to the current setup)
15:10 kowey the model of conflicting-patches-cancel-out may just be inherently confusing
15:11 kowey but that's a different story, sorry not being helpful :-)
15:15 iago a minor question, somehow "related", did you considered providing context diffs when recording hunks?
15:19 gh_ iago, you mean unified diff?
15:19 raichoo joined #darcs
15:20 dolio joined #darcs
15:20 iago gh_, yes, as far as unified diffs are a form of context diffs
15:21 iago I mean, when you're recording hunks Darcs only tells you lines to be removed/added, but gives you no context of the lines around
15:21 gh_ iago, record -u
15:22 gh_ iago, patches are welcome to make it really usable :-) (you'll soon know why I tell this)
15:22 iago ok, I was going to ask why isn't it the default choice
15:23 iago but you've answered my question
16:02 sm_ g'day all
16:07 kowey I suspect a lot of defaults in Darcs may be “nobody made a case for it one way or another” - metadefault
16:08 dolio kowey: From what I can tell, people that use git are frequently thinking about policies for using git to work around the fact that git sucks at most things. :)
16:08 kowey I doubt you'll be able to get away from version control policy
16:08 kowey eg. “don't name all your patches foo”
16:08 dolio I don't think that's good.
16:08 kowey on the other hand, having complicated workflows about merging etc
16:08 kowey may be a smell indeed
16:09 kowey but we must be careful
16:09 kowey policy discussions may reflect user goals
16:09 kowey as much as they reflect system shortcomings
16:09 kowey so the hard work is in staring real hard at the policies
16:09 kowey to figure out what the user goals really are
16:09 kowey without writing them off
16:10 dolio I'm talking about stuff like: rebase all your work into mega-patches because git can't show branching history in a good way, and does auto-merges with bad names and stuff.
16:10 * kowey nods
16:11 dolio Manually rewrite history to look nice instead of making the tool display the messy history in a good way.
16:11 kowey I suspect the right people to talk to about history
16:11 kowey are the Mercurial people
16:11 kowey just an instinct
16:11 kowey that they get archeology more
16:11 dolio Bazaar is probably better, if you ask the version control afficionado I work with.
16:11 mtp how about Fossil?
16:12 kowey good question: would be nice to see some more about these
16:12 kowey there's also (sorry to add to the stew) Veracity
16:12 kowey maybe they have some thinking about how history should work
16:13 kowey darcs doesn't even begin to deal with history
16:13 kowey here! have some patches
16:13 dolio Yeah.
16:14 kowey what? sure yeah a patch from June can come after a patch from November, what's so surprising about that?
16:14 mtp history is subjective anyway :)
16:14 kowey well duh, of course patches are going to be different depending on who's repo they're in
16:15 kowey [which is I suspect one thing people don't even realise is happening under the hood because not paying enough attention, so too blissfully ignorant to be scared/surprised by it]
16:15 kowey [if they were scared/surprised, we have a good reason why we do it this way (because we reorder using exact methods, so it's OK)]
16:15 kowey s/exact/known-reversible/
16:15 mtp i deal with p4 at work and everyone else is scared of me because i a) know how to use it and b) like using it
16:16 kowey :-)
16:16 kowey will be good to pick your brain about it in future
16:16 * kowey runs off
16:19 dolio Some things you probably can't fix with policies, even.
16:19 dolio Like, I don't know what policy you could have to fix the fact that git doesn't really know about file renames.
16:19 dolio Maybe someone's come up with something.
16:20 mtp darcs doesn't know about renames either afaik
16:20 dolio And some are probably too simple to be called 'policies.' Like, add 'foo/.empty' if you want to create an 'empty' foo directory, because git doesn't track directories directly.
16:22 dolio Does darcs move not record information about moving files in the repository?
16:23 dolio I'm not sure what the point in having it would be if it didn't.
16:24 whaletechno joined #darcs
16:28 mtp Darcs cannot reliably distinguish between a file being deleted and a new one added, and a file being moved.  Therefore Darcs always assumes the former, and provides the `darcs mv' command to let Darcs know when you want the latter.  This command will also move the file in the working tree (unlike `darcs remove'), unless it has already been moved.
16:28 dolio At least, darcs help move suggest that there's something different about move vs. deleting a file and adding a new one with the same contents.
16:29 dolio I'm not talking about darcs recognizing when you've moved things without darcs.
16:30 dolio I mean actually knowing about file moves.
16:30 * mtp nods
16:30 dolio As I understand it, git only knows about deleting and adding files.
16:30 mtp "darcs only knows about file moves when you tell it; git doesn't know about file moves at all." is my new understanding
16:30 dolio And it uses a heuristic to detect moves if an add is similar enough to a delete.
16:30 dolio But, if the files get to different, it doesn't recognize it as a move anymore.
16:31 mtp right
16:31 dolio So if you try to forward merge some changes, and the files are too different, tough luck.
16:34 dolio I would hope darcs can handle that, since it even has replace for properly handling search-and-replace.
16:34 dolio Not that I've ever had occasion to use that one.
16:47 Heffalump there is a GSoC project proposal to get darcs to detect moves automatically
16:48 Heffalump which would be at record time, rather than merge time like git does
16:48 Heffalump (as I understand git - I could be wrong)
16:55 dolio Well, git can detect them at record time, but the only thing that does is give you nicer status info.
16:55 dolio It doesn't remember it anywhere.
16:56 dolio The auto-detecting is the less important part, really.
17:09 Heffalump remembering it means that even with subsequent edits, you know about it when you merge
17:09 Heffalump I dunno if git looks through the individual change history to see that one of them was an edit, or if it just does three-way diff with the tree as it sees it
17:17 dolio Considering it's one of the main things a guy I work with gripes about with respect to git, it's probably the latter.
17:17 dolio Because if it did something that was good, it wouldn't be a problem.
17:18 amgarchIn9 joined #darcs
17:19 dolio Also looking through the entire history of the branches you're going to merge sounds slow, and you can't do anything slow.
17:20 dolio Even if the results are better.
17:22 amgarchIn9 joined #darcs
17:23 stulli joined #darcs
17:35 Heffalump as opposed to darcs :-)
17:40 mizu_no_oto joined #darcs
18:13 lelit joined #darcs
18:17 nomeata joined #darcs
18:19 amgarchIn9 joined #darcs
18:24 carter joined #darcs
18:43 edwardk joined #darcs
18:51 donri joined #darcs
19:01 gbeshers joined #darcs
19:34 carter joined #darcs
19:35 amgarchIn9 joined #darcs
21:17 amgarchIn9 joined #darcs
21:25 amgarchIn9 joined #darcs
21:26 mizu_no_oto joined #darcs
21:33 amgarchIn9 joined #darcs
22:07 dolio joined #darcs
23:05 mizu_no_oto joined #darcs
23:18 konundra joined #darcs
23:56 dolio joined #darcs

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