Camelia, the Perl 6 bug

IRC log for #darcs, 2011-04-29

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

All times shown according to UTC.

Time Nick Message
01:02 Riastradh joined #darcs
01:19 secorp left #darcs
01:28 copumpkin left #darcs
01:43 ManateeLazyCat joined #darcs
01:47 owst joined #darcs
01:54 exlevan left #darcs
01:54 exlevan joined #darcs
02:00 exlevan left #darcs
02:03 iago left #darcs
02:13 ManateeLazyCat left #darcs
02:18 owst left #darcs
02:30 intripoon joined #darcs
02:33 intripoon_ left #darcs
02:52 gwern left #darcs
02:52 gwern joined #darcs
02:52 gwern left #darcs
02:52 gwern joined #darcs
03:33 sm left #darcs
03:44 copumpkin joined #darcs
03:45 lispy_ joined #darcs
05:18 intripoon_ joined #darcs
05:21 intripoon left #darcs
05:25 lispy_ left #darcs
05:33 intripoon joined #darcs
05:36 intripoon_ left #darcs
06:03 copumpkin left #darcs
06:24 exlevan joined #darcs
06:37 exlevan left #darcs
07:39 Weltraumschaf joined #darcs
08:39 gal_bolle joined #darcs
08:43 jeltsch joined #darcs
08:48 shenshei joined #darcs
09:35 McManiaC joined #darcs
09:36 McManiaC can you convert a patch file to a regular diff?
10:09 raichoo joined #darcs
10:11 Gowilla not directly, but if you apply it to a repo then you can do darcs diff
10:31 Gowilla is now known as Heffalump
10:43 exlevan joined #darcs
11:02 gal_bolle left #darcs
11:03 gal_bolle joined #darcs
11:48 owst joined #darcs
11:57 dixie that windows installer is very nice
12:12 kowey joined #darcs
12:32 Heffalump hi
12:33 owst 'lo
12:35 exlevan hi
12:53 kowey http://neilmitchell.blogspot.com​/2011/04/darcs-for-my-wife.html
12:58 gal_bolle plop all
12:59 gal_bolle mornfall: in hashed-storage, if I have a (Tree m) object, is there a way to know how the directory hashes were computed?
13:13 copumpkin joined #darcs
13:56 Weltraumschaf left #darcs
14:11 McManiaC Heffalump: well I don't have a darcs repo anymore which I could apply it to (ghc repo)
14:11 contra-sh joined #darcs
14:11 Heffalump McManiaC: I think the ghc wiki has some instructions on porting darcs patches
14:11 McManiaC ok, thanks. I'll have a look
14:12 Heffalump *if* the darcs patch is just hunk patches, it would be theoretically possible to turn it into a diff directly. But I'm not aware of any tools that do it.
14:33 secorp joined #darcs
14:36 copumpkin left #darcs
14:38 Riastradh left #darcs
15:00 copumpkin joined #darcs
15:12 McManiaC http://hpaste.org/46164/darcs_markconflicts <- how exactly do I read this conflict thing? the code below "v v v" is the old version, the code above "^ ^ ^" the new one - but what about the part between "===" and "***" ?
15:15 owst McManiaC: Have a peek at `darcs mark-conflicts --help`, essentially, between "v v v" and "===" is the base (original) code, and the next two sections show the two conflicting changes.
15:16 McManiaC but why are there two conflicts?
15:17 owst That's what a conflict is - two changes that have been made from the same "original" state.
15:17 gwern left #darcs
15:18 McManiaC ooooh I see
15:18 McManiaC kk thanks
15:18 owst Darcs can't pick which is "right" so asks you to do so.
15:19 raichoo left #darcs
15:23 Heffalump the mark-conflicts is out of date, isn't it?
15:23 owst ?
15:25 Heffalump I omitted the vital word "help" :-)
15:26 owst Heh, I wondered if that's what you meant ;). Do you mean since you added the ==== section? (that's in the help I have, at least).
15:26 owst Oh, but that's built from screen.
15:26 owst screened*
15:26 Heffalump oh, perhaps we did update the help sometime later.
15:27 owst Ah, 2.5.2 has out-of-date help.
15:27 Heffalump the 2.7.3 I appear to have built today also has out-of-date help
15:28 Heffalump but I'm continually flipping around what tree I have my installed darcs from
15:28 Heffalump ah, Guillaume fixed it a few weeks ago
15:28 sm joined #darcs
15:30 shenshei left #darcs
15:32 navaati joined #darcs
15:33 navaati hello
15:33 owst hi
15:33 gwern joined #darcs
15:33 gwern left #darcs
15:33 gwern joined #darcs
15:33 sm_ joined #darcs
15:34 sm_ morning,
15:34 sm left #darcs
15:34 sm_ is now known as sm
15:35 owst Hi sm, thanks for adding my blog to planet darcs, I had to rush off after sending the patch.
15:35 sm no problem owst
15:36 Heffalump grmph, darcswatch is on strike :-(
15:38 sm what do you guys use darcswatch for ?
15:41 Heffalump it updates statuses in the patch tracker
15:41 Heffalump it's a real pain doing that by hand
15:41 Riastradh joined #darcs
15:44 raichoo joined #darcs
15:47 sm darcswatch reaches into our bugtracker ? interesting, I'd like to hear more about that
15:48 navaati how can I want to record, darcs show me a hunk patch that I would like to broke into 2 parts, how can I achieve that ?
15:48 navaati s/how can i want/when i want/
15:48 sm navaati: use 'e'
15:50 Heffalump sm: ask kowey for the details
15:50 kowey documented on http://wiki.darcs.net/BugTracker/Infrastructure
15:50 sm thanks!
15:52 sm a nice page
15:53 sm I'd love to be able to work on darcs infrastructure a bit
15:54 sm I had a nice chat with thomas hartmann yesterday, kicking around the idea of updating patch-tag
15:54 jderque joined #darcs
15:56 sm I wonder what scenarios could allow a darcs hosting service to generate income, if any
15:57 Heffalump a miracle?
15:58 Heffalump well, apart from tip-jar money
15:58 sm well even that would be better than the present status
15:59 sm (I assume)
16:00 kowey flattr integration would be nice for one of these systems (for the hosted projects and the project itself), but I wouldn't bet on this helping much
16:00 Heffalump who is available to work on anything like that?
16:01 sm I might be. I'm looking for new work and thomas might have a little funding.
16:02 Heffalump can you imagine it being able to actually pay for development?
16:02 * kowey brainstorms randomly: one idea might be to look beyond source control
16:03 jeltsch left #darcs
16:03 kowey sure something that's effectively a darcs host under the hood
16:03 kowey but for which the darcs hosting is actually just incidental, besides the point
16:03 kowey a weak example is something like wikia
16:04 sm well I can *imagine* a lot, but yes it's hard (understatement) to see that paying a full-time developer any time soon. Even a lowish rate for part-time work would allow some sustained effort though
16:04 Heffalump what sort of income level is needed, including hosting?
16:05 shenshei joined #darcs
16:05 sm I don't know. Just that >0 income would allow >0 continuous work
16:06 Heffalump well, there's hosting to pay for too. So that's not quite true.
16:07 sm I think that's cheap enough it can be ignored for now. TH is already paying that. Probably several of us could fit a patch-tag into hosting we're already paying for
16:08 sm you're right in a whole business sense, I'm not trying to be too hard-headed, just trying to think of some new possibilities
16:08 Heffalump yeah, it'd be great to get some movement
16:08 sm a kickstarter (fund drive) for particular goals might raise a couple of hundred $ now and then
16:09 Heffalump do you have any idea what the features roadmap would be?
16:09 raichoo left #darcs
16:09 gal_bolle left #darcs
16:09 Heffalump e.g. I think there was talking of bringing darcsden and patch-tag together. The last time we discussed it, I had the impression darcsden was more viable, but that mainly depends on where the developer effort is available
16:10 sm for patch-tag ? he listed a few, let's see.. update it to latest libs, make it easy to deploy (cabal install).. longer term, try to merge the best of darcsden and patch-tag
16:10 sm and the other one: make it bring in *some* income
16:10 sm yeah exactly.. I think I was talking about it
16:12 sm massive UI overhaul, and slick ecommerce bits would also be required
16:13 sm and if it could evolve features that make darcs development easier, that would be good
16:13 sm I mean darcs itself
16:14 Heffalump not quite sure what you mean
16:18 sm sorry.. it would be nice if some improved darcs-hub could make development of darcs easier and faster, eg complementing or improving some of the services we currently use
16:19 Heffalump ah, right.
16:19 sm features like that would have a high priority
16:19 Heffalump I think what we have at the moment is basically good enough, if a bit unreliable (e.g. darcswatch)
16:19 Heffalump though of course there are plenty of workflow improvements that are always possible
16:22 sm yes
16:23 Heffalump actually, something that supported patch review nicely would be great
16:23 owst +1
16:23 sm yes, that's an obvious one
16:25 Heffalump does github do that well?
16:25 sm in my happy imaginary world, darcs-hub has this advantage over the almight github: darcs is small and close enough to co-evolve with it. darcs and darcs-hub each could change in ways useful to support the other
16:26 Heffalump not convinced by that argument, cos is my happy imaginary world, darcs is bigger than git :-p
16:26 sm github *does* do that well, at least much better than I'm used to. Pull requests are nicely handled, with appropriate notifications, and line-by-line commenting
16:27 copumpkin (fwiw, I'd happily pay for darcs hosting if I found something I liked)
16:32 Heffalump sm: btw, if it helps with deciding it's worth to get started, I'd be happy to pledge say $50
16:33 lispy I wonder if you could find funding via kickstarter or one of those other microlending sites
16:33 sm cool - thanks Heffalump, that kind of thing *is* helpful
16:34 alexsuraci i'm doing some darcsden ssh debugging today probably
16:35 sm I think "slick ecommerce bits" are also interesting.. I think a given project could have dramatically different levels of donations depending on how that stuff is handled
16:35 sm minecraft and the humble bundle were some good examples of making contributing fun
16:35 sm cool alexsuraci
16:37 sm alexsuraci: TH said you said you had studied ssh quite a bit before you wrote that stuff.. true ?
16:49 alexsuraci sm: I read through the RFCs while implementing it, yeah
16:49 alexsuraci and compared to openSSH
16:50 sm and it's working pretty well eh
16:50 alexsuraci yeah, if you're me :P
16:50 sm heh
16:50 alexsuraci http://darcsden.com/alex/s​sh/browse/src/SSH.hs#L-320
16:50 alexsuraci apparently that's getting something that's not 0 for someone, and failing
16:51 alexsuraci which is odd because the spec says boolean FALSE for that byte
16:51 alexsuraci so i'll be debugging that later
16:51 mornfall Evening.
16:51 owst hi mornfall
16:51 copumpkin lispy: good idea
16:51 copumpkin some friends of mine made craploads o' startup cash with kickstarter
16:51 copumpkin it'd also give me a link to tweet
16:52 kowey "works with github" might be a useful feature for this sort of thing to have
16:53 kowey dunno if this requires any sort of co-operation on their part
16:53 Igloo copumpkin: What were they starting up?
16:54 copumpkin Igloo: http://www.kickstarter.com/projects/106623684​/openvizsla-open-source-usb-protocol-analyzer
16:54 copumpkin a pretty "niche" product
16:54 copumpkin probably helped by the fact that they have lots of nerds who respect them :)
16:55 Igloo In that case you also get something that you would otherwise have to buy later
16:55 Igloo Which wouldn't be the case with a free-for-open-source darcs hosting site
16:56 copumpkin sure, but sm could give away early beta access to people who give $X, and maybe a year of private hosting for $(X + k) and lifetime hosting for $(X + 5 * k) or something
16:56 copumpkin I dunno
16:56 copumpkin it doesn't necessarily have to be anything concrete
16:56 raichoo joined #darcs
16:56 copumpkin that kind of stuff is surprisingly successful
16:58 copumpkin (it also can't hurt to try, because if you don't meet the goal, you have no commitment to the people who have pledged money, and they have none to you)
16:59 Igloo Sure
17:00 Heffalump I'd suggest coming up with a concrete proposal before starting publicity though.
17:00 copumpkin yeah, definitely
17:00 sm certainly
17:01 copumpkin sm: poke me if/when you want some twitter exposure
17:01 sm thanks copumpkin !
17:02 sm do you think a series of smaller goals and fund-raising events can work ? In some ways that would be preferable to big ones
17:03 copumpkin not really sure :) I have no business experience whatsoever, sadly
17:03 sm eg start with "modernise patch-tag and make it easy to contribute to"
17:05 Heffalump easy to install is an important part of that
17:05 sm assuming that goes well, next month it might be "UI refresh" (a new kickstarter drive). After that, could be "integrate with darcsden". And so on. Or maybe smaller goals than these
17:05 Heffalump (that's worth mentioning explicitly, because it's an immediate benefit to users)
17:05 sm absolutely
17:06 Heffalump what would the target be each time?
17:06 alexsuraci sm: so are you working on patch-tag?
17:06 alexsuraci sort of out of the loop here
17:06 sm bad timing, but I'm being called to breakfast.. sorry, biab
17:06 alexsuraci heh, np
17:07 sm alexsuraci: if you can, read scrollback or log.. was just kicking around some ideas with thomas hartmann yesterday, about getting some funding to update it
17:07 alexsuraci sure
17:07 burp left #darcs
17:07 burp joined #darcs
17:07 sm who btw is in #haskell right now
17:07 sm afk
17:07 alexsuraci ah
17:34 secorp left #darcs
17:41 sm back
17:54 thorkilnaur left #darcs
17:54 thorkilnaur joined #darcs
17:56 navaati i've got a problem with darcs and UTF8
17:59 Heffalump navaati: what version? (and what's the problem)
18:00 navaati at the first patch création, it asked my email, so i typed it in the format "Name <email>", but there is an é in my name
18:01 navaati and now I get 'L<U+00E9>o' in the patch descriptions of the 1st one and 'L<U+00C3><U+00A9>o' in the next patches
18:01 navaati darcs 2.5.2
18:01 Riastradh left #darcs
18:05 * Heffalump tries to reproduce
18:06 Heffalump what OS?
18:06 navaati gentoo linux
18:06 navaati LANG=fr_FR.UTF-8
18:09 navaati when i open the _darcs/prefs/author in a text editor, it appears as 'Léo'… i'll try to correct it and see if new patches are ok
18:10 kowey hopefully this just means something superficial like darcs is neglecting to write prefs files in locale
18:10 kowey but is reading them in that way
18:14 kowey left #darcs
18:14 navaati fail : when editing _darcs/prefs/author, i get again 'L<U+00E9>o' in darcs changes
18:15 navaati so it wrote it wrong in prefs/author when it hasked me my name, and then read it wrong when printing darcs changes
18:16 navaati i can correct the first issue, but not the second…
18:37 shenshei left #darcs
18:40 Heffalump with en_GB.UTF-8, I get the right thing in _darcs/prefs/author, but it does the wrong thing displaying the ptch
18:40 Heffalump but the on-disk structures are correct
18:42 shenshei joined #darcs
18:44 ManateeLazyCat joined #darcs
18:45 Heffalump navaati: can you confirm that darcs is writing out a bad _darcs/prefs/author for you? I can't reproduce that. Is it possible the text editor displayed it incorrectly?
18:45 gbeshers left #darcs
18:46 secorp joined #darcs
18:51 alexsuraci When was the DARCS_PATCHES_XML env var introduced?
18:51 alexsuraci (roughly)
18:52 owst left #darcs
18:55 Heffalump test was added in March 2010
18:55 Heffalump still looking for the actual code change :-)
18:56 Heffalump Feb 2008
18:57 Heffalump looks like it's capped at 1K, too
18:57 Heffalump which might be problematic..
19:01 jeltsch joined #darcs
19:09 gbeshers joined #darcs
19:09 Riastradh joined #darcs
19:13 alexsuraci Thanks
19:13 JaffaCake joined #darcs
19:14 navaati how can i modify a patch that is not on the head of the patch list ? (I made an error in the patch name)
19:15 Heffalump does darcs amend-record offer you the patch?
19:15 Heffalump If not, something else depends on it, and you can't easily modify it until rebase becomes available. If you're very keen you can use the experimental rebase branch.
19:15 Heffalump If only a few things depend on it, you can unrecord those things first
19:16 ManateeLazyCat left #darcs
19:17 JaffaCake1 left #darcs
19:18 iago joined #darcs
19:20 navaati i think many things depend on it : it's my initial commit -_-
19:21 navaati can't i edit the patch file in _darcs ?
19:23 Heffalump lots of other things will have the same data in them, e.g. _darcs/hashed_inventory
19:23 Heffalump and hashes are based on it
19:23 Heffalump so not practically, no
19:23 navaati ok… doesn't really matter
19:24 Heffalump the name is part of the identity of the patch, so changing it isn't as trivial as you'd expect
19:26 mornfall Which is both good and bad.
19:38 iago I add a new file and darcs inferred this addition as one 'addfile' patch and to 'hunk' patches, the first hunk just add the first line whilst the second adds the rest of the content, is this ok ?
19:38 iago s/add/added s/to/two
19:39 Heffalump iago: it's a slightly funny thing about the way darcs handles new files - addfile actually gives you a file with a single newline in it
19:39 Heffalump so the first diff often contains content to add before that newline and some more to add after it
19:39 Heffalump mornfall: should review that for patches v3
19:39 iago oh
19:40 mornfall Indeed, good idea.
19:40 mornfall There's also another mess-up with newlines in hunks.
19:40 Heffalump do you have a list somewhere?
19:40 mornfall All because, presumably, the list-of-lines representation of files used internally in a few places.
19:41 mornfall List of things to fix about prims?
19:41 Heffalump well, everything :-)
19:42 mornfall diff_to_empty p x | BLC.last x == '\n' = line_diff p (init $ linesB x) []
19:42 mornfall | otherwise = line_diff p (linesB x) [BS.empty]
19:42 mornfall diff_from_empty p x = invert (diff_to_empty p x)
19:42 mornfall ^^ I meant this here.
19:42 mornfall Something's wrong about that.
19:43 mornfall (And this text_diff thing probably belongs into the prim implementation anyway.)
19:44 Heffalump I guess I do mean specifically a list of things to fix/improve about prims.
19:44 mornfall I shall make one.
19:44 mornfall Probably in the wiki?
19:45 mornfall Although the newline stuff should fall out automatically out of the blocky representation of hunks.
19:46 Heffalump yeah, wiki. And yes, hopefully.
19:46 Heffalump but it should have tests too
19:50 mornfall Oh, btw.
19:50 mornfall What exactly does "old text" in conflicts mean?
19:50 mornfall And, can we get a label to that effect in the marker?
19:50 mornfall I did a merge recently and was pretty confused about what it meant.
19:51 iago mornfall, if you end understanding that stuff maybe you should document it properly
19:51 iago in that way it will stop to be a "secret" known by 2-3 people
19:51 iago :P
19:51 mornfall iago: About hunks?
19:51 iago about conflicts specially
19:51 mornfall Ah.
19:51 iago but about everything in general
19:51 mornfall Conflicts will be... different. Eventually.
19:51 iago although I think that Prim's are pretty well understood
19:52 mornfall I'll probably try to track the patchids inside v3 prims.
19:52 iago well, you are right, you will change stuff
19:53 mornfall Well, I won't change conflicts. Not this summer, anyway.
19:53 Heffalump what do you mean by "what exactly"? It's the text from before either patch is applied.
19:53 mornfall Heffalump: Ah!
19:54 mornfall So it's the merge base, of sorts.
19:54 Heffalump yeah
19:54 mornfall Which would be good to point out in the marker itself.
19:54 Heffalump we don't normally put explanatory text in conflict markers
19:54 mornfall I think.
19:54 mornfall I know.
19:54 mornfall But we should start.
19:54 Heffalump I'm not averse to changing it though
19:54 Heffalump if it's not too ugly.
19:55 mornfall So we have v v v v, = = = = =, * * * *, ^ ^ ^ ^ now, right?
19:55 mornfall Or something like that, anyway.
19:55 Heffalump it might fit in with adding the patchids in the conflict (prototype of that here: http://urchin.earth.li/darcs/ganes​h/darcs-branches/marking-20110222/ )
19:55 mornfall Maybe just appending (merge base) to = = = = would work?
19:55 Heffalump how would that indicate that it was the bit above?
19:56 mornfall Oh, it's the above?
19:56 mornfall I am confused now.
19:56 Heffalump first bit is merge base, next bits are conflicting options
19:56 mornfall So the base is between v v v and = = =?
19:56 Heffalump yes
19:56 mornfall Dang!
19:56 mornfall That's why I was so confused.
19:57 Heffalump I'm actually amazed you're so confused, because everyone else I've encountered seemed to find it fairly obvious (unless my memory is faulty)
19:57 mornfall I have somehow assumed it was between = = = and * * *, with v v v and ^ ^ ^ being the branches of the conflict.
19:58 mornfall Dunno, I always interpreted the v v v / ^ ^ ^ markers as representing the "sides" of the conflict.
19:58 mornfall Multi-way conflicts are rare enough.
19:58 Heffalump I don't think they could, because a three-way conflict would have the middle side separated by *** on both sides
19:58 mornfall Yes of course.
19:58 Heffalump and I think the merge base has to come first for the same reason.
19:59 mornfall Well, I would definitely welcome something clearer.
19:59 mornfall Even if we don't have patchids yet.
20:00 mornfall (I wouldn't be opposed to an extension of merge(1) style of markers.)
20:01 Riastradh I have never understood conflict markers...  I always look at the before and after for the two branches and piece together what the right thing to do is by hand.
20:01 Riastradh (This applies not just to Darcs but also to Git, CVS, &c.)
20:02 mornfall Curiously enough, merge(1) does not include base in conflicts.
20:02 mornfall I would still expect to find it in the middle, in a 2-way conflict.
20:02 mornfall Darcs has a problem in that department, though.
20:02 jderque left #darcs
20:04 mornfall (And darcs get does take a forever and then a while. :( ...)
20:05 mornfall +        Just (p' :> _) -> -- note that doing a merge via commute doesn't work for conflictors: why not?
20:05 mornfall I don't like the sound of that.
20:07 Heffalump I think the quick answer is that conflictors are weirdly broken.
20:08 mornfall I recall David telling me that darcs didn't really use commute-merge, maybe he did mean that it doesn't work for merger/conflictor patches.
20:36 contra-sh left #darcs
21:24 dcolish left #darcs
21:24 dcolish joined #darcs
21:38 navaati is there a way to display a dependency graph of a repository ?
21:39 Heffalump navaati: http://lists.osuosl.org/pipermail/​darcs-users/2011-April/026106.html is the current state of knowledge on it
21:47 navaati "doesn't use the darcs library but shells out to the darcs executable, doing dry-runs of pulls into an empty, temporary repository, parsing the output and generating the dependency graph from that." this guy is mad oO
22:01 sm current conflict markers are utterly confusing
22:02 sm I'm amazed you thought otherwise Heffalump  :)
22:09 Riastradh left #darcs
22:23 gwern Heffalump: eh, it's no worse than filestore, which does everything through the shell...
22:27 Heffalump sm: oh, ok :-)
22:27 Heffalump navaati: the darcs library has an unstable API, and you need to use Haskell to use it
22:28 gwern also looks pretty unusable to me. to do even something like get the changelog summaries, I'd've'd rewrite most of 'darcs changes'
22:33 pumpkin joined #darcs
22:33 copumpkin left #darcs
22:39 navaati left #darcs
22:43 shenshei left #darcs
22:50 raichoo left #darcs
22:51 rks joined #darcs
22:51 rks left #darcs
22:54 owst joined #darcs
23:53 owst left #darcs

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