Camelia, the Perl 6 bug

IRC log for #darcs, 2013-02-15

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

All times shown according to UTC.

Time Nick Message
00:12 favonia joined #darcs
00:13 schlaftier joined #darcs
00:15 donri have any of darcs's performance problems stemmed from the use of haskell, or have they all been about time complexity of algorithms?
00:16 donri unrelated: i wonder if a fully functional darcs bridge could be used to make darcshub repos double as git repos that you could clone and push to and hub seamlessly updating the darcs repo
00:17 donri then github mirroring becomes a matter of git mirroring, but maybe more importantly we wouldn't even need github mirroring to appease git users unless we specifically want a github presence as well
00:18 donri sm: thoughts?
00:37 mizu_no_oto joined #darcs
01:53 mizu_no_oto joined #darcs
04:32 favonia joined #darcs
04:51 preflex_ joined #darcs
05:12 carter joined #darcs
06:03 schlaftier joined #darcs
07:25 lelit joined #darcs
07:56 owst joined #darcs
08:31 raichoo joined #darcs
09:07 amgarching joined #darcs
09:09 amgarching Hi, how do you merge ~/.darcs/caches's? I seem to have populated ~/.darcs/cache before moving my home from old laptop. Now I have two of them.
09:16 Heffalump you can just copy them on top of each other
09:16 Heffalump there's no index beyond the presence of files in a directory and identically named files should be identical, as the names are a hash of the contents
09:21 owst joined #darcs
09:21 owst Hello from the Sprint!
09:32 iago joined #darcs
09:38 Heffalump hi!
09:46 gh_ joined #darcs
09:52 gh_ hi from the sprint
09:57 marmoute joined #darcs
09:57 marmoute hi guys
09:57 marmoute I heard you were sprinting in Paris
09:58 dixie hi from the sprint
09:58 marmoute I'm a mercurial developer working and living 1 kilometer away
09:58 gh_ hi marmoute
09:59 Heffalump hi :-)
09:59 gh_ marmoute, right now we are discussing darcs/git interaction
10:00 marmoute I'm considering coming a few hour late today or during the week-end
10:00 gh_ marmoute, great!
10:00 marmoute I guess we can have interresting exchange about version control ^^
10:02 gh_ marmoute, ok, talk to me en private I'll explain you how to come
10:06 gh_ currently we are Radoslav, Owen, Ganesh and me (Guillaume) in the room
10:06 gh_ Owen is filling the whiteboard with bridge-related drawings
10:10 marmoute Bridge idea looks something interresting for everybody.
10:10 owst It will be good if/when it works
10:11 marmoute I'll probably come saturday or sunday.
10:14 Heffalump it's git-specific at the moment, but based around the fastexport format, so should generalise
10:18 gh__ joined #darcs
10:31 owst_ joined #darcs
10:33 owst joined #darcs
10:38 owst joined #darcs
10:45 gh__ talking about darcs' manual
10:59 gh__ I wrote on the sprint backlog: "what we should ensure: 1) darcs.net/manual has something 2) this something corresponds to the current stable branch of darcs (as of now 2.8), because sometimes commands and flags change"
11:26 gal_bolle joined #darcs
11:26 gal_bolle hi all
11:26 gh__ hi gal_bolle
11:26 gal_bolle not having lunch yet?
11:30 Heffalump soon
11:33 gal_bolle so how did the first morning of sprint go?
11:38 gh__ gal_bolle, good, heating up the engines
11:38 gal_bolle ok, I'll go feed mine
11:47 Heffalump http://hub.darcs.net/ganesh/darcs-repo-internals
11:47 Heffalump (relevant to a in-person discussion about the contents of _darcs0
11:53 gh_ joined #darcs
11:55 gh__ joined #darcs
11:59 gh__ sm, just so you know, this is now in HEAD http://bugs.darcs.net/patch1010 and http://bugs.darcs.net/patch1011 and http://bugs.darcs.net/patch354, so darcsden is out of sync  now
12:03 lelit joined #darcs
12:09 bsrk joined #darcs
12:27 kowey joined #darcs
12:27 * kowey waves (but in a meeting shortly)
12:35 MasseR What's that bugtracker? Custom solution?
12:35 bsrk *aditya waves
12:36 gal_bolle definitely easy but important: issue904
12:58 Heffalump MasseR: yes, on top of roundup
12:59 gh__ joined #darcs
13:00 Heffalump so is the conclusion that we should apply the patch?
13:03 raichoo joined #darcs
13:06 gh__ indeed record on sshfs still fails with a recent darcs
13:07 owst joined #darcs
13:07 bsrk joined #darcs
13:07 * owst enjoys free coffee
13:12 gh__ related to http://bugs.darcs.net/issue904 , I created a stub page http://darcs.net/Internals/Locks
13:13 owst joined #darcs
13:13 * kowey wishes the issueXXXX plugin for gitit still was working
13:13 gh__ hi kowey !
13:13 kowey some silly GHC version-cant-find-the-right-dir type thing
13:13 kowey sure it still builds
13:13 kowey (it's a plugin that converts issueXXXX to a BTS link)
13:14 gh__ in the process, we say that darcs help environment lacks a lot of environment variables
13:14 gh__ s/say/saw
13:14 kowey (hi!)
13:14 Heffalump kowey!
13:15 gh__ kowey, maybe it works... I just wanted to be sure by typing the whole url…
13:16 kowey maybe wiser (nah, it's definitely not loading)
13:17 kowey jch contributed quite a bit to the locks code
13:17 kowey in case he's in town/free
13:18 gh__ ok
13:18 kowey but I think it was bitty low level stuff with lots of C
13:18 kowey which may/may not be the stuff you're most interested in
13:19 gh__ patch of issue906 does make darcs record work on sshfs
13:21 gpiero joined #darcs
13:29 mizu_no_oto joined #darcs
13:33 mizu_no_oto joined #darcs
13:34 gh__ gal_bolle, it seems the discussion on the bug tracker is enough to accept  the sshfs patch
13:41 Heffalump do we have an actual procedure for deciding to accept new features? http://darcs.net/Development/NewFeature lists the things to consider, but do we just discuss on a single bug and reach a conclusion, or more explicitly on darcs-devel, or on darcs-users?
13:41 Kindaro joined #darcs
13:44 Kindaro left #darcs
13:45 * dixie enjoys free coffee
13:45 gh__ Heffalump, gal_bolle I'm going to push an amended version of the patch, so remove it from your repos if you have it
13:51 gh__ 4 years and a half before accepting a patch
13:51 gh__ which cleanly applies on HEAD
13:58 gh__ (for people who wonder: I amended the patch name and author name so that it is better tracked by DWN / release scripts)
14:05 mizu_no_oto joined #darcs
14:13 gh__ kowey, there was a repository living at darcs.net/repos/franchise , used in one network test
14:14 gh__ kowey, what other small repo could we use ?
14:16 gh__ I'll go with http://hub.darcs.net/kowey/tabular
14:27 lelit joined #darcs
14:35 raichoo joined #darcs
14:37 * gh__ looking at http://bugs.darcs.net/issue2284
15:37 sm yay cleanup! go sprinters!
15:41 mizu_no_oto joined #darcs
15:47 owst joined #darcs
15:57 favonia joined #darcs
16:02 bsrk joined #darcs
16:10 mizu_no_oto joined #darcs
16:15 owst joined #darcs
16:17 whaletechno joined #darcs
16:54 Heffalump iago: are you around? I'm looking at aPrimPair in the darcs test harness and I want to check the coverage data listed there. Are there any instructions on how to calculate the data?
17:00 iago Hi Heffalump, well, it was kind of dirty and I never committed the code that collects that data. For example, to collect commute cases I had to duplicate the commute function structure
17:00 iago also, I had to make some modules to export functions that were hidden
17:01 iago the code is still there (at home), but I don't know if that's going to be helpful
17:02 Heffalump yes, I think it would be useful
17:02 Heffalump I can always improve it but if there's nothing then I just can't keep it up to date
17:05 iago well, I guess the question is whether you want to keep that code in the repo
17:06 iago it introduces duplication and modifies module interfaces
17:07 iago at the time I though that a good solution may be to paste the "diff" in the wiki, so you could use it as a template to test your generators
17:07 iago I never did that :-P
17:08 iago in any case I'll send you the code when I get home
17:10 Heffalump thanks
17:10 Heffalump the reason I care right now is that QuickCheck or test-framework changed to fail any tests if the arguments are exhausted
17:10 Heffalump exhausted without getting any passes at all
17:10 Heffalump so I need to tweak generators to stop that happening (or disable the tests)
17:13 kowey (sorry, not paying attention much, but I saw QC and generators)
17:13 kowey (and thought I should mention Lee Pike's been looking for people to play with his SmartCheck stuff)
17:13 kowey which I think does something clever about shrinking
17:16 Heffalump I don't feel like rewriting the tests right now, just fixing the "failures" :-)
17:21 iago is that happening too often? (i.e. arguments being exhausted with no test passed)
17:23 iago kowey, I don't think it's gonna help here
17:23 Heffalump iago: it's happening consistently in 4 tests
17:23 kowey ah apologies, like I said, not really paying proper attention
17:23 kowey just being a keyword-jumping-in guy, Not Helpful
17:24 iago Heffalump, could you point me to the module where those properties are defined, and the properties themselves?
17:28 Heffalump Darcs.Test.Patch.Arbitrary.PrimV1 - aPrim and aPrimPair
17:31 Heffalump mornfall: any idea what happened to the buildbot? I'm not sure if it's worth resurrecting given that it's a fair amount of work to actually keep it up, though.
17:33 iago Heffalump, those are the generators, not the properties. I presume you have a property P THEN Q such that aPrimPair is not generating any patch satisfying P
17:33 iago so I would like to know what Ps are those
17:36 Heffalump oh, sorry
17:36 Heffalump Darcs.Test.Patch has qc_V1P1
17:36 Heffalump which calls Prop.checkSubcommutes Prop.subcommutesNontrivialInverse "nontrivial commutes are correct"
17:36 Heffalump the Prop.* are defined in Darcs.Test.Patch.Properties.V1Set2
17:37 Heffalump and it gets exhausted on the speedyCommute case
17:53 raichoo joined #darcs
17:54 mornfall Heffalump: You mean buildbot or hydra? http://divine.fi.muni.cz/hyd​ra/build/385104/nixlog/1/raw
17:55 mornfall Heffalump: Dependencies missing, I'd say.
17:58 mornfall Heffalump: I vaguely recall offering to put release.nix into the darcs darcs repo so that it can be updated as needed.
18:00 Heffalump mornfall: I was wondering about http://buildbot.darcs.net - is that dead and the DNS alias should be removed?
18:00 Heffalump iago: turns out the exhausted things could _never_ work, it's not a distribution problem
18:00 mornfall Yes, it's thoroughly dead, the machine it was running on no longer exists.
18:01 Heffalump so actually I don't need the coverage reproduction code right now, though I would still like to document how to redo it inline if possible
18:02 Heffalump mornfall: what does release.nix look like? Sounds like a sensible idea though.
18:02 iago good to know, I was investigating myself
18:03 Heffalump iago: thanks, sorry for wasting your time
18:03 mornfall Heffalump: http://pastebin.dqd.cz/F4Ez/
18:03 Heffalump there are a couple of other arguments exhausted ones that I haven't looked at properly yet
18:03 mornfall Heffalump: (It has some stuff that could be dropped, presumably.)
18:04 mornfall Hm, and the 4TB volume ran out of space, too.
18:05 Heffalump how does one obtain the right sha256 for a new dependency?
18:06 Heffalump I'm surprised it doesn't have to explicitly list more package dependencies, is there some implicit set being picked up too?
18:06 iago no problem!
18:06 mornfall extensibleExceptions filepath hashedStorage haskeline_0_6_4_7 html HTTP
18:06 mornfall mmap mtl network parsec random regexCompat tar terminfo text vector
18:06 mornfall zlib dataOrdlist shellish
18:06 mornfall testFramework testFrameworkHunit testFrameworkQuickcheck2
18:06 mornfall cmdlib base16Bytestring cryptohash findBin
18:07 mornfall Heffalump: ^^ this is the list actually
18:07 mornfall Heffalump: The definitions are only for packages that are not in nixpkgs.
18:07 Heffalump oh, I see
18:08 Heffalump so we can just add things and if there's something completely fresh then we have to figure out its definition occasionally?
18:09 mornfall Heffalump: http://hydra.nixos.org/jobset/nixpkgs/trunk has a list of "what's available" (look under Jobs)
18:10 mornfall Heffalump: I think the sha256 is just sha256 of the tarball from hackage.
18:10 mornfall (I use nix-prefetch-url to get it.)
18:11 Heffalump ah, ok
18:11 mornfall You can use base16 from sha256sum.
18:11 mornfall (nix-prefetch-url uses base64 which is shorter...)
18:22 bos joined #darcs
18:23 bos hullo, darcs people. y u no build with modern ghc?
18:24 kowey uh-oh, darcs people are largely in Paris at the moment (1920) and may be off to dinner
18:24 kowey didn't we just add GHC 7.6 support?
18:24 kowey does modern mean more than that?
18:24 bos ah, it looks like darcs-fastconvert is the problem
18:24 bos i'm trying to create a fastimport file to slurp into git
18:25 bos and darcs-fastonvert has bitrotted dependencies
18:25 kowey ah yes, @tell owst bos points out darcs-fastconvert has bitrotted deps
18:25 kowey oh
18:25 bos does darcs natively support those fastimport files yet perchance?
18:25 kowey @tell owst darcs-fastcovert has bitrotted deps
18:25 lambdabot Consider it noted.
18:25 kowey not yet in the darcs client itself, if that's what you mean
18:26 bos yes
18:26 bos ok
18:42 Heffalump kowey: is the Thanks page deliberately not editable online?
18:42 kowey I think so
18:42 kowey but we could relax that
18:42 kowey it's a config thing in wiki
18:42 Heffalump I'm not fussed, I can update it offline easily enough
18:43 kowey I think it's inherited from the idea that we wanted to have static content (ie. stuff imported from the homepage [man, what a good idea that consolidation was, good work to the person who pushed through all the inertia, forget if it was gh_ or sm or somebody else])
18:45 bos joined #darcs
18:45 bos kowey: darcs 2.8.4 doesn't build on fedora 17
18:45 kowey Heffalump: that something we know about?
18:45 kowey (thanks, bos)
18:45 Heffalump is it an iconv problem?
18:45 bos yes
18:45 Heffalump install the development files for libcurl
18:46 bos libcurl?
18:47 Heffalump all I know about the problem at the moment is what's in this conversation: http://irclog.perlgeek.de/​darcs/2013-02-13#i_6446470
18:47 bos wow, that worked. how random.
18:49 gpiero1 joined #darcs
18:50 Heffalump I copied the relevant setup code from haskeline, so I'm a bit surprised this hasn't showed up lots there in the past.
18:50 Heffalump but I haven't debugged it or investigated yet
18:53 Heffalump kowey: didn't we do GSoC in 2008?
18:53 kowey don't think so
18:53 Heffalump but we did in 2010, 2011 and 2012, right?
18:54 Heffalump (just a bti surprised that it currently says 2007 and 2009 - I didnt' remember there being a gap)
19:01 gh__ ending day 1 of sprint
19:58 donri joined #darcs
20:13 lelit joined #darcs
22:18 Heffalump donri: mostly the fundamental algorithms
22:18 Heffalump laziness and haskell constant factors have probably played some part e.g. in the speed of simple operations like get etc.
22:19 Heffalump but the real problems came from merging and that's an algorithmic issue
22:19 Heffalump re the hub bridge stuff, that's the plan
22:19 donri cool
22:19 Heffalump we just need a working bridge, which we are actively working on
22:19 donri i wasn't sure the exact plan for hub was to make repos transparently double as git repos
22:19 Heffalump well, that's not concretely discussed, but it seems like an obvious thing to do if sm is amenable
22:20 Heffalump it's certainly my plan :-)
22:20 donri :)
22:21 Igloo Have the darcs folk seen this thread? http://www.haskell.org/pipermail/l​ibraries/2013-February/019440.html
22:21 Heffalump the bytestring thing?
22:21 Heffalump yeah
22:21 Igloo Well, the darcshub conversation that followed it, yes
22:21 Igloo OK
22:22 Heffalump these things happen. Until we have a convincng alternative story I see little point in intervening
22:22 Igloo Just though I'd mention it in case you wanted to discuss it while all together
22:23 Heffalump btw, have you made any progress with camp/the proofs lately?
22:23 Igloo no
22:26 owst joined #darcs
22:28 dixie aaa
22:28 dixie oops, mistake
22:36 iago joined #darcs
22:54 sm Heffalump: if you or anyone wants to make and keep some kind of darcshub git support working, I'd of course be interested
22:57 gpiero1 joined #darcs
23:04 sm stiff pro-git(hub) pressure mounting on haskell-libraries
23:06 Heffalump it's disheartening, but it doesn't really alter the fundamental position that darcs is a minority system that some people still find useful: I work on darcs because I like it, and while it would be nice to have the network effects of being widely used, such is life.
23:06 Heffalump and of course a working bridge will be very useful
23:13 sm as a community we seem to be maxed out just maintaining what we have, so I take all bigger plans with a grain of salt
23:20 Heffalump bigger plans like what?
23:20 sm like anything other than "maintaining what we have", basically
23:21 sm darcs bridge, eg
23:22 Heffalump rebase, patch index, hub are all new things we've done recently...
23:22 sm are rebase and patch index in an official release ?
23:22 Heffalump no, but there's little doubt they will be
23:23 sm and do they need more work ? (yes)
23:23 sm I take your point and don't mean to sound negative
23:23 Heffalump yes, but we could release them right now and it'd be ok, just a bit unpolished
23:24 sm we make progress in interesting areas but we seem to regress in others and find out about it a year later
23:24 Heffalump if I thought it would say take a year to polish them and release, I would personally be ok with shipping what we have today
23:29 iago Heffalump, well, rebase was "almost ready" since I attended the Sprint 2 years ago
23:29 Heffalump true, but not merged, and now it is
23:30 Heffalump same for patch index, that's been around for ages, but it's in HEAD now
23:30 iago well, I suppose it takes more time
23:31 iago IMO darcs-bridge will be top priority :D
23:31 iago I basically moved to mercurial because I don't trust any darcs hosting
23:31 iago they appear, promise to last forever, and a few month later... dead
23:32 sm wait, which ones died ?
23:32 iago ok, I'm being a bit dramatic so virtually unmaintained = dead
23:32 sm gotcha :)
23:33 sm I mean: I get you
23:36 Heffalump the thing is, we're hosting a *D*VCS
23:36 Heffalump it should be so trivial to make hosting redundant, but people don't seem to
23:36 Heffalump and even darcshub isn't really because issues etc aren't in the repo
23:38 sm Hmm ? hosting isn't redundant, if we're talking about a featureful central hub - that's a valuable (these days, essential) service
23:38 Heffalump what do you mean by featureful and central?
23:39 sm Eg github. Back in the day, sourceforge
23:39 sm these provide many benefits above the level of just hosting files
23:39 Heffalump github isn't redundant because it's (a) proprietary code and (b) it hosts data that isn't trivially cloned elsewhere (issue tracker etc)
23:39 Heffalump if it weren't for that, you could spin one up elsewhere, clone, and you're away
23:39 sm I may be misunderstanding you
23:41 Heffalump the features are code+data
23:42 iago well, not exclusively
23:42 Heffalump the hosting itself is running a server
23:42 iago I'm making good use of bitbucket teams
23:42 iago I don't have to configure a server and access control myself
23:42 Heffalump I don't know exactly what thsoe are, but access permissions are just data
23:43 Heffalump I'm not suggesting you should have to run your own server, but that hosting should be redudnant in that it can easily be moved from place to place.
23:45 iago yep, of course, everything is data
23:45 iago the problem is put it to work
23:46 iago as sm, I'm not sure if I'm understanding your point
23:46 Heffalump why is everyone engaged in building non-distributed infrastructure around a distributed vcs?
23:47 Heffalump where infrastructure = issue trackers, access control, etc.
23:47 sm yeah! alexsuraci, why'd you do that ? :)
23:48 iago well, bitbucket's wiki is just a repo that you can clone etc
23:48 iago I don't know about the issue tracker
23:49 sm Heffalump: I think because usually some of those things don't fit well in a VCS generally, or in darcs specifically. Eg because of high volume of change, or because of security implications..
23:53 Heffalump ok, so it's not trivial as I claimed above. But it ought to be possible. e.g. access control by openid rather than username/password.

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