Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2014-07-29

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

All times shown according to UTC.

Time Nick Message
00:23 mizu_no_oto joined #darcs
01:16 Rastus_Vernon joined #darcs
02:00 edwardk joined #darcs
02:23 mizu_no_oto joined #darcs
04:02 dolio joined #darcs
09:10 mdiaz joined #darcs
09:28 edwardk joined #darcs
11:00 mizu_no_oto joined #darcs
11:19 mizu_no_oto joined #darcs
11:42 edwardk joined #darcs
12:29 edwardk joined #darcs
13:07 mizu_no_oto joined #darcs
14:18 lelf joined #darcs
14:18 Heffalump joined #darcs
14:20 mizu_no_oto joined #darcs
14:26 edwardk joined #darcs
14:46 joe9 joined #darcs
14:47 joe9 I am installing my cabal packages and do not have darcs installed yet. Is there any way I can clone the darcs repo using git? I cannot do "darcs get http://darcs.net --lazy" as I do not have darcs installed and want to install it from source.
14:48 joe9 probably a darcs repo cloned on github, etc..
14:51 joe9 or, if I could rsync/wget to clone that darcs repo.
15:13 lelf joined #darcs
15:32 jochu joined #darcs
15:48 thorkilnaur joined #darcs
16:07 c74d joined #darcs
16:10 mornfall joe9: cabal install darcs will install from source
16:12 joe9 mornfall: but, it is installing from 2.8.4 and that has a few breakages on my system. I want to install from the source repo.
16:13 joe9 I want to download the source repo, so, I can do cabal install there.
16:13 mornfall I don't think you can do that easily.
16:13 mornfall There are source tarballs somewhere, though.
16:14 mornfall You can also tell cabal to install a different version of darcs, obviously.
16:14 mornfall cabal install darcs-2.8.1 say, or whatever version you want
16:15 joe9 mornfall: I want to install from the latest source repo as the versioned darcs install is trying to pull in some old libraries.
16:16 mornfall tough luck then, I don't have a good solution
16:16 mornfall I was trying to push for read-only git mirrors on hub.darcs.net, but the maintainers disagree
16:17 joe9 mornfall: sorry for the bother. Yes, that is a good idea to have read-only mirror that can be cloned with rsync or git..
16:17 mornfall (it's not hard to do, but they don't want to do it on principle)
16:18 mornfall rsync is problematic as it uses a special protocol = sysadmin headaches
16:18 mornfall but git clone works over http
16:22 joe9 mornfall: this is what happens with "cabal install darcs": http://codepad.org/WoasGLda
16:23 mornfall oh dear :-)
16:24 mornfall I'm about to run into a mess, upgrading to GHC 7.8.3...
16:24 joe9 I just upgrade to 7.8.3 as some libraries were starting to break.
16:24 joe9 s/upgrade/upgraded/
16:24 joe9 and I cannot get darcs to install now.
16:25 joe9 this is my world http://codepad.org/Gf5YToOg
16:25 mornfall yeah, that's kind of expected
16:25 joe9 are there any solutions for it?
16:25 mornfall I wouldn't say darcs is exactly maintained these days
16:26 mornfall sort of limping behind
16:27 mornfall luckily I can just use stable darcs built by GHC 7.6.3...
16:28 mornfall for the forseeable future anyway
16:50 stepkut joined #darcs
16:59 Heffalump joe9: you need to use GHC 7.6 with darcs 2.8
17:06 sm just install GHC 7.6 in addition to what you have and cabal install -w ghc-7.6.3 darcs
17:17 sm mornfall, are you referring to the conversation on http://irclog.perlgeek.de/darcs/2014-04-03, or some other ?
17:18 diginet joined #darcs
17:18 lelf joe9: use the binary of darcs, then download&install the version you want
17:26 joe9 lelf: good idea. thanks a lot.
17:29 joe9 lelf: that is working like a charm. Thanks again.
17:32 amgarchIn9 joined #darcs
17:33 mornfall sm: I think Heffalump was at some point concerned about user expectations -- say being able to push into the git repo or branch off and somehow co-operate, don't really remember and can't find it quickly
17:34 mornfall sm: and the fact that you may want to re-generate the git repos on hub from scratch at some point, making git pull impossible... I don't think any of this is a big deal, but well
17:34 sm never mind.. if someone wants it enough, I guess it'll get done
17:34 Heffalump I didn't want to implement it because the current git export code loses information, but gh has done it
17:34 Heffalump it'd still need to be integrated into hub.darcs.net though
17:36 pseudolio Is the only problem with the current release and 7.8 that NOINLINE bug?
17:37 Heffalump pseudolio: I don't think it builds either
17:38 Heffalump I'm relucantantly concluding that we'll have to backport the necessary fixes :-(
17:38 pseudolio Oh.
17:38 Heffalump also, the iconv/libcurl detection problem is a pretty big annoyance
17:38 Heffalump that's again fixed in HEAD but not in 2.8 (I think)
17:39 pseudolio And head isn't close to a release?
17:40 Heffalump doesn't feel that way :-(
17:40 sm :(
17:40 diginet joined #darcs
17:41 Heffalump it's the vicious circle of having had too much stuff land without a release so it'll take qsuscs uite a bit of stabilisng
17:41 Heffalump sqsuscs uite/quite/
17:41 sm maybe it's better to just throw it over the wall
17:42 sm will there ever be a better time for a release ?
17:42 sm release and go on a long vacation :)
17:43 sm wait, where is my mission critical web scale enterprise hat..
17:43 Heffalump I've just come back from my vacation...
17:43 sm damn
17:44 sm it was good I trust
17:49 sm what are the essential things required before release ? how about freezing today's trunk and we all build and use it exclusively for a week
17:52 edwardk joined #darcs
17:56 Heffalump rebase is still missing some stuff like a way to see the changes in the rebase patch (mea culpa), right now we have the GSoC work landing that has some gaps in it, plus various UI tweaks need confirming
18:01 sm maybe we should pick a date then, and throw it over the wall then
18:03 Heffalump I guess what we really need is someone who will actually run the release.
18:03 sm a must-release-by date, might help us pull together
18:03 sm ack are you stepping down ? :)
18:03 Heffalump I'm very happy to be replaced if there's a volunteer...
18:04 sm who wants to taste the raw power of being darcs release manager ?
18:04 edwardk joined #darcs
18:05 sm Heffalump, what are the minimum tasks that person will need to do ?
18:07 Heffalump http://darcs.net/Development/ReleaseManagement documents the technical administrative side of things
18:07 Heffalump apart from that it's really just identifying what needs to be done,chivvying people along, etc.
18:13 sm so the release manager can basically decide what they need to do, as long as it results in a release some time soon
18:13 sm raw power, see ?
18:14 Heffalump I guess I'd hope the release is usable and takes darcs forward in some broad sense, but yes, basically :-)
18:16 amgarchIn9 joined #darcs
18:17 sm could put out the call on the mail list I guess
18:18 sm there's a bunch of folks out there right now who could do a great job
18:19 sm well, at least a good-enough job, which would get us past this black hole, which would be great
18:21 sm it's sad to see the stream of folks passing through here unable to benefit from all the hard work put into HEAD
18:34 gh joined #darcs
18:35 gh__ hi
18:35 gh__ hi
18:41 gh__ I know next to nothing about release management, but I wanted to propose myself and aim at a release 2.10 after the end of the GSoC projects and before the end of the year
18:42 sm Yay gh__ !
18:47 edwardk joined #darcs
18:48 sm ok with you Heffalump, all ?
18:51 gh__ there's no hurry I guess, nothing is going to happen before the end of the GSoC projects
18:52 sm true, though it'd be cool to have a plan with dates to tell people
18:54 gh__ then I guess it will be 20th of October :D
18:54 gh__ Darcs' 12th birthday
18:55 sm nice
18:57 Heffalump :-)
18:57 Heffalump I don't know anything about release management either...
18:59 gh__ I suspect the hardest part to be "get bugs fixed on all platforms"
19:00 gh__ which is usually implemented as "fix all bugs yourself", if I recall some other RM's experience :D
19:01 Heffalump right :-)
19:01 Heffalump I particularly remember Unicod.
19:01 Heffalump e.
19:02 sm a RM shouldn't get into that
19:03 kowey joined #darcs
19:03 kowey you around, sm?
19:03 Heffalump fixed release date, release free of major bugs, hands-off RM: pick 2
19:03 Heffalump sm: quick, hide, kowey's here
19:03 kowey :-)
19:04 kowey sm: I've got a lock file on http://hub.darcs.net/kowey/kbgen from when I interrupted a slow darcs push
19:04 kowey I wonder if there needs to be some hidden repo cleanup only-if-you-know-what-you're-doing button
19:05 sm aieeeee
19:05 sm yes there should
19:05 sm lacking that, let me go fix that. I have cron scripts which are supposed to
19:06 kowey the sort of button that says “don't push me”
19:06 kowey oh, maybe it's already been cleared away
19:07 gh__ an interesting kind of buttons
19:07 * kowey sighs wistfully at his non-darcsing
19:08 sm kowey: yes it seems to have been cleared
19:08 kowey I think the main change is that once I started freelancing (and then not), I'd gotten rid of the idea that I ought to work on kowey stuff one day a week
19:08 kowey because you know... kinda hard to charge for that
19:08 kowey and now I'm treating my academic work like freelance work
19:08 kowey sorry, don't mean to be making excuses, just something on my mind :-)
19:09 kowey thanks!
19:09 mornfall ha, it's kowey
19:09 kowey I think this situation would change if I ever found myself working a full-time long-term academic job
19:10 kowey where doing kowey-stuff is quite clearly an long term investment and justifiable cos academics gotta academic
19:13 * sm fixes the lock clearing job to run hourly instead of nightly
19:14 sm it clears locks more than 60 minutes old, so that should be ok
19:14 kowey so in my case,
19:14 kowey and I hope I'm not abusing hub.darcs.net here
19:14 kowey I had patches which for some reason or another seemed to involve uploading 20MB of stuff over ssh
19:15 kowey maybe had  a tarball in it or osmething
19:15 kowey and I think at some point the upload rate slowed to a trickle
19:15 kowey but dunno what the cause of that trickle might be
19:15 sm ah.. your isp maybe ?
19:15 sm mine does that (within seconds)
19:15 kowey that's why internet stuff is so mysterious to me
19:16 kowey I partly don't just wrap my head round all the invisibles, eg. isp
19:16 kowey so i dunno if that hourly lock clearing strategy would make sense there
19:16 sm if something takes more than an hour, hub just won't support it
19:16 kowey you might have an hourly cleanup with longer cleanup ah-should-be-fine
19:17 kowey that probably makes sense
19:17 sm even now, you could easily DDOS it
19:17 kowey btw, shipping a 200ish patch repo like 10 patches at a time is a bit of a pain
19:17 kowey but sort of hard to imagine this being a priority for darcs
19:18 * sm checks koweys disk usage.. #2 at 335M, just behind me
19:18 kowey eep!
19:18 kowey please do let me know if I start to be a pain
19:18 sm will do
19:18 kowey maybe actually I'm being dumb and I should just put this thing on erickow.com
19:19 kowey hmm
19:19 kowey yes
19:20 sm as I've come to expect :(, darcs check on your kbgen repo reports hash mismatches/bad index
19:20 kowey sm: sheesh, I was being dumb, sorry will delete that repo
19:20 kowey oh
19:20 sm I don't know if you see that locally
19:20 kowey should I keep that around for your forensics?
19:21 sm nah, I don't have time - thanks
19:21 kowey ok, nuking
19:21 kowey sorry for all the trouble!
19:21 sm no trouble
19:21 sm if anybody gets too large I'll let them know
19:22 sm incidentally patch-tag is shutting down on monday
19:22 mornfall hmm, right
19:23 sm hub might or might not increase in size.. not much change since he announced that
19:24 amgarchIn9 joined #darcs
19:24 kowey I wonder if people are being like me and just going “ah, I better migrate my repos, do I have any repos? (five days later... oh shoot, too late)
19:26 sm (to their secret relief)
19:26 kowey heh
19:27 carter o/ kowey
19:27 sm ack I better go see you kowey, all o/
19:27 kowey see you around, thanks!
19:27 kowey heya carter!
19:27 carter kowey: how've you been?
19:27 carter i've been on a very long running yak shave to have nice sparse matrices
19:28 carter (nlp uses those right?)
19:28 kowey could be worse
19:28 kowey depends on what goes in them
19:29 kowey if it's a bunch of scalar not-necessarily-numerical values then I could see this being used to hold a bunch of features for some machine-learning stuff
19:29 kowey with non-presence being some default/false value for any given feature
19:29 * kowey has been doing Python and Java lately and getting anxious about deHaskelling
19:30 kowey I hope I shall find my way back to the Light
19:30 carter kowey: its parametic
19:30 carter i use (M)Vector as my storage layer
19:30 carter and can also do Maybe/Either things
19:30 kowey oh that sounds handy
19:31 carter right now i'm finally getting to writing a bunch of examples and such
19:31 carter + plus juggling consuling
19:31 carter kowey: a key thing, is the api for sparse and dense matrices is the same
19:31 carter i have some Dense Only operations, but they're less generic (by their very nature)
19:32 kowey you productive people and your big-picture long-view investments
19:32 carter i want tools i'll be happy using 5-10 years from now
19:32 kowey I see that as Lower Learning Curve
19:32 carter and next week
19:32 carter oh?
19:32 kowey and so think I'm happy with that key thing
19:32 carter whichkey thing?
19:32 kowey same API for dense/sparse
19:32 carter oh yeah
19:33 carter even better, i've set it up so you can add new (n-dim) array formats
19:33 carter in user land
19:33 kowey unless of course it's one these super-abstract I-have-no-idea-what-alll-these-type-vars-are things
19:33 carter oh
19:33 carter its pretty gortty
19:33 carter https://github.com/wellposed/numerical/blob/master/src/Numerical/Array/Mutable.hs#L326-L455
19:34 carter will be the most generic api
19:34 carter end users could use
19:34 carter if they were to use the prmiiops directly
19:34 * kowey has to tend to his lentil-enhanced rice
19:34 carter ok
19:34 carter enjoy
19:34 carter and come lurk by #numerical-haskell and i'll show you more sometime
19:35 kowey :-)
19:35 carter hopefully soon i'll make it safe for hackage
19:35 carter ttyl
19:35 kowey thanks! I don't think I'll manage to do so, but I'm glad to see you're working to make things better for us all
19:36 kowey one day if Haskell has a nice response to scipy, nltk, etc...
19:36 carter oh
19:36 kowey and just becomes the path of least resistance
19:36 kowey (right now I have to Python so that random non-programmer PhD students can just whip stuff up with my code)
19:36 carter once i cut alpha, give me ~ 6 months and i'll have numpy/scipy beat
19:37 carter i suppose you could foreign export stuff
19:37 mornfall random non-programmer PhD students
19:37 mornfall what a nightmare
19:37 carter ew
19:37 kowey budding computational linguists
19:38 carter CLP
19:38 carter not just constraint logic program
19:38 mornfall I wouldn't say a thing if they were physicists, I guess...
19:39 carter all the theoretical physics folks i know an code
19:39 carter and cope with fortran
19:39 kowey I should really stop wandering in and out of the kitchen
19:39 mornfall well, some cope *only* with fortran :P
19:39 carter mornfall: yeah that too
19:39 kowey I do like the idea of not-strictly-programmer types being empowered though
19:39 carter i suspect once i cut alpha, i'll have 2-5 phsycis users in the first 2 weeks
19:39 carter sparse matrices are all they care about
19:41 mornfall if you have fusion, anything goes... :-)
19:41 mornfall no fusion no dice :P
19:43 mornfall also, fftw integration ;-)
19:46 carter fusion?
19:46 carter fusion is a massive troll on FP array computation
19:46 carter in that its only good for a few classes of algorithms
19:47 carter and is a grand research challenge
19:47 carter fftw integeration ? we've got that on hackage
19:48 mornfall yeah, but if you use a different array type you don't want to be converting stuff by hand, do you?
19:50 carter who said it'd be different
19:50 carter fftw on hackage uses storable arrays
19:50 carter and ii've support for storable vectors and pretty much any vector instance
19:51 carter *literally
19:51 carter well, i wrap em up a teeny bit
19:51 carter but easy to re/unwrap
19:52 mornfall :-) ok ...
19:53 mornfall too bad I suck at linear algebra
19:54 carter its easy
19:54 carter just start with dot product
19:54 carter and then think of all geometricly
19:54 carter mornfall: many folks learn lin alg by writing that simple 3d game they've always wanted
19:58 mornfall well, my geometric intuition breaks down quickly with increasing dimensions
20:00 carter mornfall: most intersting lin alg stuff boils down to "make a bunch of vectors orthogonal"
20:00 carter which is easier and easier in higher dim :)
20:10 edwardk joined #darcs
21:29 alegadea joined #darcs
21:52 edwardk joined #darcs
22:25 mdiaz joined #darcs
22:48 mizu_no_oto joined #darcs
23:16 edwardk_ joined #darcs
23:43 dolio joined #darcs
23:50 amgarchIn9 joined #darcs

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