Camelia, the Perl 6 bug

IRC log for #darcs, 2009-08-13

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

All times shown according to UTC.

Time Nick Message
00:38 Igloo Does anyone know where the Windows equivalent of ~/.darcs/defaults is?
01:01 darcscommitbot joined #darcs
01:22 twb joined #darcs
02:18 twb joined #darcs
03:14 sm joined #darcs
03:25 gwern @ask kowey how are you editing gitit? I notice no email seems to be showing up for your patches, just 'kowey <>'
03:25 lambdabot Consider it noted.
03:25 twb I assumed he was using the web UI
04:50 sm joined #darcs
04:56 copumpkin joined #darcs
05:56 gour joined #darcs
06:36 lelit joined #darcs
07:01 darcscommitbot joined #darcs
07:03 josephholsten joined #darcs
07:34 intripoon joined #darcs
08:16 intripoon joined #darcs
08:16 ertai joined #darcs
08:16 kolmodin joined #darcs
08:16 Riastradh joined #darcs
08:16 nwf joined #darcs
08:18 Beelsebob joined #darcs
08:29 lelit hi all
08:30 lelit yesterday I got a strange behaviour with darcs mv, one of those cases where darcs reorder the hunks in a surprising way
08:31 lelit here is the recipe: http://pastebin.com/m462df8c3
08:31 lelit I was restructuring one repo, moving stuff around; in particular, I need to "promote" a "sub-sub" folder up to the top level, and then add a few other files to it
08:33 lelit why the two different ways darcs handle the move of files in the renamed folder?
08:35 lelit all in all, it works, but these are the cases where tailor (and trac-darcs) needs to be tricky
08:37 twb lelit: if it's reproducible, I suggest reporting to bugs@darcs.net
08:37 lelit ok
09:21 lelit twb, sent, with an even more surprising finding...
09:25 lelit while cleaning up the script for the mail, suddenly it stopped to be surprising... so the mail comes with *two* scripts, almost identical, that at one point creates two different named subdirs: one shows the behaviour I said, the other not...
09:33 lelit a doubt: is darcs sorting the hunks in alphabetic order??
09:34 lelit later
09:34 twb IME darcs optimize --reorder will use date ordering
09:35 twb Or do you mean ordering within a patch?
10:07 lelit` joined #darcs
10:08 lelit` twb, no, I meant: the two different behaviours of the scripts I sent, I notice that the order of the moves in "darcs changes --last=1" is alphabetical...
10:09 lelit` so, in one example, "another/file -> folder/subfolder/file" comes first, while in the other case "third/file -> promoted/" comes last
11:40 gh_ joined #darcs
12:50 simplesimon2k joined #darcs
13:01 darcscommitbot joined #darcs
13:50 Beelsebob1 joined #darcs
14:04 simplesimon2k left #darcs
14:09 kpreid_ joined #darcs
14:52 gbeshers_ joined #darcs
14:57 gbeshers_ joined #darcs
15:06 gbeshers_ joined #darcs
15:15 jbe joined #darcs
15:21 sm joined #darcs
15:22 copumpkin joined #darcs
16:01 jbe` joined #darcs
16:05 jbe`` joined #darcs
16:18 gour joined #darcs
16:33 kowey joined #darcs
16:45 mornfall Re kowey.
16:45 kowey hey mornfall
16:45 lambdabot kowey: You have 1 new message. '/msg lambdabot @messages' to read it.
16:46 kowey gwern: I tend to edit gitit through the web interface
16:46 kowey mainly because I haven't seem to have memorised the rules for RST yet, so I tend to rely on preview
16:56 * mornfall is exploring GADTs, witness types and type families...
16:57 * copumpkin prefers to leave no witnesses
17:01 arjanb joined #darcs
17:20 kowey hi sm... quick question about roundup
17:20 sm shoot
17:20 kowey should I worry about passwords when I merge users?
17:20 kowey I guess I'm vaguely concerned about leaving somebody without the ability to log in
17:21 kowey I guess the answer is to merge away and then let the pw reset mechanism take care of the rest (with an explicit notification)
17:21 sm kowey: sounds good
17:23 kowey OK, thanks! I guess I'll collect messages and send notifications
17:23 kowey *collect emails
17:23 sm no point spending your valuable time avoiding every possible glitch, if it's not losing any data
18:15 dino- joined #darcs
18:18 dino- After lot of trouble, I was able to get darcs to build using cabal-install and the Haskell Platform on WinXP. But the binary always immediately exits with no output and an exit code of 128.
18:19 dino- The trouble I allude to was all libcurl related.
18:19 dino- Does this sound familiar to anyone?
18:52 fophillips joined #darcs
19:01 darcscommitbot joined #darcs
19:41 waern joined #darcs
19:52 kowey joined #darcs
19:53 kowey hi dino-, I noticed in the IRC logs that you were having Windows trouble
19:54 kowey it might be worth sending an email to bugs@darcs.net with full logs
19:54 kowey our Windows expert is on holiday for the month, but maybe the rest of us can try and help
19:55 kowey perhaps one thing to try: how about installing without libcurl support (using HTTP instead)?
20:12 mornfall Hm. This whole type proof business is a little unwieldy.
20:12 mornfall Brb.
20:25 mornfall Back.
20:26 mornfall Hm, admittedly, I'm not really in the right state to be writing this final report. Bah. But I really should get it done sooner rather than later.
20:32 ronny joined #darcs
20:32 ronny hi
20:33 ronny whats a good way to add new patches to a repo without touching the workdir beforehand
20:34 mornfall What kind of patch do you have in mind?
20:35 ronny im not exactly sure what kind of patch i would need
20:35 ronny im making an api to create "commits" for different vcs's without touching the workdir
20:36 mornfall Dunno. You can create a synthetic patch bundle and darcs apply it.
20:37 mornfall But that will affect workdir in the end anyway.
20:37 mornfall Hmm.
20:37 ronny affecting it in the end is ok
20:39 ronny i'll go for figuring how to make those patch bundles later
20:55 mornfall kowey: Do you have a moment? Any advice about http://pastebin.com/m5dd2202e ?
20:56 mornfall It's clearly not complete, but I'm a little lost as to what to add (and maybe what to remove).
20:57 kowey mornfall: I'll braindump comments as I read
20:58 kowey aye... that's the Darcs attitude: we will have to live with it for a long time, and mistakes now would be expensive later
20:59 mornfall Well, it doesn't apply equally to all things. But repoformats, that's the bane of all the VCSes around.
20:59 kowey i'd clarify the "bunch of work that needs to be done"
21:00 kowey also be clearer about what people reading this (final) report can do to help, for example: try darcs-benchmark
21:01 kowey maybe a sort of mini-roadmap to summarise things - for example, get hashed-storage to state where we're happy to merge into darcs 2.4 and then work on packs
21:01 mornfall I'm not very clear on the roadmap myself.
21:01 kowey I'd also tack on one last paragraph for the future work with a broader perspective, not just what you feel you need to work on, but what others could build off your work
21:02 kowey (sorry that's just me trying to apply researcher propaganda that they've been dumping on us students outside the academic field)
21:03 kowey maybe in the beginning it would be useful to try putting yourself in the perspective of Darcs user, hasn't really been following this work, just excitedly knows it's supposed to make darcs faster
21:04 kowey so concretely: what's going to happen? OK whatsnew on hashed repos is now fast; when we manage to merge in the darcs-hs stuff, similar improvements; in the future as you complete the work on packs, fetching repos and pulling patches will go faster because we're no longer trying to fetch a ton of tiny files
21:05 kowey (I'm not saying that you can implement all of these ideas for the write-up, but I'm hoping you can cherry-pick the thoughts that you think will have the most impact)
21:05 mornfall Hm. What is the deadline on this, anyway? I know, yesterday was already late, but...
21:06 mornfall To make some positive impact, I need to believe that things are positive, and it doesn't seem to be working right now.
21:07 kowey well, it's not about sales here, just about knowing which perspective to tell things from
21:07 mornfall I have to force myself into writing a little optimistically, and it tends to show.
21:07 kowey never force it
21:07 kowey it's not so much a question of tone, but of audience
21:07 kowey when talking to audience X, emphasise Y
21:08 kowey see what I'm saying? it's really just a specialisation of "increase signal to noise ratio"
21:11 kowey "(that's for another goal)." I think you may be able to cut
21:12 mornfall I don't even know what is the audience, I guess.
21:12 mornfall It felt that I should justify what I have done, for some reason.
21:12 mornfall But it may not be the right place to do that.
21:12 kowey right, focus hard on how to make what you write as useful as possible to the person that's reading it
21:13 kowey one example is this... whenever I go to talks, and I ask a question
21:13 kowey I always feel so stupid that I can't stop myself from saying crap like "oh, this may be a stupid question, but..."
21:13 kowey and I should NOT do that, because that crap is just a manifestation of my own insecurity
21:14 kowey and for the people listening to my question, and the speaker who took the trouble to deliver his talk, that's just a bunch of random noise
21:14 mornfall Hm. : - ) Yes, it seems that I feel guilty for not doing enough on the project.
21:14 kowey so you want to be merciless about going through the document and cutting out everything that's going to be noise
21:14 sm +1 for know your audience and give them the clear essentials they are interested in. I love writers who do that
21:15 * kowey is only talking in theory :-)
21:15 * sm aspires to do it
21:16 kowey mornfall: I think maybe another day to work on it should be good
21:17 kowey I hope there's stuff you can use in this dump
21:17 mornfall kowey: Oh, darcs-benchmark. What is the status of maybench? It comes with an identically-named binary, and I haven't checked what it does yet.
21:17 kowey forget about maybench basically
21:17 kowey I may work on it in September for some not-darcs stuff
21:18 kowey darcs-benchmark in maybench tries to generate repos of size n, n+1... etc and then do performance stuff on them, if I recall correctly
21:18 kowey so it's something to do think about for the future of darcs-benchmark darcs-benchmark ;-)
21:18 mornfall Interesting. Is that something that maybench in general does?
21:18 kowey right now, not really
21:18 kowey there's tiny bits and pieces of library for darcs-benchmark
21:19 kowey the idea is that in the future, it would make it easier for me, say, to benchmark GenI (some other software I work on)
21:19 mornfall Ok. I'll just skim over it. Would it make sense to upload a maybench version that renames the executable, so the packages don't overwrite each other?
21:19 kowey and for other people to do timing tests on their software, but I don't really know how to design this stuff
21:19 kowey that's a good thought... I can get on that
21:20 mornfall I do a lot of benchmarking for research -- but we also have a lot of dedicated hardware to do it.
21:21 kowey there seems to be a lot of common infrastructure which could be useful
21:21 mornfall And these benchmarks easily take a week to complete.
21:21 kowey one thing we do, for example is benchmark and generate a latex report with graphs and stuff
21:21 kowey and stuff like "run this thing on this batch of inputs; collect the outputs, etc"
21:22 mornfall Yes. The reporting code in darcs-benchmark is virtually non-existent. And persistent storage for the measured data.
21:22 kowey I guess there's no way to get around a custom harness, but if we could minimise the duplication, it'd be good
21:22 kowey maybench, incidentally, runs on a commit-bit-for-all model, unless I misremember
21:23 mornfall The approach I take with our research model checker is that the tool itself reports timing and memory use data in a common format (just a name: value list) and I store output of each run in a file.
21:23 mornfall Then I have reports that extract various bits of that data.
21:23 mornfall It *may* work for darcs, if we add enough reporting to it.
21:23 mornfall darcs foo --performance-report
21:24 mornfall wall-time: 2.3s
21:24 mornfall user-time: 2.2s
21:24 mornfall memory-used: 12M
21:24 mornfall etc.
21:26 mornfall I'm wondering if we can extract GC data from the runtime somehow.
21:27 kowey you mean by parsing +RTS -sstderr output?
21:30 mornfall No, I meant having darcs itself, by some sort of introspection, extract this data from the GC, and report it itself.
21:33 mornfall We already chop +RTS -sfoo output with regexps, but I'm not very happy about that.
21:38 kowey i've pushed patches to maybench renaming the executable to darcs-benchmark-maybench
21:40 kowey and now I'm going to head to bed... good night!
21:40 mornfall Thanks.
21:40 mornfall Yeah, me too.
21:40 mornfall Goodnight.
21:58 copumpkin joined #darcs
22:40 darcscommitbot 11 Aug 14:37 - Remove optimize --checkpoint cruft. (Eric Kow )
22:40 darcscommitbot 09 Aug 16:30 - Remove tabs from src/Exec.hs (Reinier Lamers )
22:46 ronny oh fun
22:47 ronny i hope you guys get darcs to the point where i'll consider switching to it
22:49 sm ronny: what must we do ?
22:51 ronny sm: much more speed, some kind of api or protocol to easyly talk to the repo content (preferably from python), something to emulate git/mercurial style branches
22:52 ronny currently im using hg and tend to have rather branchy histories
22:52 ronny i dont like hacks like rebase
22:52 ronny (ie evil things git invents cause its just a stupid dag of snapshoots at its heart)
22:54 ronny basically as soon as i can write a anyvc backend for darcs and it will be speedy im allmost happy
22:54 sm max battcher has a python api for darcs repos
22:54 sm just fyi
22:55 ronny you mean that thing for darcsweb?
22:55 sm no
22:55 ronny oh, ok
22:55 ronny got an url for me?
22:55 sm but maybe yes. His project is darcsforge I believe
22:57 ronny well, lets take a look if it has anything reasonable
22:59 ronny sm: btw, any reason why the controll dir is _darcs instead of .darcs?
22:59 sm ronny: yes, windows compatibility
22:59 sm a pity, isn't it
23:00 ronny hg doesnt do that, works fine
23:00 ronny so why?
23:02 ronny hmm, ok, the darcsforge code for interfacing darcs isnt really nice :(
23:02 sm perhaps not on older windows versions
23:03 ronny sm: hmm, maybe win9x is fail

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