Camelia, the Perl 6 bug

IRC log for #darcs, 2011-04-19

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

All times shown according to UTC.

Time Nick Message
00:08 twb joined #darcs
00:25 balor left #darcs
00:32 kowey bah too tired to finish this post
00:33 owst Too tired to finish reading the patch-theory pdf... soldiering on though :)
00:34 kowey points were (i) consistency is a usabilty issue (ii) darcs is imperfect (iii) patch theory is simple at heart (iv) something about performance
00:34 owst kowey: cool.
00:34 kowey left #darcs
00:55 sm catching up with latest http://wiki.darcs.net/Development/GettingStarted
00:56 sm "screened" sounds exactly like "reviewed" to me, and I have seen other projects (eg squeak) use "inbox" for this purpose, fwiw
00:57 sm also the nice overview table is undermined by the following paragraph talking about "stable" and "unstable". I won't mess with it myself
00:58 twb "moderated"?
00:58 twb (I haven't been following the conversation.)
01:01 sm also I'm unsure what is "trunk", familiar from other projects. I guess it's "reviewed"
01:02 sm or not. Maybe there isn't a single trunk
01:06 lispy left #darcs
01:27 * owst has finally finished Igloo's patch-theory doc, can't say I fully absorbed it all, but it's whet my appetite :)
01:55 owst left #darcs
02:54 intripoon joined #darcs
02:58 intripoon_ left #darcs
03:36 dleverton left #darcs
03:36 shachaf left #darcs
03:37 jeffz joined #darcs
03:37 shachaf joined #darcs
03:37 dleverton joined #darcs
03:44 sm left #darcs
03:52 sm joined #darcs
03:53 sm left #darcs
06:17 roconnor awww darn.  I was in Paris April 1 - April 3.
07:11 lelit joined #darcs
07:13 shenshei joined #darcs
07:14 lispy_ left #darcs
07:22 JaffaCake joined #darcs
07:22 JaffaCake1 left #darcs
07:23 roconnor left #darcs
07:39 Weltraumschaf joined #darcs
07:42 jeltsch joined #darcs
08:04 jeltsch left #darcs
08:25 bsrk2 joined #darcs
08:32 bsrk2 left #darcs
08:35 jeltsch joined #darcs
09:11 Weltraumschaf left #darcs
09:11 Weltraumschaf joined #darcs
09:33 jonkri joined #darcs
09:40 amgarching joined #darcs
09:43 amgarching Hi, I was testing darcs-fasconvert for interoprability and noted that some tags are converted to git tags, some are converted to empty patches with TAG as a prefix. Is there rule? Am I doing something wrong. The repo is old. Started in 2005 with darcs 1, now most of staff is done by darcs 2.3.
09:47 mornfall You should get WARNINGs for the empty patches.
09:47 mornfall Try optimize --reorder before converting.
09:57 amgarching To get an idea, if I pull an old tag VERYOLD from a stale repo ontop of a recent repo with thousand new patches that are tagged by VERYOLD. What is darcs-fastconvert supposed to do with this tag? The fact it is on top does not mean it is tagging the recent snapshot (not in darcs).
09:58 amgarching thousand new patches that are *not* tagged by VERYOLD
10:03 amgarching tried optimize --reorder, still a few old tags appear in output of "git tag -l" others appear in "git log | grep TAG"
10:11 jeffz left #darcs
10:17 jonkri left #darcs
10:41 ManateeLazyCat joined #darcs
10:48 balor joined #darcs
10:49 mornfall amgarching: Do you get warnings from darcs-fastconvert? If so, can you paste them somewhere?
10:49 mornfall And the output of git tag / git log | grep TAG.
10:54 balor left #darcs
10:57 amgarching mornfall: (cd ../ttfs-mac; ~/.cabal/bin/darcs-fastconvert export) | git fast-import > http://pastebin.com/GBXxRUkd STDERR seems to be git-only: http://pastebin.com/B71A9x0M
10:58 amgarching Note the difference between TAG and a "regular patch" with a name prefixed by TAG
10:58 amgarching git tag > http://pastebin.com/bWkQKzrB , got log | grep TAG > http://pastebin.com/3YnFzmeK
11:00 mornfall Well, presumably the tags that become TAG commits are not "clean".
11:00 mornfall In darcs terminology.
11:00 mornfall Meaning they aren't in their minimal context.
11:00 mornfall I would expect optimize --reorder to fix that though.
11:00 mornfall You should have the same number of tags and inventories in the repo.
11:00 mornfall Probably.
11:01 mornfall Right, I am getting optimizedTags, which may omit *some* clean tags.
11:01 amgarching mornfall: but is not alwyas possible two tags that are exlusive cannot be "ordered"
11:02 mornfall amgarching: Yes, which means that you won't get a 1:1 conversion in those cases.
11:02 mornfall Branching is not very well supported by fastconvert currently.
11:02 mornfall Maybe owst will improve that.
11:03 amgarching so this is regular behaviour of darcs fastconvert and not  an artifact of old repo (went through darcs1/darcs2)?
11:12 kowey joined #darcs
11:12 kowey first draft of "why darcs users care about consistency" https://docs.google.com/document/d/1SJ​N6omk2bF_t7Lp0S12nhwJME2k63Bsf1IOaOzkR​72I/edit?hl=en_GB&authkey=CPv49IsH#
11:12 kowey factual errors, misleading statements, excessive wordiness, tactical errors etc please
11:12 kowey trim trim trim is the name of the game
11:18 mornfall Artistic license duh? :))
11:18 mornfall Eye of the beholder.
11:19 kowey hmm? I'll bet there are people who swear up and down that Git has great usability
11:19 mornfall how refreshing*? (type mismatch)
11:20 kowey refreshingly simple!
11:20 mornfall Ah!
11:22 ManateeLazyCat left #darcs
11:23 amgarching snaptshots /= set of patches. A set of N patches woudl be roughly equiv 2^N snapshots which is prohibitively many.
11:28 kowey is that number based on the set of possible subsets, or something else?
11:28 mornfall It's actually permutations in practice.
11:29 amgarching imagine they all commute, the a patch is either in (1) or not (0) --> bitmask
11:29 mornfall amgarching: But it doesn't matter.
11:29 mornfall amgarching: Because set of patches has to be represented as a sequence anyway.
11:29 mornfall amgarching: Precisely because not all patches commute.
11:30 mornfall amgarching: And you permute as needed.
11:30 mornfall You can do the same with snapshots.
11:30 kowey (although interestingly, implementing minimal context turned out to be more expensive than expected, because patches commute a lot more than we thought!)
11:32 raichoo joined #darcs
11:33 owst joined #darcs
11:37 mornfall For one, I would be far from delighted if git incorporated commute-merge.
11:38 kowey oh really? would that make git worse in a way?
11:38 mornfall Well, it wouldn't fix anything, just add to complexity.
11:38 mornfall Because it would just be an option that few people use.
11:39 mornfall (3-way will stay the default no matter what, for all I can guess.)
11:39 kowey doesn't git automatically switch to octopus (whatever that is) in certain contexts?
11:40 kowey it might also be a question of porcelain
11:40 mornfall 3-way doesn't handle more than 2 parents.
11:40 mornfall (Partly because the result would depend on the order in which you merge.)
11:42 mornfall (You can turn a binary operation into N-ary as long as the original is associative. 3-way isn't.)
11:42 mornfall Well, you still can, but it is even more hairy that way.
11:44 mornfall And about annotate: it's not entirely true.
11:44 mornfall Most of the speedup comes from the implementation not from the order flip.
11:44 mornfall At least for some files.
11:44 mornfall If your file still has lines from day one, it doesn't help, but the new implementation is still substantially faster.
11:44 mornfall Hard call, either way.
11:45 mornfall And I think while it's admirable to try to be all friendly, I suspect you are falling off on the other side, kind of.
11:45 mornfall I would worry that you are bordering on sleazy/ironic on the "git is great" theme.
11:46 mornfall (Precisely because it isn't. :P)
11:46 kowey ahh that's interesting
11:46 kowey I don't want to go over the top, but I do think people are partly justified in their enthusiasm for Git
11:46 kowey and I really do agree about performance being a usability issue
11:47 kowey hmm, let's see what I can do
11:48 mornfall Sure, performance is an usability issue, first of all.
11:53 kowey fixed annotate paragraph, I think
11:53 kowey it may falsely imply that the speedup comes from the order flip though... hrm
11:53 owst kowey: missing a word in "This patch set mental model may"
11:53 kowey I've also toned down the Git language, no need to be exclamation mark this and amazing that
11:54 owst I think it's not be *for* everybody, but I'm not sure
11:54 kowey silly American
11:54 kowey ah, thanks! owst
11:54 * owst just woke up, so my tired eyes may miss some things :)
11:55 kowey a good proofreader is a blind proofreader if you should be fortunate enough to meet one
11:55 kowey had a friend who by force of reading 40x1 braille cell at a time was really amazing at catching this sort of thing
11:56 owst "darcs patch theory is incomplete" this sentence scares me a little :)
11:57 kowey gotta face the facts
11:57 owst True.
11:57 kowey if you can find another way to say it, though...
11:57 Igloo kowey: You mean you can't do the same sort of pattern matching stuff with a braille reader?
11:57 kowey preserving the directness of the message
11:58 kowey Igloo: hmm? I'm just saying he seems to read more linearly, without jumping around text like I do
11:58 kowey Igloo: so things like "you missed a word" tend to bubble up much more easily for him
11:59 kowey Igloo: it may also just be that he's a very careful reader, told me he used to read man pages from top to bottom for entertainment
11:59 owst Is history aware merging really possible in git?
11:59 owst Surely it'd be super expensive given the snapshot model?
12:00 kowey if a git merge algorithm was given point A and point B in the dag
12:00 kowey and it already knows how to go from A and B to a Least Common Ancestor
12:01 kowey surely it should be able to look at each of the commits between LCA and A, and between LCA and B as patches
12:01 kowey (or the spaces between the commits)
12:01 exlevan left #darcs
12:01 kowey I'm assuming here a git merge algorithm can be some arbitrary program with the right to ask anything it wants of the git repo
12:02 kowey and not being tied to any one merge algorithm must be kind of a nice thing
12:02 owst Right.
12:02 owst I don't think I'd be delighted to see it done in Git, it just wouldn't be "right".
12:02 owst But hey, that's me :)
12:02 kowey it wouldn't feel like Git
12:03 kowey hmm, maybe I should revise that if two actual Git users say "uhh, no I don't think i'd like that very much!"
12:03 kowey two darcs-aware Git users, that is
12:04 owst "If darcs gets a good git-bridge, I'd switch back straight away. Glad to see they've got this on the agenda." Cool!
12:05 kowey revised a bit, less rah, more curious
12:05 * owst is off to collect a parcel from the PostOffice, *sigh*
12:10 owst left #darcs
12:28 kowey is it safe for me to publish this post?
12:29 owst joined #darcs
12:30 owst I'd say so, yes.
12:39 mornfall owst: You forget that doing a snapshot diff in git is very cheap (tens of milliseconds).
12:39 mornfall About the time that darcs takes to read the patch from disk, I would guess.
12:45 exlevan joined #darcs
12:49 kowey left #darcs
12:50 owst mornfall: read the newer commit, read it's parent commit, read the two corresponding trees, then read the blobs/trees and diff?
12:50 owst Or at least, something like that...
13:04 roconnor joined #darcs
13:29 int-e left #darcs
13:45 jeltsch left #darcs
14:12 kowey joined #darcs
14:18 jeltsch joined #darcs
14:19 int-e joined #darcs
14:21 mornfall owst: Right. But you only actually diff files with differing hashes.
14:21 mornfall Like what git diff / darcs whatsnew does.
14:21 mornfall Which there aren't many.
14:21 twb left #darcs
14:21 mornfall And the time to do that is, mostly, negligible.
14:44 jonkri joined #darcs
14:55 copumpkin joined #darcs
14:58 copumpkin left #darcs
14:59 roconnor left #darcs
15:02 jderque left #darcs
15:08 raichoo left #darcs
15:13 jeltsch left #darcs
15:14 jonkri left #darcs
15:43 bsrk2 joined #darcs
15:52 lispy joined #darcs
15:58 * kowey bites the bullet and does one more pass before posting "why darcs users care about consistency"
16:13 kowey hmm, maybe should tune that title a bit
16:14 kowey darcs: consistency is a usability issue?
16:14 kowey on darcs and why consistency matters?
16:16 kowey consistency is a usability issue (why we work on darcs)
16:18 lispy left #darcs
16:25 secorp left #darcs
16:26 shenshei left #darcs
16:29 * dixie do not like wxwidgets compilation on windows xp ...
16:39 mornfall kowey: The original was best of those, IMHO.
16:39 kowey OK
16:42 jeltsch joined #darcs
16:42 kowey here goes nothing! http://koweycode.blogspot.com/2011/04/why​-darcs-users-care-about-consistency.html
16:46 owst \o/
16:47 owst kowey: "you must appreciate refreshingly simple"
16:47 owst missing "how"
16:48 kowey gah!
16:48 kowey stupid non-magically-thought-predicting Blog Engine
16:49 owst "to make sure that you include the patch later on," missing "if"
16:49 kowey thanks
16:49 owst "Darcs is a very much a work in progress." perhaps remove teh first "a"
16:49 owst Sorry, I'm being picky ;)
16:50 kowey be picky! picky is good!
16:50 kowey people tend to notice stupid details like typos and it sticks in their heads much more than the meat of the argument
16:51 kowey I have a very clever friend who can't spell for the life of him...
16:52 kowey I do hope that this has the intended effect of more people getting what Darcs is about
16:53 owst Yeah, it's a good post - good job kowey! :)
16:53 owst is it reddit ready?
16:53 * kowey has that lurchy feeling in his stomach
16:53 owst My second Darcs post in 2 days... :)
16:53 kowey good thing I won't be around, because I'm off to Functional Brighton!
16:53 kowey (post away)
16:53 owst I think noise is good - this is good noise!
16:53 owst Ah, cool. If I was in Soton, I would have come along to that.
16:54 owst kowey: !
16:54 owst "you must appreciate refreshingly how simple"
16:54 owst :p
16:54 owst swap "how" and "refreshingly"
16:55 kowey I wonder how that one happened!
16:55 owst Well, it made me chuckle :)
16:55 jderque joined #darcs
16:58 jderque left #darcs
17:00 kowey left #darcs
17:03 jonkri joined #darcs
17:04 lispy joined #darcs
17:10 teratorn nice post :)
17:12 secorp joined #darcs
17:16 dixie yes :) I like it too
17:17 jeltsch left #darcs
17:19 sm_ joined #darcs
17:30 sm_ left #darcs
17:36 jonkri left #darcs
17:39 Weltraumschaf left #darcs
17:44 Gwern-away joined #darcs
17:45 jeltsch joined #darcs
17:46 gwern left #darcs
17:50 sm joined #darcs
17:51 sm_ joined #darcs
17:57 sm_ left #darcs
18:00 sm I let two copies of a repo get out of sync, with about a dozen conflicting patches
18:02 sm I think the simplest way to get out of this is unrecord all on side, after viewing their details, pull, and redo those patches
18:07 gbeshers left #darcs
18:21 owst left #darcs
18:37 bsrk2 left #darcs
18:41 sm where's the best doc about the various ways of using gui diff and merge tools with darcs ?
18:43 sm what I'd like is for darcs pull to pop up xxdiff and let me resolve the conflict there and then, instead of marking it. I don't think that's supported
18:44 sm aha, what's this...
18:46 sm wasn't there a one-page version of the manual ? can't find it now
18:47 sm http://darcs.net/manual/Darcs_comman​ds.html#SECTION00611020000000000000 , nice
18:50 sm darcs pull --external-merge 'xxdiff -m -O -M %o %1 %a %2' .. this is a lot better!
18:53 bsrk2 joined #darcs
18:54 sm or opendiff
18:54 sm what do version 1, version 2 and the ancestor correspond to I wonder
19:16 jonkri joined #darcs
19:22 sm using xxdiff, after selecting all hunks I assumed I should exit by pressing m (accept merged) - darcs reported an ExitFailure, I hit return and it seemed ok
19:23 sm using opendiff, I did save, quit, and hit return
19:24 Gwern-away is now known as gwern
19:43 lelit left #darcs
20:02 roconnor joined #darcs
20:33 raichoo joined #darcs
20:34 iago joined #darcs
20:41 jonkri left #darcs
20:54 jonkri joined #darcs
20:55 bsrk2 left #darcs
20:55 bsrk2 joined #darcs
20:55 bsrk2 left #darcs
21:12 lispy left #darcs
21:12 jonkri left #darcs
21:48 lispy joined #darcs
21:53 owst joined #darcs
22:02 kyagrd joined #darcs
22:09 raichoo left #darcs
22:11 sm kowey: *really* nice blog post
22:35 kyagrd left #darcs
23:17 ivanm joined #darcs
23:17 ivanm left #darcs
23:18 sm_ joined #darcs
23:20 sm left #darcs
23:20 sm_ is now known as sm
23:40 roconnor left #darcs
23:47 jeffz joined #darcs
23:48 secorp left #darcs
23:50 preflex left #darcs
23:52 kowey joined #darcs
23:53 kowey hello! thanks all! note this interesting question about it: http://www.reddit.com/r/programming/​comments/gtq7b/why_darcs_users_care_​about_consistency_in_version/c1q9cw5
23:53 lambdabot kowey: You have 1 new message. '/msg lambdabot @messages' to read it.
23:56 preflex joined #darcs
23:57 kowey phew, too much Functional Brighton :-) good night!
23:57 kowey left #darcs

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