Camelia, the Perl 6 bug

IRC log for #darcs, 2011-02-11

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

All times shown according to UTC.

Time Nick Message
00:07 sm yay!
00:14 raichoo left #darcs
00:37 gh_ left #darcs
00:37 alexsuraci Nice
00:40 dankna joined #darcs
00:56 copumpkin left #darcs
01:38 copumpkin joined #darcs
02:21 secorp joined #darcs
03:05 intripoon_ joined #darcs
03:08 intripoon left #darcs
03:13 owst left #darcs
03:30 darcswikibot left #darcs
03:30 Igloo left #darcs
03:30 kolmodin left #darcs
03:35 gwern joined #darcs
03:35 gwern left #darcs
03:35 gwern joined #darcs
04:01 darcswikibot joined #darcs
04:01 Igloo joined #darcs
04:01 kolmodin joined #darcs
04:17 gwern you know, the pristine GC problem thing can be pretty awful. today I was messing with my backups repo where I store copies of things like google reader feeds or crontab or `find ~/`, and found my _darcs went from 6.4GB to 179M
04:22 * gwern had been wondering where half my spare gigabytes went to. found out, too
04:54 alexsuraci anyone remember that presentation system kowey used?
04:54 alexsuraci had the zooming in/out and all that
04:55 alexsuraci found it (prezi.com)
05:16 Gwern-away joined #darcs
05:16 Gwern-away left #darcs
05:16 Gwern-away joined #darcs
05:17 gwern left #darcs
05:50 sm left #darcs
06:20 ManateeLazyCat joined #darcs
06:43 ManateeLazyCat left #darcs
07:55 balor joined #darcs
08:01 darcscommitbot left #darcs
08:02 darcswikibot left #darcs
08:02 darcscommitbot joined #darcs
08:03 darcswikibot joined #darcs
08:14 lelit joined #darcs
08:21 twb left #darcs
08:50 Weltraumschaf joined #darcs
09:11 gal_bolle joined #darcs
09:37 gal_bolle Heffalump: congrats on 2.5.1!
10:15 lelit` joined #darcs
12:12 Funktorsalat joined #darcs
12:16 FunctorSalad left #darcs
12:28 iago joined #darcs
12:44 iago hi
12:46 dcoutts mornfall: btw, about this you showed me the other day http://pastebin.com/s4HY3gF6
12:47 dcoutts mornfall: oh, sorry, the paste answers my question :-) you did clean first
12:47 dcoutts so it could not have been an .hi file from the local project
12:47 dcoutts must have been from an already registered package
12:48 dcoutts mornfall: you don't tend to use inplace registration do you? runghc Setup register --inplace
13:05 owst joined #darcs
13:12 gal_bolle left #darcs
13:49 mornfall dcoutts: No, no --inplace.
13:49 dcoutts mornfall: do you still have the saved ghc-borked db?
13:50 mornfall dcoutts: The database yes, but not the .hi files, sadly.
13:50 dcoutts mornfall: JaffaCake wonders if we might be able to try -ddump-if-trace
13:50 dcoutts ah oh well
13:50 Gwern-away left #darcs
13:50 dcoutts mornfall: filed http://hackage.haskell.org/trac/ghc/ticket/4955
13:51 mornfall I run into the problem often, but with "local" .hi files (i.e. running a build after an update, without rm -rf'ing the build directory). I think this is the first time I ran into it in installed .hi files.
13:51 dcoutts yeah, it's more perplexing for registered packages
13:52 dcoutts if you're not using --inplace ever
13:52 mornfall Unless cabal does it automatically, I definitely didn't do it.
13:52 dcoutts or registers into a local db in the build tree, not into the global/user ones
13:53 dcoutts or/it
13:53 mornfall Right.
13:53 mornfall So I don't think this is the case.
13:53 dcoutts mm
13:54 mornfall I think doing a cabal copy and not a cabal register could be a problem, too. But I don't think I ever did this to anything but darcs itself.
13:54 mornfall Oh, hm.
13:54 mornfall Wait a minute. :)
13:54 dcoutts ?
13:54 mornfall Ok, I think I know what happened.
13:54 dcoutts ah, cabal copy to an installed location but not then registering?
13:54 mornfall This is fastconvert, so it actually does depend on libdarcs.
13:55 mornfall So what could happen is that I only cabal copied a newer build of darcs, without calling a register.
13:55 dcoutts mornfall: if you have any more detals perhaps you can append them to the ticket
13:55 dcoutts and I can think about how to make cabal copy/register more foot-bullet proof
13:56 mornfall Well, having an equivalent of "make install" would be probably enough.
13:56 mornfall (I.e. copy and register in one step.)
13:56 dcoutts yes, it's annoying that we ended up having cabal "install" clash
13:56 dcoutts do everything when we still need a "just install please"
13:57 mornfall Exactly.
13:57 mornfall (Which is why I ran into this in the first place...)
13:57 dcoutts there's a secret cabal install --only
13:57 mornfall Oh?
13:57 dcoutts from a UI pov it's a hack
13:57 gwern joined #darcs
13:57 gwern left #darcs
13:57 gwern joined #darcs
13:58 mornfall Well, certainly.
13:58 dcoutts which is why it's not advertised, we're rather sort it out properly, but still not sure what to do
13:58 dcoutts suggestions welcome
13:59 dcoutts make install of course also builds, but it knows when no rebuild is needed
13:59 mornfall I think the (re)configure is the problem not the build.
13:59 dcoutts and if ./configure were also needed due to changes in the environment then make / make install will just fail mysteriously
13:59 mornfall True.
14:00 dcoutts mornfall: perhaps instead of the reconfigure we knew what would force a reconfigure and only did it then
14:00 dcoutts I mean if there are changes in packages that the build depends on
14:00 mornfall Well, that would be good.
14:01 mornfall Another option would be to have copy also do a registration, with a --dont-register perhaps to avoid that.
14:01 dcoutts well the original point of copy was install - register :-)
14:01 mornfall Another option would be to have copy --register, and have regular copy refuse to overwrite interface files.
14:01 mornfall (At least without --force.)
14:02 dcoutts mm
14:02 mornfall Yet another option would be to never overwrite .hi files by including hashes in their (or package directory's) name.
14:03 dcoutts yes I think I like the idea of not allowing a copy into the location of an existing registered package
14:03 dcoutts copy when unregistered is safe
14:03 raichoo joined #darcs
14:03 dcoutts otherwise we should copy + register at the same time
14:03 mornfall (I.e. do the installation's à la nix.)
14:03 dcoutts well, that's the plan in the long term certainly
14:04 dcoutts mornfall: though nix deals mostly with "complete" packages, rather than when people are hacking about
14:04 dcoutts we probably don't want to keep full copies of every cabal build
14:04 mornfall As long as you can GC them quickly, I would almost think it won't hurt that much.
14:05 dcoutts mm
14:05 mornfall Installing over is not considerably cheaper than installing alongside and rm -rf'ing.
14:07 gh_ joined #darcs
14:08 mornfall I guess you don't remember my GHC trac password, duh? :P
14:10 dcoutts mornfall: I'll add a comment
14:10 raichoo left #darcs
14:11 mornfall Oh. It seems I don't have an account in the first place. Thanks. :)
14:12 mornfall (bbl)
14:19 raichoo joined #darcs
14:49 JaffaCake1 joined #darcs
14:51 JaffaCake left #darcs
15:02 Weltraumschaf left #darcs
15:02 raichoo left #darcs
15:07 gh_ left #darcs
15:14 Igloo left #darcs
15:14 kolmodin left #darcs
15:14 gwern left #darcs
15:14 owst left #darcs
15:14 balor left #darcs
15:14 dcolish left #darcs
15:14 dleverton left #darcs
15:14 kerneis left #darcs
15:14 dixie left #darcs
15:14 lelit` left #darcs
15:14 intripoon_ left #darcs
15:14 exlevan left #darcs
15:14 jben2 left #darcs
15:14 darcsbb left #darcs
15:14 dcoutts left #darcs
15:14 burp left #darcs
15:14 int-e left #darcs
15:14 mornfall left #darcs
15:14 knewt2 left #darcs
15:14 Jaak left #darcs
15:14 nwf left #darcs
15:14 darcscommitbot left #darcs
15:14 npouillard left #darcs
15:14 rntz left #darcs
15:18 iago Is that normal?
15:18 iago $ darcs changes | less
15:18 iago Exception thrown by an atexit registered action:
15:18 iago SignalException 13
15:19 burp joined #darcs
15:22 dcoutts joined #darcs
15:22 gwern joined #darcs
15:22 owst joined #darcs
15:22 lelit` joined #darcs
15:22 darcscommitbot joined #darcs
15:22 balor joined #darcs
15:22 kolmodin joined #darcs
15:22 Igloo joined #darcs
15:22 intripoon_ joined #darcs
15:22 exlevan joined #darcs
15:22 darcsbb joined #darcs
15:22 npouillard joined #darcs
15:22 nwf joined #darcs
15:22 dixie joined #darcs
15:22 knewt2 joined #darcs
15:22 rntz joined #darcs
15:22 dcolish joined #darcs
15:22 jben2 joined #darcs
15:22 dleverton joined #darcs
15:22 kerneis joined #darcs
15:22 Jaak joined #darcs
15:22 mornfall joined #darcs
15:22 int-e joined #darcs
15:25 Jaak_ joined #darcs
15:25 darcscommitbot left #darcs
15:25 npouillard left #darcs
15:25 rntz left #darcs
15:25 knewt2_ joined #darcs
15:27 darcscommitbot joined #darcs
15:27 npouillard joined #darcs
15:27 rntz joined #darcs
15:28 copumpkin left #darcs
15:30 knewt2 left #darcs
15:30 Jaak left #darcs
15:33 gwern left #darcs
15:39 gwern joined #darcs
15:44 lispy joined #darcs
15:50 shenshei left #darcs
15:50 shenshei joined #darcs
15:54 gwern left #darcs
15:56 copumpkin joined #darcs
16:01 owst left #darcs
16:04 gwern joined #darcs
16:04 gwern left #darcs
16:04 gwern joined #darcs
16:20 jben2 left #darcs
16:30 sm joined #darcs
16:57 lelit` left #darcs
17:15 shenshei left #darcs
17:18 raichoo joined #darcs
17:21 shenshei joined #darcs
17:25 owst joined #darcs
17:25 secorp left #darcs
17:49 preflex left #darcs
17:51 dcoutts left #darcs
17:51 int-e left #darcs
17:51 int-e joined #darcs
17:51 preflex_ joined #darcs
17:51 lelit left #darcs
17:52 dcoutts joined #darcs
17:52 preflex_ is now known as preflex
18:23 balor left #darcs
18:29 secorp joined #darcs
18:30 iago someone has some haskell-src-exts experience?
18:33 dankna indubitably, but not me
18:34 iago that's not helpful :P
18:35 dankna perhaps not, but at least you got a response!
18:35 dankna heh, sorry
18:44 iago I'm not able to parse Darcs sources due to the use of Literate Haskell and CPP, mainly Literate Haskell
18:45 iago haskell-src-exts is supposed to handle lit-haskell and CPP, but I don't find how
18:46 lispy iago: does this help?
18:46 lispy http://www.mail-archive.com/haske​ll-cafe@haskell.org/msg57921.html
18:47 dankna look also as though you can turn on CPP explicitly in the ParseMode, if you use parseWithMode instead of parse.
18:47 dankna http://hackage.haskell.org/packages/archi​ve/haskell-src-exts/1.10.1/doc/html/Langu​age-Haskell-Exts-Parser.html#t:ParseMode
18:48 dankna really, de-lit ought to also be an option to the parseMode.
18:48 dankna but I guess the author hasn't figured out a good interface for it yet.
18:52 iago ok I see
18:52 iago lispy, thanks for the link, I had read it before
18:52 iago but I though that delit was something that was done *also* in parseModule* family of functions
18:52 iago and now I realize that it is not
18:53 iago dankna, I'm specifying CPP as a desired extension if you refer to that
18:53 iago it does not work
18:53 dankna I see :(
18:53 iago though what I'm doing now is to run hscpp independently
18:53 dankna gotcha.
18:54 iago cpphs
19:03 kolmodin left #darcs
19:17 dleverton left #darcs
19:18 dleverton joined #darcs
19:27 secorp left #darcs
19:27 secorp joined #darcs
20:15 alexsuraci left #darcs
20:15 alexsuraci joined #darcs
20:20 Funktorsalat the proper way to make a 'branch' in darcs is to just darcs get the repo to some other folder right? (someone is asking me because I said I prefer darcs to git :))
20:20 Funktorsalat (without being an expert on either ;))
20:20 sm Funktorsalat: that's right, optionally adding --lazy for speed
20:22 iago left #darcs
20:28 dankna that is correct
20:29 dankna although it comes with a caution that branching is not as universally appropriate for all use-cases as it is in git
20:29 dankna because it's not as streamlined merging them back later
20:35 Funktorsalat hmm, wouldn't you just pull one branch from the other? or do you mean there will usually be merge conflicts
20:38 sm there can be conflicts if both branches work on the same things. darcs can hold conflicting patches in one repo, but it may be wiser to avoid that, which could mean re-recording patches from one of the branches
20:40 ppavelV6 joined #darcs
20:42 ppavelV6 left #darcs
20:52 raichoo hi, when are 2.5.1 binaries are coming for mac os?
20:52 Heffalump when someone makes them :-)
20:52 raichoo ^^
20:53 Heffalump dankna: I'm fairly sure I got haskell-src-exts to deal with de-litting without any hassle.
20:53 Heffalump I had to make a few local fixes to work round things it didn't support properly though.
20:53 raichoo Does it build with ghc 7 yet?
20:54 dankna Heffalump, I wasn't the one trying to do that, I was just offering help to iago, but thanks
20:57 Heffalump dankna: yeah, I know, but iago has gone and I felt the need to address my comments to someone :-)
20:57 Heffalump raichoo: yes
20:57 dankna oh, haha, okay :)
20:58 raichoo Heffalump: Great :) I might give building it myself a try then :)
21:05 owst left #darcs
21:07 ppavelV6 joined #darcs
21:08 ppavelV6 left #darcs
21:18 Gwern-away joined #darcs
21:18 Gwern-away left #darcs
21:18 Gwern-away joined #darcs
21:21 gwern left #darcs
21:22 Gwern-away is now known as gwern
21:29 sm alexsuraci, what's the ssh address of a darcsden repo ? I can't find it
21:39 pumpkin joined #darcs
21:42 copumpkin left #darcs
22:15 alexsuraci sm: owner@darcsden.com:reponame
22:16 sm alexsuraci: thanks, it was giving a confusing error (reported)
22:16 alexsuraci K
22:17 sm I just set up my first public project on darcsden.. was close to using github and then remembered dd
22:17 sm so far so good - thanks
22:23 iago joined #darcs
22:31 Heffalump hi iago
22:32 Heffalump I'm sure I got haskell-src-exts to manage cpp and delitting itself
22:32 Heffalump but I did have to hack a fair bit of other stuff to get it to read the darcs code
22:59 Heffalump iago: re your patch "Add missing extensions to darcs.cabal", how come darcs built without it?
23:01 lispy Heffalump: My guess: Those extensions were listed in the individual files
23:01 iago hi Heffalump
23:02 iago Heffalump, to read full Darcs code I had to remove some stuff, but literate haskell was the main problem
23:02 lispy I have some repos that I need to move off of community and I may move some to darcsden.
23:03 iago Heffalump, second question... eeh, I don't know, I just added them because haskell-src-exts complains about them
23:03 iago I check the source, I see it is true they were used... and I add them
23:06 iago Heffalump, for instance, in Lcs.hs line 439 you find "aLen :: (IArray a e) => a Int e -> Int", whose type signature make use of multiparametrictypeclasses...
23:07 iago maybe that use is not a problem for GHC
23:07 iago (LANGUAGE pragma just declare CPP)
23:08 lispy iago: each file gets the extensions defined in .cabal and defined in LANGUAGE.  Perhaps MultiParameterTypeClasses is in the .cabal /me checks
23:08 iago lispy, I could have a fault but I check if it was before add it
23:09 lispy iago: hmm...I don't see it in the cabal file.  It's possible some extension there implies it...that's harder to check
23:09 lispy So maybe a difference between GHC and haskell-src-exts
23:09 Heffalump iago: oh, I see.
23:09 Heffalump Yeah, I discovered that too. I think it's a bug in HSE>
23:10 iago I think haskell-src-exts now manages some extension dependencies... but that support may be incomplete
23:10 Heffalump http://trac.haskell.org/ha​skell-src-exts/ticket/202
23:10 Heffalump if that darcs.cabal patch is only needed to keep haskell-src-exts happy, I suggest we leave it out for now.
23:11 iago Heffalump, that's ok for me, though I think the bug is in GHC
23:11 iago it is clear that syntax should be rejected in std haskell
23:12 iago (or "it is clear" from my POV)
23:12 Heffalump true
23:12 * Heffalump boots his laptop into Linux to find the work he did with HSE before.
23:13 lispy iago: There are arguments both ways.  It's nice to be able to reference code that uses MPT without enabling the extension, but it's also nice that NOT using extensions gives you Haskell98
23:14 Heffalump also, you can't use the syntax without importing a non H98 module
23:15 iago I see
23:15 secorp left #darcs
23:15 iago and what about foreign-function-interface?
23:15 iago ok I see, it is in the LANGUAGE pragma
23:16 iago I don't know if you want to have all extensions listed in the cabal file, despite they were also mentioned in individual files
23:16 iago (well, I remember Heffalump to point me to do that, so I expect a "yes, we want that")
23:16 lispy iago: BTW, what were you doing with HSE?  Anything cool you can talk about?
23:16 Heffalump the current strategy (decided unilaterally by me) is to have commonly used extensions listed in the cabal file, and anything uncommon listed in individual files
23:17 Heffalump if I said something different, I didn't mean it, sorry!
23:17 lispy iago: The downside of listing them ALL in the .cabal file is that all modules get them, even if they don't need them.  The upside is consistency, easy auditing, and well everything compiles with the same extensions.
23:18 lispy Heffalump: That's reasonable.  Do you have any metric (informal or otherwise) for common vs. uncommon?
23:18 Heffalump iago: the two local changes to darcs code I had to make for HSE were (a) remove the GADT record syntax in Darcs/Patch/Choices.hs, and get rid of a unicode character literal in Darcs.Test.Patch.Info
23:18 Heffalump lispy: not really.
23:19 Heffalump also, to some extent there's an element of "I don't see any problem with using this whenever needed"
23:20 iago lispy, I am computing source metrics for Darcs, volume and cyclomatic complexity (or something similar :) for both modules and functions/methods, and number of incoming imports for each module...
23:20 Heffalump e.g. BangPatterns and PatternGuards
23:20 lispy iago: cool.  Please publish your results when you get time :)
23:20 iago lispy, I will in the report, I will compare Darcs with XMonad
23:21 lispy iago: oh, lots of us here at Galois would love to see that
23:21 iago Heffalump, well I asked you to use some syntactic extension like recordpuns (though finally I didn't use it) and you point me to add extensions to the cabal file
23:21 lispy Well, by lots, I mean I'd be interested, I'm sure dons would, and I suspect some of our program analysis people like Aaron Tomb would be
23:22 Heffalump iago: I guess that one I thought was in the "ok to use anywhere without thinking" category
23:22 iago Heffalump, I had to provide baseFixities to ParseMode, and I also had to remove use of BangPatterns in GADTs
23:22 iago well, not bangpatterns excuse me
23:22 iago strictness anotations
23:23 Heffalump would you like a copy of the tool I wrote, the changes I made to HSE and the changes I made to darcs?
23:23 iago they work in a common data type declaration, but HSE compains when found them in GADTs
23:23 iago Heffalump, yeah please
23:23 Heffalump I can't remember the details of everything I had to hack but it might be useful
23:24 iago lispy, some of you work in XMonad?
23:24 iago in->on
23:24 lispy iago: dons is a core author of xmona
23:24 lispy I use xmonad but I don't work on it
23:26 iago Heffalump, that's ok, I'm almost done but it maybe useful anyway, it may be helpful parsing XMonad too
23:27 iago lispy, well, we will see what results are :P though I had to add xmonad-contrib e xmonad-extras, and even doing that XMonad code is about the half of Darcs's
23:27 iago so maybe there is some code that was not done by the XMonad main contributors
23:28 lispy iago: ah, interesting.  I think your analysis should try to separate them.  I've heard from dons that the standards are quite different between core and contrib
23:28 lispy iago: core has some isabelle proofs or something but with contrib almost anything can get accepted
23:28 iago well, I will try to consider it, though it won't be faith since xmonad core is 1/20 the size of Darcs
23:29 iago lispy, does not matter, it is a maintainability analysis
23:30 iago I just pick XMonad because there is no useful info about how a good Haskell code is, so the only way is to compare code with a project that it is known (or has the famous) to be of good quality
23:31 lispy iago: GHC vs darcs?
23:31 iago GHC is not famous to be of good quality
23:31 iago (AFAIK)
23:32 iago though if you have some special interest... since I did not spend almost no time with XMonad I can change
23:32 lispy iago: I'd like to see all 3 :)
23:32 lispy iago: XMonad is good
23:32 Heffalump iago: http://urchin.earth.li/~ganesh/temp/hse-darcs/
23:32 lispy Just that xmonad contrib is not known for good style
23:32 lispy But, xmonad core is
23:33 iago Heffalump, "You don't have permission to access /~ganesh/temp/hse-darcs/ on this server."
23:34 iago lispy, well, I could just analyze XMonad core and GHC
23:35 lispy iago: Do the analysis you think best for your class but then make your tools available on haskell-cafe and then let others analyze stuff :)
23:37 Heffalump iago: hmm, that's annoying - just a permissions problem with browsing directories. There are three darcs repos under there; darcs-cleanup, hs-export-check, and haskell-src-exts
23:37 intripoon joined #darcs
23:37 Heffalump you should be able to darcs get those three.
23:38 iago Heffalump, yep it seems to work fine, thanks
23:39 iago lispy, I'm using a quick&dirty tool with HSE uniplate and syb ... :P I don't know if I would like someone to see that work
23:40 intripoon_ left #darcs
23:40 iago in fact my "tool" is not very useful since just provice raw data, instead of provide "rates"
23:40 iago raw data is usually too boring to look on it, you prefer to see rates and when a rate is low just go deep
23:40 lispy iago: Maybe share your scripts at some point and it can be cleaned up and then shared publicly?
23:41 iago but to give rates a regression work would be needed
23:41 raichoo left #darcs
23:42 iago uhm, is that ok?
23:42 iago Finished getting.
23:42 iago It's too late to use atexit
23:42 iago darcs: /home/iago/.darcs/cache/patches/0000005885-c26a52​e245cb509243f5ca7c847fc2868eab992200a7e2c59f0418a​5ee16f2ca-new_9c4af91422f860c7c0eb8974457c2f42: getSymbolicLinkStatus: does not exist (No such file or directory
23:42 iago I recall someone complains about that before
23:42 sm I just got that
23:43 sm the first time I got a repo
23:43 sm harmless I guess
23:44 iago lispy, do you use some quality metrics at Galois?
23:46 lispy iago: not like you're measuring for Haskell
23:47 iago could you be more specific ?
23:47 iago I'm asking only about Haskell
23:48 gwern incidentally, if anyone cares, I did some poking yesterday with a shell script to investigate 7zip's lzma compression compared to gzip; turns out that on my wiki repo, switching all the .gz files with .7z files saved about 10% of the space taken by the _darcs
23:50 iago gwern, I usually get also better compression rates with bzip2 than with gzip
23:51 gwern probably could've gotten a few more percent if all the big files like PDF were 7zipped too
23:51 gwern iago: well, naturally. but 7zip is better than bzip too
23:52 iago I know
23:53 iago I just use 7z in some special cases
23:55 pumpkin is now known as djtiesto
23:55 gwern it does much better for textual stuff. I was keeping a tarball of the static HTML version of gwern.net with the repo included, and 7z cut it in half from gzip's 120M or whatever
23:56 Heffalump iago: how long do you expect the test harness to take to run with patch539 applied?
23:56 gwern got an even bigger improvement over bzip when I ran it on the tarball of the mbox of my gmail
23:56 Heffalump nontrivial permutivity is currently crawling
23:56 djtiesto is now known as copumpkin
23:57 gwern darcs isn't doing something odd like shelling out to gzip is it? maybe future repository formats will use 7zip instead of gzip
23:57 iago I think it takes, at least, 2x the old time
23:57 Heffalump no, it uses zlib
23:58 Heffalump iago: this particular test is way worse than that
23:58 Heffalump (to the point of being unacceptably slow)
23:58 gwern iago: the important thing would be decomprssion speed
23:58 lispy If someone delete's their _darcs/inventory or _darcs/hashed_inventory by mistake, are they just hosed?
23:58 lispy No way to recreate it?
23:58 Heffalump lispy: pretty much
23:58 iago Heffalump, when testConditional nontrivialCommute is used... it could becomes very slow
23:58 lispy Heffalump: that seems a bit short sighted on our part, doesn't it?
23:59 Heffalump if someone deletes _darcs by mistake they're also hosted.
23:59 Heffalump s/hosted/hosed/
23:59 Heffalump if they delete a particular patch file they're probably also hosed
23:59 iago let me see
23:59 lispy Heffalump: Yeah, but if I were say, helping dons with something I might mistake inventory for index and tell him to delete it.  And then where would he be?
23:59 Heffalump there's some stuff they might get lucky with, but basically, that stuff matters.
23:59 gwern http://blogs.reucon.com/srt/2008/02/18/​compression_gzip_vs_bzip2_vs_7_zip.html puts 7zip at faster decompression than bzip, but around half as fast as gzip
23:59 Heffalump lispy: hosed.

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