IRCloggy #git 2007-04-12

Logs Search ←Prev date Next date→ Channels Documentation

Provider of IRC logs since 2005.
WARNING: As Freenode became unjoinable and lost all warnings in topics, we cannot log channels on Freenode anymore.

2007-04-12

→ beu joined00:18
Tali I'm really impressed on what Linus already achived, his implementation is really nice.00:26
But now I have to go to bed...00:26
mugwump night00:27
clee linus is a ninja00:28
aeruder_ heh, yea00:28
Tali he reimplemented the core of my submodule support in a few hours, and did it much better than me...00:31
mugwump but probably couldn't have done it like that without your implementation coming first00:32
hpa Hm00:33
Has anyone heard from Petr Baudis lately (I don't know what nick he uses?)00:33
repo.or.cz has had permission problems, and he hasn't answered my emails on the subject...00:34
mugwump kampasky00:34
beu hpa: he's usually on here as pasky.00:41
yashi i haven't read linus' gitlink patch yet, but does git still allow us to remove entire working dir and rebuild it with checkout?01:14
maybe with alternate object store?01:14
kampasky hpa: I'll answer later today, sorry01:15
hpa: kind of busy01:15
(answer + look at it, of course)01:15
hpa kampasky: :)01:16
kampasky hpa: I've done it now :)01:20
hpa kampasky: *d'oh*01:24
kampasky: I presumed creating the project would add myself to it...01:24
(The text on the website kind of implied it.)01:24
kampasky oh, I'll revise it then, thanks01:25
it actually has no idea about who "you" are while you are creating the project01:25
hpa Ah01:28
→ yashi joined01:35
→ blabli joined01:39
→ alb joined01:41
→ lcapitulino joined01:49
→ Dravont joined01:59
→ duncanm joined02:20
duncanm anyone using guilt here?02:20
→ iano joined02:24
← blabli left02:24
→ Avochelm joined02:28
← Dravont left02:34
→ iano_ joined03:00
→ dkowis joined03:02
dkowis → SupremeCommander03:03
← SupremeCommander left03:03
→ dst_ joined03:13
→ Romster joined03:18
→ Ori_B_ joined03:35
→ robtaylor joined03:40
→ Tv joined03:45
Ori_B_ → Ori04:05
Ori → Ori_B04:05
→ Bobby joined04:22
← Bobby left04:22
→ rkaway2 joined04:27
→ spearce joined04:50
← Avochelm left04:58
→ fultilt joined05:14
spearce anyone see their windows system just eat (delete) a loose object? one that's needed?05:37
gitster what does that mean?05:42
→ oMish__ joined05:42
spearce so a few times I've seen a loose object just suddenly go *poof*. never to me, always to another user.05:42
guy today had git-write-tree barf because a blob in the index didn't exist anymore as an object.05:43
oMish__ How can I print list of all files/dirs (under .) that exist in filesystem but not added to the git ?05:43
mugwump filesystem?05:43
gitster You mean by "windows" the Cygwin port?05:43
spearce months ago a different guy (different windows system) had a few loose objects go poof too. quite old ones, that had been around for at least a week or two.05:43
gitster I always run "clean -n" for that ;-)05:43
spearce NTFS, and yes, Cygwin port.05:43
gitster Windows "Disk Cleanup" (remove unused files), perhaps?05:44
mugwump virus scanner?05:44
spearce one object, out of thousands loose?05:44
yea, i'm suspecting the virus scanner over disk cleanup.05:45
mugwump some of them are heuristic05:45
but heuristic combined with random deletion and no warnings sounds positively crazy05:45
gitster positively windows, you mean?05:45
spearce yea, positively windows. :-)05:46
mugwump <== 100% windows free since 199805:46
spearce the guy it just happened to today is freaking out, because every time he touches "that git thing" it never works for him. this is just his latest problem...05:46
oMish__ git ls-files --others05:47
spearce the earlier one (months back) found it very funny that the object went missing, because he created that version of that file on that system just a few days before... and pushed it around to coworkers. we had at least 30 copies of that thing in various places before his copy suddenly went *poof*. and he never rewinds.05:48
aeruder_ spearce: "that git thing" heh05:48
aeruder_ knows exactly what kind of people you mean05:48
aeruder_ just off those 3 words05:48
;)05:48
mugwump I have simply *looked* at a Windows machine in the past and had it crash06:01
Just as a friend and I were talking about my effect on Windows machines, too06:02
spearce sadly too many people think windows is the best thing since punchcards06:05
z3ro I've got a question about bisecting a tree where you have a commit that changes things that you don't want in the middle of the bisect, for example: (good)---(unrelated commit)---X many commits---(bad) HEAD06:14
so when I'm bisecting and I end up somewhere between (unrelated commit) and (bad) how can I get rid of the unrelated commit. I guess git-revert?06:15
but I don't want to actually commit anything. just change my working copy.06:15
or would it be better to just git format-patch on the unrelated commit, then patch -R after every git bisect good/bad.06:16
gitster do you mean "I do not want to test that commit"?06:16
it is very unclear what you mean by "unrelated".06:17
z3ro gitster: kind of... that commit changes the method of calculation used (this is some 3d stuff. that commit moves the calculation from the GPU to the CPU)06:17
I want to bisect the tree as if that commit never happened.06:17
long story short, some of the debugging stuff won't work when the calculation is done on the CPU. and I need the debugging stuff to determine if the commit is good or bad.06:18
gitster you do not have to test the commit bisect suggests. you can just "git reset --hard some-other-commit" that is between good and bad.06:18
z3ro gitster: yes, but then I'd still be after the commit that moved the calculation to the CPU.06:19
basically I want to bisect as if that commit never occured.06:19
I guess the easiest way to do this is just patch -R each time.06:19
I just wanted to see if there was some magical git command that did this automatically. I guess there isn't.06:20
gitster Ah, I see. Yeah, I think that can be scripted (reset --hard && patch -R && run test && reset --hard) and fed to "git bisect run".06:20
The random tree manipulation such as "as if change X and Z never happened" is application specific, so there is not much point to build that in "the git thing", but you can script that "your application specific" part.06:22
z3ro gitster: yeah. that's what I thought. thanks. :)06:22
→ orospakr joined06:24
Dodj|gone → Dodji07:02
→ sgrimm joined07:05
→ freakabcd joined07:18
freakabcd hi all07:18
i obtained xorg/xserver with this command: git clone git://anongit.freedesktop.org/git/xorg/xserver07:19
then went into the directory and checked out a branch: cd xserver; git checkout server-1.3-branch07:20
now how do i go back to master?07:20
Ori_B the same way you went to server-1.3-branch07:20
hint: git-branch lists the branches.07:21
freakabcd so just git checkout master07:21
?07:21
Ori_B perhaps you could try it and see; it's not as though it'll mess anything up if it's incorrect.07:22
(in other words, yes.)07:22
freakabcd heh, ok. is there a way to list the branches in accordance with their creation/modification ?07:23
like how gitweb shows them at the bottom?07:23
gitster See git-for-each-ref(1).07:29
← maio left07:59
→ dkagedal joined08:07
→ nud joined08:40
← dduncan left08:48
→ ferdy joined09:13
→ not-xjjk joined09:15
→ z3ro joined09:21
→ dst_ joined10:00
→ gabor1023 joined10:28
→ GyrosGeier joined10:44
→ tahseen joined10:47
tahseen hi10:47
I want to setup a git repository on "X" where I have ssh access10:48
I put an existing repository at my home directory at "X" and used git-clone ssh://X/my-repo/ and it doesn't seem to work10:49
Do I need to setup things specially on "X" to serve the repository over ssh10:50
mugwump did you specify full path after X/ ?10:53
and install git on X?10:53
tahseen yup10:55
mugwump nopaste the errors somewhere perhaps... eg rafb.net10:55
GyrosGeier shouldn't it be "git+ssh://"10:57
djpig GyrosGeier: nope, this isn't svn :)10:58
mugwump git+ssh:// is correct, I just thought ssh:// worked too10:58
tahseen i tried both10:59
djpig mugwump: git-fetch disagrees with you...10:59
mugwump: the man page that is10:59
tahseen ssh: 192.168.1.52:: Name or service not known10:59
fatal: The remote end hung up unexpectedly10:59
fetch-pack from 'git+ssh://192.168.1.52:/home/tahseen/test/' failed10:59
mugwump there's an extra : there10:59
tahseen removing that make the first line of error "bash: line 1: git-upload-pack: command not found"11:00
rest is same11:00
GyrosGeier I've used git+ssh:// all the time11:00
tahseen, then it cannot find git on the remote machine11:00
mugwump ok, so git-upload-pack isn't available on the remote system. did you install it into your user dir?11:00
(ie, ran make install as your own user rather than root)11:01
tahseen not in user directory, but the path is exported in .bashrc11:01
would that cause problem11:01
mugwump try: ssh 192.168.1.52 which git11:01
that needs to work11:02
tahseen yup, that doesn't work11:02
anders_ As for git+ssh/ssh+git/whatever I just love the second part of http://article.gmane.org/gmane.comp.version-control.git/3967611:02
tahseen ok, let me try to resolve that.11:02
djpig anders_: thanks for the link :)11:03
tahseen interesting to know11:04
mugwump clearly, both git and ssh are required, so git+ssh:// seems appropriate11:05
but yeah, I really only started using it because of cogito11:05
right, that's it, not having to specify a full path has convinced me to start using the "evil" implicit host syntax straight away11:08
instead of making /git symlinks11:08
fonseca .. and live with the stupid warnings. ;)11:10
mugwump only from cogito! :)11:10
→ chris2 joined11:19
Tali tahseen: ssh does not start a login shell if you provide a command, thus .bashrc is not sourced for remote git commands11:32
→ Dodji joined11:33
mugwump probably need to make a .ssh/environment11:33
Tali mugwump: but that is not enabled in the server by default11:34
mugwump iirc11:34
Tali so of course the best thing is to have git installed in /usr -- everybody needs it, anyway ;-)11:35
→ lcapitulino joined11:37
→ robfitz joined11:41
→ Oejet joined11:42
tahseen Tail, thats what I'm seeing11:44
→ GyrosGeier joined11:45
yashi gitster: are you around?11:48
→ GyrosGei1r joined12:13
yashi → yashi_away12:17
GyrosGei1r → GyrosGeier12:20
→ nud joined12:23
jwb_gone → jwb12:38
→ pa_in joined13:06
→ ivan_ joined13:08
→ eMish joined13:11
eMish With which commands I clone remote ssh:// repo into some archive form, single file, from which I can later open the repo again ?13:12
Also, which command prints me sha1 of current state, so that I can print same of the other repo to comapre is states are the same ?13:17
→ orospakr joined13:25
→ Kristoffer joined13:28
moh eMish: I don't think you can clone in such a way that you create an archive on-the-fly.13:37
clone a bare repository to a tmp directory and then tar it up?13:38
→ Romster joined13:46
← Oejet left13:52
kampasky gitster: hmh, I finally get to update the homepage with new version number and guess what, in *minutes* you announce 1.5.1.1 ;)13:58
gitster: is there an easy way to determine the latest released git version?13:59
tomprince git-describe --abbrev=0 maint14:00
eMish tomprince, whas this to me ?14:12
kampasky tomprince: that's a good idea, I like it14:15
eMish which command prints me sha1 of current state, so that I can print same of the other repo to comapre is states are the same ?14:18
kampasky I wonder if git status does it14:18
in cogito, cg-status :)14:18
(or cg-object-id will print the raw info)14:18
in git, git-ref-parse HEAD, but that's very "raw"14:19
er, git-rev-parse14:19
matled kampasky: see GIT-VERSION-GEN for the script to get the verison of the latest git release, there are other tags (git-gui) which should not be used for that :)14:20
kampasky my original idea was to extract it from kernel.org but yes, having an auto-updated clone of maint will be simpler14:21
matled mh, it seems there is nothing special in GIT-VERSION-GEN for git-gui tags, perhaps it was just a problem with old git-describe, but I dont know what the problem was..14:25
→ rphillips joined14:27
→ albertito joined14:37
→ Romster joined15:06
corecode catches up on 2000 git mails15:17
corecode any big news?15:17
oh, index v215:21
nice15:21
→ troyt joined15:25
matled corecode: 6 small patches for subproject support15:27
corecode yah, i saw linus' mails15:27
ribas mugwump: pity i couldn't get soc :(15:27
corecode did packv4 arrive?15:28
v3?15:28
whatever15:28
duh, asciidoc/xmlto is slow :/15:30
building the docs takes longer than building the sources15:30
kampasky -*-Mutt: Mail/git [Msgs:39878 New:12191 Post:7 Inc:55 160M]15:31
200015:31
I envy you ;)15:31
→ hein joined15:31
Tali *g*15:32
corecode possibly you have a different list to read :)15:32
Tali kampasky: no new/old distinction in your mutt?15:33
hein → Sho_15:33
→ orospakr joined15:34
ribas Randal: so you're at fisl too :p15:35
Tali kampasky: mine is longer: [Msgs:43823 New:12 Old:29415 Inc:11] ;-)15:37
kampasky Tali: no, I've always found the 'old' marker very annoying and just useless for my mail workflow :)15:37
corecode what kind of list is this?15:38
kampasky [email@hidden.address]15:40
are you talking about a different one?15:40
corecode you have 30k mails from [email@hidden.address]15:40
Tali corecode: from 43k messages I have ~30k unread, and 12 of them have newly arrived since I looked into the folder the last time.15:40
corecode: yes, its been pretty active since the beginning ;-)15:41
corecode i guess you don't intend to read all of them15:41
kampasky corecode: I have ~40k in my current mailbox15:42
oh, the archive mailbox has just another 4k15:42
boring15:42
corecode hm. if i do a git repack -a -d, i can remove the alternates?15:44
or not?15:44
the pack is only 100MB, that's irritating me15:45
matled move the alternates file away, do a fsck, if it is successfull everything is ok :)15:46
corecode no, doesn't work.15:48
it doesn't repack the remote branches15:48
wtf?15:48
→ spearce joined16:00
ribas mugwump: ping16:01
corecode git-repack didn't work but git-pack-objects did16:05
lcapitulino spearce: hey :)16:15
spearce morning. congrats. :)16:15
→ Oejet joined16:15
→ krh joined16:16
lcapitulino thanks16:16
spearce: I'm reading the latest e-mails from the list (just to get updated), and doing the basic setup (cloning the repo, putting to run the latest version and etc)16:16
spearce good.16:17
lcapitulino next, I plan to write a more detailed schedule and will have to think about what to do first16:18
spearce i'm a little behind on some of the mailing list traffic myself, i've been utterly swamped with things not-related-to-git. :-)16:18
agreed. there's a lot we could do, but we want to do it smartly, so the end result isn't an insane mess.16:19
lcapitulino ok, no problem16:19
yeah16:19
spearce i'll probably be more sane in another week or two. i hope. damn deadlines.16:19
lcapitulino okay16:20
spearce you are roughly in my timezone, aren't you?16:23
lcapitulino I don't think so16:23
UTC-3 here I think16:23
spearce ah, so an hour ahead of me i think. that's somewhat close. :-)16:23
lcapitulino ah, ok16:23
13:23 here16:23
→ robin joined16:24
spearce ok, so only an hour ahead.16:24
lcapitulino there's a recent e-mail from Linus where he says that it isn't smart to change all the functions to return error codes16:25
spearce fyi, my irc times are rather flaky right now. i'm not always with a network connection that can get onto irc during daytime hours.16:25
yea, i know which thread you are talking about. he might be right, and we might be better off using setjmp/longjmp in at least a few places.16:25
lcapitulino he said that it's better to make the caller provide the xmalloc() (and family) functions16:25
yeah, something to think about16:26
→ nud_ joined16:27
spearce the biggest problem with a longjmp, aside from restoring register allocated-variables to pre-setjmp values, is its hard for intermediate functions in the stack to perform cleanup action if the xmalloc decides he's going to jump out and return an error at the API level.16:28
robin want exceptions...use a language that supports it16:29
spearce heh, indeed. :-)16:29
gah, its pouring outside... at least it isn't snow.16:32
Dodji → Dodj|gone16:33
robin git mem.mgt is messy, introducing longjmp etc is going to make it truly messy16:34
so you'll need garbage collection too16:35
Oejet lcapitulino: Are you a GSOC-student?16:35
lcapitulino Oejet: yup16:35
Oejet: why? :)16:35
Oejet lcapitulino: In any case welcome to the community. We will try to be helpful.16:36
lcapitulino Oejet: thanks!16:36
Oejet: I'm a bit impressed with the GIT community BTW, I'm more used to... hmm.. Let's say, not-so-lovely people :)16:40
Oejet lcapitulino: Ah, you're an OpenBSD chap.16:41
(...sorry, couldn't help it)16:41
lcapitulino Oejet: no, no. Not so bad :))16:41
spearce gitster just said he's going to get tougher. ;-)16:42
lcapitulino :)16:43
robin hopefully not nastier16:43
spearce no, not nastier. just more picky about the quality of work he takes.16:44
Oejet Hm, hopefully he will throw some peanuts at us trivial-patchers.16:46
robin I didn't mean to imply he's nasty... :)16:46
spearce gister just got bitten by a series of patches that weren't tested as well as they needed before they hit the mainline branches. and some of them problems snowballed up at the same time. i think he got a little too trusting of the patches he was accepting. :)16:48
ok, i must shove off to a part of the world that doesn't know what irc is... deadlines to meet and all that.16:50
→ Roomster joined17:02
→ Eludias joined17:07
Roomster → Romster17:21
bartman is it possible to remove changes from a commit using --amend?17:24
I mistakenly used commit -a, but meant to commit only 1 of the files17:24
$ git commit --amend arch/i386/kernel/reboot.c17:24
.. still picks up all 6 files17:25
bartman goes the git-format-patch, edit, git-am route17:27
matled hooray, % git test master17:42
*** glibc detected *** realloc(): invalid pointer: 0x0811d14c ***17:42
→ orospakr joined17:44
matled is there some way to make gdb break at this point?17:48
mountie bartman: with git-format-patch you could use a path-limiter to only get the one file changed you want17:48
bartman ah, so true17:49
mountie git-format-patch <...> HEAD~1..HEAD -- arch/i386/kernel/reboot.c17:49
bartman I have used it before, just forgot17:49
james_w bartman: also git commit --amend commits the index, so get the index to reflect the state of those files at the previous commit and they will be gone.17:50
matled ah, got a SIGABRT this time17:52
bartman james_w: that makes sense. I don't however see a way to have git-update-index forget changes17:52
matled git-format-patch -1 is also quite nice17:52
(-N will give patches for the last N commits)17:53
bartman yes, use -# for format-patch, log and gitk all the time17:53
james_w bartman: git rm --cached, or git reset depending on the state of the tree.17:55
git reset will probably be the one you want in this case. git reset --mixed HEAD^ -- paths will get the index to have the versions in the parent of HEAD. Then git commit --amend should change HEAD to record no modifications to these files.17:56
either that, or I'm completely wrong.17:57
bartman thanks.17:58
next time it comes up, I will try reset --mixed and --amend17:59
→ lyakh joined18:01
→ bronson joined18:11
→ pflanze joined18:35
→ jeffpc joined18:35
pflanze hello. I'd like to use git for an application, so after seeing that libgit isn't very well suitable for it I'm looking into scripting the git tools.18:39
Has anyone already created any such interfaces for e.g. perl?18:39
(apparently not, nothing on cpan)18:39
Or maybe I should use libgit anyway and run it in a subprocess so that it's no problem when it exits. Any documentation/suggestions?18:40
matled pflanze: corecode and me played a bit with ruby, we can already read objects but not much more :)18:44
pflanze reading objects efficiently would already put the base for the rest (no need to use git-log if one can reconstruct the log from single objects..)18:45
matled: any place I can see your code?18:45
matled http://repo.or.cz/w/gitrb.git18:46
jrockway i've been considering doing good perl bindings18:46
i hate having to exec stuff18:46
i'm working on a trac workalike in perl... but no longer have any interest in touching svn, so i'll probably be writing the git version first18:47
or i'll just steal gitweb, because it's good enough :)18:47
matled perhaps the git library project (soc) will help, but this will take some time..18:48
jrockway the basic plan i had in mind was wraping the executables minus main() and then calling the functions from perl18:49
haven't looked at any code, though18:49
so that is possibly a terrible idea18:49
ENOTENOUGHFREECYCLES18:49
ribas mugwump: ping19:01
robin pflanze: wrap jgit :)19:29
pflanze what's jgit?19:30
robin java git19:30
pflanze well, I'm not working with java19:31
robin you can invoke java code from non-java programs19:34
robin isn't really serious about the suggestion though19:35
robin otoh if you're really desperate19:39
I made a java component for use by VB because our VB "guru" couldn't read fast enough from a socket19:40
horrors19:40
→ dduncan joined19:44
← dduncan left19:50
→ dduncan joined19:55
→ fultilt joined20:01
→ iano joined20:13
pflanze anyone here from cogito fame (pasky?)?20:26
I've got a tiny patch.20:26
huhh, how comes a "git clone git://git.kernel.org/pub/scm/cogito/cogito.git" leaves me with a cogito calling itself version 0.15 ?20:35
matled what does git describe 24ae4936d2 show?20:38
pflanze cogito-0.18.2-122-g24ae49320:40
fatal: Not a valid object name show20:40
well gitk even shows a tag (I think - yellow kind of flag) with "cogito-0.18.2" further down in the history.20:41
Am I still somehow in the wrong checkout, or is the compiled-in version number plain wrong?20:42
matled where does it tell you it is 0.15?20:42
pflanze The generated debian packages contain this number20:42
in their name.20:42
and debian/cogito/usr/share/doc/cogito/changelog.Debian.gz is from 19 Sep 2005 hm20:43
This is from the debian subdir from the repo.20:43
Maybe debian packages don't use this anymore?20:43
pflanze checks with debian subdir from current etch package20:45
pflanze yeah, debian has moved away from those.20:46
pflanze will just go for local install instead20:46
matled you probably just have to add another entry in debian/changelog (for example with dch -v 0.18.2)20:46
pflanze yeah but what else will be outdated still then?20:46
matled I haven't used cogito in a long time, no idea how well all this is maintained20:47
Randal cg is "mature" :)20:49
it still gets patches20:49
but rarely20:49
I like cg-status better than git-status20:49
matled sure, but the debian subdirectory lacks updates in the version20:49
Randal debian is evil. pure evil. I'm now convinced of it.20:50
pflanze debian people just seem to have moved away from using the cg-provided debian infrastructure.20:50
matled if someone adds a debian subdirectory I think he should keep it up-to-date or remove it20:50
jwb → jwb_gone20:51
matled I could even imagine that they never used it20:51
pflanze Randal: what did convince you20:54
?20:54
Randal just the number of times now that I've heard that they have their own forked version of some public distro20:54
including Perl20:54
matled in case of cogito they just don't use the scripts that are shipped with cogito to create package..20:55
*the package20:55
→ sgrimm joined20:55
pflanze well, it never occurred to me there is any much forking. Sure many packages use some patches for debian-related or not-upstream bug fixes and the like, but I think that's ok.20:56
from the point of a user.20:57
# s/of/of view of/20:57
→ riddochc joined21:04
→ yann joined21:10
yann hi all21:11
yann changed the topic to: 1.5.0.7/1.5.1.1 | Everyone asleep or clueless? Try [email@hidden.address] | Channel log http://colabti.de/irclogger/irclogger_log/git21:11
pflanze hi yann!21:12
I'm remembering you from bigloo times.21:12
yann Anyone knows if it is possible to remove remote refs using git protocol ? I have pushed a couple of things too much on a repo...21:12
hi pflanze21:13
jrockway pretty sure that's not possible21:13
yann pflanze: how the world is small :)21:13
jrockway since someone could have pulled those changes between when you added them and now21:13
you can only (cleanly) rewrite history before you push21:13
OTOH, you could try doing git reset --hard <revision of last good commit> manually21:14
on the shared copy21:14
yann jrockway: that would not be a good reason - I can "git-push --force" anyway (and this is useful for pushing eg. an stgit stack)21:14
my problem is rather that I have pushed all the refs instead of just 2 branches (first time using git-push ;)21:15
well, I can still probably nuke the remote repo and restart the push from scratch, but it would have been useful to be able to cleanup my mess21:16
gitster "git push $URL :refs/heads/i-do-not-want-this-branch-anymore"21:17
You need at least 1.5.0 on both ends, I think.21:17
yann oh, cool - thanks21:17
pflanze How could this happen?: http://pastebin.ca/437073 <- matled: is this a problem with your repository?21:18
matled very strange21:22
I had this error twice and now it is gone21:22
jrockway i see21:23
matled do you still have problems to clone?21:23
pflanze how, it's gone? I'm still getting it.21:23
jrockway (so many terms for the same thing)21:23
matled pflanze: try this one: git://igit.ath.cx/~matled/gitrb21:24
pflanze that one worked, matled21:24
riddochc I'm feeling stupid. I accidentally committed a database dump to a repo. Several versions ago. :p21:25
Fortunately, nobody else has the repo yet. I assume rebase will help me recover from stupidity?21:25
Or, *this* particular thinko anyway? I suppose the *root* cause of the problem requires more work. ;)21:28
matled pflanze: would be interesting to find out what is going wrong anyway, perhaps there is also some bug hidden21:28
james_w riddochc: see the git list from yesterday, Junio posted instructions on how to revise history with rebase.21:30
riddochc Go figure. I'm about four days behind on the list.21:31
Thanks.21:32
pflanze (matled: well sounds plausible but I'm not having oodles of time atm, so someone else would have to track it down.)21:32
gitster http://pastebin.ca/437073 has repository URL wrong. try s|/w/|/r/|;21:32
pflanze matled: does the ruby code reimplement the git functionality from scratch? It seems so.21:32
matled pflanze: yep21:33
pflanze matled: I was expecting it to bind to the c library.21:33
I don't think it's feasible for me to go that route, I need pretty much the full git functionality, "right now".21:34
So I'm still looking for examples which are using libgit, if they exist at all.21:34
matled gitster: ah, the error message is not very helpful :)21:34
riddochc The project in question is a rails-based personal finance double-entry accounting system. As far as version control goes, I've started using git for everything I have control over.21:35
james_w pflanze: you could look at stgit, it has python wrappers for the git command line.21:35
yann it looks like the 1st remote defined is taken as default for both "push" and "fetch", even if it has only "fetch" entries - is that normal ?21:36
riddochc I've also come to the conclusion that svn is a bit less than mediocre.21:36
mugwump (stg uncommit -n X)++21:37
riddochc Besides showing up to cheerlead, I tried using git-svn to clone Bricolage's svn repository, which in turn once used git2svn when they migrated to svn. git svn multi-fetch ran for several *days* before I killed it.21:37
mugwump git2svn?21:37
riddochc Sorry, not git2svn, cvs2svn. Speaking of not thinking...21:37
Brain fuzz.21:38
yann it also looks like a parameter for setting a default remote for git-push could be useful - or did I miss it ?21:39
riddochc It seems to have trouble with svn branches. It starts reading every branch from r1. I haven't looked into whether there's a way to query svn to find the branch point.21:39
I think it downloaded the first few thousand commits several dozen times before even getting to anything more recent than 2004.21:40
I should apologize to their svn host's maintainer. :/21:41
That said, git-svn is working marvelously for me at work, where it's just one trunk, no branches, no tags yet.21:42
mugwump riddochc: use Eric's git-svn branch21:42
git://git.bogomips.org/git-svn.git21:42
normalperson mugwump: it's all been merged into master21:42
a long time ago, even21:43
mugwump oh ok21:43
well, you could mirror with svk and then do the git conversion locally21:43
riddochc That would probably be more efficient, yeah.21:43
Though having the initial fetch find some way to use fast-import would be nice.21:44
mugwump yeah, ideally you'd pipe svn_ra_replay into a converter21:44
riddochc svn_ra_replay? At this point, I know a lot more about git than I do about svn...21:45
normalperson replay support for git-svn would be nice to have, but I've already been sufficiently traumatized from working with libsvn that I'll wait for somebody else to do it21:45
riddochc Hm. Back in 10.21:45
normalperson fast-import support for git-svn, too21:46
mugwump the svn_ra_replay refers to a SVN 1.4+ feature that AIUI lets you just grab the entire revision as some kind of more efficient chunk than the regular file-based-update model21:48
→ freeb joined21:51
← tcoppi left21:53
→ tcoppi joined21:53
freeb I'm getting an error on git-svnimport, while git-svn clone seems to be handling the repository fine. Any ideas? "RA layer request failed: REPORT request failed on '/rails/!svn/bc/1001': REPORT of '/rails/!svn/bc/1001': Response exceeded maximum number of header fields. (http://svn.rubyonrails.org) at /usr/bin/git-svnimport line 955"21:56
that's for git-svnimport -v http://svn.rubyonrails.org/rails and git 1.5.121:56
riddochc Isn't there a git mirror of rails somewhere?21:58
freeb not that I know of - perhaps.21:59
riddochc Because I thought I stumbled across it once. Maybe not. Google isn't helping me much yet today.22:00
freeb as an aside, does anyone know if SVN::Mirror that git-svnimport uses makes use of the repository mirroring api introduced in subversion 1.4?22:01
well git-svn seems to be doing the job fine anyway, just wondering if this is a bug or if git-svn is doing something for me automatically that I should be doing with my usage of git-svnimport22:02
riddochc mugwump: That sounds really useful.22:02
We were just talking about how to improve git-svn, yeah.22:02
freeb I wonder if it would have much of an impact on speed. I was going to see how fast svnsync -> local repository -> import via git-svn or even using the svn importer from fast-export.git, seeing as I can't seem to just svnsync $REMOTE $DEST or anything remotely similar, I can't be bothered with it right now.22:08
→ roxana joined22:08
roxana Hello22:09
I was a student that didn't make it in GSoC 200722:09
my project was to enhance egit22:09
If you are allowed please give me some feedback about my proposal22:10
I don't want the rank, just some informations about what was the reason I didn't got accepted22:11
freeb roxana: while you're waiting for someone to answer, have you seen this? http://marc.info/?l=git&m=117639594224235&w=222:12
It seems that git was unlucky to be given so few projects.22:13
roxana thanks for the link to the message22:14
it's unfortunate that git got so few slots22:14
mugwump You were very close, constantin22:14
one more slot and you might have been in22:15
roxana but, it is possible to know if say Google would have gave 10 slots, I would be in GSoC now?22:15
mugwump People were very impressed by your fractal plug-in22:15
roxana or my proposal and skills should be trainned a little bit more :)22:16
ok, thanks for the feedback22:16
mugwump perhaps there will be a development house using eclipse and git who would be interested in sponsoring the work22:18
roxana what do you mean by "development house"?22:19
riddochc That would've probably tipped the balance from svn to git, where I am.22:19
But we don't do primarily development work.22:20
sgrimm I'm very happy to see the Perl/shell -> C project got accepted, though.22:22
I wonder if that will include git-svn.22:23
mugwump You're using the wrong article in that statement, sgrimm22:23
*a* Perl/shell -> C project got accepted22:23
sgrimm True, true22:23
freeb I've wondered about a pure C git-svn22:24
mainly because svn-fast-export.c looked fairly clean22:24
sgrimm The Perl git-svn is extremely useful (I probably wouldn't be using git every day without it) but it chews tons of CPU time when I'm importing a large repo, and takes forever to finish.22:25
And I have to admit I have a hard time following the code, though that's probably mostly me not being a Perl expert.22:25
freeb I've switched to git-svn from svk. For my usage, it's much better. However, I never really deal with complex merge situations or anything too non-standard22:26
sgrimm It actually deals with complex merges fairly well. I've used it to merge two svn branches because git's merging is so much better than svn's.22:27
mugwump well, that's lucky, because if you try that stuff on svk it presents the illusion of working22:27
sgrimm (Not that two branches is all that complex, of course.)22:27
→ zedek joined22:28
sgrimm I just wish it paid attention to .git/info/exclude when importing from svn.22:28
mugwump roxana: by "development house" I mean a company who do lots of development for many different clients22:28
roxana mugwump, ok I get it, thanks22:29
riddochc Well, I'm off for a while. Seeya, folks.22:34
freeb anyone have experience with tailor?22:34
sgrimm freeb: A tiny bit22:36
mugwump hey Subversion got 4 slots!22:43
Silverstripe got all 10 they asked for22:43
Siggy's gonna be busy22:44
freeb anyone know how the ikiwiki git backend works - I mean a wiki seems to map obviously to a single-file format like sccs or rcs22:49
→ orospakr joined23:00
→ jbellis joined23:14
← jbellis left23:14
→ DrNick joined23:15
aeruder_ freeb: yea, i'm a svk deserter for git-svn23:19
→ orospakr joined23:24
mugwump really must get that article published23:26
→ tcoppi joined23:27
→ Gitzilla joined23:28
→ spearce joined23:41
→ yorgen15 joined23:43
spearce yea, those 4 slots for svn - they must have had more students apply than we did.23:47
→ alb joined23:48

Logs Search ←Prev date Next date→ Channels Documentation