IRCloggy #git 2008-12-05

Provider of IRC logs since 2005.

WARNING: As Freenode became unjoinable and lost all warnings in topics, we cannot log channels on Freenode anymore.

	2008-12-05
	→ ibuffy joined	00:01
offby1	doener: exactly -- if there's nothing to stash, Something Bad happens.	00:01
	lemme see the current version	00:01
	hmm, I'm sure I whined about that more recently than May	00:02
doener	no, with "&&" it still fails	00:02
	stash does not set an exit code != 0 when there's nothing to stash	00:02
	← bbuss left	00:03
offby1	yup	00:03
	nor, I suspect, should it	00:03
doener	right	00:03
offby1	But it'd be nice if it emitted predictable output	00:03
	maybe it does	00:03
	anyway.	00:03
	now we know.	00:03
doener	well, "git diff --quiet HEAD && git stash && git stash apply" should do	00:04
gitte	offby1: if it were plumbing, yes, I'd like it to have predictable output.	00:04
	← LiamH left	00:04
doener	or rather, "(git diff --quiet \|\| git diff --cached) && git stash ..."	00:04
	gargh... the second diff should also have --quiet	00:04
	← jackdempsey left	00:05
doener	and it needs to be " \|\| git stash"	00:05
	doener should just stop trying	00:05
	→ bbuss joined	00:06
	← cilly left	00:06
doener	ok, "git diff --quiet && git diff --quiet --cached \|\| (git stash && git stash apply)	00:08
	that should do...	00:08
	← kuadrosx left	00:09
	← edwardam left	00:12
	← willb left	00:12
offby1	doener: ooh, clever	00:13
	→ hidek_ joined	00:13
doener	now I wonder if there was sarcasm in that... That's why I sometimes hate text-only conversations ;-)	00:14
offby1	there was no sarcasm.	00:15
doener	ok :-)	00:15
offby1	in fact I just send a memo to johnw suggesting he incorporate that	00:15
doener	thanks!	00:15
	← etherael left	00:16
	← gilimanjaro left	00:17
	→ ben_h joined	00:18
	← lolage0 left	00:20
	→ geckosen1tor joined	00:23
	→ Flathead joined	00:23
	→ selckin joined	00:25
	← gitte left	00:26
	→ sgarza joined	00:27
	← sgarza left	00:28
	← ankit9 left	00:28
	→ ankit9 joined	00:28
	← lresende left	00:29
	← branstrom left	00:31
	← ankit9 left	00:37
	← bdiego left	00:37
	→ ankit9 joined	00:37
	→ branstrom joined	00:37
	→ jstad_ joined	00:38
	← drewr left	00:39
	← aroben left	00:41
	→ gretch joined	00:44
	← ph^ left	00:45
	← __ash__ left	00:45
	← Yuuhi left	00:48
	→ edwardam joined	00:49
	→ univate joined	00:50
	← supercali left	00:51
	→ supercali joined	00:52
	← branstrom left	00:52
	→ qrush joined	00:54
	mw → mw\|out	00:55
	← hobodave left	00:55
	→ Jippi joined	00:55
	→ hobodave joined	00:56
Jippi	Hi. How do I delet a remote branch completely ? I have tried git branch -r -d 1.0 and git bracnh -D 1.0 - but its still on the server interface (unfuddle)	00:56
doener	git push <remote> :<branch>	00:56
	that pushes "nothing" into the branch, and thus deletes it	00:57
	"git branch" is purely local	00:57
	IIRC "git push" is the only command that modifies a different repo than the one you're in	00:57
	← markelikalderon left	00:58
Jippi	ah - Thanks !	00:58
ben_h	doener: so it's the colon that specifies you want to push nothing?	00:58
	or do you first have to delete it locally?	00:58
	← gretch left	00:58
	→ metajack_ joined	00:58
doener	ben_h: no, the fact that there's nothing on the left side specifies that you push "nothing"	00:59
	ben_h: no, no need to delete it	00:59
ben_h	ahh right	00:59
Jippi	doener: is it possible to retrive the tree again once deleted?	00:59
	*branch	00:59
doener	ben_h: the full refspec format is <committish>:<ref>	00:59
ben_h	hmm	00:59
doener	ben_h: <branch> is DWIMmed to refs/heads/<branch>:refs/heads/<branch>	01:00
ben_h	so something like git push origin HEAD:branch	01:00
doener	ben_h: same for a tag, but with refs/tags	01:00
ben_h	ahhh right	01:00
doener	the default refspec for push is just :	01:00
ben_h	hmm that explains a lot	01:00
doener	and that means "look for branches that have the same name in this repo and the one I'm pushing to, and update them"	01:00
	→ __ash__ joined	01:00
	→ miggilinme joined	01:01
	← pcapriotti left	01:01
doener	Jippi: the _tree_ might still be referenced by other commits. The _branch_ can be resurrected by pushing it to the other repo again.	01:01
miggilinme	(Very new to git) I just pulled a branch from the git server, but when I enter "git branch" it doesn't show it	01:01
	is there something I need to do to add the nre branch?	01:01
	new*	01:01
Pieter	miggilinme: git branch -r	01:02
cbreak	if you pull from a branch	01:02
	it usually merges into the current branch	01:02
doener	Jippi: whether you still have the respective commit depends on whether you still have the branch locally. If so, no problem. If not, there's a chance that the reflog for HEAD still has an entry for that commit	01:02
miggilinme	Pieter: still not showing	01:02
	cbreak, that's odd, because I don't see the new files	01:02
	→ altfive joined	01:02
cbreak	well. I am not sure, I never did it myself :)	01:03
doener	miggilinme: "git pull <remote> <branch>" means "fetch this branch and merge it into the one I have checked out"	01:03
	miggilinme: what's the exact command you used?	01:03
	→ jepler joined	01:03
	← elmex left	01:03
miggilinme	git pull server experimental	01:03
doener	Jippi: and if even the reflog doesn't have it anymore, and it has not been merged into another branch (which would of course keep the commit alive anyway), the objects might still be around, but dangling	01:04
	→ lukedal joined	01:04
miggilinme	doener: if you say so, I can check again to make sure it wasn't merged, I could be wrong	01:04
	doener: ah, there it is	01:04
cbreak	do a git log	01:04
miggilinme	that wasn't what I had hoped for however	01:04
cbreak	it should show	01:04
doener	miggilinme: you can reset back to before the merge	01:04
Jippi	doener: okay. Thanks :)	01:04
miggilinme	doener: how?	01:04
ben_h	miggilinme: you could pastie the output of the pull command for us to look at	01:05
miggilinme	and then how do I get the branch as a new branch?	01:05
doener	miggilinme: check "git reflog show <your_current_branch>"	01:05
miggilinme	ben_h, i did it yesterday, long gone	01:05
doener	miggilinme: likely <your_branch>@{1} (the second entry} is the commit where you have been before the merge	01:05
cbreak	to get a remote branch, you would first fetch it, and then checkout -b it?	01:05
ben_h	miggilinme: i would first create a local branch and then pull into it.	01:05
	might be a quicker way though.	01:05
doener	miggilinme: then you can do "git reset --hard <your_branch>@{1}"	01:05
	miggilinme: now to your other problem. Do you want to repeatedly fetch (all) branches from that other repo, or just once?	01:06
pastie	ben_h: http://pastie.org/331479 by miggilinme.	01:06
doener	miggilinme: ah, so you pulled and then made a commit?	01:07
miggilinme	yes, but that commit is borderline useless	01:07
	I'm a little confused by the sudden flow of responses, not sure what I should do	01:07
ben_h	miggilinme: panic	01:07
doener	ok, if it's useless, just "git reset --hard master@{2}". You'll be back at the initial commit then, ie. before the merge	01:07
miggilinme	should I reset to the initial commit?	01:07
	ok	01:08
	doener goes to order THHGTG for ben_h	01:08
ben_h	lol	01:08
miggilinme	ok, great, back to the head	01:08
	and then create the experimental branch, switch to it and then pull?	01:08
	← parasti left	01:08
doener	miggilinme: no :-)	01:08
	miggilinme: first, answer my question about your usage pattern for the other repo	01:09
	← zerok left	01:09
	← supercali left	01:09
miggilinme	what then?	01:09
doener	miggilinme: is that a repo from which you intend to fetch "regularly"?	01:09
	alb → albertito	01:09
miggilinme	well yes, I imagine so	01:09
	it's my project :)	01:09
	← withanx left	01:10
doener	miggilinme: then you'll want to have a "remote" for that repository	01:10
	miggilinme: git remote add <name> <url-to-repo>	01:10
miggilinme	ejsy	01:10
	erm, sorry	01:10
	what's that for	01:10
	?	01:10
doener	miggilinme: that adds a "remote" section to the config. One aspect is that you can use the name of the remote as an alias for the url	01:10
	miggilinme: the other is that the remote gets a default "fetch refspec"	01:11
miggilinme	in that case I think I do have that	01:11
doener	you already have a remote for that repo?	01:11
miggilinme	and a fetch refspec is?	01:11
	yes	01:11
ben_h	I was thinking about the implementation of git's hashing the other night. Anyone here familiar with it?	01:11
	→ xaiki joined	01:12
ben_h	e.g. what is it that gets hashed to product a commit id - the literal patch, or the concatenated objects, or what?	01:12
	→ Bass10 joined	01:12
doener	"git fetch" (which does the fetch part of "git pull") takes a remote repo and some refspecs as arguments	01:12
	← lukedal left	01:12
doener	miggilinme: the refspec tells it what to fetch and where to store it	01:12
	miggilinme: the default refspec for a remote fetch all branches from the other repository and stores them in refs/remotes/<remote>/<branch>	01:13
miggilinme	doener: so I want to make a refspec to tell it to fetch experimental and store it in a new branch called experimental?	01:13
doener	miggilinme: the default (fetching all branches) should be fine. And you'll also want the default behaviour of storing the stuff in so called "remote tracking branches"	01:13
miggilinme	so I want to enter "fetch refspec"?	01:13
doener	miggilinme: those branches just represent how a branch in another repo looks	01:14
	miggilinme: what does "git remote" tell you?	01:14
miggilinme	<servername>	01:14
doener	ok, then "git fetch <servername>"	01:14
	that should fetch the branches from <servername> and store them in refs/remotes/<servername>/<branch>	01:15
	eg. refs/remotes/<servername>/experimental	01:15
miggilinme	so what's the difference between this and pull?	01:15
	I mean, it gets the branches I guess	01:15
	← univate left	01:15
doener	pull is fetch + merge	01:15
miggilinme	but more technically	01:15
	ah, ok	01:16
	so this is no merge	01:16
doener	so pull is just a shorthand, pure convenience when you often just merge a branch	01:16
miggilinme	got it	01:16
doener	and fetch just gets the stuff and stores it somewhere	01:16
	ok, now after the fetch, you should get some output from "git branch -r"	01:16
miggilinme	ok, it tells me now that it added the branch	01:16
doener	the -r means that "git branch" lists the remote tracking branches	01:16
	← __ash__ left	01:16
miggilinme	yes, I see the branches listed	01:17
	ok, I guess I can figure from here then	01:17
	thanks	01:17
doener	ok, now, you can use those remote tracking branches "as usual". You can for example do "gitk <servername>/experimental"	01:17
	→ markelikalderon joined	01:17
doener	or "git log master..<servername>/experimental" to see what the remote branch has, that your master branch doesn't	01:18
miggilinme	alright, thanks	01:18
doener	you're welcome	01:18
	← geckosen1tor left	01:18
	→ univate joined	01:19
	→ vbabiy joined	01:20
	← cgardner left	01:22
	→ geckosenator joined	01:23
	→ gretch joined	01:26
	← miggilinme left	01:27
	→ ebzzry joined	01:28
	→ agenteo joined	01:30
	← xaiki left	01:31
	← scsiraider left	01:33
	→ scsiraider joined	01:34
	← hidek_ left	01:34
ben_h	flazz: yep - http://gems.github.com/	01:35
	bah wrong chan	01:35
flazz	ben_h: thanks	01:36
ben_h	:)	01:36
	← edwardam left	01:36
	→ edwardam_ joined	01:36
doener	ben_h: for a second, I thought you were psychic or something ;-)	01:36
ben_h	doener: i am, but that's unrelated	01:37
	→ bobmcw joined	01:37
doener	:-)	01:37
ben_h	doener: lol, your comment before + that reminded me of this: http://en.wikipedia.org/wiki/List_of_minor_characters_from_The_Hitchhiker's_Guide_to_the_Galaxy#Dr._Dan_Streetmentioner	01:38
jast	ben_h, how many fingers am i holding up right now then?	01:38
ben_h	jast: π	01:38
jast	not quite, i'm afraid	01:38
ben_h	:)	01:38
flazz	ben_h: so say i have a gemspec, is there a way to have git put the current tag in the version field?	01:39
doener	flazz: git describe	01:39
pfharlock	how does one do the equivalent of bzr missing in git?	01:39
ben_h	not sure sorry flaz	01:39
doener	pfharlock: what does "bzr missing" do?	01:40
pfharlock	sorry, it looks at two branches and will display any revisions missing from one or the other	01:40
doener	pfharlock: git log <branch1>...<branch2>	01:40
flazz	doener: thanks	01:40
	← ebzzry left	01:41
doener	pfharlock: three dots mean "commits reachable from one side or the other, but not both"	01:41
	pfharlock: you'll probably want --left-right with that, to see from which side each commit comes	01:41
	→ ebzzry joined	01:41
doener	pfharlock: especially nice with gitk	01:41
pfharlock	cool, thankyou very much, I'll give it a test drive	01:42
doener	pfharlock: to see stuff that's only on one side use A..B == B --not A == stuff reachable from B but not from A	01:42
	→ sdboyer_ joined	01:42
doener	(that's often more useful if you, for example, want to see what you're going to merge from some other branch)	01:43
	← univate left	01:43
albertito	I've got a repository on machine A, where I do (very minor) kernel development. I have another repository (this one bare) on machine B which I use for backups. The upload link from A to B is very slow. The link between B and kernel.org is very good (continued below)	01:44
offby1	do go on	01:44
	→ univate joined	01:45
albertito	Now, instead of A pushing everything to B, I want B to get everything it can from kernel.org, so A pushes as little as possible	01:45
offby1	sure	01:45
doener	albertito: fetch in the bare repo	01:45
albertito	I obviously cloned from k.o, worked great	01:45
	Pushed once, worked great. But the fetch seems to make no difference	01:45
	It wants to transfer what I think it's too much. But I'm not quite sure, and I don't know how can I tell	01:45
	Any ideas on how can I see some list of ids it wants to transfer, or something like that, so I can get a clue on what might be going on?	01:46
doener	albertito: how did you fetch in the bare repo?	01:46
	albertito: did you add a remote for k.o?	01:47
albertito	doener: git fetch git://k.o	01:47
	doener: no	01:47
doener	that would store stuff in FETCH_HEAD, not sure if is considered for pushes to determine what needs to be pushed	01:48
albertito	doener: I thought it could make a difference, if it didn't took FETCH_HEAD into account. Should I?	01:48
	→ harinath joined	01:48
albertito	doener: =)	01:48
	doener: I'll try the remote	01:48
	→ PhilK joined	01:48
offby1	I'd be very surprised if refs on the bare repository mattered. I'd assume that "git push" simply asks the remote repository what objects it has, and only pushes those that it doesn't already have.	01:50
	whether there are any refs pointing to those remote objects seems, to me, to be irrelevant	01:50
albertito	doener: it worked, thanks! However, now I have a "linus/master" branch on my backup repo. I'm not worried about it because it's a backup, but it doesn't seem like it should be this way	01:50
	← PhilK left	01:51
albertito	offby1: maybe it needs some tip where to start looking for things	01:51
doener	offby1: you need something efficient to figure out which objects the remote has	01:51
	offby1: I guess it just iterates the branches/tags	01:51
offby1	hmm	01:51
	bummer	01:51
doener	looking at the code right now	01:51
offby1	this question could easily be resolved by an experiment. But I'm too lazy to do it :)	01:51
albertito	offby1: in any case, it worked as expected adding the remote. But I think it should consider FETCH_HEAD too, unless I'm missing something (which is probably the case =)	01:51
doener	albertito: you could also do "git fetch k.o master:linus" to have a "linus" branch instead	01:52
offby1	well, as you can see, _I_ think it should consider _all_ objects	01:52
albertito	doener: good idea. However, in this particular case, having a remote _is_ conveniant to me, I'll just do git remote update before pushing, it's easier than copy-paste the k.o url	01:53
doener	albertito: you can change the "fetch = " line in the config for that remote	01:53
albertito	offby1: that was what I was expecting in the first place, and when it seemed not to behave that way I asked here =)	01:53
	← edwardam_ left	01:53
	→ ebzzry_ joined	01:54
	← blithe left	01:54
doener	offby1: hm, how would you look for common objects then? Bisecting the whole local history asking the remote for each one? (Just wondering, brain doesn't spit out anything smart for me)	01:54
albertito	offby1: thanks! I'm actually happy (in this case) with the "linus/" prefix, but I'll keep that in mind	01:54
	← harinath left	01:54
	→ perezd88 joined	01:55
	→ harinath joined	01:55
	← selckin left	01:55
	→ blithe joined	01:55
offby1	doener: I dunno. "I have <<very long list of objects>> ... tell me which you want"	01:56
	deskin wonders if it'd be relatively easy to use all commits which aren't ancestors, instead of just refs	01:56
deskin	though that doesn't feel quite right; some refs are could be duplicates, but maybe it really just ignores the overhead	01:57
	← harinath left	01:57
albertito	deskin: aren't those non-trivial to determine?	01:57
	← sdboyer left	01:57
albertito	bbiam	01:57
deskin	probably :/ I'm trying to think of a sane way	01:57
	→ harinath joined	01:57
deskin	which doesn't involve looking at all history	01:57
	doener is still trying to find out what it actually does now	01:57
offby1	maybe if you just run 'git-upload-pack' and examine the output ...	01:58
doener	yeah, but there's also the non-fast-forward case, I wonder if that does some smart cross-repo merge base search then	01:59
deskin	isn't this find_common()? I recall looking at Thomas Rast's log-n transmission patch, but that was in my early days of git development, and that patch is deep magic	02:00
	← Zimm left	02:01
doener	ah, thanks! Been looking at the wrong end	02:01
offby1	that's what my doctor said	02:02
	rimshot	02:02
ben_h	lol	02:02
jast	http://instantrimshot.com/	02:02
	→ threeve joined	02:03
deskin	anyone know thomas's IRC handle? IIRC he's been on here but I don't know if he frequents #git	02:03
	← joydivider left	02:04
	← jstad left	02:04
	← jstad_ left	02:04
	← scsiraider left	02:05
flazz	is it safe to set spec.version = `git describe` ?	02:06
	→ scsiraider joined	02:08
	← ebzzry left	02:08
	← softdrink left	02:08
	ToxicFrog` → ToxicFrog	02:09
	→ softdrink joined	02:09
	← ankit9 left	02:10
	← harinath left	02:13
	→ harinath joined	02:13
	← unreal left	02:14
	← ahupp\|fb left	02:15
	→ unreal joined	02:15
	→ ahupp\|fb joined	02:16
	→ sdboyer-laptop joined	02:18
	← softdrink left	02:19
	← bobmcw left	02:20
	← stephens left	02:21
	← carllerche left	02:21
	→ Richo99 joined	02:23
	← ciaran left	02:23
Richo99	Hi, I want to move a commit on main to a branch, how do I do that?	02:23
	→ _VVSiz_ joined	02:24
	← perezd88 left	02:25
	→ perezd88 joined	02:25
	← mw\|out left	02:25
	→ lordpil joined	02:26
	← markelikalderon left	02:27
doener	deskin: ok, out of boredom, I did some mail/irc log search. Seems that his handle is "charon"	02:27
	→ ezyang joined	02:27
deskin	ah, that sounds rigth	02:28
doener	deskin: http://colabti.org/irclogger/irclogger_log/git?date=2008-08-29,Fri#l890	02:28
deskin	not on now unfortunately; though I wouldn't expect it	02:28
doener	that's how I found out :-)	02:28
deskin	nice	02:29
	← Flathead left	02:29
deskin	/whois for the win; but memory for the lose in my case	02:29
	→ Flathead joined	02:30
	deskin needs to write that IRC bot that learns people's nick/ realname mappings so deskin doesn't have to remember	02:31
	→ ahupp\|fb_ joined	02:32
	← lordpil- left	02:39
	→ ceej joined	02:39
ceej	hey all, whats the command to roll back one as the last git update just messed things up	02:40
	← ijcd left	02:40
	→ Tv joined	02:40
	← VVSiz_ left	02:41
geckosenator	ceej: revert	02:41
ceej	do i need to enter a parent number though?	02:42
	_kukks_ → kukks	02:42
ceej	no way to go backj just one without looking that up?	02:42
	:/	02:43
doener	ceej: do you mean "git commit" by "git update"?	02:43
ceej	No I just pulled the latest master to my production server	02:44
	and now it's erroing out, is there away to go back to the previous state?	02:44
	cbrake → cbrake_away	02:44
doener	if you didn't do anything else yet, "git reset --hard master@{1}"	02:44
	but it's safer to first lookup the right reflog entry, "git reflog show master"	02:45
	← jesselucas left	02:45
doener	(both commands assuming that you pulled into the master branch)	02:45
ceej	ok, thank you	02:45
	that didn't appear to go back one	02:46
	still the same as master	02:46
	← ahupp\|fb left	02:46
doener	which one didn't go back?	02:46
	← flaguy left	02:46
	→ neuro_damage joined	02:47
	← bbuss left	02:47
neuro_damage	how do I have the same repo hosted at multiple different locations	02:47
ceej	hmmm still erroring out :/	02:47
doener	ceej: so you had "master" checked out, did a "git pull" and then "git reset --hard master@{1}", right?	02:48
ceej	ye	02:48
	← ebzzry_ left	02:48
	← gambler left	02:48
doener	well, that should have resetted master to before the merge and also made the working tree equal to the tree for that commit	02:49
	ceej: "git branch" actually says you have master checked out right now, right?	02:49
ceej	yes	02:50
	it gives me messages like HEAD is now at c1a4266 deleted files that are not suposed to be here.	02:50
	← neuro_damage left	02:50
doener	so master was reset to the commit	02:50
	note: master@{1} _changes_ when you do a reset...	02:51
	use "git reflog show master" to lookup which reflog entry is the one that refers to the pre-pull state	02:51
	most likely, it's the first one _after_ the first "pull : ..." line	02:52
ben_h	qrush: lolol	02:52
	oh god damn it	02:52
	qrush: lolol	02:52
	← sgrimm left	02:52
	ben_h apologises	02:52
doener	and then use that entry with "git reset --hard master@{n}", replacing n with the right number	02:52
ceej	doener:your correct but it doesn't seem to be putting the files back to the privous state	02:53
qrush	ben_h: what	02:53
ben_h	qrush: #github, i failed to select the right chan	02:53
qrush	i saw ;)	02:53
ben_h	:)	02:53
doener	ceej: but you used --hard, right? Does "git log -1" show the right thing? And if so, is "git status" clean?	02:53
	→ Tv1 joined	02:55
doener	ok, 4am, off to bed... good luck :-)	02:55
ceej	doener: thank you :)	02:56
	← kukks left	02:56
albertito	doener, offby1: I'm off to bed, thanks a lot!	02:58
	← unreal left	03:00
offby1	yay	03:01
	→ unreal joined	03:01
	→ qrush_ joined	03:02
	← Tv left	03:03
	← unreal left	03:04
	← bts- left	03:05
	→ unreal joined	03:06
	→ quoin joined	03:07
	← brett_h left	03:18
	← spearce left	03:20
	← vbabiy left	03:22
	← ahupp\|fb_ left	03:23
	→ loincloth joined	03:24
	→ EmilMedve joined	03:27
	← univate left	03:32
	← unreal left	03:34
	→ unreal joined	03:35
	← pygi left	03:38
	→ Mario__ joined	03:38
	→ ben_h_ joined	03:38
	→ imyousuf joined	03:39
	← rtomayko left	03:41
	→ nice_burger joined	03:42
	← agile left	03:43
	→ agile joined	03:46
	→ nothingHappens_ joined	03:46
	← ben_h left	03:46
	→ gambler joined	03:48
	→ ceej_ joined	03:48
	→ EiNZTEiN joined	03:48
	→ __ash__ joined	03:55
	← pfharlock left	03:56
	← ceej left	03:56
	→ offby1` joined	03:58
	← schlort left	04:02
	→ schlort_ joined	04:02
	→ univate joined	04:07
	← Tv1 left	04:11
	← Modius left	04:11
	← offby1 left	04:13
	→ IRSeekBot joined	04:15
	← qrush_ left	04:15
	offby1` → offby1	04:20
	← schlort_ left	04:20
	← qrush left	04:21
	→ nikolasco joined	04:23
	→ jonshea joined	04:24
	→ ahupp\|fb joined	04:25
	← k0001 left	04:25
	→ ebzzry joined	04:26
	→ schlort joined	04:28
	→ muthu_ joined	04:29
	← scientes left	04:33
	→ fhobia joined	04:36
	→ ben_h joined	04:38
	← nothingHappens_ left	04:40
	← dreiss left	04:41
	→ ben_h- joined	04:41
	→ girishr joined	04:42
	← ferdy left	04:42
	→ sgrimm joined	04:44
	→ Tv joined	04:46
	→ carllerche joined	04:48
	→ ferdy joined	04:51
	← Flathead left	04:52
	← murphl left	04:52
	→ jackdempsey joined	04:53
	→ qrush joined	04:54
	→ cedricv joined	04:55
	← Bass10 left	05:03
	→ mithro joined	05:03
	→ cgardner joined	05:08
	← cgardner left	05:09
	→ cgardner joined	05:09
	→ tjafk2 joined	05:10
	← sdboyer-laptop left	05:12
	→ tpiep joined	05:14
	→ girishr_ joined	05:14
	→ murphl joined	05:16
	← girishr left	05:16
	← crab left	05:17
Ryan52	offby1: what did I do wrong here? http://slexy.org/view/s20n55kjaA	05:22
	← bartman left	05:22
offby1	hole on	05:23
Ryan52	offby1, I didn't actually test what you told me the other day :p	05:23
offby1	that's OK, I have no recollection of what I told you the other day	05:23
	and am too lazy to look in my logs	05:23
	← tpiep left	05:23
offby1	ah, that's easy.	05:23
	you forgot to prepend the directory name: f6	05:24
	git cat-file -t _f6_53bbeb8482959d6131424204e57c9b75bf9b1b	05:24
	so that's the 25 Mb blob, eh?	05:24
Ryan52	ah. you didn't explain that part to me :)	05:24
	yup. but there are more :p	05:24
	hrm. fatal: Not a valid object name _f6_53bbeb8482959d6131424204e57c9b75bf9b1b	05:25
offby1	DUDE! Isn't it OBVIOUS that SHA1s are always 40 characters?	05:25
	I mean REALLY	05:25
	Ryan52: the _ were just for emphasis!	05:25
Ryan52	ohhhh.	05:25
	hehe :p	05:25
	offby1 whistles innocently	05:25
	← Mario__ left	05:25
Ryan52	you think that I was actually smart enough to count the number of characters? :p	05:25
	well, obviously I wasn't.	05:26
	but I mean a normal person :)	05:26
	hrm. blob. that was helpful.	05:26
	reads man page	05:26
offby1	that just means it's a file.	05:26
	← tjafk1 left	05:26
offby1	git show f653bbeb8482959d6131424204e57c9b75bf9b1b > wot-is-it	05:26
	→ loinclot_ joined	05:26
offby1	less wot-is-it	05:26
	it could have been a commit, a tag, or a tree.	05:27
	(Although all of those are _very_ unlikely to be 25 Mb.)	05:27
Ryan52	hrm. looks like a binary file...what do I do now? :)	05:28
	← harinath left	05:28
offby1	file wot-is-it	05:28
Ryan52	I need to figure out what the file name is so I can filter-tree it out.	05:28
offby1	that'll try to guess what it was	05:28
	oh, the name shouldn't be hard.	05:28
	just find every tree object in your repository, and then run "git ls-tree" on it, and grep for the SHA1 ... :-\|	05:29
	I hope there's an easier way but that's all I can think of	05:29
	→ harinath joined	05:30
	→ not-xjjk joined	05:30
Ryan52	hrm. okay, I guess that's fairly easily scriptable...thanks.	05:31
	→ zwu joined	05:31
	← xjjk left	05:31
	→ joshdavey joined	05:31
	← loincloth left	05:31
	← sdboyer_ left	05:34
	→ sdboyer_ joined	05:35
	sdboyer_ → sdboyer	05:35
	Ryan52 filter-branches	05:35
feindbild	does 'git log -S' work on character bases for the changes it matches?	05:36
	s/bases/basis/	05:36
offby1	as opposed to byte? I doubt it.	05:36
	git seems aggressively byte-oriented	05:37
	that's the sort of question that could be easily resolved with a simple experiment, I'd think	05:37
feindbild	no ... sorry ... I meant as opposed to diff lines	05:37
offby1	now I _really_ don't understand.	05:37
feindbild	:/	05:39
	ok, take this as example pseudo-diff: http://pastebin.ca/1276624	05:40
	→ context joined	05:40
context	hmm, is it easily possible to re-write history removing the git-svn-id's from commit's (yes i realize all hashes change)	05:41
	← pete__c left	05:41
tpope	context: it's possible. it's easier to just disable them on import	05:42
offby1	feindbild: ok, I've looked at that, but ... it doesn't help me understand your question	05:42
feindbild	it seems: git log -S'12346' finds the commit, but I want to do: git log -S'this.getById' - which doesn't ... I want to match the whole line of the diff ... if that is possible :(	05:42
	→ pete__c joined	05:42
offby1	really? git log -S'this.getById' doesn't find that commit?	05:42
	hmm	05:42
	→ Marmouset joined	05:44
	← girishr_ left	05:44
offby1	interesting	05:44
	I think I'm starting to understand feindbild's question	05:44
	and have no answer.	05:45
	all I can say is: maybe it internally stores diffs, not as the line-by-line changes that we're used to from the output of "diff", but rather in some more-efficient format ... and if a particular commit's "diff" doesn't mention part of the line, then you won't find the commit by spelling out that line	05:47
	so what you want to do is: find all commits that changed any line that itself includes "this.getById", even if the change itself didn't affect that particular bit of the line.	05:47
feindbild	yes	05:48
offby1	you'll probably have to get each diff separately, and grep 'em :-\|	05:48
feindbild	:D	05:48
	yay!	05:48
offby1	maybe --pickkaxe-regexp works differently ... tried it?	05:48
feindbild	not yet ^^	05:49
	→ Rivelli joined	05:49
offby1	don't think it'll help	05:49
	sounds like a question for a real expert.	05:50
	→ hobodave_ joined	05:51
	→ AaronLBorg joined	05:51
	← __ash__ left	05:52
AaronLBorg	noob git question alert.......	05:52
	that noob would be me.	05:52
	offby1 races to the "noob git question" station and readies himself	05:53
AaronLBorg	ok....	05:53
offby1	Shields engaged, sir!	05:53
AaronLBorg	now you got me all nervous!	05:53
offby1	RELAX!!!!	05:53
AaronLBorg	....make me forget my question......	05:53
	ok...	05:53
	offby1 helpfully blows an air horn	05:53
AaronLBorg	i have a site on a remote server	05:53
offby1	mm hm	05:53
AaronLBorg	[air horn snaps him outuvit]	05:54
	offby1 assumes the psychoanalyst pose, behind the couch, with a pad and pen	05:54
offby1	do go on.	05:54
AaronLBorg	it is one of many sites	05:54
	and all of the other sites have git installed	05:54
	i would like to install it in MY site now	05:54
	and I need to know the best way to do that server-wise	05:55
	offby1 scratches head	05:55
	→ priidu joined	05:55
AaronLBorg	is it any different than a local install?	05:55
offby1	what does "server-wise" mean here?	05:55
	Does it mean you cannot get a shell on your site?	05:55
AaronLBorg	ah ha....doc. you ask great questions!	05:55
	yes, i got a shell	05:55
offby1	please, let's keep the focus on you.	05:55
AaronLBorg	rather, i have access to a shell	05:55
	yes......after all...it is my session	05:56
	offby1 turns his family portrait over, face-down	05:56
AaronLBorg	OR..............	05:56
	offby1 subtly gives his receptionist the signal that means "in two minutes, knock and tell me I have an important phone call"	05:56
AaronLBorg	[deletes stuff]	05:56
	can i just download a tar locally and then scp up there and untar it?	05:57
	seems likely, doc	05:57
offby1	seriously, I'm not sure what the problem is.	05:57
	← ezyang left	05:57
AaronLBorg	well, would that work [says sheepishly]	05:57
	?	05:57
offby1	The usual way to get something via git is to type "git clone url://server.com/path/to/git-repository"	05:57
AaronLBorg	is it important where that repository goes? i'm assuming it is	05:58
offby1	that's up to you.	05:58
	→ ezyang joined	05:58
offby1	or to the software running on your server.	05:58
AaronLBorg	well, if this is a drupal site, right?	05:58
offby1	if it's a web site, I assume it needs to go in /var/www or someplace similar	05:58
	I've never used drupal but I have a vague notion of what it is.	05:58
AaronLBorg	yes, that's what i'm thinking	05:58
offby1	so ... let me get this straight	05:58
	there are a number of sites, all running drupal, and all serving up the _same_ site.	05:59
AaronLBorg	but lets say my site is located in /var/www/gitidiot.com	05:59
offby1	Or at least, close variants of the same site	05:59
	sure, ok	05:59
AaronLBorg	no serving up different sites!	05:59
	lemme clear this out.........	05:59
offby1	are you using git to keep a bunch of separate servers in sync?	05:59
AaronLBorg	no!	05:59
offby1	oh	05:59
	then	05:59
	what?	05:59
AaronLBorg	this is a server that has many different sites on it	05:59
	← context left	06:00
offby1	oK	06:00
	like a hosted server.	06:00
	one where some company runs it, and you install your little site on it, right next to 200 other people, each with their own sites.	06:00
AaronLBorg	like this: /var/www/public_html/gitidiot.com AND /var/www/public_html/idontknowgit.com ETC....	06:00
offby1	sure	06:00
AaronLBorg	yes, but i work at the company ;-)	06:00
	never had to install git myself	06:00
offby1	ok, so you're the hosting guy	06:00
AaronLBorg	i'm a dev	06:01
offby1	well, _installing_ git is easy	06:01
AaronLBorg	hangs head in shame	06:01
	→ girishr_ joined	06:01
offby1	likely your *nix distro has a package.	06:01
	e.g., on Ubuntu it's "aptitude install git-core"	06:01
	on Redhat or CentOS I guess it'd be "yum install git"	06:01
AaronLBorg	ok...but you sure there isn't a specific place for it to go?	06:01
offby1	failing that, it's pretty easy to build from source	06:01
	AaronLBorg: er?	06:01
AaronLBorg	yes?	06:01
offby1	like any software, it can go anywhere,	06:01
	there are conventions ... /usr/bin/git and /usr/local/bin/git are common	06:02
	are we talkin about the same thing?	06:02
AaronLBorg	i think so	06:02
offby1	"What we have here is a ... failure to communicate"	06:02
AaronLBorg	no, i think it's going good	06:02
	getting me thinking anyway	06:02
offby1	so you're not _really_ asking a question about git; you're more asking about installing software in general ... and git just happens to be the thing you have in mind to install at the moment.	06:02
AaronLBorg	if i want to clone that site to my local though.........	06:02
	well, i'm in the git room no?	06:03
offby1	ok, now you've changed the subject, and are talking about _using_ git.	06:03
	← loinclot_ left	06:03
AaronLBorg	if i want to clone....does it matter where it is?	06:03
offby1	nothin' wrong with that	06:03
	it surely doesn't matter where the "git" binary is in order for you to use it.	06:03
	← madewokherd left	06:03
AaronLBorg	that's what i needed to know, amigo!	06:03
	← threeve left	06:03
offby1	i.e., you might type "/usr/bin/git clone git://server/repository.git" or you might instead type "/usr/local/bin/git clone git://server/repository.git"	06:04
	but the effect will be the same.	06:04
	→ karmi joined	06:04
AaronLBorg	thank you!	06:04
offby1	may I ask what kind of development you do?	06:04
	← hobodave left	06:07
	← hobodave_ left	06:12
	← mithro left	06:12
	→ reithi joined	06:13
offby1	gap	06:14
	gasp	06:14
	Tukwila, WA?	06:14
	Why that's ... just down the street	06:14
	Kenmore, WA?!	06:16
Ryan52	offby1, what are you talking about? ;)	06:16
offby1	AaronLBorg.	06:16
	I tend to /whois people.	06:17
	He lives surprisingly near me.	06:17
Ryan52	oh :)	06:17
offby1	I also run off at the mouth.	06:17
	← japhb left	06:17
	→ japhb joined	06:17
AaronLBorg	what's been going on behind my back!	06:18
	?	06:18
	clear this line out...	06:18
	there	06:18
	← qrush left	06:18
offby1	I've /whois'd you down to the block	06:18
AaronLBorg	man.....	06:18
	is nothing sacred	06:18
	?	06:18
Ryan52	19:22 < robin> git verify -v <pack> can list all objects in a pack	06:19
	git: 'verify' is not a git-command. See 'git --help'.	06:19
	hrm.	06:20
	offby1, any ideas?	06:20
	I accidently repacked my stuff..	06:20
	so now the tricks you taught me don't work.	06:20
offby1	sob	06:20
	I think you're omitting a hyphen	06:20
	git verify-pack	06:20
Ryan52	okay.	06:21
	thanks	06:21
	← ahupp\|fb left	06:24
	→ vbabiy joined	06:25
	← vbabiy left	06:28
robinr	Ryan52: oops, git verify-pack	06:30
	→ crab joined	06:34
Ryan52	robinr: any idea how to figure out what filename and/or commit a blob is associated with? :)	06:35
	I'm getting by by looking at it, and thinking a bit, but...ya.	06:35
	If I weren't so familiar with this project that would be hard.	06:36
	← bgerber left	06:38
	→ carllerche_ joined	06:39
offby1	I imagine you can enumerate all the objects in the repository(!) ... filter out those that aren't trees ... then run "git ls-tree" on each, and grep the output for your SHA1.	06:39
	← carllerche left	06:40
Ryan52	I can't figure out how to get rid of the stray objects now. I tried "git repack -A" and then "git gc --prune", but that didn't make any difference...and I know I killed off most of the big objects,	06:41
	→ bgerber joined	06:41
Tv	Ryan52: you could rev-list to get all the commits, then run ls-tree -r on each one, grepping for your blob	06:41
	Ryan52: prune will only remove objects that are old enough	06:41
	→ hobodave joined	06:42
Ryan52	so I want to set gc.reflogExpireUnreachable to 0?	06:42
	← agenteo left	06:42
Tv	Ryan52: well you can do that on the command line.. that'd be for git reflog expire	06:43
	Ryan52: and then the gc thingie can kick in, and it has it's own age minimum	06:43
	← hobodave left	06:44
Ryan52	so how do I do this correctly?	06:44
Tv	oh you might need gc.reflogexpire, not just expireunreachable	06:44
	Ryan52: throw away all unused objects, even if they aren't old enough yet?	06:44
Ryan52	yes.	06:44
Tv	Ryan52: 1) make sure the repo is not being used while you do it	06:45
Ryan52	okay.	06:45
Tv	2) git reflog expire --expire=now --expire-unreachable=now --all	06:46
Ryan52	okay.	06:46
Tv	3) git prune --expire=now	06:46
	that should do it	06:46
Ryan52	hrm. should I run "git repack -A" before step 3?	06:47
	cause that did almost nothing.	06:47
Tv	yeah	06:47
	you should	06:47
Ryan52	okay, thanks!	06:47
Tv	Ryan52: afterwards you should have nothing in .git/objects/??	06:48
	oh and repack -A -d	06:48
Ryan52	ugh.fatal: sha1 file '.git/objects/pack/tmp_pack_bRLxWZ' write error (No space left on device)	06:49
Tv	or a separate run of git prune-packed	06:49
	→ Voker57 joined	06:49
Ryan52	sweet!	06:51
	18 .git/	06:51
	18 megabytes.	06:51
	it was 400 when I started :)	06:51
	thanks so much Tv and offby1 and robinr! :)	06:51
	→ moccuo joined	06:53
moccuo	when i clone a git repositository, and the repo i cloned from gets changed. how can i update my cloned repo to get the new changes?	06:54
	(from github)	06:54
	← Jacolyte left	06:54
Tv	moccuo: try reading the tutorial..	06:55
offby1	Ryan52: repay us by writing up what you did and getting it put into the FAQ :-\|	06:55
	moccuo: "git pull"	06:55
	Tv: I hope I haven't violated an implicit Code of Pedagogy :-\|	06:56
Tv	offby1: nah, but you're now going to be the go-to guy for any questions answered by the tutorial	06:56
Ryan52	gah. I shoulda done this in "script", then.	06:57
	I shoulda for myself, too..	06:57
Tv	Ryan52: well the commands are pretty much in the channel backlog now, just write it up nicely & add to faq	06:57
Ryan52	ya, I know :)	06:57
moccuo	offby1: hmm, it says its already up to date.	06:57
Ryan52	where does this FAQ live?	06:58
moccuo	that's strange. considering i just pushed some changes to github	06:58
	← harinath left	06:58
Tv	moccuo: from that same repository? or before the clone?	06:59
Ryan52	oh, found it :p	06:59
moccuo	hmm, it seems that they didnt get pushed..	06:59
offby1	moccuo: probably the branch you're on locally doesn't automatically pull from the branch to which you just pushed.	06:59
Ryan52	I'll add it sometime in the next week, I promise :). Now that I almost have successfully moved from svn to git, I need to do some actual programming :p	07:00
offby1	naw!	07:00
	_real_ programmers fiddle with their tool configuration eight hours a day.	07:00
moccuo	offby1: im trying to pull from my server that im ssh'd into. i pushed from my local system onto github.	07:00
Ryan52	lol	07:00
	offby1 makes a sweeping gesture towards his .emacs, which stretches to the horizon	07:00
offby1	moccuo: you could look at the file .git/config on your server, and see if it says that the branch you're on is supposed to pull from github.	07:01
Tv	offby1: you should wrap your lines	07:01
moccuo	offby1: well i just checked github and it appears that the push i did didnt go throught fro some reason	07:02
	so the server is actually up to date.	07:02
offby1	Tv: for Christmas?	07:02
Tv	offby1: no to avoid them stretching to the horizon	07:02
offby1	aah.	07:02
	← ben_h- left	07:02
moccuo	oooh no. i did something bad	07:02
	:(	07:03
offby1	I heard a baby scream. Was that you?	07:03
	murderer	07:03
moccuo	haha	07:03
	i freezed rails in my repo. and then did a "git push" thinking it would go to github. but it went to my development repository on the same system, which should NOT have rails frozen. ugggh	07:04
	← reithi left	07:05
moccuo	this is my first time deploying rails, hehe	07:05
	→ reithi joined	07:06
	← up_the_irons left	07:07
	→ up_the_irons joined	07:07
	← ben_h_ left	07:08
	← Arafangion left	07:08
	← doener left	07:08
	→ doener joined	07:09
	← ben_h left	07:11
	Ryan52 filter-branches again...this is starting to get fun ;)	07:11
Ryan52	how many different places could these people put database dumps? I'm on the third filter-branch, one of which made use of a glob. -_-	07:12
	→ aspotashev joined	07:12
	← sverrej left	07:13
	← murphl left	07:15
	← karmi left	07:17
Ryan52	now it's down to 6 megabytes!!!! :)	07:19
	← esden left	07:22
	→ Weaselweb joined	07:33
	→ bentob0x joined	07:38
	→ ashleyw joined	07:40
	← aspotashev left	07:41
	→ xjjk joined	07:42
	← cliffstah left	07:44
	→ cliffstah joined	07:44
	→ ijcd joined	07:46
	← ijcd left	07:47
	← Arachnid left	07:47
	← PerlJam left	07:47
	→ PerlJam joined	07:47
	→ Arachnid joined	07:48
	→ thijso joined	07:48
	→ reel_ joined	07:50
	← bobesponja left	07:51
	→ tjogin joined	07:53
	← not-xjjk left	07:53
	→ doiuf joined	07:54
	→ quoin- joined	07:54
	← nice_burger left	07:55
	→ ben_h joined	07:55
	← japhb left	07:56
	← sgrimm left	07:57
	→ japhb joined	07:58
	← ben_h left	08:01
	→ ben_h joined	08:03
	← ironfroggy_ left	08:04
	← ezyang left	08:04
	← ben_h left	08:05
	→ harinath joined	08:09
	→ jmesnil joined	08:10
	← robinr left	08:10
	← quoin left	08:12
	← Richo99 left	08:12
	→ naeu joined	08:14
	→ mithro joined	08:17
	→ mcella joined	08:18
	← jackdempsey left	08:19
	← altfive left	08:20
	→ karmi joined	08:21
	→ zerok joined	08:24
	← naeu left	08:24
	→ gilimanjaro joined	08:26
	← girishr_ left	08:27
	→ girishr_ joined	08:27
	→ JonathanT joined	08:29
	→ lucsky joined	08:31
lucsky	'morning	08:31
Ryan52	oh, shit, that bug that spearce said would be fine ruined my whole switchover. I can't rebase at all.	08:39
	error: Untracked working tree file 'vendor/plugins/railspgprocs/README' would be removed by merge.	08:39
	could not detach HEAD	08:39
	anybody know how I can tell it to add -f to that? :p	08:39
	git-rebase has no -f..	08:39
	Ryan52 reads teh source code	08:40
	Ryan52 hacks teh source code :(	08:41
	→ solydzajs joined	08:41
drizzd	Ryan52: if you have an untracked file, you obviously have a dirty working directory	08:42
	git will not allow you to randomly overwrite that with any command	08:42
	→ elmex joined	08:42
drizzd	so either move the file out of the way, or add and commit it	08:42
Ryan52	drizzd: I added -f to the git-checkout in git-rebase, it's taken care of. :)	08:42
	it's not that the file is untracked.	08:42
	it's a bug in git.	08:42
	trust me on this, I had 3 of the smart people in here look and agree with me :)	08:43
	at least three..	08:43
drizzd	ok, I've known to be wrong ;-)	08:43
	→ drizzd_ joined	08:44
Ryan52	oh, crap, but I did have another untracked working tree file...not the one it was erroring about, tho.	08:45
	← ibuffy left	08:45
	Ryan52 hates bugs, especially when he's too stupid to fix them :p	08:45
Ryan52	another == a real one :)	08:45
	← tango_ left	08:46
	→ ibuffy joined	08:46
	→ naeu joined	08:47
	→ charon joined	08:48
	← gretch left	08:49
	→ sverrej joined	08:49
	← mithro left	08:50
	→ radarek joined	08:52
	← carllerche_ left	08:54
Ryan52	uhhhhhhh....does git not handle renames of files? or is that just git-svn that doesn't?	08:54
mxey	Ryan52: Git does not track renaming.	08:54
Ryan52	so if I rename filea to fileb and run "git-log fileb" it won't show me the history of when it was filea?	08:55
	← karmi left	08:55
Ryan52	I know it doesn't "track" it, but, I mean, it is aware of it in some cases, I know.	08:55
wanders	Try git log --follow	08:55
Ryan52	oooooooooo....that's a bug :p	08:56
	so I had had two branches with a file named the same thing.	08:56
	and I was gonna merge them together.	08:56
	so I had to rename the file in one of the branches first.	08:56
	so I did, and I merged.	08:57
	now if I use --follow it shows me the history of both files :(	08:57
	← CodeOfficer left	08:57
Ryan52	whatever. I'll get over it. :p	08:57
	→ ankit9 joined	08:58
Ryan52	sadly svn handles this right...	08:59
	Ryan52 gets over it and moves on	08:59
Voker57	svn doea not have real branches	09:00
Ryan52	I know. that's why I'm moving to git :)	09:00
	← drizzd left	09:03
	← bronson left	09:04
	→ rtomayko joined	09:08
	← ankit9 left	09:08
	← naeu left	09:08
	→ ankit9 joined	09:09
	← xjjk left	09:10
	← univate left	09:14
	→ xjjk joined	09:14
	tjogin → tjogin\|away	09:16
	← ibuffy left	09:17
	→ ibuffy joined	09:17
	→ tango_ joined	09:20
	→ markelikalderon joined	09:23
	← chalky left	09:23
	→ twiinz joined	09:28
twiinz	Hi there	09:28
	→ univate joined	09:28
	→ hurikhan\|Work joined	09:28
twiinz	is it possible to target a single folder from a repository ? Say i've got a remote repository that has a few folders, each having a set of scripts inside, can I just pull 1 subfolder and not the whole repository?	09:29
mxey	twiinz: no.	09:30
twiinz	fair enough hehe	09:30
	← gilimanjaro left	09:30
	tjogin\|away → tjogin	09:33
	← markelikalderon left	09:38
	← ankit9 left	09:40
	→ nud joined	09:40
	→ ankit9 joined	09:41
	← synchris left	09:42
	→ ceyusa joined	09:43
	→ synchris joined	09:43
	← perezd88 left	09:44
	← Marmouset left	09:47
	→ pcapriotti joined	09:50
	← mjc left	09:51
	← drizzd_ left	09:54
	← moccuo left	09:56
	← ceej_ left	09:58
	→ chalky joined	10:00
	← feindbild left	10:03
	→ joevandyk joined	10:04
	→ vaskaren joined	10:05
	← ia left	10:07
	← joshdavey left	10:09
	← reel_ left	10:15
	← muthu_ left	10:17
	→ reel_ joined	10:17
	→ reel__ joined	10:17
	← reel__ left	10:18
	→ ia joined	10:23
	← fhobia left	10:25
	← univate left	10:25
	→ univate joined	10:31
	→ davik joined	10:32
	→ eddyp joined	10:33
	→ m3y3ring joined	10:36
	← m3y3ring left	10:37
	→ m3y3ring joined	10:37
	← ankit9 left	10:40
	← synchris left	10:40
	→ ankit9 joined	10:40
	→ synchris joined	10:40
	→ jfontan joined	10:48
	← univate left	10:49
	→ univate joined	10:50
	→ \|kukks\| joined	10:51
	→ bartocc joined	10:56
	→ sping_work joined	10:59
sping_work	hello	10:59
	← vmiklos left	11:02
	→ vmiklos joined	11:02
	← sping_work left	11:05
	→ sping_work joined	11:07
sping_work	Hello! Is there a way to specify a default remote for push and pull other than origin?	11:08
	← rtomayko left	11:08
	→ not-xjjk joined	11:10
	→ aspotashev joined	11:11
	→ xlq joined	11:15
xlq	What to do if you've just messed up your first git commit in the repository? (Can I make another orphan?)	11:16
sping_work	My setup is a read-only origin that I pull from and a remote "backup" that I only push to. That's why I'm wondering. Any ideas?	11:16
xlq	I suppose I'll start again.	11:18
	← imyousuf left	11:19
	→ lorandi joined	11:19
DrNick	xlq: git commit --amend	11:19
xlq	oh right	11:20
	← xjjk left	11:21
	← kanru left	11:23
	→ Angel2009 joined	11:23
Angel2009	Buenos dias	11:23
	← Angel2009 left	11:24
	→ kanru joined	11:27
	→ xjjk joined	11:28
	← ceyusa left	11:31
	← metajack_ left	11:34
	→ tzafrir_laptop joined	11:36
tzafrir_laptop	how can I add a git submodule from the directory ../ ?	11:36
	→ ceyusa joined	11:36
tzafrir_laptop	I have two local repositories that were made with git-svn	11:37
	One of them is an external of the other	11:37
	← not-xjjk left	11:39
	→ grahal joined	11:39
tzafrir_laptop	wow, working with submodules to replicate svn:externals is so much work	11:41
	I can't see myself writing simple instructions to explain someone else how to do that	11:41
	→ asdfiest joined	11:42
	← vaskaren left	11:42
	← ebzzry left	11:44
	→ ebzzry joined	11:45
	← davik left	11:49
	← dancor left	11:49
	→ not-xjjk joined	11:51
	→ lroar joined	11:52
lroar	Hi -- I'm doing "git diff" and there's one file in git's index I don't want to be part of the resultant patch I want to create from the git diff output -- can i tell git to forget this file when generating this patch?	11:53
	← jmesnil left	11:57
	→ joshdavey joined	11:57
	→ scientes joined	11:57
	tzafrir_laptop uses a symlink instead :-(	11:59
	← xjjk left	12:02
	← Gitzilla left	12:07
telmich	lroar: maybe use git diff $(git diff --name-only \| grep -v your file)	12:07
	→ Gitzilla joined	12:10
	→ temoto joined	12:12
temoto	Hello. I want to import commit from another repo. That is, i want to replay commit changes and comment and keep author from one just separate repo.	12:13
	← lroar left	12:14
temoto	more precisely i want to import one repo into another commit-by-commit.	12:14
	so repo2 will become subdir of repo1	12:14
	← LotR left	12:16
	→ LotR joined	12:16
	cbrake_away → cbrake	12:19
	→ simmerz joined	12:19
	→ asodm joined	12:20
asodm	hi!	12:20
simmerz	Hi. I'm trying to set up a gitosis server, and I thought I'd done it, but in trying to add one of my developers to the config, I had an error saying: "WARNING:gitosis.ssh:Unsafe SSH username in keyfile: '<keyfilename>'". when that user tries to clone the repo, it just asks him for the git user password on the server rather than granting access.	12:20
asodm	is there some way to grep through all revisions and not just head?	12:20
	→ knobo joined	12:22
	→ naeu joined	12:22
	→ kyle___ joined	12:24
kyle___	i keep getting could not find any SCM named `git'	12:24
	when doing cap deploy:check	12:24
	but i know i have git installed	12:24
knobo	Is it possible to make git handle cvs "$Id$" tags some way?	12:24
temoto	kyle___, git-core on ubuntu.	12:25
	kyle___, cap is capistrano?	12:25
kyle___	yes	12:25
temoto	kyle___, nice thing?	12:25
	← twiinz left	12:25
kyle___	i don't know, i'm a n00b	12:26
temoto	k :)	12:26
kyle___	what do you mean git-core on ubuntu	12:26
temoto	kyle___, git package is named git-core in ubuntu distribution.	12:26
	kyle___, basically, i mean try checking for git-core.	12:26
kyle___	i know ive done sudo apt-get install git-core like 2 times	12:27
	but the version numbers are different on ubuntu and os x, is that right?	12:28
xlq	asodm: git-grep?	12:29
asodm	asodm: right. but by default it only searches through the current revision, no?	12:29
	→ rolfb joined	12:34
knobo	Did nobody have issuses with magic keywords like $Id$ from rcs/cvs/svn?	12:36
	← LotR left	12:36
doener	asodm: what exactly are you looking for? A string that once existed but is gone now?	12:36
asodm	doener: yep.. and i wanna find out in which revision(s) it existed and where.	12:37
	→ ashleyw_ joined	12:37
doener	git log -S'string'	12:37
	shows commits that added or removed "string"	12:37
	(only add/remove, not move)	12:38
	← tzafrir_laptop left	12:38
doener	(per file though IIRC, so if "string" moved from file A to file B, that would be shown)	12:38
	← eddyp left	12:39
temoto	kyle___, of course they could be and most probaly are different.	12:39
	→ Ryback_ joined	12:39
	→ d0k joined	12:42
	← univate left	12:42
	→ gitte joined	12:42
	→ LotR joined	12:45
asodm	doener: perfect. thanks!	12:47
	→ netoman joined	12:47
	← asodm left	12:48
	→ _Vi joined	12:51
	← ashleyw left	12:52
knobo	You are ignoring me :(	12:52
	I could make a perl skript that solves my problem. Is it possible to make some hooks that runs my script on every checkout and commit?	12:54
broonie	knobo: http://thread.gmane.org/gmane.comp.version-control.git/44750	12:54
	← EmilMedve left	12:54
Pieter	knobo: see man git-attributes	12:54
	oh	12:55
	man gitattributes	12:55
Gitbot	Pieter: the gitattributes manpage can be found at http://git.or.cz/man/gitattributes	12:55
	→ xjjk joined	12:56
	→ lukedal joined	12:58
knobo	thanx. Is it possible to customize the indent option some how?	13:00
	← _Vi left	13:00
	→ PhilK joined	13:00
	→ _Vi joined	13:00
	→ univate joined	13:01
	→ d0k_ joined	13:03
	→ bobmcw joined	13:04
	→ jmesnil joined	13:08
	← awarde left	13:09
	← Voker57 left	13:09
	→ Voker57 joined	13:09
	→ snitko joined	13:11
	← not-xjjk left	13:12
snitko	hi, how do I get a remote branch to the local machine? just can't find that simple thing in google	13:12
	← d0k left	13:12
doener	snitko: git fetch, if you want to repeatedly fetch from that repo, add a remote for it	13:13
snitko	hm, that's not exactly what I want. I have master locally. I want to also have a 'xxx' branch locally. exactly, what should I do?	13:14
doener	so you just want to create a new branch? How does a remote branch come into play there?	13:15
temoto	git branch xxx	13:15
doener	Maybe you're just missing the concept of remote tracking branches?	13:15
	→ vaskaren joined	13:15
temoto	doener, i'm missing if you're going to cast that.	13:15
doener	temoto: sorry, could you rephrase that? I'm not a native speaker and don't understand the "to cast that" part :-/	13:16
	← xjjk left	13:16
doener	snitko: My guess right now is: You cloned a repo, got a master branch, but on the remote repo, there's also a xxx branch, and you're looking for that one?	13:17
snitko	right	13:17
temoto	doener, if you're going to explain.	13:18
	→ xjjk joined	13:18
doener	snitko: ok. To provide truly disconnected operations, git stores the state of branches on a remote in so called remote tracking branches.	13:18
temoto	doener, i'm not a native speaker also and i'm feeling free to invent new words sometimes :)	13:18
doener	snitko: by default, if you fetch from a remote called "origin" (default name for the remote for the repo you cloned), all remote branches are stored locally in refs/remotes/origin/<remote-branch-name>	13:19
temoto	git pull remote xxx xxx	13:19
doener	snitko: those remote tracking branches are listed by "git branch -r"	13:19
xlq	Is there a way to have two working copies from one .git dir?	13:19
doener	temoto: no, that would merge xxx twice into the current branch	13:19
xlq	i.e. to save space	13:19
doener	xlq: if you clone locally, git uses hardlinks	13:20
xlq	oh right, didn't notice :D	13:20
snitko	doener: ok, that I got.	13:20
	→ cmarcelo joined	13:20
temoto	oh well, i don't know then... git branch xxx ; git checkout xxx ; git pull remote xxx	13:20
doener	xlq: you can also use alternates and something else I don't remember, see man git-clone	13:20
Gitbot	xlq: the git-clone manpage can be found at http://git.or.cz/man/git-clone	13:20
doener	xlq: and finally, there's a git-new-workdir script somewhere in contrib/	13:20
	← quoin- left	13:21
doener	xlq: that does exactly what you're asking for, but it has some drawbacks, IIRC basically the same that are mentioned for the alternates setup in the clone manpage	13:21
temoto	I used git-clone with hardlinks and it was very fine experience. Quite no problems. there is --no-hardlinks switch also	13:21
doener	snitko: ok. Those remote tracking branches are updated by "git fetch origin" (thanks to the default config for that remote)	13:21
	← univate left	13:22
doener	snitko: now, you can basically use those remote tracking branches like any other ref.	13:22
xlq	O_o ok thans doener	13:22
doener	snitko: eg. you can do "git log origin/master..master" to see commits that your local master branch has, but the remote tracking branch doesn't have	13:22
	snitko: in the same way, you can use the remote tracking branch to specify a starting point for a branch you create	13:22
	snitko: git branch xxx origin/xxx	13:23
	snitko: that creates a local branch called xxx that starts at the same commit as origin/xxx	13:23
	snitko: local branches are generally only needed if you intend to actually work on them (ie. create commits)	13:23
	snitko: if you just want to _look_ at how things are, there's no need to create a local branch, you can just use the remote tracking branches instead	13:24
	snitko: after creating the local branch, you can switch to it using "git checkout xxx". And you can also use the "git checkout -b xxx origin/xxx" shorthand, which combines the two commands	13:24
	snitko: that is, it creates a xxx branch and checks it out	13:25
	→ Zimm joined	13:25
doener	snitko: everything clear as mud now? ;-)	13:25
snitko	doener: I got everything, except the part about not having them locally. How can I look at the changes in those branches and not have them locally?	13:25
	← asdfiest left	13:26
	→ univate joined	13:26
doener	snitko: you _do_ have them locally. The "local branch" term is just what I happen to use to differentiate between them and the remote tracking branches	13:26
eMBee	run git branch -a; to see all branches (local and remote)	13:27
doener	snitko: remote tracking branches are just as local, all required objects are there. But they represent the state of remote branches, at the time of the last fetch	13:27
	→ asn joined	13:28
asn	So, I'm editing my local kernel tree and I make some changes (add some files, edit some others, etc.). I then want to return it to it's starting state. I do a git checkout -f, but the files that I created in my local copy are still there? Shouldn't they be deleted?	13:29
doener	asn: git clean	13:29
	asn: git isn't very proud of deleting stuff it cannot restore	13:29
	← jonaskoelker left	13:29
doener	so there's an explicit command for that	13:29
xlq	;)	13:29
asn	I see!	13:30
	thanks :)	13:30
	← bentob0x left	13:30
doener	asn: well, to be correct, "git checkout -f" already destroyed some stuff git cannot restore (the uncommitted changes to the tracked files), but well, you already said --force there	13:31
	→ timonator joined	13:31
	→ alb joined	13:32
timonator	how does the git community justify the lack of metadata storage like svns properties?	13:32
asn	doener: thanks, again	13:32
	→ name joined	13:32
	← lamont left	13:33
	→ jonaskoelker joined	13:33
	→ lamont joined	13:35
eMBee	timonator: to keep it simple?	13:36
timonator	hm, not really satisfying	13:36
	i'm currently trying to justify that i like git in front of an svn user that says git is overhyped and will die down in a year or so	13:36
	→ bartman joined	13:37
	→ __name__ joined	13:37
	← name left	13:37
	← LotR left	13:37
	__name__ → name	13:38
	← _Vi left	13:38
mxey	timonator: wait a sec.	13:38
eMBee	tough job, what does the guy know about git?	13:38
timonator	mxey to the rescue!	13:39
mxey	timonator: http://whygitisbetterthanx.com/ was designed exactly for this task	13:39
timonator	mxey: "the domain name alone disqualifies clicking the url" he said	13:39
	because it hints at fanboyism	13:39
mxey	Looks like he is the fanboy. A Subversion fanboy. Tell him to stfu from me.	13:40
timonator	well, parts of his arguments are solid. like that git still has to get a lot more support from IDEs and other things...	13:41
	→ boto joined	13:41
spb	vim is the only IDE you need	13:41
	and that has perfectly decent git support	13:41
timonator	not to everyone.	13:41
mxey	For Emacs there is magit.	13:41
timonator	he uses eclipse for his massive java stuff	13:42
mxey	There is some Git stuff for Eclipse.	13:42
timonator	yes, it's very, very lacking	13:42
mxey	I was told there is no computer that is fast enough for Eclipse.	13:42
timonator	heh heh	13:42
	→ chris2 joined	13:42
	← temoto left	13:42
mxey	timonator: So he likes Eclipse, Java and Subversion. Why are you even talking to him? ;)	13:43
	→ xanonus joined	13:44
	← d0k_ left	13:44
timonator	he is a very dear friend to me, that's why	13:44
deskin	timonator: editorializing, so others here may not agree with me; but first, I'd point out that git has analogues of most of the svn: properties of interest, like ignoring files, or using clean/smudge filters; and for other metadata, git lets you put it in files where it belongs. I don't know of a problem arising from lack of git metadata tracking that can't be solved using clean/smudge filters and putting information in other files	13:44
andres	And the eclipse stuff suffers a bit from the "lets reimplement everything in java" approach - although I have to admit the reasons are better than usual.	13:44
deskin	ask him to show you a real problem that svn's metadata wins at which git actually cannot do	13:44
timonator	what are those clean/smudge filters you're talking about? like having \n in the repo and \n\r in the working copy? with hooks?	13:44
	← nikolasco left	13:45
deskin	it could be newline stuff, though there's other ways to do that since it's so common; basically, clean/smudge filters let your working copy update with a non-canonical form of the file in the repository, and convert back to the canonical form when you check in	13:46
	→ LiamH joined	13:46
deskin	one example would be expanding $Date$, $Author$ or whatever CVS-style keywords	13:46
	← albertito left	13:47
timonator	ah, that's neat	13:47
deskin	but git doesn't give you a 'keyword' property to apply; it says 'write your clean/smudge filter to do that' :)	13:47
andres	And: "Be prepared to handle some pain resulting of nearly anybody using those"	13:47
deskin	true, they aren't a silver bullet; for one thing, branch switching becomes slower since they have to run in cases where files might not have changed before, but could now	13:48
andres	svn has a way much better library. Thats by far its biggest advantage I think.	13:50
	(and its windows support. Git is slow on windows. That reminds me. I wanted to send some windows straces...)	13:50
bobmcw	but if you like to branch, git is the best thing ever	13:50
	branching alone sold me on git	13:50
	→ EmilMedve joined	13:50
bobmcw	well, the merging after I branch	13:50
	SVN can branch, it's merging that's painful	13:51
timonator	and svn branches have to be pushed to the server	13:51
	for everyone to see	13:51
eMBee	but branching in svn is much more expensive compared to git	13:51
andres	And its merge tracking is a bad joke because of multiple reasons (no repeated merges, slow slow slow)	13:51
	← univate left	13:52
eMBee	and the svn interface to branches and tags, making them look like directories, is just horrible	13:52
	andres: are you talking about merge tracking in svn 1.5?	13:54
	→ loincloth joined	13:54
	→ univate joined	13:54
andres	eMBee: yes	13:54
	← jonaskoelker left	13:55
	→ jonaskoelker joined	13:55
eMBee	andres: do you have experience with that or a pointer to some text explaining the problems/limitations?	13:55
	→ Yuuhi joined	13:55
andres	eMBee: I have some experience and no pointers I am aware of without searching	13:56
timonator	what happens to a file that is in revision control and later someone adds a .gitignore that matches it?	13:56
Pieter	nothing	13:56
timonator	so it stays in there?	13:56
Pieter	.gitignore only matches untracked files	13:56
	→ Cacheaway joined	13:57
Pieter	yes	13:57
	→ nikolasco joined	13:57
timonator	what does svn do in that case?	13:58
	→ _Vi joined	13:58
	← loincloth left	13:59
timonator	i mean, add svn:ignore after the file has been checked in	13:59
andres	timonator: the same	14:01
	→ loincloth joined	14:02
	← jds left	14:03
andres	timonator: everything else is bogus I think.	14:03
timonator	he says "gits merges aren't much better than any merges, because merges are always messy"	14:04
	>_>	14:04
	he says having to pull before you can push is a great feature	14:04
deskin	I think mxey is right: sounds like he's the fanboy here :)	14:05
timonator	:>	14:05
mxey	I'm glad my friends all use Git :)	14:05
deskin	I'd ask him, if merges are so messy why do gits do them so often :)	14:05
mxey	although it took some time to convert one of them from Mercurial.	14:06
dsop	git, mercurial everything the same	14:06
	just a matter of taste	14:06
mxey	I recently read the Mercurial 1.1 changelog… a lot of Git mentions… git-like features…	14:06
timonator	the argument is getting heated	14:06
mxey	dsop: Not really, Git is much more powerful.	14:06
dsop	mxey: not really	14:07
mxey	does Mercurial have a reflog?	14:07
dsop	mxey: you can do a lot of the stuff with mercurial too, if not go and write an etension	14:07
	mxey: hg's history is immutable	14:07
mxey	Git's history is as well.	14:07
dsop	depends on the view. Well nevermind, I guess not.	14:07
	← solydzajs left	14:08
dsop	what I mean is: both have advantages and disadvantages it's up to you that you chose	14:08
	→ vbabiy joined	14:08
mxey	If you think so, tell me what Mercurial's advantage is.	14:08
dsop	e.g. git is so hard to interface with	14:08
mxey	I never had a problem using Git in shell scripts…	14:09
dsop	try to add git to an IDE and you think different	14:09
	← ashleyw_ left	14:09
mxey	IDEs are a waste of screen space.	14:09
timonator	dsop: have you added hg to an IDE yet?	14:09
dsop	timonator: no but I know the implementation, pretty easy actually.	14:09
timonator	does it interface the python classes directly?	14:10
dsop	timonator: for sure	14:10
mxey	I am aware that Mercurial is more library-based than Git. Which means you have to code Python instead of sticking commands together on your shell.	14:10
dsop	also mercurial has hg serve	14:10
	and zeroconf	14:10
mxey	git daemon exists as well.	14:10
telmich	zeroconf for mercurial?	14:10
dsop	yes, but you have to give 100 of options before it really oes what you want	14:10
telmich	that announces what?	14:10
mxey	dsop: Git supports aliases.	14:10
Voker57	dsop: not really, one or two	14:11
mxey	alias.serve=!git daemon --verbose --reuseaddr --base-path=. --export-all ./.git	14:11
dsop	mxey: but it is not very user friendly	14:11
mxey	Aliases are user friendly.	14:11
dsop	mxey: but not git daemon	14:11
	telmich: it announce hg repositories	14:11
mxey	yes, hg serve is more straightforward. but setting up a serve alias for Git is hardly a problem.	14:11
	andres prefers if opening a port is not dead easy.	14:11
telmich	dsop: hmmm...interesting...which program can be used to display those?	14:11
dsop	mxey: it is a problem if you want to use a version control system and don't want to know all it's details	14:12
	telmich: mercurial :)	14:12
deskin	dsop: you mean, using mdns like apple/ avahi-type stuff?	14:12
mxey	Not wanting to learn is a problem, yes. Just learn.	14:12
telmich	dsop: nice	14:12
dsop	mxey: no it's just the common case, mercurial is for most uses just easier to learn and deal with. that's why I say: it's a matter of what you like	14:12
	it's not a technial think: if you prefer the unix-style with a little bit more power use git. if not use hg	14:13
	← chalky left	14:13
deskin	hmm, one in every seven commits in git.git is a merge; no way I'd want to do that in svn	14:13
mxey	Power is more important than beginner-friendlines… you are a power user for a longer time than you are a beginner	14:13
dsop	I stick to the 80/20 rule and therefore hg is as perfect as git is. you use 20% of the features	14:13
	→ jm joined	14:13
dsop	mxey: I actually don't find something that I want to have in hg that's not implemented yet	14:14
mxey	Maybe you should reach beyond that arbitrary barrier and discover Git.	14:14
	→ ashleyw joined	14:14
dsop	mxey: I use git for 1 1/2 years know, I know it well and I like it, but it's just that it annoys me if people say: hg/bzr is just too bad	14:14
	they all have their points and their reasons to use them.	14:15
	→ esden joined	14:15
mxey	I used Mercurial before I used Git.	14:15
	Actually it took a long time to decide on this matter.	14:15
dsop	and particularly for hg/git there is no better in general there is just: I need this and other people need other things. So that's why both have their points	14:15
	typical thing which hg is bad: reflog, rebasing, etc.	14:15
	← synchris left	14:16
mxey	You are free to believe that, but stop bothering me with it.	14:16
dsop	typical thing git is bad: interface, simplicity, windows	14:16
mxey	Do you really mean simplicity or do you mean newbie-friendly?	14:16
telmich	hmm, is git support for windows still bad?	14:17
dsop	telmich: at least for usual windows user that want to have a nice gui	14:17
	← univate left	14:17
telmich	dsop: ahh, in that way	14:17
mxey	Usual Windows users should that away from version control.	14:17
	s/that/stay/	14:17
	← vbabiy left	14:17
telmich	dsop: but I mean -- the "normal" way (commandline) should work without problems, doesn't it?	14:17
	→ vbabiy joined	14:17
dsop	telmich: yes, for sure	14:17
telmich	mxey: well, I'm not sure whether this would help... the real solution is to get rid of windows at all	14:18
dsop	telmich: but if a company decides which vcs to use, they often do not use git because of the lack of a good gui	14:18
mxey	telmich: Well, Ubuntu GUI users are not much better.	14:18
azuk	telmich, there are problems even in command line use	14:18
	← _Vi left	14:19
	→ jds joined	14:19
	← bartman left	14:20
telmich	azuk: which ones?	14:20
	→ _Vi joined	14:20
	→ muthu_ joined	14:21
	← vbabiy left	14:24
	→ vbabiy joined	14:24
timonator	my friend says, that the world doesn't need git, that it's only good for special cases like for example google android	14:24
	but for any non-huge-scale project it's nonsense, as people will just adapt the SVN workflow of one central repo	14:25
telmich	timonator: not true	14:25
timonator	yeah, that's what i said	14:25
telmich	timonator: I was just creating some t-shirts for a local hackers group	14:25
timonator	he doesn't even agree to disagree :\|	14:25
mxey	telmich: t-shirts? ;)	14:26
telmich	timonator: and everybody had to submit his/her eps and svg via git	14:26
	mxey uses Git for almost everything	14:26
telmich	mxey: yep, you can find them there: http://git.schottelius.org/?p=EOF/shirts-2008;a=summary	14:26
timonator	heh	14:26
mxey	oh, !eof?	14:26
telmich	yep	14:26
mxey	I think timonator and me should be able to see these shirts live in a few weeks ;)	14:27
telmich	mxey: you are also at the 25C3?	14:27
	→ qrush joined	14:27
mxey	yes	14:27
telmich	nice	14:27
	where are you from?	14:27
dsop	hmm 25C3.good option	14:27
mxey	first one	14:27
	telmich: Vicinity of Düsseldorf.	14:27
telmich	mxey: ahh, german	14:28
	mxey nickt	14:28
	← scientes left	14:28
	telmich = ch,zh	14:28
telmich	hmm, let's make a git workshop	14:29
	← vaskaren left	14:29
	← vbabiy left	14:29
	→ vaskaren joined	14:29
mxey	I'm giving Git introductions about every week at work now :D	14:29
	→ vbabiy joined	14:29
	← vbabiy left	14:29
pflanze	Is there a way to run gitk with a file argument (to show only commits touching that file) but still show the complete diffs, of all files touched by the selected commits?	14:30
telmich	mxey: was the same in my old company	14:30
	→ softdrink joined	14:31
	← _Vi left	14:33
charon	pflanze: untick "limit diffs to listed paths" in gitk's settings	14:33
	← softdrink left	14:33
	← snitko left	14:33
telmich	mxey: http://events.ccc.de/congress/2008/wiki/!eof#Using_the_distributed_version_control_system_git	14:34
	← grahal left	14:34
timonator	telmich: we could come visit you :)	14:34
mxey	agreed.	14:34
	the chat system sounds interesting as well	14:34
telmich	timonator: sure, you are invited to have a look at the chaotic, friendly hackers group	14:35
pflanze	charon: thanks, that's it; although I would have preferred a commandline option, so as to not affect every future invocation / not having to click that item all the time.	14:35
	← unreal left	14:35
mxey	timonator: Craig also reserved a CCCD table ;)	14:35
pflanze	Maybe I should add one.. (but I don't know tcl)	14:35
charon	pflanze: make a patch to recognize --full-diff as with git-log ;)	14:35
timonator	mxey: we've got lojban space	14:35
mxey	telmich: Tell me more about the "Distributed communities"	14:36
	timonator: There is also a Debian table ;)	14:36
timonator	where you are going to sit and pick up girls ;)	14:36
mxey	But I if lurk with Craig I will end up at the Gentoo table.	14:36
	I need more Debian shirts so I can wear one of them at the Gentoo party…	14:36
timonator	heh heh	14:37
	→ shep-werk joined	14:37
shep-werk	Howdy all	14:37
mxey	moin shep-werk	14:37
timonator	why aren't we talking in #git-social or something?	14:37
mxey	Are we on-topic fascists? ;)	14:37
timonator	everything is logged!	14:37
mxey	oh dear	14:38
	they are gonna hunt down all Git users!	14:38
	← EiNZTEiN left	14:38
timonator	no, all slackers that discourage on-topic talk on #git	14:38
mxey	yeah, I should hack some more JavaScript.	14:38
	→ bartman joined	14:39
shep-werk	timonator: you want on-topic discussion? Does that mean you will answer my question? :-D	14:39
telmich	mxey: it's mostly about how to survive as a community, though you are not living next to each other	14:39
timonator	if i can, yes. ask	14:39
mxey	telmich: ah	14:39
shep-werk	So, we are going to branch a release version (7.0) from our development branch (development)	14:39
	and at the beginning, there are going to be many bug fixes that need to go into both branches	14:40
	obviously, as time goes on, more will be pure development, and less will be fixes that will go on that branch	14:40
timonator	yeah	14:41
	→ parasti joined	14:41
shep-werk	my question is: are there any neat git tricks I can use to help prevent fixes in dev from not making it to 7.0	14:41
	we use other tracking things (bugzilla flags, e.g.) but git really is the authoritative place	14:41
deskin	shep-werk: you might want to look at man gitworkflows; the 'apply fixes to the lowest branch that needs them, and merge up' approach seems to work pretty nicely	14:43
Gitbot	shep-werk: the gitworkflows manpage can be found at http://git.or.cz/man/gitworkflows	14:43
	→ cannonball joined	14:43
charon	also, make topic branches for every fix, and name them accordingly if 7.0 needs it; then 'git branch --no-merged $stable' can tell you the topics that are still not on 7.0	14:44
	← qrush left	14:44
	→ qrush joined	14:45
	← wsuff left	14:46
shep-werk	deskin: yeah, basically we do the downwards cherry-picking now	14:46
	→ [RIT]Rawn027 joined	14:46
shep-werk	which can be painful	14:46
deskin	yes	14:46
shep-werk	I guess the root cause it that we will fix a bug in dev	14:46
	then realize that it is broken in 7.0 as well	14:47
	→ jackdempsey joined	14:47
shep-werk	maybe that means that our workflow should be the other way around	14:47
deskin	well, that can be addressed by careful testing for the bug in both places	14:47
shep-werk	find a bug	14:47
	yeah	14:47
	eMBee would generally want to test a fix in dev first, before adding it to stable	14:47
mxey	If you introduce a regression test per bug, you can run them on all your branches ;)	14:47
pflanze	Any gitk knowledgable person helping me understand how option parsing in gitk works?	14:48
shep-werk	mxey: woah - testing? now you are talking crazy	14:48
deskin	one can do that (test fix in dev percolates to stable), but topic branches are a better way than cherry picking	14:48
	→ aroben joined	14:48
mxey	shep-werk: Automated testing is not crazy.	14:48
shep-werk	mxey: I know, just poking fun at you	14:49
mxey	Thought so	14:49
eMBee	unfortunately automated testing does not include automated test-writing	14:49
	deskin realised just the other day that 'make test' in git.git could benefit from the -j option; now patches come out 3x sooner	14:49
timonator	hey, where did i see that comparison of hg, svn, bzr and git in terms of speed and repository size?	14:49
shep-werk	mxey: our full QA quite takes ~ 5 days, so it can be a bear to use	14:49
	suite*	14:49
mxey	hrhr	14:50
	← peeja left	14:50
mxey	can't you run single tests?	14:50
	← xjjk left	14:50
shep-werk	mxey: the person who wrote out testsuite did not strive for user friendliness....	14:50
eMBee	shep-werk: 5 days??? and here we are complaining that our testsuite runs 40 minutes and investing a lot of work paralellizing it to get it down to less than 5 minutes...	14:50
shep-werk	it's possible to do, but not easy	14:50
	eMBee: our testsuite includes solaris 8, on a not-too-fast machine :-(	14:51
	→ xjjk joined	14:52
shep-werk	eMBee: if you like parallelizing things, you should check out non-recursive make ;-)	14:52
eMBee	how do you use that testsuite? here we run tests before every commit to trunk	14:52
shep-werk	eMBee: I guess the scope is just different, as that wouldn't be very feasible here	14:53
	← ankit9 left	14:53
eMBee	shep-werk: that's what i wonder about, such a testsuite would not work at all here	14:54
andres	eMBee: do that for performance regression tests...	14:54
eMBee	andres: do what?	14:54
shep-werk	eMBee: sometimes our problems only occur at crazy amounts of data ;-)	14:54
andres	eMBee: running regression tests before every commit	14:55
	eMBee: when a singly query takes some hours ;-)	14:55
shep-werk	deskin: what about cases where dev and 7.0 have diverged so much that a fix doesn't apply in both?	14:55
	→ _Vi joined	14:56
shep-werk	it seems once a single commit like that occurs, you can no longer do merges	14:56
	→ absentia joined	14:56
timonator	well, you'd cherry-pick anyway, wouldn't you?	14:56
	after a while, i mean	14:56
absentia	grrr. I did a rm -rf dir in my local files.. and when I do a git pull, it says i'm up to date... how do I tell get to actaully do something useful and GET the files that i need?	14:57
deskin	shep-werk: you mean, doesn't apply cleanly? Or, isn't broken in both?	14:57
timonator	and if the dev tree diverged so far, that the fix doesn't apply to the original, you might be better up rewriting that fix or checking if the fix even applies	14:57
shep-werk	timonator: I guess - so the suggestion is do merges while they are close together, then cherry-pick once the rush has dies down?	14:57
	deskin: i meant doesnt apply cleanly, but I guess both cases are important	14:58
absentia	?	14:58
parasti	absentia: so what do you want to do? get your files back or pull?	14:58
timonator	i've never been able to test this in a real-world situation...	14:58
absentia	I want to pull the latest files from the master	14:58
deskin	you can hack the merge result to be whatever you want; so even if something in stable 'shouldn't be merged', you can do the merge as a sort of no-op, but honestly I don't think it's a case that will come up a lot	14:58
shep-werk	absentia: git reset --hard will put files back to the current HEAD	14:58
absentia	thank you	14:58
timonator	or whatever ref-like you put there	14:58
absentia	thank you	14:59
	→ EiNZTEiN_ joined	14:59
	→ EiNZTEiN joined	14:59
deskin	it's really a better situation than svn: svn ci -m 'DO NOT MERGE: fix bug foobar'	15:00
shep-werk	OK, I'll see if doing merges makes sense for us - some of our devs are still used to CVS and SVN, so there can be confusion	15:00
	deskin: preaching to the choir there ;-)	15:00
absentia	fvck	15:01
	← absentia left	15:01
	← reithi left	15:01
	→ scientes joined	15:01
	→ withanx joined	15:02
	← thiago_home left	15:03
	→ thadwheeler joined	15:03
pflanze	Hm, I've removed any reference to "full-diff" from current gitk, but it still doesn't complain about unknown argument, why?	15:04
	← _Vi left	15:05
Zathrus	anyone have experience w/ git on Windows? Wondering if cygwin or mysysgit is preferable; needs to be able to talk over ssh to Linux boxes (if it could talk over putty's ssh, then even better, since I have the keys setup properly via pageant)	15:06
	→ _Vi joined	15:06
thadwheeler	quick question - I have gotten git installed on Mac 10.5 - when I try to git commit -a I get the error after entering my commit message in Vim: error: There was a problem with the editor 'vi'.	15:07
	Please supply the message using either -m or -F option.	15:07
	any thoughts?	15:07
cbreak	weird	15:07
xlq	weird	15:07
deskin	thadwheeler: I've seen this one before; default vi exits with an error code on some mac setups	15:07
shep-werk	thadwheeler: use emacs ;-)	15:07
	shep-werk starts flamewar	15:08
cbreak	I use os x 10.5 and it works	15:08
xlq	shep-werk: :(	15:08
cbreak	(although, I do use vim, not vi)	15:08
metajack	thadwheeler: what is your $EDITOR ?	15:08
	→ Hendu joined	15:08
	← Weaselweb left	15:08
thadwheeler	metajack, nothing? how do I set that?	15:08
shep-werk	deskin, timonator: So the suggestion is basically "use merge, until it doesn't make sense anymore", right?	15:08
Hendu	Hello. I'm using git on mac os x and would like to change the app that helps me file merge. Can this be done?	15:09
deskin	one quick fix might be to do echo -e '#!/bin/bash\nvi $@\nexit 0' > myvim; EDITOR=myvim git commit...	15:09
timonator	shep-werk: that's how i'd do it	15:09
	but ymmv	15:09
	Hendu: very much so, see http://www.kernel.org/pub/software/scm/git/docs/git-merge.html the configuration section	15:10
Hendu	timonator: thanks	15:10
deskin	shep-werk: more or less; but merging isn't a substitute for thinking about your commits and workflow :) Grok the change being made, and you'll know where to apply it, and how to propagate it	15:11
shep-werk	deskin: thats the tricky thing that my CTO brought up - how do you prevent someone else from merging your commit that should be in 7.0 and not in development?	15:11
deskin	you rewrite the merge result to not contain those changes	15:12
cbreak	artefact:~ cbreak$ vi --help	15:12
	VIM - Vi IMproved 7.2 (2008 Aug 9, compiled Sep 25 2008 19:37:57)	15:12
deskin	look at merge strategies; -s ours exists for precisely this reason	15:13
cbreak	that works for me with git	15:13
thadwheeler	deskin, reading thought the command you sent - and I am a little confused..	15:13
deskin	ok... what can I explain?	15:13
	I left out lots of stuff, like, chmod +x, but I think the spirit is clear	15:14
	deskin is not a mac user, so there could be some translation issues	15:15
thadwheeler	is this basically writing a config for VIM? is there an $EDITOR variable I should be setting some where?	15:15
xlq	thadwheeler: it's a shell script that pretends to be vim, so that you can change the value it retuns.	15:15
deskin	well, you could do git config core.editor = myvim (or emacs or ...)	15:15
thadwheeler	whats the purpose of hte command I guess is what I am getting at? sorry..	15:15
cbreak	the mac vi is really a vim	15:15
	← esden left	15:15
cbreak	and it works for me, which is weird	15:15
deskin	try this: 'vi'; then ':q'; then at the shell, echo $?	15:16
cbreak	with git version 1.6.0.3.523.g304d0	15:16
thadwheeler	so I can change it in the git config core.editor	15:16
deskin	vi is returning an error code, which git interprets correctly as failure; so my script ignores vi and exits with 0 (success)	15:16
thadwheeler	mine exits with a 1	15:17
deskin	exactly	15:18
thadwheeler	ls	15:18
	oops	15:18
	→ threeve joined	15:18
deskin	now try './myvim'; ':q'; 'echo $?'	15:19
	→ ashleyw_ joined	15:20
cbreak	how do you exit vim?	15:21
	→ esden joined	15:21
Rhonda	cbreak: :q!	15:21
xlq	:q	15:21
	or ZZ	15:21
	\|kukks\| → kukks	15:21
cbreak	:q! should not work	15:21
deskin	the ! doesn't make a difference	15:21
Rhonda	cbreak: Why shouldn't it work?	15:21
cbreak	it will not save the commit message	15:21
bobmcw	if you've :w first, :q! is fine	15:21
cbreak	yeah, but if not, it won't :)	15:21
thadwheeler	ls -al	15:22
xlq	'!' means to discard changes	15:22
Rhonda	Was about exiting. Noone mentioned saving changes. :)	15:22
deskin	no, ! means to really quit	15:22
xlq	...when applied to :q	15:22
	yes	15:22
	well, it actually really closes a window	15:22
PerlJam	! means to emphatically do whatever operation it is attached to.	15:22
Rhonda	If you want to save and quit, I suggest :x	15:22
deskin	ah, you're right, discard anything since last written	15:22
PerlJam	e.g., :w! foo means to really write foo even if that overwrites and existing foo	15:23
	s/and/an/	15:23
deskin	but it doesn't change exit status	15:23
xlq	and :help! tries to calm you down	15:23
cbreak	my vi gives an exit value of 0	15:23
Rhonda	xlq: haha, cool easter egg :)	15:23
cbreak	what does "which vi" return?	15:24
	← xjjk left	15:24
	← jackdempsey left	15:24
thadwheeler	deskin, getting 0 as return now.. but I a gussing I needed to use vim not myvim in the code...	15:24
cbreak	the os x vi is in /usr/bin, maybe you have a fink vi or something	15:24
deskin	well, if vim exits with 0 already, just use that, not my script	15:24
shep-werk	deskin: do you have a good reference for the 'ours' strategy? The manpage is a little short...	15:25
thadwheeler	nope regular vim exits with 1	15:25
cbreak	which is your regular one?	15:25
	because I know for sure that the one from mac os x returns 0	15:26
xlq	Right there, next to your regular one	15:26
deskin	shep-werk: it's quite simple: 'make a commit as having merged this branch as a parent, but secretly use the same tree of the branch I'm on'	15:26
	→ xjjk joined	15:26
deskin	i.e. whatever the other branch has in terms of content doesn't matter; the result tree is always just what you have	15:26
thadwheeler	cbreak, in run vim - exit and get a 1 as a return code	15:27
shep-werk	deskin: sadly, I'm still not getting it - would playing with a test git repo make it clear for me?	15:27
cbreak	thadwheeler: which vim?	15:27
deskin	but now that other commit is an ancestor of your branch, so you can merge again later without reintroducing that change	15:27
thadwheeler	which vim gets me /usr/bin/vim	15:27
cbreak	weird	15:27
deskin	shep-werk: I highly recommend it	15:27
cbreak	what version?	15:27
deskin	the test repo thing, that is	15:27
shep-werk	deskin: it sounds kind of like git itself - can be hard to understand, but once you do it it clicks ;-)	15:28
	← ashleyw_ left	15:28
thadwheeler	cbreak, I wonder if its because I installed git without ports or fink.. .	15:28
cbreak	I installed git from git	15:28
	but my version is a bit older. and either way	15:29
xlq	let's hope that wasn't your first install, cbrake	15:29
cbreak	it should not change the return code from vim	15:29
xlq	let's hope that wasn't your first install, cbreak	15:29
deskin	I've said before, other VCSs try to hide how things are done from the user; git goes the other way, and requires you to understand how things are done to use it effectively	15:29
thadwheeler	cbreak, vim is 7.2.22	15:29
cbreak	very weird. I have 7.2	15:29
	→ ashleyw_ joined	15:29
cbreak	so, the same	15:30
	← tjogin left	15:30
	← ashleyw left	15:30
	← ashleyw_ left	15:30
thadwheeler	not a big deal really - I can still use the Textmate Bundle or citool	15:30
cbreak	xlq: I stopped counting :)	15:31
	← [RIT]Rawn027 left	15:31
	→ unreal joined	15:31
	← came0 left	15:33
	← ia left	15:34
shep-werk	deskin: I think I am starting to see what is happening here	15:34
	← _Vi left	15:34
shep-werk	so, if I make a change in 7.0 that should not be in dev	15:34
	i commit in 7.0	15:34
	→ jackdempsey joined	15:34
shep-werk	then I merge 7.0 into development with -s ours	15:35
	which just does not change the files	15:35
	but, when I do a future change in 7.0 that should be in dev	15:35
	then I can do a normal merge, and still get that change	15:36
deskin	shep-werk: you got it exactly	15:36
	← jackdempsey left	15:36
shep-werk	im not as dumb as I look ;-)	15:36
deskin	that, and you're learning now; watch out world :)	15:36
	← zjason_ left	15:37
deskin	I'm off to $work, have fun with your merges	15:37
shep-werk	thanks for the help	15:37
	have fun	15:37
xlq	Oh, I get it!	15:38
	→ powerfox joined	15:41
	→ bdiego joined	15:42
powerfox	Hi! Is there any simple way to tell git-merge, that remote repo is always "right", so if any conflict occur then to use file from remote?	15:42
	→ vbabiy joined	15:42
xlq	that's not very wise	15:43
	← Hendu left	15:46
	→ akitada joined	15:46
	→ murphl joined	15:47
	← jonaskoelker left	15:47
Zathrus	if you're doing that, then you may as well just use rsync	15:48
knobo	Can filter get to know about filename?	15:49
ciaranm	where's the list of env vars for git filter-branch these days? the man page points to git-commit, but they're not there	15:49
knobo	git config filter...	15:49
parasti	ciaranm: man git-commit-tree	15:51
Gitbot	ciaranm: the git-commit-tree manpage can be found at http://git.or.cz/man/git-commit-tree	15:51
ciaranm	ta	15:51
	← xanonus left	15:52
	← girishr_ left	15:53
	→ Ramblurr joined	15:54
Ramblurr	i have a git repo that isn't using git-svn, and i need to publish this git repo to an svn repo	15:55
	ideally keeping the history	15:56
	all the git-svn articles i'm finding talk about going from svn->git-svn	15:56
	→ ahupp\|fb joined	15:57
dsop	Ramblurr: I created a svn module, imported it via git-svn than created a graft, merged/rebase my git branch on top of the svn branch and than do a dcommit	15:57
knobo	Is there a way to pass information about the filename to the filter that runs befor commit and after checkout?	15:57
dsop	Ramblurr: maybe there is a better way to do it, but it worked.	15:57
	→ LotR joined	15:58
	← Cacheaway left	15:58
	→ Cacheaway joined	15:58
	→ willb joined	15:58
	→ JackTheRabbit joined	16:00
	← JackTheRabbit left	16:00
	→ JackTheRabbit joined	16:01
	← JackTheRabbit left	16:01
	→ JackTheRabbit joined	16:02
xlq	boingboingboing	16:03
	→ metajack_ joined	16:04
	← esden left	16:04
	→ esden joined	16:05
JackTheRabbit	Hi esden!	16:05
	→ bobmcw_ joined	16:05
	→ lil-murph joined	16:05
esden	ahh hi JackTheRabbit	16:06
	← bobmcw left	16:06
	bobmcw_ → bobmcw	16:07
	→ bronson joined	16:07
	rolfb → rolfb_away	16:07
pflanze	Is there still no way to revert multiple patches in one go? Except doing it manually, or using git rebase -i afterwards?	16:08
	← murphl left	16:08
	→ mw joined	16:11
	← metajack left	16:13
	← sverrej left	16:13
parasti	rebase -i doesn't exactly qualify as "revert"	16:13
madduck	but it can squash	16:13
	pflanze: no. :)	16:13
parasti	ah, squash the manual reverts? ok	16:13
	← Voker57 left	16:13
	← cgardner left	16:14
pflanze	I'm still waiting for the case where I want to remove a whole feature consisting of a series of 200 patches or so.	16:14
	→ metajack joined	16:14
	← bartocc left	16:15
	→ grahal joined	16:15
madduck	which would have been in a branch, right? You can revert the merge...	16:15
	→ iulian joined	16:19
	← Ryback_ left	16:19
	← metajack left	16:20
	→ sgrimm joined	16:22
	→ Ilari joined	16:23
	→ mkdread joined	16:26
	← asn left	16:26
	→ bieneff joined	16:27
mkdread	hi, I have deleted a lot of files in git but I don't know how to do to track the changes without "git rm" one by one. "git add ." does't seem to work. Any idea?	16:28
	← vaskaren left	16:28
	→ stephens joined	16:28
	→ vaskaren joined	16:28
	← metajack_ left	16:28
	→ jackdempsey joined	16:29
	→ not-xjjk joined	16:29
halfline	mkdread: does git-status tell you what you want to know?	16:29
parasti	mkdread: git add -u?	16:29
mkdread	yes, in "Changed but not updated:" it appears	16:30
	thanks parasti , it worked	16:30
parasti	just make sure it didn't add any other changes you didn't want to add	16:31
thadwheeler	deskin, thanks for the help... I switched my editor to textmate. for git.. I will look into the problems I am having later.	16:33
mkdread	ok, and another question, how can I take out some files from the index or all of them without modifiing my working directory?	16:34
Ilari	mkdread: 'git rm --cached'.	16:34
	→ not-xjjk_ joined	16:35
	← JackTheRabbit left	16:37
	← reel_ left	16:40
	→ cgTobi joined	16:40
	→ kumbayo joined	16:41
	← kyle___ left	16:41
	← xjjk left	16:41
	→ johnflux joined	16:42
johnflux	sorry for stupid question - how do I change to specific revision?	16:43
	← vaskaren left	16:43
cbreak	change how?	16:43
johnflux	I know the SHA of a particular commit, so I want to check out the tree at that moment	16:43
cbreak	you can use git rebase -i	16:43
	oh, you can do git checkout <hash>	16:43
parasti	johnflux: well, you answered that yourself :)	16:43
cbreak	but I am not sure if you can commit from a detached head	16:43
johnflux	do I use checkout even if I've already got it checked out?	16:44
Ilari	johnflux: 'git checkout <ID>' (but don't make commits on top of that or those can be lost if you are not careful).	16:44
nadim	pflanze: hi, had any chance to pack what you were going to send me?	16:44
	→ Ryback_ joined	16:44
cgTobi	johnflux: you might be confusing 'git checkout' with 'git clone'	16:44
pflanze	nadim: not yet; I wonder whether you're interested in my packages history?	16:45
johnflux	it gives me a warning 'Note moving to XXX which isnt' a local branch	16:45
Ilari	johnflux: It's just a warning.	16:45
nadim	pflanze: I'm always interrested in others input	16:45
	→ drewr joined	16:45
pflanze	nadim: it would cut time a bit down, although shouldn't matter that much; some time will be finding which scripts there are and which dependencies they have.	16:45
	← not-xjjk left	16:46
	← IRSeekBot left	16:46
nadim	I'm gathering input from multiple sources right now so you input is welcome.	16:46
johnflux	is it possible to checkout for a given date?	16:47
cbreak	no	16:47
johnflux	I need to checkout lots of modules. using the date seems to be the best way to make sure all the modules compile with each other	16:47
cbreak	git is not linear	16:47
johnflux	cbreak: doesn't it still have a concept of which is the 'main' trunk?	16:49
cbreak	what's a trunk?	16:49
johnflux	hmm	16:49
cbreak	if you have a branch that contains merges, it could have two concurrent streams	16:49
	or more	16:49
johnflux	yeah true	16:49
	hmm, how can I do this... I want to bisect Xorg which has lots of modules	16:49
cbreak	do they use git submodules? I think those are bound to the main by revision	16:50
johnflux	how can I tell ?	16:50
	cbreak: no they don't	16:51
	:)	16:51
	← sgrimm left	16:52
	← Yuuhi left	16:52
	→ sgrimm joined	16:53
	← lil-murph left	16:54
	← Strogg left	16:54
	→ Strogg joined	16:55
johnflux	cbreak: it seems that you can do what I want	16:55
	git checkout -d '4 months ago'	16:55
	→ xjjk joined	16:56
	→ fhobia joined	16:56
	→ lil-murph joined	16:57
	→ ceej joined	16:57
	← tango_ left	16:57
	← mkdread left	16:57
cbreak	hmm...	17:00
johnflux	cbreak: sorry wait :)	17:00
cbreak	I wonder how it resolves the ambiguity. /me goes read man page	17:00
johnflux	cbreak: sorry the syntax seems to be something like: git checkout branch@{yesterday} for example	17:01
	but I can't get it to wokr	17:01
	work	17:01
cbreak	that's the reflog	17:01
	imagine you update your repository from the central every month	17:01
	after the update, you do a checkout of half a month ago. you get how your local repository was half a month ago	17:02
	not the central one	17:02
nadim	wow! that's too often ;)	17:02
	→ hobodave joined	17:02
	→ edwardam joined	17:03
johnflux	hum	17:03
	← scientes left	17:03
	← Ramblurr left	17:03
	← bartman left	17:03
	← EmilMedve left	17:04
	→ scientes joined	17:04
	→ EmilMedve joined	17:04
knobo	I'm trying to find out if there is any environment variables I can use with filter (as in gitattributes). So I made a filter wich just cats env(1) to a file. But I don't get any git variables. Is there non?	17:04
	→ not-xjjk joined	17:05
	→ bartman joined	17:06
	→ xanonus joined	17:06
Ilari	knobo: Filter is designed to act as function from set of files to set of files. Doing anything else would break things badly...	17:06
	knobo: s/files/file contents/g	17:07
	knobo: So, two files with identical contents need to map identically.	17:07
	← not-xjjk_ left	17:08
	→ m3y3ring` joined	17:08
johnflux	fatal: ref HEAD is not a symbolic ref. won't bisect on seeked tree	17:09
	what does this mean?	17:09
cbreak	that means you have a detached head	17:09
	you possibly did a checkout on a hash value directly	17:10
	or a reflog entry	17:10
	(I think :)	17:10
johnflux	I did 'checkout HEAD' again...	17:10
	how do I fix it? :)	17:10
Ilari	johnflux: 'git checkout <branch'.	17:10
johnflux	I guess I could give the SHA of the most recent commit	17:10
Ilari	johnflux: 'git checkout <branch>'.	17:10
johnflux	okay I did 'git checkout master'	17:11
parasti	johnflux: using a hash will just do another detached checkout	17:11
johnflux	It says now (when I bisect) won't bisect on seeked tree	17:11
	oh, I need to reset my last bisect	17:12
	why doesn't it just say so!	17:12
	someone make these errors more clear please :-)	17:13
	→ Jacolyte joined	17:13
	← charon left	17:13
johnflux	It should say "Bisect already in progress. Use 'git bisect reset' to reset.'	17:13
Ilari	johnflux: What git version? "seeked tree" sounds like some pretty old version...	17:14
johnflux	git version 1.5.4.3	17:14
	is that old?	17:14
	latest is 1.6.0.4	17:14
	hmm	17:14
Ilari	johnflux: Not terrible, but old...	17:14
	← Ryback_ left	17:15
knobo	Ilari: could you give me an example of what you mean filter is supposed to be used to. I don't understand.	17:15
	← Trepanator left	17:16
	→ albertito joined	17:16
	← xjjk left	17:17
Ilari	knobo: Lets say you have file that uses some pretty odd characters. You can't seem to find editor that can edit those characters properly AND supports UTF-8, but you want files in repo to be encoded in UTF-8. So you set up filters to convert character set from UTF-8 to whatever on checkout and from whatever to UTF-8 on add.	17:19
	→ jeckel joined	17:19
	→ cgardner joined	17:19
Ilari	knobo: Another example is to set filter to decompress files on add so they can be deltified against each other.	17:19
johnflux	Ilari: like ODF files :)	17:20
Ilari	johnflux: Yup. I even have written filter for that...	17:21
johnflux	awesome :)	17:21
	btw git bisect is really cool	17:21
	now that I've got it going :)	17:21
knobo	Ilari: how about rcs $Id$ tags?	17:21
	← radarek left	17:21
Ilari	knobo: Git has export-subst.	17:22
	← m3y3ring left	17:22
	← meyering left	17:22
Ilari	knobo: The idea is: While content is within SCM, use SCM services, and only write explicit version information on export from SCM.	17:22
	← powerfox left	17:24
knobo	Ilari: aha.. It does not fit for everybody though. But you can not please everyone..	17:24
	→ carllerche joined	17:24
knobo	Maybe I have to say with subversion :(	17:25
	→ sverrej joined	17:25
knobo	I'll cry now......	17:25
	I have spent to much time trying to make this work.	17:25
Ilari	knobo: You have some stuff that doesn't require any building?	17:25
	knobo: Maybe look at how git.git (the git itself) gets its version number.	17:26
	knobo: What sort of project you would want that RCS $ID$ braindamage for anyway? :->	17:28
	← hurikhan\|Work left	17:28
knobo	Ilari: it's a lisp project.	17:30
	Ilari: we reacently switched from rcs to subversion. But I thought I could use git via. git-svn.	17:30
	The lisp system reads the $Id$, and remembers which version each function was compiled from.	17:31
	So you can ask it in runtime. which version does this function come from.	17:31
johnflux	knobo: that's considered bad practise	17:31
	knobo: it introduces merge conflicts :)	17:31
	← gitte left	17:31
knobo	subversion introduces merge conflicts.	17:32
Ilari	knobo: Maybe put special file into tarball that tells the version. Otherwise the lisp code could ask Git about the version...	17:32
	← jmesnil left	17:32
johnflux	yeah have your make system get the version from git and put that into a header file	17:33
	then have your files include that header file	17:33
	← alb left	17:34
	→ willb1 joined	17:34
Ilari	knobo: For just ID of version 'git rev-parse --verify --quiet HEAD'.	17:35
	→ tango_ joined	17:35
	→ willb2 joined	17:35
	← nud left	17:36
Ilari	knobo: That dumps 40 hex digit version ID to stdout.	17:36
knobo	Ilari: I could use the ident function, and rewrite the code that reads the $Id, just to make it compile in my development repository.	17:37
	Then deploy from somwhere else.	17:37
	← cedricv left	17:37
	→ pkern joined	17:37
pkern	Hi there. git alternates could only refer to other local repositories, right? I can't specify on a repo exported by dump http that one might find more objects at another repo exported with git-daemon?	17:38
Ilari	knobo: 'ident' basically writes (modified) SHA-1 of FILE into the file.	17:38
	→ cedricv joined	17:38
	← willb1 left	17:38
Ilari	knobo: Its not revision ID!	17:39
	→ gilimanjaro joined	17:40
	← cedricv left	17:40
	← cgardner left	17:40
	→ cory joined	17:40
	cory → bantic	17:40
	← tango_ left	17:43
	→ ijcd_ joined	17:44
	← Cacheaway left	17:45
	→ Cacheaway joined	17:45
Ilari	pkern: Yes, alternates can only refer to local repositories. There's also HTTP alternates, but it is only used by HTTP fetching.	17:46
	← fhobia left	17:47
	→ dfarnsworth joined	17:48
	← willb left	17:49
Ilari	pkern: And I think that even those HTTP alternates can only refer to other repositories on the same server.	17:49
dfarnsworth	What's a good way, using existing git commands, to test if a commit is in a branch?	17:50
Ilari	dfarnsworth: 'git branch --contains <ID>'.	17:51
	← ceej left	17:51
	→ ankit9 joined	17:51
dfarnsworth	Ilari, hey! that' easy. In fact it rings a bell. Thanks.	17:52
	→ ceej joined	17:52
pkern	Ilari: Do you know how HTTP alternates would be specified, by chance?	17:53
	→ ia joined	17:53
Ilari	pkern: Actually, http-alternates can refer to other servers. But it is only read by fetch over HTTP (from REMOTE side!)	17:56
	← muthu_ left	17:57
dfarnsworth	Ilari, 'git branch --contains <ID>' shows me all branches that contain ID. So, it's slow. Any easy way to just check one branch, so I can speed it up? I can do it comparing the results of a couple of 'git-rev-list --max-count=1 ...'. I'm wondering if there is a better way.	18:01
	→ cgardner joined	18:03
	→ kyle__ joined	18:04
Ilari	dfarnsworth: 'git rev-list <branch> \| grep `git rev-parse <id>`'?	18:04
	← mcella left	18:04
Ilari	dfarnsworth: Actually, 'git rev-list <branch> \| grep `git rev-parse <id>` \| head -1'. The head should quit after first line, killing grep with SIGPIPE, which should kill rev-list with SIGPIPE.	18:06
	→ tango_ joined	18:08
	← JonathanT left	18:09
	← lil-murph left	18:12
knobo	I dont understand what's wrong with rcs $Id$. It's been working for us since 1998 perfectly.	18:12
pkern	Ilari: How? (And yeah, I do have a http server but I want to spread the load by telling the clone/fetch client to get most of the stuff elsewhere.)	18:12
pflanze	Is it possible to merge a branch from another repository with a small depth (i.e. 1)?	18:13
	Probably my question could be rephrased: is it possible to create grafts which work for cloning by other people?	18:13
	(I'm assuming that the shallow clone feature is implemented using grafts, maybe I'm mistaken)	18:14
flazz	is #github now forwarded here?	18:14
	← tango_ left	18:14
	→ kRutOn joined	18:15
kRutOn	hello	18:15
pflanze	What I want to do is create a new repo which merges stuff from other repositories as subtree merges; but I only want to provide the history of that other projects from a particular point on;	18:15
Ilari	pflanze: Pull that branch?	18:15
albertito	knobo: man gitattributes	18:15
Gitbot	knobo: the gitattributes manpage can be found at http://git.or.cz/man/gitattributes	18:15
	← johnflux left	18:16
pflanze	and I want to still make it possible to incrementally merge subsequent changes done to these other repositories.	18:16
albertito	knobo: faq keyword expansion (I don't know how the faq bot works)	18:16
Ilari	pflanze: I don't think that's possible...	18:16
	← nibbo left	18:16
albertito	knobo: apparently not that way =) look for "Does git have keyword expansion?"	18:16
pflanze	So I'm back to rebasing / cherry picking somehow?	18:17
dfarnsworth	Ilari, the best I've found is 'git rev-list --reverse <id>^..<branch> \| head -1'. I think that works. :)	18:17
	← sping_work left	18:17
	← agile left	18:17
	← ankit9 left	18:19
	→ Voker57 joined	18:19
	→ bobesponja joined	18:20
	→ murphl joined	18:20
pflanze	Is there a tool to take new patches from one branch and apply them to another? Remembering the last-applied patch.	18:20
	And modify the commit message to contain a "upstream $repo $commitid" like line.	18:20
	Sort of like the stable kernel branch is made.	18:21
	Actually pretty much exactly the same thing.	18:21
knobo	albertito: I know about gitattributes.	18:22
pflanze	Any offering?	18:22
	→ vbabiy_ joined	18:22
	→ nibbo joined	18:23
	← not-xjjk left	18:24
flazz	if i want a gemspec picked up by github need i name it in the gemspec user-projname?	18:26
	→ xjjk joined	18:28
cgTobi	I try to use git in a mixed platform environment (Linux/OSX/WindowsXP). It works pretty well so far, except that I can't find a way to pull from windows machines. Can somebody give me a hint how to do that?	18:30
	→ boto_ joined	18:31
	→ tango_ joined	18:31
	← thadwheeler left	18:33
xlq	cgTobi: oh I see, because git-instaweb doesn't work on Windows?	18:33
	← naeu left	18:33
cgTobi	xlq: yeah, for example	18:34
	xlq: git-deamon is also not available	18:34
	→ david_koontz joined	18:34
	→ alley_cat joined	18:34
	→ maronnax joined	18:35
xlq	cgTobi: really? Oh :(	18:35
cgTobi	xlq: I also tried to setup a ssh server on windows, but I couldn't make it prompt to the git-bash	18:35
	xlq: I should note that I use msysGit on the windows machine	18:35
maronnax	This is a total noob question, but is there a good way to push the equivalent of a git-trunk to a subversion server? I know hardly anything about git but I'm interested in trying it out. However, the project I work on (it's just me really) is hosted on sourceforge and I use their svn servers. So I need some good way to push the changes I would make in the 'real' git repository to subversion. Does anyone have good advice here?	18:37
	→ ankit9 joined	18:37
	→ Sigma` joined	18:37
	← vbabiy left	18:38
parasti	maronnax: hmm, are you just looking for git-svn?	18:39
	→ putrycy joined	18:39
maronnax	parasti, cool, I think that's totally what I'm looking for.	18:40
parasti	if you have Git installed, you might have git-svn already (git svn --help to check)	18:42
	← gilimanjaro left	18:42
parasti	otherwise, look for it in your distros package repo or compile it	18:42
	← simmerz left	18:42
parasti	though I guess you just need Perl and there's nothing to compile	18:43
	→ joydivider joined	18:43
	← boto left	18:46
	← EiNZTEiN left	18:46
	Sigma` → Sigma	18:47
maronnax	parasti, well thanks	18:49
	I still have a fair amount of reading as to how best to locally administer git and set it up and all that, but knowing this, I can jump into it, you know?	18:50
	← jfontan left	18:50
	→ CodeOfficer joined	18:51
	→ geckosen1tor joined	18:51
Ilari	cgTobi: If you could mount the filesystem containing the git repository somehow...	18:51
xlq	cgTobi: can you do it over normal http?	18:53
	→ nud joined	18:53
	← geckosenator left	18:53
	→ perezd88 joined	18:53
	→ ReinH joined	18:53
cgTobi	Ilari: I'll probably have to do just that or stick with pushing	18:54
	xlq: you can't push over http AFAIK	18:55
xlq	cgTobi: oh, you want pushing too	18:55
	→ juanjoc joined	18:55
	→ eddyp joined	18:55
cgTobi	xlq: sorry, no, pushing works fine via ssh, my fault. I mean I don't have a webserver on the windows machines	18:56
	xlq: and its not supported by msysgit as far as I can tell	18:57
pflanze	maronnax, parasti: I'm not sure git-svn is suited for pushing git history to svn; rather the other way around; anyway I think that question has been asked on the git ml at least once.	18:57
	→ tokkee joined	18:57
	→ altrux_ joined	18:57
pflanze	The "normal" usage of git-svn is to import svn into git, then add linear history with Git on top and only push that linear history back.	18:58
	rolfb_away → rolfb	18:58
pflanze	In any case you loose information with the pushing back (author, date, and non-linear history).	18:58
altrux_	howdy, i have a repo that had some of the production api keys commiited to it osme time ago, andi need to remove those, preferably retroactivly	18:59
	how do i go about removing the history of a file	18:59
Ilari	The usual usage for Windows machines is to keep them on only when working on them and then turn them off... While Unix machines tend to be up 24/7...	18:59
	That's the classical usage anyhow...	18:59
	altrux_: man git-filter-branch	18:59
Gitbot	altrux_: the git-filter-branch manpage can be found at http://git.or.cz/man/git-filter-branch	18:59
altrux_	thanks so much!	18:59
	i didn't know what to search on, my google_fu is weak today	19:00
	← eddyp left	19:00
pflanze	maronnax: also, I've seen people writing about how to mis-use google code svn repositories for publishing Git projects, but I don't remember exactly how they did it; it will be hackery in any case.	19:00
	maronnax: note that google seems to be going to add Git to their offerings some time.	19:01
	deskin wonders why you'd want to abuse google code when you can use github	19:01
	→ mauricio joined	19:01
Voker57	deskin: because github has no bugtracker, wiki and downloads?	19:01
mauricio	guys	19:02
	i've checked out a remote branch	19:02
	and set a local branch to track it	19:03
	now i get this after some work	19:03
	"Your branch is ahead of the tracked remote branch 'origin/css_changes_shaun' by 3 commits.	19:03
	"	19:03
	and it doesn't really seems to make sense to me	19:03
xlq	I appear to be unable to clone from an instaweb	19:03
mauricio	i've used this tutorial to get this working -> http://www.zorched.net/2008/04/14/start-a-new-branch-on-your-remote-git-repository/	19:04
	boto_ → boto	19:04
pflanze	mauricio: this is exactly what you should expect, isn't it?	19:04
	you did 3 commits, right?	19:04
mauricio	yep	19:04
	but if i push	19:04
	i get noshing	19:04
	"already up to date"	19:05
	but the commits aren't being pushed	19:05
pflanze	ah, you need to mention the branch probably	19:05
	← ahupp\|fb left	19:05
pflanze	git push origin css_changes_shaun	19:05
	maybe you aren't on that branch locally, btw; check with git branch -av	19:06
mauricio	* css_changes_shaun_local 374983f Free news items	19:06
	master 0af7241 specs	19:06
	origin/HEAD 0af7241 specs	19:06
	origin/before_new_design f0e3810 Revert "Beta invites implementation"	19:06
	origin/css_changes 4982e21 Merge branch 'css_shanges_shaun' of ssh://mauricio@dev.talkies.de/repos/talkies.git into css_changes	19:06
	origin/css_changes_shaun ab43ee3 Refactor CSS. More components added to the styleguide.	19:06
	origin/css_changes_shaun_local 374983f Free news items	19:06
	origin/css_changes_shaun_test f554636 Merge branch 'css_changes_shaun' of ssh://ivor@dev.talkies.de/repos/talkies into css_changes_shaun_test	19:06
	origin/css_shanges_shaun 374983f Free news items	19:06
	origin/master 0af7241 specs	19:06
	origin/new_design 9fa8cfc Tweaking actor detail page	19:06
	origin/rails_pre_2.1 f79d97f adding host to test mailer	19:06
	this is the output	19:06
	right now i'm at css_changes_shaun_local	19:06
pflanze	please don't paste more than about 3 lines.	19:06
mauricio	ops	19:07
	sorry	19:07
pflanze	if necessary, use a paste service for it.	19:07
	→ ahupp\|fb joined	19:07
mauricio	[branch "css_changes_shaun_local"]	19:07
	remote = origin	19:07
	merge = refs/heads/css_changes_shaun	19:07
	this is the branch config	19:07
	which seems to be fine for me	19:07
	(not that i'm specifically good with git anyway...)	19:07
	← ahupp\|fb left	19:08
pflanze	yep, and your remote and local are the same commit, 374983f	19:08
	so, where exactly does git (still?) say you're 3 commits ahead?	19:08
parasti	pflanze: that's a good point -- I only asked about git-svn because the OP mentioned an existing project, but it makes no sense to host a Git project in Subversion	19:08
mauricio	holy shit	19:08
	→ synchris joined	19:09
mauricio	pflanze, well, there was no css_changes_shaun_local at the remote branch	19:09
	→ dreiss joined	19:09
	→ unreal_ joined	19:09
xlq	Git-instaweb is working (I can see localhost:1234 in a web browser). How can I git-clone from that?	19:09
mauricio	pflanze, or at least there shoudn't be one, this is my local branch that tracks css_changes_shaun	19:10
	→ jfontan joined	19:10
pflanze	mauricio: yes, you've probably created that one now	19:10
	← nibbo left	19:10
mauricio	pflanze, how?	19:10
pflanze	did you run my command? if so, that's the culprit, sorry	19:11
mauricio	pflanze, :( no prob	19:11
	pflanze, but do you have any idea about wha'ts happening?	19:12
	why is it creating a new branch when it should push to the tracked branch?	19:12
pflanze	two solutions: either get the git config right, or use correct push syntax;	19:12
	since I'm not using the latter frequently maybe someone else can tell.	19:12
mauricio	pflanze, and how do i get git config right?	19:13
pflanze	well, man git-push says, refspec can contain both the source and target parts	19:13
Gitbot	well: the git-push manpage can be found at http://git.or.cz/man/git-push	19:13
mauricio	pflanze, it seems to be ok	19:13
pflanze	your config only contains a merge, not push entry, it seems	19:13
mauricio	hum	19:13
pflanze	it only has remote = origin, and assumes the name of the remote branch is the same as the local	19:14
	you really probably want to merge back your local branch to the css_changes_shaun branch	19:14
	i.e. "git checkout css_changes_shaun; git merge css_changes_shaun_local; git push"	19:14
	that would be the +-normal way of doing things.	19:15
	You can configure git to use different remove branch names for local branches;	19:15
	but it may only confuse you in the long term.	19:15
	→ nud_ joined	19:15
	← xlq left	19:16
	→ dkerschner joined	19:17
mauricio	pflanze, if i git checkout css_changes_shaun i get this:	19:17
	try error: pathspec 'css_changes_shaun' did not match any file(s) known to git.	19:17
	→ evalsaidem joined	19:17
mauricio	pflanze, if i try git checkout origin/css_changes_shaun	19:18
	it seems to be ok	19:18
	→ brandon joined	19:18
mauricio	i can merge and it seems to work	19:18
pflanze	mauricio: ah yep, you don't have such a branch actually.	19:18
mauricio	but when i pull i get another 'its already upt do date'	19:18
	*when i push	19:18
brandon	what does 'fatal: impossibly long personal identifier' mean if git svn rebase fails?	19:18
pflanze	mauricio: you need to rename your local branch	19:18
	git checkout master; git branch -m css_changes_shaun_local css_changes_shaun	19:18
	→ nice_burger joined	19:18
	→ charon joined	19:19
	→ naeu joined	19:19
pflanze	then git checkout css_changes_shaun; git push	19:19
	Maybe the "git branch -m" works also while you'r on that branch, I don't remember.	19:20
mauricio	OMG!	19:20
	man	19:20
	← willb2 left	19:20
mauricio	it seems to work	19:20
	pflanze, thanks!	19:20
	← nud left	19:20
mauricio	i've been struggling with this since morning	19:20
brandon	anyone have an idea about my git svn issue?	19:20
pflanze	mauricio: for some reason, you checked out that branch under a name which differs from the remote branch name; you should not do that :)	19:21
mauricio	pflanze, the tutorials said it wasn't a problem :(	19:21
	← putrycy left	19:21
pflanze	well, in which way not a problem?	19:21
	→ nibbo joined	19:21
pflanze	can you tell the URL?	19:21
	→ niki joined	19:21
	→ not-xjjk joined	19:21
pflanze	ah, got it	19:21
	→ SyntaxRebel joined	19:22
mauricio	pflanze, http://www.zorched.net/2008/04/14/start-a-new-branch-on-your-remote-git-repository/	19:22
	look at use branch from another location	19:22
	← ankit9 left	19:22
mauricio	pflanze, git checkout --track -b new_branch origin/new_feature_name	19:22
SyntaxRebel	why does git always use "more" as its pager? even why i set GIT_PAGER=less -R, it still uses more! :S	19:22
	← qrush left	19:22
pflanze	mauricio: he says "git checkout --track -b new_feature_name origin/new_feature_name";	19:23
	and of course you can choose another name; but then, as you've seen,	19:23
mauricio	if you've created the branch yourself	19:23
	look after this	19:23
pflanze	you need to tell Git where to push to.	19:23
mauricio	at "use branch from another location"	19:23
	i'm really hoping my pragmatic version control using git arrives before christhmas	19:24
	hopefully these problems will not appear again :)	19:24
pflanze	It's not a problem.	19:24
mauricio	pflanze, for me that doesn't know what's happening, it is :)	19:24
pflanze	He doesn't pretend that after this command Git knows where to push it back to, from what I can see.	19:24
	Why not write him a reply and ask for how he'd set it up so that pushing works too?	19:25
	← unreal left	19:26
pflanze	mauricio: btw mention your Git version, since newer Gits might have changed behaviour to add push too upon that --track option, dunno.	19:27
Ilari	Later git versions recognize push refspec ':' which acts as the default push all matching.	19:27
	→ jberg- joined	19:27
mauricio	1.5.6	19:27
SyntaxRebel	does git use GIT_PAGER for help pages?	19:28
	→ rtomayko joined	19:28
mauricio	pflanze, that's what i'm about to do, just checking out that the push ocurred correctly	19:28
pflanze	Ilari: hm, not sure what "all matching" would mean here; he's got (or: he had) a local branch name differing from the remote branch name.	19:29
charon	SyntaxRebel: 'git foo --help' is just a fancy way to spell 'man git-foo'	19:29
Ilari	mauricio: Maybe look at 'git ls-remote' output and see what it prints for pointer value for that branch and then compare to what was pushed?	19:29
SyntaxRebel	charon: as it "git help foo"?	19:29
	as is*	19:29
pflanze	Ilari: git branch -av should be as good for this, right?	19:29
	← kumbayo left	19:29
SyntaxRebel	yep, i need to change PAGER :)	19:29
charon	SyntaxRebel: yes.	19:30
Ilari	pflanze: ls-remote looks at the remote repository, git branch only looks at local repository...	19:30
charon	actually, there appear to be configuration options for both the format (man, info, html) and viewer used, but it defaults to man using 'man'.	19:30
pflanze	But after a push, the remote and local will be the same.	19:30
	His problem is that the branch was of a different name.	19:30
	And .git/config only had a merge entry, none for push.	19:31
	For that local branch.	19:31
mauricio	pflanze, yep, both branches pointing to the same commit :)	19:31
pflanze	It seems that "git checkout --track -b new_branch origin/new_feature_name" doesn't set up a push entry.	19:31
	Only a merge one.	19:32
	Or at least, it did this for his git version.	19:32
	← niki left	19:32
SyntaxRebel	charon: thanks btw. :)	19:32
	← xjjk left	19:32
pflanze	mauricio: anyway, this new_branch example doesn't make any sense to me; if he wanted to demonstrate adding multiple remotes, i.e. another one than "origin", then using a different local branch name would make sense.	19:33
mauricio	pflanze, i guess the idea is that you can track the new branch from anywhere	19:34
pflanze	Otherwise, you check the remote branch out under the same name; you can always create new branches locally from that afterwards,	19:34
mauricio	pflanze, and looking at the comments, there's a guy saying that it wasn't working for him too	19:34
pflanze	but you usually merge back changes to the original branch then before pushing; this way things stay clear.	19:34
	Yes, but then his example would have to look like:	19:35
	git remote add $someurl	19:35
	sorry	19:35
	git remote add otherorigin $someurl	19:35
	and then	19:35
	git checkout --track -b new_branch otherorigin/new_feature_name"	19:35
mauricio	yeah	19:35
pflanze	only then it would make sense to use a different local name,	19:35
	since you could conflict with the same branch name from another origin server.	19:36
Ilari	The use for ':' is as 'remote.foo.push' value and as push refspec when other refspecs are explicitly specified...	19:37
	← kyle__ left	19:38
	← evalsaidem left	19:38
pflanze	Ilari: I'm not sure I understand your sentence. You'd do "git config remote.foo.push :"? Would this make git look at the "merge" refspec and then figure out from that where to push to?	19:40
	(mauricio: just in case you're running Debian or Ubuntu and want to try a newer Git: http://repo.or.cz/w/gitcj-debian.git)	19:42
	→ ankit9 joined	19:42
mauricio	i'm at openSUSE :)	19:42
	if i get a newer version locally	19:43
	do i need to upgrade the server?	19:43
	pflanze, just to be sure :D	19:43
pflanze	nah, if the version matters at all for this case, it will be the local one.	19:43
	← jonshea left	19:47
	← ceyusa left	19:48
	→ pygi joined	19:48
	→ xjjk joined	19:49
	→ qrush joined	19:51
cgardner	Are there any tutorials on writing hooks for git? Aside from the man pages?	19:52
mauricio	cgardner, do you mean hooks like post-commit?	19:52
pflanze	Look at the example scripts in .git/hooks/, I think they are all documented	19:52
cgardner	mauricio: yes.	19:52
mauricio	cgardner, do as pflanze said, they are well documented	19:53
altrux_	ok so i had 2 git-filter-branch commands to run, and the first one was flawless, now im getting a 'Namespace refs/original/ not empty' error, do i just need to find this Namespace refs/original/ and delete it?	19:53
	sorry kind of worried about fubaring my archive	19:53
	← synchris left	19:53
pflanze	altrux_: yes, trash .git/refs/original/$thefile	19:53
mauricio	cgardner, they'll give you the variables you want, it's just a matter of writting what you want in your preferred language	19:54
pflanze	altrux_: if you're on a recent enough git, it has "git reflog" and you can use that to get back to previous state anyway	19:54
altrux_	ok cool	19:54
	i got backups before is tarted too	19:54
	→ orkim joined	19:54
	← mauricio left	19:56
	← bartman left	19:56
cgardner	mauricio: Is this the only documentation for the hooks? http://www.kernel.org/pub/software/scm/git/docs/githooks.html	19:56
	← murphl left	19:56
cgardner	that's all I could find in the online documentation	19:56
	err... on the kernel.org documentation	19:57
pflanze	cgardner: have you looked at the example scripts?	19:57
cgardner	yeah, they don't really help a lot	19:57
	← sdboyer left	19:57
cgardner	I've figured it out, I just was wondering if there was any other documentation...	19:57
pflanze	What did you figure out which was not documented?	19:57
	I suggest you just write a patch to either the scripts or the docs, and send it to the git ml.	19:58
	→ niki joined	19:58
	← not-xjjk left	19:59
	→ bartman joined	19:59
	→ _Vi joined	20:00
	← lorandi left	20:00
	← AaronLBorg left	20:02
	→ sdboyer joined	20:03
	← nice_burger left	20:04
	← DrNick left	20:07
	← flazz left	20:10
	→ flazz joined	20:11
	EiNZTEiN_ → EiNZTEiN	20:11
	→ synchris joined	20:11
	→ __ash__ joined	20:11
	→ _psychic_ joined	20:11
__ash__	um.... anyone in here a guru of git?	20:12
	← grahal left	20:12
Ilari	__ash__: Well, shoot anyway... :->	20:12
	← brandon left	20:12
	← agib left	20:13
	← iulian left	20:13
__ash__	error: remote 'refs/heads/master' is not an ancestor of	20:13
	local 'refs/heads/master'.	20:13
	whenever i do a push i get a	20:13
	i get that ^^	20:13
	and i was wondering what is the right way to fix this issue	20:13
Ilari	__ash__: Maybe somebody else has updated the remote repository?	20:14
__ash__	its sorta steming from a bad suggestion by someone to do a git reset --soft HEAD	20:14
	no, i am the only one that has	20:14
pflanze	If nobody fetched from your repository, use git push --force	20:14
__ash__	is there a way i can just tell it to override whats on the remote?	20:14
	gotcha	20:15
pflanze	If somebody did fetch, you'll screw up his history a bit.	20:15
	→ DrNick joined	20:16
	← ebzzry left	20:16
	→ lanwin joined	20:17
pflanze	(and/or give him merge conflicts, depending on your local changes)	20:18
	→ agile joined	20:18
	→ kennethlove joined	20:19
jogla	where can i find documentation about submodule?	20:20
	man git-submodule does only explain some comments	20:20
Gitbot	jogla: the git-submodule manpage can be found at http://git.or.cz/man/git-submodule	20:20
	← ankit9 left	20:21
jogla	hmm.. the description section is missing in my manpage	20:21
	→ ahupp\|fb joined	20:22
	← cgTobi left	20:22
	edwardam → edwardam_lunch	20:23
	→ ezyang joined	20:23
	← nud_ left	20:24
	← hobodave left	20:24
	→ hobodave joined	20:25
	→ cgTobi joined	20:27
	→ krunk-\|work joined	20:28
krunk-\|work	I was getting these errors when trying to push/pull from a remote repo: zsh: command not found: git-receive-pack	20:30
	some googling indicates this is because only .bashrc/.zshrc is sourced and not, say, .zshenv. is there a way to change this behavior so that the environment file is also sourced?	20:31
cgTobi	is there a big difference between making a copy of a git repo and cloning it (locally)	20:32
	?	20:32
jogla	when you clone it, your origin will be set	20:33
azuk	krunk-\|work, that doesn't sound credible	20:33
	← _psychic_ left	20:33
deskin	cgTobi: minor difference: cloning will hard-link the pack files, so you'll start off with minor additional disk usage	20:33
Ilari	krunk-\|work: Put PATH redefintion in .zshrc?	20:33
cgTobi	jogla & deskin: cheers	20:34
	← bartman left	20:34
krunk-\|work	Ilari: I have a really extensive zshenv, it'd be kludgy to do so and is not recommended as the "most correct way" by the zsh docs	20:35
	azuk: hm, what could it be then? I'll put the needed path in the rc just to test	20:35
	→ bartman joined	20:36
	→ c0sin joined	20:36
	← jeckel left	20:37
azuk	I assume you are pushing/pulling over ssh? what exactly would run zsh in such a way that .zshenv isn't read? it just doesn't sound sensible to me	20:37
krunk-\|work	azuk: sounded weird to me too, honestly	20:39
azuk	actually, it's precisely .zshenv that I would expect to be read	20:39
	→ b_jonas joined	20:41
krunk-\|work	ohhh I think I know what it is, I think I ran into this before. need to check something.	20:41
	← lukedal left	20:41
	→ drizzd joined	20:42
b_jonas	I think I'll have to try git one of these days because I hate svn more and more as I use it	20:42
	← Voker57 left	20:45
	← nothinghappens left	20:45
Ilari	b_jonas: Why won't you do little test project to learn the basics of how to use it locally?	20:46
drizzd	b_jonas: make; make install; mkdir test; cd test; git init; touch file; git add file; git commit -m myfirstcommit; that's it!	20:47
krunk-\|work	10m and a cheat sheet and you're ready to go with git	20:48
b_jonas	drizzd: yeah, that much is ok	20:49
	if I only had to do that svn worked too	20:49
	it's the more complicated parts where I get annoyed with svn	20:49
krunk-\|work	pretty sure this is a due to an error being output by one of the scripts sourced in /etc/profile on the server. the error causes ssh to abort	20:49
	I think I ran into it with a non-git related issue	20:50
b_jonas	the slow speed is the least annoyance -- recently I did a mass renaming of twelve files and each 'svn rename' took like three seconds	20:50
	but there are also more important problems	20:50
	→ willb joined	20:50
Weiss	b_jonas: i found git hardgoing at first when i started doing anything more trivial than "commit", but after a few days it felt very natural. it's best to have someone patiently explain it to you, and read that "git for computer scientists" document	20:51
	← alikins left	20:51
bmalee	We've had so much trouble with svn branching at work recently...makes me even more glad of git. Don't think I'll manage to get the rest of the department converted, though...	20:51
b_jonas	there does not seem to be an easy way with svn to make it treat a file as a copy of another (as in svn copy or svn move) if I already have copied the file in the work dirs -- it insists on wanting to actually copy the file	20:51
	→ perezd88_ joined	20:53
b_jonas	I'm using svn mostly for managing projects only I work on on my home computer btw	20:54
	← perezd88_ left	20:54
b_jonas	and I'd use git for the same as well	20:54
PerlJam	b_jonas: what?	20:59
	b_jonas: svn groks symlinks	20:59
	→ nud_ joined	20:59
	← pkern left	21:00
b_jonas	PerlJam: it does, but you can only create them in two ways, and neither helps for me:	21:01
	either you svn copy a file in your working copy, which it does not want to do if you already have created the target file	21:02
	so in that case you have to move the target file, do the svn copy, and move it back to that name, and commit	21:02
	which suks	21:02
	(the other way is to directly svn copy the file in the repo, but then you need to check it out which gives the same problem and it also requires another commit for the editing)	21:03
	→ kyle__ joined	21:03
	← bieneff left	21:05
Ilari	b_jonas: Git works in terms of snapshots. That's why it doesn't have 'git cp'. And 'git mv' is purely for convience...	21:05
	→ RichiH joined	21:06
RichiH	ok, question	21:06
	i have a repo, foo.git	21:06
	it pulls from a reomte source	21:07
	into a branch called upstream	21:07
Ilari	RichiH: Branch, not a remote?	21:07
	← perezd88 left	21:08
Ilari	RichiH: Ah, its bare repo, so probably a branch...	21:08
RichiH	Ilari: does not really matter, it's hypothetical anyway. as i am not very sure of git's nomenclature, assume whatever makes more sense ;)	21:08
	three branches, feature1, feature2 and fix1 are based on upstream	21:09
Ilari	RichiH: Or a namespace (would also make sense in bare repo).	21:09
RichiH	the master branch merges feature1, feature2 and fix1 (think topgit)	21:09
b_jonas	if I understand right, svn copy (and svn move) works by setting the parent of a file another file in the same repo, so at next commit it will diff from that file, and the logs and history later will follow the copies	21:09
	but there doesn't seem to be a way to just directly set the parent of a new file when I add it	21:09
	so the interface is inconvenient	21:10
RichiH	now, i tag the current state of master	21:10
	and develop in feature2, create fix2 and merge it into master, as well	21:10
	master is updated with the contents of feature2 and fix2	21:11
	← _Vi left	21:11
RichiH	1) will that update of master happen automagically?	21:11
	2) is there a way to create a tar.gz from tagged state of the master branch even though it's not the most current, any more?	21:11
Ilari	RichiH: Well, somebody has to create the new state...	21:12
uau	b_jonas: if you care about that in svn then you'll like git even less	21:12
	because git lacks any functionality to express where a file comes from	21:12
Ilari	RichiH: 'git archive <committish>' can accept any valid commit ID to produce archive from. Tag names are valid committishes.	21:13
RichiH	Ilari: is that an answer to 1)?	21:13
	ok, perfect	21:13
Ilari	RichiH: I think its not updated automatically.	21:14
uau	RichiH: branches basically never change automatically	21:14
b_jonas	uau: really?	21:14
uau	you have to explicitly check out that branch and merge stuff into to incorporate changes from other branchs	21:14
b_jonas	uau: how would it store representations efficently?	21:14
RichiH	ok, thanks	21:14
	makes sense, that way	21:14
b_jonas	uau: I mean, they say they store the repo compactly	21:14
	so it has to handle file moves	21:14
	→ kumbayo joined	21:14
	← cgTobi left	21:15
RichiH	but even if i updated the master branch, i will always be able to create a tar from the state it had some time in the past?	21:15
uau	b_jonas: the storage is in a lower layer; how it works has no direct impact to anything visible in the UI	21:15
Ilari	b_jonas: It collapses identical copies using hashes. And it has heuristics to figure out what objects to delta against each other.	21:15
	→ joshdavey_ joined	21:15
uau	except performance	21:15
	← joshdavey left	21:15
b_jonas	Ilari: I see	21:15
Ilari	RichiH: The master branch pins those states, so they are available for archive creation.	21:16
drizzd	essentially it's like creating an archive for all objects, just smarter	21:16
uau	b_jonas: git has some heuristics to guess what were supposed to be moves/copies after the fact (when looking at the history later)	21:16
RichiH	Ilari: perfect, thanks	21:16
uau	but they're neither particularly reliable nor work consistently (in the sense of exactly how each tool uses those heuristics to affect its output)	21:17
	→ tjogin joined	21:17
b_jonas	pity	21:17
	the moves are not only convenient because of the efficent representation	21:18
RichiH	hmm	21:18
b_jonas	but because this way when you query the log of a moved file you'll get the logs of the old file too	21:18
RichiH	can i use the same tag several times?	21:18
	i.e. can i tag master STABLE several times?	21:18
drizzd	b_jonas: that usually also works with git	21:18
RichiH	to create a moving target which can be used to generate tar.gz?	21:18
drizzd	b_jonas: furtheremore, with git, you're not usually interested in the history of a file	21:18
	b_jonas: you're either interested in the history of the entire project, a subdirectory, or maybe just a few lines of a file. But rarely do you care about any file in particular.	21:19
uau	RichiH: you can delete tags and create new ones using the same name	21:19
RichiH	b_jonas: git is different from svn. it does not really know/care about files. all it does is manage diffsets	21:19
	uau: can i tag one state with two names?	21:19
uau	but IIRC that does not work well with cloning (outdated tags are not automatically updated on repositories that pull from the updated one)	21:20
RichiH	i.e. STABLE and STABLE_081205	21:20
uau	yes there are no limitations on that	21:20
b_jonas	well, svn also cares less about files than cvs	21:20
	← esden left	21:20
	→ alikins joined	21:20
drizzd	b_jonas: how so?	21:20
	→ aknowles joined	21:20
b_jonas	drizzd: cvs does not have moves, so each file is separate, also cvs has separate version numbers for each file so there are no versions which represent a version of the whole project unless you use a tag or a date,	21:21
RichiH	uau: k	21:21
	thanks	21:21
drizzd	b_jonas: I see. Well that's similar to git then.	21:22
b_jonas	also cvs commit handles each file as a separate commit, with separate commit messages and no relations.	21:22
	edwardam_lunch → edwardam	21:22
b_jonas	I don't really care about this because I don't use cvs anymore for some time	21:22
	I used to work with cvs for a little time	21:22
	in a larger project	21:23
	and I also tried it at home back when I didn't really know about the other version control systems	21:23
RichiH	can i tag a state that is in the past?	21:23
uau	RichiH: you can tag arbitrary objects	21:23
	even things other than commits	21:23
RichiH	uau: ok	21:23
Ilari	RichiH: Anything that has object ID can be tagged.	21:24
uau	like individual file contents or a directory tree	21:24
RichiH	but due to me being able to delete tags, i can create inconsitent states, no?	21:24
uau	what do you mean by "inconsistent"?	21:24
b_jonas	uau: ah, so in git tags and branches are different things?	21:24
uau	consistency of what?	21:24
Ilari	RichiH: Even other (annotated) tags can be tagged. :-)	21:24
RichiH	uau: well, if i refer to another git repo by tag and that tag moves	21:24
uau	b_jonas: well... depends on how you view it	21:25
b_jonas	iirc svn handles branches as a special kind of tag which have a default future history line too, not just a past history line	21:25
	s/svn/cvs/	21:25
	sorry	21:25
RichiH	or some other contruction which relies on a tag	21:25
Ilari	RichiH: Well, to name a state in globally signaficant manner, you need to use its commit ID.	21:25
b_jonas	branching is one thing that's I think done better in cvs than svn	21:25
Ilari	RichiH: Tags are only 'semi-global'.	21:25
RichiH	Ilari: that's what i meant, yes	21:25
b_jonas	but the other differences outweigh that	21:25
	← shep-werk left	21:25
RichiH	so that is by design, then	21:25
b_jonas	though the cvs way isn't that good either	21:26
RichiH	not a decission i would make, but OK. now that i know, it can be accounted for	21:26
uau	b_jonas: in git both branches and tags are at their most basic just simple pointers to objects, "name"->SHA1_HASH	21:26
b_jonas	uau: but a branch has to know it future lineage	21:26
uau	a branch points to its current state	21:26
b_jonas	that is, if I understand correctly, in git you can have separate commits that have the same commits as parents	21:26
Ilari	RichiH: In git, the only real difference between branches and tags is that HEAD can only point to branch (or be detached) and thus only branches can be committed on.	21:27
b_jonas	I See	21:27
uau	when you make a commit in a branch you first create a new commit object representing the new state, then update the branch pointer to point to that state	21:27
b_jonas	so a branch works by having a "current" branch and making that point to the new version when you commit?	21:27
	that makes sense actually	21:27
	← bdiego left	21:27
uau	so creating a new branch is basically just creating a pointer that tells the value of one SHA1 hash	21:28
Ilari	b_jonas: HEAD is usually pointer to current branch. When you commit, that branch is updated.	21:28
uau	(though in practice there are extra convenience features like branch reflogs which tell how the branch has been edited over time)	21:28
b_jonas	yes	21:28
	whereas svn has linear time so you can't have separate branches of the same file at a time, and cvs has linear time per branch, so you can only have one next version of a version in each branch	21:29
	or something like that	21:29
azuk	RichiH, note that git pull doesn't change move/delete tags	21:29
Ilari	b_jonas: Those are called symrefs (symbolic references) they pretty much function like symlinks (in fact, Git interprets ref file being symlink by treating it as symref to corresponding place).	21:29
RichiH	so are tags valid within, but ot across repos?	21:30
Ilari	RichiH: Tags are valid only in repo that they are made in until they are transferred to other repos.	21:30
	RichiH: You can push and fetch tags.	21:30
	→ mosno joined	21:31
RichiH	so while i pull code, i fetch tags?	21:31
	→ mjc joined	21:32
RichiH	i.e. use different commands?	21:32
uau	yes by default you fetch tags from the remote repo too (though there are options which affect that behavior)	21:32
	but by default tags which have changed in the remote repo are NOT updated later	21:33
azuk	RichiH, no, you get tags with pull as well, pull is fetch + merge	21:33
	→ esden joined	21:33
azuk	RichiH, I suggest you read some tutorial / user's manual	21:34
krunk-\|work	hm, I see what's going on now. looks like git 1.6.x does not put the git-upload-pack and such binaries in $path/bin but in $path/libexec but the push/pull operation is looking for them in bin	21:34
	er I need to add the $path/libexec to my path	21:35
Ilari	RichiH: And tags don't belong to any 'branch' they are essentially 'branch-like' themselves (for instance, you can create new branch starting from tag).	21:35
doener	krunk-\|work: actually, upload-pack and such should be in bindir	21:35
	krunk-\|work: http://git.pastebin.com/m2082a9f8	21:36
	→ not-xjjk joined	21:36
RichiH	azuk: i am doing that, but bouncing ideas off other people's minds is faster/better in some circumstances	21:37
krunk-\|work	doener: huh, wonder what went awry with my install	21:37
	doener: just checked another system with git installed by macports, they're in bin. but on this one where I installed by hand, they're not	21:37
RichiH	Ilari: wouldn't the tag need to be persistant once i branch off from it?	21:37
krunk-\|work	maybe a config option	21:37
RichiH	or is the tag i branch from just the commit in branch x? i.e. a softlink?	21:38
Ilari	There are three really important things in Git repository: Pointer to current branch, refname to commit ID mapping and object ID to object contents mapping.	21:38
	→ ajonat joined	21:38
Ilari	RichiH: No, you don't need to keep the tag after you branch from it (unless you want to).	21:39
	In Git, even single branches can have nonlinear time.	21:39
	And the set of commits that belongs to some branch is always transitively (following parent pointers) closed.	21:40
	← pencilo left	21:41
	← chris2 left	21:42
	← netoman left	21:42
Ilari	In fact, the defintion of set of commits on branch is: The smallest set that 1) Contains commit designated by branch pointer and 2) is transitively (following parent pointers) closed.	21:42
	← kRutOn left	21:43
	← c0sin left	21:44
RichiH	Ilari: so i can basically pull the rug from under a branches' feet and it doesn't care in the least	21:45
	← vbabiy_ left	21:45
	→ vbabiy joined	21:45
Ilari	RichiH: In fact, 'parent of branch' is only recorded in config file and pretty much the only command that cares is 'git pull' (when run without arguments).	21:47
	← xjjk left	21:47
RichiH	makes sense	21:47
	RichiH has to say that the underlying design of git appears to be extremely clean	21:48
	← orkim left	21:48
	← jfontan left	21:48
	← aspotashev left	21:49
Ilari	RichiH: Yes it is. Git for instance only has four types of objects: Blobs store file contents, trees build them into directories, commits bind them into history and tags introduce trust into the system.	21:49
	→ jfontan joined	21:50
RichiH	tags as in the tags i can set myself?	21:50
	← boto left	21:50
Ilari	RichiH: 'tags' as in tag objects ('git tag -a' or 'git tag -s' creates those).	21:50
RichiH	Ilari: so each commit gets one tag by default?	21:52
	the commit info tag, so to speak	21:52
Ilari	RichiH: Nope. Tag objects have to be explicitly created. Maybe look at 'git for-each-ref' output and see entries with type 'tag'.	21:53
	RichiH: Ordinary branches always have type 'commit' in that output.	21:53
drizzd	Ilari: what you said above could be the slogan for a git ad	21:54
	← PhilK left	21:56
Ilari	RichiH: 'tags' can have both 'commit' for lightweight tag or 'tag' (the ID printed is then tag ID) for annotated/signed tag as type...	21:56
	← nibbo left	21:59
	← unreal_ left	22:00
RichiH	Ilari: k, thanks :)	22:01
	→ unreal joined	22:01
Ilari	RichiH: Basically, tag objects store freeform message about what is tagged (just like commit messages contain freeform description of the commit).	22:03
	→ pkern joined	22:04
pkern	Could I filter some subtree from a git repo, even if I have to rewrite all the history for it?	22:04
	→ JonathanT joined	22:05
	→ nibbo joined	22:05
RichiH	pkern: afaik, partial checkouts are not supported	22:05
	← synchris left	22:05
pkern	RichiH: I want to get rid of it in the repo.	22:06
	← loincloth left	22:08
	← LiamH left	22:08
	→ loincloth joined	22:09
	→ flaguy joined	22:10
	← ceej left	22:11
	← loincloth left	22:15
	← kumbayo left	22:16
	→ sanity_ joined	22:17
	← bobmcw left	22:17
	← ReinH left	22:18
	← rtomayko left	22:18
	← petteri left	22:18
	→ petteri joined	22:18
	→ rtomayko joined	22:18
	← rphillips left	22:20
	→ rphillips joined	22:20
	← MrHeavy left	22:21
	→ bobmcw joined	22:25
	← cmarcelo left	22:26
	← kyle__ left	22:28
	→ metajack joined	22:28
	← vbabiy left	22:29
	← up_the_irons left	22:29
	→ up_the_irons joined	22:30
	→ ceej joined	22:30
	← ahupp\|fb left	22:30
	← Cacheaway left	22:31
	← b_jonas left	22:32
	← ezyang left	22:32
	← metajack left	22:33
	→ ReinH joined	22:34
	← kennethlove left	22:35
	cbrake → cbrake_away	22:38
	→ ahupp\|fb joined	22:39
	→ alikins_ joined	22:43
	← alikins left	22:43
	→ cops joined	22:50
	← parasti left	22:50
	← cops left	22:50
	← hobodave left	22:51
	← juanjoc left	22:52
	← JonathanT left	22:55
	← jfontan left	22:55
	→ xjjk joined	22:56
	→ Cacheaway joined	22:57
	← smtlaissezfaire` left	23:00
	→ thiago_home joined	23:00
	← mosno left	23:01
	← incidence left	23:01
	← aknowles left	23:02
	→ not-xjjk_ joined	23:03
	← bantic left	23:04
bremner	pkern: you saw man git-filter-branch ?	23:04
Gitbot	pkern: the git-filter-branch manpage can be found at http://git.or.cz/man/git-filter-branch	23:04
	→ parasti joined	23:06
	← lucsky left	23:06
	→ heda joined	23:07
	← not-xjjk left	23:07
mutex	all of the sudden when I see git diffs	23:08
	it is full of lines like the following	23:09
heda	hi all, i want to run a project whereby certain files are provided by a remote git repository, is there a way i can set it up so that i can easily get the latest version of all the remote files but leave the rest of the local files as they were - an svn externals sort of effect	23:09
mutex	ESC[32mif (parse_args(argc, argv, &args) != 0) {ESC[m	23:09
	i guess less isn't processing colors ?	23:09
thiago_home	yes	23:09
	what's $LESSO ?	23:09
	$LESS ?	23:10
mutex	-R -M --shift 5	23:10
	also my .gitconfig says	23:10
thiago_home	the -R says process ANSI colours	23:10
mutex	pager = less -+RSX	23:11
	← agile left	23:11
mutex	yeah so then why isn't my display rendering them is the question	23:11
	→ schacon joined	23:11
thiago_home	and -+R turns it back off	23:11
	you need -R	23:11
mutex	oh	23:11
	ah ha!	23:12
	works now ;-)	23:12
	← priidu left	23:13
	→ bdiego joined	23:14
	→ rem7 joined	23:15
	← xjjk left	23:15
	← rem7 left	23:16
pflanze	git filter-branch is somehow horribly slow with a little longer history for me.	23:18
	It's taking 25 seconds to handle a single commit right now.	23:19
	And it's only a few hundred files.	23:19
	→ xjjk joined	23:19
pflanze	I want a tool which can take a list of paths that I want to preserve,	23:20
	and removes all other paths, and removes commits which don't touch any of the paths I want to keep.	23:20
	What do you suggest?	23:20
	← ibuffy left	23:21
thiago_home	filter-branch should do that quite fast if you use --index-filter	23:21
	← SyntaxRebel left	23:22
thiago_home	what are you using?	23:22
pflanze	git filter-branch --index-filter 'git ls-files -z \| perl -wn -0 -l0 -e '\''BEGIN { open my $in, "<","../_bin_with_renames_" or die; %want=map { chomp; $_=>1} <$in>; close $in or die;} print unless $want{$_}'\'' --commit-filter 'if [ z"$1" = z"$(git rev-parse "$3"^{tree})" ]; then skip_commit "$@"; else git commit-tree "$@"; fi'	23:25
	well instead of ../_bin_with_renames_ I'm using an absolute path since it changes cwd.	23:25
	It is now at about 2 minutes to treat one commit.	23:26
thiago_home	what process is taking so long CPU time?	23:26
	you can see it in top?	23:26
pflanze	git-filter-branch	23:26
	99% cpu	23:26
thiago_home	disk I/O or just CPU?	23:26
pflanze	just cpu	23:26
thiago_home	hmm	23:26
	no clue then, sorry	23:27
pflanze	hm strange is that strace -p $pidofgitfilterbranch -f shows:	23:27
	rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0	23:27
	rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0	23:27
	brk(0xbad000) = 0xbad000	23:27
	brk(0xb2d000) = 0xb2d000	23:27
	brk(0xbad000) = 0xbad000	23:27
	brk(0xb2d000) = 0xb2d000	23:27
	brk(0xbad000) = 0xbad000	23:27
	brk(0xb2d000) = 0xb2d000	23:27
	brk(0xbad000) = 0xbad000	23:27
	brk(0xb2d000) = 0xb2d000	23:27
	in a loop.	23:27
	and almost nothing else.	23:27
	← name left	23:29
pflanze	so it looks like bash is using up all cpu.	23:29
cehteh	pflanze: allocating memory for datastructures in a loop	23:29
pflanze	yeah but what datastructure is bash using here?	23:29
cehteh	does it end eventually?	23:29
pflanze	it's still running.	23:30
	for 13 minutes now	23:30
cehteh	strange .. how many memory does it consume? .. and if thats bash then it looks fishy	23:30
ciaranm	pflanze: did you forget to disable the swedish chef?	23:30
cehteh	i dont know the filter branch code .. does it use arrays or other fancy bash features?	23:30
pflanze	when it started, it went on processing commits very quickly as I can see the "rm" messages from git rm; now it takes >2' per commit.	23:31
	PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND	23:31
	9978 chris 20 0 15072 5316 1132 R 100 0.3 13:15.95 git-filter-bran	23:31
	← not-xjjk_ left	23:31
pflanze	ciaranm: what's that?	23:31
ciaranm	it's a muppet who says bork bork bork a lot	23:32
	it's funny because bork sounds like brk!	23:32
pflanze	aha	23:32
	← willb left	23:32
pflanze	That repo has 697 commits.	23:36
	I mean that's tiny with what other people have right?	23:36
	pasky: are you there, do you have an idea?	23:37
	← lanwin left	23:38
pflanze	btw this is GNU bash, version 3.2.39(1)-release (x86_64-pc-linux-gnu)	23:38
	→ aFlag joined	23:39
aFlag	hi. When I try to push something git yields an error and advise me to use [email@hidden.address]	23:39
	if I use push like that, it works, but isn't there a way to save that so push always push there?	23:40
	instead of trying to push to git://github.com/user/repo.git	23:40
pflanze	you need to configure your remotes correctly	23:41
	man git-remote	23:41
Gitbot	pflanze: the git-remote manpage can be found at http://git.or.cz/man/git-remote	23:41
pflanze	and maybe you need some other command; I still usually edit .git/config directly.	23:41
	Which is probably not the right thing, but that's why I can't tell you better.	23:42
	← nud_ left	23:42
pflanze	I'd assume that github has some docs on how to set up pushing, really!	23:43
	So go RTFM and then tell me how it works ;)	23:43
aFlag	hm, I'm not really able to access github web site from here, it doesn't work too well on lynx	23:44
parasti	also, #github	23:44
pflanze	aFlag: there are also the "links" and "w3m" console browsers, maybe those work better.	23:45
	← rolfb left	23:45
	→ raalkml joined	23:46
	← raalkml left	23:47
	← aroben left	23:47
	→ ibuffy joined	23:47
pflanze	pasky: .git-rewrite/map/ contains 278 files right now (I guess that means 419 of my 697 commits still to go, and it's getting slower..)	23:48
	← xanonus left	23:48
pflanze	pasky: the interesting thing is that these files grew much larger, and:	23:49
	chris@novo:~/repo.or.cz/cj-git-perl-tools/.git-rewrite/map$ cat da0053442bc6dfbd6c530e76f75e25a3256148f5\|wc -l	23:49
	4800	23:49
	chris@novo:~/repo.or.cz/cj-git-perl-tools/.git-rewrite/map$ cat da0053442bc6dfbd6c530e76f75e25a3256148f5\|sort -u\|wc -l	23:49
	1	23:49
	→ aspotashev joined	23:49
pflanze	So I guess bash is searching through these lists of identical lines?	23:49
	← sdboyer left	23:49
	→ sdboyer joined	23:50
pflanze	heh actually: chris@novo:~/repo.or.cz/cj-git-perl-tools/.git-rewrite/map$ cat *\|sort -u\|wc -l	23:50
	1	23:50
	← naeu left	23:50
	→ gilimanjaro joined	23:51
pflanze	nadim: are you there? I'm pondering writing a Perl based filter-branch..	23:52
	What do you think? (I guess, using the gitweb gsoc code)	23:53
	I mean, by the time that bash finishes here I've probably done it :)	23:54
	PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND	23:55
	9978 chris 20 0 15584 5972 1132 R 100 0.3 35:34.58 git-filter-bran	23:55
	← ahupp\|fb left	23:55
	→ _Vi joined	23:56
	← bobmcw left	23:57
	← Sigma left	23:57
	← alley_cat left	23:57
pflanze	btw, filter-branch from git version 1.6.0.4	23:58
	← gilimanjaro left	23:59
	→ gilimanjaro joined	23:59

Logs Search ←Prev date Next date→

Channels Documentation