IRCloggy #git 2008-01-30

Provider of IRC logs since 2005.

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

	2008-01-30
wjlee	how should I let proj1 keep track new commits?	00:02
	← Sonderblade left	00:03
	→ glommer joined	00:03
	→ Pupeno joined	00:06
	← bentob0x left	00:07
	← saimon left	00:09
	← trapni left	00:09
	→ HG` joined	00:14
	← Yuuhi left	00:16
	← adante left	00:19
	← loswillios left	00:21
	→ loswillios joined	00:21
	← lukass left	00:22
corecode_	TGEN: you need to run with -i	00:23
	TGEN: git execs more in the parent and does the work in the background	00:23
	TGEN: you can also set GIT_PAGER=cat or something like that	00:23
	TGEN: then it won't fork	00:23
	← nud_ left	00:24
TGEN	k	00:26
duncanm	sigh, is it the case that none of the emacs ifaces to git (vc-git.el, git.el) support branches?	00:27
TGEN	45314 git CALL mmap(0,0x2000000,0x1,0x2,0x3,0,0x2d869000,0)	00:29
	45314 git RET mmap -1 errno 12 Cannot allocate memory	00:29
	45314 git CALL mmap(0,0x2000000,0x1,0x2,0x3,0,0x2d869000,0)	00:29
	45314 git RET mmap -1 errno 12 Cannot allocate memory	00:29
	so yeah	00:29
	← HG` left	00:29
	→ HG` joined	00:29
TGEN	(git -i didn't work on 1.5.0)	00:29
	anyhow, it's likely the default kernel MAXDSIZ value is to blame	00:30
mugwump	duncanm: vc-git.el displays the current branch name in the status bar	00:32
duncanm	mugwump: can i change the branch?	00:32
mugwump	TGEN: check on the mmap man page what those arguments imply	00:32
	see if git is being a bad program or not	00:32
	duncanm: dunno. I guess that would be more of a git.el thing	00:32
	I'm not sure what being able to switch branches from my editor would win me	00:33
duncanm	i don't think git.el knows how to change branches	00:33
	← HG` left	00:34
duncanm	mugwump: if i change branches from the terminal, and edit/commit from emacs, it'll still go to master, wouldn't it?	00:34
	→ HG`` joined	00:34
TGEN	mugwump: nothing wrong with the arguments, just the number of consecutive mmap()s which together make the datasize limit trip over, which is hardwired in the kernel to be 512Mibyte max \|:(	00:34
duncanm	because emacs only knows to commit to the master branch, afaict	00:34
careo	no. it'll go to whatever branch you're on regardless	00:34
duncanm	but it still says 'master' on the modeline	00:34
mugwump	"show other version" will work	00:34
wagle_	duncanm, emacs might get confused over what its got in its buffers and whats in the files	00:35
mugwump	you'll probably need to kill and reload the buffer after switching branches	00:35
	which is arguably a bug	00:35
duncanm	hmm	00:35
	it'd be nice if i can just switch branches from without emacs/vc/git.el	00:35
	s/without/within/	00:36
mugwump	well, it's just lisp ... go write it :)	00:36
wagle_	it often notices that the underlying file changed out from under the buffer (and offer to update buffer, or save to the file), but i don't know how much it checks to detect that	00:36
careo	could you just run the command in the shell from within emacs?	00:37
	← kzak__ left	00:37
	← loswillios left	00:39
	→ loswillios joined	00:39
duncanm	i don't even see the code in vc-git.el or git.el for setting the modeline	00:39
	→ ahx joined	00:41
mithro	is there a way to create "aliases" for branches in git?	00:43
	I would like to have a "stable" branch alias which maps to a given branch	00:44
	← kukks left	00:44
wagle_	have tags, but they wouldnt follow the branch head around	00:44
	→ kukks joined	00:45
ahx	hi. When i do a pull ,is it possible to have it accept all incoming changes? - I dont want to resolve any given conflicts, instead i want accept every change that pull introduces (only this time).	00:45
wagle_	could make a foo-<fullversionnumber>-stable tag every so often	00:46
mithro	well, we already tag when a release occurs	00:46
wagle_	others might have more ideas than i	00:46
mugwump	mithro: see if git symbolic-ref is your answer	00:47
Eridius	ahx: maybe you should submit a patch to add a new merge strategy "theirs"? There's already an existing strategy "ours"	00:48
ahx	ah ok. i didnt know of "ours"	00:49
	thanks	00:50
duncanm	Eridius: i thought rebase is that way to go for things like that	00:50
Eridius	duncanm: normally yes, but it seems like ahx wants to blow away any of his local changes that conflict with upstream	00:50
gitster	why not "fetch && reset --hard origin"?	00:51
duncanm	i'm trying to learn to use branches in my workflow, so far, i've just been using master	00:52
mithro	mugwump: I'm looking at that, and I'm not sure how I would use it	00:52
	Eridius is assuming ahx doesn't want to lose non-conflicting changes. ahx? What is it you're trying to do here?	00:52
mugwump	mithro: on the server, git symbolic-ref refs/heads/stable refs/heads/master	00:52
	would point "stable" to "master"	00:52
duncanm	Eridius: so i push my master branch to my public server, but if i want to setup a devel branch, and i push to that, i can't rewrite history locally and still keep in sync with the server, right?	00:53
Eridius	if you've pushed your own changes out to a remote repo, then no, you don't want to rebase	00:53
duncanm	hmm	00:54
	← ReinH left	00:55
mithro	mugwump: that seems to be what I want	00:56
	← cmarcelo left	00:56
	← KirinDave left	00:57
mithro	mugwump: thanks	00:57
ahx	Eridius, yes. I dont want to loose my non-conflicting changes. Just accept all conflicting changes.	00:57
duncanm	Eridius: come to think of it, i guess that's okay, because i can do the commit merging when i merge the devel branch back to master	00:58
Eridius	ahx: you could create a patch from your changes, reset --hard, pull, then attempt to reapply your patch	00:58
	→ KirinDave joined	00:59
	← glommer left	00:59
ahx	Eridius, ok. I understand. Thanks for help and good night.	01:08
	← ahx left	01:08
	← KirinDave left	01:09
	→ reval joined	01:10
	→ adante joined	01:13
	→ FunkeeMonk joined	01:14
	← HG`` left	01:24
	← flaguy left	01:24
	← mithro left	01:29
	→ mithro joined	01:30
	← devogon left	01:30
	→ adante_ joined	01:33
	→ KirinDave joined	01:34
	→ glommer joined	01:34
	← GSaldana left	01:38
ebzzry	Hi! For small personal projects, is it ok to make small frequent commits?	01:45
tpope	sure	01:46
halorgium	ebzzry: it is ok for all projects :)	01:46
tpope	that's okay on big projects too	01:46
halorgium	as long as they are atomic	01:46
	← FunkeeMonk left	01:46
ebzzry	Ohh, so changes, no matter how small, are ok to commit, even in rapid succession?	01:47
halorgium	commits are local remember	01:48
tpope	ebzzry: on a larger project, you might consider going back and cleaning them up again before pushing them	01:48
	← adante left	01:48
	adante_ → adante	01:48
tpope	but generally, "commit early, commit often" is the way to go	01:48
ebzzry	Ok.	01:49
	→ garrin joined	01:53
	← zapnap left	01:53
	← glommer left	01:56
	→ kylem_ joined	01:57
	← reval left	01:57
Ori_B	oh, question	02:03
	is there a short cut for "git-rebase origin/mybranch"?	02:03
	→ glommer joined	02:03
Ori_B	er. the git-fetch; git-rebase sequence, rather.	02:04
	something like git-update?	02:04
doener_	in 1.5.4 there's git pull --rebase	02:04
Ori_B	ah. still a bit verbose, though.	02:05
	(ie, not tab-completable)	02:05
doener_	you can configure your branch to default to rebasing instead of merging	02:05
tpope	in a good shell it is (or will be)	02:05
Ori_B	tpope: heh. fair enough.	02:05
	← kanru left	02:07
	→ fowlduck joined	02:11
fowlduck	der	02:11
	← kylem left	02:13
	→ reval joined	02:14
	← tvachon left	02:14
	← timocratic left	02:16
	← KirinDave left	02:17
	→ imyousuf joined	02:17
	← Strogg left	02:19
	→ antares joined	02:21
	← fowlduck left	02:21
	→ Strogg joined	02:21
	→ jj- joined	02:21
	→ moh_ joined	02:23
	← garrin left	02:23
	← gebi left	02:28
	→ gebi joined	02:28
	← wjlee left	02:32
	← railk left	02:33
	← RobotDeathSquad left	02:37
	← reval left	02:38
imyousuf	under what license is git released?	02:38
tpope	gpl 2	02:39
laj	http://git.kernel.org/?p=git/git.git;a=blob;f=COPYING;h=6ff87c4664981e4397625791c8ea3bbb5f2279a3;hb=HEAD	02:40
tpope	and yes, all I did was look for that file	02:40
	though I had a very strong suspicion	02:40
	→ garrin joined	02:42
imyousuf	thanks	02:43
	→ reval joined	02:43
	← reval left	02:49
imyousuf	A supplementary question regarding licensing - If I use git to develop a system which uses it for VCS and distribute it under commercial license (not GPL) will that be a problem?	02:50
DrNick	consult your lawyer	02:52
imyousuf	DrNick: before I do that just wanted to know the usual practice	02:53
	and that will be done at a later stage	02:53
tpope	imyousuf: sounds very sketchy. it might could be done in the same way nvidia makes linux kernel modules	02:55
	with clearly drawn boundaries etc	02:56
	ianal obviously	02:56
imyousuf	tpope: we will not be modifying any code of git, but use the git commands	02:56
mugwump	imyousuf: absolutely not.	02:56
imyousuf	and distribute them	02:56
mugwump	read the GPL	02:57
	it specifically grants blanket use of the application for any purpose	02:57
tpope	yeah the fact you'd have to distribute git yourself sounds like the deal breaker	02:57
mugwump	oh, right. sure, if what you're building is a derived work	02:58
	(as shaky a term as that is internationally)	02:58
imyousuf	yes mugwump mine is a TOTALLY usage or derived work	02:58
	the tricky part is distribution, am I allowed to distribute it with a (not under a) commercial license, the GIT license will be intact	02:59
cehteh	mhm ... how the other way around, i work on this webfrontend which i would like to be under the Affero GPL3 .. but it links libgit	02:59
mugwump	cehteh: GPLv2 is compatible with GPLv3, that's fine	03:00
cehteh	i hope so	03:00
mugwump	well, imyousuf, if you are building a program that uses git as a library, the requirements of the GPL may apply	03:01
cehteh	its agpl ... which is more proper for a webfrontend i think	03:01
imyousuf	GIT is rather used as a software whose commands the executed by the program, not the library	03:01
mugwump	yes, but it's not so clear-cut, because you'll likely be using the "plumbing" commands	03:02
cehteh	imyousuf: depends ...	03:02
	i hope libgit becomes more useable over time	03:02
mugwump	the general principle is that you can't defeat the GPL just by wrapping a network or CLI layer around them	03:02
imyousuf	I see	03:03
mugwump	However that doesn't stop you distributing git source and your programs to people	03:03
	← pombreda left	03:03
mugwump	I think	03:04
	AIUI you will be safe by avoiding plumbing commands, or by using a simple enough subset of commands that you could re-implement those portions for your release	03:05
imyousuf	hmm, my understanding was if I mention explicitly about GIT's licenses and user rights and distribute the source of git along with my program (under a commercial license) that should not be a problem	03:06
mugwump	If it was LGPL that would be the case	03:06
imyousuf	mugwump: can you give me an example of a plumbing command?	03:06
mugwump	'git-cat-file' is plumbing; 'git-show' is porcelain	03:06
	← tongueroo left	03:07
imyousuf	I see, so git-commit, git-submodule, git-diff, they are porcelain?	03:07
mugwump	right	03:07
	all the commands whose output is difficult to parse because it's designed for people :)	03:07
imyousuf	great, because I will be using the porcelain commands only	03:07
mugwump	well, the regular "ruling" (actually, statement from the FSF) on copyright of output from GPL commands applies	03:08
cehteh	mugwump: sorry to tell you GPL3 isnt compatible with GPL2 .. or actually i think i have a problem	03:08
mugwump	cehteh: do you understand why or can you cite a reference for that?	03:09
cehteh	http://www.gnu.org/philosophy/license-list.html	03:09
	Please note that GPLv3 is not compatible with GPLv2 by itself. However, most software released under GPLv2 allows you to use the terms of later versions of the GPL as well.	03:09
	... and the AGPL3 is even more special, but has provisions to be compatible with GPL3 ... not really so with GPL2	03:10
mugwump	more at http://www.gnu.org/licenses/rms-why-gplv3.html	03:11
	→ kanru joined	03:12
cehteh	well i guess git wont be upgraded to gpl3 anytime soon	03:12
	(i mean git is not the kernel imo here is gpl3 more proper than for the kernel, but well)	03:13
	madewokherd reads the first few paragraphs and thinks "wow, that would be a mess if things started using gpl3"	03:13
	→ devogon joined	03:14
mugwump	Interesting, I thought that the clause was not that restrictive	03:16
	This file is licensed under the GPL v2, or a later version at the discretion of Linus.	03:16
	^^ What a Bonehead	03:16
cehteh	hehe	03:16
imyousuf	can we actually request the git team (or Linus) to use a little lesser restrictive license, such as LGPL or CDDL? or even dual license git?	03:17
cehteh	imyousuf: i think you dont get the idea about free software :)	03:18
	→ aroben joined	03:18
tpope	imyousuf: you can try all you want, wouldn't expect much	03:18
cehteh	actually i have a problem to use a more strict license :)	03:18
imyousuf	:) oh ok, I am actually wanting to learn :)	03:19
tpope	it's not like he picked it out of the blue, he's been using it with the kernel for a dozen plus years	03:19
	← jj- left	03:19
	cehteh wonders if a "may i use your humble gpl2 project as library in a agpl3 project" has any chance	03:20
mugwump	Don't use the CDDL	03:21
	That's a stupid license	03:21
cehteh	tpope: well kernel and git have different objectives, as i saied before, gpl3 for git wont do any problem i think, but its not my decision	03:21
mugwump	The person who wrote it got up on stage and admitted they included terms specifically designed to make it incompatible with the GPL	03:21
tpope	cehteh: I think he'd be more amenable to your case than imyousuf	03:22
	but, perhaps not enough to switch	03:22
imyousuf	mugwump: I do agree it is horribly incompatible with GPL	03:22
cehteh	tpope: i dont want a switch, just an agreement/exception	03:22
	→ eMBee joined	03:23
mugwump	no doubt those retarded and pointless clauses were invented by twats like Joerg "you guys are all idiots" Schilling	03:23
	s/invented/inspired/	03:23
cehteh	haha	03:24
mugwump	cehteh: you'd need to get all contributors to agree	03:24
imyousuf	btw can I use the output of a git porcelain in a product other than GPLed?	03:24
mugwump	it's not just Linus	03:24
	→ jj- joined	03:24
cehteh	mugwump: no copyright assignment for git?	03:24
tpope	to save you a trip to wikipedia that's the cdrecord guy	03:25
mugwump	imyousuf: "use" of GPL programs is generally considered to be fine - there's a FSF statement on this	03:25
	cehteh: no, there isn't. Just the signed-off-by: lines	03:25
	← tcoppi left	03:25
imyousuf	ah ok, thats in fact great	03:25
cehteh	i've previously done 'gpl2 only' things too .. but contributions should be gpl2-or-any-later, or copyright assignment	03:25
imyousuf	than I guess I can put forward of using GIT as the product for VCS in this case :)	03:26
tpope	now you've got me worried	03:26
mugwump	imyousuf: sure, and don't forget, reimplementing plumbing commands isn't that difficult	03:27
tpope	I wrote some glue that basically helps you coordinate a git-svn repo with a trac website	03:27
imyousuf	yup I can always have a look at the existing ones :), but the porcelains will suffice my requirements	03:27
	→ KirinDave joined	03:28
imyousuf	actually so that I can use GIT (and not Mercurial) in the product I will have to spend hours coming up with a NetBeans VCS plugin for GIT :) but I will happily do it to be able to use GIT :)	03:28
tpope	lots of running things like git-apply and git-rev-parse behind the scenes	03:28
	I licensed it MIT, now I'm wondering if that's an issue	03:28
	→ tcoppi joined	03:28
tpope	my understanding is it'd be okay but I'm not real clear on the details	03:28
mugwump	MIT grants permission to relicense, so it's fine	03:29
tpope	I just wasn't sure if the gpl would "infect" my work	03:29
	not that I'm all that picky about the license for a petty project like this, I'm just flowing with the community	03:30
imyousuf	another question :) : I want to come with a Java Native Interface (another project not the one mentioned above) API for git, can I dual license it under CDDL and GPLv2?	03:31
	the reason for stupid CDDL is	03:31
	→ adante_ joined	03:32
imyousuf	my company are afraid of using GPL libraries (without Classpath exception) in our commercial products.	03:32
tpope	lots of sun projects are dual licensed under the gpl and cddl	03:32
	← KirinDave left	03:32
imyousuf	yeah, but I will be using GPLv2ed git as lib for this (though in this case I will not be distributing git), that is why I am asking whether I can use CDDL or simple single license it under LGPL	03:33
cehteh	no only gpl2	03:34
mugwump	in general the more permissions you grant the better	03:35
	assuming they're good permissions :)	03:35
imyousuf	:)	03:35
	so which one actually grants the most "good" permissions :)	03:35
	?	03:35
cehteh	mugwump: the better yes .. but still incompatible	03:35
mugwump	Well, dual CDDL / GPLv2+ seems reasonable	03:35
cehteh	when you include a gpl2ed library you cant release your code as lgpl	03:36
mugwump	sure you can	03:36
	you can license any code under anything you want	03:36
	← bobesponja left	03:36
mugwump	(any code you are the copyright holder for, that is)	03:37
	→ adante__ joined	03:37
mugwump	whether that impacts on other people's licensing, that's the issue	03:37
cehteh	ok yes .. but then you give it away with a incompatible license which translating users have no license at all	03:38
mugwump	hence ... the more licenses you grant permission under, the more likely it is the users will be able to find a combination that works	03:38
imyousuf	I get it mugwump and cehteh	03:39
cehteh	mhm	03:39
	and i am more confused :)	03:39
imyousuf	so now the question is can a CDDLed product use GPL2 libs?	03:39
	→ jerbear joined	03:40
mugwump	heh, not if it's only CDDL	03:40
cehteh	actually you can make a GPL3'ed program linking a propietary lib .. so i see no reason why i cant link a gpl2'ed lib to my gpl3 program	03:40
imyousuf	I see, so when someone wants to use my product under CDDL will that be possible with a GPL2ed lib in the classpath?	03:41
madewokherd	everyone just use something public-domain equivalent, it's easier >_<	03:41
imyousuf	btw the product is dual licensed under GPLv2 and CDDL	03:41
mugwump	If it exposes the programming interface (ie, the plumbing) of the GPLv2 app out the other side, then the program using it will need the GPLv2 as well	03:41
	If it's just doing abstract things, ie a model of operation invented by your library, then probably not	03:42
cehteh	madewokherd: not really, here in germany, public domain is legally not possible	03:42
imyousuf	so I can actually plumb the plumbing to avoid the issues :)	03:42
mugwump	sure. say your library used the plumbing, and the exposed a generic, VCS-agnostic interface on the other side	03:43
	You're not using the combined program as a library	03:43
tpope	cehteh: really, how so?	03:43
	I've never heard that	03:43
mugwump	Well, actually, maybe you are. hrmm.	03:43
	Well, I hope we've helped enough for you to write a nice e-mail to the FSF lawyers to ask :)	03:44
imyousuf	yeah, mugwump :) now I at least know what to ask :)	03:44
	thanks mugwump, cehteh and tpope	03:45
cehteh	tpope: we differentiate between copyright which is transferable (ownership?) and creatorship which is not, when you created something you did it you cant tansfer this fact	03:45
	PD in the US sense means you drop all rights/privileges you have ..	03:46
	but in germany creatorship has some state you cant drop	03:47
tpope	is there a practical difference?	03:47
	← adante left	03:48
	adante__ → adante	03:48
cehteh	not much, rather obscure cases	03:48
	→ pergesu joined	03:50
tpope	interesting to know	03:51
	cehteh is not a lawyer	03:51
pvh	so I think I just really screwed up my "git rebase -i"	03:51
doener_	there's always the WTFPL	03:51
	← mithro left	03:51
mugwump	WTFPL is dangerous	03:51
	→ mithro joined	03:52
mugwump	it doesn't disclaim responsibility	03:52
	← adante_ left	03:52
madewokherd	I was thinking of MIT	03:52
	mugwump not a lawyer either, but was deeply involved with preparing this: http://www.e.govt.nz/policy/open-source/open-source-legal2	03:53
pvh	At the end of my rebase I got the cryptic error message "cannot lock the ref". What gives?	03:53
mugwump	pvh: you just need to delete the lock and continue probably	03:54
pvh	mugwump: where would i find the lock?	03:54
mugwump	it'll be under .git/refs	03:54
pvh	is it a .file?	03:54
	imyousuf just send an email to FSF :)	03:54
pvh	mugwump: i don't see anything that looks like a lock file using find .git \| grep lock or grep lock .git -R	03:56
	← kukks left	03:56
jerbear	is anyone here familiar with PIL?	03:57
mugwump	the parrot language?	03:57
jerbear	oops, wrong channel	03:58
mugwump	pvh: shrugs ... strace it to see what it's trying	03:58
	← lamont left	03:58
	→ lamont joined	03:58
	← lamont left	03:59
	→ KirinDave joined	04:02
	→ lamont joined	04:02
	→ hharrison joined	04:04
	← Strogg left	04:06
	→ Strogg joined	04:06
	→ kylem joined	04:07
	← sgrimm left	04:07
	← csc` left	04:10
Toad	anyone know how one comes by a beta password for github?	04:13
	← aroben left	04:20
	← jerbear left	04:20
Toad	anyone here use repo.or.cz?	04:22
mugwump	plenty	04:22
	← glommer left	04:22
	← kylem_ left	04:23
	← lamont left	04:24
imyousuf	Toad: is there anything in particular you want to know on repo.or.cz?	04:24
Toad	imyousuf: I just registered an account, but it doesn't seem to be working properly with my public key	04:25
	it's prompting for a password	04:25
	→ kevwil joined	04:25
imyousuf	I see, I faced the same problem when I did not copy my key correctly	04:25
Toad	hm, it was probably hard wrapped	04:26
imyousuf	but later when I copied the again (the whole file content) it worked ok	04:26
Ilari	Toad: Did you register your public key? Did you copy the keytype ID (ssh2-...) too (somebody didn't)?	04:26
imyousuf	yeah thats what I missed first time :)	04:27
Toad	Ilari: yep, just copied my entire id_dsa.pub	04:27
	just tried again and made sure there were no hard wraps, same deal	04:27
	now I've got todd, tlipcon, and tlipcon2	04:27
imyousuf	Toad: you will be prompted for the password of the key if you have not done ssh add	04:27
Toad	imyousuf: nope, it's in my ssh agent	04:27
	1024 fa:36:bf:e1:77:37:7b:2f:bd:17:b5:93:25:fb:90:88 /home/todd/.ssh/id_dsa (DSA)	04:27
imyousuf	I am using ubuntu and I am asked for the password if I have not done ssh-add	04:29
Ilari	Toad: Did you check that it tries to use correct user ID? Somebody used incorrect one...	04:29
Toad	oh, I think it did actually wrap it... stupid xterm	04:29
	→ mediogre joined	04:29
Toad	Aha, yep, that was it. wonder if I can email them and have them merge my now 4 accounts	04:30
	← antares left	04:33
	← careo left	04:33
	← devogon left	04:34
Toad	hrm, now I'm getting errors with permission denied trying to create temporary file	04:34
	not sure if that's on my end or theirs	04:34
	Writing 4014 objects...	04:35
	fatal: Unable to create temporary file: Permission denied	04:35
	error: pack-objects died with strange error	04:35
	oh, wonder if I didn't give access to this project to my last user	04:35
	→ lamont joined	04:35
imyousuf	and also add yourself to the project :)	04:35
	← Pupeno left	04:36
	Ilari wonders why there isn't even simple regex check on repo.or.cz to reject obiviously bogus public keys...	04:36
	→ doener joined	04:37
	→ Pupeno joined	04:39
imyousuf	Toad: also when pushing a new repo for the first time use --all	04:44
Toad	imyousuf: ah, too late. I just used master:refs/heads/master	04:44
imyousuf	:)	04:44
	as long as it works :)	04:44
Toad	appears to	04:45
	just needed an easy way to share a git repo with people outside the companyh	04:45
Ilari	Something like: 'ssh2-(dss\|rsa) [a-zA-Z0-9\+/=]+( .*)?'	04:45
imyousuf	try cloning it to another location to verify	04:45
Toad	gitweb seems to be showing it nicely. I'll try a clone to make sure	04:45
	→ FunkeeMonk joined	04:45
Ilari	Toad: Just 'master' on push should work...	04:46
imyousuf	Ilari: adding that regex would come in extremely handy for newcomers like myself	04:46
Ilari	imyousuf: I don't know if it actually works but...	04:46
Toad	yep, cloned just fine	04:46
Ilari	Nope, it does't.	04:47
	Sightly fixed version: 'ssh-(dss\|rsa) [a-zA-Z0-9\+/=]+( .*)?'. That at least accepts my public key.	04:47
imyousuf	ah yup	04:48
	Toad: if you have any other branch other than master that you want to push git push master will not push them, AFAIK	04:48
	so you will need to push them individually	04:48
Toad	imyousuf: yea, I pushed the ones I wanted individually	04:48
	imyousuf: I've been using git for several months now, just hadn't used repo.or.cz yet	04:48
imyousuf	oh ok :)	04:49
Toad	(we use git at work for version control, but it's closed source)	04:49
Ilari	Actual algorithm is as follows: If no refspecs are specified, take set of local branch names, set of remote branch names, do intersection and push those branches in intersection.	04:49
	IIRC, one can specify multiple refspecs on one push command, so 'git push foo bar baz' should push bar and baz to foo.	04:50
imyousuf	yup	04:50
Ilari	imyousuf: Checked if that regex thingy works? It is in the pattern syntax used by egrep.	04:51
Toad	headed home from work now. thanks guys	04:51
	(man, should stop staying til midnight!)	04:51
imyousuf	anytime Toad	04:52
	← mithro left	04:52
	← doener_ left	04:53
imyousuf	Ilari: sed -e 's/ssh-(dss\|rsa) [a-zA-Z0-9\+/=]+( .*)?/true/p' /home/imyousuf/.ssh/id_rsa.pub this did not work	04:53
	imyousuf is trying to figure out why	04:53
Ilari	imyousuf: There's embedded / in the expression.	04:54
doener	sed wants $...$ not (...)	04:54
imyousuf	yes and ( as well	04:54
	→ njero joined	04:55
Ilari	command I used: 'egrep "ssh-(dss\|rsa) [a-zA-Z0-9\+/=]+( .*)?" .ssh/id_dsa.pub'.	04:55
	That BASE64 checking could be better...	04:56
	← lamont left	04:57
doener	Ilari: s/dss/dsa/?	04:57
Ilari	Sightly improved version: 'ssh-(dss\|rsa) [a-zA-Z0-9\+/]+=( .)?'	04:57
	doener: It's DSS in the key I have.	04:57
doener	right, just created a dsa key to check. interesting	04:57
	→ lamont joined	04:57
imyousuf	sed -e 's/ssh-$dss\\|rsa$ [a-zA-Z0-9+\/\=]\+$ .*$\?/true/p' /home/imyousuf/.ssh/id_rsa.pub this worked	04:58
Ilari	IIRC, Digital Singnature Scheme	04:58
	imyousuf: Why does it give true twice?	04:59
imyousuf	s/-e/-ne/	04:59
doener	or just drop the p	04:59
ribas	mugwump: awake!?	04:59
imyousuf	yeah that works too, thanks doener	05:00
mugwump	ribas: sure	05:00
ribas	mugwump: =)	05:00
	→ sgrimm joined	05:01
	→ tvachon joined	05:01
	← tvachon left	05:01
ribas	mugwump: can you read one paragraph for me?! And help make it better? it is about my README (patch) for gitweb	05:01
	← madewokherd left	05:04
	← lamont left	05:04
	→ lamont joined	05:06
ribas	mugwump: here we go, if anyone more is awake may help to =)	05:06
	* owner (or gitweb.owner)	05:06
	File with project (of a repository) owner, single lined. By default set to	05:06
	filesystem's directory owner. You can use gitweb.owner repo configuration	05:06
	variable, but file takes precedence.	05:06
	← lamont left	05:06
	→ mithro joined	05:08
mugwump	the possession should move to the end of the compound noun - ie, filesystem directory's	05:08
	"single lined" sounds awkward, perhaps "a single line"	05:08
	← kevwil left	05:09
mugwump	also, you're missing articles in a couple of places	05:09
	→ srid joined	05:10
	← KirinDave left	05:10
	→ pombreda joined	05:10
mugwump	File with the owner of the repository, on a single line. By default set to hte filesystem directory's owner. You can use the gitweb.owner repo configuration variable, but the file takes precedence.	05:10
	← FunkeeMonk left	05:12
ribas	mugwump: thanks a lot =)	05:15
	→ KirinDave joined	05:15
	→ lamont joined	05:17
ribas	now, going to sleep =) soon i'll be going to Rio de Janeiro =D	05:30
	← srid left	05:32
	→ spearce joined	05:34
	← mithro left	05:35
	← njero left	05:41
spearce	hah. 70 line insertion, 43 line commit message. i'm horrible.	05:43
	how did i ever get by with "cvs -m foo" in the early days?	05:43
	← Ingmar left	05:45
	→ thiago_h1me joined	05:45
	→ Ingmar joined	05:45
	← KirinDave left	05:46
	← thiago_home left	05:46
	← fhobia left	05:46
	← PerlJam left	05:47
	→ PerlJam joined	05:47
	← faheem left	05:53
	→ faheem joined	05:53
	← lamont left	05:54
	→ lamont joined	05:56
	← peper left	05:57
	→ peper joined	05:57
gitster	Yeah, you should strive for a log message that is about twice as long as the patch text ;-)	05:58
spearce	hey, at least it describes wtf i'm doing.	05:58
	now if only the code actually matched the commit message.... sigh	05:58
	stupid f'ing bugs.	05:58
	damn.	06:00
	ok, so i gotta choice.	06:00
	always use a pipe between the hook and receive-pack for stderr, even if sideband wasn't supported on the client, and have receive-pack dump the data to stderr... or try to realize we didn't enable sideband so allow receive-pack's stderr to inherit into the hook, like we use do to.	06:00
	the code looks worse with the latter option. lots of "if (use sideband)" tests.	06:01
	→ reval joined	06:01
kmap	Is there any way to set up a push/pull only git repository on a remote machine which doesn't have git installed (without installing git there of course)	06:03
gitster	I do not recall the details of how I did it on the upload-pack side...	06:03
	No.	06:03
kmap	I guess it'd make sense to install git there...	06:03
	gitster: Right. Thanks	06:03
spearce	gitster: looks like you did the first case, always use the pipe.	06:03
	kmap: you can publish a repository over standard HTTP, but to get the repository there you'd have to use scp or rsync or something manually. its really annoying. its really best if you can just install git.	06:04
kmap	spearce: I have scp access.	06:07
	spearce: But the point is, yes... I better get git there. :-)	06:08
spearce	so git on it. :)	06:08
kmap	spearce: Right! :-)	06:08
	spearce was mr. create garbage today.	06:08
spearce	i kept rebasing the same 68 patches over and over again. generated 6000 loose objects in only a few hours.	06:09
	insane.	06:09
	→ saimon joined	06:09
spearce	at best that project usually only grows about 200 objects/week. if that.	06:09
	so gitster, you said you were perhaps more comfortable with this pgp based authentication in git-daemon than some of the other approaches we've kicked around?	06:11
	← reval left	06:19
gitster	perhaps more comfortable is an overstatement, I should say.	06:20
spearce	less queasy? :)	06:22
gitster	But it surely is much less crazy compared to the other stuff.	06:22
	→ jeremyb joined	06:23
spearce	unfortunately all us smart people haven't come up with too many great solutions.	06:24
	← nkallen left	06:29
	→ nkallen joined	06:30
	→ ken_ joined	06:33
	→ robinr joined	06:33
mugwump	spearce: will you be around in an hour or so to chat about this? I'm about to walk home	06:40
spearce	no, probably not, i'll be heading to bed about then.	06:40
mugwump	ah, ok	06:40
	well... anyway, I was going to mention, that the design comes from trying to distill out all the essential parts of the cpan6 project I was briefly involved in	06:41
spearce	hmm.	06:41
mugwump	so, we discussed this problem space pretty thoroughly	06:42
spearce	30 page cpan6-design.pdf... that you authored. worth reading parts of it? :)	06:43
mugwump	however I really didn't like the other guy's insistence that all the messages be specced in xmlschema	06:43
	the top level stuff, sure	06:43
	hang on, I'll review it	06:43
spearce	i don't want to keep you. i'll skim this then head to bed. i think i need to chew on it a few hours in the back of my mind before i can write anything intelligent again on this subject. :)	06:44
mugwump	cool, sounds about right :)	06:44
	sleep well!	06:44
	← Sho_ left	06:45
mugwump	oh, the results of the review - that design document isn't all irrelevant, but the namespaces stuff (ie, higher level stuff) would just sit in an "index" project, ie the basic system doesn't have to care	06:45
	but enough... I'm off home	06:46
	→ namenlos joined	06:59
	← dysinger left	07:03
	← robinr left	07:07
	Dodj\|away → Dodji	07:14
	→ pmezard joined	07:23
	← EricWinchell left	07:25
	thiago_h1me → thiago_home	07:29
	← Ademan left	07:29
	→ Ademan joined	07:29
hachi	anyone know how to get git-cvsimport to authenticate with the cvs server?	07:30
	I'm doing a cvs login ahead of time, but it's not sending the password to the server when it tries to log in... according to the strace I'm reading	07:30
	← Phoop left	07:30
	→ bentob0x joined	07:34
up_the_irons	testing..	07:39
	← ken_ left	07:43
	← pmezard left	07:46
	← capisce left	07:48
up_the_irons	testing (again)	07:49
	← loops left	07:51
	→ johan-s joined	07:55
	→ KirinDave joined	07:59
	→ langenberg joined	08:15
	← spearce left	08:16
	← pergesu left	08:26
	→ jpetso joined	08:29
jpetso	Tv: ping	08:29
up_the_irons	jpetso: he's probably resting after a long night of team fortress 2	08:30
jpetso	heh	08:30
	up_the_irons: with you participating?	08:30
up_the_irons	jpetso: no, but he mentioned he was getting into it	08:31
jpetso	mmkay	08:31
	up_the_irons: ok, thanks for the info	08:31
up_the_irons	jpetso: i try to stay away from addictive video games, b/c i sure can get addicted...	08:31
jpetso	up_the_irons: nothing better than a state of trance looking at the ever-moving screen	08:32
up_the_irons	jpetso: :)	08:32
jpetso	:P	08:32
	← KirinDave left	08:32
jpetso	Tv: (when you get up.) seems like our release system overlord dww still considers branch and tag restrictions necessary: http://groups.drupal.org/node/8526#comment-26347	08:35
up_the_irons	but who cares about the CVS deploy module, lol	08:37
	→ ferdy joined	08:39
	→ pmezard joined	08:39
	← somnium left	08:41
	→ dsop joined	08:44
	→ Maik joined	08:45
	← Maik left	08:45
	→ Maik joined	08:45
	← Maik left	08:45
	← gebi left	08:52
	→ gebi joined	08:52
	→ FunkeeMonk joined	08:54
	← saimon left	08:57
jpetso	up_the_irons: if Drupal would switch to git, the CVS deploy module would certainly get converted into a Git deploy module, so... yah, it's a valid example	09:04
hlisv561	jpetso: What does the CVS deploy module even do?	09:04
jpetso	hlisv561: i'm not totally sure what it does exactly, but it helps rolling out Drupals using CVS checkouts instead of release tarballs	09:05
up_the_irons	jpetso: oh ok	09:05
	← kanru left	09:07
	→ Sho_ joined	09:12
	← sverrej left	09:12
	→ kzak__ joined	09:23
	→ saimon joined	09:24
	→ robewald\|work joined	09:44
	→ sverrej joined	09:49
	→ pmezard_ joined	09:55
	← pmezard left	09:55
	→ Fullmoon joined	10:01
	→ devogon joined	10:07
	→ octopod joined	10:07
context	jpetso: id rather hope deploying any software wouldnt depend on a scm of any type	10:08
	cvs or git or any other	10:08
	i dont want my production system depending on my scm software	10:08
	← kzak__ left	10:09
jpetso	context: it doesn't, but 1. you can do it, and 2. using the CVS deploy module, while not required, makes things easier	10:09
	→ kzak__ joined	10:10
jpetso	context: if i understand correctly, it just quiets the automatic update notifier and gets the version from the VCS tag instead	10:10
	context: nothing that should prevent you from doing it your way, though	10:11
	← vzctl left	10:12
is_null	hello everybody, how is it possible to rename the "master" branch please ?	10:12
	← jpetso left	10:13
is_null	also : why doesn't git remove from the working directory, the files that were created and added in another branch ? is it a normal behavior ?	10:16
Debolaz2	is_null: Yes, that is normal. It leaves untracked files alone.	10:17
is_null	Debolaz2: thanks	10:21
	← FunkeeMonk left	10:22
	→ ebel joined	10:24
	→ priidu_ joined	10:25
	→ tjafk2 joined	10:26
Tv	is_null: renaming master: git branch -m master foo	10:37
	← imyousuf left	10:39
is_null	Tv: thanks i had figured with the manual	10:41
	(hopefully)	10:41
	← timj left	10:42
	→ rubydiamond joined	10:42
	← namenlos left	10:50
	→ namenlos joined	10:54
is_null	is it possible to checkout and specify how deep should the history be ? For example, i'd like te checkout with only the 500 last commits	10:56
Mikachu	yes	10:57
up_the_irons	is_null: use --depth	10:58
is_null	thanks	10:58
	is_null deserves to be RTM'ed	10:59
up_the_irons	hehe	10:59
Mikachu	my "yes" was meant to imply that :)	10:59
	← johan-s left	11:04
is_null	what does this error mean please ? http://rafb.net/p/D8CFCK66.html i've found a thread about it on internet but it's 40-messages long and relates to cygwin and winxp which is not revelant in my case	11:07
	→ janm joined	11:07
context	is_null: well find the real command out of your ant file	11:09
	so you can get a better mesage possibly	11:10
	← namenlos left	11:10
context	that or its completely related to ant itself and has nothing to do with git	11:10
Tv	is_null: someone tried to cat a git ref file directly? that's broken	11:11
	is_null: they should use git rev-parse	11:11
	even packed refs will make that fail	11:11
context	heh	11:12
	→ kanru joined	11:12
is_null	context: i experience the same problem without ant	11:12
context	is_null: so yeah, you gave us 3 lines of ant output, no /real/ command, no /real/ output, and no /real/ context	11:12
is_null	my bad, let me prepare a description of the problem	11:13
context	the cat: line is compltely understandable, thats standard	11:13
	so 2 lines of output	11:13
Tv	the fault is in whoever uses cat for that purpose	11:13
context	is_null: dont use non git tools to get sha1 hashes from a git repo	11:14
	→ stillLotR joined	11:14
	← fultilt left	11:15
	→ johan-s joined	11:21
	← garrin left	11:21
	→ namenlos joined	11:22
	kmap → should	11:25
	should → kmap	11:25
	← LotR left	11:33
	→ lcapitulino joined	11:34
	→ Phoop joined	11:35
	→ glommer joined	11:35
Ademan	i'm looking at the man page for git-push, you can push things to git across http? O_o is there any sort of authentication that goes on?	11:38
is_null	ok, i got my report done : http://rafb.net/p/FtDzmn19.html any tip or idea please ?	11:40
	← robewald\|work left	11:44
hlisv561	Ademan: Authentication on server side is implemented by HTTP daemon. What is supported in client side is dictated by curl library used.	11:49
Ademan	hlisv561: so in the case of apache would .htaccess dictate that?	11:49
hlisv561	Ademan: The authetication employed is matter of Curl library in client side and Apache on server side. As far as I know, there is no specific authentication support in git itself for HTTP push.	11:50
	→ mithro joined	11:50
hlisv561	Ademan: Probably apache can be configured in this regard using .htaccess...	11:51
Ademan	alright thanks	11:51
	→ cedricv joined	11:52
hlisv561	is_null: Maybe default branch on remote side is set to nonexistent one?	11:52
is_null	hlisv561: indeed, thanks !	11:52
	← kelvie left	11:54
	→ vbgunz joined	11:58
	stillLotR → LotR	12:04
	→ dsaxena joined	12:10
	→ njero joined	12:19
	← mithro left	12:20
	→ mithro joined	12:20
	→ Yuuhi joined	12:24
	→ kelvie joined	12:28
	← Cap_J_L_Picard left	12:33
	→ Cap_J_L_Picard joined	12:34
	→ cmarcelo joined	12:39
octopod	if I have two git-svn clones of the same repo on two different machines, is it possible to set up so I can git push/pull between them before dcomitting from one? Or am I best doing a normal git clone of one of them from the other?	12:44
hlisv561	octopod: I think that using git-svn with repository setup like that is not very advisable due to limited branch model of SVN. And one has to be careful doing pushes into repositories with working trees due to push-to-HEAD problems.	12:51
	← ironfroggy left	12:52
	→ FunkeeMonk joined	12:52
hlisv561	octopod: And neither is doing git clone of one from other necressarily a good idea for repo that has branches to dcommit to SVN.	12:54
octopod	hlisv561: ok, thanks... basically what I want to do is be able to move local changes to a development server to let others preview them before they're committed to svn	12:55
hlisv561	octopod: DVCS (such as Git) easily can take development model that is hierarchial with peering. CVCS (such as SVN) pretty much assumes two-layer model (central repository and its users).	12:56
	→ destruct_ joined	12:56
	→ tiagosab joined	12:56
tiagosab	Hi! How do I find out in which branches a commit is?	12:57
hlisv561	octopod: If you just want to offer previews, make copy of the preview branch (rebased on top of latest from SVN) somewhere where it is accessable. But make clear to other users that that branch is not stable.	12:57
octopod	hlisv561: no one would be using git on the preview branch apart from me, I just really want to be able to push a local branch to the remote server, I'm not sure what you mean by "make a copy"	12:59
hlisv561	octopod: Pushing a branch makes copy of a branch since branches are local.	13:02
namenlos	hi, is there a possibility to interact with mercurial users using git?	13:03
	→ Leonidas joined	13:04
	← destruct left	13:07
	destruct_ → destruct	13:08
	← glommer left	13:08
	moh_ → moh	13:09
	→ Lash\| joined	13:12
	→ robfitz_ joined	13:17
	← Cerebon left	13:22
maks_	namenlos: hg-to-git	13:26
	← barthus left	13:27
	← robfitz left	13:28
	robfitz_ → robfitz	13:31
	→ loops joined	13:31
	→ careo joined	13:33
namenlos	maks_: thx. so i am able to convert from hg to git. but am i right if i say, that there is no such thing like git-svn?	13:34
	← johan-s left	13:36
	← Ademan left	13:43
	→ Ademan joined	13:44
	→ ironfroggy joined	13:46
	← ironfroggy left	13:46
	→ ironfroggy joined	13:47
	→ jj-_ joined	13:48
	← jj- left	13:48
	→ kukks joined	13:51
	→ TuxCM- joined	13:51
	→ Ryback_ joined	13:53
	→ bdiego joined	13:57
	→ EmilMedv1 joined	13:58
	EmilMedv1 → emilmedv1	13:59
	emilmedv1 → EmilMedv1	13:59
	← EmilMedv1 left	13:59
	← EmilMedve left	14:02
	→ EmilMedve joined	14:03
	→ Cerebon_ joined	14:05
	← Tux__ left	14:07
	→ unreal joined	14:07
	← kukks left	14:11
	→ kukks joined	14:12
is_null	what command to set the default branch please ?	14:15
	→ johan-s joined	14:16
noodl	hi. can anyone tell me what this error means, please? (the local machine is mac os 10.5 running 1.5.3.7 and the remote is linux running 1.5.3.8) http://dpaste.com/33136/	14:17
	Cerebon_ → Cerebon	14:17
	← ironfroggy left	14:20
	→ ironfroggy joined	14:20
hlisv561	is_null: If it is repository without working tree, I think one has to either use 'git symbolic-ref' or hand-edit the repository. The symbolic reference to default branch is called HEAD.	14:22
is_null	hlisv561: i have a working tree but i'm unsure if git is aware of it's name	14:27
	HEAD has "ref: refs/heads/merged", which is my working tree	14:29
hlisv561	is_null: I was talking about the repository on the server...	14:30
loops	noodl, perhaps Git isn't set up properly on web3.. or not in path ?	14:30
is_null	hlisv561: thanks, i'm quiet confused since the repos i work on is on the server	14:31
	*as woll	14:32
noodl	hmm, it's in /uar/local/bin but i'll make sure that's in PATH for a non-interactive shell	14:32
	← kukks left	14:33
	→ kukks joined	14:34
	jj-_ → jj-	14:34
loops	is_null, i'm joining this conversation late; can you restate the issue you're having easily?	14:34
is_null	it was caused by a misunderstanding of mine and lack of reading the manual, causing two repos to cohabit on the server	14:36
loops	okay.. Git shouldn't have a problem with 2 repos on a server.. have you sorted it out for yourself?	14:38
noodl	loops: good call. it's not an issue with $PATH but the remote git installation seems incomplete (has roughly half the git-* executables as my local one)	14:40
is_null	i did sort the http hosting, thanks hlisv561. But i just figured that it's not possible to clone with --depth via HTTP; though that's something i want to be part for public deployment - which hosting solution allow it please ?	14:41
loops	noodl, it may have been installed with the gitexecdir option, moving some of the lower level plumbing commands out of standard PATH.	14:42
	is_null, git-daemon.	14:42
noodl	loops: make install complains about: gtar: This does not look like a tar archive	14:43
	← priidu_ left	14:44
loops	noodl, sorry that's a new one on me..	14:45
noodl	lemme try the build from scratch	14:46
	failed again. here's the complete configure/make/make install output in case anyone has seen this, http://dpaste.com/33141/	14:53
loops	noodl, do you have a copy of gnu tar installed on that machine instead of gtar ?	14:54
	← johan-s left	14:54
noodl	tar (GNU tar) 1.14 << that's the only one. (gtar links to it)	14:55
loops	noodl, looking at the Makefile, it seems gtar should only be invoked on SunOS.. odd	14:55
	← IRSeekBot left	14:56
	→ johan-s joined	14:57
loops	noodl, does the command gtar cf - . \| gtar tf -	14:57
	typed at the command line work for you?	14:57
	→ cannonball joined	14:58
noodl	yep	14:58
	→ mjc joined	14:58
loops	hmmm	14:58
	→ EmilMedv1 joined	14:58
noodl	shows the list of the directories from tar stream on stdin, and works in the templates/blt directory. same works with xf -	14:58
	EmilMedv1 → emilmedv1	14:58
	← EmilMedve left	14:59
	← emilmedv1 left	14:59
	→ EmilMedve joined	15:00
is_null	i'm failing to fetch to repo and i don't understand why, this is the log from git-deamon : http://nopaste.com/p/a4nyGqi5qb	15:01
johannes	touch /var/www/devangels/git/.git/git-daemon-export-ok might help	15:02
loops	is_null, take a look at the git-daemon man page.. you need to either use --export-all, or use a flag file in each repo to mark it for export via git-deamon	15:02
noodl	loops: gah, it fails because i have CDPATH set. not sure why, with '.' in the path, but anyway it's installed now. sorry :(	15:03
is_null	ah, i've read it but i didn't understand, thanks	15:03
	→ stick joined	15:03
loops	noodl, okay.. (btw you should never "export" CDPATH, it should only ever affect the shell, nothing executed by it)	15:03
	cbrake_away → cbrake	15:04
	← lcapitulino left	15:04
noodl	loops: ah! my bad. and woohoo, i finally have the original clone command working. thanks a bunch for your time :)	15:04
loops	np.. glad you got it working.	15:04
	→ dysinger joined	15:05
	→ lcapitulino joined	15:07
	→ GrimRC joined	15:10
	→ pedro__ joined	15:10
GrimRC	hi all; I have kernel 2.6.24, and I've used git-init to turn it in to a repository; I've done make distclean, but there are still files that are (git)ignored; how do I list the untracked files?	15:11
EmilMedve	Did you try 'git ls-files -i'?	15:13
GrimRC	no, but I get this: ls-files: --ignored needs some exclude pattern	15:14
	I'm new to git; I must be using that command wrongly	15:16
EmilMedve	Is adding --exclude-standard helping?	15:16
	If all you want to do is to remove _all_ the untracked files, try 'git clean -d -x -f`...	15:16
GrimRC	it just dumps usage info with -i and --exclude-standard	15:17
	I want to unignore all the ignored files, so that I can merge wireless-2.6 without fatal errors	15:17
	I mean, I want to unignore the ignored files, because there shouldn't be any left after make distclean, if I understand it right; from Googling, it seems sometimes vmlinux* files get accidentally ignored	15:18
loops	GrimRC, if this is a brand new repo that you created, nothing should be ignored by Git.	15:18
GrimRC	Linux kernel comes with its own .gitignore files	15:19
loops	ahh.. so can you just edit those to get the results you want?	15:19
GrimRC	yeah, but I don't know how to get a list of untracked files so I can fix them all in one go	15:19
EmilMedve	At this point I'm not sure I understand what exactly do you want to do...	15:20
	→ zapnap joined	15:20
EmilMedve	git clean -x -d -f should clean up all the untracked files...	15:21
	← zapnap left	15:21
Tv	GrimRC: man git-ls-files	15:22
	← kanru left	15:22
loops	git ls-files is pretty awful when it comes to just asking for something simple like... give me a list of all ignored files..	15:23
GrimRC	I can't work out the command-line to do it	15:24
loops	GrimRC, git ls-files -o seems to do it	15:26
	← kukks left	15:26
Tv	loops: how is that pretty awful?	15:26
GrimRC	what I'm ultimately doing is, turning my 2.6.24 kernel in to a repository, adding a branch called wifi and the remote wireless 2.6 repository, doing a git-fetch of the remote repository, then a git-merge so I have a modified 2.6.24 kernel - is this even right?	15:26
loops	but the documentation and UI sucks bad	15:26
	→ kukks joined	15:27
loops	Tv: to get ignored files you'd expect to be able to say "git ls-files --ignored"	15:27
Tv	yeah, and the docs lie about it needing an argument too	15:27
	but still, that's an extra whooping 10 seconds time spent	15:27
loops	tv: and when you read the documentation.. you ask yourself what the --other option does and it says list "other" files.. not very helpful.	15:27
Tv	anything not covered above ;)	15:27
GrimRC	loops: thanks for that - it works	15:27
loops	Tv, sure.. it's not the end of the world.. just a blemish	15:28
GrimRC	doing git-fetch from a remote repository only fetches all that meta-data stuff right? so that when I do git-merge it only downloads what's needed? right??	15:28
	← pedro__ left	15:29
Arjen	No	15:29
loops	No..	15:29
GrimRC	is there any sensible way to do that?	15:29
Arjen	git fetch gets all that the local repository doesn't have yet, compared to the remote	15:29
GrimRC	so I have master which is 2.6.24 - I should have made a new branch called wireless-2.6 (the same name as the remote repository), then git-pulled?	15:30
	I mean merge local master in to local wireless-2.6, then git-pull the real wireless-2.6?	15:31
Tv	GrimRC: more like git fetch && git checkout -b wireless-2.6 remotes/origin/wireless-2.6	15:31
	i guess	15:31
	you're not being very precise, so I can't be very precise ;)	15:31
	← rubydiamond left	15:31
GrimRC	so git-fetch downloads objects, meaning it downloads all the data?	15:32
Arjen	Yes	15:32
GrimRC	how do I do a diff/rsync across the internet, of my kernel and theirs?	15:32
	→ chris2 joined	15:33
Tv	GrimRC: you really don't	15:33
	GrimRC: you fetch, then do everything locally	15:33
GrimRC	but that'd mean I download everything from them, making my copy of the Linux kernel useless?	15:34
Tv	shared bits are reused	15:34
GrimRC	maybe the shared bits were reused, I can't tell	15:34
	so I don't need to merge my kernel and theirs, I just check theirs out now	15:35
	I mean, I fix these broken ignores and then do that	15:36
	← Lash\| left	15:36
Tv	GrimRC: you sound quite confused about git fundamentals -- don't know if you have the right background for this, but can't hurt to try: http://eagain.net/articles/git-for-computer-scientists/	15:36
	→ aroben joined	15:36
	Dodji → Dodji\|away	15:39
GrimRC	yeah so this stuff about hashing with SHA-1 means that I can pull from other people's repository in a sane manner, as long as I have my Linux kernel included with git-init	15:39
Tv	fetching a subsystem-specific tree is internally no different from fetching a new version from linus' tree	15:41
	→ kumbayo joined	15:41
GrimRC	so that should mean it's downloaded efficiently as possible (like rsync)?	15:41
Tv	better than rsync most of the time ;)	15:42
	→ bobesponja joined	15:45
GrimRC	du -ch !(.git) = 308MB and du -ch .git = 446MB --- I have Linux 2.6.24 in my working directory (and in master branch) and I have wireless-2.6 git repository git-fetched too	15:46
	→ pedro__ joined	15:47
GrimRC	.git/objects = 443MB - seems like I've done something wrong, coz wireless-2.6 is basically modified kernel	15:48
arw_	hm? i don't understand, why do you think there is something wrong?	15:50
GrimRC	when I commited my copy of the kernel to master, it seemed to take up much less space	15:51
	are objects compressed?	15:51
semi_	hey, does git work with jira? (jira issue tracker http://www.atlassian.com/software/jira/ )	15:51
	svn apparently does	15:52
	but couldn't find any info from the web	15:52
arw_	GrimRC: yes and no.	15:52
	GrimRC: running 'git gc' compresses the uncompressed ones	15:53
	← EmilMedve left	15:55
	← johan-s left	15:55
	→ EmilMedve joined	16:05
	→ kanru joined	16:16
	← kumbayo left	16:19
	← kukks left	16:22
	→ kukks joined	16:23
	← namenlos left	16:24
	← dsaxena left	16:24
	← langenberg left	16:28
	← GrimRC left	16:28
	→ namenlos joined	16:34
	tjafk2 → timj	16:34
	← namenlos left	16:38
	→ agorf joined	16:39
agorf	hello. anyone knows the git equivalent of svn copy?	16:39
vmiklos	just copy and git add	16:40
agorf	won't this prevent git from knowing where the copy came from?	16:40
Debolaz2	agorf: git detects this automagically.	16:41
agorf	Debolaz2: oh. ok, thanks	16:42
arw_	no. git detects the copy by the files content, not any special data that something like 'svn copy' records.	16:42
Debolaz2	agorf: The reason why this works is that git sees the file, as arw_ pinted out, has the same content, and hence is related to the original file.	16:42
agorf	Debolaz2: so i need to commit the copy before doing any changes to id	16:43
	it*	16:43
	Debolaz2: in subversion, you can svn copy foo bar and then make changes to bar, and subversion remembers that bar came from foo.	16:44
vmiklos	git log --follow does the same	16:44
	you can do the same here	16:45
jrockway	agorf: no, you can change the copy on the same commit where you make the copy	16:45
vmiklos	then git will say "foo.c is copied from bar.c: they are 85% similar"	16:45
jrockway	yup	16:45
	→ chrizz joined	16:46
arw_	unless of course the similarities are so small that the copy will become undetectable, but then the copy was useless anyways :)	16:46
agorf	what i basically am doing is moving a big block of code from foo to bar	16:47
vmiklos	just do it, then see that git annotate -C will recognize it's copied from foo	16:47
	→ lol_h joined	16:49
	→ KirinDave joined	16:50
agorf	vmiklos: hm, it doesnt	16:52
vmiklos	you can try --find-copies-harder :)	16:54
	→ IRSeekBot joined	16:54
	→ HG` joined	16:54
	← cehteh left	16:55
	← djpig left	16:55
	← bentob0x left	16:59
	→ alley_cat joined	16:59
	→ johan-s joined	17:02
agorf	vmiklos: http://rafb.net/p/DJihCg62.html	17:03
	here's a short example that doesnt work for me	17:03
drizzd_	agorf: By default there need to be at least 40 or so matching characters. You can change the default by supplying it to -C, e.g. git annotate -C -C10 copied_file	17:04
	← mediogre left	17:05
	→ L33trDaNu joined	17:06
L33trDaNu	is there a way to see what files were affected with git log ?	17:06
agorf	drizzd_: hm, i have no luck with that. maybe im doing something wrong	17:06
	← blix_ left	17:06
	← z3ro_ left	17:07
drizzd_	L33trDaNu: git log --name-status	17:07
L33trDaNu	drizzd_: thanks	17:08
drizzd_	agorf: In your case the similarities are so small, you'd probably have to use -C -C3 or something	17:08
	← Phoop left	17:09
	→ reval joined	17:09
agorf	drizzd_: yeah, i tried with C ranging from 1 to 10	17:09
Tv	I don't think -C means the number of lines	17:11
	the number afterwards is a more abstract "score" from in tens of percents	17:12
	the 40 char limit is harcoded IIRC	17:12
	make a less minimal minimal failing example ;)	17:12
tiagosab	I am using git to track an upstream cvs, but I just found that the stable branch has duplicated commits (only in git, not in cvs). I checked it out again with git-cvsimport and I found the same result. Any hints?	17:13
agorf	Tv: in my normal example, i moved 66 lines from foo to bar, and it still doesnt work	17:13
Tv	agorf: please just reiterate that "short example" so it has more data	17:13
	← tbecker left	17:14
vmiklos	i just checked, i moved a functom from foo.c to bar.c, i modified the function in May 2007. after copying the function, git annotate -C still says May 2007, and not Jan 2008, so the copy detection works properly here	17:14
agorf	Tv: yes, what im saying is that in my normal case, i did the exact same thing as with this small example, with the only difference that i had such enough data	17:14
vmiklos	function*	17:14
Tv	agorf: sure but please redo the short example with enough data	17:15
vmiklos	(the function is 141 lines long)	17:15
	→ desegnis joined	17:15
	→ z3ro joined	17:15
Tv	agorf: what happens to me usually at that point is i see the short example work, and realize what the difference is between it and the real case, and understand what happened and why	17:15
agorf	vmiklos: what is the output i should expect from git annotate? will it contain the name of the source file?	17:16
	Tv: i see your point	17:16
vmiklos	without -C git annotate will say you modified the function today	17:16
	← sverrej left	17:17
agorf	vmiklos: ok my bad. it works	17:17
vmiklos	with -C, it will realize today you just copied it and it will show the date of the last _real_ modification	17:17
agorf	vmiklos: i was expecting to see a source filename of shorts	17:17
	thank you both for your help	17:17
vmiklos	the commit name is shown, you can see the commit using git show <hash>, and then you'll probably see the source of the function as well	17:18
agorf	vmiklos: ack. thanks :)	17:19
vmiklos	np :)	17:19
drizzd_	agorf: git blame will show you the filename	17:19
	→ ReinH joined	17:19
agorf	drizzd_: indeed	17:19
	← reval left	17:20
	← FunkeeMonk left	17:23
	← alley_cat left	17:33
	→ lukass joined	17:34
	→ sunblush joined	17:35
	← nkallen left	17:37
	→ alley_cat joined	17:38
	← KirinDave left	17:38
	→ bogner joined	17:40
bogner	is it possible to write a hook to change my .gitignore file before it's read by git? i want to ignore files based on the existence of other files (ignore foo.ps and foo.dvi if foo.tex exists, for example)	17:43
	i can write a script to find out which files to ignore and update the .gitignore, but i don't know how to get that script to be called at the right times	17:44
	← sgrimm left	17:47
	→ sgrimm joined	17:47
jrockway	bogner: why not ignore .ps and .dvi? do you ever want those checked in?	17:47
	← ironfroggy left	17:47
bogner	well, i can ignore dvi, but i happen to have a few ps files i want in the repository (they never change, but i like them to be cloned with the repo)	17:48
	right now i just have .gitignores in the directories i have tex files in, which works well enough, but i'm curious about whether something like this is possible	17:50
jrockway	i don't think a hook exists right now... but it would be trivial to implement the functionality you desire	17:50
	everything is shell/perl scripts. just hack something in :)	17:50
	s/everything/nearly everything/ ;)	17:50
	← Catfish left	17:52
bogner	yeah. i suppose git-ls-files would be the place to hack it in, then?	17:52
	← harinath left	17:52
	meyering → jim-errands	17:55
	→ harinath joined	17:55
	← alley_cat left	17:56
	← ebel left	17:57
	→ arwed joined	17:58
	← bogner left	17:59
	→ bogner joined	17:59
	→ reval joined	18:01
	→ langenberg joined	18:01
drizzd_	bogner: you'll probably be better off using exceptions. And you _can_ add ignored files I think.	18:02
	← pmezard_ left	18:03
bogner	drizzd_: add ignored files in a hook, you mean?	18:03
Tv	bogner: forget the hook part ;)	18:04
	→ KirinDave joined	18:05
Tv	bogner: and you really probably just want to put exceptions in .gitignore	18:05
	→ KirinDav joined	18:05
	→ alley_cat__ joined	18:06
	→ drizzd joined	18:08
	← kukks left	18:12
	→ kukks joined	18:13
	→ cehteh joined	18:13
	→ tvachon joined	18:14
	→ djpig joined	18:14
	→ docelic joined	18:16
	← langenberg left	18:16
	→ langenberg_ joined	18:17
	→ ironfroggy joined	18:17
	← arwed left	18:18
	→ xpgs joined	18:18
	← xpgs left	18:19
	← alley_cat__ left	18:19
	→ pgssgp joined	18:19
	← KirinDave left	18:21
	KirinDav → KirinDave	18:21
	← drizzd_ left	18:22
	← cedricv left	18:22
	← janm left	18:24
	← Leonidas left	18:25
	jwb → jwb_gone	18:28
langenberg_	Odd thing just happened, I did a: git svn rebase; git checkout -b new_feature; commit, commit, commit; git checkout master; git merge master..new_feature; git svn dcommit. And it anded up as a single SVN commit.	18:37
	ended*	18:37
	It shouldn't be doing that right?	18:37
	Erh, and just a 'git merge new_feature'	18:38
	→ Sonderblade joined	18:39
	→ alley_cat joined	18:39
doener	langenberg_: that's how a real merge is supposed to end up	18:41
langenberg_	I just tried it again, now I get a serie of small commits.	18:42
	Could it be because one of my commits was a git revert ?	18:42
doener	probably because you didn't get a merge commit, but a fast-forward	18:42
	→ kevwil joined	18:43
langenberg_	When does that happen?	18:45
	← bogner left	18:45
doener	when there's no new commit in master since you branched	18:46
langenberg_	Ah, then it squashes all the changes from the branch?	18:46
	→ timocratic joined	18:47
doener	no, then a fast-forward happens	18:47
	say you do: git checkout -b foo; edit; git commit -m foo; git checkout master; git merge foo	18:47
	→ yeowman joined	18:47
doener	then you have: X (master) -> Y (foo)	18:48
	→ mithraic joined	18:48
doener	obviously, merging the changes from foo can be achieved by just moving master forward	18:49
	but if you have this:	18:49
	X --> Y (master)	18:49
	\	18:49
	-> Z (foo)	18:49
mithraic	I seem to have wedged myself pretty good, and could use some help un-wedging.	18:49
	Upon git push, I get this error:	18:49
	→ aazar joined	18:50
mithraic	error: remote 'refs/heads/master' is not a strict subset of local ref 'refs/heads/master'. maybe you are not up-to-date and need to pull first?	18:50
doener	then the merge must creates a new commit M, that joins the two lines of history	18:50
	s/creates/create/	18:50
	M has two parents, Y and Z	18:50
mithraic	This is on a local branch that points to master on the remote	18:51
	← pombreda left	18:51
	→ Zariel joined	18:52
doener	SVN cannot represent such a history. So what git-svn does, is ignoring Z and just creating a new revision with the changes from Y to M	18:52
Zariel	right im trying to setup GIT on a webdav share, but i cant access the webdav in winxp =/	18:52
doener	so the changes the happened in the merged-in branches are squashed in _SVN_, git still retains the correct history	18:53
	← L33trDaNu left	18:53
	→ L33trDaNu joined	18:53
doener	langenberg_: in general, unless you really _want_ such a squashed merge (e.g. when the branch also exists in SVN), you should rebase instead of merging	18:54
langenberg_	doener: Ah, I really don't want a squashed merge. All branches are locally.	18:54
	I should rebase with the branch?	18:54
	→ pmezard joined	18:55
doener	langenberg_: yeah, just rebase, no need to merge at all. Just "git svn rebase" the branch and then dcommit it directly. If you then "git svn rebase" your master, it will get the commits from svn and you can drop your branch	18:56
	langenberg_: or you can do "git rebase master my_branch; git checkout master; git merge my_branch"	18:56
	jwb_gone → jwb	18:57
doener	that way, you're guaranteed to get a fast-forward and can just dcommit from master	18:57
	which way you prefer is up to you. The latter is probably less confusing from the git point of view	18:57
	→ L33trDaN1 joined	18:58
	→ Ademan_ joined	19:01
	← docelic left	19:02
langenberg_	Okay	19:03
hachi	anyone know how to get git-cvsimport to authenticate with the cvs server?	19:03
	← Man-erg left	19:03
langenberg_	doener: Does that mean that all small commits I make during my branch, get also to be small commits in SVN? (That's what I prefer)	19:03
hachi	I'm stracing it and the comm with the server shows it not sending the password	19:04
doener	langenberg_: yep	19:04
	langenberg_: rebasing creates a linear history without any merge commits	19:04
	→ arwed joined	19:04
langenberg_	doener: What about: git svn rebase; git checkout -b feature; modify; commit; git checkout master; git svn rebase; git rebase feature; git svn dcommit?	19:05
doener	langenberg_: nope, that will rebase master with "feature" as upstream	19:06
	← alley_cat left	19:06
doener	you can do "git rebase master feature" instead, but that basically ends up the same as my first suggestion	19:06
langenberg_	Ah, okay :)	19:07
doener	with master still requiring a "git svn rebase" to get the changes you dcommitted from feature	19:07
langenberg_	I'll try that the next time, thanks.\	19:07
mithraic	Any suggestions for my little problem?	19:08
doener	langenberg_: to see how git-svn works, try looking at your master in gitk	19:08
	← Ademan left	19:08
langenberg_	doener: Well what I don't get about git-svn in gitk, is that I'm not seeing my local branches, why is that?	19:09
doener	langenberg_: were they merged yet?	19:10
	langenberg_: gitk by default only shows the history of the current branch	19:10
	try: gitk --all	19:10
	langenberg_: if you look at your master's history in gitk, you should notice one branch that branched from master and then joins back into master	19:11
	that join point is the merge commit that you see squashed in svn	19:11
	git-svn only follows the primary line of history, creating SVN revision that are equal to the commits in that line of history.	19:13
	← L33trDaN1 left	19:13
doener	and that branch was outside the primary line of history, thus SVN only got two revisions: before the merge and after the merge	19:14
	← L33trDaNu left	19:14
	← kevwil left	19:16
	← sgrimm left	19:17
langenberg_	aah	19:21
	lemme check	19:21
	→ p4tux joined	19:26
	← tiagosab left	19:29
	→ Leonidas joined	19:33
	← reval left	19:33
mugwump	you don't have to restart gitk, you can go "view...edit" and put "--all" in there	19:40
	→ capisce joined	19:43
	→ pombreda joined	19:49
	← lol_h left	19:55
	→ alley_cat__ joined	19:58
	← kukks left	19:59
	→ kukks joined	19:59
	→ mithraic_prime joined	19:59
semi_	can I sort all blobs in a repo by size?	20:01
	← arwed left	20:01
semi_	and if I were to remove some accidentally versioned unnecessary binary files from the repo after git-svn clone, would git-svn still be able to talk to the svn repo?	20:02
	I mean, doing git-filter-branch to remove something from history	20:03
	→ antares joined	20:03
	→ sgrimm joined	20:09
kergoth	whats the easiest way to copy a commit & its associated tree from one git repo to another, that dont share history? copy it over in the filesystem and cherry-pick it?	20:10
	→ peso joined	20:11
kergoth	wait, nm, alternates worked easier	20:13
vmiklos	you can fetch the tree object	20:13
	and then create a commit that points to the tree but has a parent from the other repo	20:14
	← mithraic_prime left	20:14
kergoth	fetch in what way?	20:15
	← mithraic left	20:15
context	kergoth: you could prolly use git-show and git-apply	20:18
	← capisce left	20:20
	← alley_cat__ left	20:30
	→ glommer joined	20:35
	→ shep-work joined	20:36
shep-work	how can i find what tags apply to a specific hash?	20:37
	more specifically, the hash for the version that I am working in	20:37
	→ alley_cat joined	20:38
shep-work	i see git-show-ref	20:43
johan-s	is there a hook that can be run when people clone a repos?	20:43
shep-work	which seems really close	20:43
	← saimon left	20:43
thiago_home	shep-work: git-describe	20:45
	→ saimon joined	20:45
shep-work	thiago_home: thanks! looking at the man page, i see that if the HEAD^1 is tagged, then it will add cruft to the end of the tag	20:49
	thiago_home: is there a way to skip that, and only show if a tag points at this version ?	20:49
	← alley_cat left	20:51
	→ madewokherd joined	20:51
semi_	is git-svn supposed to detect svn merges?	20:54
drizzd	shep-work: git show-ref -d <tagname> \| grep `git rev-parse HEAD`	20:54
	If <tagname> points at HEAD, this will succeed.	20:55
	→ mikearthur joined	20:55
mikearthur	hey	20:57
	http://git.fredemmott.co.uk/?p=yanihp;a=summary	20:57
	http://mikearthur.co.uk/gitweb/?p=mikepete.git;a=summary	20:58
	the prior has a URL field whereas mine (the latter) doesn't	20:58
	how do I add that?	20:58
shep-work	drizzd: thanks. that is close to what I want	20:59
	drizzd: what I really want to do is tag the code with a version, then use this bit of code to see if i am building a tagged version	20:59
	drizzd: so i dont already know the version, as you suggest	21:00
	also, where can i file bug reports for git documentation?	21:02
	and feature requests	21:02
drizzd	shep-work: then you can just do git-describe --abbrev=0	21:03
shep-work	drizzd: but, say I tag now, then change a file	21:04
	drizzd: git-describe would report that the current state is tagged	21:04
thiago_home	semi_: no	21:04
semi_	how about tags?	21:04
	→ kevwil joined	21:05
	← lcapitulino left	21:05
semi_	it asks for the tags dir but still makes a tags/foo branch instead of a tag	21:05
	with --stdlayout	21:05
	s/asks for the tags dir but still//	21:05
	→ sNov joined	21:06
	→ langenberg joined	21:07
	← kevwil left	21:08
thiago_home	semi_: tags in Subversion are actually branches. For all purposes, that is.	21:09
	why git-svn does it like that, I don't know.	21:09
semi_	thiago_home: yes, I'm aware that svn does not actually have real branches or tags, just copies	21:10
	but still git-svn has options for setting tags dir	21:10
	so it might do something special to them	21:10
vmiklos	so subversion tags are branches. why git-svn would create a tag for a branch?	21:10
thiago_home	in git, a tag is something that appears in the middle of your branch	21:11
semi_	I'm just asking why does it take tags dir as option	21:11
	if it matters squat	21:11
thiago_home	so your current tip is a child of that tag	21:11
	in Subversion, a tag is a tip of a branch.	21:11
	even if it tagged, very seldom would your current HEAD be a child of that tag	21:11
	→ nzkoz joined	21:12
drizzd	semi_: You may want to have a look at http://repo.or.cz/w/git.git?a=blob_plain;f=GIT-VERSION-GEN	21:13
	← duncanm left	21:13
semi_	drizzd: looking, but I am not sure which discussion you refer to	21:15
	← langenberg_ left	21:15
drizzd	semi_: It checks if the repo is dirty, for example. You were asking about that.	21:15
doener	semi_: it does matter. Without the tags dir, it wouldn't be able to find the tags	21:15
	semi_: and to avoid name clashes, it prefixes the remote branches for the svn tags with tags/	21:16
drizzd	semi_: I guess I lost you somewhere	21:16
	← pombreda left	21:17
semi_	drizzd: I'm not sure if I was asking that. Am I just having a brain deadlock or am I the right person?	21:17
doener	semi_: doing real tags is impossible as soon as someone comes up with the great idea of actually modifying a "tag-branch" in svn	21:17
drizzd	semi_: No, I confused you with shep-work...	21:18
semi_	drizzd: ok np	21:18
	doener: so I figured, but couldn't it do things optimistically with some special command line parameter and if that fails then redo the regular way	21:18
doener	semi_: personally, I'd find that even more confusing, because suddenly tags could disappear	21:19
semi_	oh, true, that could happen in the future after you did the clone	21:20
	didn't realize	21:20
	now I see	21:21
	thanks	21:21
doener	and finding a matching commit in some of the real branches - if such a thing exists at all - is probably way too costly	21:21
	→ GrimRC joined	21:21
shep-work	drizzd: heh, i was wondering if some of that would be useful to me ;-)	21:21
doener	and then, I went out of ideas when I originally thought about improving the situation ;-)	21:21
GrimRC	hi again; does anybody know if I can merge just 1 file from another branch in to my working copy?	21:22
	→ dsaxena joined	21:22
	← unreal left	21:22
doener	GrimRC: you can incorporate the changes, but git won't store any metadata about such a "merge"	21:23
GrimRC	doener: oh cool yeah; how do I do that?	21:23
	→ nud joined	21:24
drizzd	shep-work: yeah, sorry about that. it's late... I didn't quite understand what you meant by the current 'state' being tagged. You were talking about the repository being dirty, right?	21:25
doener	easiest is probably: git checkout $dstBranch; git diff $(git merge-base $dstBranch $srcBranch) $srcBranch -- $file \| patch -p1	21:25
	thiago_home has the code to find the original commit	21:26
thiago_home	if the svn tag is a perfect copy of another branch and is never modified	21:26
	this breaks if there are any file modifications or if the copy isn't of the full and entire branch	21:27
doener	GrimRC: another possibility is: git show $(git merge-base $dstBranch $srcBranch):$file > baseFile; git show $srcBranch:$file > newFile; git merge-file $file baseFile newFile; rm baseFile newFile	21:27
	← careo left	21:30
Tv	i'd imagine git show BRANCH FILE \| git apply	21:31
	← dsaxena left	21:32
doener	Tv: s/show/diff/ ?	21:32
Tv	doener: pretty much the same thing	21:33
GrimRC	doener: git-diff HEAD other_branch_here -- path/to/file seems to work	21:33
shep-work	drizzd: been there :-)	21:33
Tv	doener: but not with that exact s//	21:33
	doener: is you show a commit, it shows the diff	21:33
	doener: and is easier to type than the right syntax to git diff	21:33
	doener: as you demonstrated ;)	21:34
shep-work	drizzd: but yeah, say i have --A--o--o-->	21:34
	drizzd: with tag 'A'	21:34
	drizzd: if i am at the HEAD of that branch	21:34
	drizzd: git-describe will say 'A-2-asdasd'	21:34
GrimRC	git-show other_branch:path/to/file > path/to/file	21:35
doener	Tv: but you don't show a commit with the command you gave. Actually "git show HEAD~20 builtin-commit.c" is empty here, while "diff" shows a patch	21:35
Tv	doener: your diff doesn't do what you think	21:35
	doener: and your HEAD~20 doesn't happen to touch that file	21:35
GrimRC	git-diff HEAD other_branch -- path-to-file \| patch -p1	21:35
	doener: need a : between path and commit I think	21:36
Tv	GrimRC: correct, lower level, more to type	21:36
	GrimRC: no	21:36
	BRANCH:FILE means show this file	21:36
	git show BRANCH means show the whole commit	21:36
	git show BRANCH FILE FILE.. means show these files of the commit	21:36
	the change to those files that is	21:37
doener	Tv: ah sorry. Still, as soon as you have more than one commit, or just know a range in which the commit lies, diff is easier	21:37
Tv	doener: yeah if you're not talking about a single commit	21:37
doener	Tv: and in what way does my diff not do what I expect?	21:37
Tv	doener: that form compares working tree to the commit you name	21:38
	doener: not the changes that commit introduced	21:38
	doener: oh and i meant the s/// thing	21:38
	doener: the full merge-base thing you said seems correct (though not very robust)	21:39
doener	Tv: yeah. That was due to the single commit vs. commit range misunderstanding.	21:39
	just didn't want to type the whole merge-base thing again. For HEAD~20, the merge base is obvious and I simply assumed a clean working directory ;-)	21:40
	← mithro left	21:41
doener	oh well, and I obviously would've merge some stuff from the current branch into the current branch... blah :-)	21:41
	s/merge/merged/	21:41
	doener should stop talking now	21:41
GrimRC	I asked the wrong thing; I was just after pulling one file out, not true merge; the git-show and git-diff work for me; surprised it's not simpler	21:44
	← EmilMedve left	21:44
doener	GrimRC: just getting a file from a certain commit is easy: git checkout $sha1 -- $file	21:44
	← peso left	21:44
GrimRC	ohhh!	21:44
	doener: so that long line was to merge properly	21:45
doener	yeah	21:45
	→ EmilMedve joined	21:45
GrimRC	hehe my bad	21:45
	← pgssgp left	21:46
drizzd	shep-work: yeah but git-describe --abbrev=0 will just show you the tag and using git-show-ref -d you can check if it points to the current HEAD	21:47
shep-work	drizzd: ah, thats the key i need	21:49
	drizzd: let me go try that	21:49
	← lukass left	21:50
	← hharrison left	21:50
	← mikearthur left	21:52
	← timocratic left	21:53
	→ sverrej joined	21:53
	← glommer left	21:55
	← GrimRC left	21:56
shenson_not_here	I have a git repository that one of the objects is missing from, how do I go about fixing that?	21:57
	→ unreal joined	21:58
	bdiego → midgar	21:59
	→ timocratic joined	22:01
	← sunblush left	22:02
mugwump	shenson_not_here: procedure in the git user's manual	22:06
	← Sonderblade left	22:06
	shenson_not_here → shenson	22:10
	→ dsaxena joined	22:10
	← EmilMedve left	22:10
nzkoz	mugwump: you going to warkworth this weekend?	22:11
mugwump	nzkoz: yes, starting the drive up tonight	22:12
	I guess I'll pitch camp at the first looking soft spot when it starts to get dark and finish the drive in the morning :)	22:12
halorgium	nzkoz: have you fallen into the git-pit? :)	22:13
	oh, kiwifoo-ing eh?	22:13
	→ lyakh joined	22:18
lyakh	hi, another simple question: how can I see from what branch I branched my current branch?	22:18
	← desegnis left	22:19
lyakh	or just any branch for that matter	22:19
nzkoz	halorgium: yeah, kiwifoo, but git-questions abound	22:19
lyakh	and without gitk preferably	22:19
	← Cerebon left	22:20
nzkoz	mugwump: you should do a session on 'getting to git' or some such, migration for us is harder than it should be, but definitely something we're considering	22:21
	← langenberg left	22:22
	→ majoh joined	22:28
	shenson → shenson_not_here	22:29
	→ rhett joined	22:29
rhett	how do I search git commits that mach a string? gitk is really confusing to me	22:29
	there are like 3 search fields, and none of them work	22:30
	← Leonidas left	22:30
	→ lamont` joined	22:30
loops	rhett, from the command line you can do git log -S'search string'	22:31
mugwump	nzkoz: sounds a very technical topic for kiwifoo :)	22:31
nzkoz	mugwump: getting to beer?	22:32
	then surprise us with git	22:32
loops	rhett, the latest version of gitk has a box you can just fill in to search	22:32
mugwump	My planned talk title was "When You Have a Hammer: applying lessons from source control to management of legislation, or: Git + Law = Westminster++"	22:32
	← octopod left	22:33
nzkoz	ok, I guess I'll have to 'buy' you a beer and talk about our specific needs :)	22:34
mugwump	sure, sounds good ... I can show off my Perforce importer as well, I'm thinking it could be adapted to work for SVN and CVS	22:35
gitster	lyakh: the answer to the question is undefined, if you are talking about the global history. Inside your local repository, you can find it out by "git log -g that-branch" and look for an Reflog entry that says "branch: created from blah".	22:35
nzkoz	people use perforce? ;)	22:36
loops	hi gitster	22:36
mugwump	google do, rampantly	22:36
loops	mugwump, is your p4 importer in contrib? or a different one?	22:36
	→ DrBehavior joined	22:36
	← bobesponja left	22:36
mugwump	loops: it's the one at http://utsl.gen.nz/gitweb/?p=git-p4raw	22:36
gitster	loops: I suspect rhett wanted to learn "git log --grep=string".	22:36
mugwump	It's not quite ready for prime time yet, but will be soon	22:36
lyakh	gitster: thanks, I think, looking for "created from" will do!	22:37
rhett	thanks guys	22:37
gitster	lyakh: The obvious reason that the question does not have any meaningful answer in the global history is this (I am not doing this only to help lyakh, by the way, but for people to look in the channel log).	22:38
	← DrBehavior left	22:38
gitster	"git branch b1 master; git branch b2 master; git checkout -b mine master; work work; git commit".	22:38
rhett	no gitster that just greps the log messages	22:38
	i want to search the source code edited	22:39
gitster	rhett: your question was ambiguous.	22:39
rhett	ok, my bad	22:39
gitster	Then -S or "git log -p \| less".	22:39
loops	mugwump, nice..	22:39
	← lamont left	22:39
gitster	Now, did the branch "mine" come from master, b1, or b2?	22:39
	You cannot tell from the topology you can see from the global history.	22:40
lyakh	hm, I think you can define it to be "master"	22:40
gitster	The best you can do is "git merge-base master mine" to see where 'mine' diverged from 'master'.	22:40
	Think again. master, b1 and b2 all point at the same commit.	22:41
lyakh	I understand, but b1, b2, and mine _start_ there	22:41
	and master brings you there	22:41
Mikachu	there is no such distinction in git	22:42
gitster	That's not recorded in anywhere in the global history.	22:42
Mikachu	a branch only points at the one commit, and that commit in turn has the history	22:42
gitster	Yes. Branch is an illusion without reflog.	22:42
	And it is designed that way.	22:43
mugwump	I like to think of that as branches are observed, not declared	22:43
	← pmezard left	22:44
gitster	If you do this "git checkout -b mine master; git commit; git commit; git checkout master; git merge mine; git checkout mine; git commit ; git commit", then, you _started_ mine from master 4 commits ago (master~2), but there is nothing in the shape of the resulting history that records that fact.	22:44
	The only thing you can tell is that mine has two more commits ahead of master.	22:45
	← aazar left	22:45
	→ aazar joined	22:45
	← agorf left	22:47
lyakh	but one can trace branches? like show log of a branch?	22:48
Mikachu	you can see what you have done with the branch with git-reflog	22:48
lyakh	ok, thanks	22:49
vmiklos	(or git log -g)	22:50
gitster	yeah, the important point to remember is that reflog is a local matter.	22:50
nzkoz	mugwump: interesting stuff in this new p4 thing. for us the challenge is on the other side, how do we maintain a backwards compatible read only mirror in svn	22:51
gitster	In the global history, it does not matter if you created those 4 commits (see example 6 minutes ago) on 'mine' branch and then fast-forward-merged them into 'master', or you just built them directly on 'master', if you show the resulting 'master' to other people to make it global.	22:52
loops	nzkoz, there was a proposed SoC project to make a svn server for git.. (like the git cvsserver)	22:52
	don't know if it was accepted or anything done tho.	22:53
nzkoz	loops: yeah, something like that could be really interesting, but it's non-existent from what mugwump said last time we spoke	22:53
	→ rxKaffee joined	22:53
nzkoz	visa issues or somethign?	22:53
lyakh	git log -g doesn't show "created" for me...	22:53
	ah, ok, it was "Created":-)	22:53
rxKaffee	anyone know of some reviews of repo.or.cz, or comparisons to other similar services?	22:54
	← Fullmoon left	22:55
halorgium	nzkoz: excellent, i'm not invited but i'm flying up to akl tonight, having lunch with a couple of guys who are going :)	22:57
	→ csc` joined	22:58
	→ pombreda joined	22:59
	← Ryback_ left	23:00
	← mjc left	23:01
	→ ilogger2 joined	23:03
	→ mithro joined	23:18
	→ flaguy joined	23:25
	→ dataangel joined	23:30
dataangel	git newb question: So when you 'add' a file it puts it in the pool of files to be in the next commit. And when you commit, it just commits to your local copy, right? And then, if you decide to have a central repository somewhere, the way that you handle committing to it is have the central repository pull yours, yes?	23:31
	dataangel is seeing if he understands this whole distributed thing	23:31
	→ nud_ joined	23:31
doener	dataangel: a distribution point (e.g. a repo at repo.or.cz) is usually being pushed to	23:33
fonseca	It is easier to push your changes to the central repo.	23:33
dataangel	what does repo.or.cz mean? I saw that written somewhere else	23:34
doener	dataangel: that's a domain name ;-) It's a git hosting service	23:34
dataangel	oh ;p	23:34
doener	ie. http://repo.or.cz/	23:34
dataangel	if you're using a repository like that though, when you commit it's still only to your own local git right? Then when you do a 'push' it can go to some remote host like repo.or.cz, but not when you commit like in cvs?	23:35
doener	yeah. commit is local. Only when you want to publish your commits you push them.	23:36
	→ sNov joined	23:38
dataangel	sweetness	23:43
	I like the idea of tracking history on your box before you have to push :D	23:43
hachi	anyone know how to get git-cvsimport to auth properly with a pserver? I'm dumping the network traffic and it's not sending the password as far as I can see	23:43
dataangel	so if I checkout a project to two different folders, A, and B, and then I make some changes in the A folder, and then commit, the B folder version will never know about it unless I tell it explicitly to merge with the A folder version right? It's not like cvs where you can be in folder A, change things, commit, and then update in B and have the changes reflected there	23:45
hachi	I'm really wanting to check out the source to openldap using git	23:45
doener	dataangel: right. But you can also just create a branch in B. Nobody will see it, unless you push it	23:48
	→ bobesponja joined	23:49
	→ jaalto joined	23:49
jaalto	I mistakenly added many files to be committed. Can the effect be cancelled by "rm .git/index" safely?	23:50
nzkoz	jaalto: git status tells you how to get out of that situation	23:52
mugwump	jaalto: try git reset --mixed	23:52
cehteh	jaalto: git gui offers a nice way to deselect them	23:52
mugwump	or use git gui	23:52
jaalto	git gui is not available (I'm logged remotely)	23:52
	nzkoz: In this case I'm seeking for sledgehammer to calcel all in one sweep. The "git reset HEAD" works for single files	23:53
	→ emiquelito joined	23:53
emiquelito	hi all! how can I change the message of an old commit?	23:54
	the commit hasn't been pushed to the shared repo yet	23:54
mugwump	emiquelito: try git rebase -i	23:54
jaalto	mugwump: The reset cleans the staging area?	23:54
mugwump	jaalto: correct	23:54
jaalto	ok	23:54
mugwump	with --mixed it does	23:54
	with --hard it does the files as well	23:54
emiquelito	mugwump thanks, but I thought rebase -i was used to remove an entire commit	23:55
	I would like to keep the commit but just change its message	23:55
mugwump	you can edit the commit message on the way through	23:55
emiquelito	cool, thanks	23:56
nzkoz	git commit --amend too	23:56
mugwump	I personally use stacked git for this; stg uncommit -n 5; stg pop -n 4; stg refresh -e; stg push -a; stg commit	23:56
	to change the commit message of a commit 4 commits ago	23:56
jaalto	mugwump: I have always been wondering why people would use quilt functionality with git? For me the branches seem more natural	23:57
mugwump	that's right, many people are quite happy with 'git rebase -i'	23:57
	It's a bit unidirectional for me	23:58
	I like to go forward, backwards, change a bit here, a bit there, etc	23:58
jaalto	I mean, separate branch for each feature: Like push/pop with quilt	23:58
	Perhaps it's the workflow question. I on the other hand like to merge features together and keep the lines of development separate (in contrast to push/pop like in quilt)	23:59

Logs Search ←Prev date Next date→

Channels Documentation