IRCloggy #git 2009-10-18

Provider of IRC logs since 2005.

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

	2009-10-18
Ilari	nick0101: Then obtain clone of that repository on server.	00:00
nick0101	sudo adduser \ --system \ --shell /bin/sh \ --gecos 'git version control' \ --group --disabled-password \ --home /home/git \ git	00:00
	I can do that	00:00
	← rudi_s left	00:00
j416	What variable to I have to set to be able to run git from outside the repo I want to work with?	00:00
Ilari	nick0101: 'git@localhost:gitosis-admin.git' or so...	00:01
loops	j416, export GIT_DIR=/path/to/git/.git	00:01
j416	GIT_DIR="/my/work/dir/.git" ?	00:01
	great	00:01
	thanks :)	00:01
nick0101	Ilari: thats right	00:01
	→ selckin joined	00:02
Ilari	nick0101: Now when you have clone, copy public key file from OS X machine to server (/tmp is good place to put it).	00:02
nick0101	ok, just a sec	00:03
	done	00:03
Ilari	nick0101: You want the new key to get the same access rights as the key you have on server?	00:04
	← ntoll left	00:04
nick0101	Yeah	00:04
Ilari	nick0101: Open the admin keyfile in keydir/ of the clone. Copy the line from copied key to be second line of that file (leave the first line alone).	00:05
	nick0101: How username of each key is determined in gitosis: If key is in file called 'foo.pub', then the username is 'foo'.	00:06
	→ jtdowney joined	00:06
Ilari	nick0101: This appiles even if you have multiple keys in the same keyfile.	00:06
	← ludde left	00:07
Ilari	nick0101: Sanity check: What does 'ls -l ~git/repositories/gitosis-admin.git/hooks/post-update' say?	00:07
	← ph^_ left	00:08
nick0101	'/media/mirror/git/repositories/gitosis-admin.git/hooks/post-update'	00:08
sitaram	try uppercase L	00:08
	← thiago_home left	00:08
sitaram	ls -L	00:08
	or try ls -al /media/mirror/git/repositories/gitosis-admin.git/hooks/post-update	00:08
	← ph^ left	00:09
	→ jrmuizel joined	00:09
	← agib left	00:09
nick0101	-rwxr-xr-x 1 git admin 69 2009-10-12 20:42 /media/mirror/git/repositories/gitosis-admin.git/hooks/post-update	00:09
sitaram	good	00:09
nick0101	sorry, didn't grab the whole line	00:09
	glad thats good :(	00:10
	← ThaDon left	00:10
nick0101	:) *	00:10
sitaram	goodno we just wanted to see those "x" bits	00:10
nick0101	alright	00:10
Ilari	nick0101: Now, add the modified keyfile, commit and push.	00:10
	→ qrush_ joined	00:10
	→ ThaDon joined	00:11
	→ hiffy joined	00:11
hiffy	hey dudes	00:11
	so I have a 'class notes' git repo, and I just started a 'assignment' repo and it ocurred to me that it would make sense to have it inside my 'class notes' repo	00:11
	→ WALoeIII joined	00:12
hiffy	is there some painless way to merge/fold one into the other?	00:12
Ilari	hiffy: Make it submodule?	00:12
nick0101	Alright, commited	00:12
hiffy	adding it as a remote and merge does not seem right	00:12
	Ilari: a submodule you say	00:12
Ilari	nick0101: Then push. After pushing try 'ssh git@<server>' from OS X box. It should complain something about SSH_ORIGINAL_COMMAND.	00:13
nick0101	correct	00:13
	→ Mezner joined	00:13
j416	hiffy: I would keep it separate, it would just be all tangled if you keep it together if class notes is not dependent of your assignment. But that's me..	00:14
	→ dsch04 joined	00:14
Ilari	nick0101: Now you should be able to clone the gitosis-admin repo on OS X box.	00:14
	← DavidKlein left	00:14
	→ ilteris_ joined	00:14
nick0101	awesome, that worked	00:14
hiffy	j416: I just wanted to have a "school" repo I have up on github	00:15
	so I don't have to create a new project for every assignment	00:15
nick0101	thanks a lot for the help! I really didn't want to use svn again	00:15
j416	hm i see	00:15
hiffy	j416: it is minor and maybe petty, but I'm curious if it's possible - thinking of it as a hierarchy of folders and all	00:16
j416	yeah	00:16
Ilari	nick0101: The case for adding new user that should be treated as its own: Add file called 'foo.pub' to keydir containing the public key(s) (one per line). Then open gitosis.conf and authorize user 'foo' to perform the wanted actions.	00:16
j416	i see what you want to achieve	00:16
hiffy	it kind of makes sense to have all my 'school related' stuff in one place :)	00:16
	← rodyaj left	00:16
	← dreiss left	00:17
nick0101	Ilari: Alright, I think I should be able to handle it, now that you've directed me to put the keys in the right place :)	00:17
j416	yeah..	00:19
	→ _jason553839 joined	00:19
j416	it would be neat if you could create repo groups in github	00:19
	that would simply correspond to a folder	00:19
	← priidu left	00:20
loops	hiffy, why not just add it as a remote into your school repo and push it up to github as a separate branch, no need to merge	00:20
hiffy	but it has virtually nothing in common	00:21
	here a link to illustrate might make more sense	00:21
loops	doesn't matter... branches aren't required to have anything in common	00:21
j416	hiffy: you could make a new branch from scratch	00:21
	from zero	00:21
hiffy	http://github.com/phillmv/notes/tree/master/fall2009/ suppose I want to slot in my assignment branch under a new folder that reads 'cps633'	00:22
loops	hiffy, i'm not saying it makes sense, i'm just saying you can dump everything into a single repo if that's _what you want_	00:22
hiffy	loops: hah, fair enough	00:22
loops	hiffy, you've already started a separate repo for it, if you fetch that into your school repo, it will be a completely separate independent branch.. no merge needed	00:23
j416	hiffy: you're indenting your notes like source code lol	00:23
hiffy	j416: it makes a lot of sense in my mind	00:24
j416	yeah, it does to me too	00:24
	very tidy :)	00:24
	→ jmatthews joined	00:24
hiffy	:D I was actually thinking of writing some form of syntax highlighter	00:24
j416	just, I never thought of that way before	00:24
	... :P	00:24
hiffy	for the branching and maybe some symbols for stuff for me to lookover later, so ## makes it RED AND BOLD	00:24
	← nick0101 left	00:25
hiffy	but... this is my last semester so not a lot of ROI	00:25
j416	I wish I had git in 2003	00:25
hiffy	if I'm going to be typing it into vim and readin	00:25
j416	(when i started univ) :P	00:25
hiffy	ya	00:25
	→ qhoxie_ joined	00:25
hiffy	so much less overhead oer svn	00:25
j416	svn didn't make sense to me	00:26
hiffy	i would have copies of all of my assignments if I had git back from day one	00:26
j416	yeah	00:26
	neat	00:26
	→ iron joined	00:26
	← fridim_ left	00:26
j416	hiffy: you could make one repo per class	00:27
	→ martianlobster joined	00:27
j416	then you have notes for that class in a 'notes' branch, and assignments in separate branches	00:27
	← kpreid_ left	00:27
martianlobster	is there a "git cp" command like in svn, "svn cp"?	00:27
hiffy	j416: Hm. True; that way I don't have such a clobbered history in case i actually need to revert it	00:28
j416	that's probably what I would do if I didn't keep it in separate repos	00:28
hiffy	for history class who cares not gonna need to revert commits	00:28
j416	hiffy: I think your notes there on github are covering two classes, right?	00:28
	I'd make those two repos if there are assignments aswell	00:28
loops	martianlobster, no.	00:28
hiffy	but for actual code it would be a pain in the ass if i needed to revert and in the process undid abunch of other notes	00:28
martianlobster	loop thanks	00:28
loops	martianlobster, just cp a b; git add b	00:29
hiffy	I think submodule is prolly the way to go	00:29
martianlobster	loops ok makes sense	00:29
j416	I would probably prefix the repo names with university name or some other tag that makes them show together in a sorted list	00:29
	← qhoxie left	00:29
	qhoxie_ → qhoxie	00:29
	← segher left	00:29
j416	hiffy: yes	00:29
hiffy	j416: Yeah... just feels lame to have a 'history essay' repo tho; it's acceptable for my 'implementation of tiny encryption algorithm'	00:29
j416	so separate branch for project stuff :)	00:29
	hiffy: well, it's an essay right	00:29
Ilari	tiny encryption algorithm? Sounds weak. :-)	00:30
j416	Ilari: lol	00:30
hiffy	Ilari: hehe	00:30
j416	so that's an assignment in one branch "essay" and notes in your branch "notes"	00:30
	→ kpreid_ joined	00:30
loops	now with two-bit keys!	00:30
hiffy	and rebase whenever it's done	00:30
j416	then you can write your essay and don't worry about deleting text you think you don't need :)	00:30
hiffy	loops, Ilari ACTUALLY it's about as strong as DES iirc :B	00:30
j416	I've been using git for writing stuff, it's great.	00:31
	'writing-stuff' that is	00:31
Ilari	hiffy: I.e. Key is too short and likely the algorithm is bad.	00:31
hiffy	Ilari: I got the joke ;). It's just an actual real algo I didn't pull out of my ass http://en.wikipedia.org/wiki/Tiny_Encryption_Algorithm	00:32
	128bit keys	00:32
	tho a theoretical attack brings it down to 117 bits or something	00:32
loops	j416, i've been looking for words and sentences in commit sha1's. waiting for divine revelation	00:32
	Tank` → stsch	00:33
Ilari	DES is 56 bits... 3DES is 112 (both two- and three-key variants).	00:34
	← jrmuizel left	00:35
	← kpreid_ left	00:36
	← krawek left	00:36
hiffy	Ilari: actually the double DES has a theoretical attack of about 57 or 58 bits if you get to choose enough plaintexts	00:36
j416	loops: o_O	00:36
	hiffy coughs	00:36
j416	:D	00:36
	← justin-george left	00:37
hiffy	I like the idea of all these security researches storing 2^57 strings of 56 bits	00:37
	*researchers	00:37
Ilari	hiffy: I think one needs only few plaintexts for meet-in-the-middle attack against double DES. That's why triple DES is used (there the same attack brings it down to 112 bits).	00:38
	← loops left	00:39
	→ loops joined	00:39
	← _jason553839 left	00:40
	← loops left	00:40
	← pheaver left	00:41
	→ pheaver joined	00:41
SRabbelier	what's wrong with my vim configuration again if hitting up-arrow in insert mode adds an 'A' instead of going up?	00:43
	← __iron left	00:43
	← Yuuhi left	00:43
	→ DIDAVISION joined	00:44
j416	SRabbelier: the #vim guys should know. :)	00:44
	→ FFighter joined	00:49
j416	I wonder how many people use git locally and never do remote stuff because they have some kind of OCD	00:53
Mikachu	SRabbelier: probably wrong $TERM set	00:53
SRabbelier	Mikachu: partially solved by 'touch ~/.vimrc', but now ctrl+w isn't working, or backspace	00:54
	→ defectiv joined	00:55
SRabbelier	Mikachu: apparently I need to read ":he 'bs"	00:55
	← fr left	00:56
	→ fr joined	00:56
	→ kpreid_ joined	00:56
Mikachu	as for why it inserts an A, the up arrow literally sends esc [ A to the program	00:57
	→ kardan_ joined	00:58
	→ D joined	00:58
	D → Guest90464	00:59
SRabbelier	Mikachu: ah, interesting; learn something new every day	01:00
	→ Adlai`` joined	01:00
sitaram	(is also why sometimes when you hit Escape to get out of insert mode it takes a fraction of a second; the terminal emulkator is waiting for any other characters which would indicate an F-key eing pressed)	01:01
^Einstein	can git accept vanilla subversion commits and/or branch manipulation, or does it only work the other way around?	01:01
sitaram	being*	01:01
SRabbelier	sitaram: ah, cool	01:02
	^Einstein: what do you mean?	01:02
	→ jtaby joined	01:05
	← mbroeker left	01:05
	→ _jason553839 joined	01:06
^Einstein	SRabbelier: git-svn enables using git with subversion upstream repos, but is there a way a novice user can use svn to talk to a git repo with the git-svn plugin?	01:07
SRabbelier	^Einstein: no	01:10
	^Einstein: that would be something you'd have to ask the svn people to create	01:10
	→ loops joined	01:10
^Einstein	SRabbelier: I didn't know if git-svn had an API that recognized when it was an svn client rather than a git client that was connecting, and behave appropriately.	01:11
	← mjf left	01:11
SRabbelier	^Einstein: no, git-svn is a part of git, it is not a server	01:11
	← riverrat66 left	01:11
	← DIDAVISION left	01:11
^Einstein	git is both a client and a server, but subversion is not.	01:11
SRabbelier	^Einstein: git is not a server	01:12
^Einstein	Just curious, that's all.	01:12
SRabbelier	^Einstein: git _has_ a server, called 'git deamon', but the 'git' command is not a server	01:12
^Einstein	SRabbelier: with git there is less of a client-server model, because all clients are servers. Am I incorrect in that description?	01:12
SRabbelier	^Einstein: that is incorrect	01:12
	← redondos left	01:12
SRabbelier	^Einstein: clients are not servers	01:12
	← rotty left	01:12
	→ rotty joined	01:12
	→ bdimcheff joined	01:12
loops	Git is considered peer to peer, but practically speaking there are git clients and git servers	01:13
	→ redondos joined	01:14
	← warthog9 left	01:15
	← kardan\| left	01:15
j416	pear-to-pear	01:17
SRabbelier	j416: I prefer apple-to-apple	01:17
j416	me too	01:17
	← Adlai` left	01:20
	← dwmw2_gone left	01:22
	→ VVSiz_ joined	01:23
	← vvsh left	01:27
	← _jason553839 left	01:27
	← Determinist left	01:27
	← iron left	01:31
	← martianlobster left	01:31
	→ Davey joined	01:33
	→ JasonWoof joined	01:35
	← orafu left	01:35
	← MisterN left	01:35
	→ orafu joined	01:35
	← lightcap_ left	01:36
	← defectiv left	01:37
	← Davey left	01:37
	→ Dashkal joined	01:38
	→ Davey joined	01:40
	← qrush_ left	01:41
	← VVSiz left	01:42
	→ warthog9 joined	01:43
	→ pheaver_ joined	01:47
	← pheaver_ left	01:48
	→ pheaver_ joined	01:48
	← pheaver left	01:48
	→ Hiron joined	01:48
	pheaver_ → pheaver	01:48
	← jtaby left	01:48
	← kpreid_ left	01:49
	→ kpreid__ joined	01:49
	→ onigiri joined	01:50
	→ yaotti joined	01:51
	← jtdowney left	01:55
	← Hiron left	01:58
	→ Adlai``` joined	02:02
djszapi	Hello!	02:02
	How can I view the differences betweem two branches ?	02:02
	for the same named file.	02:02
	→ jrmuizel joined	02:04
loops	djszapi, git diff branch1:filename branch2:filename	02:05
	djszapi, or git checkout branch1 ; git diff branch2 -- filename	02:06
	← jrmuizel left	02:06
	→ _jason553839 joined	02:06
chuck	how do I remove a folder from version control without actually removing the files	02:11
loops	chuck, git rm --cache <dir>	02:12
chuck	loops: thanks	02:14
	→ dreiss joined	02:14
sitaram	usually followed by adding it to .gitignore	02:14
Ilari	'git diff branch1 branch2 -- filename'	02:15
chuck	sitaram: yes	02:15
	removing the build directory from my iphone app's repo	02:15
loops	djszapi, ^^ Ilari	02:16
	← tazle_ left	02:17
	← tetha left	02:17
	→ tetha joined	02:18
	← WALoeIII left	02:21
	→ pheaver_ joined	02:22
	← Adlai`` left	02:22
	→ eno joined	02:24
	← _jason553839 left	02:25
	→ cytrinox_ joined	02:25
	→ JKac3BEq joined	02:26
	← JakeSays left	02:28
	→ JakeSays joined	02:32
	← onigiri left	02:35
	← eno__ left	02:35
	← cytrinox left	02:37
	cytrinox_ → cytrinox	02:37
	← pheaver left	02:39
	pheaver_ → pheaver	02:39
	← Buey left	02:39
	→ coolcat joined	02:42
	← kpreid__ left	02:42
	→ kpreid_ joined	02:44
	→ WALoeIII joined	02:51
	← Guest90464 left	02:52
	→ tjaway joined	02:52
	← kpreid_ left	02:55
	→ godsflaw joined	02:56
	→ godsflaw_ joined	03:05
	← tjafk left	03:08
	→ kpreid_ joined	03:08
	← tjaway left	03:09
	→ tjaway joined	03:09
	← Chillance left	03:09
	→ ryoma joined	03:10
	← Crewe left	03:11
	← LiamH left	03:11
	← godsflaw left	03:13
dngor	Is there a technical doc on git-svn fetch?	03:15
SRabbelier	dngor: technical doc?	03:15
	→ aresnick joined	03:16
dngor	I want to know how it works, and why it does something in particular.	03:16
	If in svn I tag a directory rather than the whole trunk, git svn fetch creates two tags: svn/tags/tagname@rev and svn/tags/tagname	03:17
	← kukks left	03:17
dngor	The next svn commit's parent is the svn/tags/tagname@rev, so the svn/tags/tagname becomes a little dead-end branch.	03:17
	← coolcat left	03:18
dngor	They look like this: http://poe.dyndns.org/~troc/tmp/poe-svn2git/dead-end-branchlets.png	03:21
loops	dngor, if you dont get an answer here try on mailing list	03:22
dngor	Thanks. I'm also hitting th' web, 'cause it knows all.	03:22
	← Mezner left	03:23
	← FFighter left	03:26
	→ hyperair joined	03:31
	← dunolie left	03:33
	← kpreid_ left	03:35
	→ kpreid_ joined	03:35
	→ davido joined	03:40
	→ dunolie joined	03:42
	→ jaysern joined	03:43
	→ Mezner joined	03:49
	→ scarabx joined	03:50
	← hyperair left	03:53
	← Mezner left	03:54
	→ lenix^ joined	03:55
	→ Crewe joined	03:56
	← Dashkal left	03:58
	← Crewe left	03:59
	← Sho_ left	04:07
	Adlai``` → Adlai	04:14
jaysern	i'm trying to pull from my git repo, and i'm getting a merge error	04:15
	→ jrmuizel joined	04:16
jaysern	error: Entry 'settings.py' not uptodate. Cannot merge.	04:16
	fatal: merging of trees 94627dd66106d09aa59d340e3398596472c2c276 and c47d12800a1ea92e6e10d91f655b1e185ec7818f failed	04:16
	Merge with strategy recursive failed.	04:16
	does that mean i need to resolve the conflicts by hand ?	04:16
sitaram	jaysern: most people avoid merging, pulling, etc on a dirty tree	04:16
	← jrmuizel left	04:17
jaysern	what is a dirty tree ?	04:17
sitaram	no conflict to resolve; the merge didnt happen -- commit your stuff first then pull	04:17
	→ jrmuizel joined	04:17
sitaram	dirty tree is where you have unstaged/uncommitted changes to tracked files	04:17
edenc	w 31	04:17
loops	you sank my battleship	04:18
sitaram	31? thats a big grid (or I forgot how that game goes in detail!)	04:18
loops	heh.. details details	04:18
jaysern	on my local dev box, i try to do a commit -a	04:18
	nothing added to commit but untracked files present (use "git add" to track)	04:18
	and then when i try to push, it says everything up-to-date	04:18
	so on my web server, i try to do a pull, but it tells me of the merge error	04:19
sitaram	jaysern: ok, this means that you have a local untracked file, caled settings.py, which the other guy has started tracking between your last pull and the current remote. Bringing it in would overwrite your local changes.	04:19
	jaysern: 'git stash' is another way to deal with it, if you're not sure you want to commit your local stuff first	04:19
jaysern	ahh	04:20
	i was tinkering with that file on the web server	04:20
	what if i do want that file over written ?	04:20
loops	jaysern, just delete it and pull	04:21
sitaram	jaysern: just rm the file then (make sure it is shown in 'git status' as untracked)	04:21
	← Tichodroma left	04:27
	← kpreid_ left	04:29
	→ kpreid__ joined	04:29
	→ SRabbelier1 joined	04:31
	← SRabbelier left	04:31
	← Akufen left	04:31
jaysern	sweet! thanks guys	04:32
	← jrmuizel left	04:35
	← j416 left	04:35
	← ryoma left	04:38
	← sh1mmer left	04:39
	→ sh1mmer joined	04:43
djszapi	git diff master s_cube/maint -- Makefile.am	04:45
	warning: refname 's_cube/maint' is ambiguous.	04:45
	@Ilari, loops. It can't work ^^^	04:45
doener	djszapi: probably you have refs/heads/s_cube/maint and refs/remotes/s_cube/maint	04:45
	djszapi: try "git for-each-ref \| grep s_cube/maint" to check	04:45
djszapi	git for-each-ref \| grep s_cube/maint	04:46
	7253e66996a192df8b4843f9481316a9507d59da commit refs/heads/s_cube/maint	04:46
	142166438c373842bc9d3cd921ce13db0e980fe8 commit refs/remotes/s_cube/maint	04:46
doener	djszapi: see? :-)	04:46
djszapi	yes, thanks doener	04:46
doener	djszapi: you can disambiguate by adding heads/ or remotes/	04:46
djszapi	okok	04:46
	I understood	04:46
	git diff master refs/remotes/s_cube/maint -- Makefile.am	04:48
	djszapi ~/Munka/ArchLinux-devel/Projects/aurman $	04:48
	Do I need for similar method in case of master too ?	04:48
	→ jrmuizel joined	04:49
doener	djszapi: no, that succeeded without ambiguity. Makefile.am just doesn't differ between these two	04:49
	← jrmuizel left	04:49
djszapi	it differs.	04:49
loops	i'm sure this has been discussed on the list and rejected, but it seems to me heads/ should be assumed to disambiguate and user can supply remotes/ to override it	04:50
djszapi	http://pastie.org/659236	04:50
doener	loops: heads/ is already assumed. That's why he got a warning, not an error	04:51
	djszapi: that shows that you have uncommitted changes to Makefile.am, not that Makefile.am differs between any two branch heads	04:51
	→ jrmuizel joined	04:51
loops	doener, ah right you are. I don't think it's warning worthy either.. .warning is too scary	04:51
doener	loops: hm, and if you actually meant remotes/whatever? Without the warning, you'd not even know that it picked the "wrong" one	04:52
djszapi	doener what can I do now without committing anything, to make a backup from a branch on my computer and vimdiff with it the other branch ?	04:52
loops	doener, if you don't know the difference, how could you ever "mean" remotes/...	04:52
doener	loops: hm? I'd not assume that the user is at all times aware that he has heads/foo and remotes/foo	04:53
loops	oh crikey... indeed..	04:53
	loops crawling back under rock	04:53
doener	loops: he might very well know the difference, but might not be aware that he has two refs that have ambiguous shortnames	04:54
	djszapi: difftool can drive vimdiff	04:54
djszapi	it's not the question	04:54
	I've set it	04:54
	the question is how can I view the difference between two branches without commiting them ?	04:55
doener	djszapi: hu? Branches consist entirely of committed stuff. And "git diff <branch_head_A> <branch_head_B>" will happily produce a diff between these two branch heads	04:56
	djszapi: and difftool will do so using the configured diff tool	04:56
	djszapi: but Makefile.am simply doesn't differ between the two refs you gave to "git diff"	04:56
	djszapi: your paste just showed that it differs between "master" and your working tree	04:57
	djszapi: you can pass just a single ref to "git diff" to diff that against the working tree, if that's what you want	04:57
djszapi	yes, how ?	04:58
doener	as I said, "git diff <single_committish>"	04:58
	e.g. "git diff master" or "git diff remotes/s_cube/maint" or "git diff HEAD~5^2~3"	04:58
djszapi	??	04:58
	What ?	04:58
	Doener..	04:59
	I've got a file in master without being in stage, and the same situation in another branch, so without staging there too, but I'd like to view the diff between them.	04:59
doener	djszapi: you only have a single worktree/index. Your uncommitted changes are in there. They're not part of any branch	05:00
tpope	git diff branch1 branch2 -- file	05:00
djszapi	The comparable file are in separate branch without any staging.	05:00
	tpope: pls. read back...	05:00
	doener: I really can't understand	05:00
	Is there only one index for Makefile.am e.g. ?	05:01
	even more branch exists with modification for that file ?	05:01
SRabbelier1	djszapi: no	05:01
doener	djszapi: there's only one index, it's neither per branch nor per file	05:01
SRabbelier1	djszapi: there can be only 'uncomitted' version of a modified file	05:01
	← dreiss left	05:01
	SRabbelier1 → SRabbelier	05:01
doener	djszapi: if you modify "foo" without committing, and then switch to another branch head, those changes are kept in your work tree and/or index	05:02
	← schmukkk left	05:03
doener	djszapi: http://git.pastebin.com/m4bbecbc7	05:03
	djszapi: the "M test" line tells that I had uncommitted changes that were carried along	05:04
	← lenix^ left	05:16
	→ lenix^ joined	05:17
	← aresnick left	05:18
	← kpreid__ left	05:21
	→ kpreid_ joined	05:22
	→ machrider joined	05:23
	← Bass10 left	05:25
machrider	are there any git commands that operate on a repository via URL, rather than fetching something locally?	05:27
	i want to do some simply queries, like get the latest commit sha and message	05:27
	simple*	05:28
doener	machrider: just fetch, push and ls-remote IIRC. Things like log, diff, blame, etc. require things to be present locally	05:28
loops	machrider, you can do git ls-remote <url> to get current sha1s and branch names	05:30
machrider	yeah, it's almost what i want :)	05:30
	→ lightcap_ joined	05:31
loops	machrider, well for more you have to follow doeners advice :o_	05:31
	:o)	05:31
doener	not much more you can do, except for fetching things, but then we break a pre-condition ;-)	05:31
machrider	fetch isn't the end of the world, but it's heavy.	05:32
	but yeah.	05:32
loops	machrider, ah, you're looking for the message as well. you might be able to scrape it, if the site provides gitweb access	05:32
doener	or if you got ssh access: ssh $host 'git --git-dir=/path/to/.git branch -v'	05:33
	→ cjs_ joined	05:35
machrider	thanks guys.	05:35
cjs_	So it's possible to have a checkout that uses the repo files (i.e., the .git directory) from another checkout, right? Do I have my terminology correct?	05:36
	← jaysern left	05:36
machrider	cjs_: s/checkout/working copy, regargind terminology	05:37
	regarding. jeez. can't type.	05:37
loops	cjs_, it's possible to have multiple working trees without needing complete copies of the repo .git in each	05:37
	← jrmuizel left	05:37
doener	cjs_: It's a rather uncommon setup, not sure if there's an "official" term, but the script that does the setup uses "workdir" for the combination of working tree + index + HEAD (+ maybe other things I forgot)	05:37
	cjs_: a more common thing is to have a repo that uses the "alternates" mechanism	05:38
	→ jaysern joined	05:38
machrider	cjs_: the git-new-workdir script works well for me.. it's in contrib/	05:38
loops	cjs_, the script doener mentioned is in the contrib directory in the official git.git repo	05:38
cjs_	Ah, ok; "repo" and "working copy" are the terms I'm looking for, I guess. And it's also possible to have a repo without a working copy, of course. So it should be possible to have a repo without a working copy, and a working copy in a separate directory based on that repo, right?	05:38
loops	repo without working tree = bare repository	05:38
cjs_	"alternates" mechanism?	05:39
doener	cjs_: with that setup, you have a full repo (i.e. independent refs), but an "alternate" that instructs git to look into other repos to find objets	05:39
	s/objets/objects/	05:39
	→ g3d joined	05:39
doener	cjs_: both setups have drawbacks. With a second workdir, you get into problems with more than one workdir has the same branch head checked out at the same time	05:39
cjs_	doener: Ah, thanks. That's not quite what I'm looking for. I'm actually kind of wanting to use (abuse?) git as a sort of backup solution.	05:40
	← lightcap_ left	05:40
doener	cjs_: sounds like you just want .git to not live at .git but at /random/path?	05:40
cjs_	doener: Mmm, I can imagine. Though things out to be ok if you're never making modifications in more than one workdir at once, right? I.e., modify, commit, and then a git co in another workdir should just get that update. Or does that not work?	05:40
	Sorry, reset, I meant.	05:41
doener	reset --hard even (or checkout -f)	05:41
	← jaysern left	05:41
doener	both drop uncommitted changes	05:41
loops	cjs_, you might want to checkout http://eigenclass.org/hiki/gibak-backup-system-introduction there are 3 backup on top of git solutions mentioned there	05:41
cjs_	loops: Ooo, thanks!	05:41
	And OCaml tools are a bonus, though not as much as Haskell tools would be. :-)	05:42
	← p_masho left	05:43
cjs_	So, I guess my questions would be, is it possible to have a working copy of only part of a repo (e.g., everything under /foo, but not anything under /bar, though /bar is in the repo), and is there some clever way to use a bare repo from across a network, without having a local copy.	05:43
	→ p_masho joined	05:44
	← g3d left	05:44
	← Davey left	05:44
cjs_	If you're curious about the application, I'm wanting to drop my entire e-book library into git, so that I can easily use it across multiple machines. However, though the interesting chunk is only a few dozen MB (at the moment), there happens to be an 11 GB gutenberg subdir that is not going to come anywhere near fitting on my laptop.	05:45
doener	cjs_: sub-checkout: not without trickery, and only read-only. Remote bare git-repo: Only if you can mount it (e.g. fuse sshfs)	05:46
	cjs_: what you could do, is creating multiple repos (one for foo/, one for bar/) and another "super-repo".	05:47
	cjs_: that super-repo could either have the other repos as submodules, or merge them using the subtree merge strategy	05:47
cjs_	Right, I was thinking of sshfs for that. As for read-only sub-checkout, hm, well, that doesn't give me anything over just running rsync.	05:47
	→ jtdowney joined	05:47
doener	cjs_: subtree merge: http://www.kernel.org/pub/software/scm/git/docs/howto/using-merge-subtree.html	05:47
	← wagle left	05:47
cjs_	Hm. I thought about splitting the repos, but I start doing that with all of my stuff and I'm going to have a repo explosion.	05:48
	→ lightcap_ joined	05:49
	→ jaysern joined	05:49
cjs_	So, I guess I could handle that to a great degree with submodules, though.	05:50
	← jaysern left	05:51
cjs_	Hm. Well, it sounds as if what I probably want is just to keep standard (non-bare) repos on a host with gobs of storage space, and then just write a couple of simple rsync scripts to "pull" and "push" the appropriate subsets of those working copies to and from my laptop.	05:53
sitaram	cjs_: sorry; just came back. e-book == binary format file of some sort (like pdf)? why version control at all... (just curious...)	05:55
	in other words, if you can't reasonably branch, merge, etc., all you need is a backup solution, not a VCS, no?	05:56
cjs_	Because I have multiple (usually 3 or 4, at least) copies of my ebook library, and when I deal with that with rsync, adds can get lost fairly easily.	05:56
	I keep a partial copy on my laptop, a full copy on my home machine, another on my work machine, and I really should have a backup copy somewhere on a different continent, too.	05:57
sitaram	cjs_: multiple copies on multiple computers or multiple versions of the files even?	05:57
	← panfist left	05:58
	→ jaysern joined	05:58
	qrush → qrush\|away	05:58
sitaram	cjs_: I'd use unison to sync them up. One "unison profile" for each pair of locations. Use unison's builtin "ignore" facility to get only what you want. etc	05:58
luke-jr	cjs_: uh, omit --delete from rsync	05:59
cjs_	luke-jr: Yeah, but then I move a file, and I end up with multiple copies of it in other repos, which I then have to sort out manually.	05:59
sitaram	far be it from me to advocate something other than git on #git, but you're not using all the real power of git anyway in an situation like this	05:59
	→ JED3 joined	05:59
sitaram	aah -- moves unison doesn't handle well; even renames... if you do that very often you may waste a lot of bandwidth	06:00
	→ jtaby joined	06:00
cjs_	sitaram: I've looked at Unison, but because each arc in the graph has to be explicitly configured, it's not so convenient as git's automatic full mesh.	06:00
	← jtaby left	06:00
cjs_	Basically, with git I can transfer between any repo, any time, with no extra configuration.	06:01
sitaram	cjs_: git has an automaic full mesh? I thought you have to still set up each remote yourself?	06:01
cjs_	Nope. Just give it a repo URL and push or pull away.	06:01
loops	a remote is just a shortcut	06:02
	So it's not required	06:02
sitaram	cjs_: yeah I see what you mean; it's a lot more cumbersome with unison -- at least a profile file with 2 URLs in it, one being the local directory and the other being the remote one	06:02
	especially ifyou can type URLs fast enough, it doesn't even save you any time to say "unison lap-home" or "unison lap-work"	06:03
cjs_	Right. And then, git's automatic compression can be helpful for things like archives of log files, I'm thinking.	06:03
	← jaysern left	06:04
sitaram	you've got log files in an e-book library? ok then I retract what I started this with	06:04
cjs_	No, no. Different thing.	06:04
dngor	I can fix my dead-end branchlets with grafts.	06:04
luke-jr	cjs_: you might look into extending my gitbackup script	06:04
cjs_	But the same sort of problem: I have the master copy of my log files on the web server, but I need to back those up, I need to copy them to other machines for analysis, etc. etc.	06:04
	luke-jr: where would I find it?	06:05
luke-jr	cjs_: Gitorious	06:05
cjs_	Ah.	06:05
luke-jr	for the log thing though, I'd probably use tail -F :p	06:05
	gitbackups assumes that on average things don't change more than once every few minutes	06:06
cjs_	luke-jr: I'm not clear on how that would work. Let me describe my situation. I have three web servers, say, one primary and two backup ones, all in an identical configuration. The primary vanishes for whatever reason, I just switch the DNS to a backup. Each generates perhaps a few GB/year of log files. (Well, some backups may generate almost nothing.) I need to merge these, back up the merged version, etc.	06:08
	I do this all with manual renaming, compression, and copying at the moment. But there's got to be a better way.	06:08
	→ bsdvodsky joined	06:08
sitaram	cjs_: merging in this sense is not the same as git's merging, I take it?	06:09
cjs_	Again, the partial working copy thing, if we had it, would seem as if it could be helpful.	06:09
luke-jr	sitaram: sounds like it is	06:09
	though he'd need a custom merging algorithm	06:09
cjs_	sitaram: No, not really.	06:09
luke-jr	cjs_: either way, you're doing it semi-wrong :p	06:10
	sounds like you should be using DNS round-robin	06:10
cjs_	I'm happy with, e.g., server1-current.log, server2-current.log, server1-2008-07.log, etc.	06:10
sitaram	cjs_: I'd just cat all the log files; each access happened, that's it; assuming timestamps are correct, maybe sort by time to get a combined access log	06:10
luke-jr	cjs_: oh, that will work easier	06:10
sitaram	aah so merge only at the directory level not file level	06:10
cjs_	luke: I could, if I wanted to distribute the load as well. But round-robin is not what you want when you're trying to deal with server failures, because if a server vanishes, you still have to remove it from the DNS.	06:11
luke-jr	cjs_: no, round-robin is also failover	06:11
	if one IP is inaccessible, browsers automatically try the next	06:11
cjs_	luke-jr: They do?	06:11
luke-jr	provided they're in the same IP namespace anyhow	06:11
	if all your IPv6 servers are down, it won't try IPv4 addresses	06:12
cjs_	Last I checked, browsers cached only a single address after a lookup, and continued to use that until it expired.	06:12
luke-jr	cjs_: if they did, they'd be broken :)	06:12
	I've never had problems with it	06:12
	and I intentially make servers in a round robin down once in a while	06:13
	intentionally*	06:13
	of course, that round robin is used by a game, not browsers much	06:13
cjs_	Hm. I'll have to have a look again. But the semantics of getting back 3 A records, in the DNS point of view, is not "use these until you find one that works," but pick any one of these and use it.	06:13
luke-jr	cjs_: and if it fails, try another	06:13
	that's why they get 3 records	06:13
	if it was merely random, the server would only send one	06:13
cjs_	luke-jr: That's an application level semantic.	06:14
luke-jr	cjs_: the standards say applications must do that, IIRC	06:14
cjs_	Do they? I don't recall that....	06:14
luke-jr	actually, it's not even really application level	06:14
cjs_	Hm. I'll have to look into that again....	06:14
luke-jr	the standard socket layer returns things in a way that applications really NEED to do that or they won't work	06:14
	← kpreid_ left	06:15
luke-jr	assuming they're not using ancient gethostbyname or such	06:15
	→ kpreid_ joined	06:15
cjs_	So, luke-jr, basically your positing that most application will call getaddrinfo, and work through that list until they find something to which they can connect successfully.	06:16
luke-jr	cjs_: anyhow, on your central backup box, ssh server1 tail -n 10000000 -F /var/log/foo.log > server1_foo.log	06:16
	cjs_: if they don't, they will break in fairly normal circumstances	06:16
	getaddrinfo will likely return the IPv6 address first, which many hosts are too stupid to connect to	06:17
cjs_	luke-jr: I run my backups the other way; servers push their backups to the backup server.	06:17
luke-jr	if they don't try the next one, they won't work with IPv4	06:17
	tail -n 10000000 -F /var/log/foo.log \| ssh backup 'cat > server1_foo.log'	06:17
cjs_	Though, hmm, I suppose I could have a secure backup pull by making the files to back up available via rsync.	06:18
luke-jr	real-time log backups save you from script kiddies covering their tracks, too	06:18
	could always aim for remote logging instead, as well	06:19
	why make local log files when you can just configure Apache to log it remotely directly? :p	06:19
	bed time for me, night!	06:20
	→ jaysern joined	06:20
	→ Sigma joined	06:25
	← yhager left	06:28
	→ savant joined	06:32
	← savant left	06:32
	← jtdowney left	06:34
	→ mbroeker joined	06:38
rtyler	hiyo folks; I see that through fast-export stuff you can export an hg repository	06:39
	is it possible to keep a git repo in sync with hg?	06:39
	→ abadr joined	06:40
abadr	What if I want a branch to become master and want to drop any changes made to master since the branching?	06:40
	← Fragsworth left	06:41
loops	abadr, one way is.. git checkout master ; git reset --hard branch ; but it will throw away any uncommitted changes	06:42
abadr	uncommitted changes in master?	06:43
	oh nm	06:43
loops	rtyler, check out hg-to-git in the contrib directory of the git.git repo, you can do incremental imports from an hg repo similar to git-svn	06:43
	abadr, in your working tree	06:43
	← TriggerB left	06:44
rtyler	loops: you mean the github project?	06:44
loops	rtyler, no the official Git repository. Some distributions package hg-to-git up in an rpm or whatever as well	06:45
abadr	loops: I guess that worked, but I'm confused about what happened	06:45
	where is the commit that got rid of all the recent stuff in the master branch?	06:45
loops	rtyler, git://git.kernel.org/pub/scm/git/git.git	06:46
	→ jrmuizel joined	06:47
loops	abadr, there was no commit made. The reset just made "master" point to the same series of commits that "branch" points to	06:47
	rtyler is cloning now	06:47
abadr	interesting	06:48
loops	abadr, this was the wrong way to go if you're working on a public/shared repo	06:48
	abadr, i should have asked first	06:48
abadr	I'm not	06:48
	it's private	06:48
	I guess I can delete the branch now. Are those post-branching commits to master anywhere now? I don't need them, just curious.	06:48
loops	abadr, yes you can delete the branch now. the commits that use to be referenced by master are now available in the reflog "git reflog master"	06:49
voker57	they are dangling, i think	06:49
loops	abadr, the reflog expires over time. Once they are no longer referenced anywhere, including the reflog, they are considered dangling and will be garbage collected	06:50
	oops.. git reflog show master	06:50
	→ Ozzy joined	06:51
	← Ozzy left	06:51
abadr	interesting	06:52
	← scarabx left	06:53
abadr	is there a way to do the same thing that creates a commit instead?	06:53
rtyler	loops: any decent guides on how to use this sucker?	06:53
loops	rtyler, unfortunately none that i know of.	06:55
	← bdimcheff left	06:55
loops	abadr, not sure what you're shooting for there	06:55
abadr	well, this may not have been the right thing to do. it's not "public" in that other people use it, but I do sync to and from a remote from multiple places	06:57
	not sure how to do that now	06:57
rtyler	loops: well drats	06:57
loops	abadr, yeah.. it's not recommended to do what we did on repos you're sharing.. even if it's only with yourself ;o)	06:59
abadr	loops: now what? :)	06:59
loops	abadr, PANIC!	06:59
	abadr runs around screaming.	06:59
loops	abadr, 1 sec.. real life is intruding here	07:00
	← pheaver left	07:00
dngor	Is there a git command to find the child(ren) of an object?	07:00
	→ justin-george joined	07:01
	→ thiago_home joined	07:03
loops	abadr, okay.. sit down..	07:06
abadr	lying down, in fact	07:06
loops	dngor, no. Git only has parent pointers	07:06
	dngor, of course it could be written by searching	07:07
	→ drizzd joined	07:07
	→ kpreid__ joined	07:08
	← kpreid_ left	07:08
	→ priidu joined	07:08
	← jrmuizel left	07:08
	→ c15c8ra1n joined	07:08
loops	abadr, you could treat this like you're operating in a shared/public repo situation	07:09
	abadr, that is you would never rewind master like we did with that reset	07:10
abadr	I'm listening	07:10
loops	abadr, instead you would use "git revert" to make _new_ commits that reverse the changes made by the commits you don't like	07:10
	abadr, and then apply new changes on top of that	07:10
	abadr, so when remote repos pull, all they see are new commits and they will be happy. even though some of those commits essentially undo or throw away existing commits	07:11
abadr	first, how do I get master back to the remote master?	07:11
	← sh1mmer left	07:12
loops	abadr, assuming you haven't made any more commits or done anything else.. git reset --hard HEAD@{1}	07:13
	→ aigon joined	07:13
loops	abadr, which reverts to the commit saved in the reflog (praise the reflog)	07:13
abadr	whew. ok	07:13
loops	abadr, did ya do that?	07:14
abadr	yep	07:14
loops	abadr, okay.. the reflog is a great safety valve for stuff like this	07:14
fujin	<3 reflog	07:14
	→ snokat joined	07:15
snokat	http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733	07:16
	http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733	07:16
	http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733 http://vkontakte.ru/reg21901733	07:16
Adlai	interesting	07:17
abadr	cool links bro	07:17
loops	not really	07:17
Adlai	I love how if you check which channels these spambots are in, they always pick #ubuntu	07:17
	← snokat left	07:18
abadr	loops: now what? :)	07:18
loops	abadr, doh.. yeah.. we're not done are we ;o)	07:18
	abadr, okay this should work... git checkout branch ; git merge -s ours master ; git checkout master ; git reset --hard branch"	07:21
	← justin-george left	07:22
loops	abadr, that assumes that branch and master have a common ancestor at some point.. ie. diverged from each other	07:22
abadr	yep	07:22
	→ hyperair joined	07:22
abadr	hey, I think that worked	07:26
loops	abadr, it should. you should double check cause i'm not in top form tonight ;o)	07:26
	→ jjuran joined	07:26
abadr	well i've learned from you anyway :) thanks for all your help	07:26
	→ siprbaum joined	07:27
loops	abadr, np, good luck with it	07:28
	← aigon left	07:30
	→ Aikawa_ joined	07:42
	→ ph^ joined	07:42
	← JED3 left	07:45
	← sdboyer left	07:48
	→ Guest33124 joined	07:51
	← Aikawa left	07:54
	← cloudhead left	07:54
	→ ciskje joined	07:55
	→ dreiss joined	08:01
	← kpreid__ left	08:01
	→ kpreid_ joined	08:01
	→ PerStrandh joined	08:02
	← machrider left	08:02
	← warthog9 left	08:03
	← Azeroth-Work left	08:04
	→ segher joined	08:06
	← qhoxie left	08:09
	← abadr left	08:09
	← yaotti left	08:10
	← PerStrandh left	08:14
	→ ludde joined	08:15
	→ jfxberns joined	08:18
	→ Yewbacca joined	08:21
	→ abadr joined	08:25
	← jaysern left	08:27
	→ ph^_ joined	08:29
	→ Adlai` joined	08:30
	← Adlai left	08:30
	Adlai` → Adlai	08:30
	← ciskje left	08:34
	→ j416 joined	08:35
	VVSiz_ → VVSiz	08:35
	→ pantsman joined	08:36
	← lightcap_ left	08:39
	→ xide joined	08:39
	→ relix joined	08:39
xide	i need 9 ppl to register at this link, that's kinda of a promotion in the russian facebook) sum1 do me a favour, spend 2 minutes of ur life and follow this link http://vk.com/reg480180	08:39
thiago_home	no	08:42
djszapi	How can I delete the files that are under revision control, but they took part in the .gitignore file, something similar like git-clean -x without removing the untracked files.	08:42
thiago_home	djszapi: git clean removes untracked files only	08:42
	djszapi: it never removes tracked files	08:42
	djszapi: to remove tracked files, use git rm	08:43
	← abadr left	08:43
	→ mmarsu joined	08:43
djszapi	Normally, only files unknown to git are removed, but if the -x option is specified, ignored files are also removed. This can, for example, be useful to remove all build products.	08:43
	git clean --help	08:44
thiago_home	djszapi: right	08:44
	djszapi: but I didn't say ignored	08:44
	djszapi: I said git clean removes untracked files	08:44
	djszapi: files that are tracked are not untracked. So git clean will not remove them.	08:44
djszapi	-X	08:45
	Remove only files ignored by git. This may be useful to rebuild everything from scratch, but keep manually created files.	08:45
thiago_home	djszapi: a file that is tracked is not ignored.	08:45
djszapi	Ah-ok	08:45
Aides	xide, get a life	08:45
djszapi	Okay, then how can I delete automatically the files from .gitignore rules ?	08:45
	→ DavidKlein joined	08:45
thiago_home	djszapi: parse the rules and execute git rm for each of the files that match	08:46
	← DavidKlein left	08:46
	→ DavidKlein joined	08:46
doener	djszapi: git ls-files -c -i --exclude-standard -z \| xargs -0r git rm --cached	08:47
djszapi	thanks.	08:47
loops	djszapi, git rm $(grep -v '#' .gitignore)	08:48
djszapi	thanks both of them.	08:48
doener	loops: that lacks the --cached part, and doesn't handle !rules	08:48
loops	doener, details details... but you're right of course ;o)	08:48
doener	(and the grep should look for '^#'	08:49
loops	bah... now yer just being pedantic	08:49
djszapi	hehe	08:50
loops	it works on the vast majority of .gitignore's, but there are a few corner cases like you say	08:50
	← dreiss left	08:50
doener	loops: makes a difference for e.g. linux-2.6.git ;-)	08:50
	loops surrenders to reality and admits it was a horrible suggestion	08:51
loops	not to mention somebody could put /* in the .gitignore	08:52
	i want it out of the archives now :oP	08:52
	← xide left	08:53
doener	loops: "fatal: '/*' is outside the repository"	08:53
loops	yeah.. saved by "git rm"	08:53
	← c15c8ra1n left	08:53
loops	i think i'll cut my losses and go fill out that form for xide	08:53
doener	nooooo	08:54
loops	heh	08:54
	doener will be more careful next time	08:54
	← pantsman left	08:54
	← kpreid_ left	08:54
	→ kpreid_ joined	08:54
	→ warthog9 joined	08:55
	← mmarsu left	08:57
	→ ixti joined	09:05
	← TheUni left	09:06
	→ TheUni joined	09:06
djszapi	How can I use git am from command line efficiently ?	09:08
selckin	man git-am	09:08
Gitbot	selckin: the git-am manpage can be found at http://git.or.cz/man/git-am	09:08
djszapi	I use it now from mutt, because the mail filenames isn't reasonable	09:08
	but if I can identify the patch mail efficiently, I'd use git am from command line, if it's possible.	09:08
doener	djszapi: hm? I don't understand that "the mail filesnames isn't reasonable" part	09:09
	djszapi: I usually just tag the mail series in mutt, copy them to a new mbox, and apply that using git am	09:09
	sitaram gets a feeling of deja vu -- someone else was explaining this yesterday, I dont recall to whom	09:10
	← hyperair left	09:11
	→ hyperair joined	09:11
	doener wasn't responsible for either side, and wouldn't remember anyway	09:11
	← hyperair left	09:12
djszapi	doener: why do you copy then ?	09:12
	If you're in mutt you can presh '\|'	09:12
	and then git am -s	09:12
	I'd like to be totally independent from any mail client.	09:13
	just getmail.	09:13
	and then I'd like to choose easily the mail, patch inside it	09:13
doener	djszapi: because I mostly use "git am -3", and if that leads to conflicts, I have to leave mutt anyway	09:13
djszapi	otherwise if i open a mail client, i can use inside that actions, like in mutt '\|'	09:13
doener	djszapi: repeat the "switch to mutt; apply; conflicts; switch to vim; fix; commit" cycle for ten mails and it's a major pain	09:14
djszapi	you can use after '\|' in mutt git am -s -3 too	09:14
doener	djszapi: any my mutt instance is usually running with $PWD being $HOME, so I'd have to chdir around, too	09:14
djszapi	yeah.	09:14
	→ lhz joined	09:15
djszapi	you use 'cd' command before opening mutt.	09:15
	not ?	09:17
	→ pantsman joined	09:18
cjs_	For some reason, when I specify --trunk, --branches and --tags to git svn, it fails with "Bad URL passed to RA layer: Unrecognized URL scheme for 'trader' at /usr/lib/git-core/git-svn line 745". Has anybody seen this sort of thing before?	09:19
	→ hyperair joined	09:19
cjs_	(I've checked to make sure that the URLs are correct and exist in the repo.)	09:19
	← j416 left	09:21
loops	cjs_, is it an http:// URL ?	09:22
cjs_	No, it's a file:/// URL.	09:22
	However, I did try it with an svn+ssh: URL, as well, and had the same issue.	09:23
	(My URL structure is /client/project/trunk, /client/project/branch/foo, etc., BTW.)	09:23
loops	cjs_, don't think it has to do with the structure, others have reported this on the list, but i don't see a good answer :o(	09:24
cjs_	Oh dear. That kind of puts a crimp in my plans to convert from subversion....	09:24
	← hyperair left	09:25
	→ hyperair joined	09:25
loops	cjs_, well i'm sure there is a way to resolve it, if mugwump doesn't come on and answer your question here, your best bet is asking on the mailing list	09:25
doener	cjs_: which git version?	09:26
loops	doener, good thought	09:26
cjs_	version 1.6.0.4	09:27
	BTW, is svn2git obsoleted by git-svn?	09:27
doener	cjs_: there were IIRC at least two fixes for such errors since then	09:27
	→ Octalot joined	09:28
	→ travisjeffery joined	09:28
doener	cjs_: though the "url scheme" part seems new to me, I'd still try a more recent git version	09:28
	→ d0k joined	09:28
cjs_	Hm. What's involved in trying out a newer version on an Ubuntu system?	09:28
loops	cjs_, no, they're separate projects	09:28
doener	cjs_: svn2git warps git-svn	09:28
	s/warps/wraps/	09:29
cjs_	It's more than just grabbing a single binary, isn't it?	09:29
doener	cjs_: there should be backports available	09:29
	← jfxberns left	09:31
	→ rudi_s joined	09:32
	← davido left	09:34
cjs_	Ah, found one. Thanks.	09:35
	→ jfkw joined	09:35
cjs_	This is 1.6.5.1.	09:35
doener	which is the most recent release	09:36
cjs_	Same issue, though.	09:38
	→ Axius joined	09:38
cjs_	Are there any workarounds? I've got only about four branches I need to import, really, at least in this one project.	09:38
doener	cjs_: could you pastebin the output you got?	09:38
loops	cjs_, what version of svn ?	09:39
	cjs_, can't say i know it matters, but am curious	09:40
cjs_	svn 1.5.4, and the only output was "Bad URL passed to RA layer: Unrecognized URL scheme for 'trader' at /usr/lib/git-core/git-svn line 940"	09:40
loops	cjs_, what command line did you use ?	09:40
cjs_	Ok, I've possibly found the issue. I was using --trunk=xxx --branch=xxx --tags=xxx projname, where the xxx was the full URL. If I give just "trunk" "branch" and "tag" as the xxx arguments (i.e., relative), and give the full URL (without /trunk appended) before the projname (the output dir name), it runs, though I've not verified that it actually does the right thing.	09:45
doener	yeah, the --branches etc. options expect paths relative to $url	09:45
cjs_	Ah, well, the docs claim you can use full URLs.	09:46
	And you can, if you use just --trunk.	09:46
	← kpreid_ left	09:47
	→ kpreid__ joined	09:47
	← Axius left	09:47
doener	cjs_: ah, you were using it without the full url as argument	09:47
	cjs_: never tried that, didn't expect that to work either	09:48
cjs_	Any thoughts on how I might check to see if it's doing the right thing before it's done? It's a rather large repo....	09:51
Mikachu	you can run gitk in another terminal at any point during the import	09:51
	→ Tuomas joined	09:52
cjs_	Ah. Actually, git branch -l should show me the branches it's bringing in, shouldn't it? But we're a ways in, and no branches, just the trunk so far. It probably does them after the trunk, though.	09:52
Mikachu	yeah	09:52
	→ mikem` joined	09:53
cjs_	It would rather want to, when you think about it. Hm. Well, perhaps I'll go grab some dinner and see where we are when I come back.	09:53
Mikachu	well, -a, not -l	09:53
	-l means create reflog :)	09:53
cjs_	Oops, yeah, git tag -l.	09:54
	→ abms1116 joined	09:54
mikem`	hi, in the git-svn manpage, the blurb about the --prefix option says, near the end: "Setting a prefix is useful if you wish to track multiple projects that share a common repository." -- could someone elaborate on this? How does this work? I'd like to track multiple projects that are living in the same repo, each in its own dir which contains its set of trunk/, branches/ and tags/ dirs.	09:54
cjs_	So I'm not clear on what svn2git does that git svn doesn't.	09:54
Mikachu	they are just different programs	09:54
doener	cjs_: It tries to add things like "convert svn tags to git tags", by being a wrapper around git-svn	09:55
cjs_	mikem`: You arrived at a very appropos time for that question. :-)	09:55
	→ mbuf joined	09:56
mikem`	haha so it seems	09:56
doener	cjs_: but it doesn't do that in exactly the smartest/fastest/most correct way	09:56
cjs_	By adding the --tags argument to the git-svn command line?	09:56
mbuf	when there is a conflict during a merge, I manually fix the changes, and then I should just add the modified file, and commit?	09:56
doener	cjs_: no. That option make git-svn create remote tracking branches for the svn tag	09:56
	s	09:56
	cjs_: as tags in svn are just branches, except for convention	09:56
	cjs_: so to play safe, git-svn treats them like branches	09:57
Mikachu	mbuf: yes	09:57
cjs_	Ah, I see. So I'll actually get out branches, not tags, for svn "tags" pulled in via --tags.	09:57
doener	cjs_: btw, we're talking about this svn2git, right? http://github.com/jcoglan/svn2git	09:57
mbuf	Mikachu, thanks!	09:57
cjs_	doener: yeah, that's the one I was looking at.	09:57
	I'm not sure that the other web pages I looked at were talking about that one.	09:58
doener	cjs_: ok. That has e.g. fix_tags here: http://github.com/jcoglan/svn2git/blob/master/lib/svn2git/migration.rb	09:58
cjs_	mikem`: So what exactly are you trying to do?	09:58
doener	cjs_: that converts the tag-related remote tracking branches into real git tag	09:58
	s	09:58
	→ BurmaSauce joined	09:58
doener	cjs_: but 99,99% of the time, that straight conversion is _not_ what you want	09:59
	cjs_: and it wastes a lot of time by doing "git checkout $foo" and then tagging what HEAD resolves to, instead of directly tagging $foo	09:59
mikem`	I have an SVN repo, with a layout like this: svn://server/projA/{trunk,branches,tags}, svn://server/projB/{trunk,branches,tags} etc. I'm not sure how to track it with git. the git-svn manpage says something about --prefix and multiple projects in the same repo, but I'm not quite clear on how exactly it works	10:00
	→ erikvold joined	10:01
doener	cjs_: http://git.pastebin.com/m6d587724	10:02
mikem`	ah, I think i get it. I'd need to do a separate `git svn clone --prefix projA/` for each project in the repo. I'd end up with as many git repos as there are projects in the SVN repo	10:02
doener	cjs_: you want the tag to reference the original commit C, not the bogus commit T that svn had to create	10:02
	→ Axius joined	10:03
doener	cjs_: And if someone was smart enough to change the tag in svn, by committing to it, you likely want some more surgery to get a meaningful history that doesn't have the tag living somewhere outside the "real" branch	10:04
cjs_	Hm. I'm not sure I see how much difference that makes, though if you trust the svn repo to use the convention correctly, you might want to drop the "tag branch" entirely.	10:04
	Fortunately I don't use tags much.	10:05
	← travisjeffery left	10:05
	→ therrg joined	10:05
doener	cjs_: yeah, the tags/1.0.0 remote tracking branch can be dropped later	10:05
	cjs_: and having the tag referencing C instead of T makes a big difference if you use e.g. "git describe"	10:06
	cjs_: "git describe E" could tell you that E is a descendant of 1.0.0, being two commits ahead of it. But only if the 1.0.0 tag references C	10:06
	← _Mica1 left	10:06
doener	cjs_: if it references T, it's outside of E's history, and thus not available to "git describe"	10:07
	← hyperair left	10:07
cjs_	Ah!	10:07
	← BurmaSauce left	10:08
	→ hyperair joined	10:10
	← vital left	10:10
	→ vital joined	10:11
	← erikvold left	10:11
	→ jfxberns joined	10:15
	→ vvsh joined	10:23
	→ patrikf joined	10:24
	← Axius left	10:24
	← therrg left	10:25
	→ therrg joined	10:26
	→ davido joined	10:28
	→ akitada joined	10:28
	← abms1116 left	10:29
	→ stamina joined	10:31
	→ Yuuhi joined	10:32
	← GodEater left	10:33
	→ GodEater joined	10:33
	→ __iron joined	10:34
akitada	Can git-svn handle more than one repository at the same time? One for getting changes and another for saving the changes?	10:34
	Getting changes in svn repo is easy. The hard part of this plan would be exporting the changes from the git repository and saving it to svn.	10:38
	← kpreid__ left	10:40
	→ kpreid_ joined	10:40
pantsman	I'd use 2 git repos, one cloned from each SVN repo	10:42
	and pull from one repo to the other before dcommit	10:42
Ilari	That doesn't work very well...	10:42
	Maybe format-patch and am to move them as patches might work...	10:43
	→ Grahack joined	10:43
akitada	Ilari: yes, manual patch work seemed to be the only way to accomplish this...	10:44
	Looking the changes occured in a repo, study easy commit and decide which patch is I wanted to include in my repository and possibly make some adjustments to the commit for my own repo and apply the modified cahnges into my own repo, etc etc...	10:47
	Ilari: this is what you suggested for me, correct?	10:47
	and I'd also need another manual patching work for exporting changes in git repo into svn.	10:49
Ilari	akitada: Well, format-patch and am are for mass patch extraction and mass patching.	10:52
	→ trochala joined	10:52
Mikachu	you can't use rewriteRoot for that?	10:53
akitada	Ilari: Let me check the format-patch manual.	10:55
	→ Blackshark joined	10:56
akitada	and git-am	10:57
charon	akitada: if you import the source svn with --no-metadata, you can just rebase the commits onto the destination svn	10:58
	the --no-metadata is important because the (wrong) git-svn-id lines would confuse git-svn as to what should go where	10:58
	(or use gfp\|am, like Ilari says; a noninteractive rebase is just that)	10:59
	→ j416 joined	11:01
	stsch → Tank`	11:03
	← Arild left	11:03
	← ph^ left	11:04
akitada	Ilari: well, it helped much. must admit I was missing big help in git here.	11:05
	charon: gfp?	11:07
charon	format-patch	11:07
akitada	Ah Git-Format-patch. ok	11:08
	← bsdvodsky left	11:10
akitada	charon: as for noMetadata, I suppose it won't work for me, because it's not one shot import but it's continuous.	11:11
charon	akitada: that doesn't matter as long as you only git svn fetch (not git svn rebase) and take care to not drop the "cache"	11:12
	→ jaggz joined	11:12
jaggz	hi	11:12
charon	(in this case it's strictly speaking not a cache as it's not expendable)	11:12
jaggz	do any git users live in Los Angeles and want me to treat them to a coffee or dinner and show me the ins and outs of git usage?	11:13
	I'm an svn user and I am sad.	11:13
	← aliceinwire_ left	11:13
	← priidu left	11:16
	→ corni joined	11:17
	→ Pupeno joined	11:17
akitada	charon: That workflow is something like: git svn fetch; git merge refs/remotes/trunk; git svn dcommit (to destination repo) ?	11:17
Pupeno	Is there a way to get the number of commits so far?	11:17
	← thijso left	11:17
	→ _jason553839 joined	11:17
j416	Pupeno: I always do something like: git log --oneline \| wc -l	11:18
charon	akitada: well, from your description i assumed you wanted to copy over the exact commits, which a merge can't do	11:18
j416	for the current branch	11:18
	but there may be a better way.	11:18
	→ thijso joined	11:18
	← vvsh left	11:18
Pupeno	j416: --oneline! exactly what I wanted. Thanks.	11:18
	→ khmarbaise joined	11:18
j416	:)	11:18
cjs_	Grr...git svn it didn't load the freaking branches.	11:18
charon	akitada: you can use a merge but it will look like a big commit bringing over all changes in the destination svn	11:18
\\steve	a git svn rebase really takes ages on a windows machine	11:21
	mbuf → mbuf\|away	11:21
	→ jwheare joined	11:22
	← kpreid_ left	11:24
	→ kpreid_ joined	11:24
jwheare	i'd like to use git-filter-history to start a new repository based on a single file in another repo	11:24
	in the past i've used subdirectory-filter, but in this case there's other files in the subdirectory i'm not interested in.	11:25
	what's the best way to do this?	11:25
cjs_	Oh, it did get them; I just needed to use "git branch -a" to see them.	11:25
	→ neldoreth joined	11:26
j416	jwheare: you want to keep the history of that single file?	11:26
jwheare	j416: yup	11:26
j416	ok	11:26
	by git-filter-history I presume you mean git-filter-branch	11:27
akitada	charon: so what does that workflow look like?	11:27
jwheare	(i meant git-filter-branch) :)	11:27
j416	and	11:27
	it's not hard at all	11:27
	if you know bash script	11:27
cjs_	Ah, but they're just remote branches? What do I need to do (that svn2git does) to get local copies of these branches into my repo?	11:28
doener	cjs_: to create a branch head, just the usual "git branch <name> <starting-point>"	11:28
j416	jwheare: I have never done that exact same thing, but if it were me I would probably use git filter-branch --tree-filter '<cmd>'	11:28
doener	cjs_: they're remote tracking branches. All the required objects are already local.	11:29
j416	where <cmd> is a bash script that removes everything but that file	11:29
	there may be simpler ways to do it.	11:29
neldoreth	hello, i have a little question: i want to push something from my branch in my remote git and get now the following error: "To prevent you from losing history, non-fast-forward updates were rejected, Merge the remote changes before pushing again. See the 'non-fast forward'" so i do a git pull, but everything is up to date (same when i say git merge origin/my_branch) - can someone give me a hint?	11:29
cjs_	doener: Ah, really?	11:29
j416	also, make the script always return 0.	11:30
	mbuf\|away → mbuf	11:30
cjs_	So if the remote branch is named remote/foo, I can just "git branch foo remote/foo"?	11:30
Ilari	neldoreth: Check what branch was rejected and then check what branch you are on. pull merges into current branch.	11:30
cjs_	I'm not quite clear on what a "remote tracking branch" is.	11:30
j416	possibly --index-filter is faster, I have never used it.	11:30
Ilari	cjs_: Mirror of branch from another repository.	11:30
doener	cjs_: git has at least three basic namespaces for refs	11:31
	cjs_: refs/heads/* for branch heads, refs/tags/* for tags, and refs/remotes/* for remote tracking branches	11:31
	cjs_: remote tracking branches are (as Ilari said) mirrors of the branch heads of a remote repository, they track a branch in the remote repo, hence the name	11:32
charon	akitada: git svn fetch -Rsource; git checkout -b tmp source/trunk; git rebase --onto destination/trunk previously_ported_over; git svn dcommit -Rdestination; git branch -f previously_ported_over source/trunk	11:32
	oh, and delete the 'tmp' branch afterwards ;)	11:32
doener	cjs_: what "git branch [-r\|-a]" shows, are just the shortnames for the [remote tracking] branches	11:33
neldoreth	Ilari: ah, right - thanks!	11:33
	← brizly left	11:33
doener	cjs_: so if "git branch" shows a branch head "master", its full name is refs/heads/master	11:33
Ilari	neldoreth: Hint: Its possible to make command prompt display name of current branch.	11:33
	← kpreid_ left	11:33
charon	akitada: untested though. the idea is, you use a rebase to "copy over" all new commits from 'source' over to be based on 'destination/trunk' so that you can dcommit there	11:33
	→ kpreid_ joined	11:33
cjs_	Ah, so when I pull from the other repo, those branches will be updated as well?	11:33
Ilari	neldoreth: (install git command completion and configure command prompt).	11:33
neldoreth	Ilari: it does, i am using the zsh with a specific config, but i totally missed that	11:34
doener	cjs_: when you fetch. "pull" is a convenience command that wraps fetch and merge/rebase	11:34
charon	akitada: and as i said, it's really important that those commits do not have any git-svn-id lines, so you must configure 'source' with --no-metadata	11:34
doener	cjs_: by default, names of remote tracking branches have the form refs/remotes/<remote>/<branch_head>	11:34
	→ brizly joined	11:35
	→ aresnick joined	11:35
doener	cjs_: e.g. the "master" branch head of the remote "origin" is tracked as refs/remotes/origin/master	11:35
	cjs_: unfortunately, the default for git-svn is to not have that <remote> infix	11:35
	cjs_: that a) pollutes the namespace and b) can easily make things ambiguous, as git allows those short names to be used	11:36
cjs_	Ah, ok. So basically, I can branch from a remote branch, hack away, and my changes will go back to that branch when I push them back.	11:36
doener	cjs_: from a remote tracking branch (or just "tracking branch", but that name usually leads to confusion, thanks to the --track option to git branch/checkout)	11:37
	cjs_: a remote branch is the actual thing in the remote repo	11:37
	cjs_: i.e. you fetch from a remote, getting mirrors of the remote branches locally as remote tracking branches	11:38
akitada	charon: cool but harder than I hoped. maybe I forgot to mention the src repo is not 100% same with project tracked on destination repo. it's a fork of src repo. so I think it's gfp & am with manual adjustments for my goal it seems.	11:39
	← _jason553839 left	11:40
j416	I'm reading in the mercurial document for git users, that if a tag is created with -a in git (to make it an object), it will be global instead of local	11:40
	I find nothing about this in man git-tag	11:40
Gitbot	j416: the git-tag manpage can be found at http://git.or.cz/man/git-tag	11:40
	→ psoo joined	11:40
doener	j416: because that's, uhm, inaccurate...	11:40
cjs_	So if I'm going to use this as my new master repo, and I want clones of it to be able to access those branches, what do I need to do?	11:40
j416	doener: thank you for confirming. :)	11:40
	so I though	11:40
	thought*	11:40
charon	akitada: sure, there are several ways to do the rebase :) e.g., you can also use an interactive rebase to rearrange stuff, or use a merge strategy if required	11:40
doener	j416: "git tag -a" creates an annotated tag, i.e. a tag object, with message, tagger, date etc.	11:41
j416	tags will always be local until you specifically choose to pull or push them, right?	11:41
	yes	11:41
Mikachu	fetch will fetch tags that are in the updated range	11:41
j416	and just "git tag" is more similar to a symbolic ref, right?	11:41
doener	j416: while just "git tag" creates just a ref that directly references a commit, no tag object created, thus no message etc.	11:41
jwheare	j416: thanks for the advice	11:41
	→ bentob0x joined	11:41
doener	j416: no! A symbolic ref references another ref. That would likely be a moving target	11:42
j416	jwheare: be careful with merges if you have those, I have never run filter-branch with merges. Maybe it works fine, I don't know :)	11:42
	oh	11:42
	thanks :)	11:42
	ok so, a head then	11:42
Ilari	Filter-branch works fine with merges since it does filtering commit by commit.	11:42
j416	Ilari: ok :)	11:42
	neat	11:42
doener	j416: a ref. By "head" one usually means a branch head, i.e. a ref in the refs/heads/ namespace	11:42
j416	ok!	11:43
charon	akitada: but in general, toying with svn histories is easy if you always remember the basic rule: the "target" SVN is always the one referenced in the first git-svn-id reachable through first-parent history	11:43
j416	but technically they're the same thing, right?	11:43
	(apart from the name and storage location)	11:43
doener	j416: "git tag" always creates a ref in the refs/tags/ namespace. Without -a, the ref will directly reference the object that is to be tagged	11:43
j416	yep, my thought. Thanks	11:43
doener	j416: with "-a", a tag object is created, that has some extra data and references the to be tagged object. And the ref will reference that tag object, instead of the object to be tagged	11:44
j416	yeah	11:44
doener	you can transport either into different repos using fetch/push. But _usually_ the light-weight tags are used locally, and public tags (like for releases) are annotated	11:45
	but there's no technical reason that would stop you from pushing a lightweight tag	11:45
	← GodEater left	11:45
akitada	charon: it's impressive you know much about git-svn.	11:46
charon	akitada: well, not enough anyway, but i used it during an svn-to-svn transition to keep the two repos in sync for a few months :)	11:46
akitada	charon: for my case, it's not a sync, but more like fork. src repo is not what I manage and I don't have the commit right. git repo and destination svn are what I manage and can commit anything I want.	11:48
	← davido left	11:49
	→ davido joined	11:49
	← psoo left	11:49
doener	cjs_: http://git.pastebin.com/m5451bb1c	11:50
akitada	charon: so git and dst svn include some changes in src svn (not all) and my haking work. I even delete some of the files in my repos so simply importing changes in src svn should fail.	11:50
	→ babilen joined	11:50
doener	cjs_: so you need to create local branch heads for everything that people cloning your repo should have (unless you want them to resort to manually adjust the configured refspec ;-))	11:51
pielgrzym	hi there :) when I push into a non-bare repository the changes remain uncommited - how to change this?	11:52
bremner	pielgrzym: faq non-bare	11:53
Gitbot	pielgrzym: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare	11:53
akitada	charon: so it'd be a little harder than your case, unfortunately.	11:53
pielgrzym	bremner: just realized it would be cooler to have a crossrepo between production eviorment and testing enviorment - not to push direclty into production server :)	11:54
	→ corni_ joined	11:55
bremner	yes. and that crossrepo should be bare	11:56
pielgrzym	havin a bare repo - how can I set the default branch to other than master?	11:56
thiago_home	write to HEAD	11:56
	→ psoo joined	11:56
pielgrzym	thiago_home: thanks!	11:57
	← doener left	11:58
	→ doener joined	11:58
bremner	I think git-symbolic-ref can do it too, if you don't like vi	11:59
cjs_	doener: Ah, so I was right to create those local branch heads. And should I decide to pull further changes on those from the svn repo, I'll need to rebase my local branches, as well as my master, right?	12:02
doener	cjs_: yeah. But that rewrites history, and quickly becomes a pain if you share that repo via git	12:03
	← corni left	12:03
doener	here's a short setup guide for a svn one-way mirror: http://git.or.cz/gitwiki/GitFaq#HowdoImirroraSVNrepositorytogit.3F	12:04
	→ aspotashev joined	12:04
cjs_	Oh, if I were going to pull further stuff, I'd make it two-way, and dcommit any changes coming in from the git side.	12:04
doener	cjs_: that's hairy, thus the setup is one-way, you should put any "not yet in svn" stuff in there	12:05
	cjs_: using git-svn makes you prone to a whole set of svn shortcomings if you dcommit	12:05
	→ malumalu joined	12:07
cjs_	Hm. I'm confused now. Is using a clone of an svn clone going to be more of an issue than just using an svn clone repo?	12:08
charon	akitada: yeah. i just used merges as i only had to work around a rename, and sufficiently high rename limits worked well	12:08
	→ BurmaSauce_ joined	12:08
doener	cjs_: yes, if there are any "not in svn" things in the intermediate repo	12:09
	cjs_: with such commits, you have to rebase, which rewrites history, and forces the downstream repo to deal with that and rewrite its history as well	12:09
cjs_	Ah, but I'll be ok if I'm not making changes in the svn repo, right?	12:11
doener	cjs_: if you don't fetch from the svn repo anymore, it's basically out of the equation, so you're git-only	12:11
cjs_	I think I see what you mean now, and I can probably handle that issue. I'm not looking for true interoperation; I'm looking to be able to switch back to svn should a client decide he just can't stand git.	12:11
doener	cjs_: well, and don't dcommit either	12:11
	cjs_: basically, as long as you interact with the svn repo, you have a central repo there, even with git-svn.	12:12
	cjs_: What git-svn offers are primarily local commits then.	12:12
	cjs_: some "advanced" workflows (that still have quite some flaws), may allow you to do a bit more, but often, it's not worth it	12:12
	cjs_: note that I'm not trying to say that you may not dcommit at all, you should just not share anything that's not in the svn repo, unless you know what you're doing	12:13
	← j416 left	12:14
	← BurmaSauce_ left	12:14
doener	cjs_: if you have that central svn mirror, others can do their initial clone from there, and from then on use "svn fetch/rebase/dcommit", ignoring the mirror	12:14
	cjs_: that way, you share things through the svn repo	12:14
	cjs_: and that works. But sharing things via "git fetch/push" is kinda dangerous	12:15
	→ _jason553839 joined	12:15
cjs_	Hm. That's a thought, but I'm actually trying to get rid of svn here. The only real question is, can I work in git for a few weeks, and bring everything back into svn if it doesn't work out for a project.	12:15
doener	cjs_: as long as you keep your history linear (no merging), that should work	12:16
akitada	charon: to be honest I still don't get it well on how I could workaround the situation where i'm currently on but I now know I need to be more familiar with git-format-patch, git-am and git-svn. that's better than not knowing where to start :p	12:16
doener	cjs_: honestly, git-svn is a very nice bandaid if you're forced to work with a svn repo	12:16
	cjs_: but if you want to look into real git usage, you're probably better off with something where all the (git-)svn restrictions don't affect you	12:16
	← therrg left	12:17
cjs_	doener: So, basically, get everybody rebasing from a central git repo before they push to it?	12:17
	→ Sonderblade joined	12:18
doener	cjs_: if you won't dcommit (or only after you declared the git attempt as a failure), then yes, that would likely work.	12:18
djszapi	How can I view about it a commit in which branch is it ?	12:19
doener	cjs_: i.e. you create a linear history with git and don't interact with the svn repo, at all, for the time being. Then you should be able to dcommit back to svn later (if you decide to drop the git repos, they'll need a fixup otherwise)	12:20
thiago_home	djszapi: can you rephrase?	12:20
djszapi	yep	12:20
	git show e77336d9a60dc9e0363569cdc3c9630aa24bd6e5	12:20
Gitbot	[git e77336d9a]: http://tinyurl.com/ylqdfpj -- send-email: confirm on empty mail subjects	12:20
	doener guesses "git branch --contains <committish>"	12:20
djszapi	e.g. I would like to get information about this commit, in which branch is it	12:20
	e.g.	12:20
thiago_home	that's what I'm guessing to	12:20
cjs_	doener: Right, that's easy enough to do, since it fits in with the svn working style we've been doing all along.	12:20
thiago_home	yeah, git branch --contains	12:20
djszapi	the output is empty prompt	12:21
	without any informations.	12:21
thiago_home	then it's in none of your local branches	12:21
djszapi	Do i need to clone first of all the branch ?	12:21
thiago_home	try git branch -a --contains	12:21
djszapi	remotes/origin/{next,pu}	12:22
doener	cjs_: it's a quite limited workflow for git though ;-)	12:22
djszapi	How can I get describtion about a branch, why is it established and what does it's name cover ?	12:22
	← g0bl1n left	12:22
thiago_home	djszapi: post to the mailing list	12:23
djszapi	it's a faq I think.	12:24
	→ BurmaSauce_ joined	12:24
djszapi	It's needed to be desribed somewhere.	12:24
	← BurmaSauce_ left	12:24
	doener didn't understand the question, especially the second part	12:25
djszapi	doener: what does it mean e.g. next/pu branch ?	12:25
	and why was it established/created ?	12:25
cjs_	doener: Limited workflow for a little while is not a problem. It still gives us everything we get from svn, and once everybody's comfortable, we can declare we're not using svn any more, and move on to using other useful git features.	12:25
	→ tbuehlmann joined	12:26
	← mikem` left	12:27
doener	cjs_: it's just more painful to rebase all the time. And it probably also leads to keeping the svn "I need to update 50 times a day" syndrome	12:27
	cjs_: anyway, good luck :-) Have to go	12:27
	djszapi: there's no "describe a ref" mechanism in git. Such things are usually in READMEs or something like that	12:28
	djszapi: for git, google "A note from the maintainer"	12:28
	→ gitte joined	12:28
cjs_	doener: We very much like the "update 50 times a day" thing; it keeps us honest, and keeps merge costs down.	12:29
	Basically, in an agile world, you want everybody working on something that's as close to production, or what's about to be rolled out to production, as possible.	12:30
	→ jstephenson joined	12:30
	→ ciskje joined	12:31
	← davido left	12:31
	← jstephenson left	12:31
doener	dunno, I rather focus on getting my 5-10 commits done to finish a thing, and then merge it once. Instead of getting interrupted over and over again	12:32
	→ jstephenson joined	12:32
djszapi	doener, thiago_home: thanks	12:32
bremner	cjs_: one thing to keep in mind is that merging works better in git than in svn	12:37
	or so they tell me. I never really used branches in svn	12:37
cjs_	bremner: Sure, but even so, working against code that no longer exists is a waste of time.	12:37
bremner	cjs_: I'm not here to manage your project for you :)	12:38
cjs_	I understand that. :-)	12:39
	← _jason553839 left	12:40
cjs_	Basically, there are certain parts of git that are designed for what I would call "slow-moving" projects, with groups that don't talk to each other a lot (i.e., constantly because they're usually all working together in the same room) and don't realize very often (i.e., not several production releases per day at times).	12:40
	So I just don't worry too much about the features that solve problems I don't have.	12:40
bremner	s/git/version control/	12:40
cjs_	Hm? Well, I was talking about some of the stuff that git makes easy that's harder in svn. Such as doing large merges.	12:42
bremner	hey, whatever works. Just be aware there are many possibilities with git that were not really there with svn. In particular, I notice really thinking about the commit history as output rather than one piece of code. I think this has important maintenence consequences	12:42
cjs_	Can you elaborate on that?	12:43
Ilari	Get large enough merges (and not just large enough in file count) and the cost of resolving conflicts dwarfs the initial command cost. I have seen merge that took 5 days to resolve.	12:43
bremner	well, by working on a sequence of commits locally for a bit, I am more able to make them into logical chunks, to support e.g. debugging by bisecting	12:44
cjs_	llari: exactly.	12:44
	→ brain0 joined	12:44
cjs_	bremner: Ah, yes, I've noticed that a bit.	12:44
Ilari	But often such merges are sign of bad workflow / application organization. Most merges should be easy to resolve.	12:46
	→ jaggz- joined	12:46
	→ _Mica_ joined	12:46
cjs_	llari: Well, I dunno; I frequently do architecture changes that touch many parts of the system. If you think of having to do those as the result of bad application organization, well, yes, that's exactly why the changes were made. :-)	12:47
	→ rudi_s_ joined	12:48
	← ixti left	12:48
	→ w0rmie joined	12:48
	← w0rmie left	12:48
	→ w0rmie joined	12:49
selckin	what kind of merge takes 5 days?	12:51
cjs_	BTW, anybody here use gitosis? Can one have repos in subdirectories of the repositories directory?	12:51
selckin	2 teams not talking to each other for a month and rewriting the entire thing?	12:51
sitaram	cjs_: I believe so; never tried it though	12:51
Ilari	Ah yeah. Agile tends to use "we can refactor this later" over "design up front". Of course, Waterfall is straw man model.	12:51
selckin	but doesn't that aslo imply you merge alot and thus keep the sizes down?	12:52
	→ corni__ joined	12:52
cjs_	llari: Agile does as much design up front as it can. And those who do a lot more design up front usually tend to get it wrong anyway. You're far safer being able to change your design than to hope you get it right the first time, no matter how much time you spend on it the first time.	12:53
	And besides, those big changes are cheap, so who cares?	12:53
	→ tazle_ joined	12:53
Ilari	selckin: 3 months of changes on one side (history in Git). 2 years of changes on other side (no known version control)... That can't end wel...	12:53
	And its pretty telling that there was 3 months of changes on one side and ~2years on the other...	12:54
	← rudi_s left	12:55
Ilari	cjs_: I would rather get the design quickly almost right. Big changes to programs aren't cheap...	12:56
	→ psoo_ joined	12:57
	← w0rmie left	12:58
	← Grahack left	12:59
cjs_	llari: So would anyone. Agile doesn't mean "don't design." It just means, once you've gotten as far as you can with the design, based on how much you know right now, start implementing rather than just going around in circles.	13:01
	→ aigon joined	13:03
	← JasonWoof left	13:04
	mbuf → mbuf\|away	13:05
	→ Bass10 joined	13:06
	→ bremner_ joined	13:06
	→ Akufen joined	13:06
	← Phurl left	13:08
	→ Phurl joined	13:08
	← corni_ left	13:08
	← babilen left	13:09
	← hummingkid left	13:10
	← psoo left	13:12
	→ yaotti joined	13:12
	← Akufen left	13:15
	← _Mica_ left	13:18
	→ MisterN joined	13:19
	→ onigiri joined	13:20
	← trochala left	13:20
neunon	Is there a way to make git always ignore changes for a specific file (and not commit them with git commit -a), unless the changes are added to the index via 'git add -f'?	13:20
selckin	gitignore works like that i think	13:21
Ilari	Except that gitignore is ignored for tracked files.	13:22
neunon	exactly.	13:22
	→ jtdowney joined	13:22
djszapi	How do I rewrite the author of commit too, git commit --amend can't handle it.	13:22
Rhonda	djszapi: It can, it has an --author option	13:23
djszapi	cool	13:23
	← aigon left	13:25
djszapi	and how can I rewrite a commit e.g. author, in the HEAD^^ commit ?	13:25
	→ alley_cat joined	13:25
djszapi	Can I do it only in the last one ?	13:25
	git cherry .. ?	13:25
gitte	djszapi: rebase -i	13:25
	djszapi: git cherry is unfortunately a pretty useless command, taking a nice and short name from more useful things.	13:26
	← brain0 left	13:26
gitte	djszapi: with rebase -i, you can rebase, reorder and modify the latest <n> commits.	13:26
	djszapi: in your case, do "git rebase -i HEAD~2", which will show you the latest two commits in an editor.	13:26
	djszapi: replace the first "pick" with an "edit".	13:27
djszapi	nope	13:27
	http://pastie.org/659443	13:27
gitte	djszapi: then write and exit, Git will stop after re-applying the 2nd latest commit.	13:27
bremner_	Ilari: Isnt there some gitattibutes trick to ignore changes to a tracked file?	13:28
djszapi	git reset ?	13:28
	gitte: I can't understand	13:28
	because i didn't get editor at all.	13:28
gitte	djszapi: you want to edit a commit when you still have uncommitted changes? That is not the safe way, so djszapi does not allow you to do it.	13:28
	djszapi: you have local changes!	13:28
Ilari	There's assume-unchanged, but it has its own problems...	13:28
	← aresnick left	13:28
djszapi	I need to use git stash	13:29
	until I edit the 2nd latest commit	13:29
	that's the best way, right ?	13:29
gitte	djszapi: or you need to commit them first. That would be the best way.	13:29
djszapi	It's a not working tree	13:29
	that I've got now, in this momment	13:29
	→ _Mica1 joined	13:30
djszapi	Is git stash safe and good in such a situation ?	13:30
Rhonda	Might be.	13:31
	→ ab3 joined	13:33
	→ schmukkk joined	13:34
	← relix left	13:35
	← jwheare left	13:36
	→ babilen joined	13:37
	← yaotti left	13:38
	→ Sho_ joined	13:39
akitada	charon: found your method is introduced in google code http://google-opensource.blogspot.com/2008/05/export-git-project-to-google-code.html	13:42
	→ parasti joined	13:43
	→ Xeross joined	13:43
Xeross	How can I undo an old commit (50 commits old or so)	13:43
gitte	akitada: it is unfortunate that Google Code blessed an inferior method and touted it as a "solution". But using git-svn to "export" Git repositories just loses too much information to be useful.	13:45
	Xeross: git revert <commit>	13:45
	→ jrmuizel joined	13:46
akitada	gitte: could you show me superior method than git-svn? I'm looking for good export method now.	13:47
gitte	akitada: no, there is no good method.	13:47
selckin	use mercurial if you really want to use google code	13:47
akitada	selckin: I don't use google code, just need to export a git repo into svn.	13:48
	gitte: even inferior method would be better than nothing :)	13:50
	→ BurmaSauce_ joined	13:51
	← BurmaSauce_ left	13:52
	akitada wish svn will learn git-am someday.	13:52
	→ Tichodroma joined	13:52
	← bentob0x left	13:52
Tichodroma	How do I restrict git log to commits that are not in some specified branch?	13:54
djszapi	gitte: I did that pick -> edit and I rebased but the 1st last commit was lost.	13:54
	→ aspotashev_ joined	13:54
djszapi	how can I take back ?	13:55
	→ Akufen joined	13:56
bremner_	akitada: Do you have merges in your history?	13:56
akitada	bremner_: yes.	13:56
bremner_	Sucks to be you	13:57
	← jfxberns left	13:57
	→ corni_ joined	13:59
akitada	bremner_: because it can't be expressed in svn?	13:59
bremner_	Afaik correct	14:00
Gitzilla	gitte: Did the last pic show up OK?	14:00
gitte	djszapi: unfortunately, Git's user interface is not intuitive and relies on the user to read the output. So maybe you missed the fact that you have to finish the rebase by "git rebase --continue".	14:00
	Gitzilla: perfect, yes. Thanks!	14:00
	← corni__ left	14:00
akitada	bremner_: I'll try to keep it flat	14:01
djszapi	gitte: Did i lost the last commit ? :(	14:02
gitte	djszapi: you tell me.	14:02
	djszapi: you can find out by looking at "gitk <branch>".	14:02
djszapi	yep, but if I use git rebase --continue it couldn't happen ?	14:02
gitte	djszapi: if your rebase was not finished yet, that will show the old history, and "git branch" will show you that you are on "(no branch)", which is Git's way to say that it is on an unnamed branch.	14:03
djszapi	* (no branch) <- yep	14:05
	→ ivenkys_ joined	14:07
	← aspotashev left	14:07
djszapi	gitte: cool, thanks	14:08
Tichodroma	anyone?	14:09
djszapi	anyone	14:09
	← Xeross left	14:09
gitte	djszapi: it would be good if you raised awareness of Git's user-unfriendliness with the core developers, because I gave up.	14:09
djszapi	hm ?	14:10
gitte	djszapi: did you not have enough problems trying to use Git that you had to ask in this channel? If the user interface of Git was better (i.e. done by somebody who knows something about user interface design), you would not have had to.	14:11
	→ mrlemao joined	14:12
	→ LiamH joined	14:12
djszapi	not the userinterface is the problem for me.	14:13
	the short of git knowledge/experience, I like use command-line, I mainly use command line for my purposes, so it's okay for me absolutely.	14:14
	→ jaysern joined	14:14
	← EricInBNE left	14:14
	← mrlemao left	14:15
	Tank` → stsch	14:17
gitte	djszapi: oh well, then people have to continue to help in this channel.	14:17
	mbuf\|away → mbuf	14:17
	← gitte left	14:17
Tichodroma	gitte: yeah I also disagree. It's very well documented and the user interface is fine. But It's a complicated technology hence the documentation is a lot to take in.	14:17
	So asking for help here is very useful.	14:17
	→ aigon joined	14:18
	→ corni__ joined	14:18
	← aigon left	14:19
djszapi	gitte, doener, loops @Ilari are all very helpful, sorry for them i didn't mention	14:19
	→ aigon joined	14:19
Gitzilla	The technology is not complicated; the concepts are different than what most SCM users are used to; the documentation could use a lot of work; the UI is inconsistent.	14:19
	← ivenkys left	14:20
	← jrmuizel left	14:20
	mbuf → ajnr	14:20
	ajnr → mbuf	14:20
	← corni_ left	14:21
	→ gunmuc joined	14:24
	→ vvsh joined	14:29
	← Tichodroma left	14:30
	→ _Mica_ joined	14:31
	→ surial joined	14:32
surial	So, I killed some branches on github via git push origin :branchName...	14:32
	but if someone else, who had those branches before, does a git pull or git fetch, 'git branch -a' still lists these branches.	14:33
	How do you fix that?	14:33
	→ nape joined	14:33
	← bremner_ left	14:34
Aides	does git branch -r list them?	14:35
	→ kukks joined	14:35
bremner	surial: man git-remote, look for prune	14:36
Gitbot	surial: the git-remote manpage can be found at http://git.or.cz/man/git-remote	14:36
	→ kampasky joined	14:37
	← aigon left	14:38
	→ hummingkid joined	14:39
surial	bremner: Thanks :) Just what I was looking for.	14:39
	Is there any easy way I can check why my git repository is relatively large?	14:39
	I used to check in some deliverables, and am no longer doing that. I can live with removing them retroactively, though, as it's been pushed out to a remote repo, I guess that's not a good plan.	14:40
	→ Axius joined	14:40
	→ mfilipe joined	14:41
	← parasti left	14:42
	→ ianw joined	14:43
	← madewokherd left	14:43
	→ j416 joined	14:44
	← _Mica1 left	14:44
	→ dirk2 joined	14:44
	→ icwiener joined	14:44
ianw	Does this make sense if I want to remove all of my merge commits from master: git fetch; git rebase origin/master master	14:45
	→ yaotti joined	14:46
	→ ianw_ joined	14:47
	← ianw left	14:47
	ianw_ → ianw	14:47
	← nape left	14:48
sitaram	should work, I think...	14:48
	→ Grahack joined	14:49
sitaram	but if any of your merges had conflict resolutions etc I'm not sure what happens	14:49
	→ bdimcheff joined	14:50
	→ jfxberns joined	14:54
bremner	surial: basically it is like you say; if you can use git-filter-branch, but that will mess up people who already pulled your repo	14:56
	→ Mezner joined	14:56
	← mbuf left	14:58
doener	ianw: all merge commits that aren't also in origin/master	14:59
	→ rafpaf` joined	14:59
doener	sitaram: he'll most likely have to solve them again, unless he has rerere enabled, and hits the same conflicts again. Then rerere should be able to automatically fix them	15:00
	sitaram: and of course, he might encounter entirely new conflicts	15:00
	→ mebus_ joined	15:01
	← j416 left	15:01
	→ madewokherd joined	15:01
doener	surial: http://git.pastebin.com/m5eb2a44b	15:01
	→ ilteris__ joined	15:02
doener	surial: you can save that as, for example, "git-find-large" somewhere in $PATH, make it executable, and then do "git find-large 20", to find the top 20 large objects in your repo	15:02
	→ corni_ joined	15:03
doener	surial: it only considers objects that are required for any ref. Objects that are prunable (either immediately or after a reflog entry expired) are ignored	15:03
surial	nice.	15:03
	those prunables go away if I 'git gc', right?	15:04
	← rafpaf left	15:04
doener	"git gc" will drop objects that are immediately prunable (not reachable through any ref, nor through the reflog), if they're older than 2 weeks (default)	15:04
	with "git gc --prune=0" it will prune all objects that aren't reachable through any ref nor through the reflogs	15:05
	and to expire your reflogs (making them empty, might not exactly be what you want), you can do "git reflog expire --expire=0 --all" (or instead of --all, maybe --branches or a specific branch head, whatever you need)	15:06
	if you know exactly which reflog entry is "bad", you can also drop just that one (though usually you have at least two, one in HEAD's reflog, one in the branch head's reflog)	15:06
	← joevano left	15:06
mebus_	is it possible to do git mv in one repository and then if I pull from it to do git mv in the local repo too?	15:07
surial	got it. This is working great.	15:07
mebus_	because now I got dublicate files.	15:07
doener	surial: note that the path shown for each size is just one name used for that blob.	15:07
	surial: i.e. if you renamed a file at some point, and run filter-branch to get rid of one version, you might then discover that the script now shows a different name	15:08
	← Pupeno left	15:08
doener	surial: looking for a paths for a given blob is too expensive	15:08
surial	bremner: So, it'll mess with other people's stuff, but what if I can tell them all explicitly to git pull? Would that help? Or do they have to toss it and re-clone?	15:10
	doener: Sure, but, in this case, I recognize the files, and they are indeed the big ones.	15:10
doener	surial: "git pull" wraps fetch + merge, so that would merge the new history, with what they have (the old history) == major mess up	15:10
	surial: "git pull --rebase" _might_ work, but of course only if they don't have merge commits locally	15:11
surial	oh, that makes sense. So they'd basically have to fetch, toss their local branches, and remake them from the remotes.	15:11
doener	surial: see git-rebase(1) "RECOVERING FROM UPSTREAM REBASE" for a discussion	15:12
surial	or do that. Okay. Huh. Well, a guesstimate is that the repo size would drop from 100MB to 20MB if I move the website files into a separate project and delete those used-to-be-checked in deliverables.	15:12
doener	surial: it's basically the same thing	15:12
	→ Pupeno joined	15:12
surial	Okay. Well, we don't have too many contributors _yet_, so now is the time, I guess. Thanks for all the help.	15:12
	→ ff joined	15:15
bremner	surial: just make sure you rebase often, that will keep the number of contributors managable :)	15:18
	← corni__ left	15:18
surial	bremner: :)	15:18
	→ aidan joined	15:20
	← babilen left	15:20
	qrush\|away → qrush	15:21
aidan	I'm a little confused .. I've got my repos on github, I'm trying to push some changes, but I need to update my repos first (the push is rejected) .. My pull fials with: http://bin.cakephp.org/view/1460823539	15:24
	→ hugoxrosa joined	15:24
cbreak	tried to specify?	15:24
LiamH	I'm trying to add a second remote and check out a branch from it, imitating the first example on the git remote man page http://www.kernel.org/pub/software/scm/git/docs/git-remote.html. However, the second git branch -r shows me the same result as the first. Is there a step missing?	15:27
	← stamina left	15:28
	→ zirpu joined	15:29
	→ babilen joined	15:30
LiamH	The git fetch returned no output	15:31
	→ belak joined	15:31
aidan	cbreak: specify what and where	15:31
	← belak left	15:31
cbreak	specify what the error message told you to specify	15:32
	the branch you want to merge with	15:32
	pull apparently is fetch and merge	15:32
	→ nevyn_ joined	15:33
	→ scarabx joined	15:33
aidan	there's only one nbranch, master	15:34
	git pull <repository> <refspec> ... what's repository and refspec?	15:35
ff	man git-filter-branch	15:37
Gitbot	ff: the git-filter-branch manpage can be found at http://git.or.cz/man/git-filter-branch	15:37
cbreak	you can try git pull someremote master	15:37
	→ HardPhuck joined	15:41
HardPhuck	i'm a git first timer so please be gentle :)	15:42
	how do i commit changes to the master repo?	15:42
cbreak	commit them locally	15:42
	then push your changes	15:42
	← jaysern left	15:43
	→ Axius_ joined	15:43
HardPhuck	ah.. push then, let's try	15:43
	error: unable to create temporary sha1 filename ./objects/fe: File exists	15:44
Axius_	I made a script for listing some files in a directoy.What do I need to put in the script to switch directly to that directory?	15:44
	← pcgod left	15:44
LiamH	OK, I found the problem; I think there's an error in the man page for git remote. It should have said "git fetch <remote-name>".	15:44
cbreak	Axius_: cd?	15:45
	← nevyn left	15:46
	← dunolie left	15:47
	← ff left	15:48
akitada	Is there any documentation that describes what each [svn-remote] attribute mean?	15:48
	← yaotti left	15:49
	→ davegb3 joined	15:49
	← khmarbaise left	15:49
	← davegb3 left	15:50
	→ yaotti_aw joined	15:50
	yaotti_aw → yaotti	15:50
	← mfilipe left	15:51
	← magcius left	15:52
loops	akitada, man git-svn has some info	15:53
Gitbot	akitada: the git-svn manpage can be found at http://git.or.cz/man/git-svn	15:53
loops	LiamH, it doesn't have to be a remote-name, it can also be a URL.	15:56
LiamH	loops: OK, but the example on the man page simply says "git fetch" and that does nothing.	15:57
	← icwiener left	15:57
	→ pcgod joined	15:57
loops	LiamH, that should actually update all your remote tracking branches	15:58
	LiamH, but it wont update any local branches	15:58
	← Axius left	15:58
	→ icwiener joined	15:58
loops	LiamH, however it will appear to do nothing if you're already up to date. it just returns to the shell prompt	15:59
	→ ianw_ joined	15:59
	← ianw left	15:59
	ianw_ → ianw	15:59
loops	LiamH, try "git fetch -v" to prove this	15:59
	→ corni__ joined	15:59
LiamH	loops: Not until I gave the remote name did it pull anything from the new remote. I was already up to date on origin.	16:00
	→ dunolie joined	16:01
loops	LiamH, perhaps there is an issue with the very first time you fetch from a new remote? If you try "git fetch -v" now (without remote name), does it work?	16:02
	→ jaysern joined	16:02
LiamH	loops: Yes, it looks fine.	16:02
	It was an issue with a first-time fetch from a new remote, but I took the example in the man page to be that situation.	16:03
	→ ianw_ joined	16:05
	← ianw left	16:05
	ianw_ → ianw	16:05
	Thell\|Away → Thell	16:06
	→ justin-george joined	16:06
HardPhuck	hmm i get errors when i try to 'push' could that have anything to do with permissions?	16:07
loops	HardPhuck, yes.	16:07
	← aidan left	16:07
HardPhuck	http://pastebin.com/md47ee3d	16:07
	this	16:07
	← MisterN left	16:08
loops	LiamH, do you have more than one remote configured ?	16:10
LiamH	loops: yes	16:11
	→ flaguy48 joined	16:11
loops	LiamH, so, it will only pull from one remote on each invocation, and it will try to pick the appropriate remote based on your currently checked out branch	16:12
	← nadim left	16:12
loops	LiamH, so until you fetch for the first time, and create a branch from what you fetch, the first remote will be being fetched (and presumably be up-to-date)	16:13
	LiamH, which instructions were you following in the docs ?	16:13
	→ mastro joined	16:14
	→ nadim joined	16:14
LiamH	git help remote, under "EXAMPLES", the first one "Add a new remote, fetch, and check out a branch from it". To the letter, of course with my names substituted.	16:15
	→ Ozzy joined	16:15
mastro	newbie question: what's the best way to abort a merge? (i always used rebase wich let me do --continue or --abort and I cant find the --abort command on git merge)	16:15
	← corni_ left	16:15
LiamH	The "git fetch" there has no argument, but the response is said to be " * refs/remotes/linux-nfs/master: storing branch ´master´ ..."	16:16
	I don't think that's right.	16:16
	← Ozzy left	16:17
	← Axius_ left	16:18
	← ianw left	16:19
	← babilen left	16:20
mastro	I did a git reset --hard HEAD	16:20
	but I don't know if this is the best way to do so....	16:21
	→ TML joined	16:23
mastro	another thing: I'm doing some testing: I created a new git repo.. then I cloned it with --bare option in another folder... added the bare-repo as remote location and cloned the bare repo in another directory: so now I have 2 directory that track the same repo (all locally)	16:24
	→ ianw joined	16:25
mastro	I noticed it created 2 branch in my remote repo: master and HEAD	16:25
	why that?	16:25
ianw	If I have a commit in master and I want to remove a bunch of merge commits with origin/master, can I do this safely: git rebase origin/master master	16:25
	→ Fikeis joined	16:26
	→ dvyjones joined	16:29
dvyjones	How do I undo a git-add?	16:29
	→ mjf joined	16:31
	→ qhoxie joined	16:31
wereHamster	dvyjones: git reset or git rm --cached	16:33
dvyjones	git rm --cached doesn't remove the file?	16:33
wereHamster	only from the index	16:33
dvyjones	Umm, wait a minute.	16:34
	Sorry, it wasn't a git-add, but a git-rm (the files are still there.	16:34
	→ JED3 joined	16:34
wereHamster	what exactly did you do?	16:34
	→ corni_ joined	16:34
dvyjones	Nothing.	16:34
	I just noticed that the commit would remove a lot of files.	16:35
wereHamster	and you want to undo 'nothing'?	16:35
	→ mfilipe joined	16:35
	← tbuehlmann left	16:35
	qrush → qrush\|away	16:39
	→ schierbeck joined	16:40
	← ianw left	16:41
	→ ianw joined	16:41
	← ianw left	16:43
	→ cloudhead joined	16:46
	steveire → steveire_	16:48
	← jtdowney left	16:49
	← corni__ left	16:50
	← jfxberns left	16:51
	→ aliceinwire joined	16:52
	→ jrmuizel joined	16:52
	→ MisterN joined	16:53
	← JED3 left	16:53
	→ schlortoo joined	16:54
	← jrmuizel left	16:54
	→ _jason553839 joined	16:56
	← compy left	16:57
	→ compy_ joined	16:57
	→ smuf joined	16:58
	→ pheaver joined	17:00
	→ GodEater joined	17:00
	← JKac3BEq left	17:07
	→ corni__ joined	17:09
	→ dreiss joined	17:10
	← objorn left	17:11
	← Mezner left	17:14
	← mastro left	17:15
	← corni_ left	17:17
	← simNIX left	17:17
	→ parasti joined	17:19
	→ ereslibre joined	17:19
	→ jfxberns joined	17:21
	→ gitte joined	17:22
	→ relix joined	17:24
	→ lresende joined	17:24
	← _jason553839 left	17:24
	Guest33124 → sdboyer	17:26
	→ jtdowney joined	17:26
	→ corni_ joined	17:26
	← akitada left	17:27
	← jtdowney left	17:27
	← corni__ left	17:28
	→ jtdowney joined	17:28
	← jtdowney left	17:29
	→ jtdowney joined	17:30
	← jtdowney left	17:31
	→ jtdowney joined	17:32
	← xenoterracide left	17:32
	→ eletuchy joined	17:33
	← jtdowney left	17:33
	→ jtdowney joined	17:34
	← jtdowney left	17:35
	→ defectiv joined	17:36
	→ jtdowney joined	17:37
	→ xenoterracide joined	17:37
	← jtdowney left	17:38
	→ jtdowney joined	17:40
	→ abms1116_ joined	17:41
	← jtdowney left	17:42
	← ciskje left	17:42
	→ jtdowney joined	17:43
	← abms1116_ left	17:44
	→ abms1116 joined	17:45
	→ g4mm4 joined	17:46
	← jfxberns left	17:47
	← pantsman left	17:49
	← smuf left	17:50
	→ smuf joined	17:51
	← abms1116 left	17:53
	← schlortoo left	17:55
	← dunolie left	18:00
	← jtdowney left	18:01
	→ ixti joined	18:01
	← mfilipe left	18:03
	rafpaf` → rafpaf	18:03
	← zirpu left	18:05
	qrush\|away → qrush	18:07
	→ mcolyer joined	18:10
	← Tuomas left	18:12
mcolyer	I am using jgit and I see that GitIndex is deprecated. I would like to be able to check things out but the only way that I can see to do that is using WorkDirCheckout which uses GitIndex. Can someone point me in the right direction?	18:12
	→ objorn joined	18:12
	← schierbeck left	18:14
	→ dunolie joined	18:14
	← objorn left	18:15
loops	mcolyer, spearce or robinr would be right people to answer here, but i think you can safely use WorkDirCheckout since it is not marked as deprecated. Likely it will be updated to use DirCache internally at some point	18:15
	→ rtl joined	18:16
mcolyer	loops: Thanks, that helps.	18:17
rtl	how would I delete a branch name prefixed with a dash?	18:18
	← smuf left	18:19
	← tazDg left	18:19
	compy_ → comps	18:19
Fissure	hm, typically one uses -- to signal end of options with gnu tools	18:19
	does branch -d -- branchname work?	18:19
loops	rtl, or git branch -d refs/heads/-branch	18:20
	← surial left	18:20
	→ tazDg joined	18:20
Fissure	why didn't i think of that	18:20
	the second step is to smack the person that created such a silly branch name	18:20
rtl	ah trueness.	18:20
	Fissure: hah, yarly.	18:21
loops	Fissure, your suggestion worked	18:21
Fissure	hm, good to know	18:22
rtl	Yea, both work, thanks peeps.	18:22
	→ rolfb joined	18:24
	→ aigon joined	18:24
	→ alvarezp joined	18:29
	→ Aikawa joined	18:32
	← Aikawa_ left	18:32
	→ jceb joined	18:32
	← nadim left	18:35
	← ilteris_ left	18:35
	← onigiri left	18:36
jaggz	do any of you live in Los Angeles?	18:36
	→ nadim joined	18:36
	→ danzat joined	18:38
	← qhoxie left	18:38
danzat	Any idea what would make git really slow when cloning/pulling?	18:38
Fissure	danzat: slow network, slow computers....	18:39
danzat	Fissure, I can assure you it's not that	18:39
Fissure	is it using cpu or sitting there?	18:39
danzat	just sitting there	18:40
	hanging	18:40
	taking its time	18:40
Fissure	over what protocol?	18:40
danzat	git://	18:40
	→ schierbeck joined	18:41
	← schierbeck left	18:41
	← jaysern left	18:41
danzat	is it using some specific port maybe?	18:41
Fissure	git:// is port 9418	18:41
danzat	I'll try to open it	18:41
loops	danzat, is the problem specific to cloning a specific repo? maybe that repo is just large and needs repacking	18:42
Fissure	heh, yeah, having the connection time out could cause that kind of slowness :P	18:42
danzat	no	18:42
	i get this from two hosts	18:42
Fissure	the repos don't have a huge number of branches/tags, do they?	18:43
danzat	no	18:43
	do you want to try yourself?	18:43
	→ enherit joined	18:43
	qrush → qrush\|away	18:43
loops	danzat, sure. what's the link	18:44
danzat	git://git.tuxfamily.org/gitroot/qucsqt4dev/qucsqt4.git	18:44
	→ bsdvodsky joined	18:44
loops	yeah, cloned in 15 seconds	18:44
danzat	still hangs here	18:45
	← defectiv left	18:45
loops	well hanging is different than slow	18:45
	likely you have some firewall issure etc	18:45
	← alley_cat left	18:45
danzat	the firewall isn't on	18:45
Ilari	danzat: Try to use telnet? If it connects, type crap there and see if it disconnects you then?	18:45
danzat	and I've opened that port in my router	18:46
rolfb	danzat: you've stopped the process and retried it?	18:46
danzat	it gave me this line: "remote: Counting objects: 11499, done." quickly enough	18:46
	→ hvoigt joined	18:46
Ilari	danzat: Then "hung"?	18:47
danzat	yes	18:48
jaggz	strace it?	18:48
Ilari	I would guess it's waiting for data from network...	18:49
	← dirk2 left	18:49
jaggz	I hate waiting too long for data from network	18:49
	→ ephcon joined	18:50
Ilari	And do idling TCP sockets really time out (the endpoints of connection won't try to maintain them)?	18:50
loops	danzat, you've tried a few times, and same thing each time?	18:51
danzat	getting the strace	18:51
	← tehbaut left	18:51
	→ tehbaut joined	18:52
	← _Mica_ left	18:52
	→ \|CheckBit\| joined	18:52
	← \|CheckBit\| left	18:52
danzat	http://pastebin.mozilla.org/677201	18:53
	→ qhoxie joined	18:54
danzat	I killed it after it hanged	18:54
	for a minute	18:54
Ilari	danzat: Completes object negotiation, then starts to wait for pack data, which doesn't arrive.	18:55
	danzat: Actually, there's child processes involved. Maybe strace with -f would be more useful?	18:55
loops	Ilari, what about the NAK at 743.. is it normal?	18:56
danzat	ok	18:56
	→ onigiri joined	18:56
	← wshimmy left	18:56
Ilari	loops: Yes, its normal.	18:57
rolfb	is there a good way to update a repo according to its .gitignore? by that I mean making the repo behave is if the gitignore was always in place even if its been changed at a later point	18:58
	i've been using 'cat .gitignore \| xargs git rm --cached	18:59
	← hvoigt left	18:59
danzat	with -f: http://pastebin.mozilla.org/677204	18:59
	← aigon left	18:59
thiago_home	rolfb: I think you need to do like you're doing: manually	19:00
	→ aigon joined	19:00
rolfb	ok	19:00
	but what will happen to people who has those files locally in their repo if they pull the repo after i did this?	19:01
	← aigon left	19:01
	→ aigon joined	19:01
Ilari	danzat: What happened to that 30620 process? It didn't seem to enter kernel to wait for anything...	19:01
	← aigon left	19:02
loops	danzat, that works but has some corner cases. doener came up with better one liner last night: git ls-files -c -i --exclude-standard -z \| xargs -0r git rm --cached	19:02
	→ aigon joined	19:02
	← aigon left	19:03
loops	ooh damn.. rolfb ^	19:03
	→ jacob_ joined	19:03
rolfb	loops: ?	19:03
	jacob_ → Guest76688	19:03
loops	rolfb, those files will be removed for them	19:03
	← jchris left	19:04
	Guest76688 → j_christ	19:04
rolfb	thats kinda of bad?	19:04
loops	rolfb, that line above was for you	19:04
rolfb	i figured :)	19:04
	thanks	19:04
loops	rolfb, well they can recover copies easily enough. But yeah, it's a pain	19:05
	→ aigon joined	19:05
lenix^	is git supported wel in eclipse?	19:05
	→ hvoigt joined	19:05
loops	lenix^, EGit is actively developed yeah	19:05
rolfb	our issue is that sqlite3 databases which contains mostly test-data doesn't version great	19:06
	so it might be better to use a remotely shared test-db?	19:06
	for our group	19:06
	← bsdvodsky left	19:06
loops	rolfb, whatever works. You might consider versioning ascii database dumps rather than the binary.	19:07
	← danzat left	19:08
rolfb	interesting	19:08
lenix^	sweet tits	19:08
rolfb	git won't fuck up the structure?	19:09
loops	git wont touch it at all	19:09
rolfb	won't touch the dump?	19:09
loops	Right.. no more than it would touch the binary database	19:09
rolfb	i'm not following	19:10
loops	me either apparently :o)	19:10
rolfb	if i dump to db to ex: yaml	19:11
loops	How would Git fuck up the structure? The structure of what?	19:11
rolfb	yaml has a conscise structure if i want to load the data into the db again	19:11
	so i'm just wondering if there is a loooooot of lines between parent structure and the new lines from a different merge, if it would fuck up the structure and put the wrong fields in the wrong table	19:12
loops	I suppose you'd have to be careful when merging, might require a custom merge strategy.	19:12
rolfb	right	19:12
	so remote db we do	19:12
DrNick	why dump to yaml when you can dump to SQL?	19:12
rolfb	:)	19:12
	→ jrmuizel joined	19:13
rolfb	better consistency between databases	19:13
	database-types	19:13
DrNick	ok, whatever, if you want to write your own YAML dumper instead of just using the sqlite3 .dump command, go ahead	19:14
rolfb	DrNick: yaml db plugin already exist	19:14
	:)	19:14
	→ priidu joined	19:15
DrNick	echo .dump \| sqlite3 blah.db > blah.sql	19:15
	← aigon left	19:16
	→ aigon joined	19:16
	← aigon left	19:17
	→ keex joined	19:17
	→ tjc joined	19:19
	← jrmuizel left	19:20
j_christ	If someone can help with this, it would be great... So, I can use git-tag to get a list of all tagged commits, but is there a way to tell if the current working directory is tagged?	19:21
	→ aigon joined	19:21
	→ danzat joined	19:22
	← aigon left	19:22
	→ aigon joined	19:22
	← mebus_ left	19:23
j_christ	Or I guess alternatively, how to get the sha1 for the working directory so that I can compaire it to sha1's in the .git/refs/tag dir?	19:23
loops	j_christ, maybe something like git describe --tags --exact-match HEAD	19:24
j_christ	loops, I think I can work with that... Thanks.	19:25
rudi_s_	j_christ: git tag --contains $(git rev-parse HEAD) would work too.	19:25
	← hvoigt left	19:25
rudi_s_	But loops one is better.	19:25
	Thell → Thell\|Away	19:26
	→ adam12 joined	19:27
	← aigon left	19:27
	→ aigon joined	19:27
j_christ	rudi_s, git tag --contains $(git rev-parse HEAD) does not seem to work as well.	19:27
rudi_s_	j_christ: Hm, works for me (but loops one is better anyway).	19:29
j_christ	rudi_s, if I'm on a tagged branch it returns the tag, but if I'm not, it returned the tag as well.	19:30
	→ reidab joined	19:30
Ilari	danzat: I noticed that that paste seems to be truncated...	19:30
rudi_s_	j_christ: Hm, no idea, works for me.	19:30
loops	j_christ, the --contains means that the tag will be displayed if it is reachable from HEAD (ie. anywhere in the history), whereas exact match means it _is_ the HEAD	19:31
	j_christ, that is --contains means if HEAD is reachable from a given tag	19:31
	→ bremner_ joined	19:32
danzat	Ilari, http://pastebin.com/m8cb3d21	19:32
j_christ	loops, I understand... was just reading something about that in man	19:32
	→ owen1 joined	19:32
	← tjc left	19:32
	→ joevano joined	19:32
owen1	i installed gitosis, typing sudo -H -u git gitosis-init < /tmp/id_rsa.pub result in: env: -H: No such file or directory any clues?	19:33
	→ c15c8ra1n joined	19:34
	← aigon left	19:34
	← c15c8ra1n left	19:34
	→ aigon joined	19:35
loops	owen1, maybe someone will pipe up with an answer, you might consider gitolite, which is more actively developed these days: http://github.com/sitaramc/gitolite	19:35
	→ bremner__ joined	19:36
	← bremner__ left	19:36
Rhonda	owen1: You used wrong commandline options, -H requires an argument, see man sudo.	19:37
owen1	loops: thanks, i'll take a look	19:37
Ilari	danzat: Okay, looking at situation: The 19 process (sideband demux?) seems to block on network input. 20 process (index-pack) blocks on output from 19 process. 18 process (clone) blocks on output of 19 process.	19:38
Rhonda	owen1: erm, forget it, did read it wrong myself. I guess your git user doesn't have a home set in your passwd file - or that directory isn't existing.	19:38
	← aigon left	19:38
	→ aigon joined	19:38
	← aigon left	19:39
	→ aigon joined	19:39
	← aigon left	19:39
owen1	Rhonda: what do u mean 'home set in my password file'?	19:40
	← p_masho left	19:40
	→ aigon joined	19:40
owen1	i have dir called /home/git	19:40
Rhonda	And is that set as home of the git user?	19:41
	getent passwd git	19:41
danzat	Ilari, what does it mean?	19:41
	← aigon left	19:41
	→ jaysern joined	19:41
	→ bremner__ joined	19:42
	→ aigon joined	19:42
Ilari	danzat: It seems that the hang is because connection hangs...	19:42
danzat	hm...	19:43
	← aigon left	19:43
	→ WALoeIII_ joined	19:43
owen1	Rhonda: i think so. i run the adduser command as described on the blog post - sudo adduser --system --shell /bin/sh --gecos 'git version control' --group --disabled-password --home /home/git git	19:44
	← pheaver left	19:44
danzat	Ilari, can I specify the listening port? maybe I got the wrong port open	19:44
	← WALoeIII_ left	19:45
	→ pheaver joined	19:45
	→ aigon joined	19:45
	← aigon left	19:45
	→ WALoeIII_ joined	19:45
Rhonda	owen1: If "getent passwd git \| cut -d: -f6" does print /home/git then it should be alright. But yes, gitolite might be more interesting these days. :)	19:45
Ilari	danzat: Definitely not wrong port. The have/want negotiations complete in that trace.	19:46
	→ aigon joined	19:46
	← aigon left	19:46
	→ aigon joined	19:47
	→ sh1mmer joined	19:48
	← eean left	19:48
owen1	Rhonda: it printed /home/git. btw, i am using my laptop machine both as server and client so my rsa.pub is not from different machine. could this be the issue?	19:48
	← aigon left	19:48
	→ aigon joined	19:48
	← aigon left	19:49
owen1	Rhonda: i read about gitolite. the main difference is permission per branch, which i don't really care about. is there anything else?	19:49
Ilari	owen1: 'sudo -H -u git /bin/bash' complains?	19:49
owen1	Ilari: yes. got the same error	19:49
Ilari	owen1: Looks like sudo is busted...	19:49
j_christ	Another question... Is there a way to tell if your working directory is dirty (other than git status)?	19:50
Ilari	owen1: I think one can work around it. 'sudo -u git /bin/bash' does work?	19:50
	→ aigon joined	19:50
	← bremner_ left	19:50
owen1	Ilari: env: -u: No such file or directory	19:50
	← aigon left	19:51
	← aspotashev_ left	19:51
Ilari	owen1: Can you get root?	19:51
	→ aspotashev joined	19:51
	→ aigon joined	19:52
	← aigon left	19:52
Rhonda	owen1: What does sudo -l give you? If it's longish, pastebin it.	19:52
	→ g0bl1n joined	19:52
owen1	Rhonda: env: -l: No such file or directory	19:52
Rhonda	which sudo	19:52
	← Blackshark left	19:52
	→ pantsman joined	19:53
	← krynnotaur left	19:53
Rhonda	I think your sudo is b0rked.	19:53
Ilari	owen1: root => root shell, that is.	19:53
owen1	/usr/bin/sudo	19:53
	Ilari: i am not sure what do u mean by that. sorry (-:	19:53
	psoo_ → psoo	19:54
Ilari	owen1: Shell as user root.	19:54
owen1	Ilari: can u tell me how?	19:54
Rhonda	owen1: If sudo -l doesn't work then you have a problem. But it's not git related.	19:54
Ilari	owen1: Maybe sudo works enough for 'sudo /bin/bash'.	19:54
	→ jtdowney joined	19:55
	← gitte left	19:56
owen1	Ilari: i am confused. what should i type? sudo /bin/bash ?	19:56
Ilari	owen1: Ah, what's the current user?	19:57
	→ bremner_ joined	19:57
	← jtdowney left	19:57
owen1	Ilari: i logged in to the machine with oren. is that what u mean?	19:57
	→ davido joined	19:57
Ilari	owen1: Well, does 'sudo /bin/bash' also complain or give a shell?	19:57
owen1	Ilari: no complain	19:58
	← bremner_ left	19:58
owen1	Ilari: i see bash again	19:58
	or shell. whatever u call this.	19:58
	→ bremner_ joined	19:58
Ilari	owen1: Try 'whoami'... It should show 'root'.	19:59
	→ Florin_ joined	19:59
owen1	Ilari: correc	19:59
	t	19:59
Ilari	owen1: Then 'su --login git' (note, su, not sudo).	19:59
	→ bcardarella joined	20:00
loops	j_christ, git ls-files --exclude-standard -o -m	20:00
	← tazDg left	20:00
	← WALoeIII left	20:00
	WALoeIII_ → WALoeIII	20:00
owen1	Ilari: done	20:00
	← DavidKlein left	20:00
j_christ	loops, cool, thanks.	20:00
	→ schierbeck joined	20:01
Ilari	owen1: Now 'whoami' output should have changed to 'git'. And 'echo $HOME' should show '/home/git', right?	20:01
	← Florin_ left	20:01
owen1	Ilari: correct	20:01
Ilari	owen1: Now try 'gitosis-init </tmp/id_rsa.pub'.	20:02
owen1	Ilari: i use ubuntu 9.04, btw	20:02
	→ aigon joined	20:02
owen1	Ilari: works! so what is the issue with my sudo?	20:03
	→ tazDg joined	20:03
Ilari	owen1: No idea.	20:03
	→ JKac3BEq joined	20:03
Ilari	owen1: Now in another shell, try 'ssh git@localhost'. It should throw complaint about SSH_ORIGINAL_COMMAND.	20:03
owen1	Ilari: Ilari correct	20:04
	← dunolie left	20:05
Ilari	owen1: One final sanity check. 'ls -l ~git/repositories/gitosis-admin.git/hooks/post-update'. What does that show (the whole line)?	20:05
owen1	Ilari: i run it as the regular user, not git - ls: cannot access /home/git/repositories/gitosis-admin.git/hooks/post-update: Permission denied	20:06
Ilari	owen1: Run it as git user.	20:07
	owen1: Since in normal operation it gets run as gitosis user.	20:07
owen1	Ilari: when u say 'run it as a git user' u mean type: su --login git ?	20:08
Ilari	owen1: Don't you have the shell as git user still open?	20:08
owen1	Ilari: ho. ok	20:09
Ilari	owen1: Well, open one (that sudo and then su should work).	20:09
	owen1: (some command line params needed, see the commands in backlog).	20:09
owen1	Ilari: -rw-r--r-- 1 git git 69 2009-10-18 13:06 /home/git/repositories/gitosis-admin.git/hooks/post-update	20:10
Ilari	owen1: Ouch. Chmod 755 that file.	20:10
	owen1: After that, it should show '-rwxr-xr-x 1 git git 69 2009-10-18 13:06 /home/git/repositories/gitosis-admin.git/hooks/post-update'	20:11
	← schambers1 left	20:12
owen1	Ilari: one sec. from some reason i don't have tab complete.	20:12
	← malumalu left	20:13
Ilari	owen1: Don't you have select and paste?	20:13
owen1	Ilari: it asked my for password for the user git. what is it?	20:14
	Ilari: unless i am suppose to do all this with the user oren or the root?	20:14
Ilari	owen1: su asks? If you run it as root, it won't.	20:14
owen1	Ilari: i run it like this: sudo chmod 755 post-update	20:15
	(i cd into the correct dir)	20:15
Ilari	owen1: What does 'whoami' show?	20:15
	← bcardarella left	20:15
owen1	Ilari: git	20:16
Ilari	owen1: Then 'chmod 755 post-update' should work.	20:16
	← reidab left	20:16
	→ Stravy joined	20:16
	← jceb left	20:16
Ilari	owen1: Now 'ls -l post-update' should show that the permissions have changed...	20:17
owen1	-rwxr-xr-x	20:17
	→ corni__ joined	20:17
	← mbroeker left	20:17
	← Fikeis left	20:18
	← bremner__ left	20:18
	← g0bl1n left	20:18
Ilari	owen1: Yup. Now you should be set to clone the gitosis admin repo (in normal user shell): 'git clone git@localhost:gitosis-admin.git'.	20:18
	→ dwmw2_gone joined	20:18
owen1	Ilari: works. thank you!!	20:19
	Ilari: i am curious to understand what the hell is wrong with my laptop	20:19
	← rleigh left	20:19
Ilari	owen1: Sudo being busted is OS-level problem...	20:20
	→ dunolie joined	20:21
	← aigon left	20:21
	← j_christ left	20:21
	→ xanonus joined	20:22
	→ jtdowney joined	20:22
	→ erikvold joined	20:22
rolfb	Ilari: not PIBCAP?	20:22
	err	20:22
	→ slabbeh joined	20:22
rolfb	pibcac	20:22
	← corni__ left	20:23
LotR	ITYM PEBKAC :)	20:23
owen1	what is the main reason for using gitolite over gitosis?	20:24
loops	owen1, per branch permissions, easier? admin	20:24
	← lenix^ left	20:24
	← corni_ left	20:25
Ilari	owen1: Well, let's say that you want that sudo fixed before you migrate from gitosis to gitolite. Gitolite is mostly intended to be admined using remote shell...	20:25
	← Pupeno left	20:26
	→ Pupeno joined	20:27
owen1	Ilari: what should i look for when googling for my issue? ask for help with sudo -l ?	20:27
	→ khmarbaise joined	20:28
Rhonda	owen1: Check what is in your sudoers.	20:28
	Content of /etc/sudoers that is.	20:28
	← kampasky left	20:28
	→ kampasky joined	20:29
owen1	Defaults env_reset	20:29
	root ALL=(ALL) ALL	20:29
	← jaysern left	20:30
owen1	%admin ALL=(ALL) ALL	20:30
Ilari	To me, it looks more like sudo does accept the command, but treats all that -H and -u and those as part of command.	20:30
	→ ilteris_ joined	20:31
patrikf	I read so much sudo -l here. shouldn't that be sudo -i?	20:32
	→ bremner__ joined	20:35
	← bremner_ left	20:37
Rhonda	Ilari: Yes, but I have no clue why it would do that, thought it might be one of the options set in there ...	20:38
	patrikf: No, sudo -l displays what you can do.	20:38
	→ eparis joined	20:38
	→ _jason553839 joined	20:39
eparis	git-am is complaining because the patch tried to change a file that was not in tree. I want to just continue and have it drop that hunk. I can't git-add a file that isn't there (can I?) I can't git-am --resolved.... how do I make it just keep going?	20:40
	(i'm reordering some patches, so this doesn't surprise me that the file doesn't exist and I'm ok with that)	20:42
	← Grahack left	20:42
Ilari	eparis: 'git rm --cached <file>' and then 'git am --resolved'?	20:42
	← ludde left	20:44
eparis	$ git rm --cached fs/notify/fanotify/fanotify.c	20:44
	fatal: pathspec 'fs/notify/fanotify/fanotify.c' did not match any files	20:44
Ilari	eparis: 'git status' lists what as unmerged or some other weird status?	20:44
eparis	is say it's clean.	20:45
	← jaggz- left	20:45
Ilari	eparis: 'git am --skip' then?	20:45
eparis	But it did make changes (or should have) to other files.....	20:45
Ilari	eparis: Well, status should show changes to those files then...	20:46
eparis	oh well, I'll just do it the hard way, revert all of it, drop that hunk from the mbox, reapply the whole mbox.	20:46
	→ reidab joined	20:47
	← qhoxie left	20:47
	← owen1 left	20:49
	← lucs left	20:49
	→ khmarbaise_ joined	20:52
	← rtl left	20:53
	→ seb_ joined	20:57
	seb_ → \|Lupin\|	20:57
	→ rtl joined	20:57
\|Lupin\|	good evening, everybody	20:57
	← loops left	21:00
	← khmarbaise left	21:00
	khmarbaise_ → khmarbaise	21:00
Ilari	\|Lupin\|: Pipe characters because they feel distinctive? :-)	21:00
	→ aidan joined	21:01
aidan	What do I do about this: html/config/core.php: needs update	21:01
	git pull (gives that)	21:02
Ilari	aidan: You have uncommitted changes to that file...	21:02
	← bremner__ left	21:02
	→ jaysern joined	21:03
	← WALoeIII left	21:03
	→ WALoeIII joined	21:03
\|Lupin\|	Ilari: \| characters for the modulus... that's a way of reminding me to be always positive... :-)	21:04
	← adam12 left	21:04
	→ Mezner joined	21:07
	→ h0nk joined	21:08
	← _jason553839 left	21:10
	→ hvoigt joined	21:10
	← h0nk left	21:12
	← ferdy left	21:13
	← tazDg left	21:13
	→ murph joined	21:15
aidan	Ilari: how can I just pull master and overwrite any changes?	21:15
	← selckin left	21:15
	← ia left	21:15
murph	Is there a good way to get a diff/diffstat of all changes in the last 24 hours?	21:15
	← keex left	21:15
Ilari	aidan: 'git stash' and then 'git pull origin master'?	21:15
\|Lupin\|	aidan: git reset ?	21:16
	(speaking under the git experts' control)	21:16
	oops I'm wrong apparently...	21:16
	→ sgh joined	21:17
sgh	Hi	21:17
\|Lupin\|	little question please	21:18
	I have done something bad with my git repo and now don't know how to make things return to a usable state..	21:18
	I have commited a change on a file	21:18
	then modified the file	21:18
	then wanted to amend the commit	21:18
sgh	How do I delete a remote branch without having the remote branch persisting in another users clone?	21:19
\|Lupin\|	and to do so I thought I had to do a git reset --soft HEAD^	21:19
	so now I think the file, he index and the commit tree are in three different states.	21:19
	I'm wondering how to go out of this trap...	21:19
	→ abadr joined	21:19
	→ ferdy joined	21:20
	← jtdowney left	21:20
	← WALoeIII left	21:20
	← ph^_ left	21:21
Ilari	\|Lupin\|: First look at 'git reflog' and back off the reset (with another git reset --soft to suitable state). Then add the changes you want to amend and then commit --amend?	21:21
\|Lupin\|	Ilari: thanks ! let me try...	21:22
	→ WALoeIII joined	21:22
\|Lupin\|	Ilari: actually, what is taken into account by the -amend option ? the content of the index ?	21:24
	→ ia joined	21:25
	← drizzd left	21:25
	→ tazDg joined	21:26
	→ _jason553839 joined	21:27
	← xanonus left	21:27
\|Lupin\|	Ilari: I did a git reset --soft HEAD@{1} which worked I think, but git diff does not show any difference for the file I'm interested in, so I'm still confused...	21:27
Ilari	\|Lupin\|: It determines what to commit like ordinary commit.	21:27
	\|Lupin\|: Maybe 'git diff --cached' does?	21:28
	← hvoigt left	21:28
	← ilteris left	21:28
	ilteris_ → ilteris	21:28
\|Lupin\|	Ilari: yeah, it wants to remov ethe file I think, because I did a git rm --cached thefile	21:29
	shall I re-add the file as a whole ?	21:29
Ilari	\|Lupin\|: Yeah, unless you want the file removed.	21:29
	← priidu left	21:29
\|Lupin\|	Ilari: no no, I really want it.	21:31
	Ilari: I re-added it iand it works... almost	21:31
	there are some changes to this file I don't want to commit as part of the ammend	21:32
	amend	21:32
	Ilari: so I guess I have to remove them and to re-add the file, so that just the amend part gets in the index...	21:32
	← eparis left	21:32
	← khmarbaise left	21:33
\|Lupin\|	Ilari: hmm there is still something I don't understand about comit -amend... it takes into account both the content of the index and the content of the last commit ?	21:35
	→ bx2_ joined	21:36
	→ smuf joined	21:37
	→ swombat joined	21:38
swombat	Hello... I need some git help...	21:38
	→ wagle joined	21:38
swombat	We've had a bit of a boo-boo... my colleague merged the master branch into the prod branch early, then undid the merge, then committed prod and deployed that... now i did a merge from prod to master, and it's overwritten my changes for the last week or so	21:38
	i think i know which commits need to be reapplied	21:39
	how do I reapply them to master? (or to prod)	21:39
	← rtl left	21:39
kusma	swombat: git cherry-pick?	21:39
swombat	kusma: that works in the same branch too?	21:39
	← patrikf left	21:39
	swombat tries	21:39
kusma	not sure, did you give it a try?	21:40
	→ Aikawa_ joined	21:40
swombat	it works!	21:40
kusma	:)	21:40
	← icwiener left	21:40
Ilari	\|Lupin\|: Only (constructed) index is used in determining what to commit.	21:40
\|Lupin\|	Ilari: constructed index ? what's that ?	21:40
Ilari	\|Lupin\|: -a, -i, -o, filenames and all that.	21:41
	→ tbuehlmann joined	21:42
	← parasti left	21:42
	→ loops joined	21:43
	← Sigma left	21:43
\|Lupin\|	Ilari: ok	21:45
	Ilari: btw, not sure it was you or someone else who suggested reading the git Parabble, Git for computer scientists and Git from the bottom up, but the parabbel was a ver nice suggestion, thanks	21:45
	← aidan left	21:45
	→ Zenopus_ joined	21:46
\|Lupin\|	Ilari: git for computer scientists has many pictures I couln't see, and git botom up was a pdf so I may ask the author if he could send me something easier to read for me, such as a LaTeX file...	21:46
	← aspotashev left	21:47
	← lhz left	21:49
	← bx2_ left	21:50
	← danzat left	21:50
	→ danzat joined	21:50
Ilari	\|Lupin\|: Running pdftotext on that PDF at least doesn't seem to produce complete garbage...	21:50
	→ btipling joined	21:51
	← kur1j left	21:52
btipling	say I have two branches working on different stuff, both are not the master branch, but both are derived from the master branch	21:52
	→ schambers joined	21:52
btipling	I have a new file in the first branch and I want it in the second branch	21:52
\|Lupin\|	Ilari: yeah it often produces something usable... the thing is I like seeing the structure, what is emphasized, etc. Well I'll see, perhaps start with a pdftotext...	21:52
btipling	how do I get that file without causing problems?	21:52
	→ bcardarella joined	21:53
\|Lupin\|	anyway... thanks to Ilari and all for so much help provied so far !	21:53
	btipling: if it's commited, cherry-pick	21:53
	← reidab left	21:53
btipling	hrm, ok I'll check that command out	21:54
	thanks \|Lupin\|	21:54
\|Lupin\|	btipling: you are vry welcome !	21:54
	btipling: you are the first person I can help on #git :)	21:54
btipling	heh cool	21:54
	I'm new to git	21:54
	→ agenteo joined	21:54
	← _jason553839 left	21:54
\|Lupin\|	btipling: so am I :)	21:55
	good night eveybody !	21:55
	bye !	21:55
	← \|Lupin\| left	21:55
	← Aikawa left	21:56
	→ Aikawa joined	22:00
	← Zenopus left	22:00
	Zenopus_ → Zenopus	22:00
	→ zacharyc_ joined	22:00
	← zacharyc_ left	22:01
	→ hvoigt joined	22:01
	← hvoigt left	22:01
	→ zacharyc_ joined	22:01
	← psoo left	22:02
	← ixti left	22:02
	← scarabx left	22:07
	→ rtl joined	22:07
	← rolfb left	22:10
	← bcardarella left	22:12
	→ akitada joined	22:13
btipling	how do I check out another local branch if I don't want to commit changes I have in my current active branch	22:16
	← gunmuc left	22:16
	← Aikawa_ left	22:17
btipling	ah -f	22:17
	→ werdnativ joined	22:19
	→ EricInBNE joined	22:19
	← tbuehlmann left	22:21
	→ simcop2387 joined	22:22
simcop2387	is there a way to setup a default repository for git-push?	22:22
	i'm sure ts in the manual i'm just having a brain fart right now	22:22
werdnativ	yes, for each branch. in .git/config	22:23
	[branch "master"]	22:23
	remote = origin	22:23
	merge = refs/heads/master	22:23
	where origin is your named remote.	22:23
simcop2387	ah thanks	22:24
	← Mezner left	22:24
Mikachu	btipling: that will also delete those changes	22:24
	→ reidab joined	22:24
werdnativ	on a related note, can a branch be linked to multiple remotes?	22:24
btipling	Mikachu: I didn't actually have any changes, I undid them all, and I didn't want to commit nothing, so it's ok	22:24
Mikachu	okay	22:24
	← sgh left	22:24
werdnativ	e.g. pull from one, push to another?	22:25
	← rudi_s_ left	22:25
	→ magcius joined	22:27
	→ JasonWoof joined	22:27
	← schmukkk left	22:27
Ilari	werdnativ: No. But one can override what to pull and where to push to.	22:29
werdnativ	right, just git push somewhere	22:29
	← Stravy left	22:30
	← simcop2387 left	22:30
	→ JEEBcz joined	22:30
	→ Azeroth-Work joined	22:31
	← slabbeh left	22:33
	→ fynn joined	22:33
fynn	Yo. What do you guys call the permanent branch in that holds the code you actually deploy in production?	22:33
	← onigiri left	22:34
	← Yuuhi left	22:34
	kardan_ → kardan	22:36
	→ aresnick joined	22:37
	← pantsman left	22:39
	← Bass10 left	22:40
chuck	fynn: master ;-)	22:40
	← JEEB left	22:43
	→ eean joined	22:46
	JEEBcz → JEEB	22:49
	← akitada left	22:50
	← thiago_home left	22:50
	→ jrmuizel joined	22:54
	← mjf left	22:54
	→ Bass10 joined	22:56
	→ oneforall joined	23:01
	→ akitada joined	23:02
oneforall	http://pastebin.ca/1628160 any idea why I get this ?	23:02
	← abadr left	23:04
rtyler	because you have a crappy xmltoman install?	23:05
	or	23:05
	you're running this somewhere where you cannot allocate as much memory as xmltoman needs and they're not handling E_NOMEM properly?	23:06
	← jrmuizel left	23:06
oneforall	seems to be something peopel are getting with glibc 2.10.1	23:06
	its in my chroot. compiled fine many times before (have 8GB and less open too than I did many times before) but the diff is glibc2.10.1, gcc4.4.1 .. hmm But I'll check and see the last time I compiled xmltoman.	23:08
	← hachi left	23:09
	← luke-jr left	23:12
	→ luke-jr joined	23:13
	→ jrmuizel joined	23:14
	← jrmuizel left	23:17
	→ pgregory joined	23:17
	← pas left	23:17
pgregory	hi	23:17
Ilari	...	23:17
pgregory	can anyone tell me why over an smbfs share I'm getting the following error when I push?	23:17
	fatal: Unable to create temporary file: Permission denied	23:18
	→ Haarg joined	23:18
	→ pas joined	23:18
pgregory	I can write to the share, and create directories, and everything, but git fails with that.	23:18
Ilari	pgregory: Sounds like permission error (remote side).	23:18
pgregory	Ilari: but I can write to the share, and even write to the folder that contains the git repository.	23:18
Ilari	pgregory: Is some kind of syscall tracer (strace, etc) available?	23:19
pgregory	I can create files, delete them, create directories etc.	23:19
	← __iron left	23:19
pgregory	Ilari: I have strace, but I have no clue how to use it.	23:19
Ilari	pgregory: 'strace -f <command to trace>'. It outputs trace to stderr...	23:20
	→ abadr joined	23:21
pgregory	the last message before the error is...	23:21
	[pid 10196] open("./objects/pack/tmp_pack_0U8wO5", O_RDWR\|O_CREAT\|O_EXCL\|O_LARGEFILE, 0600) = -1 EACCES (Permission denied)	23:21
	→ krynnotaur joined	23:22
pgregory	aha, I don't have permission to access the .git folder for some reason.	23:22
	← mithro left	23:23
pgregory	bit it shows up as drwxrwxrwx 1 pgregory root	23:23
	same as all other files	23:23
	→ sergiodj_ joined	23:24
pgregory	I mean folders	23:24
	← justin-george left	23:25
pgregory	this is weird, all I want to do is push and go to bed :(	23:25
Ilari	pgregory: Share permissions on server?	23:27
pgregory	possibly, however, I haven't actually 'done' anything to forcibly stop access to .git	23:27
	so unless git itself does that when it creates it, I can't see how.	23:27
Ilari	pgregory: 'whoami' in shell where push fails prints what? 'pgregory'?	23:29
pgregory	yes	23:29
Ilari	pgregory: ... So no rootsquash or so involved...	23:30
	← abadr left	23:30
pgregory	nope	23:30
	→ jrmuizel joined	23:30
pgregory	I'll have to send a patch now, it's beaten me	23:30
fynn	OK, again I forgot: how do I create a local foo branch, that would also be tracking a remote branch?	23:31
	so when I "git push", the remote would have a foo branch as well, with everything my foo branch has.	23:31
	→ hachi_ joined	23:32
pgregory	cheers	23:32
	← pgregory left	23:32
	← d0k left	23:32
Ilari	fynn: 'git push <remote> foo' (except that doesn't set up the info for pull).	23:33
fynn	Ilari: yeah, but don't I need to specify anything at foo's creation time?	23:33
	hachi_ → hachi	23:33
	← sergiodj_ left	23:35
Ilari	fynn: Just specify the correct starting point (if current commit isn't correct).	23:36
	← Haarg left	23:36
	← Sonderblade left	23:36
	dwmw2_gone → dwmw2_NRT	23:37
	← akitada left	23:39
fynn	Ilari: OK, I did: "git checkout -b foo; git push origin foo"	23:39
	that worked, but foo isn't tracking origin/foo	23:39
Ilari	fynn: Yeah. Needs 'git config branch.foo.remote origin', 'git config branch.foo.merge refs/heads/foo'.	23:41
fynn	Ilari: hm, I thought there was some paramter in the branching command to make foo automatically track origin/foo?	23:42
	→ plediii joined	23:42
	← reidab left	23:43
	→ owen1 joined	23:44
Ilari	fynn: There is, but the branch to track must exist...	23:45
fynn	Ilari: nod, that's what confused me, thanks :)	23:46
	← schierbeck left	23:46
	→ mithro joined	23:46
sitaram	one trick I have used in the past (easier to remember) is after that "push origin/foo", I'd "git branch -D foo; git checkout -t origin/foo"	23:47
	→ defectiv joined	23:47
fynn	oooh that's very clever	23:48
sitaram	well I dont know about clever but it gets the job done with stuff my somewhat feebler mind can remember :)	23:49
	← relix left	23:49
owen1	i try to add new repo to gitosis. git remote add origin git@localhost:cool_project.git and get: fatal: remote origin already exists. what's the problem?	23:50
	→ _ringlej joined	23:50
	← ringlej left	23:51
Ilari	owen1: The name 'origin' is already in use. Pick another name for remote.	23:51
	← ia left	23:52
owen1	Ilari: what does it mean 'name for remote' is this the name i'll use to identify this specific repo?	23:52
	← sergiodj left	23:52
Ilari	owen1: Short name to designate the repo (that also appears in names of remote tracking branches associated with that repo).	23:53
owen1	Ilari: ok, i'll use the project's name, i guess.	23:53
sitaram	owen1: one normally uses a name referring to the server when naming an origin; the fact that many servers host onnly one repo of interest to us notwithstanding :)	23:55
	sya tomorrow you want to get a second project from there, you'd end up calling it "git clone project-A:project-B.git"	23:56
	say*	23:56
	← defectiv left	23:57
owen1	sitaram: so origin should be an identifier for my server? and will i be able to use it on more projects?	23:57
	→ ia joined	23:57
sitaram	if the server has a related project you want to fetch a branch or something from, sure...	23:58
	nothing stops you from naming it whatever you want, of course -- I was just saying it might make more sense to name ifafter the server	23:58
	it	23:59
	say git.kernel.org has multiple linux repos (various people's linux trees) and you're tracking linus's tree from it	23:59
	→ aresnick_ joined	23:59

Logs Search ←Prev date Next date→

Channels Documentation