IRCloggy #git 2008-08-19

Provider of IRC logs since 2005.

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

	2008-08-19
	← spearce left	00:00
	→ faheem joined	00:00
faheem	Hi. I'm trying to check out the django branch of the following repos online, http://github.com/kwe/passenger-django-wsgi-example/tree/django but having some difficulty. If I do git clone git://github.com/kwe/passenger-django-wsgi-example.git do I get both branches listed or just one?	00:01
	→ mithro joined	00:01
	→ cnk joined	00:01
faheem	It looks to me like I just get master.	00:01
	→ toxx_ joined	00:01
johnw	does "git log ..foo" show me all commits in foo which are not in HEAD?	00:04
	→ spearce joined	00:04
doener	yes	00:04
johnw	cool, thanks	00:04
doener	faheem: git branch -r	00:04
	← spearce left	00:04
	→ tiglionabbit joined	00:05
	→ smartcat99s joined	00:05
	→ jackdempsey joined	00:05
smartcat99s	is there a way to force git-bisect to checkout a revision? (getting conflicts on Untracked working tree files)	00:06
johnw	stash first?	00:06
	→ davidfetter joined	00:06
smartcat99s	in a bisect of a clean kernel tree?	00:07
	← loinclot_ left	00:07
	← igorgue left	00:08
smartcat99s	for reference: git bisect start v2.6.27-rc3-171-gb635ace v2.6.27-rc2	00:08
j_engelh	git bisect start; git bisect bad <thisone>; git bisect good <thatone>;	00:09
	easier to remember than trying to figure out which of the two arguments to 'start' are bad and which is good	00:09
smartcat99s	I did that	00:09
	Bisecting: 347 revisions left to test after this	00:09
	error: Untracked working tree file 'Documentation/usb/auerswald.txt' would be overwritten by merge.	00:09
	← deavid left	00:14
	← ereslibre left	00:14
Mikachu	why can git bisect good take several refs while bad only can take one?	00:17
j_engelh	why can git remote rm take only one remote?	00:18
Mikachu	at least it's self-consistent ;)	00:18
	← trochala left	00:22
	← schacon left	00:22
	→ trochala joined	00:22
	→ morii_ joined	00:23
	→ toxx__ joined	00:24
	→ johnw- joined	00:24
	→ igorgue joined	00:24
	→ hobodave joined	00:24
j_engelh	> ./vxrandpw -J	00:25
	ujirorimo	00:25
	← priidu left	00:29
	queso\|away → queso	00:30
	← mw left	00:30
	→ doener_ joined	00:30
	← pygi left	00:31
	← toxx_ left	00:31
	→ quix joined	00:31
	← toxx__ left	00:31
	← johnw left	00:32
	← tiglionabbit left	00:33
	← zachinglis left	00:34
	← morii left	00:35
	morii_ → morii	00:35
quix	I want someone to clone my repository and have a working build right away. But with the submodule thing, they have to clone and then do some steps to get the submodules. Is there a way for clone to grab the submodules in one step?	00:35
	→ hobodave_ joined	00:36
faheem	doener: Can you elaborate, please? How do I get the django branch? git branch -r django doesn't appear to do anything.	00:39
	→ gcw\|secution joined	00:39
doener_	"git branch -r" shows the remote tracking branches	00:39
	→ rmh3093 joined	00:39
doener_	you can use them like any other ref for any command	00:40
faheem	doener: I just want to know the command to check out the django branch. I don't actually use git.	00:40
doener_	for exampl, you can show its log, diff it, or even create a local branch starting from it, so you can do your own work	00:40
	well, "git checkout"	00:40
	→ pygi joined	00:40
quix	I guess I'll use the subtree strategy. The submodule system seems undesirable since it requires extra steps which appear needless.	00:41
	← aidy left	00:41
quix	Maybe there is something I'm not understanding -- do people actually use submodules? Or do most people follow the "subtree strategy" outlined?	00:43
	← doener left	00:43
faheem	doener: Sorry, still not sure what I need to do. So I'm doing git clone git://github.com/kwe/passenger-django-wsgi-example.git . Then what do I need to do to see the django branch?	00:43
doener_	"git branch -r" to see the remote branches, and "git checkout $branchname" to check out a branch	00:44
	or "git checkout -b $new_local_branch $starting_point" to create and checkout a new local branch, starting at $starting_point	00:45
	eg. "git checkout -b foo origin/foo"	00:45
	← Paf left	00:46
	→ rezxcers joined	00:46
rezxcers	hey, all, i tried to git clone a repos, but i got error like this:Permission denied (publickey).	00:47
	fatal: The remote end hung up unexpectedly, why?	00:47
faheem	doener_: Well, I guess I don't understand the theory here, but I seem to have the right directory now. What's the recommended intro/tutorial these days?	00:47
	← hobodave_ left	00:47
	← rmh3093 left	00:48
	→ sebell joined	00:49
doener_	faheem: No idea, I don't think I really ever read one. http://git.or.cz/gitwiki/GitDocumentation has a few links	00:49
	→ bloopletech joined	00:49
	← smartcat99s left	00:50
doener_	faheem: the user's manual, git from bottom up, git for computer scientists, git magic, gitcasts, git community book	00:50
bloopletech	I'm relatively new to git, and I'm having a wierd problem. I am working on a Ruby on Rails project, and I'm using GitHub.com as my reopsitory host.	00:50
doener_	faheem: those are the ones I've at least heard something about	00:51
	rezxcers: look like you don't have permission to ssh to that server	00:51
bloopletech	I have plugins in ./vendor/plugins using git clone plugin_url plugin_folder	00:51
	→ tiglionabbit joined	00:51
	hobodave → Guest85694	00:51
faheem	doener_: Thanks.	00:52
rezxcers	yep, i need a SSH key, but i didn't know how to generate it.	00:52
	→ rmh3093 joined	00:52
bloopletech	When I'm getting ready to commit, I do git add . in the project root, then I check the results with git status. But the result is all the files in the plugin being shown as deleted, and new file (plugin folder) in the status results	00:52
hacim	how do I fast-forward a remote tracked branch?	00:53
	← \ask- left	00:53
doener_	hacim: hm? That makes no sense to me	00:53
bloopletech	Usuallly, it sorts itself out if I do git add vender/plugins/... again	00:53
	← Plouj left	00:54
bloopletech	But I'd like to work out why it happens, as it's a bit of a pain.	00:54
doener_	bloopletech: if you add a git repo, that gets added as a submodules	00:55
	s/s$//	00:55
	→ el_isma_ joined	00:55
hacim	doener_: i switched to a branch based off of a remote branch, and git tells me: "Your branch is behind the tracked remote branch 'jrollins/master' by 409 commits, and can be fast-forwarded"	00:56
	→ bogolisk joined	00:57
doener_	bloopletech: please keep the discussion on the channel... http://www.kernel.org/pub/software/scm/git/docs/user-manual.html#submodules	00:57
	← el_isma_ left	00:58
hacim	doener_: I guess i just have to git pull in the branch	00:58
doener_	hacim: that means that you local branch is only missing changes from the remote branch and has no changes that are not in the remote branch.	00:58
	so when you attempt a merge, git will (by default) no create a merge commit, but just update the branch to point to the new stuff	00:59
bloopletech	doener_: So, do you think I should be using submodules instead of just git clone ... inside my project?	01:00
	→ loincloth joined	01:00
doener_	"git pull", "git merge jrollins/master", "git pull --rebase", foobar... they'll all end up fast-forwarding	01:00
	→ pjhyett joined	01:01
	hansengel\|away → hansengel	01:01
	← loincloth left	01:01
	→ loinclot_ joined	01:01
doener_	bloopletech: sounds likely, but so far, I don't really know what you're doing, for example why you're cloning a repo into your working tree, if you already had files in that directory	01:01
bloopletech	doener_: Sorry, I'll try to explain better.	01:02
	← Guest85694 left	01:02
doener_	but maybe it's just my lack of caffeine and the fact that my heart is still going at 180bpm or so... halogen bulb exploded a few minutes ago...	01:02
bloopletech	doener_: I have a project at PROJROOT/ . Inside is PROJROOT/vendor/radiant. PROJROOT/vendor/radiant was created by: git clone radiant_url PROJROOT/vendor/plugins	01:03
	doener_: So I have a git project, and inside it, a cloned copy of another project.	01:04
doener_	yeah, that sounds like a submodule candidate	01:04
bloopletech	doener_: Halogen bulbs can be nasty	01:04
	doener_: And the end result at the moment is crazy stuff happening when I do git add .	01:04
	doener_: Ok, I'll try it out	01:04
	← yofel left	01:07
	→ polpak joined	01:07
quix	It looks like the remotes are only things you add locally, and are not something that's committed?	01:07
doener_	yep	01:07
	→ yofel joined	01:07
quix	ok thanks, I'm starting to get this a little	01:08
bogolisk	what is '^0' as in 'branch_name^0' ?	01:09
j_engelh	tag resolution	01:09
bogolisk	ok! ty	01:10
quix	But it's the other end of the problem of submodules. Submodules are recorded, but then the user has to manually get them after getting the main branch. With remote subtrees, the whole tree is obtained, but the remotes are not recorded and require manual work to sync.	01:10
	I don't see a decent solution for tracking another repository.	01:11
bogolisk	why in git-base: to start, "git checkout $onto^0" was done instead of "git reset --hard $onto^0" ?	01:11
	← rezxcers left	01:11
	← zdennis left	01:12
quix	Is there such thing as a "first-time clone" hook where I can do "git submodule init && git submodule clone" ?	01:12
doener_	bloopletech: you mean s/base/rebase/, right?	01:12
	← jackdempsey left	01:12
bloopletech	doener_: huh?	01:13
j_engelh	bogolisk: "ref^{type}" dereferences ref repeatedly until you cannot dereference anymore, or you get an object of given type, whichever comes first. "ref^{}" means just dereference tags until you get a non-tag. "ref^0" is a shorthand for "ref^{commit}". "commit^{tree}" could be used to find the top-level tree of the given commit.	01:13
	← joydivider left	01:13
doener_	bloopletech: sorry, I meant bogolisk... stupid tab-completion	01:13
bloopletech	doener_: thought so, :)	01:13
j_engelh	straight from sha1_name.c	01:13
bogolisk	doener_: yeah, typo	01:13
	j_engelh: thanx, will re-read 10 times!	01:14
doener_	bogolisk: because reset --hard would not get you a detached HEAD, but instead mess with your current branch	01:14
j_engelh	git-rebase does git checkout $onto^0	01:15
	if it only did git checkout, like doener set, you move to the branch, not the detached commit	01:15
	s/set/said/	01:15
	but it gets more interesting even	01:15
bogolisk	doener_: ic!	01:15
j_engelh	you can tag a tag.	01:15
	→ capisce_ joined	01:15
j_engelh	so, if you use git-tag newtagname oldtagname^0, you make sure you don't tag the tag, but the commit.	01:15
doener_	j_engelh: I did not say checkout _at_all_	01:16
bogolisk	ok, let me try: not do a "reset --hard" cuz it move the ref, but "checkout with a commit sha1" give a detached head...	01:17
bloopletech	doener_: Hmm, the issue with submodules is that the dependancies have to be loaded from the remote repo's, whereas I would like to have the files actually in my repo.	01:17
bogolisk	but where the result of the step's merge will be if head was detached?	01:18
	← yofel left	01:18
doener_	bogolisk: try "cat .git/HEAD" after "git checkout master" and after "git checkout master^0" (or whatever branch you feel like using)	01:19
	bogolisk: in the first case, HEAD is a symbolic ref to the branch, in the second one it's a plain ref to a commit	01:19
bogolisk	doener_: I understand that, but I'm just trying to see what's happending between each replay during a rebase	01:20
doener_	bogolisk: when you do commit, merge, reset, ... and HEAD is a symbolic ref, git dereferences that and updates the branch	01:20
bogolisk	happening	01:20
doener_	when HEAD is not a symbolic ref, HEAD itself is updated	01:20
bogolisk	doener_: ok!!!!	01:21
	doener_ takes the exclamation mark storm as a sign to shut up	01:21
bogolisk	interesting?! so you mean you cannot commit with detached HEAD but merge is ok!	01:21
Mikachu	you can commit with a detached HEAD	01:22
bogolisk	doener_: com'on. (see no exclam)	01:22
	bogolisk is confused	01:22
	← SjB left	01:23
	← quix left	01:24
johnw-	hi bogolisk	01:24
Bryson	i typed a command incorrectly to track a remote branch. i now have a branch NAMED '--track' and when i try to delete it, the branch name is treated as an option and nothing happens. help.	01:25
Mikachu	Bryson: -- --track	01:25
	← capisce left	01:26
	→ Sylia joined	01:26
bogolisk	johnw- hi	01:26
johnw-	bogolisk: I saw you momentarily forked one of my GitHub projects :)	01:27
hacim	when I did a git push to my remote, it pushed all my local branches, how do I 'undo' that and not push those in the future?	01:27
	← loinclot_ left	01:27
bogolisk	johnw- : you mean magit?	01:27
johnw-	yeah	01:27
Mikachu	hacim: it will push all branches that have the same names as ones in the remote	01:27
	hacim: you can specify an exact set in .git/config	01:27
johnw-	i was going to write my own Emacs mode, but then found magit on the same day, and I'm now working with the author to merge our efforts	01:27
bogolisk	johnw-: yeah I tried to publish my own changes on top of marius and thinking cloning yours might be the fastest way :)	01:28
	johnw-: cool	01:28
johnw-	for now I'm keeping my changes as a series of commits that I keep rebasing on his master, so that he can always see my "pending" commits	01:28
	we're still hashing out questions of design and interface	01:28
hacim	Mikachu: it created these new	01:29
bogolisk	johnw-: if you're curious you can take a look at my changes on the branch magit_ld in bogolis/magit repo	01:29
Mikachu	hacim: that should only happen if you use --all	01:29
johnw-	but now I'm excited, he's already halfway to where I wanted to be	01:29
	what's your repo URL?	01:29
Bryson	Mikachu: that worked. thanks. why does that work?	01:29
Mikachu	hacim: or if you configured it that way	01:29
	Bryson: because -- means no more options	01:29
johnw-	Bryson: the -- convention is almost universal in the UNIX world	01:29
hacim	Mikachu: I guess because I've got: push = +refs/heads/*	01:29
bogolisk	johnw-: : git://github.com/bogolisk/magit.git	01:29
Mikachu	which is confusing, because in a lot of commands, it means no more branches ;)	01:30
hacim	Mikachu: how do I remote the remote branches?	01:30
johnw-	oh, I thought you deleted your fork	01:30
	adding you as a remote now...	01:30
Mikachu	hacim: git push remotename :branchtodelete	01:30
bogolisk	johnw-: I did. then I cloned marius and push it on bogolisk/magit. master branch is all his, magit_ld branch is mine	01:30
johnw-	ah, I see	01:31
hacim	Mikachu: thanks	01:32
johnw-	cool, I hope he grabs your changes	01:33
	ah, you're using SPC too, I bound that to scroll-down	01:33
	→ mikem joined	01:35
johnw-	bogolisk: ok, I'm moving my own work to the 'johnw' branch so that 'master' is a mirror	01:35
mikem	hi, i'm trying to configure git to send updates to buildbot on each commit. i'm not sure what to put in the post-receive hook in order to launch this script properly: http://buildbot.net/trac/browser/contrib/git_buildbot.py	01:36
johnw-	mikem: I do this	01:37
	i put a copy of my post-receive (based on that .py) in ftp://ftp.newartisans.com/pub/post-receive	01:39
	bogolisk: ok, I made a branch johnw_bogo that merges you and me together. There were only two changes of your I didn't take: I bound magit-toggle-hs-thing to "h", and I use magit: PROJ as the magit buffer name.	01:40
	ooh, very colorful	01:42
bogolisk	johnw-: I use dark background os I have no idea how the colours look on light bg	01:44
johnw-	well, the non-change text is a bit too light grey	01:44
	← kanru left	01:44
mikem	johnw-: so you replace .git/hooks/post-receive with that python file?	01:45
johnw-	yep	01:45
	and make it +x	01:45
	bogolisk: ok, check out my johnw_bogo branch on github	01:45
mikem	johnw-: ah i see. is it possible to call it from post-receive instead?	01:45
johnw-	i like it, it's pretty	01:45
bogolisk	johnw-: what color do you suggest?	01:45
johnw-	mikem: well, I guess you could	01:45
bogolisk	johnw-: did you try the hide-show thing?	01:46
johnw-	bogolisk: grey is good, just use grey25 instead of grey90 or whatever you're using	01:46
	bogolisk: no, about to	01:46
	i like your minor-mode idea too	01:46
	all of these faces are customizable, right?	01:46
bogolisk	johnw-: yes!	01:46
	customize-group magit	01:46
johnw-	grey50 is good	01:47
bogolisk	johnw-: I think there should be an option to whether to display the logs in the status buffer or not	01:47
johnw-	let me show you a screenshot	01:47
bogolisk	johnw-: ok, default will be grey50 for light background	01:47
johnw-	http://ftp.newartisans.com/pub/magit.png	01:49
bogolisk	johnw-: I like it	01:50
	→ gretch joined	01:51
johnw-	the hiding is perfect	01:51
	← tiglionabbit left	01:51
	← Bryson left	01:51
johnw-	the next thing I want is the ability to start out with everything hidden, and the filenames indented with status prefix, so the buffer looks like the old git.el when all files are hidden	01:52
	bogolisk: I'm not sure how magit-mode's C-c C-v = is better than using C-x v = with VC? I'm rather partial to VC's keybindings. I think it might be better to develop a vc-magit.el than to have a minor-mode	01:53
	→ loincloth joined	01:58
hansengel	Hi guys, I'm trying to set up a local Git server on my Mac with git-daemon. I got the daemon running, but when I try to clone a repo from the client, the server outputs this error:	01:59
	← rmh3093 left	01:59
	← ijcd_ left	01:59
hansengel	'/myrepo.git': unable to chdir or not a git archive	01:59
	here's the command I used to run the server, as user git, group git:	01:59
	→ rmh3093 joined	02:00
Mikachu	the paths can be a bit confusing with git-daemon before you understand what happens	02:00
hansengel	git-daemon --verbose --export-all /Users/git	02:00
Mikachu	you probably want to give --base-path	02:00
hansengel	inside /Users/git, I have a folder called myrepo.git	02:00
	Mikachu: what would that do?	02:00
Mikachu	otherwise, you need to git clone git://blah/Users/git/myrepo.git	02:00
hansengel	aaaah... got it	02:00
	okay, new problem	02:01
	just cloned.. no errors on the server, I can see the request	02:01
	but on the client:	02:01
	fatal: no matching remote head	02:01
	the repo is empty, is this why?	02:01
Mikachu	yes	02:01
hansengel	ah, so I'll try pushing first	02:02
	sweet! think I got it working	02:03
	so now where is the data located on the server?	02:03
	(the file I just pushed)	02:03
Mikachu	uh, in /Users/git/myrepo.git?	02:03
bogolisk	johnw-: do you have a list of colors for the defaults with light bg	02:04
hansengel	yeah, but shouldn't I see the file I just pushed? ran `touch empty`, added, pushed	02:04
Mikachu	see where?	02:04
hansengel	but don't see it on the server in /Users/git/myrepo.git	02:04
Mikachu	of course you don't, it's a bare repo	02:04
bogolisk	johnw-: we might get away with letting the user to choose the prefix: C-c v or C-c C-v	02:04
hansengel	oh, okay.	02:04
	← foca left	02:05
hansengel	one more thing.. is there any way to run this as a startup process, @mac guys?	02:06
	so I don't need to run it via terminal each time	02:06
johnw-	bogolisk: but why have your own minor-mode instead of just writing a VC backend? if you do that, you get to plug in to all of VC's features, like invoking ediff	02:07
	→ swhitt joined	02:07
	← sgrimm left	02:07
bogolisk	johnw-: vc already does git! but it does it the "wrong way"!	02:08
johnw-	how is that?	02:08
bogolisk	johnw-: vc also doesn't have the concept of staging/unstaging	02:08
	→ quix joined	02:08
johnw-	i'll use magit.el for my staging/unstaging needs	02:09
bogolisk	johnw-: AFAIR, vc only commit 1 file at a time	02:09
johnw-	i never use VC for committing, btw	02:09
	only for pulling up diffs, logs, visiting old versions of the file, etc.	02:09
	C-u C-x v ~ is burned into my brain	02:09
bogolisk	johnw- :)	02:10
	so you're attached to the key-binding not the code	02:10
quix	Is it possible for git to run hooks stored in the repository? So for example, after someone clones a repository for the first time, a hook is run?	02:10
johnw-	i'm most attached to the keybinding and the behavior	02:10
	quix: no	02:10
Mikachu	quix: no, that would be a horrible security bug	02:10
bogolisk	johnw-: so what C-x v v should do in a vc-git ?	02:11
johnw-	like if I were to visit a Web page and it ran code... oh	02:11
quix	Right --- I thought of that --- but I want to issue git commands, not arbitrary commands.	02:11
	→ kanru joined	02:11
Mikachu	git commands are arbitrary commands :)	02:11
johnw-	bogolisk: it should offer to run magit-status :)	02:11
Mikachu	git filter-branch --tree-filter 'rm -rf ~'	02:11
bogolisk	hmmmm C-c C-v C-v in the current magit-mode :)	02:11
quix	Is there a git command which will rm -rf / ?	02:11
	→ jackdempsey joined	02:11
quix	oops beat me to it :)	02:12
johnw-	bogolisk: also, magit has no easy way right now of doing what C-x v u does (equivalent of "git checkout -- FILE")	02:12
	→ vbabiy joined	02:12
bogolisk	C-c C-v u	02:12
johnw-	yeah, like I said, I'd rather see you write vc-magit :)	02:12
quix	Here is the problem: I want to use a submodule, but I don't want users to have to "git submodule init && git submodule pull" or whatever after cloning. I want it to just work.	02:13
johnw-	having a non-VC minor-mode which emulates VC just feels, very odd	02:13
	← igorgue left	02:13
quix	I could merge in the tree, but I don't want to --- it would be inappropriate to mix the histories.	02:13
johnw-	and C-x v d should invoke magit-status in vc-magit	02:13
	← bartman` left	02:14
bogolisk	that I like	02:14
quix	So I really don't see good solution to submodules vs merged subtrees.	02:15
	Surely others have this problem? Unless I'm missing something	02:15
Mikachu	quix: i guess you could make a new default branch that has just a script 'runme' that switches branch and does the submodule stuff	02:16
	sort of an ugly solution	02:16
quix	What if the repository had a hook which could only run certain commands, such as "git submodule init" etc	02:16
	→ waltman joined	02:17
johnw-	quix: i find that most developer trees have a preperatory script of some kind, like "autogen.sh". It just becomes the first thing you need to run before you can use the tree.	02:17
	My own project has "./acprep" for doing all this housework stuff.	02:17
quix	Well, I want someone to clone the branch and have everything working immediately --- any manual labor is out of the question.	02:17
Mikachu	that's not how it works, it seems	02:17
johnw-	they can't even bother to read a README that suns "first, run ./RUNME"?	02:17
	s/suns/says	02:17
waltman	I want to undo the entire latest ccommit. I tried "git checkout HEAD^ --", but git complains "Note: moving to "HEAD^" which isn't a local branch" and puts me in a state where I'm not in any branch.	02:18
j_engelh	people are not used to run anything but ./configure	02:18
	← david_koontz left	02:18
waltman	What do I do?	02:18
johnw-	j_engelh: then they should download releases, not source trees	02:18
j_engelh	heh	02:19
	still, even the configure-generating scripts have a somewhat standard name :p	02:19
	→ david_koontz joined	02:19
quix	Everything should be automated. Down with manual labor. It annoys people, and it makes us stupider and monkey-like.	02:19
johnw-	quix: security and convenience are at two ends of a spectrum. Everyone prefers a difference place along this line.	02:20
waltman	sigh	02:22
quix	I think that's a red herring. The reason we are talking about security is because git do multiple clones at once. If it could, there would be no need to run scripts.	02:22
	s/do/can't do/	02:22
johnw-	i agree that automatic fetching of submodules is desirable	02:23
	← david_koontz left	02:23
waltman	how do you undo an entire bad commit?	02:23
johnw-	I'd like to be able to pass "--auto-fetch" to "git submodule add"	02:23
quix	The user is already cloning a repository -- why not clone others?	02:23
	makes no sense.	02:23
johnw-	waltman: git revert, if it's a public commit	02:23
waltman	what's the difference between that and git co HEAD^ ?	02:24
johnw-	quix: what you're suggesting is entirely valid, but it should be made a part of the clone operation itself, not done by external scripts that might do anything	02:24
	quix: maybe bring this up on the mailing list?	02:24
quix	john -- yes that's what I've been meaning.	02:24
johnw-	waltman: git co HEAD^ is not what you want	02:24
	waltman: if it's in your local tree, you want git reset --hard HEAD^	02:24
	← sebell left	02:24
johnw-	waltman: or git reset --soft HEAD^ followed by git checkout -- FILE	02:25
	or even: git commit --amend	02:25
	then undo the change, and then commit again	02:25
	← MattDiPasquale left	02:25
johnw-	(you may have to undo the change before commit --amend, depends on your interface)	02:25
quix	Before I ask on the ML, is there some reason for the current git-submodule behavior? The requirement to run git-submodule init && git-submodule clone ?	02:26
johnw-	quix: I think you're just looking at the historical necessities of an after-thought	02:26
waltman	thanks, think I finally got things back where I wanted :)	02:26
Mikachu	because you can run git clone && git submodule init, but you can't run git clone && oops, don't clone the submodule	02:26
	(that is a guess)	02:27
johnw-	Mikachu: the repo maintainer should be able to indicate in .gitmodules which submodules should auto-fetch on clone, and which shouldn't	02:27
bogolisk	johnw-: something I want to add to magit. from the list of "unstaged" hunks listed in the status buffer, let the user reverse-apply that hunk the the workdir. (basically a selective git-checkout)	02:27
	→ Arrowmaster` joined	02:27
	← tango_ left	02:27
Mikachu	i don't use submodules myself so i don't have an actual opinion :)	02:27
bogolisk	johnw-: s/the the workdir/to the work dir/g	02:27
johnw-	bogolisk: yep, that's exactly what I want too, I've already asked Marius for it	02:27
bogolisk	johnw-: ok	02:27
	→ tango_ joined	02:27
bogolisk	johnw-: don't know if you noticed but I change the apply code to NOT use the .git/magit-tmp	02:28
	← GuyFromHell left	02:28
	← ricky left	02:28
johnw-	so far, I've used submodules for optional external stuff, things only a select few will want to grab in addition to the main sources	02:28
	→ ricky joined	02:28
	← thijso left	02:28
	→ ddollar joined	02:28
	Arrowmaster` → Arrowmaster	02:28
	→ GuyFromHell joined	02:28
	→ thijso joined	02:28
	← jackdempsey left	02:28
	→ telmich joined	02:28
	← krawek left	02:28
	→ krawek joined	02:28
johnw-	but especially for recursive submodules (where the submodule refers to an alternative branch in the repo itself), auto-fetching would be nice, and safe	02:28
	→ tiglionabbit joined	02:28
	→ tewk joined	02:28
	→ xhanjian joined	02:29
	→ jackdempsey joined	02:29
	→ jsimonss joined	02:29
Mikachu	what happens if the submodule is the branch itself? :) then you would get infinite recursion	02:29
johnw-	for recursive submodules, I'd also like the ability to specify making a shallow clone instead of a regular clone	02:29
	Mikachu: i avoid that :)	02:29
	usually a recursive submodule refers to a so-called "empty branch", a data side-band	02:29
quix	I don't quite understand the need to garbage collect. It's basically a defrag of the "git filesystem". Surely there is an algorithm which eliminates this need via piecewise application, such as how UNIX filesystems work?	02:30
johnw-	like the way the git project includes generated documentation	02:30
	quix: it's not a defrag	02:30
	quix: there is no fragmentation where immutable objects are concerned	02:30
Mikachu	if anything, gc causes fragmentation :)	02:30
quix	Well I was speaking metaphorically -- nonetheless it can be eliminated, no?	02:30
johnw-	the reason for "gc" is that if you delete a branch, at some point its objects should get cleaned up	02:31
j_engelh	git prune defragments the gitspace :)	02:31
johnw-	and if there's already inside of a pack, you don't want git repacking after every reflog expiration	02:31
Mikachu	you don't want to rewrite a 500MB packfile every time you git add myfile.c	02:31
	→ jtong joined	02:32
quix	johnw, but can't this be done by git? There's got to be an acceptable algorithm to do it incrementally.	02:32
	→ zdennis joined	02:32
Mikachu	quix: there is already git gc --auto that most commands invoke for you	02:32
johnw-	quix: why does this even bother you?	02:33
	quix: I'm perfectly content with git gc --auto	02:33
quix	It bothers me because it's something that's unnecessarily put on the user.	02:34
johnw-	it's the way git is setup out of the box	02:34
quix	git could have easily handled this and we would have avoided the conversation.	02:35
johnw-	i was happy to avoid it before :)	02:35
quix	I guess philosophical flaws bother me unnecessarily.	02:35
johnw-	i think you may be inveighing about issues that aren't actually real problems, once you've used git for a while	02:35
Mikachu	quix: you aren't content with gc --auto?	02:36
quix	Mikachu: I would only be content if the issue were solved permanently by removing the need for gc commands altogether. But I agree that the problem is mostly philosophical, not practical.	02:38
Mikachu	that makes almost no sense at all	02:38
johnw-	i think he's saying that git-gc makes him feel all icky inside	02:38
quix	Well, it's sort of like a noise in your car engine which doesn't really make a difference, but shouldn't be there.	02:39
Mikachu	you don't know how packs work in git, do you?	02:40
quix	Aha, that is exactly the point. I shouldn't have to know how packs work.	02:40
	← gcw\|secution left	02:41
Mikachu	well, if you don't know why gc is needed, you shouldn't make any statements about whether it's needed	02:41
johnw-	Mikachu: he's just trolling now, wishing for a perfection that meets his idea of how things should be	02:41
Mikachu	(i don't think using the word 'trolling' is fair here)	02:42
	→ bartman joined	02:42
johnw-	he's not making idle comments in order to poke us into defending something?	02:42
Mikachu	no i don't think so	02:43
quix	My point isn't a very big one. It's been artificially enlarged since that's mostly what I've talked about so far.	02:43
johnw-	ok, I retract then	02:43
quix	My apologies.	02:43
	But the point is nonetheless not an idle point. I could imagine a future git implementation which would remove the need for anyone to care about calling git-gc. That would be an improvement.	02:44
	← fafhrd left	02:44
	→ fafhrd joined	02:45
	← swhitt left	02:45
	← zawir left	02:47
johnw-	quix: I apologize as well, but recommend further research on the question.	02:50
	← cbreak left	02:51
	→ zdennis_ joined	02:54
	← quix left	02:55
	← pygi left	02:59
	→ pygi joined	03:00
bogolisk	johnw-: do you know emacs well?	03:11
	← zdennis left	03:11
johnw-	yes, very	03:11
	ever used Eshell?	03:12
bogolisk	johnw-: how do define a custom for a key sequence?	03:12
johnw-	bogolisk: key mappings are not customizable	03:12
bogolisk	johnw-: try it a long time agot	03:12
	ago	03:12
	johnw-: then it would a string I guess	03:13
	← LiamH left	03:13
bogolisk	johnw-: I want to define a keyma with just the last keystroke. let the user select the prefix they want	03:13
	johnw-: "C-c C-v" vs "C-c v"	03:13
	← nn left	03:14
johnw-	you could customize which keymap variable your bindings are applied to	03:14
	example: (define-key mode-specific-map [(control ?v)] 'magit-mode-map)	03:15
	all of your bindings ride on `magit-mode-map' as single keybindings	03:15
	the link into the C-c space would happen using that define-key command	03:15
	→ paltman joined	03:16
johnw-	C-c v is reserved for the user, btw	03:16
bogolisk	johnw-: I mean C-x v	03:16
johnw-	you are still going to make it a separate minor-mode that rides on top of VC's keybindings? may I ask why? Do you not like the VC architecture?	03:18
bogolisk	johnw-: it's more like (define-key magit-mode-map (kbd "=") 'magit-diff-current-file) .... (define-key magit-minor-mode-map (read-kbd-macro custom-key-squence) magit-mode-map)	03:19
	johnw-: I'll have to re-read vc code, but the last time I did the front-end/back-end link was too limiting.	03:20
	btw, vc-git.el already exists	03:20
johnw-	yes, but it doesn't call magit-status on C-x v d :)	03:20
	but perhaps you're right	03:20
	← desrt left	03:20
johnw-	i use C-x C-g to run magit	03:20
	→ jcummins joined	03:21
	→ xshelf joined	03:21
	→ boink__ joined	03:22
	← tcoppi left	03:23
	→ sfullenwider joined	03:24
jcummins	I have a repo at /proj and some open source libraries located at /proj/lib/. One of the libraries was recently updated. Is there an easy way to update my repo with the latest code from their subversion repository? Would I just 'cd /proj/lib/foo ; git svn clone http://svn.foo.com' ?	03:24
johnw-	cd /proj/lib ; git svn clone URL foo	03:25
jcummins	ah. cool. Thanks johnw-	03:25
xshelf	anyone with git-p4 background here?	03:26
	I am trying to figure out, how to avoid "$Id:/some/file/path" from getting replaced by "$Id$" when I do a 'git-p4 clone //depot/some/project'	03:27
	→ BrianTheCoder joined	03:29
johnw-	i'm afraid i have no idea on that one xshelf	03:32
	← Oompa left	03:33
	← vbabiy left	03:33
	← cmarcelo left	03:33
	← raz left	03:34
	boink__ → raz	03:34
xshelf	let me start reading the best documentation... the code itself!	03:34
	I am trying to avoid learning Python, looks like there is no alternative... why didn't people stick to good old perl :)	03:36
	→ ben_h joined	03:36
johnw-	egads, please no	03:37
offby1	I've used both; Python certainly makes more readable code, at least for small-scale stuff	03:39
	once it gets big and complex, it doesn't much matter what you use :-\|	03:39
xshelf	johnw-: Are you John Wiegley (seen you on emacs lists)	03:39
bogolisk	johnw-: I've add the ability to customize the key prefix for magit minor-mode. You can choose "C-x v" if you want:)	03:41
xshelf	not getting into any flame war, just trying to make up my mind. I have jumped into learning Python and got out more often than I can count! I just feel I can accomplish almost everything in perl	03:42
	← jackdempsey left	03:45
johnw-	xshelf: yes, I am	03:46
	xshelf: Ruby might be a better fit	03:46
	→ schacon joined	03:46
	polpak bites his tongue.	03:46
johnw-	bogolisk: I think I'll stick with vc-git, but thanks	03:46
	offby1 bites polpak's tongue	03:47
johnw-	bogolisk: i'll keep an eye on magit-mode, though, will probably switch later on	03:47
polpak	eh, kinky..	03:47
offby1	salty.	03:47
	→ spearce joined	03:47
johnw-	WRONG	03:47
xshelf	johnw-: I am so confused, when I decide to learn a language, I buy the books (when I put my money, I am forced to learn). It is with Python that I am wavering! Languages are becoming like editors or dscm. They all almost have the same features, with minor pros/cons. Very difficult to choose. Waiting for parrot to run python/perl on same vm	03:49
johnw-	i've been using Python for about 10 years now, and I like it very much. But it is not my ideal language, it's just very useful	03:50
	conversely, I love Common Lisp to death, but finding almost no practical uses for it	03:50
	so, I use Python for my scripting, and write code for Emacs to scratch at least a part of my Common Lisp itch	03:51
	← spearce left	03:51
	← kukks left	03:52
	← hansengel left	03:52
xshelf	The only motivation for me to learn Python is to hack on git-p4/bzr/hg. I find it really odd, because I like some tools, I have to learn a whole new language to be able to hack on it. Tomorrow, some good tool will be in Ruby/lua/haskell... where is an end to this language learning!	03:53
	→ Beket joined	03:53
johnw-	are you a programmer? is there ever a possible end?	03:53
xshelf	I am a programmer and I find myself learning languages rather than solving complex issues	03:54
	← flaguy left	03:54
johnw-	it's a balance	03:54
	← mithro left	03:54
	→ alenoosh joined	03:55
johnw-	i don't code in Ruby yet, but I'm 1/3 of the way through Programming in Ruby	03:55
xshelf	sorry for all the off topic tal on #git, just that I find interesting people here	03:55
johnw-	and I really want a professional reason to learn Erlang, so far no personal reasons	03:55
	← schacon left	03:55
xshelf	i guess I will just have to accept the fact and learn them on demand	03:56
johnw-	yep	03:56
bogolisk	johnw-: I saw that Marius merged your changes	03:56
johnw-	in the last three years I had to learn both ASP.NET and J2EE, for different projects. How I would love to settle on one piece of technology that just works and does it all, but such is not life.	03:56
	bogolisk: he merged some, but not all	03:56
bogolisk	johnw-: my branch (magit_ld) is already "public" on git hub, should rebase or merge (with his new head)	03:57
johnw-	xshelf: you did inspire me to give another look at http://ecls.sourceforge.net/, thanks :)	03:57
bogolisk	should I...	03:57
johnw-	bogolisk: unless people make branches based on your branch, rebasing is ok	03:57
	→ mithro joined	03:57
	→ gcw\|secution joined	03:57
johnw-	what will happen is that when they fetch, it will just reset their bogolisk/master_ld	03:57
bogolisk	johnw-: "people" = you? so is rebase ok	03:57
johnw-	yeah, I'm not branching based on your code	03:58
bogolisk	johnw-: ok	03:58
johnw-	I'll just tear down my johnw_bogo branch and re-merge it	03:58
bogolisk	johnw-: ok, ty	03:59
johnw-	interesting, your rebase didn't change your commits	04:00
	i just fetched your latest, and all before HEAD~3 are the same commit ids	04:00
sfullenwider	I have a situation where there are three coders working on a project (with me as one of them). None of us want to share our repositories with the outside world via git. Is it possible to set up repositories that can be fetched by only specific people?	04:01
johnw-	sfullenwider: set it up on a private server that you can all ssh to	04:01
	also, GitHub sells private accounts	04:01
sfullenwider	is that the only option?	04:01
johnw-	you could e-mail patches to each other	04:01
	git-daemon has no authentication	04:01
	you could probably use HTTP Digest authenication with the WebDAV support, but then you'd lose the niceties of the git-daemon transfer protocol	04:02
	git+ssh is by far the simplest	04:02
	lastly, you could each pull from each other's repositories, directly off of each other's machines	04:02
	→ imyousuf joined	04:02
xshelf	johnw-: You have always inspired me (I had exchanged mails with you a while ago, sure you will not remember)	04:03
johnw-	what's your name?	04:03
sfullenwider	johnw-: how could we pull it directly off of each others machines? by providing ssh access to each other?	04:03
	← frq1_dave left	04:03
xshelf	johnw-: dhruva	04:03
johnw-	sfullenwider: yes, or you could use git-daemon by restrict access via firewall	04:04
	s/by/but	04:04
sfullenwider	ah	04:04
	→ alb joined	04:04
sfullenwider	ok, tnx	04:04
jcummins	hmmm.... still having some SVN problems here. I'm trying to get a subfolder of my project to track an SVN repo.	04:05
johnw-	what kind of problems?	04:05
jcummins	so I go to /proj/lib, type 'git svn clone URL foo'	04:05
	that works fine	04:05
	I commit my changes and push to my server	04:05
	← sfullenwider left	04:06
jcummins	then I try to pull the repo from another location	04:06
	← albertito left	04:06
jcummins	it does not contain the SVN repo that I cloned earlier.	04:06
	→ tcoppi joined	04:06
jcummins	Does that make sense?	04:07
johnw-	yes	04:07
	I think you want to use git submodules	04:07
	so that after cloning, your user type "git submodule update --init", and it will fetch a copy of the SVN repo mirror, once you make that mirror accessible to them	04:07
	xshelf: wow, you wrote me on 16 Oct 2002, a while back now!	04:08
	← bts- left	04:08
johnw-	nice to meet you on IRC :)	04:08
xshelf	johnw-: Amazing, you are well organized, I am happy to meet you too	04:09
	→ srid joined	04:09
	→ tjafk1 joined	04:09
johnw-	yes, that time you wrote me was a very hectic time in my life indeed	04:09
	sorry I was not more available	04:09
jcummins	johnw-: Interesting. So every time I want to clone my repo, I'll have to go into that folder after cloning and type 'git submodule update --init" ??	04:10
johnw-	jcummins: the submodule command is run at the top-level	04:10
	here's what you do	04:10
	after you make your git-svn mirror in lib/proj, you type:	04:11
	git submodule add PUBLIC-URL lib/proj	04:11
	where PUBLIC-URL is a publically accessible place you are going to host a copy of your git-svn mirror	04:11
	then you check in .gitmodules and lib/project	04:11
	lib/proj	04:11
	but the commit of lib/proj will be a submodule reference, not the whole directory tree	04:11
xshelf	not a problem, I started learning things and got more closer to free source development. Hacked on samba on vms, did some minor contributions to mercurial, made some noise in emacs	04:11
johnw-	then when others clone your project, they will run submodule update --init, which fetches the git-svn mirror from PUBLIC-URL and populates lib/proj on their side	04:12
jcummins	I see	04:12
	← CodeOfficer left	04:12
johnw-	if your Makefile depends on having lib/proj pulled down, you could add a Makefile rule to run git submodule update --init	04:12
jcummins	yeah, I think that's what I'll have to do.	04:12
	I was hoping it would be an easier process.	04:12
johnw-	have the Makefile depend on a file that you know is in lib/proj	04:13
	jcummins: the alternative is copying sources into your project, but I think this is the "right" now	04:13
	here's hoping that git-submodules mature in their interface	04:13
	jcummins: btw, I do exactly what I've suggested to you in my own project	04:13
	→ cedricv joined	04:14
	offby1 has a Makefile rule that simply runs "git clone" ... can't remember why	04:14
jcummins	hmmm.... I'm wondering if I want to just copy all the code and add it manually.	04:14
	← eikonos left	04:14
jcummins	I'll lose the history, but in this case it doesn't really matter.	04:14
johnw-	jcummins: just for the convenience?	04:14
xshelf	the lack of easy infrastructure to extend git is the one of the few shortcomings other dscm complain about	04:14
johnw-	your Makefile can automate the pulling	04:14
bogolisk	johnw-: to reverse-apply a hunk to the workdir, should I use git-apply or just ...diff?	04:15
johnw-	bogolisk: i'd use git-apply	04:15
bogolisk	johnw-: fine, buy why?	04:15
johnw-	well, diff won't do it, and patch is an external tool	04:15
jcummins	johnw-: the makefile can automate the process, yes, but I'll still be dependent on their repository being up.	04:16
bogolisk	johnw-: ok:) I mean s/diff/patch/	04:16
johnw-	jcummins: use GitHub to host your mirror	04:16
	jcummins: unless this is private code	04:16
	I host all of my mirrors on http://git.newartisans.com	04:16
	there are several of them by now	04:16
	bogolisk: git-apply is pretty much == patch, but it's available when git is	04:16
bogolisk	johnw-: ok.	04:17
johnw-	bogolisk: i merged in your latest 3 commits, thank you	04:17
bogolisk	johnw-: for the xxx-thing-at-point commands, would it be better to use "keymap" text attribute instead?	04:18
	johnw-:)	04:18
j_engelh	johnw-: uhuh	04:18
johnw-	bogolisk: for a minor-mode, yes, for a major-mode, no	04:18
j_engelh	cgit hm	04:18
bogolisk	johnw-: uh... ok, buy why?	04:20
johnw-	well a major-mode by definition governs the modality of the buffer, so using keymap text properties is just splitting hairs	04:20
	for a minor-mode, though, it's a nice way to localize changes in local key behavior based on content	04:20
	← paltman left	04:21
bogolisk	johnw-: ok, I'll think about it. I honestly have no opinion on this minor vs major thing.	04:21
johnw-	the elisp info manual has good sections on the difference	04:22
	→ igorgue joined	04:22
johnw-	minor modes are like auxiliary features that might be applicable within multiple major modes	04:23
	like VC	04:23
bogolisk	johnw-: I just thing that for stage/unstage in the status buffer should use the same keystroke (let's say s). If the hunk was a unstaged one then it would be staged and vice-versa! An easy way to do it is to have different keymap text attribute for the hunks in the staged section vs the one in the unstaged section.	04:25
	s/thing/think/	04:25
	← tjafk left	04:25
	← bloopletech left	04:25
johnw-	bogolisk: i agree that it should use the same key	04:25
	bogolisk: and I think it should be "s"	04:26
	bogolisk: in which case the command should be name magit-toggle-hunk-staged	04:26
	there are already text properties to identify if it's a staged or unstaged hunk	04:26
bogolisk	johnw-: it would be magit-toggle-thing-at-point-state because if it was the diff header then the whole diff would be staged/unstaged.... different keymap text properties would be simpler :)	04:27
johnw-	ah, you're right	04:28
	well, perhaps you are right then	04:28
	i bow to the rationality of fresh eyes :)	04:28
bogolisk	johnw-:)	04:28
	→ gretch` joined	04:29
	→ rdavila joined	04:30
	← warthog9 left	04:30
bogolisk	johnw-: I can code that up but it'd be nice if Marius was convinced to do it because he might have strong opinion about the UI in the status buffer	04:30
	johnw-: i.e. It'd be nice if you can convince Marius to code that up (single key for stage/unstage) :)	04:31
	bogolisk sleepy	04:31
	bogolisk go to bed	04:31
	← bogolisk left	04:31
johnw-	night bogolisk	04:31
	maybe you will dream of basilisks	04:31
	the nice thing about s/u is that it prevents you making a mistake if you're in the wrong part of the buffer	04:32
	← bdiego left	04:33
	← xshelf left	04:35
	→ jmettraux joined	04:36
	← queso left	04:40
	← bronson left	04:42
	← madewokherd left	04:46
	← alenoosh left	04:47
	← gretch left	04:47
	→ alenoosh joined	04:47
	→ warthog9 joined	04:53
	← srid left	04:55
	→ SJS joined	04:59
	← SJS left	05:01
	→ SJS joined	05:01
	→ srid joined	05:02
	gretch` → gretch	05:02
	← srid left	05:02
	← cedricv left	05:03
	← alenoosh left	05:10
	→ ph^ joined	05:12
	← rmh3093 left	05:12
	← rdavila left	05:15
	→ schacon joined	05:15
	← schacon left	05:15
	← perezd88 left	05:15
	→ rmh3093 joined	05:15
	← mithro left	05:15
	← ThunderChicken left	05:16
	→ mithro joined	05:16
	SJS → ThunderChicken	05:17
	→ alenoosh joined	05:18
	→ robinr joined	05:19
	← igorgue left	05:19
	→ igorgue_ joined	05:19
	igorgue_ → igorgue	05:19
	→ Aikawa joined	05:22
	→ sgrimm joined	05:23
	← ruphy left	05:24
	→ adante joined	05:29
	← jstad left	05:30
khc	what's this para-walk branch found on an email 2 years ago?	05:31
	→ eikonos joined	05:32
	→ perezd88 joined	05:35
Mikachu	apparently something that was discarded for some reason :)	05:37
khc	I was hoping that it would help over nfs	05:37
	← robinr left	05:38
	khc wonders who's Junio C Hamano here	05:39
Mikachu	gitster	05:39
khc	gitster: why did you drop para-walk ;-)	05:39
	← xinming left	05:42
	→ xinming joined	05:42
	→ bugfux joined	05:48
	← xinming left	05:48
	→ jackdempsey joined	05:48
	← jmettraux left	05:48
	← Beket left	05:48
	→ xinming joined	05:55
	← LostInTheWorld left	05:56
	← krawek left	06:00
	← zdennis_ left	06:05
	→ mstormo joined	06:05
	→ schacon joined	06:06
	→ kixx joined	06:06
	← jcummins left	06:11
	← alenoosh left	06:12
	→ xinming_ joined	06:15
	→ l0verb0y joined	06:21
	→ daaku joined	06:22
	← ciaran left	06:23
zjason_	git-cat-file -e can tell whether a commit exist. How to do this in gitweb interface?	06:23
daaku	i did something stupid - i commited - pushed - reset - commited. i know no one else has pulled after me, can i force the push and get around the 'rejected - non fast forward' error?	06:23
	nm. force. so obvious!	06:24
	← aquabu left	06:26
resmo	i heard about that git can act as a svn server, git running on server, clients can use their svn-clients. But didn't find any information about on the web.	06:27
	← xinming left	06:29
johnw-	it can run as a cvsserver, not svn	06:29
	← trochala left	06:32
	← mikem left	06:34
	→ trochala joined	06:35
	→ capisce joined	06:36
	← daaku left	06:37
	← igorgue left	06:39
	→ xinming joined	06:39
	← tcoppi left	06:43
resmo	johnw-: ah ok. Got this wrong	06:43
	← perezd88 left	06:48
	← meyering left	06:49
	→ meyering joined	06:49
	← capisce_ left	06:49
	→ Adri2000 joined	06:49
	← xinming_ left	06:50
	→ BabelO joined	06:50
	← schlort left	06:52
	alb → albertito	06:58
	albertito → Maxy	06:59
	Maxy → albertito	06:59
	← bobesponja left	07:00
	← trochala left	07:00
	→ trochala joined	07:01
	→ CodeOfficer joined	07:01
	→ alenoosh joined	07:03
	← mstormo left	07:04
	← Aikawa left	07:04
	← mithro left	07:04
	← bartman left	07:04
	← kanru left	07:04
	← mtodd left	07:04
	← rtyler left	07:04
	← greedo left	07:04
	← univate left	07:04
	← pflanze left	07:04
	← dancor left	07:04
	← comp left	07:04
	← corecode left	07:04
	← xjjk left	07:04
	← hacim left	07:04
	← Magnus` left	07:04
	← johannes left	07:04
	← stick left	07:04
	← metze left	07:04
	← tmm1 left	07:04
	← pmcnabb left	07:04
	← GyrosGeier left	07:04
	← cods left	07:04
	← ruskie left	07:04
	← yann left	07:04
	← tarbo_ left	07:04
	← eMBee left	07:04
	→ GyrosGeier joined	07:05
	→ hacim joined	07:05
	→ cods joined	07:05
	→ metze joined	07:05
	→ Magnus` joined	07:05
	→ rtyler joined	07:05
	→ dancor joined	07:06
	→ corecode joined	07:06
	→ bartman joined	07:06
	→ tarbo joined	07:06
	→ pmcnabb joined	07:06
	→ mstormo joined	07:06
	→ comp joined	07:06
	→ xjjk joined	07:06
	→ RichGuk joined	07:08
	→ eMBee joined	07:09
	→ stick joined	07:09
	→ mtodd joined	07:10
	← neoeinstein left	07:11
	→ jcummins joined	07:11
Arjen	What would be the 'cleanest' way to get the sha1 of the current branch head?	07:13
	→ mithro joined	07:13
	→ ruskie joined	07:13
thiago_home	git rev-parse EHAD	07:14
	HEAD	07:14
Arjen	Of course, thanks :-)	07:14
	← CIA-2 left	07:16
	→ kanru joined	07:18
	→ univate joined	07:18
	→ greedo joined	07:18
	→ pflanze joined	07:18
	→ yann joined	07:18
	→ Aikawa joined	07:19
	→ FunkeeMonk joined	07:20
	← pjhyett left	07:21
	→ johnw_ joined	07:22
	← davidfetter left	07:25
	→ ciaran joined	07:25
	← jackdempsey left	07:27
	→ johannes joined	07:31
	→ tmm1 joined	07:32
	← johnw- left	07:32
	→ jackdempsey joined	07:33
	← jackdempsey left	07:36
	→ nutrimatt joined	07:37
	→ redondos joined	07:40
	← machrider left	07:41
	→ bartocc joined	07:43
	→ cilly joined	07:46
	→ priidu joined	07:48
	→ reel_ joined	07:50
	← janneke left	07:55
	→ janneke1 joined	07:55
	→ redondos_ joined	07:57
	← bartocc left	07:58
	→ bartocc joined	07:59
	→ fgeek joined	08:00
	← sgrimm left	08:02
	← ciaran left	08:04
	jdq_ → jdq	08:04
	→ sgrimm joined	08:05
reuss	when installing git 1.6.0 is git-sh-setup supposed to go away also? .. if yes, how do you get the require_work_tree and cd_to_toplevel functionality in your scripts?	08:08
Mikachu	reuss: s/git-/git / in all your scripts	08:08
reuss	yeah, i did that - but it doesn't work..	08:09
	i still get that those two commands are not found ..	08:09
	git sh-setup; require_work_tree; cd_to_toplevel; doesn't seem to work in my script	08:10
	← redondos left	08:11
	redondos_ → redondos	08:11
reuss	and i get the right git version in my script	08:11
osfameron	I'd like to push a git repo from my laptop to the server	08:11
	I've done git init-db on my server	08:11
	but the manpage for "git-push" doesn't seem to refer to this use case (from a quick skim, do I need to rtfm this in more depth?)	08:12
	→ HFactor joined	08:13
Mikachu	reuss: maybe you should read the manpage	08:13
HFactor	hi all,	08:13
	do any have a small presentation (slide) for basic git ...	08:13
mithro	hey guys, I'm having problem with the git-cvsserver bridge not working properly - it's adding empty commits	08:13
osfameron	ah, I have to "git remote add origin" ?	08:14
reuss	Mikachu: yeah, did that - theres a "right" way of sourcing it it seems .. and i was doing it wrong :) .. . "$(git --exec-path)/git-sh-setup" works..	08:14
	← polpak left	08:15
	→ icwiener joined	08:15
	← HFactor left	08:16
	← FunkeeMonk left	08:17
	← soultan left	08:17
	→ bartocc_ joined	08:17
	← l0verb0y left	08:17
mithro	I'm using git 1.5.6 and there is nothing strange in the gitcvs log	08:17
	← khc left	08:17
mithro	anyone got any idea how to debug this problem?	08:19
	← Tv left	08:20
	→ snitko joined	08:20
	→ jmesnil joined	08:20
	← lu_zero_ left	08:21
	→ offby1` joined	08:24
	← kixx left	08:24
	→ soultan joined	08:25
	← cods left	08:26
	← schacon left	08:26
	→ Shilin joined	08:27
Shilin	hi, I have a git repository and inside it there is another git repository. Seems like the sub-git repository cannot be added and commited. How do I handle that ?	08:29
up_the_irons	Shilin: that's correct, git will not add another .git directory within a repo	08:30
	Shilin: I have no idea how to handle that though, sorry	08:30
Shilin	anyway, thank you. I have been struggling with that and only now I have noticed that subfolder is not commited. Now I know the problem and I will use a different approach.	08:31
	← RichGuk left	08:32
	← bartocc left	08:33
	← bugfux left	08:34
	→ cods joined	08:34
	→ volante_ joined	08:35
	→ jmesnil_ joined	08:35
	→ jtoy joined	08:36
	→ m3y3ring joined	08:36
	← CodeOfficer left	08:37
jtoy	dumb question, how can I know what revision i am on?	08:38
	← johnw_ left	08:38
	← cilly left	08:39
	← offby1 left	08:39
	← jmesnil left	08:39
jtoy	is there first line of git show the same thing as current version?	08:40
osfameron	jtoy: doesn't sound like a dumb question. (And I don't know either, but would like to)	08:41
	← meyering left	08:41
Mikachu	if you want only the number, git rev-parse HEAD	08:41
	→ f0rkit joined	08:42
jtoy	ahhh, nice, having a hard time on google finding that	08:42
osfameron	ah, I woul dnever have thought of that :-)	08:43
	I did try "git info" on the lines of "svn info" but no joy	08:43
Arjen	It's one of those commands you overlook	08:43
	(I had the same question earlier :-)	08:43
	m3y3ring → meyering	08:44
	← jcummins left	08:46
	→ danten joined	08:56
	→ schlort joined	08:56
	→ Tv joined	08:59
	→ samjam joined	09:00
	← samjam left	09:00
	→ advany joined	09:00
	→ samjam joined	09:01
samjam	how can I list only tags of a branch?	09:01
thiago	there's no such concept	09:02
	tags don't belong to a branch	09:03
	what you want is to find all tags that are in the past history of a given branch	09:03
	I don't know of any commands for doing that, but it can be scripted	09:04
osfameron	git+- # you can do everything, but it might hurt	09:04
thiago	git tag \| while read tag; do test $(git merge-base $tag master)== $(git rev-parse $tag) && echo $tag; done	09:05
	change "master" to the branch you want	09:05
	and add a space before ==	09:05
	samjam pauses to grok that command	09:05
wildfire	thiago, should it be 'git tag -l' ?	09:06
thiago	wildfire: same thing	09:06
	samjam: Did you get the merge-base and rev-parse ?	09:06
samjam	I'm just "getting" it	09:06
	thiago knows of no other way of telling whether $commit1 is an ancestor of $commit2	09:07
	← sverrej_ left	09:07
thiago	though such a command would be extremely useful, since merge-base is more expensive than necessary if there is no ancestor relationship	09:07
samjam	so if the best merge between tag and master is the tag, then the tag is there...	09:07
	← SeJo left	09:07
samjam	thats cool, thanks	09:08
thiago	"is $commit1 an ancestor of $commit2" == "is the merge-base of $commit1 and $commit2 equal to $commit1"	09:08
	that's a property of a merge-base	09:08
samjam	I shall credit you with that jagh (just-another-git-hacker)	09:09
	→ johnw joined	09:09
doener_	test -z "$(git rev-list -1 ..$tag)"	09:09
wildfire	pasky, is your gitweb 'graphic log' integrated in to mainline at all?	09:09
doener_	though I doubt that that is any faster	09:09
thiago	doener_: what's that supposed to do?	09:10
	that will work if $tag is in the future of HEAD, but not if it's branching from the past	09:10
doener_	same thing as the merge-base thing (well, for HEAD instead of master)	09:10
	← BarryCarlyon left	09:11
thiago	it works if $commit2 is a parent of $commit1, but not if $commit1 and $commit2 are not ancestors of each other	09:11
samjam	I need only tags which are direct ancestors, not also in the future	09:12
thiago	exactly	09:12
	that's what the command I gave you does	09:12
doener_	hm? "a..b" is empty exactly when b is an ancestor of a	09:13
	← eddyp left	09:13
samjam	it's brilliant; just so you know it's part of an srpm builder which extracts "pristine" source for the last tagged release and patches since that	09:13
	hmmm.... although it seems the darn'd samba team branch and then tag the branch; so it can't help there... ah well...	09:15
pasky	wildfire: no	09:15
	wildfire: I have some patches upcoming that will at least make it easier to integrate	09:16
doener_	thiago: what do you mean by "branching from the past"?	09:16
thiago	doener_: a..b is also empty when a isn't an ancestor of b, but there's a third commit c such that c is ancestor of both a and b	09:16
	i.e. merge-base a b == c	09:16
doener_	no	09:16
thiago	yes	09:16
	← trochala left	09:17
thiago	$ git rev-list -1 ..networkaccess-branchpoint \| wc -l	09:17
	0	09:17
	what do you take that to mean?	09:17
doener_	that networkaccess-branchpoint is an ancestor of your HEAD	09:18
thiago	$ git rev-parse HEAD	09:18
	0121852471196486340d20cdcc53b78dbe3dd52a	09:18
doener_	http://git.pastebin.com/m1786adee	09:18
thiago	$ git rev-parse networkaccess-branchpoint	09:19
	632ee4fce4e7505c77627362cd306e30c42aeca8	09:19
	$ git merge-base HEAD networkaccess-branchpoint	09:19
	9c3495b69f719005fbda1bfd860c4d7c7b3c4cd3	09:19
	what does that tell you now?	09:19
samjam	I guess I could look for where the common ancestor is the tag or a parent of the tag	09:19
pasky	for the love of god, please teach git branch --contains	09:21
doener_	thiago: that tells me that you merged networkaccess-branchpoint into your HEAD, but that networkaccess-branchpoint itself is still pointing to the commit _before_ that merge commit, I'd say	09:21
pasky	instead of heavy-grade plumbing :)	09:21
	though, hmm	09:22
thiago	pasky: it's the inverse of git branch --contains	09:22
	← m-vo left	09:22
pasky	well, its output is the inverse	09:22
	not the function	09:22
	but yes, this could be more elegant	09:23
doener_	thiago: so networkaccess-branchpoint _is_ an ancestor of HEAD, and the merge-base trick failed	09:23
	but that makes no sense to me either	09:24
thiago	doener_: ah, found my mistake	09:24
	rev-parse returned the SHA-1 of the tag, not of the commit the tag points to	09:24
doener_	heh :-)	09:25
thiago	$ git rev-parse networkaccess-branchpoint^{commit}	09:25
	9c3495b69f719005fbda1bfd860c4d7c7b3c4cd3	09:25
	so, yeah, it is an ancestor	09:25
pasky	:)	09:25
thiago	pasky: git branch --contains will tell you which branches contain the tag	09:25
	samjam: wants the tags that are contained by a given branch	09:25
samjam	adding ^{commit} to your jagh gave me what I expected	09:26
thiago	samjam: yep	09:26
doener_	ok, so we agree that "a..b" being empty also signifies ancestry, right?	09:26
thiago	hint: put that inside ''	09:26
pasky	thiago: I meant to just use git branch --contains instead of the merge-base stuff	09:26
	it's slower and less elegant, yes	09:26
	but more understandable :)	09:26
thiago	doener_: agreed	09:27
	pasky: git branch --contains $tag \| grep -q $branch ?	09:27
pasky	kind of	09:27
samjam	So I have:	09:27
	git tag -l "$PATTERN" \| while read tag; do test "$(git merge-base $tag HEAD)" = "$(git rev-parse $tag^{commit})" && echo $tag \|\| echo .; done	09:27
	→ sverrej joined	09:28
thiago	pasky: it's a lot less efficient, since for each tag it'll scan all branches	09:28
pasky	as I said	09:28
	I think git rev-list -1 ^HEAD $tag is more reasonable than merge-base	09:28
thiago	yeah, but I thought you meant "a little less efficient", not "a lot less efficient"	09:28
	→ trochala joined	09:28
pasky	shrug :)	09:28
thiago	let me put it this way:	09:29
	← Sput left	09:29
thiago	if you're writing a pre-update hook script that determines if a forced push is allowed or not, you'd use git rev-list -1 $orig $new ?	09:29
	→ Sput joined	09:30
	→ deavid joined	09:30
pasky	thiago: no, I'd use ^$orig $new	09:31
	;)	09:31
	but yeah	09:31
	thiago updates his hook script	09:32
pasky	now, I have a feeling that I've written such a hook before and in that case I almost certainly used merge-base there	09:32
	but that's simply because -1 is relatively new invention	09:32
	← mtodd left	09:32
thiago	can we have git is-ancestor $rev1 $rev2	09:33
	→ mtodd joined	09:33
	dwmw2_gone → dwmw2	09:33
thiago	returns 0 if they are the same, 1 rev1 is ancestor of rev2, 255 if rev2 is ancestor of rev1 and 128 if they are not related?	09:33
	← GyrosGeier left	09:33
	← jelmer left	09:33
	← hegge left	09:33
pasky	I think having such a specialized command would be fairly unnatural	09:34
	since it's so trivial to do with rev-list	09:34
	it might make sense ti include this in some rev-list(1) cookbook though	09:35
	→ GyrosGeier joined	09:35
	→ hegge joined	09:35
	→ jelmer joined	09:35
thiago	wtf	09:38
	$ git rev-list -1 ^master HEAD	09:38
	fatal: ambiguous argument 'HEAD': unknown revision or path not in the working tree.	09:38
samjam	so I should use git-merge-base or git-rev-list ?	09:38
	← johnw left	09:39
	thiago sets the hook script and waits to see if anyone complains	09:41
	→ gebi joined	09:45
doener_	samjam: you could also try to parse this: git log --decorate --pretty=oneline \| grep 'tag:'	09:45
	samjam: that's probably quite hard and fragile though, but possibly a log faster, since it walks the history only once	09:46
	→ chris2 joined	09:46
	← selckin left	09:46
samjam	hey, thats really good; whats fragile about it, the output format may change?	09:47
doener_	extracting the refs/tags/whatever part	09:47
	there might be more than one entry per commit, and you have to cope with whitespace and stuff	09:48
	but it's a hell of a lot faster. Just looking up the merge-bases for HEAD and v0.99 and v1.0.0 in git.git already takes longer than the whole log command	09:49
	→ ereslibre joined	09:49
	← trochala left	09:50
	→ ^matthieu^ joined	09:50
doener_	maybe you could convince someone to at least provide a pretty=format option that allows to print the decorations, or something like that	09:50
	would probably make the job easier	09:50
	→ trochala joined	09:51
	→ JanK joined	09:51
JanK	can i search for files in the history?	09:51
	(i'm looking for a file in a deleted folder, but i dont know the rev when it was added and removed)	09:52
doener_	git log -- directory/file	09:53
	← snitko left	09:57
	→ skarab joined	09:58
	→ Lash\| joined	09:59
	→ cilly joined	10:00
	← pygi left	10:01
	← dancor left	10:01
	dwmw2 → dwmw2_gone	10:02
	← skarab left	10:03
	← jtoy left	10:03
	janneke1 → janneke	10:06
	← tjafk1 left	10:06
	→ tjafk joined	10:06
	→ tcoppi joined	10:08
	← BabelO left	10:10
vuf	uhm, I want the rebase -i functionality (for squashing), but rebase says there is nothing to do because master did not advance	10:10
	→ astubbs joined	10:11
	← ereslibre left	10:13
doener_	maybe you wanted "git rebase -i HEAD~5" or something?	10:13
vuf	doener_: yes it seems like it. Why is specifying HEAD~5 and master different, it's the same commit?	10:15
doener_	are master and HEAD the same, or master and HEAD~5?	10:15
	← astubbs left	10:15
doener_	if master and HEAD~5 really reference the same commit, then there's no difference	10:16
vuf	do'h!	10:16
	→ dragonball_ joined	10:16
doener_	I assumed that you are currently on master and that you want to change some commits that are also on master	10:16
	← trochala left	10:16
vuf	sorry, I forgot to change away from master before trying the rebase	10:16
	doener_: you are right, and I am stupid	10:16
doener_	that's what you said ;-)	10:16
vuf	thanks	10:17
	→ janm joined	10:21
	→ trochala joined	10:22
	→ xshelf joined	10:22
xshelf	i want to use lighttpd to show gitweb, what is the minimum conf I need to get it going (just a read only web UI for repo)	10:25
	→ parasti joined	10:27
	→ ysis joined	10:28
	← Method left	10:29
doener_	xshelf: try starting "git instaweb", that starts a lighttpd running gitweb for the current repo	10:29
	xshelf: the used config file will be placed in .git/gitweb/httpd.conf	10:29
xshelf	if I want to host 2 repos, can I do from the common root folder?	10:29
	→ bcarlyon\|laptop joined	10:29
	→ l0verb0y joined	10:31
doener_	xshelf: yes. See README and INSTALL of gitweb	10:32
	→ Method joined	10:36
	← cilly left	10:38
	← trochala left	10:40
	→ trochala joined	10:41
	→ Yuuhi joined	10:41
	→ cilly joined	10:41
BBHoss	anyone know why in os x git instaweb would give a 500 and an error on the console: (mod_cgi.c.584) cgi died	10:42
	(mod_cgi.c.998) CGI failed: Exec format error /Users/hoss/Documents/synergyeoc/.git/gitweb/gitweb/gitweb.cgi	10:42
	← JanK left	10:42
	← l0verb0y left	10:45
xshelf	doener_: I apologize for my impulsive questions, I did the easier way. I had cloned the git repo, I too a peek into the httpd.conf under the .git/gitweb. Thanks	10:46
	→ vbabiy joined	10:47
	→ Weasel[DK]_work joined	10:49
doener_	xshelf: no need to apologize. Bote that "instaweb" also modifies gitweb.cgi to set the project root, and maybe other stuff as well. I'm not that familiar with it	10:52
	uhm, s/Bote/Note/	10:52
ciaran29d	doener_: do you have a cold? :)	10:52
	→ tvw joined	10:53
doener_	not anymore, I had one during most of July	10:53
	→ m-vo joined	10:55
	→ bentob0x joined	10:56
	→ langenberg joined	10:56
	← pflanze left	10:57
langenberg	I'm trying to write a post-receive hook, to notify people with log old...new. Now I added one line to the file: echo Old:$1 New:$2 Ref:$3 >> ~/support_git.log, but the result output is: Old: New: Ref:	10:57
	post-receive should get called with <old-value> SP <new-value> SP <ref-name> LF, right?	10:58
doener_	langenberg: yeah, but on stdin	10:59
langenberg	What does that mean?	10:59
	$n is for parameters and not stdin?	10:59
	→ yatiohi joined	11:00
	← tcoppi left	11:00
doener_	it's "echo old new ref \| post-receive", not "post-receive old new ref"	11:00
	and $n refers to the positional parameters	11:00
langenberg	Aah I see	11:01
doener_	"read old new ref" should do	11:01
	← trochala left	11:01
doener_	probably with a while loop to process every line	11:02
	→ botanicus joined	11:03
	← Lash\| left	11:04
	← Cap_J_L_Picard left	11:04
	→ Cap_J_L_Picard joined	11:04
	→ Lash\| joined	11:09
	← cilly left	11:10
	← imyousuf left	11:11
	→ ereslibre joined	11:12
	→ cilly joined	11:14
	← waltman left	11:14
	→ eddyp_work joined	11:19
	bcarlyon\|laptop → BCarlyon\|Laptop	11:20
eddyp_work	is there a way to get rid of a remote branch which was packed? I got rid of the remote and I don't have any (remote) refs anymore, except a packed ref	11:21
	should I just remove the line in packed-refs and repack?	11:22
	err, s/repack/git gc/	11:22
	← Tv left	11:23
	→ davi joined	11:24
	→ selckin joined	11:24
	← botanicus left	11:25
	← BBHoss left	11:29
uau	eddyp_work: any reason you can't delete the ref normally (without manually editing packed-refs)?	11:32
eddyp_work	uau: with git-??? which command?	11:32
uau	and exactly what do you mean by "getting rid" of the branch? that you have no more references to it? or freeing the disk space used by objects only belonging to it?	11:33
	how to delete references depends on what kind they are	11:33
	(branch or tag)	11:33
eddyp_work	uau: both... is a remote branch	11:34
uau	if you already deleted the remote that should delete associated remote branches IIRC?	11:34
eddyp_work	apparently I managed to manually mess it up so it doesn't see any remote, but the remote branch is visible	11:35
uau	you can delete remote branches with "git branch" too	11:36
	with the -r switch	11:36
	← soultan left	11:37
eddyp_work	uau: thanks, branch deleted	11:37
	← brendon_work left	11:38
	eddyp_work hopes git gc will remove the objects kept in place by that branch, too	11:38
	→ brendon_work joined	11:38
uau	if you had the branch checked out recently there can still be references to it in the HEAD reflog	11:38
eddyp_work	and it seems indeed that there are less objects now	11:38
	→ soultan joined	11:39
eddyp_work	.. but more wasted space ... oh, well ....	11:42
	→ BBHoss joined	11:42
Weasel[DK]_work	Trying to create a GIT repo from CVS via git-cvsimport, but i get error on files which were deletet once (Attic), now it complaints about the files not longer int repo. What can be done ?	11:43
	← jmspeex left	11:43
	← tvw left	11:44
	→ yofel joined	11:44
	→ tvw joined	11:44
	→ ebel joined	11:46
	← mithro left	11:48
	→ lea_w joined	11:50
	→ LostInTheWorld joined	11:50
	→ diask joined	11:51
	← brendon_work left	11:51
	→ Tv joined	11:52
	→ brendon_work joined	11:52
broonie	Hrmpf. Why do I still have loose objects after git repack -a -d?	11:53
	→ mithro joined	11:57
	→ cbreak joined	11:59
	← langenberg left	12:00
vuf	broonie: I think that unreferenced objects are retained for 30 days, see git-gc configuration	12:01
	→ joelmoss joined	12:01
joelmoss	hi all	12:01
	→ pygi joined	12:03
	deavid → xdeavid-away	12:04
	→ yofel_ joined	12:05
	← tiglionabbit left	12:05
	→ tiglionabbit joined	12:06
	→ cannonball joined	12:07
	→ j_manu joined	12:08
	← ysis left	12:09
j_manu	newbie here. If i create a new branch with "git checkout -b test" and then edit a file and when i do "git status" I see the edited file as mmodified. All fine and good. Now when i do "git checkout master" and then do "git status" I don't expect to see the modified file	12:10
	in there because I though branches were totally independent.	12:10
doener_	you didn't commit yet	12:10
	→ dancor joined	12:10
doener_	so the changes are in your working tree and/or index only	12:11
j_manu	so if i don't commit the changes are global across all branches ?	12:11
doener_	branches are just pointers to commits	12:11
dancor	how do i run gitweb with gitosis? just running git instaweb from my gitosis-admin repo doesn't fill in the descriptions, nor does it hide the repos without 'gitweb = yes'	12:11
j_manu	hmm.. sorry for the svn lingo.. so if both master and test point to revision 88 initially	12:12
doener_	branches are just references to commits. Until you commit, your changes are not in any commit.	12:12
j_manu	I need to commit the changes to bump up the version of test and point it to a different commit ?	12:12
doener_	minus some remaining svn lingo, yes	12:13
	cbrake_away → cbrake	12:13
doener_	if you commit, git creates a new commit object and adjusts your current branch to point to that new commit	12:13
j_manu	and whn i do svn merge test, master resets to that particular commit ?	12:13
	sorry git merge test	12:13
doener_	depends.	12:14
	← Cap_J_L_Picard left	12:14
doener_	if there are no commits in master that are not in test, then yes, master can be fast-forwarded	12:14
	→ Cap_J_L_Picard joined	12:14
j_manu	what happens when there are independent commits in both ?	12:14
doener_	if there are commits in master that are not in test, then a merge commit is created and master gets updated to point to that merge commit	12:14
j_manu	ok.	12:15
	and test will still point to the same commit as before right ?	12:15
doener_	yep	12:15
	→ lorandi joined	12:15
doener_	"gitk master test" nicely visaulizes that	12:15
	s/au/ua/	12:16
j_manu	{unfortunately can't access gui. working using ssh]. how do i disregard all changes in the working copy ? git reset HEAD^ ?	12:16
doener_	uncommitted changes? just "git reset --hard"	12:16
j_manu	oh!. then i'm confused. What does git reset HEAD^ do then ?	12:17
doener_	HEAD^ is the first parent of HEAD, so "git reset HEAD^" would move your branch back to the previous commit	12:17
eddyp_work	j_manu: git reset HEAD^ would loose you latest committed entry	12:17
j_manu	oh.	12:17
eddyp_work	but your working copy would be unaffected	12:17
doener_	and because it defaults to --mixed, the changes would not be dropped at all, the working tree would remain unchanged	12:17
	"git reset --hard" == "git reset --hard HEAD" == reset the index and working tree to match HEAD	12:18
eddyp_work	actually, in git nothing is usually lost until you 'git gc'	12:18
j_manu	but if i do git status then all the changes which was in the dropped commit will show up right ?	12:18
eddyp_work	yes	12:18
j_manu	so it is useful to drop a commit and then redo it again ?	12:19
doener_	(unless you use --hard)	12:19
j_manu	--hard will disregard all working copy changes also ?	12:19
doener_	eddyp_work: since 1.6.0, reflogs don't expire anymore be default. So even gc won't drop anything	12:19
eddyp_work	if you:1) never published that change 2) you think you should reorganize history	12:20
	j_manu: ^^^	12:20
	doener_: so how do you make them be dropped? git gc --prune?	12:20
doener_	"git reset --hard HEAD^" == move the current branch to HEAD^ and reset the index and working tree to match that	12:20
	eddyp_work: no, you need to drop the reflog entries first	12:20
	← tvw left	12:21
doener_	pruning only affects unreferenced objects anyway	12:21
	s/unreferenced/unreachable/	12:21
eddyp_work	I see	12:21
	← yofel left	12:22
	→ advany_ joined	12:23
	← advany left	12:23
j_manu	got it..	12:24
	← soultan left	12:24
eddyp_work	j_manu: I think you shoul dalso read this: http://gitfu.wordpress.com/2008/04/09/links-about-the-git-index/	12:25
j_manu	what does HEAD and HEAD^1 and HEAD^2 refers to ? is it trunk, trunk minus 1 commit, trunk misu 2 commits ? [pardon svn lingo]	12:25
	I will look into it.. thanks	12:25
	→ juanjoc joined	12:25
	→ soultan joined	12:26
doener_	juanjoc: HEAD is a reference that points to what have checked out at the moment	12:26
eddyp_work	j_manu: HEAD^2 is not anything, but HEAD~2 is correct	12:26
doener_	argh... stupid tab-completion	12:26
pasky	HEAD^2 is second parent of head	12:26
	→ BabelO joined	12:26
	→ capisce_ joined	12:26
eddyp_work	pasky: is really ^2 valid?	12:26
doener_	j_manu: (the above was for you)	12:26
pasky	eddyp_work: yes	12:26
	but it is meaningful only for merge commits	12:27
	→ langenberg joined	12:27
jast	HEAD~2 means take the first parent of the first parent (i.e. go up to the first parent twice)	12:27
doener_	j_manu: most of the time it will be a symbolic ref to some branch	12:27
pasky	^ chooses parents horizontally, ~ dives into history vertically ;)	12:27
doener_	j_manu: that can be resolved to the commit that the branch points to. And the ^ and ~ syntax walks along the ancestry of commits	12:28
eddyp_work	pasky: but most of the times you'd want ~	12:28
doener_	j_manu: git-rev-parse(1) has more information	12:28
j_manu	ok.. in svn terms if i checkout revision 88, HEAD points to revision 88, HEAD~1 is 87 HEAD~2 is 86 etc ?	12:28
eddyp_work	j_manu: correct	12:28
doener_	"kind of"	12:28
j_manu	ok :). kind of means ?	12:28
Weiss	doener_: what is 'origin/HEAD' in that case? especially since 'origin' is a bare repository..	12:29
eddyp_work	for heads with mutiple parents the thing is more complicated, I guess	12:29
doener_	Weiss: bare repos also have a HEAD ref. Most of the time, a symbolic ref to the master branch	12:30
broonie	vuf: Oh, bah - it doesn't pack the old objects. Thankls.	12:30
doener_	Weiss: that, for example, determines which local branch "git clone" creates for you	12:30
jast	j_manu, main thing that makes it only "kind of" is that you can't really number revisions in git that way. when merges happen, a commit can be based on more than one 'parent' commit, unlike in svn where commit N+1 is always based only on commit N (and its history)	12:30
	→ LiamH joined	12:31
jast	also, of course there's the fact that these revision numbers could differ between various copies of the repository since we're all decentralized	12:31
doener_	Weiss: and "origin/HEAD" is also a symbolic ref, most of the time pointing to origin/master	12:32
	→ Lemurnomicon joined	12:33
j_manu	what you are trying to say is that there are different paths to HEAD for each repo and on each repo HEAD~1 and HEAD~1 may result in different object database ?	12:33
Weiss	doener_: ok... so, more for consistency than anything you'd ever want to use explicitly?	12:33
doener_	j_manu: and when you compare svn to git, in git terms, you don't even have branches in svn. Just one huge versioned directory.	12:33
	Weiss: you could eg. have a master and a maint branch, and decide that clones should start with a maint branch instead of a master branch (for whatever reason)	12:34
	Weiss: and it makes a plain "git log" possible in a bare repo	12:34
pl0pix	why do I have to clean up (git reset --hard d	12:34
doener_	Weiss: and yeah, it helps consistency and bare<->non-bare conversions ;-)	12:34
pl0pix	or git clean -f after a git checkout	12:34
jast	pl0pix, that's not true in the general case	12:35
doener_	pl0pix: because you didn't commit your changes	12:35
	(and maybe lost some work now)	12:35
j_manu	thanks a lot for the help :). I will now go try out all that i have learned.	12:36
	← samjam left	12:36
Weiss	doener_: ok, thanks	12:36
pl0pix	no, I didn't lost files. I try to clone multiple branches and had a directory in a branch where it schouldn't exist	12:40
	← alenoosh left	12:40
	← j_manu left	12:41
	← capisce left	12:42
	← Sput left	12:44
	→ Sput joined	12:45
	← Sput left	12:46
	→ cmarcelo joined	12:48
	→ ysis joined	12:48
	→ madewokherd joined	12:49
	← vbabiy left	12:51
	→ vbabiy joined	12:54
j_engelh	Heh, there was a "stupid" merge strategy in git?	12:55
	→ Sput joined	12:55
	← gretch left	12:56
vuf	so if reflogs don't expire, will the loose objects stay around forever by default?	12:56
j_engelh	you can pack them together.	12:57
	← ebel left	13:01
vuf	how?	13:01
	← ph^ left	13:02
	jmesnil_ → jmesnil_lunch	13:03
	→ hansengel joined	13:04
hansengel	Hi, I'm having some problems with Git.tmbundle.. is this a good place to ask?	13:04
	← brendon_work left	13:05
	→ kmr83 joined	13:05
cehteh	there is prolly no better ...	13:05
kmr83	hi guys	13:05
	i've got a newbie question about git	13:05
	→ brendon_work joined	13:06
hansengel	cehteh: Okay, well whenever I open any of the dialogs this error is shown at the top:	13:06
cehteh	i should add "dont ask me" :) at least not perl related things	13:06
	→ kukks joined	13:06
cehteh	but someone else here might help	13:06
hansengel	sh: line 1: git: command not found	13:07
	cehteh: lol, k	13:07
pasky	<HEAD:git-gui/lib/choose_rev.tcl>	13:07
kmr83	can i checkout the source from a git repo, specifically tuxonice.git, and compile it to get a working 2.6.27 kernel?	13:07
Gitbot	[git choose_rev.tcl]: http://tinyurl.com/6f8ctg [blob]	13:07
	→ swhitt joined	13:07
cehteh	hansengel: well .. check $PATH .. and if git is isntalled at all	13:07
	→ koke joined	13:07
hansengel	cehteh: it is, and I can use it from the command line fine	13:07
	that's why I'm confused	13:07
Weiss	kmr83: that's the idea :) - though exactly what you get when you checkout the source depends entirely on the project..	13:08
Lemurnomicon	the path TextMate sees may not match what your shell sees.	13:08
cehteh	does the stuff sanitize the PATH? .. maybe	13:08
hansengel	yeah, possibly	13:08
kmr83	thanks Weiss, its my first tie using git :)	13:08
Lemurnomicon	not that it sanitizes, but the OS X loginwindow may have a different notion of PATH than what your shell configs set and so any GUI apps get its path not your shell's	13:09
doener_	no need to wear a tie, it's all casual here	13:09
	;-)	13:09
Lemurnomicon	I think the bundle looks for a TM_GIT environment variable, so maybe set that via the TM pref pane	13:09
hansengel	Lemurnomicon: okay, I'll try	13:09
	← bartocc_ left	13:09
pasky	anyone with tk skills here?	13:09
hansengel	and I set this to /usr/local/bin?	13:09
	or /usr/local/bin/git?	13:09
Lemurnomicon	to the path to git, so if your git's /usr/local/bin/git yeah	13:09
kmr83	this probly is not the place to ask this, but has anyone tried compiling the latest 2.6.27-rc3?	13:09
	Weiss used to do Tcl/tk, but has all but forgotten it..	13:09
pasky	http://repo.or.cz/w/git.git?a=blob;f=git-gui/lib/choose_rev.tcl#l6	13:10
	I just can't figure out (i) what the hell is the binary data (ii) how to decode it (iii) how to encode it	13:10
	I mean, it's obviously something in base64	13:10
j_engelh	kmr83: yes, and?	13:10
kmr83	well i'm compiling it right now, its my first time compiling from a git source, not sure if it will work...has anyone compiled the latesst kernel source from git?	13:11
pasky	but according to tutorials I assumed it's a GIF but it doesn't really look so; maybe it's raw data, but then how to display it out of tk and where did it come from?	13:11
	and what pallete etc does it use?	13:11
doener_	pasky: looks like a gif	13:12
j_engelh	kmr83: less metaquestions!	13:12
doener_	pasky: ... in base64	13:12
hansengel	Lemurnomicon: woot! that worked	13:12
	thanks.	13:12
kmr83	well, sorry	13:12
Lemurnomicon	huzzah, I R USEFUL TODAY	13:12
	← soultan left	13:12
	→ frq1_dave joined	13:12
hansengel	:P	13:12
pasky	doener_: well, for one, it's quite hard to decode it - openssl thinks it's invalid base64, uudecode complains about "short input"	13:12
	and the decoded form looks _nothing_ like GIF	13:13
	← ben_h left	13:13
j_engelh	burn the gifs	13:13
kmr83	i guess it should work....i hope it works	13:13
Weiss	pasky: did you run 'file' on the raw data?	13:13
j_engelh	why should not it?	13:13
Weiss	or maybe 'identify'	13:13
	→ soultan joined	13:14
vuf	pasky: it's a gif of a magnifying glass	13:14
pasky	weiss: of course	13:14
Weiss	..and it just said 'data'?	13:14
pasky	vuf: how did you decode it?	13:15
vuf	pasky: base64-decode-region in Emacs	13:15
pasky	hum	13:15
vuf	there are Perl one-liners to do that as well :)	13:16
pasky	how do I trigger that command?	13:16
	pasky doesn't know emacs commands :)	13:16
Lemurnomicon	Mark the region and then M-x base64-decode-region. ctrl-space at the beginning, move to the end with cursor keys or C-n, C-f, etc.	13:17
resmo	hmm what is the normal procedure for init a server / client repo, git init --bare on server, git init ... add . ... commit on client and push to server?	13:17
vuf	pasky: Ctrl-Space, Esc >, Esc x base64-decode-region	13:17
pasky	wow	13:17
doener_	pasky: base64 decodes it for me	13:17
pasky	that indeed worked	13:18
	→ JimmyB_ joined	13:18
pasky	now I'm really puzzled	13:18
	→ zachinglis joined	13:18
pasky	why doesn't uudecode or openssl base64 grok it?	13:18
	→ ph^ joined	13:19
thiago	echo '...' \| openssl base64 -d \| file -	13:19
	/dev/stdin: writable, no read permission	13:20
	how useful :-P	13:20
	→ MK_FG joined	13:20
j_engelh	uh	13:21
	file -s -	13:21
thiago	openssl base64 -d doesn't decode long lines	13:22
doener_	same thing. The problem is that openssl doesn't generate any output	13:22
thiago	$ wc -c /tmp/foo.gif	13:22
	712 /tmp/foo.gif	13:22
doener_	(for the file "problem")	13:22
thiago	$ openssl base64 < /tmp/foo.gif \| tr -d '\n' \| openssl base64 -d \| wc-c	13:22
pasky	aaaaaaaah	13:22
thiago	0	13:22
pasky	yeah	13:22
j_engelh	doener_: -out	13:22
	no?	13:22
pasky	we all had the same idea at once :)	13:22
	with short lines it works	13:22
	damn openssl	13:23
j_engelh	well	13:23
pasky	so only thing left puzzling me is uudecode behaviour, but that tool is uberstrange anyway	13:23
doener_	j_engelh: -out what? The input is invalid, so there's no output	13:23
j_engelh	sigh	13:23
doener_	$ echo Li4uCg== \| openssl base64 -d \| file -	13:23
	/dev/stdin: ASCII text	13:23
	works just fine	13:23
	→ mw joined	13:23
j_engelh	what I do find strange is that openssl, interactively, requires two EOFs	13:24
	→ neoeinstein joined	13:24
	← hansengel left	13:25
vuf	pasky: I think MIME/PEM has a line length limit below 80	13:25
pasky	is there some nice repository of free toolbar icons?	13:25
jast	try the gnome and kde artwork pages :)	13:25
j_engelh	/usr/share/icons/	13:25
	dwmw2_gone → dwmw2	13:26
	→ Oompa joined	13:26
	← davi left	13:26
pasky	hmm, I'd like something more windows-style ;)	13:27
resmo	hmm made a git init --bare repo, when i try to fetch git says atal: Couldn't find remote ref HEAD	13:27
pasky	well, I'll keep looking	13:27
j_engelh	pasky: that would possibly not be free	13:27
resmo	what do i wrong	13:27
pasky	...	13:27
doener_	resmo: you can't fetch from an empty repo	13:27
	resmo: push something first	13:27
vuf	j_engelh: ^D flushes the line buffer if you have typed something (try with cat)	13:27
j_engelh	vuf: no i mean like openssl base64 -d <enter> Li4uCg== <enter> <^d> <^d>	13:28
	or I need base64 -d <enter> Li4uCg== <^d> <^d> <^d>	13:28
	3 even...	13:28
	→ mithro_ joined	13:28
resmo	doener_: but when i try to push it says No refs in common and none specified; doing nothing.	13:29
	Perhaps you should specify a branch such as 'master'	13:29
vuf	j_engelh: oh, okay ... I didn't test :)	13:29
	→ davi joined	13:30
resmo	doener_: damn sorry wrong permissions...	13:31
	← fafhrd left	13:31
joelmoss	quit	13:31
	← joelmoss left	13:31
	→ Plouj joined	13:33
Plouj	hi	13:33
	how do I push a local branch to a remote repositry's master branch?	13:33
	→ irotas joined	13:34
	→ bdiego joined	13:34
	← mithro left	13:34
GarethAdams_	Plouj: you don't	13:34
thiago	git push remote localbranch:master	13:34
	the server may reject that	13:34
Plouj	yeah, howcome it rejects it?	13:35
	I get:	13:35
cehteh	:)	13:35
Plouj	error: src refspec localbranch does not match any.	13:35
Weiss	kmr83: i guess you know a little about compiling kernels from source to begin with?	13:35
Plouj	fatal: The remote end hung up unexpectedly	13:35
	← davi left	13:35
Plouj	GarethAdams_: why not?	13:35
	← kmr83 left	13:36
Plouj	what else is the simplest way to update a remote repository from a local branch?	13:36
bmalee	Merge it to your local master, then push that?	13:36
	mithro_ → mithro	13:36
Plouj	bmalee: but I don't want to mix my local master with the changes in a local branch	13:36
	there are some things in the localbranch that don't belong in the maste branch	13:36
GarethAdams_	Plouj: http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	13:37
	→ bryanl joined	13:37
Plouj	that doesn't really deal with my issue	13:38
	basically I just don't want to rsync my workdir to another machine	13:40
	because I don't know what that might do to the internal git files on the remote machine	13:40
	→ bongiojp joined	13:41
	→ foca_ joined	13:44
	← soultan left	13:44
	← ereslibre left	13:45
	← irotas left	13:46
	← kanru left	13:46
	→ irotas joined	13:47
	jmesnil_lunch → jmesnil	13:49
	→ soultan joined	13:49
	→ paltman joined	13:50
	→ lu_zero_ joined	13:55
Plouj	ok	13:56
	I'll try pulling from the remote computer my localbranch	13:56
	← ferdy left	13:57
	← soultan left	13:57
	← pygi left	13:58
	→ qrush joined	13:59
	→ soultan joined	14:00
	→ andyjgill joined	14:00
	→ kumbayo joined	14:01
andyjgill	Can someone explain the difference between HEAD and master?	14:01
	→ schacon joined	14:02
shd	andyjgill: HEAD is the latest commit of the current branch	14:02
	andyjgill: and master is the default branch created by git	14:02
	→ rdavila joined	14:02
shd	andyjgill: usually, people have their latest stuff in master	14:02
	andyjgill: and master is usually the most important branch, though not necessarily	14:02
andyjgill	So a commit automatically moves HEAD?	14:02
shd	yes	14:02
andyjgill	Ahh. And master is just a name, but one created at init time.	14:03
shd	and git reset can be used to adjust HEAD arbitrarily	14:03
Plouj	much better!	14:03
shd	for example, you can destroy 3 latest commits from history with git reset --hard HEAD~3	14:03
doener_	when you checkout a branch, HEAD becomes a symbolic ref to that branch	14:03
andyjgill	So is master a branch? (Thing of branch type)	14:04
shd	andyjgill: yes	14:04
jast	which means that HEAD normally just says: hey, if you want to know what's checked out right now, look at (e.g.) master	14:04
doener_	so when you use HEAD, that's derefenced to the branch (and then eventually further to the commit that the branch points to and then maybe even further)	14:04
shd	andyjgill: you could use any other name, git doesn't really care about the name	14:04
andyjgill	This helps. Thanks.	14:05
	Has anyone got a good link to a "git for darcs users" tutorial?	14:05
doener_	eg. "git commit" looks at HEAD, and if it's a symbolic ref, it dereferences that to get the branch it needs to update	14:05
	and if you're on a detached HEAD, HEAD is not a symbolic ref, but just references a commit directly	14:05
	and then, "git commit" would update HEAD itself	14:06
	← advany_ left	14:06
andyjgill	Ahh. Thats smart.	14:06
doener_	so "HEAD is the latest commit of the current branch" is not quite correct. Most commands will derefence HEAD to get to a commit, but HEAD itself is just a ref	14:07
jast	otoh, HEAD is always what's currently checked out (unless you do evil things)	14:07
	→ kanru joined	14:08
doener_	what evil things would that be? HEAD define "what is checked out"	14:08
	s/define/defines/	14:08
	→ spearce joined	14:08
jast	an update-ref on HEAD, for example :)	14:08
	yeah, well, okay	14:09
	→ charon joined	14:09
	← ysis left	14:09
jast	i see what you mean	14:09
doener_	that like a checkout + read-tree + whatever. It still changes HEAD, and that's the only notion of "what's checked out" that git has	14:09
vuf	can diff be made whitespace aware? e.g. to display a pretty patch in case I indent a block due to it getting into a conditional?	14:10
	← Sho_ left	14:11
jast	what would you have it do? diff can only describe completed lines getting removed and added	14:11
doener_	vuf: you mean you want diff to ignore whitespace changes? Or even as far as reindenting on the fly?	14:11
	the former can be done with -b/-w, the latter would produce broken patches	14:11
	← dragonball_ left	14:11
doener_	jast: there's --color-words, but yeah, that probably won't help here	14:12
	→ MentalMaelstrom joined	14:13
	→ krh joined	14:13
vuf	doener_: -w is probably what I want, thanks ... it looks nicer in a side-by-side comparison, but there is only so much you can do in a unified diff	14:13
	← Sput left	14:14
	← Plouj left	14:15
	→ Sput joined	14:15
	← schacon left	14:16
	← Sput left	14:16
	→ Sput joined	14:16
resmo	hmm no matter what i try i can't push to a git server (git init --bare), it says "No refs in common and none specified; doing nothing. Perhaps you should specify a branch such as 'master'."	14:19
cbreak	how did you push it?	14:20
resmo	ssh	14:20
cbreak	git push remoteserver?	14:20
shd	resmo: git push -f on the first time	14:20
resmo	git push ssh://user@example.com/path/to/git-repo.git	14:21
shd	resmo: when you create a repo, it doesn't have commits, and therefore the first push alarms about that	14:21
ToxicFrog	You need to tell it what branches to push	14:21
shd	resmo: push by default allows only straight continuation of history, and but that can be overriden with -f	14:21
ToxicFrog	Normally it pushes only branches you have in common	14:21
cbreak	you can do a git remote add to make it easier in the future	14:21
ToxicFrog	Since the remote is bare, you have no branches in common with it	14:21
	so try something like "git push url master" to push the master branch to it explicitly.	14:22
resmo	cool thanks guys	14:22
ToxicFrog	You can also use git-remote to make this more concise: git remote add name url; git push name branch	14:23
cbreak	weird, "git remote -h" does only work in a repository	14:23
shd	resmo: ah, correction, -f was not required here	14:23
ToxicFrog	Yeah, -f isn't needed for pushing new branches.	14:23
shd	my bad..	14:23
	← BrianTheCoder left	14:23
ToxicFrog	It just won't push new branches by default, so that, for example, if you have a dozen topic branches these don't end up getting pushed without your request.	14:24
resmo	this shoud really be written in the quickstart docu :)	14:24
	← lu_zero_ left	14:25
	→ smartcat99s joined	14:26
	→ davidfetter joined	14:27
	→ sergiodj joined	14:28
	← bdiego left	14:29
	← camcorder left	14:29
	→ draco_ joined	14:29
	metze → metze_weg	14:29
	metze_weg → metze_away	14:32
	→ ferdy joined	14:32
	→ Sho_ joined	14:32
	foca_ → foca	14:33
	→ fafhrd joined	14:38
	xdeavid-away → deavid	14:38
	→ ondrej joined	14:39
ondrej	hi, is there a way to put "--color-words" into a .gitconfig file? I know how to put regular colors by default, but I haven't figured out this.	14:41
lea_w	Am I correct that grafts usually only occur for shallow clones?	14:42
	→ vbabiy_ joined	14:42
	→ botanicus joined	14:42
lea_w	ondrej: Looking at 'man git-config', it doesn't look like it. The best idea I have is using an alias in your shell.	14:42
Tv	lea_w: that is one use for grafts, probably easiest to realize	14:43
ondrej	lea_w -- thanks, that's a good idea	14:43
	→ d0k joined	14:43
lea_w	Tv: Are there any other uses (apart from people creating grafts manually for obscure purposes)?	14:43
ondrej	I am looking at the git commits if they implemented it	14:44
	→ acmoore joined	14:44
Mikachu	re the ohhhh jeeee message i got from git-svn, there's an updated version of gnutls now that fixes it	14:46
Tv	lea_w: you can fake history in general, with grafts	14:48
	lea_w: chopping history off is just one special case	14:48
acmoore	is there some technical reason why there's no hooks available for git-send-email? Or, is that just something that hasn't been written (yet) because it's not commonly requested?	14:48
lea_w	Tv: Yes, but how exactly? Where's TFM to read?	14:48
Tv	lea_w: e.g. you could attach a bigger history to a newer, partial, history	14:49
lea_w	Ah. Hm, I see.	14:49
	→ mehemiah joined	14:50
Tv	lea_w: grafts are kinda messy internals, with lots of traps..	14:50
Mikachu	i never found any file that documented the format of the grafts file	14:50
Tv	lea_w: but basically apart from understanding the object graph, there's not much to read -- just enlightenment	14:51
	Mikachu: how about Documentation/gitrepository-layout.txt	14:51
lea_w	Tv: So what do they do though? Is it just something like "internally substitute new_SHA1 for old_SHA1"?	14:51
Mikachu	yep, never found that ;)	14:51
Tv	lea_w: how about reading ;)	14:51
	ToxicFrog eyes gitk with suspicious	14:51
lea_w	Tv: That's my question. Reading what?	14:52
ToxicFrog	ME: check out branch "releases"	14:52
lea_w	ah...	14:52
Tv	lea_w: i just pasted a filename	14:52
ToxicFrog	GITK: ERROR! checked out branch "releases"	14:52
lea_w	that file	14:52
	Thanks everyone. :)	14:52
Mikachu	<HEAD:Documentation/gitrepository-layout.txt>	14:52
Tv	lea_w: but in general, if you're not dying for need of grafts, stay the hell away	14:52
Gitbot	[git gitrepository-layout.txt]: http://tinyurl.com/5gt7p4 [blob]	14:52
Tv	lea_w: also, learn to grep.. took me 10 seconds to find that file	14:52
	→ tcoppi joined	14:55
	← vbabiy left	14:56
	→ Pistos joined	14:56
Pistos	Definition of WTH:	14:56
	% git branch --no-m	14:56
	error: Ambiguous option: no-m (could be --no-merged or --no-merged)	14:56
	← cmarcelo left	14:57
cbreak	that looks very ambiguous	14:57
Pistos	Yep.	14:57
cbreak	I can't even tell them appart	14:57
Pistos	lol	14:58
Mikachu	i think it thinks there's both the --no-merged option, and the --merged option with a --no- prefix, somehow	14:58
	→ morphir joined	14:59
Pistos	I just compiled from repo version of git, FWIW.	14:59
	← morphir left	14:59
Mikachu	builtin-branch does define the options with OPT_NONEG, but that seems to be ignored in at least one place in the option parser	15:00
	← vbabiy_ left	15:03
	→ vbabiy joined	15:04
acmoore	I'm interested in updating my bugtacker (bugzilla) when I submit patches. my workflow is that I use git-send-email to submit the patches to the group mailing list. Is there a hook for send-email or something I could use to tie in my script to post to bugzilla?	15:06
segher	just cc: bugzilla?	15:07
	← d0k left	15:07
vmiklos	http://tw.apinc.org/weblog/2007/01/16#using-git-with-bugzilla	15:07
	← Pistos left	15:07
	→ kazim59 joined	15:07
vmiklos	eh, it's 404 ;/	15:07
	that hosted a simple git-send-bugzilla script	15:08
acmoore	segher: that would be great. my bugzilla doesn't accept email, though. Perhaps I should try to get someone to fix that.	15:08
kazim59	What do I do to undo a commit? I don't want any signs that it ever happened.	15:08
acmoore	vmiklos: Oh. I'll see if I can get google cache or archive.org or something to find that for me.	15:08
	← MK_FG left	15:08
vmiklos	kazim59: is it the latest commit?	15:09
kazim59	vmiklos: yes it is	15:09
	← GyrosGeier left	15:09
vmiklos	git reset --hard HEAD^	15:09
kazim59	vmiklos: oh thanks!	15:09
cbreak	did you push it?	15:09
kazim59	vmiklos: okay, and what if its not latest?	15:09
cbreak	git rebase -i	15:10
kazim59	cbreak: no... not yet.	15:10
cbreak	good :) if you pushed it, doing reset or rebase is not a good idea	15:10
kazim59	cbreak: I understand that :)	15:10
	→ jackdempsey joined	15:11
	← paltman left	15:12
	→ bugfux joined	15:15
	← langenberg left	15:16
	→ ezraburgh joined	15:16
kazim59	when I apply a patch my friend sends me, I get trailing whitespace errors. What can I do?	15:16
aeruder	kazim59: you sure they are errors and not just warnings?	15:17
doener_	normally, those are just warnings (unless you configured it to make errors out of them)	15:18
aeruder	if they are in fact errors, that means for whatever reason you probably made your hooks executable	15:18
doener_	if you want the whitespace noise to be fixed, use --whitesapce=fix	15:18
	s/sap/spa/	15:18
	→ brasko joined	15:18
kazim59	aeruder: it leaves a .dotest directory, and wants me to do something	15:19
	→ machrider joined	15:19
aeruder	kazim59: how are you applying the patch?	15:20
machrider	so, when you checkout a branch in git, it modifies whatever files in your working tree are different	15:20
kazim59	aeruder: git am < patchfile	15:20
machrider	has anyone ever modified git to instead make each branch a subdir?	15:20
	so you can switch around without rebuilding the whole tree?	15:20
	→ rwaite joined	15:21
doener_	machrider: see git-new-workdir in contrib/	15:21
	→ fafhrd_ joined	15:21
aeruder	kazim59: is the patch currently staged? (i.e. git status shows things ready to commit) ?	15:21
machrider	doener_: interesting... is it safe to use?	15:22
brasko	doener_: git-1.5.6.4/contrib doesn't have that file	15:22
aeruder	machrider: a lot of people use it, so it is kept up to date	15:22
brasko	doener_: is it in a newer version?	15:22
machrider	awesome	15:22
aeruder	brasko: it is actually contrib/workdir/git-new-workdir	15:23
	← gcw\|secution left	15:23
pasky	<v1.5.6.4:contrib/workdir/git-new-workdir>	15:23
Gitbot	[git git-new-workdir]: http://tinyurl.com/5ormw8 [blob]	15:23
pasky	it's there	15:23
	← BBHoss left	15:23
	← draco_ left	15:23
aeruder	kazim59: if so, use git am --resolved	15:23
	doener_ is just too lazy to be a find(1) replacement	15:24
	← mehemiah left	15:24
	→ NewMonarch_ joined	15:26
NewMonarch_	Is it bad to have use a hyphen in a branch name?	15:26
cehteh	no	15:26
	starting with one might be inconvinient :)	15:26
aeruder	might be a bad idea to make the first letter a hyphen ;)	15:26
	→ Weasel[DK] joined	15:27
	→ alley_cat joined	15:27
	← f0rkit left	15:27
	← rmh3093 left	15:27
NewMonarch_	I've got a branch called intl-orders and I can't check it out.	15:27
cehteh	git branch '\`rm -rf\`'	15:27
	→ f0rkit joined	15:27
	→ bronson joined	15:27
	→ rmh3093 joined	15:27
NewMonarch_	it wasn't a local branch to begin with but I first pulled and now I'm trying to check it out	15:27
	→ netoman joined	15:28
pasky	and why do you think the dash is the culprit?	15:28
	← brasko left	15:28
	← fafhrd left	15:28
NewMonarch_	Cause I didn't have any other ideas. :D	15:28
aeruder	i'm still waiting for the clarification of "I can't check it out."	15:29
NewMonarch_	No, I've been using Git for months but haven't gotten into branching yet.	15:29
	git checkout intl-orders => error: pathspec 'intl-orders' did not match any file(s) known to git.	15:29
machrider	git branch -a	15:29
lea_w	git am doesn't like being fed a MIME patch: "fatal: `pos + len' is too far after the end of the buffer" The patch is here: http://pastebin.ca/1178740 What can I do (apart from editing the patch)?	15:29
NewMonarch_	I did git checkout origin/intl-orders but that seemed to freak it out. My branch was "no branch" or something.	15:30
pasky	didn't you say the branch is not local?	15:30
machrider	you should create a local branch that tracks the remote one	15:30
NewMonarch_	pasky: yeah. Do I need to add the branch?	15:30
machrider	if you want to work on it	15:30
NewMonarch_	ok didn't realize that	15:31
	I've only pushed branches until now.	15:31
	How do I go about that?	15:31
pasky	git checkout -t -b intl-orders origin/intl-orders	15:31
	I wonder if the -t is redundant now?	15:31
jast	has been for ages	15:32
	tracking is the default for this type of command	15:32
	→ zawir joined	15:33
jast	in git 1.6.1 it will probably be even easier: git checkout -t origin/intl-orders	15:33
NewMonarch_	got it... thanks. Wish that were a little more intuitive. :D	15:33
	→ schacon joined	15:33
kazim59	aeruder: alright!	15:35
	→ ijcd joined	15:35
kazim59	How do I rename a branch?	15:35
jast	git branch -m. see the git-branch manpage.	15:35
	→ BBHoss joined	15:36
doener_	lea_w: hm, maybe message id [email@hidden.address]	15:36
	← foca left	15:36
lea_w	doener_: thanks!	15:37
	→ foca_ joined	15:38
	← keseldude left	15:39
	← ^matthieu^ left	15:39
lea_w	doener_: it doesn't actually apply on next for me. well, at least people are aware of the problem then	15:39
doener_	oh well, I only remembered having seen that, I'm a master using pansy	15:41
lea_w	doener_: hm, that patch is in next already	15:41
	doener_: weird that it still doesn't work	15:41
	foca_ → foca	15:41
lea_w	i'll post a note to the list i think	15:41
doener_	did you install the new version yet? Maybe some old dashed executable still in bindir?	15:42
	→ rompus joined	15:42
	→ ijcd_ joined	15:42
doener_	(just guessing, feel free to ignore me ;-))	15:43
	← ciaran29d left	15:43
	→ jeffpc joined	15:43
	→ ciaran29d joined	15:44
	→ igorgue joined	15:44
heptadeca	I'm trying to educate my co-workers about git; we're using it for a small project on a team of five people. I'm under the impression that it's generally a good idea to create branches for local work, then merge them and push.	15:45
	← qrush left	15:45
	→ advany joined	15:45
heptadeca	One of my co-workers is of the opinion "the way I worked in CVS was to just do all my work on the thing I checked out, then push that up. git supports it, so why do I need to branch?" Should I try and convince him otherwise?	15:45
doener_	yep, either merge or rebase (-i),	15:45
	definitely	15:45
	branches help to keep independent work separated and gets you a way saner history	15:46
	sounds like you should also convince him not to produce huge "here's my last month of work" commits	15:46
heptadeca	But this creates more work for him, which turns him off to git. (CVS-only knowledge)	15:47
doener_	well, it pays off the first time you need to make use of the history you created	15:48
	← igorgue left	15:48
heptadeca	How does branching make a saner history? Because it's linear, rather than merge-y?	15:48
doener_	if you use your scm as some write-only backup mechanism, of course the history seems to have little value. But when you actually _use_ your history, for example for debugging purposes, it's quite worthwhile to keep that clean, at least to some degree	15:49
	no, branching on its own doesn't create a linear history. But if eg. your current task needs 20 commits and you get something else to do in between, you can just switch to a new branch, starting _before_ your new stuff	15:50
	then you do that other thing and switch back to the other tasks	15:50
	← bentob0x left	15:51
doener_	so commits are kind of "grouped together" by topic	15:51
heptadeca	Hm. Part of this might be due to long-term training in CVS, because he claims not to work on more than one thing at a time. And when he does, it's because he wants to see both at the same time.	15:52
doener_	and not some A-B-C-C-A-A-B-B-D-D-A-A-A-B-C mix, where every other commit deals with a topic totally different from the previous one	15:52
	← NewMonarch_ left	15:52
	← advany left	15:52
	← sgrimm left	15:52
heptadeca	Man, that sounds awesome from my perspective, but it does sound tricky as to convince CVS-only developers as to the advantage of it, and to train them on how to do it.	15:52
doener_	well, then he only needs to create one branch	15:52
	← BBHoss left	15:53
	→ sgrimm joined	15:53
heptadeca	Ah, then his argument is "why create the branch? Why not just work in what I checked out?" Which is one I don't know the answer to.	15:53
doener_	if he's totally strict about the "one topic only" thing, the one branch he needs is already there, master	15:54
	← Sput left	15:54
	← EgS left	15:54
	→ EgS joined	15:54
heptadeca	OK, so his current method is correct, I'm just being anal. Second question: With this workflow of his, working in master, is it safe for him to rebase and thus clean up his history?	15:55
	→ Sput joined	15:55
doener_	as long as he doesn't rebase anything that he pushed or which has been pulled from him, sure	15:55
heptadeca	Additionally, (I don't expect to understand the answer) "why can't a rewritten history get shipped down when I pull?"	15:56
doener_	it does	15:56
heptadeca	But I thought if I rebased a branch and rewrote its history, I couldn't do it to a publicly-used branch.	15:56
	← Sput left	15:57
	→ Sput joined	15:57
doener_	"rewritten" means "create a new series of commits, that probably do similar stuff (or even the same)"	15:57
	but still, they're new commits	15:57
	but those that already pulled the old commits will still have those old commits	15:57
	they don't get magically removed	15:58
heptadeca	So, why would that be a problem that other people have old commits?	15:58
doener_	they're basically the same as any other local commit	15:58
	← ijcd left	15:58
doener_	pull is fetch + merge.	15:58
heptadeca	Wouldn't their history be updated to point at the new ones?	15:58
	And the old commits just become things to be garbage-collected?	15:58
doener_	so it first fetches the new history, and then merges that with the old one	15:58
jast	not by default because the new ones no longer have common history with the old ones	15:59
doener_	so the created merge commit has the old history as one parent, and the new history as the other	15:59
	← rajasaurus left	15:59
	→ ereslibre joined	15:59
jast	so fast-forward is out at any rate	15:59
	→ bts- joined	15:59
doener_	heptadeca: the history is "replaced" for the remote tracking branch (eg. origin/master)	15:59
jast	what happens is that it tries to merge the old commits with the new ones and that'll get you heaps of conflicts because both do pretty much the same	16:00
	← jmesnil left	16:00
doener_	heptadeca: but your local branch (eg. master) still has it	16:00
jast	if fetch doesn't complain and abort first	16:00
doener_	heptadeca: it's pretty much the same as creating two branches, doing mostly the same thing twice (once on each branch) and then merging the two	16:01
kazim59	How do I rename a branch name?	16:01
doener_	17:35:28 <kazim59> How do I rename a branch?	16:01
	17:35:49 <jast> git branch -m. see the git-branch manpage.	16:01
heptadeca	OK, now I'm starting to understand. Thanks, doener_ & jast!	16:03
doener_	jast: somewhat recent versions of git place a + in the fetch line by default (though it was broken for some versions)	16:03
jast	that's a bit evil, i think	16:03
doener_	jast: and "git pull remote branch" uses FETCH_HEAD anyway, so that wouldn't complain either	16:03
jast	yeah, but wouldn't the fetch part complain (when there's no +) if no fast-forward update is possible?	16:04
	→ bobesponja joined	16:04
doener_	only for "git pull", not for "git pull remote branch"	16:04
	FETCH_HEAD is a temporary fetch-only ref	16:04
	ie. that doesn't affect the remote tracking branches at all	16:05
	← f0rkit left	16:05
jast	hmm, how come i didn't know that :(	16:06
	→ Eludias joined	16:06
	← sverrej left	16:07
	← BCarlyon\|Laptop left	16:07
	→ bdiego joined	16:10
	→ lu_zero_ joined	16:18
	→ CIA-2 joined	16:20
	← Lash\| left	16:21
	← Whtiger left	16:22
	← diask left	16:22
	→ NewMonarch joined	16:23
	→ Whtiger joined	16:24
	← MentalMaelstrom left	16:25
	← reel_ left	16:25
	→ peeja_ joined	16:25
peeja_	is it possible to git-reset to the index, blowing away unstaged changes?	16:26
doener_	git checkout -- .	16:27
	← priidu left	16:28
	← sgrimm left	16:28
	→ sgrimm joined	16:28
peeja_	doener_: hmm, that left my local changes intact	16:28
	← kazim59 left	16:29
doener_	changes to tracked file, or new files?	16:29
	tracked files should be reverted to their state in the index	16:30
peeja_	tracked files	16:30
doener_	or did you delete them from the index?	16:30
peeja_	it gave me "M <filename>" for each file	16:30
	no, they're just files with unstaged changes	16:30
Mikachu	peeja_: are you sure you ran git checkout -- . and not git checkout -- ?	16:31
peeja_	oh	16:31
	heh	16:31
Mikachu	where the . is part of the command, but the ? isn't ;)	16:31
peeja_	thanks :)	16:31
	I read that as the end of the sentence	16:31
	yeah, that works. many thanks :)	16:31
doener_	thanks Mikachu! I would have asked a gazillion of other things, but not have thought of that :-)	16:31
Mikachu	hehe, it already happened to me once, and i asked those gazillion other questions that time :)	16:32
	→ jones_ joined	16:33
	→ igorgue joined	16:38
	← rdavila left	16:41
	→ krawek joined	16:43
	← bdiego left	16:43
	← Yuuhi left	16:44
	→ Yuuhi joined	16:44
	→ bdiego joined	16:46
	→ perezd88 joined	16:46
	← Yuuhi left	16:47
	→ hurikhan\|Work joined	16:47
	→ langenberg joined	16:48
	← redondos left	16:48
	→ thannoy joined	16:49
hurikhan\|Work	hey there... I've just set up gitosis which worked great. i added new repositories to the config, pushed it and the config was updated on the server as expected. nowi wanted to push my still local-only repos to the remote but git just tells me "No refs in common and none specified; doing nothing." - however it created an empty repo on the remote end...	16:49
Mikachu	hurikhan\|Work: --all	16:50
	← jones- left	16:50
hurikhan\|Work	Mikachu, still the same error	16:50
	replacing --all by "master" yields another error: "src refspec master does not match any." so it doesn't help either	16:51
	← bryanl left	16:51
hurikhan\|Work	can i rely on gitosis inititating an empty repo on itself correctly?	16:51
rwaite	no more git-* woo.	16:53
hurikhan\|Work	is it ok to use subdirs like so: "writable = customer/project1 customer/project2" in gitosis.conf?	16:54
doener_	are you trying to push from an empty repo?	16:57
	→ ciaran joined	16:59
	→ WorkingGeier joined	17:01
hurikhan\|Work	doener_, i suppose it was empty because on first try to push it was created by gitosis	17:03
	oh wait, doener_, not _from_ an empty repo	17:04
	my local repo already contains data	17:04
	→ _john joined	17:04
	← _john left	17:04
doener_	and there's a branch called master in that repo?	17:05
hurikhan\|Work	"git branch -a" yields nothing	17:06
	should i create a master branch?	17:06
doener_	well, then the repo is empty (I'm still talking about the repo _from_ which you push)	17:06
	← bugfux left	17:07
	→ joydivider joined	17:07
	← langenberg left	17:07
hurikhan\|Work	doener_, oh yes you are right... not only i confused to repositories - i also forgot to commit files to the other :-(	17:08
	← NewMonarch left	17:08
hurikhan\|Work	thanx for the advice, pushing from the correct repo worked	17:08
	→ qrush joined	17:09
	→ Lash\| joined	17:19
	→ bcarlyon\|laptop joined	17:20
	← koke left	17:22
	→ pygi joined	17:24
	← eikonos left	17:26
	→ queso joined	17:27
	← bobesponja left	17:27
	→ bobesponja joined	17:28
	← eddyp_work left	17:28
	peeja_ → peeja	17:31
	→ ciaran_ joined	17:31
	← EgS left	17:31
	← Sput left	17:31
	→ EgS joined	17:31
	→ Sput joined	17:32
	→ brendon_work_ joined	17:36
	← brendon_work left	17:36
	→ eikonos joined	17:37
	← foca left	17:41
	→ foca_ joined	17:42
	→ langenberg joined	17:42
	← Whtiger left	17:43
	→ gmaxwell joined	17:43
	→ lanaer joined	17:44
	→ _Vi joined	17:45
	→ brendon_work joined	17:46
	← brendon_work_ left	17:46
	← ondrej left	17:47
	→ flaguy joined	17:47
	mw → mw\|lunch	17:48
gmaxwell	Has anyone produced a date-proofing service for GIT? The idea is that you send them the hash of a commit, and they respond back with a signed git tag containing a timestamp for that commit. The application would be providing independent proof that a particular version existed on or before a specific point in time.	17:48
	← greedo left	17:48
Tv	gmaxwell: you can use a generic timestamping service	17:49
	→ greedo joined	17:49
	← lanaer left	17:49
gmaxwell	Tv, sure, For workflow reasons it would be nice if it integrated directly into git.	17:49
	← sbahra left	17:50
	→ CodeOfficer joined	17:51
_Vi	How to tell to git-merge to always resolve conflicting places to one fixed side? (E.g. instead of ' <<< aa === bb >>>' just write ' aa ' and report that there's no conflicts?	17:52
	← ciaran left	17:52
	foca_ → foca	17:52
	← bcarlyon\|laptop left	17:53
	← Aikawa left	17:54
	→ bcarlyon\|laptop joined	17:54
	→ robinr joined	17:54
	→ Aikawa joined	17:55
	← langenberg left	17:56
	→ Whtiger joined	18:00
drizzd	_Vi: git merge -s ours	18:00
doener_	that goes for _all_ files though, not just conflicting ones	18:01
_Vi	I've just read what '-s ours' do mean, it tell that it changes no files at all, just records that is is formally merged.	18:01
	← warthog9 left	18:01
com4	The ip address of my svn repository changed. I updated .git/config and .git/svn/metadata but I get the error "Unable to determine upstream SVN information from working tree history" when i try to rebase. How do I fix this?	18:02
drizzd	ok, that I don't know	18:02
Mikachu	wasn't some option added recently that might help with that?	18:02
	dunno what it actually was though	18:02
	→ foca_ joined	18:03
_Vi	Something like 'if(no_conflict(chunk){ auto_merge(base,chunk); }else{ /* Just ignore it */ }'.	18:03
doener_	com4: try setting rewriteRoot as well (to the old url)	18:04
	com4: if you're lucky, that might do	18:04
com4	doener_: where?	18:04
doener_	in the config, see git-svn(1)	18:04
drizzd	maybe it's possible to extract the partly auto-merged version from one of the stages	18:04
	have you tried looking at git ls-files -u?	18:05
	← brendon_work left	18:05
com4	doener_: nope :(	18:06
parasti	com4: or maybe use git filter-branch to rewrite the git-svn-id lines	18:06
doener_	com4: if it is an option for you, do a new clone	18:06
	com4: if you have not yet dcommitted commits, you can transfer them to the new clone	18:07
	→ snitko joined	18:07
	RandalSchwartz averts his eyes at SVN-related messes	18:08
RandalSchwartz	_Vi - "merge one direction if a conflict" is almost always a sign that you're using version control to do something else's work	18:08
	why wouldn't you want to decide which of two independent edits are more correct?	18:09
	especially if that means that only half of a correlated edit is accepted, and the other half not. that's broken.	18:09
_Vi	RandalSchwartz, There's a lot of git-unfriendly and not-quite-necessary auto-generated data bundled with file which data should be tracked. /* Probably some other way will solve this problem better */	18:10
drizzd	what if you limit it to rejecting only conflicting commits, that's like automatic cherry picking	18:10
RandalSchwartz	Yeah - stop version-controlling generated stuff	18:10
_Vi	(The file is OpenOffice .fodt document)	18:10
	→ polpak joined	18:11
RandalSchwartz	git is for sources, not products	18:11
	drizzd - same problem	18:11
_Vi	Git works for html files. Is html file a source of product?	18:11
RandalSchwartz	suppose file fred had a conflict on A and B, but wilma accepted B and betty accepted A.	18:11
	if you bias fred towards A, then the wilma edit is wrong	18:12
	if you bias fred towards B, then the betty edit is wrong	18:12
	→ tetraedr joined	18:12
RandalSchwartz	you really need to figure out why the conflict is there, and resolve everythign globally in step	18:12
tetraedr	hi	18:12
_Vi	OK, probably if I need to git OOo documents I should write a little script that purifies them.	18:12
tetraedr	is there any manual or sort of comprehensive guide on git?	18:12
RandalSchwartz	tetraedr - besides all the docs that come with it?	18:13
tetraedr	RandalSchwartz: sort of	18:13
	:)	18:13
	← Weasel[DK] left	18:13
tetraedr	just I like to read books	18:13
drizzd	tetraedr: http://git.or.cz/#documentation	18:13
tetraedr	and stick around them	18:13
RandalSchwartz	not getting it. maybe I'm dense	18:13
	what are you missing?	18:14
tetraedr	RandalSchwartz: I just don	18:14
drizzd	that comes with git, but it's online ;-)	18:14
tetraedr	't like to run around whole google to look for something	18:14
RandalSchwartz	yeah - stop googling	18:14
	read the stuff that comes with git	18:14
	if you still want more, look at the wiki	18:14
	← foca left	18:14
tetraedr	git - the stupid content tracker - goot start :)	18:14
	→ langenberg joined	18:15
_Vi	Where I should read about how I should organise my commits and when I should branch/merge, what comments I should write for them, patterns/anti-patterns of using RCSes, etc.?	18:15
	← ciaran_ left	18:15
	→ Weasel[DK] joined	18:16
	→ lacky33 joined	18:16
	→ aquabu joined	18:16
	← Whtiger left	18:16
drizzd	tetraedr: you probably don't want to read the manpage, but the user manual	18:16
lacky33	Anyone here able to lend a hand getting gitosis working?	18:16
	→ Whtiger joined	18:17
	← BabelO left	18:17
	→ advany joined	18:17
tetraedr	drizzd: sort of :)	18:18
	→ MattDiPasquale joined	18:18
	foca_ → foca	18:18
	← EgS left	18:19
	← Sput left	18:19
	→ EgS joined	18:19
	→ Sput joined	18:19
	← queso left	18:20
	→ warthog9 joined	18:20
	→ BabelO joined	18:21
robinr	tetraedr: its the content that is stupid, not the tracker	18:23
tetraedr	robinr: I've got it :)	18:23
	← gmaxwell left	18:24
	→ ciaran joined	18:24
	← BabelO left	18:24
	→ david_koontz joined	18:24
	→ brendon_work joined	18:24
	→ almostautomated joined	18:25
	→ BabelO joined	18:26
	← adante left	18:28
	→ EmilMedve joined	18:30
lacky33	anyone here have any idea what this means? http://gist.github.com/6213	18:31
	→ drizzd_ joined	18:31
pasky	maybe incomplete gitosis installation or something?	18:31
	← EmilMedve left	18:32
lacky33	install didn't raise any errors :\	18:32
	← Whtiger left	18:33
	← zachinglis left	18:34
	→ jstad joined	18:34
	→ Whtiger joined	18:35
	→ boink__ joined	18:37
	← flaguy left	18:37
	→ chuckr joined	18:37
	← igorgue left	18:38
chuckr	I need to find a way to get a tag from a git archive that wouold be approximately 6 months old, how would I do something like that?	18:38
	Oh, I left off, after I found the tag, I need to be able to update my git archive to that level	18:38
j_engelh	chuckr: git describe $(git log --until="6 months ago" --pretty=format:%H -1)	18:39
chuckr	I guess there isn't any reason I couldn't say that I'm talking about the Xorg archive. They have a nice script (git_xorg.sh) that updates all of the dirs, I just need to know how to modify the command that updates	18:40
j_engelh	once you know the tag, git checkout is easy	18:40
chuckr	that's going to get me the tag, that's half of it ...	18:40
	(I know cvs really wellm git far less well)	18:40
	→ alenoosh joined	18:40
j_engelh	git checkout <yourtagname>	18:42
	shush shush now go checkout :p	18:42
	→ adante joined	18:42
alenoosh	hi all , I have noticed different announcements about the latest stable release of git : one says 1.5.6.5 and the other 1.6.0 , which one is right ?	18:42
j_engelh	both.	18:42
alenoosh	j_engelh, ?	18:42
chuckr	thanks, got the tag and now I'm going to do the checkout, do appreciate it. For some reason, I'm having a really bad time witht he git syntax	18:43
	← drizzd left	18:43
j_engelh	alenoosh: if both are declared 'stable' (rather than say, oldstable or whatever you want to call it), then the one with the higher version number wins	18:43
	→ foca_ joined	18:43
alenoosh	in git.or.cz , it is 1.5.6.5 but in git-scm.com it is 1.6.0	18:44
j_engelh	the git.or.cz webpage was not updated, 1.6.0 is on the ftps.	18:44
Mikachu	git.or.cz says 1.6.0 here	18:44
alenoosh	j_engelh, oh really	18:45
j_engelh	yeah yesterday it was not updated ;)	18:45
Mikachu	iirc git.or.cz just looks at the maint branch	18:45
	← hurikhan\|Work left	18:47
chuckr	j_engleh, just to let you know, I got my script modified, and everything worked ok (once i realized you had to have been referring to a bash syntax, I usually use tcsh as a user shell, ksh as a programming shell)	18:47
	mw\|lunch → mw	18:47
alenoosh	yes you are right , I cleared my history , now it's 1.6.0 , thanks	18:48
	← raz left	18:48
	boink__ → raz	18:48
	← Whtiger left	18:52
	← ferdy left	18:54
	→ Whtiger joined	18:56
	← sergiodj left	18:58
	→ ferdy joined	18:59
	← Cap_J_L_Picard left	18:59
	→ Cap_J_L_Picard joined	18:59
alenoosh	j_engelh, will upgrading to new version of git , cause problems for repositories that have used previous version of git ?	19:00
j_engelh	I have not had any.	19:00
Mikachu	you could get duplicate hooks in new repos if you don't use a package manager	19:00
lacky33	anyone here have any idea what this means? http://gist.github.com/6213	19:00
godlygeek	Mikachu: why git-for-each-ref to get branch names, and not git-rev-parse --symbolic-full-name --branches --remotes ?	19:01
	← tiglionabbit left	19:01
	→ jmalonzo joined	19:01
	← mithro left	19:02
	→ mithro joined	19:02
	← _Vi left	19:03
Mikachu	--branches is sort of new, i think	19:03
godlygeek	ok. figured there was probably a reason, just figured i'd ask since it looks a lot simpler.	19:04
almostautomated	anyone feel up to helping an msysgit update question (first time user of git)? Installed msysgit-fullinstall-1.5.4...., noticed that that didn't install the shell extension for GUI, so then did git fetch which gave an origin error, then did git fetch git://repo.or.cz/msysgit.git followed by cd git && make install... still showed 1.5.4 so...	19:04
	→ sbahra joined	19:05
almostautomated	grabbed the 1.5.6.1 preview installer, installed it to the same dir I had used for 1.5.4	19:05
	→ zawir_ joined	19:05
	→ gcw\|secution joined	19:05
	→ redondos joined	19:06
	← bdiego left	19:06
almostautomated	the git shell showed 1.5.6.1 and the msys shell showed 1.5.4 still so I did make install again... bad idea :)	19:06
	← Whtiger left	19:06
almostautomated	now both the git shell and msys shell show 1.5.4 and the gui shows 1.5.6.1	19:06
	← zawir left	19:07
Mikachu	you have to uninstall the old binaries first probably	19:07
	→ Whtiger joined	19:07
	← alenoosh left	19:07
	→ bdiego joined	19:07
	← BabelO left	19:08
xinming	WHenever I use a git-<command>, It will bring up pager automatically, How to avoid this pleaese?	19:09
Mikachu	man git-config	19:09
	→ queso joined	19:09
Mikachu	/pager	19:09
almostautomated	hrrmm... I was thinking that I did the pull make install incorrectly, shouldn't I be able to pull the latest and make install to update git? or a better question would be how would I update msysgit to the latest?	19:10
	→ Bryson joined	19:10
hachi	If I try something silly like: git svn clone -treleases/ -b/	19:10
	does anyone know if it will do 'the right thing' ? :)	19:10
	← Whtiger left	19:10
	→ dv_ joined	19:10
dv_	hi	19:10
Mikachu	almostautomated: 1.6 stopped shipping the git-* binaries in $path, so it won't overwrite the old ones	19:10
dv_	if I type git branch <branch>, a new branch is created based on the contents of <branch>	19:11
Mikachu	dv_: no	19:11
dv_	but what if I want a completely empty branch, where I can put fresh new files?	19:11
	← adante left	19:11
	→ _graham_ joined	19:11
dv_	sorry, I meant create <branch> based on HEAD	19:11
Mikachu	dv_: git symbolic-ref HEAD refs/heads/newbranch; rm .git/index; git add files; git commit -m hooray	19:11
	→ Whtiger joined	19:12
xinming	Mikachu: What do you mean please?	19:12
	→ ciaran_ joined	19:12
dv_	wow. this looks ... scary	19:13
Mikachu	xinming: man git-status, look for "core.pager"	19:13
	xinming: err, git-config	19:13
	Mikachu decides 3 people is too many to help at the same time	19:13
xinming	Mikachu: one is done. :-)	19:14
almostautomated	Mikachu: thanks :) make that two	19:14
	← janm left	19:14
	→ Sonderblade joined	19:15
	lu_zero_ → lu_zero	19:16
	← gcw\|secution left	19:17
	← frq1_dave left	19:19
	→ univate_ joined	19:19
hachi	error: unable to create directory for .git/refs/remotes/tags/ircd-hybrid-7.2.0	19:19
	← sbahra left	19:20
hachi	tags is an existing file	19:20
	why is git trying to create a directory the same name as a file it already created?	19:20
dv_	Mikachu, so you would suggest to branch it and then get rid of all the files in the new branch?	19:21
Mikachu	dv_: i don't know if i was suggesting anything, as much as i was telling you exactly what to do	19:21
dv_	well I dont do rm anything until I fully understand it	19:22
Mikachu	removing .git/index is the same as git rm --cached -r ., but easier	19:22
dv_	alright	19:23
Mikachu	and if someone tells you rm on irc and you don't trust them, just use mv instead :)	19:23
dv_	git symbolic-ref HEAD refs/heads/newbranch <- this still puzzles me	19:25
	I copied a local repository as a testing ground,	19:25
jast	hachi, what did you do that resulted in that message?	19:25
dv_	and destroyed it using this command	19:25
	→ BabelO joined	19:25
hachi	earlier paste	19:25
Mikachu	then you didn't use that command	19:25
dv_	"fatal: Not a git repository"	19:25
hachi	-treleases/ -b/	19:25
Mikachu	you probably misspelled refs/heads/	19:26
dv_	probably	19:26
	but there is no way back, is there?	19:26
Mikachu	of course there is, just edit .git/HEAD :)	19:26
dv_	git symbolic-ref HEAD refs/heads/master refuses to work	19:26
	→ EmilMedve joined	19:26
hachi	I would blame my incantation, but I see no reason why it would force git's hand in creating a directory for a file that already exists	19:26
	← EmilMedve left	19:26
	← peeja left	19:27
	→ sbahra joined	19:27
jast	hachi, does that file have any interesting contents?	19:27
	← lacky33 left	19:27
	← univate left	19:27
hachi	it has a SHA1	19:28
j_engelh	I haz a SHA1, but I committed it.	19:28
hachi	k/ick j_engelh	19:28
jast	I made you a SHA1	19:28
Mikachu	but i garbage collected it	19:28
jast	:(	19:28
	← ciaran left	19:29
	→ tiglionabbit joined	19:29
jast	it's a bit surprising that refs/remotes would contain a ref directly	19:29
	especially when the repository you cloned from doesn't have that	19:29
	you could check if it has broken packed refs	19:29
	→ jerbear joined	19:29
dv_	hmm	19:30
	← krawek left	19:30
robinr	omg, my egit patch bucket is overflowing	19:30
dv_	if I modify a file in branch A,	19:30
	→ krawek joined	19:31
dv_	and commit it, then checkout branch B,	19:31
	← bdiego left	19:31
dv_	the file isnt marked as modified	19:31
	oh wait.	19:31
	dv_ stupid	19:31
	→ bdiego joined	19:32
	→ towski joined	19:32
	→ d0k joined	19:34
dv_	hmmm any chance on seeing the differences between branch A and B?	19:35
j_engelh	git diff A..B	19:35
dv_	ah	19:36
	I tried git diff A B	19:36
	thanks	19:36
	→ clarby_ joined	19:39
j_engelh	dv_: there is no file called 'B' :)	19:39
drizzd_	git diff A..B and git diff A B are synonymous	19:41
j_engelh	git diff A B sounds to me like git diff A..workingcopy B	19:41
clarby_	Ok, so .. I have a repository on my computer, my friend one on his. If he wants to receive my updates, does he not need access to my account credentials? (we're using ssh). How do people go about this? Add users to your system that you work with, and make the repository readable bya common group or something?	19:41
j_engelh	and git diff --cached A B would git diff A..index B	19:41
	← pygi left	19:45
	→ pygi joined	19:45
Eludias	clarby_: You can either push your updates, or he can pull your updates. Pulling requires read permissions on the 'server', pushing requires write permissions on the 'server'.	19:45
	clarby_: A 3rd option is to have a real server to which both of you push/pull.	19:46
clarby_	Eludias: Ok, we'll go with the readable permission stuff thing :)	19:46
Eludias	clarby_: Only work by pulling is the easiest. In that case you just 'read' the others repo.	19:46
dv_	ok now that I created and committed to branches,	19:47
	→ ez joined	19:47
dv_	a simple git push will push all new branches as well?	19:47
Eludias	clarby_: In that case your friend needs to be able to read your files. Which probably means he has an account on your machine.	19:48
dv_	or do I need --all here? (I dont know what the manpage means by "refs")	19:48
drizzd_	refs are branches and tags	19:48
dv_	okay so I do need --all	19:48
	right?	19:49
clarby_	How about his one: We've both checked out a copy of a repository using git-svn. We've been working on different things, comitting to our local git repositories. What happens when I try to pull his version? :)	19:49
	→ adante joined	19:50
	← bgerber left	19:50
jast	sometimes it works well, sometimes it doesn't. anyway, if it works and produces merges, committing those back to svn is usually not a good idea. you'll want to squash stuff together etc., and then the other guy's repository and yours will definitely be out of sync. so even if everything works well you'll have to talk to each other for some operations	19:51
dv_	guys? git push --all is ok?	19:51
	I don't want to fsck up other's repositories	19:52
jast	push --all is required if you want to push all of your local branches, even those not existing on the other side	19:52
	← advany left	19:52
jast	push without arguments only considers branches that exist on both sides	19:52
dv_	which is the case here; I just created one	19:52
jast	you can also specify a single branch to push	19:53
	→ WorkingG1ier joined	19:53
jast	personally i think it's best not to get into the habit of using --all if all you did was to create a single new branch	19:54
	→ botanicus_ joined	19:54
	→ jaalto joined	19:54
drizzd_	I believe git push --all is equivalent to git push origin 'refs/heads/*'	19:55
	← snitko left	19:55
	← botanicus left	19:55
jast	not exactly, i think. refs/heads/* is probably not a good refspec.	19:55
	→ sexy_man_XD joined	19:55
jast	oh well, it probably is, usually	19:55
	→ t joined	19:55
drizzd_	I think that's again short for 'refs/heads/:refs/heads/'	19:56
	← jtong left	19:56
	→ redondos_ joined	19:57
jast	yeah, it probably is. i was thinking about something else but that was a dead end	19:57
	← WorkingGeier left	19:57
	WorkingG1ier → WorkingGeier	19:57
	← yatiohi left	19:59
	← dv_ left	19:59
	→ yatiohi joined	20:00
	← sexy_man_XD left	20:00
	← t left	20:01
	meyering → jim-away	20:01
	→ sverrej joined	20:02
	← Lash\| left	20:02
	← nostromo left	20:05
	← alley_cat left	20:06
	← eikonos left	20:06
	→ rdavila joined	20:07
	→ nostromo joined	20:07
clarby_	jast: Ok. So now he pulled mine, got two conflicts. Should I then let him fix those, and I'll pull, and then we can dcommit to the svn repository?	20:08
jast	you don't both have to merge	20:09
	because when you dcommit, in the svn repo the merges pretty much disappear, so if you now each have slightly different merges, i can't guarantee git-svn won't get confused on one side	20:09
	what i did was this, roughly:	20:09
	→ toxx_ joined	20:11
jast	as soon as all of your non-svn commits are integrated in one repository, create a series of linear commits from them. in fact i avoided merges altogether to make this easier. then dcommit, throw away the non-svn commits and git-svn rebase.	20:12
	it's ugly, admittedly, but i've had wildly different experiences with dcommiting git merges	20:12
	← redondos left	20:12
	redondos_ → redondos	20:12
jast	at least even with crippling your workflow that way, using git-svn is more pleasant than using svn itself	20:14
	→ BrianTheCoder joined	20:16
	← Eludias left	20:18
	← rompus left	20:22
	← rdavila left	20:22
	← MattDiPasquale left	20:23
	← Aikawa left	20:25
	← langenberg left	20:26
	→ Aikawa joined	20:27
	← jmalonzo left	20:27
	→ ams joined	20:28
ams	any quick way to check if you are in a git repository or not? (like hg root or some such)?	20:28
robinr	git rev-parse --git-dir	20:29
ams	lovley	20:29
	thanks	20:29
	← ams left	20:29
	← tetraedr left	20:30
	→ ams joined	20:30
ams	hi again! :-)	20:30
	← d0k left	20:31
	← schlort left	20:31
	→ Lash\| joined	20:32
	→ capisce joined	20:32
	→ GITNoob joined	20:33
GITNoob	does anyone have experience with the submodule capabilities?	20:33
	i've been having issues with pushing a submodule back to our master repository and I can't seem to find any documentation on pushing submodules back to origin repo	20:34
	→ schlort joined	20:36
GITNoob	can anyone see this?	20:38
	→ _Vi joined	20:39
	→ waltman joined	20:39
waltman	I want to pull a single file from a git branch to master. is that "git checkout master; git checkout mybranch foo.pl; git commit"?	20:40
	that doesn't seem to get its history though.	20:40
ams	any way to check if a file is registered in git?	20:40
vmiklos	many :)	20:41
	try git ls-files	20:41
kergoth`work	waltman: you cant just pull one file with history, it doesnt work that way, not without using filter-branch to rewrite those commits to include just that file, and then pull those over, anyway..	20:42
	well, afaik.	20:42
	→ MattDiPasquale joined	20:42
waltman	kergoth`work: can I pull just the revision where that file changed?	20:43
GITNoob	sorry to spam.. but am i voiced (can anyone besides me see this messsage?)	20:43
ams	vmiklos: thank you	20:43
kergoth`work	GITNoob: the channel isnt moderated, you dont need to be voiced	20:44
GITNoob	cool. thanks	20:44
	→ methods joined	20:44
methods	what's the difference from a git push to commit ?	20:44
	→ zachinglis joined	20:45
	← capisce_ left	20:45
smartcat99s	methods: commits happen on the local machine, and push shares commits out to other repositories	20:45
jast	methods, i suggest you read http://en.wikipedia.org/wiki/Distributed_Version_Control_System	20:46
up_the_irons	methods: commit does not talk to any server, it commits locally. Push shares your local commits to a server	20:46
methods	i use svn allot	20:46
GITNoob	looking for help with submodule push to master repository. I've become an expert with submodule adds but am having all kinds of trouble pushing a submodule to the master repo. any help appreciated thanks	20:46
up_the_irons	methods: and yes, you need to read about distributed VCS's :)	20:46
methods	is svn not a dvcs ?	20:47
up_the_irons	methods: svn is centralized, git is distributed.	20:47
methods	ah	20:47
parasti	that's the d in dvcs	20:47
methods	yea it's like everyone works in a branch of their own right	20:47
	then pushes are really like merges	20:47
up_the_irons	methods: if you try to apply svn principles to git, you will be very confused. To use git properly, you need to learn the differences between distributed and centralized systems	20:48
methods	i'm just saying	20:48
jast	hey, i just found my first xhtml document in the wild that gecko refuses to render because it's invalid	20:48
up_the_irons	methods: actually, the "pull" is the merge part. push just copies the objects/refs to the server	20:48
jast	and that's why you shouldn't use xhtml :)	20:48
	(it was supposed to be about the merits of DVCS)	20:49
up_the_irons	GITNoob: any time i pondered using submodules, I found that using a simple sym-link did the trick just the same	20:50
	→ precision joined	20:51
bmalee	jast, surely that's only why you shouldn't use invalid xhtml?	20:51
	← tango_ left	20:52
	← ph^ left	20:52
jast	bmalee, check out the 'SPECIFIC PROBLEMS' section in http://www.hixie.ch/advocacy/xhtml	20:53
	← methods left	20:53
	→ dsyzling joined	20:53
GITNoob	up_the_iro: we wanted to avoid having to log on to our central GIT repository when checking in "subprojects." That's the only reason we were pursuing the subproject approach. Is there a sym-link procedure that would accomplish the same thing?	20:54
	→ bugfux joined	20:54
GITNoob	ideally we want the concept of submodule commits in GIT to be as simple to perform as CVS project imports were to your target repo	20:55
j_engelh	imports are evil :p	20:56
	← spearce left	20:57
	← pygi left	20:58
	→ ph^ joined	20:58
up_the_irons	GITNoob: i don't understand why you need to log onto the central git repo server	20:58
	GITNoob: each symlink points to a first class git repo on it's own	20:59
	GITNoob: and you updated those repos as you see fit. I know, a little more work, but it does the job	20:59
GITNoob	if we're talking vanilla symlinks then I understand why they would be easier, but doesn't think require a login to our central GIT repo (our repo where all the devs officially send commits).. and manually create symlinks?	21:00
	← GITNoob left	21:00
	← cannonball left	21:01
	← lorandi left	21:02
	→ GITNoob joined	21:06
	← netoman left	21:07
	→ mattf joined	21:07
up_the_irons	GITNoob: no problem	21:07
	GITNoob: yeah, you need to create the symlinks manually, but only once	21:08
mattf	is it possible to manage connections to multiple svn repos within a single git repo via git-svn?	21:08
up_the_irons	mattf: yes, but not really worth the pain it causes	21:08
	mattf: i did it once, then abandoned it	21:08
	mattf: just made separate repos	21:09
mattf	ok. i'm hoping to consolidate repos	21:09
GITNoob	up_the_irons.. gotcha. we could create the symlinks manually also, but if we were to take that approach, then using submodule is just as viable. We haven't experienced any problems with them as long as we create the submodules from our "master repo"	21:10
up_the_irons	GITNoob: ok	21:10
GITNoob	i guess the long and short of it is.. there's no way to get around having to log onto our central GIT repo to handle this issue. is that a valid statement?	21:10
Arrowmaster	yes if you use something that allows you to remotely create new repos, like gitosis	21:11
	← BabelO left	21:11
	→ eikonos joined	21:11
up_the_irons	well, i don't really understand why you need to log onto the central git server. Is this for continuous integration or something?	21:11
doener_	GITNoob: maybe you could explain your trouble with pushing submodules?	21:13
	→ BabelO joined	21:13
Arrowmaster	hes creating new submodules right? meaning new repos that dont yet exist on their central server so they cant be pushed to	21:14
doener_	ah, yeah, that would explain the login stuff, adding completely new subprojects	21:14
GITNoob	sure. our problem is this. We create a bare CENTRAL_repo, and clone the project to a WORKING_repo... from the WORKING_repo we add new submodules via "mkdir module_a; touch foo; git add .; git commit; cd ../; git submodule add module_a; git submodule commit; git push"	21:15
	→ tibbe joined	21:15
tibbe	could someone please clarify how to use .gitignore to ignore all files and directories under dist/ ?	21:15
	← waltman left	21:16
Arrowmaster	tibbe: create dist/.gitignore and add * to it	21:16
	← bongiojp left	21:16
Auris-	just to day I ignore a directory with: /dir/	21:16
tibbe	Arrowmaster: dist/ doesn't exist in the directory, it's created as a part of the build	21:16
GITNoob	the push appears to work.. but when attempting to clone the new subproject from CENTRAL_repo/module_a, there are complains about the folder not existing on CENTRAL_repo	21:16
parasti	tibbe: add /dist/* to the top-level .gitignore?	21:17
Auris-	parasti, why *?	21:17
tibbe	parasti: I'm trying variations on that, haven't tried that particular one. Will do	21:17
doener_	GITNoob: yeah, you should: mkdir new_submodule; cd new_submodule; touch foo; git add foo; git commit; git push ssh://server/path/to/new_submodule_repo master;	21:17
tibbe	I tried dist/ /dist/ /dist/*/	21:18
GITNoob	git-submodule status yields results for module_a from WORKING_repo.. but gives an empty list from CENTRAL_REPO... even though the push to CENTRAL_repo appears to have worked	21:18
doener_	GITNoob: and then clone that new repo into your superproject (from the server)	21:18
Auris-	tibbe, have you added dist/ into the repo?	21:18
tibbe	Auris-: no, I don't want it in the repo	21:18
	← Method left	21:18
jast	the shortest way is to put '/dist' in the toplevel .gitignore file	21:18
Auris-	tibbe, well, sometimes people try to ignore things they have already commited. of course it doesn't work.	21:18
	← bdiego left	21:18
	← mithro left	21:19
Arrowmaster	GITNoob: submodules are not added to a repo, only information that they exist	21:19
	← bcarlyon\|laptop left	21:19
tibbe	Auris-: ah OK, no I haven't done that, thanks for the warning though	21:19
GITNoob	doener_: so you're saying we should be cloning from WORKING_repo/module_a... into CENTRAL_repo/module_a?	21:19
	→ mithro joined	21:19
	← tibbe left	21:20
	← rwaite left	21:20
	← icwiener left	21:21
doener_	GITNoob: no. You create a repo for the submodule on the central server. And then you add the submodule in your local repo with "git submodule add ssh://server/module_a module_a", and then you can commit that and push it otu	21:22
	← ph^ left	21:22
doener_	s/otu/out/	21:22
	GITNoob: the above commands were about creating that repo for the submodule	21:23
	→ bdiego joined	21:23
GITNoob	I see	21:24
	so in practice, I should always create the submodules on the central server	21:24
doener_	GITNoob: your "git submodule add module_a" recorded your local path to the module_a repo, so the others try to clone from "module_a/", and of course that doesn't exist yet	21:24
Arrowmaster	the only way to avoid having to login to the server to create it is by using an automated solution such as gitosis	21:24
	→ pjhyett joined	21:24
doener_	the superproject only stores data about what and from where to fetch. Not the whole subproject repos	21:25
GITNoob	gotcha. thanks for the information	21:25
doener_	http://www.kernel.org/pub/software/scm/git/docs/user-manual.html#submodules	21:25
	did you read that?	21:25
	→ kixx joined	21:25
GITNoob	yep. i was reading that this morning	21:26
doener_	hm, ok, it's only a small note that you should not use local URLs, I thought it was a bit more elaborate	21:27
vmiklos	btw, if you want so (storing the whole subproject) then you can use the subtree merge strategy.	21:27
	← Weasel[DK] left	21:27
GITNoob	i think i see what needs to take place now. thanks for all of the help	21:29
	→ alb joined	21:30
	← kumbayo left	21:31
	→ pipegeek joined	21:32
	→ sergiodj joined	21:33
	→ david_koontz_ joined	21:33
pipegeek	how does one get the full hashes produced by git-reflog? It cuts them off after seven characters with an ellipsis	21:33
j_engelh	where?	21:34
	← Sonderblade left	21:34
pipegeek	hmm?	21:34
doener_	git reflog --abbrev=40	21:34
j_engelh	you are seeing seven chars where..?	21:34
pipegeek	thank you	21:34
j_engelh	oh, reflog	21:34
	dwmw2 → dwmw2_gone	21:35
	→ ebel joined	21:35
	← qrush left	21:36
	→ wagle joined	21:37
	← bugfux left	21:37
	→ codeshot joined	21:40
	← Aikawa left	21:40
	← tcoppi left	21:43
	→ tcoppi joined	21:45
	← albertito left	21:46
	← david_koontz left	21:49
	← jstad left	21:49
	← EgS left	21:49
	← Sput left	21:49
	→ EgS joined	21:49
	← ez left	21:50
	→ jstad joined	21:50
	→ Sput joined	21:50
	← LiamH left	21:51
	← Sput left	21:51
	→ Sput joined	21:51
	← vbabiy left	21:52
	← krh left	21:53
	→ bcarlyon\|laptop joined	21:54
	→ tommorris joined	21:54
	cbrake → cbrake_away	22:00
	← andyjgill left	22:01
	← Amorphous left	22:01
	← BabelO left	22:02
	← jackdempsey left	22:02
	← toxx_ left	22:03
	→ Amorphous joined	22:04
	← BrianTheCoder left	22:04
	← GITNoob left	22:07
	→ BabelO joined	22:07
	→ pygi joined	22:07
	→ spearce joined	22:08
	← dsyzling left	22:10
	→ dv_ joined	22:11
dv_	hi	22:11
	so I've pushed a new branch into a repository	22:11
	now I want to pull from it, but I always get only the master one!	22:12
doener_	pull is a shortcut for fetch+merge	22:12
dv_	do I have to do git clone <url> <branch> ?	22:12
doener_	it merges into your current branch whatever it is configured to	22:12
	just "git fetch" and then "git branch -r" to see the remote tracking branches	22:13
	← charon left	22:13
dv_	ah alright	22:14
	← EgS left	22:14
	← Sput left	22:14
	→ EgS joined	22:14
	← EgS left	22:14
	→ EgS joined	22:14
dv_	now to actually get that branch on my local repository	22:14
	→ Sput joined	22:15
dv_	git checkout origin/<branch> is enoughß	22:15
	← zachinglis left	22:15
dv_	?	22:15
doener_	you already have that branch. Otherwise "git branch -r" wouldn't list it	22:15
	→ fixUp joined	22:15
	← _graham_ left	22:15
dv_	but git branch doesnt	22:15
	→ bugfux joined	22:16
doener_	because "git branch" lists your local branches. Those on which you work	22:16
dv_	well	22:16
	this is what I want!	22:16
	kergoth`work restrains his impulse to yell RTFM repeatedly..	22:16
dv_	i want to continue to work on that branch	22:16
doener_	if you want to create a branch to work on, it's "git checkout -b new_name origin/whatever"	22:16
dv_	I RTFMed. it confused me	22:16
	doener_ is in some kind of trance, actually typing most stuff without looking at the screen ;-)	22:16
	← Lash\| left	22:16
	← JimmyB_ left	22:17
dv_	in git I always create new branches?	22:17
	oh wait	22:17
	decentralized	22:17
	← jstad left	22:17
doener_	only if you want to... in this case, you want one, didn't you say that just a moment ago?	22:17
	← fafhrd_ left	22:18
doener_	remote tracking branches represent the state of the remote repository	22:18
dv_	yes, I just was under the impression that I need to give git a command to pull the branch over to my computer	22:18
	as in a copy	22:18
doener_	so when you try to check one out, you get a detached HEAD instead. Because otherwise committing would alter the branch and defeat its purpose	22:18
	you already did that with "git fetch"	22:18
	"git fetch" updates the remote tracking branches	22:19
	→ dragonball_ joined	22:20
	→ gasbag joined	22:20
	→ johnw joined	22:20
dv_	"Because otherwise committing would alter the branch and defeat its purpose"	22:21
	you mean, trying to commit when the branch is remote	22:22
	?	22:22
	→ adante_ joined	22:22
doener_	yes. The remote tracking branch represents the state of the remote	22:22
dv_	well yes, its not local,	22:22
	so a commit would affect a non-local repo	22:22
	← neoeinstein left	22:22
doener_	no. You have the whole branch and all of its history locally available. "git fetch" fetched that for you	22:22
dv_	hmm and what if I do something like git checkout -b somename1 origin/anothername2?	22:24
	i.e. when I change the name while checking out?	22:24
doener_	that creates a new branch somename1 that initially points to the same commit as origin/anothername2	22:24
	(and checks that branch out)	22:24
	branches are just references. They reference some commit	22:25
dv_	all pulls and pushes then go to origin/anothername2?	22:25
doener_	no	22:25
	somename1 is (by default) setup so that "git pull" will fetch and merge the branch "anothername2" from the remote "origin"	22:26
	but push has no per-branch settings	22:26
	the default operation is just to update all branches on the remote for which you have local branches with a matching name	22:26
	← tcoppi left	22:27
	→ bugfux_ joined	22:28
	← bugfux left	22:28
doener_	kergoth`work: you still ok? ;-)	22:30
dv_	so somename1 wont be uploaded	22:30
robinr	spearce: dammit, we broke eclipse 3.3 compatibility	22:30
kergoth`work	i stopped looking at the channel before i exploded	22:30
dv_	unless I specify --all	22:30
spearce	robinr: in egit master?	22:30
robinr	yep	22:30
spearce	dammit.	22:30
doener_	or just somename1, no need to push everything else as well...	22:30
robinr	pushed it /before/ i tried the batch build	22:30
spearce	well, fix up patch it is then.	22:31
raz	is there a quick way to have 'git status'-like output but only for the current directory (and below)? 'git status .' does unfortunately not limit	22:31
robinr	spearce: provider.getJSch().addIdentity(path);	22:32
	← _Vi left	22:32
robinr	getJSch isn't in 3.3	22:32
spearce	argh.	22:32
robinr	shall we ditch 3.3 ?	22:32
spearce	i'm starting to think so. there's UI bugs in 3.3 that marek says they won't even fix.	22:32
robinr	yep	22:32
	← schlort left	22:32
spearce	i mean i'm not surprised they say that, they want to stay focused on the latest stuff, just like everyone else.	22:33
	its just hard for us to continue to support the aging 3.3 when its maintainers aren't really doing it either.	22:33
robinr	they started to drop off 3.5 milestone releases already	22:33
	← juanjoc left	22:34
robinr	and noone is paying us to support 3.3...	22:34
spearce	correct	22:34
	→ eventualbuddha joined	22:34
spearce	so maybe its just time.	22:34
doener_	raz: closest thing would be "git diff --name-only -- ." I think	22:34
eventualbuddha	if I have two commit ids, abc and def, is there a way to tell which one came first, assuming they are in an ancestor-descendant relationship?	22:35
doener_	raz: that's for the "changed but not updated" part. For "changes to be committed" add --cached	22:35
	→ rdd joined	22:35
j_engelh	eventualbuddha: git rev-list ID1 \| grep ID2 && echo "ID2 is a descendant of ID1";	22:35
robinr	that flickering from Mareks patches... is it really bad?	22:36
	← joydivider left	22:36
eventualbuddha	j_engelh: cool, thanks	22:36
doener_	eventualbuddha: do you already know that one is the ancestor or the other?	22:36
spearce	robinr: its so annoying to me on mac os x i can't stand to test his code, let alone use it. it looks like utter crap.	22:36
robinr	i.e. the alternativ commented out code	22:36
	← adante left	22:37
	adante_ → adante	22:37
eventualbuddha	doener_: yes	22:37
	→ schlort joined	22:38
eventualbuddha	doener_: the situation is this: I have multiple machines which have a deployed revision of the codebase. I read the revision from each. they may not match (or even exist). I want to show which revision is deployed on which machines and what changes will be made to bring them up to date	22:38
	← bdiego left	22:38
j_engelh	eventualbuddha: there's a faster one I think even	22:38
doener_	test -z "$(git rev-list -1 a..b)" && echo "b is an ancestor of a"	22:38
j_engelh	git log --pretty=oneline $ID1..$ID2; if it outputs zero lines, $ID1 is not an ancestor of $ID2.	22:39
raz	doener_, well, i actually like the "modified/deleted/renamed/untracked" groupings... i wonder why "git status ." doesn't do what i would have expected?	22:40
	← danten left	22:40
doener_	eventualbuddha: hm, "they may not match (or even exist)" doesn't sound like a guaranteed ancestor relationship	22:40
	raz: because "git status ." shows what would happen if you do "git commit ."	22:41
	→ flaguy joined	22:42
raz	doener_, well, not really. git status always shows "Changed but not updated" also from parent directory. even though that stuff would never commit in a "git commit .".	22:42
	← chris2 left	22:42
eventualbuddha	doener_: well, they're always deployed from the same branch (master, at this point), so if a deployed revision exists, it is guaranteed to be an ancestor (or equal) of the current master	22:43
doener_	raz: well, "changed but not updated" stuff does not go into a commit	22:43
	← LostInTheWorld left	22:43
doener_	raz: only the "changes to be committed" part	22:43
eventualbuddha	j_engelh: cool, thanks again	22:43
raz	doener_, then why does "git status ." show it at all? when i'm asking for the status of "." then i'm most certainly not interested in stuff that's not gonna be committed in some parent directory...	22:44
j_engelh	you can combine that with rev-list even	22:44
	git rev-list id1..id2 \| wc -l; # greater than 0: ancestorship yes	22:44
	to avoid the potential overhead of log	22:44
	→ clamothe joined	22:44
doener_	raz: you're not asking for the status of ".", you're asking for the status that is relevant for "git commit ."	22:45
	raz: and that means "commit everything in ., but nothing else"	22:45
	→ abizern joined	22:45
doener_	raz: so all changes in your current directory and below should show up as staged	22:45
clamothe	I'm using git-svn. If I create a tag, then svn dcommit, will the tag be created in the subversion repository?	22:45
	← CelticSoul left	22:45
doener_	raz: and everything outside your current directory shows up as unstaged, even if you did a "git add" on it before	22:46
	← david_koontz_ left	22:46
doener_	j_engelh: well, and using my rev-list version from a few minutes ago, you also avoid outputting and counting the lines, except for the first one	22:47
raz	doener_, well, i guess i understand what git is trying to achieve but imho that's bad UI design. when i ask for the status of '.' then i clearly want the stage/unstaged list for '.'. not for the whole repository.	22:47
j_engelh	yeah you can use -1 :)	22:47
	→ david_koontz joined	22:48
doener_	eventualbuddha: hm, what do you do after determining which one is the ancestor?	22:48
	eventualbuddha: I have a feeling that you might want the a...b syntax	22:49
eventualbuddha	doener_: I plan on showing a list of commits, similar to how tig does it, but instead of branches in brackets (e.g. [master] [origin/master] Whatever) it'd show servers: [ww1.example.com] Some commit \n Another commit \n [ww2.example.com] The commit the other server is on	22:51
	doener_ goes looking for what tig output looks like	22:51
	← bugfux_ left	22:52
doener_	eventualbuddha: http://jonas.nitro.dk/tig/screenshots/main-view.png	22:52
	eventualbuddha: like that?	22:52
eventualbuddha	doener_: yes, but more stripped down (like no date). imagine that instead of the branch name in pink ([cogito-0.16.3]) it'd be the server that is at that revision	22:53
	or servers, as the case may be	22:54
doener_	eventualbuddha: with the full log, or just a list of the states of the servers?	22:56
eventualbuddha	doener_: the log from the oldest deployed revision to the newest HEAD, with deployments marked in brackets as branches are in that screenshot	22:57
robinr	spearce: tried http://www.jgit.org/update-site ?	22:57
	night	22:57
	→ ajonat joined	22:57
spearce	not recently. i'll give it a shot.	22:57
	→ patmaddox joined	22:57
	→ CelticSoul joined	22:57
eventualbuddha	doener_: this is part of a larger project. I'm not looking to do this all simply as part of a quick n' dirty git command	22:57
doener_	eventualbuddha: there's only one HEAD per repo	22:57
	← WorkingGeier left	22:58
eventualbuddha	doener_: sure. I'm being imprecise. but you get the idea	22:58
doener_	eventualbuddha: I'd probably have refs that represent the various servers, pointing to the respective commits	22:59
	← schlort left	22:59
doener_	eventualbuddha: then you should be able to reuse more existing stuff and can probably avoid a good deal of walking the history over and over again	22:59
eventualbuddha	doener_: you mean you'd actually record in the repo where the servers are?	22:59
	→ emlprime joined	22:59
doener_	eventualbuddha: yeah, maybe something like refs/servers/timmy, refs/servers/jimmy, refs/servers/kenny, ...	23:00
eventualbuddha	interesting idea...	23:01
	doener_: how would I use that in git?	23:01
	→ schlort joined	23:02
doener_	you can just run "git update-ref refs/servers/jimmy $whatever" to create/update the ref, and then use it as servers/jimmy	23:03
	→ LiamH joined	23:03
	→ BrianTheCoder joined	23:03
eventualbuddha	huh. that's neat	23:03
doener_	though at least the --decorate option for "log" doesn't seem to look at such refs	23:03
	← mithro left	23:03
eventualbuddha	doener_: tig seems to	23:03
	but that doesn't help me all that much	23:03
abizern	Okay, so I've done a git reset --hard in a fit of anger	23:03
	I know that the sha of each of my commits still exists in the repo	23:04
	Two questions:	23:04
doener_	abizern: git reflog	23:04
eventualbuddha	doener_: git log --decorate does show them	23:04
doener_	eventualbuddha: in the "commit" line after the hash, in parentheses? Doesn't do that here	23:05
abizern	doener_	23:05
doener_	that's me	23:05
eventualbuddha	doener_: yep. commit f342d72c391253171d2a77ad143daec1cd292565 (refs/servers/ssu2.prod, refs/servers/ssu1.prod, refs/remotes/origin/master, refs/remotes/origi	23:05
	n/HEAD, refs/heads/master)	23:05
abizern	doener_: thanks for answering my question before I even asked it.	23:05
doener_	eventualbuddha: hm, weird, doesn't do that here. Which git version is that?	23:06
eventualbuddha	git version 1.5.3.7 on OS X	23:06
doener_	ah, crap, I forgot the "refs/" part when I created it	23:07
	works here as well now	23:07
	→ HenrikV joined	23:10
	← mtd left	23:11
	← ereslibre left	23:11
HenrikV	My setup has been working fine, but now when I do a 'git push s1', I get	23:12
	error: unable to create temporary sha1 filename ./objects/tmp_obj_qdqH6Z: Permission denied	23:12
	fatal: failed to write object	23:12
	error: unpack failed: unpacker exited with error code	23:12
	← BabelO left	23:12
HenrikV	I don't recall changing any permissions ?!?!	23:12
doener_	abizern: oh well, it was not so hard to guess what you were looking for when you started to describe your situation ;-)	23:12
	← thannoy left	23:13
abizern	doener_: I still feel like an idiot. - I can see the sha I need, but I can't think how to move the tip of my branch to that sha.	23:13
doener_	abizern: either reset again, or merge, or rebase	23:14
	→ BabelO joined	23:14
	→ WorkingGeier joined	23:15
	← robinr left	23:15
	← emlprime left	23:15
	→ jstad joined	23:16
abizern	doener_: thanks again. Rebase worked a treat.	23:17
	← tiglionabbit left	23:20
	→ jmspeex joined	23:21
	← BabelO left	23:22
	← univate_ left	23:23
HenrikV	It seems there are some file locks interfering. I haven't found any running processes that should have any relation to the repository. Is there a way to recover the repos?	23:23
	← tommorris left	23:23
	→ univate_ joined	23:23
	→ ereslibre joined	23:23
	→ BabelO joined	23:24
	→ tcoppi joined	23:25
	← ereslibre left	23:27
HenrikV	When you push to a repository, what ref is it filed under?	23:27
	← xinming left	23:27
doener_	whatever you specified	23:27
	the default is to update matching branches	23:28
	that is, all branches on the remote for which a local branch with the same name exists	23:28
HenrikV	so a 'git push someone@somewhere' pushes changes to master from master, right?	23:30
doener_	yes	23:30
	→ mithro joined	23:30
doener_	unless you have setup some push refspecs in your config I think	23:30
HenrikV	ah	23:30
doener_	then it should probably do whatever your config says instead	23:31
	← ebel left	23:32
	→ tiglionabbit joined	23:36
	← kixx left	23:36
	← codeshot left	23:36
	→ iqson716_ joined	23:37
	jstad → jstad\|away	23:37
	← iqson716_ left	23:37
	← tiglionabbit left	23:38
HenrikV	Is there a document somewhere describing how to use git with user/group. My repos stopped working within anything but root access	23:39
	← cbreak left	23:39
	jstad\|away → jstad	23:40
	→ rdavila joined	23:41
	← BabelO left	23:43
	→ zachinglis joined	23:48
HenrikV	Guess I'm fuxored	23:48
doener_	hm? sounds like you chmod need some chmod	23:49
	s/chmod/just/1	23:49
	← HenrikV left	23:49
doener_	or maybe some patience...	23:49
	← fixUp left	23:50
	alb → albertito	23:51
	← halfline left	23:52
	← parasti left	23:53
	← pete__c left	23:53
	→ Pasquale joined	23:54
	→ tiglionabbit joined	23:54
	jstad → jstad\|away	23:55
	← zachinglis left	23:55
	→ mjc joined	23:55
	← Pasquale left	23:56
	→ zachinglis joined	23:56
	← zachinglis left	23:57
	→ zachinglis joined	23:58
	→ markkalderon joined	23:59

Logs Search ←Prev date Next date→

Channels Documentation