IRCloggy #git 2008-09-18

Provider of IRC logs since 2005.

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

	2008-09-18
raz	i must say i'm less than impressed by git's legendary merging abilities. admittedly this was my first merge, but it was quite a lot more painful than with svn :\	00:00
	i'm especially confused as to why git insisted on giving me a 3-way merge when there were only two differing versions around.	00:00
	do better merge-tools use the "ancestor" information to assist, e.g. will they display two panes?	00:01
	raz is tired, plz forgive me using e.g. instead of i.e. all the time ;)	00:01
	← p4tux left	00:01
Eridius	raz: better tools actually use the ancestor	00:02
	vimdiff doesn't appear to do that	00:02
	← LostInTheWorld left	00:02
raz	well, the best merge tool i know is "meld". but i cant get it to work on this box.. grrr.. pygtk screwups	00:03
	is there a way to have git-pull simply put a ".new" file instead of scattering merge-markers over the existing file?	00:04
	since git-mergetool fails on me i'd prefer to do it by hand next time	00:04
	vimdiff works quite well when only two files are involved :)	00:04
	mugwump just used git-commit -m blah filename	00:05
mugwump	raz: write a script? see git-ls-files -u	00:05
raz	hmmm	00:06
	← RaceCondition left	00:06
raz	mugwump, in what order would i perform my actions? git-fetch, then git-ls-files -u to see conflicts?	00:06
	until now i simply used git-pull all the time, so i don't know much about other ways to do it	00:06
	→ containmentfield joined	00:07
	← mithro left	00:07
mugwump	I mean, when you see an failed merge	00:07
	after git-pull (which is the same as git-fetch ; git-merge xxx)	00:07
raz	but that's too late. git pull already garbles the file	00:07
containmentfield	Hi all, so I would like to take two commits out of the current branch and make a new branch out of that, from the point at which the branch was created. What would be the best way to do that?	00:08
	→ twb joined	00:08
containmentfield	I was thinking about creating a branch from master and just applying those two commits to it. Does that make sense?	00:08
twb	Can I simplify "cd foo && git pull", so I don't need a shell? e.g. "git pull --chdir foo"	00:09
	← svend_ left	00:09
	→ rafaelbandeira3 joined	00:10
	→ cods joined	00:11
Pieter	twb: git --git-dir pull perhaps	00:12
	← zachinglis left	00:13
	← clamothe left	00:13
twb	Hmm; git --git-dir midori pull --error--> fatal: Not a git repository: 'midori'	00:13
	← imakethings left	00:13
Pieter	i think git-dir has to point to the .git	00:14
	and then you probably also need to give the path to the workt tree	00:14
twb	Hmm, that doesn't give an error. I wonder, though, if it will try to modify the wrong working directory.	00:14
	→ jackdempsey joined	00:15
Pieter	try git --git-dir=midori/.git --work-tree=midori pull or so	00:15
twb	Haha	00:15
	$ git --git-dir=midori/.git --work-tree=midori pull	00:15
	fatal: /usr/bin/git-pull cannot be used without a working tree.	00:15
	→ narflekw joined	00:16
Pieter	--work-tree=..	00:16
twb	OK. Now to generate a repo that's out of date again, so pull will try to edit the working tree	00:16
	i.e. to test that it DTRT	00:17
Pieter	git reset --hard HEAD^	00:17
raz	humm. how do i see the revision of a file in my working copy?	00:17
Eridius	raz: what do you mean?	00:17
	do you want the sha1 hash for the blob?	00:17
twb	Pieter: yeah, that's just written shit all over my parent directory	00:18
raz	Eridius, well, i said "git checkout $rev myfile"	00:18
Eridius	raz: the $rev there is the commit	00:18
raz	Eridius, yes, but how do i see that info when i come back later?	00:18
Eridius	you don't. when you checkout a specific file from a commit, git doesn't track that you did that.	00:19
	→ dewogalang joined	00:19
Eridius	well, I mean the checkout will update the working tree and the index. but it won't record what commit it used	00:19
	← dewogalang left	00:20
	→ joshbuddy joined	00:20
raz	hmmm	00:20
	feels strange. so i have a file from a past revision in my wc but git-diff shows nothing.	00:20
Eridius	why should it show anything? your working tree matches your index	00:20
	→ dewogalang joined	00:21
Eridius	try git diff --cached	00:21
raz	hummm	00:21
	← codeshot left	00:21
DrNick	or git status	00:21
raz	what does --cached do?	00:22
Eridius	it diffs the index vs HEAD	00:22
	rather than the working tree vs the index	00:22
	← joshbuddy left	00:23
raz	how do you guys work with that stuff without getting horribly confused all the time? ;)	00:23
Eridius	hah	00:23
Pieter	it's pretty logical once you get into it	00:23
raz	hmm	00:23
	← patmaddox left	00:24
raz	i don't get the advantage of this "index" thing...	00:24
	so far it has only managed to confuse me	00:24
Eridius	it's incredibly useful	00:24
raz	any use-cases?	00:25
mugwump	http://www.linuxmagazine.com/id/2547	00:26
	← containmentfield left	00:26
Eridius	sure. when you don't want to commit everything in your tree	00:26
raz	Eridius, in other VCS you just pick what you want and what you don't want...	00:26
	well, yes, staging files one by one is nice. but so far it has not yet outweighed the troubles that this concept gave me :\	00:26
Eridius	in git, you can add a file to the index, continue to make changes, then commit without picking up the new changes to the file	00:27
mugwump	http://www.scribd.com/doc/2545829/Embrace-The-Git-Index # free online version of that article	00:27
Eridius	why is it giving you any trouble at all?	00:27
raz	so i assume there must be a huge use-case	00:27
Eridius	`git commit -a`	00:27
	→ ben_h joined	00:27
raz	Eridius, yea, that part is easy, i'm doing it all the time. but everything involving hopping over different revisions is just strange.	00:27
	like that "git diff" thing	00:27
	Eridius wishes he saved an article he ran across a while ago explaining the uses of the index to subversion users	00:27
	← clamothe_ left	00:27
raz	well, gonna read that article, maybe it enlightens me	00:28
Eridius	http://tomayko.com/writings/the-thing-about-git	00:28
twb	If I just do "git-archive HEAD", it emits a tarbomb -- extracting stuff to $PWD rather than $PWD/project-name.	00:29
	Ah, --prefix	00:30
	← pjhyett left	00:31
	← igorgue left	00:31
	← Bryson left	00:32
	← ben_h left	00:32
twb	Is there a way to get a (more or less) monotonically increasing natural number, along the lines of svn's "revision number"?	00:33
Eridius	twb: that concept doesn't make sense in the distributed world	00:34
	← eddyp left	00:34
twb	How about the analog of "darcs changes --count" ?	00:34
	→ patmaddox joined	00:34
twb	Listing the number of changesets in the repo.	00:35
Eridius	git rev-list HEAD \| wc -l ?	00:35
	note that this will count merges too	00:35
	git rev-list --no-merges HEAD \| wc -l ?	00:36
	→ archis joined	00:36
Eridius	but the point is any sort of number like that is at best transient	00:36
	because the only way to have a revision number like that in a decentralized world is to designate one particular repo as the "authoritative" one and only give that repo the power to assign numbers	00:36
mugwump	twb: 'git describe' is normally what people use for that	00:37
	← kanru left	00:37
raz	Eridius, i think comment 16 in that article (http://tomayko.com/writings/the-thing-about-git) sums up my opinion	00:37
Eridius	raz: you should have linked to the comment	00:37
	→ softdrink joined	00:38
twb	Eridius: I'm aware of that.	00:38
raz	Eridius, http://tomayko.com/writings/the-thing-about-git#comment-74399 sorry	00:38
Eridius	twb: what do you want this for?	00:38
twb	Eridius: creating snapshot Debian packages for personal use from a canonical upstream git repsitory.	00:38
Eridius	twb: use the timestamp of the latest commit?	00:38
mugwump	raz: you can still work like that if you want, just use 'git commit -a'	00:38
twb	Eridius: that would be a rather long string.	00:39
	Eridius: more to the point, I assume that git can backdate timestamps, since darcs and hg can.	00:39
raz	mugwump, the comment actually explains why the "tangled working copy problem" does not justify the index - it has been solved in other ways already	00:39
	mugwump, thus i'll keep on searching for index use cases :)	00:39
Eridius	twb: sure, but you have to explicitly tell it to	00:39
mugwump	look, it's the way it works, it has many advantages you don't understand yet, get over it	00:40
Eridius	raz: git supports basically every workflow under the sun	00:40
twb	Eridius: meaning that the monotonicity of the timstamp is not guaranteed.	00:40
Eridius	your argument is basically "but there's another way to do it". That doesn't mean the index is useless. It just means, well, that there's another way to do it	00:40
mugwump	twb: that's why you have topological sorting	00:40
raz	mugwump, i'm just trying to grasp these advantages. why is it so hard to find information about them if there are so many? i'm honestly curious, not a VCS zealot here.	00:40
mugwump	it's not hard	00:40
Eridius	raz: you're rejecting every possible use of the index that's suggested	00:40
	that's why it's so hard for you	00:40
	"no, I don't like that one" -- "why won't anybody tell me a single use?!"	00:41
mugwump	there's just a lot of them :)	00:41
raz	Eridius, ehm. which uses have you suggested?	00:41
mugwump	eg, the difference between git reset --soft vs git reset	00:41
	git filter-branch --index-filter	00:41
Eridius	raz: dude, I gave you an article to read about the index	00:41
mugwump	merging	00:41
twb	mugwump: $ git describe HEAD --error--> fatal: cannot describe '0943ac252e2f199c1c0c1d383201f698277c0fc1'	00:41
mugwump	twb: you need tags first (annotated ones, ie made with with git tag -m)	00:42
	any tag, tag the first commit ever if you like	00:42
	→ ben_h joined	00:42
twb	mugwump: there were some tags a ways back	00:42
	← dewogalang left	00:42
mugwump	they're probably not annotated	00:43
twb	git tag lists eight of them	00:43
raz	Eridius, well, your one article was debunked in comment #16. and the other article explains in depth how git works but not why (i.e. for what).	00:43
mugwump	if you went 'git cat-file -t tagname' it would say 'commit'	00:43
twb	mugwump: hmm, what does annotation mean?	00:43
raz	anyways, i won't lose sleep over this, just wondering ;)	00:43
mugwump	heh, "debunked"	00:43
Eridius	raz: no, it wasn't debunked	00:43
	comment #16 was just somebody bitching about the fact that you're not testing exactly what you're committing, and saying it's possible to use another workflow	00:44
	that's not the same as debunking	00:44
raz	Eridius, well, it justifies the existance of the index with the "tangled working copy problem". but this problem has been solved by mercurial and bzr without using the index concept at all. i'd call that debunked, no?	00:44
mugwump	that comment does make a valid point - that you need to make sure that you are aware that making changes piecemeal might lead to broken commits	00:44
	← MarcWeber left	00:44
	→ Bryson joined	00:44
Eridius	you can always reset to the commit and test later, before you push. Or you can use a CI server. Or you can just give up the idea that every single commit is going to have the full test suite run on it	00:44
mugwump	but it does not go the whole way to showing that this is impeded because of the index	00:44
Eridius	raz: you're missing the whole concept of MULTIPLE WORKFLOWS	00:45
twb	Eridius: incidentally, Darcs' commit-time test code will make a separate copy of the working tree without any unrecorded changes, and run the test command on that, rather than the potentially "dirty" working tree.	00:45
Eridius	and you can use the same "solution" in git too, with something like stgit	00:45
mugwump	or just a commit hook	00:45
	mercurial and bzr are simply less flexible, for many reasons including the lack of a staging area	00:46
	you can't script as well with those tools	00:46
raz	Eridius, hmm. so git is kinda the "do-it-yourself" vcs? self-assembly, batteries-not-included? ;) .. nah, i think most people use git in a similar way...	00:46
Eridius	raz: go ahead and ignore the index if you want. Many many people find it to be an extremely useful tool	00:46
	not the only tool, but a tool	00:46
	raz: the original git wasn't even billed as a VCS, it was billed as the tools to build your own VCS	00:46
	→ bts- joined	00:46
Eridius	but yeah, most people tend to use many of the common features	00:47
	like, say, the index	00:47
raz	Eridius, well, if it could just be ignored that would be nice. but it can't. pretty much every operation assumes knowledge of the index and how it works.	00:47
mugwump	have you seen git-gui yet?	00:47
Eridius	what's so hard about ignoring it? `git commit -a`	00:47
	→ saturnflyer joined	00:47
RandalSchwartz	now try a merge. :)	00:47
	← harinath left	00:47
RandalSchwartz	and when it blows up, yoru index is your friend.	00:47
mugwump	isn't it nice that you're not forced to make the commit with one command, that you can build it up, make it look right, commit, perhaps notice you've made a mistake, amend it...	00:47
raz	ah well...	00:48
mugwump	you just don't get that with the others, not to the same degree	00:48
	→ kanru joined	00:48
Eridius	hell, I never even run `git commit -a`, I use `git add -u` first so I can run `git status` and make sure it looks correct	00:48
raz	mugwump, i simply don't see that as an advantage at all. and i don't get your other use-cases either. so i'll just shutup now :)	00:48
	← kef left	00:50
	← achew22 left	00:50
	→ mithro joined	00:52
	mw → mw\|out	00:53
	→ kef joined	00:54
twb	10:46 <Eridius> raz: the original git wasn't even billed as a VCS, it was billed as the tools to build your own VCS	00:54
	→ ajonat joined	00:54
twb	Eridius: so it's another TLA! ;-P	00:54
Eridius	git is a TLA?	00:54
twb	TLA as in GNU Arch	00:54
Eridius	what? "GNU Arch" isn't a TLA. "GNU" is	00:54
twb	TLA = Tom Lord's Arch [implementation]	00:55
	twb is nonplussed by the apparent lack of background training in VCS hackers.	00:55
Eridius	TLA = Three Letter Acronym	00:56
	← tongueroo left	00:56
	→ Jacolyte joined	00:56
	← schacon left	00:59
	→ rmh3093 joined	00:59
narflekw	i want to tag every commit pushed to my remote origin, but adding a line to my post-update hook doesn't seem to do anything after a push from my working repo	00:59
	running the hook by hand on teh server woks	00:59
mugwump	chmod +x ?	00:59
Eridius	tag every commit? why?	01:00
narflekw	er not commit	01:00
	push	01:00
Eridius	ah	01:00
narflekw	sorry	01:00
	added `git tag -a -m 'testing' "autotag_`date +%Y%m%d%I%M%S`"` before git-update-server-info	01:00
Eridius	narflekw: did you try mugwump's suggestion?	01:01
narflekw	Erdius: it's already executable, and running ./hooks/post-update does exactly what i want	01:01
	it just doesn't appear to happen when i push	01:02
	→ Lars_G joined	01:02
narflekw	tho it does say it's updated the server info	01:02
mugwump	you need to specify what to tag	01:02
narflekw	ahhhhh	01:02
mugwump	it will be $1 or $2 or something	01:02
Lars_G	In case my friends fail me over there (points) I'm failing at a git basic use horribly and need some help.. Basically I pulled from a repo. I did some changes in a file. I pull again and git is telling me it can't merge (auto merge I bet). How can I manually merge?	01:03
ben_h	Lars_G: the changes you made aren't committed?	01:03
Eridius	Lars_G: it's probably telling you the file is no uptodate	01:03
	*not	01:03
	that means you have a dirty working tree	01:04
Lars_G	hhmmm let me check I think I commited... locally you mean?	01:04
Eridius	you need to commit first	01:04
Lars_G	not a push	01:04
	← rmh3093 left	01:04
Eridius	right, locally	01:04
	→ pjhyett joined	01:05
Lars_G	doh	01:05
	I've used centralized CSV for so long, that I can't distinguish between commit and push	01:05
	Lars_G slaps himself	01:05
Lars_G	CVS	01:05
	← saturnflyer left	01:06
	← kbingham left	01:06
ben_h	you can still do it without pushing	01:06
	if you want to keep your changes local.	01:07
	→ saturnflyer joined	01:07
Lars_G	ben_h: I have to, no permissions on the remote	01:07
	;)	01:07
ben_h	:)	01:07
Lars_G	it's just bad habits from svn	01:07
ben_h	one way (and i don't know if this is considered the best/easiest) is	01:08
	git checkout -b lol && git pull origin master && git checkout master && git merge lol	01:09
	→ warthog9 joined	01:09
ben_h	or you could rebase lol	01:09
	if appropriate	01:09
	← saturnflyer left	01:10
Lars_G	lol	01:11
	→ joshbuddy joined	01:12
johnw	if I merge lol, will get tell me "I haz no commitz for uz!"	01:13
	→ clamothe joined	01:14
clamothe	I am using git-svn. I created a branch of master, and made changes. Some other people made changes to trunk on the svn server.	01:15
	When I was finished with my branch, I svn-rebased my master branch. Then I merged my master branch into my personal branch.	01:15
	Now I want to bring my changes upstream	01:15
	← irotas left	01:15
clamothe	I just merged master into my branch, so my branch should take complete precident.	01:16
	any suggestions?	01:16
uau	you shouldn't do merges if you intend to put the stuff back to svn	01:16
	rebases only	01:16
clamothe	gotcha	01:16
	any way to do it now then?	01:16
	→ marj joined	01:17
uau	well if you haven't done anything after the merge you can just delete it from the history	01:17
clamothe	I've done a few merges actually	01:17
	have 3 merges over time	01:17
	I could just make a diff of my branch and master	01:17
	and apply that to master	01:18
	right?	01:18
	it wouldn't preserve my commit logs	01:18
	but that's okay, I can compound them	01:18
uau	you cannot push history with merges with svn	01:18
	so somehow you have to construct a branch that's just additional non-merge commits on top of what git-svn has	01:18
	how to best do that depends on the details of what you've done	01:19
clamothe	I think it will be easiest just do `git diff master mybranch ... git apply mypatch` to master	01:19
uau	an easier way to do the same is "git checkout mybranch ." while on master	01:21
	← parasti left	01:22
uau	that won't delete files i think, but otherwise it should set every file to the state they have in the checked-out version	01:22
	→ gretch_ joined	01:23
	→ nn joined	01:24
Lars_G	johnw: http://show.degopro.com/catz.jpg	01:25
	→ _VVSiz_ joined	01:26
	← univate left	01:27
	→ univate joined	01:27
	← ereslibre left	01:29
	gretch_ → gretchen	01:29
	→ tongueroo joined	01:30
	→ ereslibre joined	01:30
	← Lars_G left	01:31
	→ mfwitten joined	01:31
mfwitten	Hello. Where can I read about grafts? Documentation/ doesn't have much from what I grep	01:31
Eridius	git-filter-branch is the only place I've actually seen documentation	01:32
	← tongueroo left	01:32
mfwitten	ok	01:32
	→ JimmyB_ joined	01:33
	← pjhyett left	01:34
mfwitten	What exactly are the technical reasons why a shallow repo has limitations?	01:35
	← spearce left	01:37
	→ aiur joined	01:37
	→ jaredj joined	01:38
	→ jaswinder joined	01:38
	→ bogolisk joined	01:38
	← bogolisk left	01:41
jaredj	noob question. i have gotten the torvalds linux 2.6 tree. i want to pull the patches from the v9fs dev tree. git pull blablabla/v9fs.git right?	01:41
	→ catbot_ joined	01:41
jaredj	oho look at this copious error message. nevermind, i'm sure it'll direct me well	01:42
	← VVSiz_ left	01:43
	← grindz left	01:43
	← david_koontz left	01:44
	← omally__ left	01:44
	← Guest84351 left	01:44
	→ xhanjian joined	01:46
	← johnw left	01:47
	← joshbuddy left	01:49
	← whee left	01:53
	← markkalderon left	01:54
	← xhanjian left	01:57
	→ xhanjian joined	01:57
	← aiur left	01:57
	← kef left	01:58
	← cbreak left	01:58
	→ kef joined	02:00
	→ bobmcw joined	02:00
	← davidfetter left	02:02
	← ereslibre left	02:03
	→ davidfetter joined	02:03
	← xhanjian left	02:04
	← clamothe left	02:07
	→ capisce joined	02:08
	← capisce_ left	02:11
	→ joshbuddy joined	02:12
	← ijcd left	02:13
	← bts- left	02:13
	→ rmh3093 joined	02:14
	← sgrimm left	02:15
	→ ^conner joined	02:15
	← joshbuddy left	02:16
^conner	I'm having trouble creating a checkout/bracnch that merges two remote linux trees	02:16
mfwitten	git svn looks like it's doing NOTHING. Is this normal while it "searches for history" ?	02:17
^conner	i thought it should work something like (after adding the remotes): git checkout -b foo treeA;git checkout -b foo treeB	02:17
	but that pukes on me saying the branch already exists	02:17
MaddHatter	do you want git merge, maybe? (is there a git svn-merge?)	02:19
	← xjrn left	02:20
	→ lima joined	02:20
^conner	MaddHatter, that could be it	02:22
	git merge netdev-2.6 tip	02:23
	→ clamothe joined	02:23
^conner	netdev-2.6 - not something we can merge	02:23
	hmm	02:23
	it doesn't work with branch names from those remotes either	02:24
MaddHatter	I don't use git+svn, so I can't help you much there	02:29
	→ jerbear joined	02:31
	→ ZorbaGama joined	02:36
^conner	MaddHatter, I'm not using SVN at all, I'm pulling linux git trees	02:37
	← pygi left	02:37
	→ pygi joined	02:37
MaddHatter	oh... I merged your comment into mfwitten's I guess	02:37
mfwitten	MaddHatter: Yeah, I noticed ;-)	02:38
	→ k776 joined	02:39
ZorbaGama	I am having a weird issue I cannot find a solution for. I cloned a git repository, then pushed that to a new repository. The new repository didn't get the tags that the old one did (I'm not sure if it should have) but I also can't figure out how to add new tags - I can add them locally, but push/pull'ing to another system doesn't result in them showing up	02:42
	neurodamage → neurodamage\|away	02:42
ZorbaGama	is there some difference with how tags have to be pushed? or do I have something wrong on my config? I am unable to find a solution	02:43
	← marj left	02:43
offby1	ZorbaGama: I don't have any specifics	02:45
	just a couple of vague hunches	02:45
	1) tags come in two flavors: the kind you can push, and the kind you can't :) (The latter are sometimes called "lightweight" tags). Make sure you've got the former kind.	02:46
	2) _some_ command(s) don't move tags by default, although I suspect the command(s) I'm thinking of are "pull". Check the "git-push" man page to see if there's some option that means "please push the tags, too"	02:46
	→ schacon joined	02:47
narflekw	git push --tags	02:49
ZorbaGama	awesome, I shall check out both of those. I think the old ones weren't lightweight, but the one I made may have been lightweight. testing shall occur	02:51
	thanks offby1/narflekw	02:51
	← twb left	02:52
	← cilly left	02:52
offby1	"There will be tests"	02:53
	← kukks left	02:54
ZorbaGama	git push --tags did the trick	02:54
narflekw	=]	02:55
	hooks don't appear to be running after a push on my remote server	02:56
	← jackdempsey left	02:56
narflekw	has anyone had problems iwth hoooks not running on a server accessed via http?	02:56
	specifically i'm trying to run a post-update hook	02:57
	← merlin83 left	02:57
	← mfwitten left	02:59
	← Yuuhi left	03:03
	→ marj joined	03:03
	→ schacon_ joined	03:05
	← schacon left	03:05
	← statim left	03:06
	← MaddHatter left	03:06
	→ statim joined	03:06
	→ MaddHatter joined	03:08
	→ schacon joined	03:09
	← crab left	03:10
	→ LostInTheWorld joined	03:13
	← jerbear left	03:13
	→ jeukku joined	03:14
	← schacon left	03:16
	→ schacon__ joined	03:18
	→ schacon joined	03:20
	→ zachinglis joined	03:20
	← zachinglis left	03:20
	← lima left	03:23
	← schacon__ left	03:24
	← mw\|out left	03:26
	← schacon_ left	03:26
	← bobmcw left	03:26
	← Starnestommy left	03:27
	→ offby1` joined	03:31
	→ joshbuddy joined	03:33
	← rafaelbandeira3 left	03:35
	← loincloth left	03:36
	← offby1 left	03:39
	offby1` → offby1	03:39
	← schacon left	03:39
	→ Flathead joined	03:40
	→ schacon joined	03:43
	← jeukku left	03:50
	← kanru left	03:50
	← schacon left	03:51
	neurodamage\|away → neurodamage	03:55
	→ kanru joined	03:56
	→ jackdempsey joined	03:58
	← joshbuddy left	03:59
	← madewokherd left	04:00
	→ imyousuf joined	04:00
	← narflekw left	04:01
	← jackdempsey left	04:03
	→ zjason_ joined	04:04
	← statim left	04:05
	→ jackdempsey joined	04:05
	← kef left	04:11
	→ kef joined	04:12
	← russells left	04:16
	→ pumpkin joined	04:18
	← tchan left	04:18
	← zjason left	04:19
	→ tongueroo joined	04:19
	→ rdavila joined	04:22
	→ priidu joined	04:25
pumpkin	a while back someone in here was very helpful in guiding me how to do something rather complicated with git, and now I've forgotten how it all worked	04:27
	I can't remember his name :/	04:27
	I'm trying to work out how I set it up because I need something like it again	04:27
	← tjafk1 left	04:28
	→ tjafk2 joined	04:28
offby1	bet it was ilari, Mikachu, or Tv	04:28
pumpkin	hmm nope	04:28
	anyway, I'll describe what's going on and see if anyone has any thoughts	04:28
	← jackdempsey left	04:28
offby1	between them they've probably accounted for 85% of my enlightenment in this channel	04:28
pumpkin	:)	04:29
	→ UVSoft joined	04:29
UVSoft	hi	04:29
pumpkin	well, I have a project that relies on a modified version of another open source project that's hosted on github	04:29
UVSoft	when i save a file in dos format, git shows a lot of ^M at the end of the lines, how to deal with it?	04:29
pumpkin	so I want to maintain my changes to this other open source project as part of mine	04:29
	and be able to update from their repo	04:30
	and keep it all in my repo, as the changes make no sense on their own	04:30
	so what I remember doing is setting up a self-referential git submodule	04:30
	after checking their project out into one of my branches	04:30
offby1	UVSoft: "don't do that then"? DOS format is, by definition, "^M at the end of the lines".	04:31
pumpkin	or something like that	04:31
offby1	sounds complex	04:31
	→ mario__ joined	04:31
offby1	you can't just have your own branch, with your changes, into which you occasionally merge from upstream?	04:31
	I do that with Emacs -- I have a few tiny tweaks of my own, but otherwise track upstream	04:32
	← pygi left	04:32
pumpkin	well, I need to track it in sync with my "enclosing" project	04:32
UVSoft	offby1, ok, suppose i have a file in dos format, than i make some changes in it and invoke git-diff, and see ^Mssss.... that's not what i want	04:32
pumpkin	which is where the submodule came in	04:32
offby1	hmm	04:32
pumpkin	offby1: i.e., it would be nice to have a consistent history for the changes to the subproject, too	04:33
offby1	UVSoft: seriously, I think there might be a way to avoid that, since it's got to be a commonm problem ... I just don't know what it is.	04:33
	pumpkin: doesn't my scheme do that?	04:33
pumpkin	hmm, I'm not sure	04:33
MaddHatter	UVSoft: you probably want to change config.autocrlf or whatever	04:33
pumpkin	I'm not that good with git yet	04:33
	I haven't ventured outside of my comfort zone much	04:33
offby1	submodules sound like the wrong idea, since their whole purpose in life (if I understand correctly) is to _not_ track the upstream, but rather to stay stuck on one commit.	04:33
pumpkin	ah	04:33
UVSoft	MaddHatter, i have to store my files in dos format	04:34
MaddHatter	then you don't want git converting them to *nix format, so disable the autocrlf conversion	04:34
pumpkin	offby1: I'm sorry, do you think you could elaborate a little on how I would set your system up?	04:34
offby1	hold on.	04:36
pumpkin	thanks	04:36
offby1	so just "git clone git://whatever-it-is".	04:36
	then "git checkout -b my-hacks"	04:36
	edit to your heart's content	04:36
UVSoft	MaddHatter, and what is its default value?	04:36
offby1	then when you feel like syncing with upstream, just do "git pull upstream my-hacks"	04:37
pumpkin	offby1: within my other git repo?	04:37
offby1	... other git repository?	04:37
	you have more than one?	04:37
MaddHatter	UVSoft: not sure, but I think it converts by default	04:37
pumpkin	well the point is	04:37
	offby1: I'll make this more concrete, sorry	04:37
	I'm making a ruby interpreter plugin for a program	04:37
offby1	just don't drop it on my foot	04:37
	sure sure	04:37
pumpkin	and I need to make some changes to ruby for it, but I also have code of my own	04:37
offby1	sure sure	04:37
pumpkin	so my own code is already in a git repo	04:38
offby1	hmm	04:38
pumpkin	I'd like to check ruby out into that git repo	04:38
	make my changes	04:38
	and still be able to update from ruby (I'm using the matzruby from github)	04:38
offby1	you want a repository that contains both your hacked-up ruby, and the plugin?	04:38
pumpkin	yeah	04:38
offby1	that sounds dubious.	04:38
pumpkin	because they don't make sense individually	04:38
offby1	I think if it were me I _would_ keep them separate	04:38
pumpkin	like, my hacked-up ruby has no point on its own	04:38
offby1	yeah	04:38
pumpkin	I'm going to make tie-ins with the plugin	04:38
offby1	well, you could add your plugin as a subdirectory of the ruby distro or something.	04:39
	or vice-versa, for that matter.	04:39
	But I think life will be easier with just one repository, not two.	04:39
	→ catbot__ joined	04:39
pumpkin	well, adding ruby as a subdir of mine	04:39
	means I can't update without pain	04:39
	unless I clone it into my other repo	04:40
	which then means I'm tracking a .git folder with git	04:40
	which feels dirty, but maybe it's not uncommon?	04:40
	← softdrink left	04:40
	← catbot_ left	04:40
offby1	doesn't sound like the right thing.	04:40
pumpkin	adding my thing as a subdir from ruby just feels wrong	04:40
	← EmilMedve left	04:40
offby1	I suggest you try it anyway :)	04:40
pumpkin	hmm	04:40
UVSoft	but git show, git diff etc still show ^M at the end of the lines....	04:40
MaddHatter	you said you wanted the ^Ms	04:41
	← k776 left	04:41
pumpkin	offby1: the solution this other person whose name I can't remember suggested worked nicely in exactly this kind of situation	04:41
offby1	UVSoft: if you're using Cygwin, try msys; if you're using msys, try Cygwin :-\|	04:41
pumpkin	just trying to remember how I set it up	04:41
UVSoft	MaddHatter, i want them in the file, but not in the output of git commands	04:41
	→ jeukku joined	04:42
UVSoft	it's rather annoying you know....	04:42
MaddHatter	well, then you _do_ want crlf conversion	04:43
	or no... you want the opposite of crlf conversion	04:43
	the only thing that comes to mind is to disable crlf conversion and use a pre-commit hook to add back the ^Ms	04:45
	(i.e. call unix2dos before the commit)	04:45
	→ SRabbelier joined	04:45
offby1	I have this vague notion that it's actually a terminal thing -- maybe he could tell his terminal to simply not display carriage-returns as ^M, but I don't know how to do that	04:46
MaddHatter	it's possible the files actually have \r\r\n in them, too	04:46
offby1	ick.	04:46
UVSoft	it seems that core.autocrlf true works, my files are checked in with only cr, and checked out with crlf, and git diff, etc doesnt show any ^M	04:47
MaddHatter	hey, if it were a Mac, it'd be \n\r	04:47
	offby1 's head spins	04:47
offby1	WHY CAN'T WE ALL JUST GET ALONG?!	04:47
MaddHatter	UVSoft: yes, but you said you had to store them with ^M	04:47
	pumpkin → pumpkin\|out	04:47
UVSoft	MaddHatter, sorry it was my fault, i ment store in my working copy of the repo	04:48
MaddHatter	ah, well then your problem is solved :-)	04:48
	→ jackdempsey_ joined	04:48
UVSoft	thanks!	04:48
pumpkin\|out	MaddHatter: no it wouldn't be :P	04:49
	not for the last decade, anyway	04:49
MaddHatter	well, I meant Mac as in Mac, not Mac as in OSX	04:51
	it's like Sun and version numbering	04:52
	you'd think a major company like that could, between them all, figure how to count	04:52
pumpkin\|out	:)	04:53
	→ sgrimm joined	04:54
	← UVSoft left	04:56
	← eikonos left	04:57
	← niki left	04:57
	→ xinming_ joined	04:58
	mario__ → pygi	04:59
	← catbot__ left	05:00
	→ catbot__ joined	05:00
	→ statim joined	05:01
	→ mbuf joined	05:03
	← xinming left	05:11
	← bobesponja left	05:13
	← tongueroo left	05:20
	→ langenberg joined	05:21
	← rdavila left	05:25
	← foca left	05:31
	← langenberg left	05:32
	→ loincloth joined	05:32
	← selckin left	05:34
	→ gittish joined	05:35
	→ theCarpenter joined	05:36
	← LostInTheWorld left	05:36
theCarpenter	i have a desktop and a laptop; i want to be able to code on both of them	05:36
offby1	aw, that's unrealistic	05:37
	don't be greedy	05:37
	offby1 whistles innocently	05:38
theCarpenter	how would i go about some sort of push-based setup with git? f.ex i want to be able to write a script that constantly pushes changes from one to the other every five minutes or so; how could git do that?	05:38
offby1	you could have git do that, sure	05:38
	"git push" would work	05:38
	you'd create a bare repository on one box (presumably the desktop)	05:38
	→ tvachon joined	05:38
offby1	then have the script run on the laptop while you're working.	05:38
	probably better to "push" deliberately though ... never heard of anyone doing it automatically like you're suggesting	05:39
	← loincloth left	05:39
	← tvachon left	05:39
theCarpenter	well i was kinda thinking that the push could be linked to git commit? is there an option for that, or a param?	05:40
offby1	nothing built in, no	05:40
theCarpenter	hm okay.. should be easy enough to atuomate	05:40
offby1	you might be able to rig up a hook on the laptop that does that though.	05:40
	I've never used git's hooks, so I don't know for sure	05:40
	← clamothe left	05:41
offby1	if all you want is automatic backup, git might be overkill.	05:41
	unison or rsync or something might make more sense	05:41
theCarpenter	well i mean i might want to work on one machine at one point and another machine at another	05:41
	→ clamothe joined	05:41
offby1	yeah. What I do is: I have a central "bare" repository, and make clones of it on all those other machines.	05:41
	→ loincloth joined	05:42
offby1	(And then I forget to "push", and forget which machines have which work)	05:42
	so I basically run git as a centralized system.	05:42
	offby1 glances around nervously.	05:42
offby1	don't tell anyone.	05:42
	→ tchan joined	05:42
offby1	the boxes are in a star topology: they all communicate via the central bare repository.	05:42
	← qrush left	05:44
	→ neunon joined	05:45
theCarpenter	so what about creating the repos themselves	05:47
	→ dewogalang joined	05:54
	← gittish left	05:55
	nn → _nn	05:55
	→ mbuf` joined	05:57
	← mbuf left	05:57
	→ namenlos joined	06:01
	← priidu left	06:05
	→ FunkeeMonk joined	06:06
	← kyofel left	06:06
	→ russells joined	06:07
	→ harinath joined	06:12
MaddHatter	theCarpenter: just git init, git commit, and git clone	06:13
	_nn → nn	06:13
MaddHatter	and you can use a commit hook to push up to the remote, assuming you're always connected to the remote when you commit	06:13
theCarpenter	but is it possible to create a whole new repo on a diff machine like that	06:14
MaddHatter	what do you mean by "whole new repo"?	06:14
	git init creates a new empty repository; git clone copies an existing repository to a new location/machine	06:14
pumpkin\|out	I think I have my epic git fu working	06:14
theCarpenter	i can push whole repos?	06:14
pumpkin\|out	that's what you push	06:14
	or branches	06:14
theCarpenter	even if it doesn't exist before?	06:15
Sput	offby1: I'd say that's common for most projects, having a central "official, public" repository every developer pushes to	06:15
MaddHatter	yes, you can push branches that don't exist on the remote repository	06:15
	you can use --all to push everything	06:15
theCarpenter	i mean suppose im with my laptop in the middle of nowhere and i want to replicate it to my desktop as a kind of backup and os i can work on my desktop later	06:15
	okay	06:15
pumpkin\|out	you'd typically init a bare repo on the other side and push to that	06:15
theCarpenter	okay	06:15
pumpkin\|out	or just use github :P	06:15
MaddHatter	you need to put the bare repository somewhere that's accessible to every place where you want to actually do work	06:17
	Call that place A.	06:17
	→ deavid joined	06:18
MaddHatter	Then you can push from B, C, D, etc. to A as you do work at each of them, and pull from A to B, C, D, etc. as you want to update to the latest changes you've made elsewhere	06:18
	and A and B can both be separate directories on your desktop computer if you like	06:18
	→ theCarpenter4 joined	06:22
	← JimmyB_ left	06:26
	→ loiseau joined	06:27
	← helge left	06:29
	→ helge joined	06:29
pumpkin\|out	hmm	06:33
	how can I make sure a remote in my .git/config stays around when I push it?	06:33
	← dewogalang left	06:34
thiago_home	git push doesn't modify .git/config	06:34
	→ bentob0x joined	06:35
	→ reel_ joined	06:35
pumpkin\|out	I mean	06:36
	oh I see	06:37
	so there's no way to have a remote that I have show up in the remote repo?	06:37
Pieter	you can push remote refs to there	06:37
	se git push --mirror	06:38
	*see	06:38
pumpkin\|out	thanks, I'll check it out	06:38
	← theCarpenter left	06:41
	→ rtomayko_ joined	06:41
	← johan-s left	06:42
	← imyousuf left	06:43
	← bronson left	06:48
pumpkin\|out	hmm, all I really want is for there to be an extra remote entry the .git/config in any cloned copy of my github repo	06:51
	is that not possible?	06:51
Pieter	no	06:51
pumpkin\|out	hmm, fair enough I guess	06:51
	not a huge deal	06:51
	→ imyousuf joined	06:53
	← schlort left	06:54
	→ schlort joined	06:54
	← ph^ left	06:56
	pumpkin\|out → pumpkin\|closet	06:57
	→ koke joined	06:59
	→ Weaselweb joined	06:59
	→ drizzd joined	07:01
	→ aiur joined	07:06
	→ bronson joined	07:07
	→ priidu joined	07:10
	→ johan-s joined	07:11
	→ BabelO joined	07:13
	→ rraasch joined	07:13
	← CodeOfficer left	07:14
	← bentob0x left	07:16
	← archis left	07:16
	→ ph^ joined	07:18
	→ vintrepid joined	07:18
	← kef left	07:21
	→ charon joined	07:21
	→ KelebekTurkce413 joined	07:25
	← neunon left	07:25
	→ tritz joined	07:26
	→ gbucher joined	07:29
	pumpkin\|closet → pumpkin	07:33
	← KelebekTurkce413 left	07:34
	← ben_h left	07:36
	← Flathead left	07:37
	→ Flathead joined	07:37
	← tarbo left	07:40
	← schlort left	07:46
	→ schlort joined	07:46
	→ nud_ joined	07:48
	← ajonat left	07:48
	→ kef joined	07:49
	→ pgokeeffe_ joined	07:51
	→ reel__ joined	07:54
	← priidu left	07:59
	→ priidu joined	08:02
catbot__	how to fix such an error, as below	08:06
	$ git add README.txt	08:06
	error: unable to create temporary sha1 filename .git/objects/b1: File exists	08:06
	error: README.txt: failed to insert into database	08:06
	error: unable to index file README.txt	08:06
	fatal: adding files failed	08:06
	← kef left	08:07
	→ drizzd_ joined	08:08
catbot__	anybody can give a hand?	08:08
seanius	hmm... some kind of wierd evil twin issue?	08:09
	← jello left	08:10
	→ christophsturm joined	08:10
	← reel_ left	08:11
	→ drizzd__ joined	08:12
	→ markkalderon joined	08:13
	→ Zarin joined	08:13
	→ capisce_ joined	08:13
	← gbucher left	08:16
	← capisce left	08:17
	→ drizzd___ joined	08:21
	← nn left	08:22
	← drizzd left	08:23
	→ nn joined	08:24
	→ reithi joined	08:24
	← aiur left	08:25
	→ aiur joined	08:26
	→ kyofel joined	08:27
	← drizzd_ left	08:27
	← rtomayko_ left	08:29
	→ drizzd joined	08:29
	nn → _nn	08:30
	← markkalderon left	08:31
	→ jmesnil joined	08:31
	← drizzd__ left	08:37
	← theCarpenter4 left	08:40
	← drizzd___ left	08:40
	← priidu left	08:42
	→ shodges joined	08:42
	← shodges left	08:42
	→ priidu joined	08:43
	← marj left	08:44
	→ marj joined	08:45
	← pumpkin left	08:46
	← mithro left	08:47
	← Flathead left	08:48
	→ Fullmoon joined	08:53
	→ archis joined	08:53
	← Fullmoon left	08:53
	← reel__ left	08:53
	→ reel_ joined	08:53
	← reithi left	08:57
	→ radarek joined	08:58
	→ hurikhan\|Work joined	09:01
	← koke left	09:01
	→ ereslibre joined	09:01
charon	catbot__: check permissions on .git/objects	09:02
	→ mcella joined	09:03
	→ not-xjjk joined	09:05
	→ gbucher joined	09:11
	→ Flathead joined	09:13
	→ deavid1 joined	09:14
	← ^conner left	09:14
	→ chris2 joined	09:16
	← reel_ left	09:16
	← pgokeeffe_ left	09:16
	← BabelO left	09:16
	← deavid left	09:16
	← thiago_home left	09:16
	← rphillips left	09:16
	← clairvy left	09:16
	← werdnativ left	09:16
	← Eridius left	09:16
	← uggabooh left	09:16
	← Tv left	09:16
	← Oompa left	09:16
	← jettero left	09:16
	← daniel-w left	09:16
	← com4 left	09:16
	← jeffpc left	09:16
	→ RaceCondition joined	09:17
	→ BabelO joined	09:17
	← gbucher left	09:17
	← xjjk left	09:17
	→ gbucher joined	09:17
	→ thiago joined	09:18
	→ jeffpc joined	09:18
	← Zarin left	09:18
	→ fujin joined	09:20
	→ reel_ joined	09:21
	→ pgokeeffe_ joined	09:21
	→ thiago_home joined	09:21
	→ rphillips joined	09:21
	→ clairvy joined	09:21
	→ werdnativ joined	09:21
	→ Eridius joined	09:21
	→ uggabooh joined	09:21
	→ com4 joined	09:21
	→ Tv joined	09:21
	→ Oompa joined	09:21
	→ jettero joined	09:21
	→ daniel-w joined	09:21
	→ serp joined	09:21
	→ stick joined	09:21
	→ andres joined	09:21
	→ tmm1 joined	09:21
	→ johannes joined	09:21
	→ spool joined	09:21
	← serp left	09:21
	← johannes left	09:21
	→ serp_ joined	09:21
	→ johannes joined	09:21
	← com4 left	09:22
	← rphillips left	09:22
	→ com4 joined	09:22
	→ rphillips joined	09:22
	← tmm1 left	09:23
	→ Sho_ joined	09:23
spool	hi, I feel like this is something I should be able to figure out, but thus far I can't: is there a way to resolve conflicts by always deffering to remote changes over local ones?	09:23
	→ tmm1 joined	09:24
spool	the canonical example is with a binary blob, for which diffs don't really make much sense	09:24
doener	binary blobs are just marked as unmerged IIRC	09:25
	there are no conflict markers added	09:25
	you're then free to replace the file with whatever version suits your needs	09:25
	→ cilly joined	09:25
spool	mmmm... I think my files isn't quite a binary (I used that as an example)	09:25
	it's a sqlite database	09:26
doener	you can use an attribute to explicitly mark it as binary	09:26
spool	oh really? how does that work?	09:26
doener	add a line like "filenameglob merge=binary" to a .gitattributes file	09:27
	→ rdavila joined	09:28
uggabooh	anyway, you can always get the proper version with "git checkout correctcommit file"	09:28
	→ mediogre joined	09:28
doener	and probably also "-diff", so that it's treated as binary for diff as well	09:28
spool	is .gitattributes placed in the root like .gitignore?	09:28
	← diask left	09:29
spool	uggabooh: is correctcommit the hash of the right version?	09:29
	← rdavila left	09:29
doener	works the same as .gitignore, yes. (You can place .git{ignore,attributes} files anywhere, not just the root of the repo)	09:29
spool	mmmm useful, thanks very much	09:29
doener	spool: the hash of the commit that contains the right version (or a ref that references that commit)	09:29
	spool: not the hash of the correct blob object itself	09:30
spool	doener: cool ok right	09:30
	doener: this has happened to me in a variety of ways, the worst being that I was tracking a remote repo, and AFAIK, I didn't change anythin locally, but then I ended up pulling a massive conflict	09:30
	← \ask-_ left	09:31
spool	doener: so what I wanted was to just accept all the recent changes	09:31
doener	most likely, you did actually change something	09:31
spool	doener: would git checkout hash . have sorted it?	09:31
doener	no	09:31
	at least not in a generally sane way	09:32
spool	doener: :( I ended up just recloning	09:32
	doener: I mean, the weird thing is that if I did change anything, it certainly wasn't in the files that were conflicting	09:32
	← imyousuf left	09:33
doener	you could also define a custom merge driver for those files, that always takes the version from the branch that you're merging into your branch, but that will be the wrong thing quite often when you merge back and forth	09:33
spool	doener: yeah I'd rather not do that. Recloning makes more sense in those cases	09:33
doener	spool: then maybe your upstream rebased (or rewrote history by other means), go and kick their a**	09:34
spool	doener: :) that's more likely	09:34
	doener: how would you deal with something like that? Is recloning the best workaround or is there a cleaner way?	09:35
	doener: without writing some crazy custom thing	09:35
doener	if you have no local commits that you care about, just reset --hard to the remote tracking branch	09:35
spool	doener: ahhhh that's what that's for	09:36
	doener: that command has always terrified me	09:36
doener	otherwise, you need to rebase your local commits onto the remote tracking branch, which can be painful	09:36
	→ Ademan joined	09:36
doener	how so? Even if you mis-use it, the reflog is there to save you	09:37
	(well, except for uncommitted changes)	09:37
spool	doener: or, if they've rebased, then would locally rebasing fix it?	09:37
	doener: It's mostly because I wasn't sure what it did, and I was afraid to screw things up	09:37
	doener: '--hard' has fairly dire connotations, purely on the basis of its meaning	09:38
doener	see above ;-) But you likely need to make use of --onto and look for the right commit to use as "upstream"	09:38
	so something like "git rebase --onto origin/master $first_commit_you_want_to_keep^ master"	09:39
	→ RaceKondition joined	09:39
	← RaceCondition left	09:39
spool	woah ok, I had no idea you could do stuff like that	09:39
	→ felipe joined	09:39
doener	the fun only starts there, it gets interesting once you add -i to the mix ;-)	09:40
spool	does that mean interactive? that's another thing I've been meaning to learn but afraid to try	09:40
doener	yep	09:40
	nothing to be afraid about... the reflog saves you, and you can always just make a copy of your whole repo and play around	09:41
spool	does it work like mergetool? I think I tried it once I got quite confused.	09:41
doener	not at all	09:41
	→ ph^_ joined	09:42
doener	http://blog.madism.org/index.php/2007/09/09/138-git-awsome-ness-git-rebase-interactive	09:42
	→ reithi joined	09:42
spool	is that your blog?	09:42
doener	no	09:42
	doener doesn't have anything worthwhile to annoy the world with in a blog	09:43
Araneidae	I want to split an earlier patch into two parts, but I don't really want to bring the patch to the head first. How should I go about this?	09:43
	Yesterday you suggested reset HEAD^ when the patch is at the top of the tree, but now I've got one that isn't...	09:44
doener	Araneidae: you'll have to... use "git rebase -i $commit_you_want_to_split^"	09:44
spool	doener: your help has been quite worthwhile, and everyone has the option of not reading it	09:44
Araneidae	Yep, I was guessing that was where I would start	09:44
doener	Araneidae: change the first line (should be the commit you want to split) from "pick" to "edit", save, quit	09:44
Araneidae	ok	09:44
doener	Araneidae: that will bring you to a state where HEAD point to that commit	09:44
	then do the "git reset HEAD^; git add; git commit; git add; git commit" dance	09:45
Araneidae	Ah, I see. Then I can reset HEAD^ and carry on as before?	09:45
doener	and then "git rebase --continue"	09:45
Araneidae	Cool	09:45
doener	spool: thanks :-) But I kinda need the "interactiveness" of irc, or at least email. "Static" docs are what I suck at.	09:45
	Araneidae realises he wants to cherry pick part of the patch into another branch...	09:46
spool	doener: well, I guess irc logs give you the best of both, though you could have comments on your blog....	09:46
doener	Araneidae: really "cherry-picking", ie. duplicating the changes, or do you want to "move" the changes from one branch to the other?	09:47
	Araneidae: and do the changes you care about already form a single commit?	09:47
Araneidae	No, there's a change to one file deep in branch A which I want to apply to branch B.	09:47
doener	and the respective commit does more than changing just that file?	09:48
Araneidae	Unfortunately that change is combined with a bunch of other stuff that's not strictly related.	09:48
	So commit C = tiny change I want + bunch of other stuff	09:48
doener	you can do "git cherry-pick -n $C", "git checkout HEAD -- all the other files that were changed" and then "git commit"	09:49
	→ kbingham joined	09:50
doener	spool: we have logs of this channel :-) http://colabti.org/irclogger//irclogger_logs/git	09:50
Araneidae	Interesting suggestion doener. In this case "all the other files that were changed" is going to be tedious!	09:50
spool	doener: yes, the channel topic tells me so ;)	09:51
Araneidae	and it won't hurt to split out the patch anyway	09:51
spool	doener: an unrelated question -> is it possible to have git use dwdiff rather than diff for prose files?	09:52
doener	Araneidae: git checkout HEAD -- $(git diff --name-only HEAD \| grep -v file_you_want_to_keep)	09:52
	← vintrepid left	09:54
doener	spool: see "Defining a custom diff driver" in gitattributes(5) as a starting point	09:54
Mikachu	spool: you already know about git diff --color-words i guess?	09:54
spool	doener: awesome, thanks	09:55
	→ koke joined	09:55
doener	Mikachu: ah, good point :) didn't know what dwdiff is	09:55
spool	Mikachu: no I don't actually	09:55
	Working with a latex manuscript can be tricky	09:56
Tv	--color-words seems to be what most word-differs do, except with color instead of {+foo+}	09:56
Mikachu	dwdiff has a -c option for coloring, i think	09:56
spool	can I get color words to work with mergetool?	09:56
Tv	still loses to meld..	09:57
Mikachu	and i don't think you can specify word separators for --color-words	09:57
spool	Mikachu: yeah, these are reasons why I wanted to use it	09:57
uggabooh	all open source diff tools suck :(	09:58
spool	Mikachu: or this -> http://code.google.com/p/google-diff-match-patch/	09:58
	← aiur left	10:01
spool	Mikachu: do you know if dwdiff is the best program for diffing latex? Any other suggestions?	10:03
Mikachu	no idea	10:03
spool	fair enough. You've all been very very helpful. Thanks so much.	10:05
	one last one :) -> what's the best way to do svn externals these days?	10:06
	← RaceKondition left	10:08
	→ ramontayag joined	10:12
ramontayag	hi all. i'm learning rebase, and i have a conflict as expected. there's this one file that exists in both master and mynewbranch. i don't want to merge what master has into mynewbranch. how do i do this?	10:13
spool	oh well, anyways, thanks very much for your help	10:14
	→ parasti joined	10:14
	→ arafangion joined	10:20
arafangion	Hey guys... Many people here use git-p4? I've made a (tiny) change that makes it use _substantially_ less memory, though it does make it significantly slower.	10:21
up_the_irons	arafangion: heh, i think people want git to run fast :)	10:21
arafangion	up_the_irons, Yes... But people also like their programs to use less than 1GB of ram. ;)	10:21
jammyd	paging is uber slow, so if you've not got much ram, then that patch might well result in it being faster	10:22
up_the_irons	arafangion: oh man, it uses 1G? wow	10:22
	← Flathead left	10:22
uggabooh	arafangion: what's the change?	10:24
arafangion	uggabooh, Shall I put the whole file onto rafb.net?	10:24
	← rraasch left	10:24
arafangion	up_the_irons, For sure. :)	10:24
	torarne <3 rebase -i	10:24
uggabooh	arafangion: well, the diff if it's tiny	10:25
arafangion	uggabooh, Sure, let me download the original again... I made the changes at work, but emailed the result.	10:26
uggabooh	arafangion: well if it's a hassle, don't bother ... I'm just curious :)	10:28
arafangion	Nah, I should figure this thing out. ;) What diff options do you prefer?	10:28
ramontayag	hi all. i'm learning rebase, and i have a conflict as expected. there's this one file that exists in both master and mynewbranch. i don't want to merge that particular in master into mynewbranch. how do i do this?	10:31
arafangion	http://rafb.net/p/d3RpI632.html	10:33
Mikachu	arafangion: "git diff" ;)	10:33
arafangion	Obviously, you can see why that'd make the new version slower, but vastly more efficient. It's still trivial to optimise much further.	10:33
	Mikachu, Ha, I'm a newbie indeed! :) I didn't feel like doing a git checkout, though. :(	10:34
	→ Lash\| joined	10:34
arafangion	Damn, how'd that tab get in there(!) Fixed on _my_ machine, now. :)	10:36
	→ elmex joined	10:36
Tv	ramontayag: git tracks the whole tree, not individual files	10:39
ramontayag	Tv, does this mean that i have to manually fix the conflict? i can't just say "ignore if there's a conflict", basically	10:40
arafangion	ramontayag, I think you could shelf it first.	10:40
ramontayag	shelf?	10:40
Tv	ramontayag: you can do anything, but just making the file look like on one of the branches, when you see the conflict, is the easy way	10:41
	ramontayag: but don't think that that wouldn't "merge" the changes to that file; it will just say that those changes merged in cause no change in the file	10:41
	→ bentob0x joined	10:41
arafangion	uggabooh, Any comments? Or did you seize up when you saw that it was python? :)	10:42
ramontayag	Tv sorry i don't understand	10:43
	→ Beket joined	10:47
	→ langenberg joined	10:49
	→ mbuf`` joined	10:52
	← cilly left	10:54
	← mediogre left	10:55
	→ clamothe_ joined	10:56
	← clamothe left	10:57
	→ warlock_mza joined	10:59
	← catbot__ left	10:59
	← gbucher left	11:00
	← mbuf`` left	11:00
	→ clamothe joined	11:01
	→ mbuf joined	11:01
uggabooh	arafangion: I am amazed that this saves a lot of memory	11:01
	→ arafangion_ joined	11:01
	→ mediogre joined	11:02
GarethAdams	I need to find out changes that have been made to the phrase "Pay recommendation" anywhere in my repository	11:03
	any file, any branch	11:03
	what's the easiest way to do it?	11:03
doener	git log -p -S'Pay recommendation' --all	11:03
	← arafangion left	11:03
GarethAdams	aha, didn't spot -S	11:04
	:)	11:04
	← koke left	11:04
GarethAdams	i'll work most of these options out one day	11:04
arafangion_	ramontayag, Have you ever used a VCS system before?	11:05
	doener, Yikes, that's next to impossible with svn! :)	11:05
ramontayag	arafangion_: yes, svn, but learned only enough	11:05
doener	arafangion_: even if it would work with svn, it would probably take ages ;-)	11:06
Mikachu	but it wouldn't be nearly as cpu intensive for the client :P	11:07
	← mbuf` left	11:08
doener	Mikachu: sure about that? My gut feeling is that the client would request a diff for each revision and do the search itself :-p	11:08
arafangion_	doener, For each and every damn revision, indeed, and svn's even slower than cvs.	11:10
spb	the one advantage that cvs has over svn is that you can grep in the repository for things like that	11:12
	← capisce_ left	11:13
	← gretchen left	11:13
uggabooh	arafangion_: are arrays really that memory expensive in Python?	11:14
	← clamothe_ left	11:15
	→ capisce joined	11:15
xinming_	hmm, googled for 'convert git to svn', But all the results are 'convert svn to git', Anyone here can give the hint for converting git repo to svn please?	11:17
Tv	xinming_: svn can't express git history	11:17
	→ clamothe_ joined	11:18
arafangion_	uggabooh, No, not at all - unless that array is a 400MB array, being converted to a 400MB string, which means that you're using 800MB of RAM, and you've _still_ got the original data!	11:18
	← clamothe_ left	11:18
	→ crab joined	11:19
xinming_	Tv: So, do you mean, If you switched to git, there is no way to go back? >_<	11:19
	→ EmilMedve joined	11:19
arafangion_	xinming_, Why would you /want/ to go back?	11:19
	→ clamothe_ joined	11:19
crab	if i'm trying to sync one bare repository with another, then i need to do something like git reset --hard remotename/master, right?	11:19
Tv	xinming_: i mean svn is kinda dumb and limited by todays standards	11:19
crab	oh, no. reset --hard won't work in a bare repository.	11:19
Tv	crab: maybe you want git clone --mirror	11:20
	→ cilly joined	11:21
crab	Tv: no, this is a temporary thing.	11:21
	← clamothe_ left	11:21
crab	Tv: (i cloned a repository, which was the master, but my repository is meant to replace it. but i have to pick up another batch of changes from it.)	11:21
xinming_	arafangion_: 10 reasons, 1) no good gui support, 10) Not as easy as subversion.	11:21
	Tv: Yes, In fact, I really loved the git clone feature, which is really handy for making backup.	11:22
	and more consistent in some sense	11:22
uggabooh	arafangion_: fair enough. And no string pre-allocation in Python, to speed the thing up?	11:23
crab	Tv: any suggestions? basically i want to fast-forward foo to origin/foo for all my local branches and then git remote rm origin.	11:23
w00t	I agree with the lack of GUI support, but I think (after having used git for a few weeks) that it's not significantly harder than subversion, depending on your workflow	11:23
	→ mugdha1 joined	11:23
w00t	what sort of stuff have you found difficult?	11:24
GarethAdams	xinming_: actually, git-archive is better for creating snapshots	11:24
mugdha1	What should one do when one gets a checksum mismatch error after issuing "git fsck --full"	11:24
spb	once you get your head around it it's no harder for the stuff svn does normally	11:24
	and significantly more capable	11:24
w00t	spb: noodle	11:24
mugdha1	We're dealing with a 13 GiB repository here and don't want to lose data.	11:24
Tv	GarethAdams: snapshot != backup	11:24
mugdha1	The repository has a lot of binary image data.	11:24
xinming_	GarethAdams: I want the history also being cloned	11:25
GarethAdams	fair enough	11:25
xinming_	So, there is no easy way to convert git repo to svn, right?	11:25
Tv	crab: just update the refs, then.. e.g. fetch with the right refspec will do that	11:25
	xinming_: that is in general impossible because of svn limitations	11:25
parasti	mugdha1: might be worth asking in the mailing list	11:26
	← clamothe left	11:26
Tv	mugdha1: do you have a good clone of it somewhere?	11:26
xinming_	Ok, then, I'll remove the history, and importing again in svn.	11:26
	← _nn left	11:26
Tv	mugdha1: you can often restore the corrupted object(s) from another clone	11:26
xinming_	Tv: thanks.	11:26
mugdha1	Tv: Cloning it takes 3 hours. How does one restore corrupted objects from another clone?	11:26
	→ _nn joined	11:26
Tv	mugdha1: you can just plop them in .git/objects	11:27
crab	git fetch git://thingy +:?	11:27
mugdha1	Tv: Alright.	11:27
Tv	crab: that would not download objects the local clone thinks it has	11:27
crab	Tv: that's ok, right?	11:27
Tv	no it's not	11:27
	he has corrupted objects he needs to replace with good copies	11:28
mugdha1	Also, I'd like to know how best to use git for very large repositories. It doesn't seem to compress image data very well.	11:28
crab	Tv: oh, i'm talking about MY problem, not his. sorry. :-)	11:28
Tv	mugdha1: "don't" ;)	11:28
mugdha1	Tv: Heh, as in?	11:28
Tv	crab: oh yeah, right -- i think that would be good, with sufficient quoting	11:28
	crab: you can make a git clone --mirror of a small repo and see how it configures it	11:29
	mugdha1: not designed for that	11:29
	← chris2 left	11:29
doener	mugdha1: http://git.or.cz/gitwiki/GitFaq#head-ac11406480d09e2df98588e800e41b7256602074	11:29
mugdha1	Tv: I see. What would you suggest instead? We'd like history tracked	11:29
crab	+: is an invalid refspec. too bad.	11:29
Tv	crab: see what --mirror does	11:30
	crab: i think you're missing refs/heads/	11:30
	mugdha1: there aren't that many version control mechanisms aimed at multi-gigabyte repos, that's just not what source code usually is	11:30
	→ bcarlyon\|laptop joined	11:30
	← andreaja left	11:30
Tv	mugdha1: maybe you want the original plan 9 venti, or some hierarchic data management scheme from IBM you pay $$$ for?	11:31
crab	it does refs/:refs/	11:31
	thanks.	11:31
	← ciaran left	11:31
doener	mugdha1: I think someone mentioned that Perforce is popular among game developers to handle large amounts of graphics stuff. But that's just what I heard...	11:32
mugdha1	I hear that	11:32
	→ dwave joined	11:32
	→ powr-toc joined	11:34
	← jmesnil left	11:37
powr-toc	I have two branches (a and b) with different histories, and I want to rebase a onto b. They should be able to merge ok as they dont share any paths. Is this possible?	11:38
arafangion_	Perforce is _very_ fast.	11:38
	Especially if you have a nice massive server.	11:38
	Damn irritating to use at times, but it is certainly very fast, and has good merging (for a traditional VCS)	11:38
	→ random8r joined	11:39
charon	powr-toc: look in man git-rebase for --onto examples	11:40
Gitbot	powr-toc: the git-rebase manpage can be found at http://git.or.cz/man/git-rebase	11:40
	→ whee joined	11:40
powr-toc	charon: cheers, I thought that might do it...	11:40
	→ ciaran joined	11:44
	→ Beket- joined	11:44
mugdha1	We're considering making submodules for sections of images according to category and separating them from the codebase. We're already neck deep into a lot of stuff and I don't think anybody would be prepared to move to another system from git.	11:44
doener	mugdha1: IIRC there was also some talk on the mailing list, about storing badly compressible stuff in files instead of packs, avoiding some overhead or so, I don't remember the details	11:45
	← morphir left	11:45
fujin	Is there anyway to make my repo stop deleting empty directories?	11:45
random8r	yes put a text file in them. that's what I do.	11:45
arafangion_	doener, How would you know if they're hard to compress, without compressing them?	11:45
random8r	then add that textfile to the .gitignore file	11:46
doener	arafangion_: I have no clue ;-) Maybe something like a gitattribute, like "*.png -compressible" :-)	11:46
Tv	random8r: or just version control it..	11:46
	← Beket left	11:47
powr-toc	charon: Can you do it if they don't share history?	11:47
Tv	mugdha1: i manage my 5Mpix digicam pics in git, it's not too much of a pain with pics that size	11:47
random8r	tv what do you mean?	11:48
mugdha1	Tv: sure. each one of our images are over 60M in size though.	11:48
random8r	tv git removes empty folders.	11:48
Tv	mugdha1: but i split them to multiple repos based on time, etc	11:48
mugdha1	is even*	11:48
Tv	random8r: hence, you put a file in; but instead of .gitignoring it, just git add it..	11:48
random8r	tv git doesn't add it.	11:48
	tv coz it's empty.	11:48
Tv	random8r: are you sure?-)	11:49
mugdha1	well, you add it after you've placed the text file in the empty directory	11:49
random8r	yeah, then you gitignore the folder	11:49
	DUH	11:49
	that's exactly what I just said.	11:49
Tv	random8r: you keep misunderstanding	11:49
	oh well	11:49
mugdha1	just git add wherever/ after touch wherever/readme.text	11:49
random8r	so explain it	11:49
	bcarlyon\|laptop → BarryCarlyon	11:49
doener	touch directory/.gitignore; git add directory;	11:50
random8r	tv - git ignores empty folders, right...?	11:50
mugdha1	yeah	11:50
Tv	random8r: but it's not empty if there's that text file in it	11:50
	→ mdalessio joined	11:50
random8r	I know.	11:50
	this is what I suggested... adding a file to the directory and then adding the folder.	11:50
Tv	random8r: you wanted to add the file or subdir to .gitignore, which is something totally different	11:51
random8r	NO	11:51
	I said add a text file to it	11:51
	then add the text file to .gitignore	11:51
Tv	oh man	11:51
	Tv gives up	11:51
random8r	<shrug> whatever.	11:51
	there's a folder let's call it hello	11:52
	→ jonkristian joined	11:52
random8r	it is empty	11:52
	so I add a file to it called empty_text_file.txt	11:52
	then I add the folder	11:52
	→ tud joined	11:52
random8r	(ie git add hello)	11:52
	then I edit my .gitignore file to include empty_text_file.txt	11:52
bremner	random8r: too late	11:52
random8r	so it doesn't track changes.	11:52
fujin	naw man.. you want to have the file in there, _not_ ignoredd	11:53
	so that it dont delete the dir	11:53
Tv	random8r: that's not the way git works	11:53
random8r	oh well it works like that for me	11:53
	<shrug> I must have a weird version or something	11:53
fujin	naw you're crazy	11:53
Tv	(14:52:27) random8r: so it doesn't track changes.	11:53
random8r	ah well. :)	11:53
Tv	that part is plain not true	11:53
parasti	random8r: when you added hello, you also added everything in that directory	11:53
random8r	yup.	11:53
	→ mjc joined	11:53
random8r	ah actually you're right I don't gitignore it.	11:54
	that's for when I want to not track certain files.	11:54
	like site-specific config files	11:54
charon	powr-toc: yes. think of rebase --onto C A B as taking all the patches in 'log A..B' and applying them on top of C	11:55
random8r	good stuff	11:55
	→ merlin83 joined	11:55
	→ Catfish_ joined	11:56
tud	Hello, I have a question about the workflow I could use with git.	11:56
	We have a group of developers working at customer site where they can access only http (no ftp/ssh/git/...)	11:56
	and other people working from office or home	11:56
	The customer would like to have the "official" repository at their site but form outside is unreachable	11:56
	How can we do? thanks	11:56
	← mugdha1 left	11:57
	← whee left	11:58
	→ whee joined	11:58
arafangion_	tud, What's the problem?	11:58
	tud, Obviously, you're going to be much more secure if you allow ssh, but what's the problem?	11:58
tud	the problem is to have the work of everybody sincronized	11:59
	→ mugdha joined	11:59
random8r	no that's the solution.	11:59
	← mugdha left	11:59
	← deavid1 left	11:59
crab	can you push sanely to http: repositories now?	11:59
tud	the central repository is at customer site but from outside is not reachable	12:00
Tv	crab: it's probably the most buggy push mechanism, and will likely stay the most buggy one, because it's so rarely used	12:00
crab	nod	12:00
arafangion_	tud, Have the central repository push its changes to an external repository.	12:00
jonkristian	This hook is invoked by git-commit. It takes no parameter <- what does that mean? that it takes no parameter?	12:00
	→ jmesnil joined	12:00
arafangion_	(And also pull).	12:00
Tv	jonkristian: no command line args	12:01
	arafangion_: you mean fetch, as it'd be a bare repo	12:01
jonkristian	Tv: if i want to automaticly pull when a push has been made, where can i put that then?	12:01
tud	so we should have two central repositories, one (the official) to custome rsite and one for people working outside	12:01
Tv	jonkristian: post-update	12:01
jonkristian	Ok	12:01
tud	and somebody at customer site should take care of merge the two repositories, correct?	12:02
Tv	tud: if everyone can't access one place, what choice do you have?	12:02
	tud: the merge can happen at either end	12:02
tud	tv: this is my question :)	12:02
arafangion_	Tv, I should probably shut up, as green as I am. :)	12:02
	← marj left	12:03
	← Catfish left	12:04
jonkristian	hmm, error: unable to create temporary sha1 filename ./objects/d5: File exists	12:05
Tv	jonkristian: check that you have read/write access	12:06
	→ smeevil joined	12:06
smeevil	hi all	12:06
	i was wondering if the following is possible : lets say i have a few commits in the master branch and i would like to branch off the last 3 commits from the master branch to a new branch... so that the master branch head is basically 3 commits back and the new branch has the latest 3 commits	12:08
	→ lorandi joined	12:08
smeevil	is that possible in any way ?	12:08
jammyd	branch, then reset master back 3 commits?	12:09
smeevil	jammyd: then what happens if you add a patch to the master branch ?	12:09
	→ retybok joined	12:09
jonkristian	Tv: now i get, 0cf4dd7..ccea67f master -> master	12:10
	fatal: Not a git repository: '.'	12:10
retybok	how do you find out the base directory of the git project you're in?	12:10
smeevil	jammyd: i never actually resetted to a previous commit , changed something and commited that, so no idea what happens then :D	12:10
jonkristian	also, if i do a git pull on my local computer, it doesn't understand it, i have to do git pull ssh://....path to repo	12:10
	isn't there a way to avoid this?	12:11
Tv	retybok: git rev-parse --show-cdup	12:11
	jonkristian: define a remote	12:11
retybok	Tv: thanks	12:11
jammyd	smeevil: by reseting you're forgetting the intermediate commits, aiui, so a new commit will be on top of [ what was previously ] the 4th commit back from master	12:11
Tv	smeevil: this may or may not help: http://eagain.net/articles/git-for-computer-scientists/	12:12
smeevil	jammyd: i see, thanks :D	12:12
jonkristian	Tv: i have [remote "origin"] url = ssh:// etc. in .git/config	12:12
	ssh:// path to that repo	12:12
smeevil	Tv: ty :) that will help	12:13
Tv	jonkristian: are you on a branch other than master when running that pull?	12:13
jonkristian	git branch shows master	12:13
	i havent really gotten around to play with branches yet	12:13
Tv	jonkristian: does "git pull origin" work?	12:14
jonkristian	ill try	12:14
	→ cbreak joined	12:14
jonkristian	no	12:14
	→ calin joined	12:15
Tv	jonkristian: so how does it fail, then?	12:17
	→ morphir joined	12:17
Tv	jonkristian: sounds like you have a different url in the config compared to typing it on the command line	12:17
jonkristian	hm	12:18
	← reel_ left	12:18
jonkristian	it's the same, it fails with this, http://rafb.net/p/tXf3gD21.html	12:19
	← iulian left	12:19
Tv	jonkristian: your config is missing [branch "master"]	12:20
jonkristian	and this is my config, http://rafb.net/p/L1LAdo76.html	12:20
	hmm	12:21
Tv	jonkristian: compare that to what git does automatically if you clone something	12:21
	→ jasam joined	12:21
smeevil	jammyd: that worked mighty fine :D thanks !	12:22
xinming_	I noticed about that when we do commit, in git-log, the author and email is from .git/config, But using git within a centralized fasion, How to prevent the user "fake" the user and email info please?	12:22
jonkristian	should i add [branch "master"] to that file?	12:22
jammyd	smeevil: good good! :)	12:22
	← mdalessio left	12:23
cehteh	xinming_: you cant do that in git itself, but you gpg sign commit messages, not very well supported tough	12:23
	one cant sign complete commit objects :/	12:24
Tv	xinming_: first, define fake	12:24
	xinming_: and think of me pushing changes made by you..	12:24
	→ diask joined	12:24
jonkristian	Tv: thanks, got it working	12:25
	i was missing the branch master stuff	12:25
xinming_	Tv: eg, I wrote a .git/config with the the same user as another user which is also be able to push to the centralize repository,	12:26
Tv	xinming_: .git/config isn't pushed	12:26
xinming_	Tv: I don't, But in log	12:26
	git log	12:26
	→ joshbuddy joined	12:26
Tv	xinming_: you mean you created a commit with someone else as author/committer?	12:26
Mikachu	xinming_: yes, you can fake commits	12:26
Tv	xinming_: as i said, think of me pushing a commit made by you	12:26
	xinming_: is that a fake or not?	12:27
xinming_	Tv: It's not.	12:27
Tv	xinming_: so how do you know	12:27
	→ giom joined	12:27
Tv	xinming_: you can't use the pusher identity, then..	12:27
	→ bobmcw joined	12:27
	← joshbuddy left	12:28
Tv	signed tags on every commit?	12:28
Mikachu	Tv: well, you could allow changing author but not committer, maybe	12:28
Tv	Mikachu: that doesn't help me push his (unedited) commits	12:28
xinming_	Tv: But If someone want to "fake" as another user, eg, I don't like someone, I changed .git/config to the name of someone, and so does email, and then push.	12:28
Mikachu	Tv: if it's a centralized repo his commits are already in the centralized repo ;)	12:29
cehteh	you can impersonate anyone when doing commits .. git has no means of authentication commits	12:29
Tv	well signed tags..	12:29
Mikachu	you could check with a hook?	12:29
	(which would then disallow pushing someone elses commits)	12:30
Tv	Mikachu: you could prevent me pushing things made by others, yes; but i explicitly asked and he didn't consider that a fake	12:30
xinming_	I mean, It's a bit like subversion, Eg, we can do ssh authentication. and the user is always identified through ssh.	12:30
cehteh	Mikachu: that woukld be ugly for merges	12:30
Tv	xinming_: svn is very different	12:30
Mikachu	cehteh: why? assuming the repo is centralized, the commits by other users are already in there	12:30
cehteh	mhm yes	12:30
	→ deavid joined	12:30
Tv	Mikachu: that's like "forced centralized" or something..	12:31
cehteh	would work	12:31
	Mikachu is speaking only theoretically, he doesn't think it's a good idea	12:31
jogla	How can i inspect changes before i update? I know i can use git fetch, but where would i find the changes then?	12:31
wereHamster	FETCH_HEAD	12:31
Tv	jogla: in remote branches	12:31
xinming_	I'm still thinking on why this can't be easily done within git.	12:31
Mikachu	jogla: if you fetch a remote ,they're in what tv said :)	12:31
	→ irotas joined	12:32
Mikachu	xinming_: for the reasons given above	12:32
jogla	Ahh, thank you	12:32
	← mbuf left	12:32
Tv	xinming_: two choices 1) forced centralization as Mikachu suggested (which is what svn has) 2) gpg signatures for every commit (easy way: signed tags; alternative: a signed chunk of data in commit message, including parents and tree sha)	12:33
gebi	are there any planes to support the author property of svn in git-svn?	12:33
	to set the author to the _real_ author not the authenticated user	12:34
Tv	huh	12:34
	gebi: svn only stores one string of author information, IIRC	12:34
	→ gbucher joined	12:35
gebi	Tv: yea, the problem is that git-svn is not able to set the author correct (it uses the authenticated user, default in svn)	12:35
Tv	gebi: authenticated user of what operation?	12:35
gebi	git svn dcommit	12:36
	e.g if i do a rebase in git, all commits belong to me after git svn dcommit (in the svn repo)	12:36
Tv	gebi: in general, you can't commit as other people in svn	12:36
gebi	you can ;)	12:36
	you just have to allow it in the hooks	12:36
	but yes, not per default	12:37
Tv	anyway, nobody on this channel has historically claimed to work on git-svn, really	12:37
	try the mailing list	12:37
xinming_	Tv: thanks for the sollutions, Could you please explain the reason why this can't be done easily please? I still not understand. need a more detailed example.	12:37
Tv	xinming_: it can't be done easily because nobody has programmed it for you	12:38
gebi	Tv: ok thx :)	12:38
	→ juanjoc joined	12:38
Mikachu	well, normalperson is eric wong	12:39
gebi	ok	12:44
	we do svn:author and svn:date adjustion for a better svn history here and no problems so far	12:44
	though wrong svn:date could be problematic	12:45
	→ ebel joined	12:47
	→ Yuuhi joined	12:50
	→ tommorris joined	12:50
	→ koke joined	12:50
	→ johnw joined	12:51
	← jeukku left	12:52
	← pgokeeffe_ left	12:52
	→ naoshige joined	12:52
arafangion_	gebi, You could ask them to put such info into the commit messages instead?	12:52
naoshige	how was it i forced a git pull?	12:52
	i get "needs update" on almost all files in this public repo	12:53
	i know they need updates thats why i wanna pull but	12:53
arafangion_	gebi, And I think svn allows hooks to allow you to change the commit as it gets made, nasty things happen when you do that, but I think you can change the commit message itself (ie, the log) fine without any issues.	12:53
	→ Guest84351 joined	12:55
wereHamster	naoshige: needs update usually means that you have a dirty working tree. So you first have to commit your changes or git-stash them	12:55
arafangion_	(Reason nasty things can happen is because svn clients expect that the files they commit are the same as what the commit is, if that makes sense, ie, they assume that the client doesn't automagically change things on the fly)	12:55
Tv	naoshige: "public repo"? http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	12:55
naoshige	wereHamster: but i haven't touched the repo since i first cloned it	12:56
	nevrmind ill just clone it again	12:59
	→ mw\|out joined	12:59
naoshige	thanks for the link Tv	12:59
gebi	arafangion_: would be an idea, but no tool works the way it should with this in svn (thats why i mess with propset --revprov svn:author)	12:59
arafangion_	gebi, I'm talking about server-side svn hooks.	13:00
	→ kef joined	13:00
	← naoshige left	13:02
	← tud left	13:02
	→ mithraic joined	13:03
arafangion_	gebi, The best solution is to give each user their own account.	13:03
gebi	arafangion_: you mean something like --add-author-from from git-svn or adjusting the svn:author property automatically on the svn server?	13:03
	arafangion_: thats not the problem	13:03
	if i do a rebase in git and after that do a git svn dcommit the author and date is wrong (which i fix with some crude scripting)	13:04
	→ ciaran_ joined	13:04
	→ kukks joined	13:05
arafangion_	gebi, Sounds like you've got a pretty good handle on things.	13:05
	→ mbuf joined	13:06
	→ etgryphon joined	13:07
gebi	arafangion_: the setup just cries for pain, 2 svn repositories (one internal and one for external devs) synched with git-svn ;), with more and more git devs	13:07
	← kbingham left	13:07
arafangion_	gebi, Why don't you have a human 'merge master'?	13:08
Tv	gebi: most git-svn users commit their own stuff to git-svn, not others; i guess that's the source of your pain	13:08
arafangion_	gebi, Who's daily task is to merge the two svn repos, in addition to his workload.	13:08
	mw\|out → mw	13:08
Tv	gebi: two svn repos sounds a recipe for pain..	13:09
gebi	arafangion_: the merge "master" does not know well about git and is using just svn	13:09
	Tv: might be, yes :)	13:09
	Tv: the problem is that there is a master branch in git too which has to be exported to svn (and those branch has commits from all the git users)	13:10
	unavoidable to commit others work as not all users have commit access to the right (internal) repo	13:11
	← mithraic left	13:11
arafangion_	gebi, Is it not possible to remove one of those svn servers, so that there is truely only one svn server, and allow people to use git as an "optimisation"?	13:11
gebi	arafangion_: no as the internal network has to be "certified"	13:12
	you really don't want every student in the internal researcher network ;)	13:12
Tv	gebi: sounds like death by middle management	13:12
arafangion_	gebi, Perhaps not, but who is going to trust those student repos? ;)	13:13
	_VVSiz_ → VVSiz	13:13
arafangion_	gebi, git lends itself rather well to such hierachial systems.	13:13
gebi	Tv: yea :)	13:13
	arafangion_: i know	13:13
	→ JimmyB_ joined	13:15
	→ mithraic joined	13:15
gebi	arafangion_: but you can't force git on all those users (as it's a quite old project)	13:15
	→ krh joined	13:15
	← ciaran left	13:15
	← EmilMedve left	13:16
	← langenberg left	13:16
arafangion_	gebi, You can force it on the students, though. ;)	13:16
	gebi, The project might be old, but very few of your users are old timers, I'd bet. ;)	13:17
	→ jeukku joined	13:17
	→ langenberg joined	13:17
Tv	hey you could make them use git-cvsserver ;)	13:18
	← johnw left	13:18
	→ joshbuddy joined	13:18
gebi	Tv: lol yea :), we have even a few ms source "safe" users :D	13:21
	arafangion_: the internal users are not all students though ;)	13:21
	← jeukku left	13:25
	→ vbabiy joined	13:26
arafangion_	gebi, Perhaps the main thing that git's missing is to have a simplified GUI that while effectively cripples git, at least lets you work in a "classic" svn-like fashion.	13:27
	→ jeukku joined	13:27
arafangion_	gebi, ie, no (user-visible) local branching, simple merging, that kind of thing.	13:28
gebi	thats why so many userse choose mercurial	13:28
	← reithi left	13:29
	→ Zarin joined	13:29
arafangion_	gebi, Why didn't you?	13:30
	→ andrewtaylor_ joined	13:30
	← jeukku left	13:30
gebi	i don't want another constraint scm	13:30
	→ softdrink joined	13:31
	→ tarbo joined	13:31
andrewtaylor_	Hi, i'm just starting with git, I need to pull a branch from an existing local repository to be a master on a new local repository. Can anyone give me some pointers?	13:32
gebi	arafangion_: every hg user i know of which doesn't use it as a better backup uses git now, so the choise was quite clear ;)	13:32
	arafangion_ is planning on using git as a better backup... once his server gets back online. :)	13:32
	← deavid left	13:33
	← Sho_ left	13:33
gebi	arafangion_: heh	13:33
arafangion_	G'night. :)	13:33
andrewtaylor_	is it git pull /home/bob/myrepo master ?	13:34
gebi	andrewtaylor_: git init; git fetch ../path_from_other_repo master:master	13:34
arafangion_	gebi, Used to use svn for that. :)	13:34
andrewtaylor_	thanks, what's the difference between pull / fetch / clone ?	13:34
arafangion_	gebi, And svn's almost ideally suited for that, if one doesn't need the RCS capabilities...	13:34
doener	clone ~= init + fetch + checkout	13:34
	pull = fetch + merge	13:34
gebi	andrewtaylor_: pull is fetch+merge, fetch is fetch and clone is init+fetch+(checkout)	13:34
	arafangion_: omg, no	13:35
andrewtaylor_	gebi: perfect sense :)	13:35
gebi	fetch transferes objects and refs from another repository to your current one	13:35
arafangion_	gebi, Well, once you DO need RCS capabililties... You need to read between my words. ;)	13:35
	→ mithro joined	13:36
	← mithraic left	13:36
gebi	arafangion_: nope, ways before	13:36
	svn is way too fragil, slow and has no automatic backup	13:37
arafangion_	gebi, If you have two machines, you do have a backup, though not with any history, and speed has improved heaps, though not up to git standards.	13:37
	← softdrink left	13:37
gebi	just got a svn error yesterday (somethin like checksum mismatch, blablabla) just because one user comitted an ntfs alternate data stream	13:38
	→ softdrink joined	13:38
arafangion_	gebi, It's especially fragile if you still use the bdb system.	13:38
gebi	arafangion_: a "backup" without the ability to restore the hole thing is pointless	13:39
arafangion_	gebi, Which brings you back to this whole "RCS" thing I mentioned. ;)	13:39
gebi	;)	13:39
arafangion_	gebi, And you can restore the "whole" thing, just not any past revision prior. ;) And many traditional backup systems are the same.	13:40
	G'night.	13:40
	← cilly left	13:40
gebi	n8	13:40
	← arafangion_ left	13:40
	→ mithraic joined	13:43
	→ gzy joined	13:45
	→ cannonball joined	13:46
	→ jas4711 joined	13:48
	→ grahal joined	13:48
	→ Beket joined	13:53
	→ EmilMedve joined	13:55
	← ph^ left	13:56
	→ jerbear joined	13:57
	← mithraic left	13:58
	→ johnw joined	14:00
	← johnw left	14:00
	← jackdempsey_ left	14:00
	→ johnw joined	14:00
	→ spearce joined	14:01
	← random8r left	14:02
	← namenlos left	14:04
	→ namenlos joined	14:05
	→ deavid joined	14:08
	← calin left	14:10
	← Beket- left	14:11
jonkristian	files created in .git/ are they all 644 and dirs 755? except for the stuff in hooks/	14:12
	← andrewtaylor_ left	14:12
RandalSchwartz	git doesn't do metadata	14:13
	that's for your installer	14:13
Tv	jonkristian: depends	14:13
	davidfetter waves to Tv	14:13
RandalSchwartz	Oh oh. in .git	14:13
	sorry	14:13
	not in "git" :)	14:14
	→ foca joined	14:14
	RandalSchwartz stares at diet coke and yells "work! work!"	14:14
Mikachu	jonkristian: objects/* is 444 i think	14:14
jonkristian	ouch, then i messed up	14:14
Mikachu	well, maybe not the info/packs file and some other things i didn't think of	14:14
	davidfetter queues up "car wash"	14:14
	← not-xjjk left	14:15
jonkristian	im having trouble seeing a solution, people on my group can't push, and using this example, http://www.linuxworld.com/community/?q=node/3057 , I can't think of a way to maintain permissions on a webfolder.	14:16
	← bobmcw left	14:16
	→ olauzon joined	14:16
Tv	jonkristian: "shared"..	14:17
jonkristian	shared	14:17
Tv	jonkristian: or gitosis	14:17
jonkristian	oh, shared	14:17
	hm	14:17
	→ zachinglis joined	14:18
	← smeevil left	14:18
	← retybok left	14:18
	raz → razZz	14:18
	razZz → raz	14:18
jonkristian	so i do git --shared=groupname inside the repository and chown -R root.groupname or something?	14:19
	← gzy left	14:25
	← tommorris left	14:27
	← blowery left	14:28
	→ blowery_ joined	14:28
	→ aroben joined	14:28
	→ blowery joined	14:28
	← namenlos left	14:28
	→ namenlos joined	14:29
	→ ijcd joined	14:29
	→ alb joined	14:29
	← nbrosnahan left	14:29
gebi	Tv: i teached gitosis to initialise all repositories from a configurable custom template directory: http://git.einsteinmg.dyndns.org/?p=own/gitosis.git;a=log;h=refs/heads/t/templates	14:36
mjc	yarr	14:36
	→ schmir joined	14:37
	mjc → the_good_pirate_	14:37
the_good_pirate_	Bleh	14:37
	the_good_pirate_ → pirate_mjc_yarr	14:37
schmir	hi all. how is the file called where I can artificially modify the revision graph?	14:37
pirate_mjc_yarr	wrong room :p	14:37
	My bad	14:37
krh	schmir: are you thinking of git rebase -i perhaps?	14:38
schmir	krh: no	14:38
doener	schmir: grafts	14:38
xinming_	Is there any other way to update a bare repository except using git push.	14:38
schmir	doener: yes, thanks	14:38
krh	xinming_: something like GIT_DIR=repo.git git fetch	14:39
xinming_	krh: Is it the same as we do on the client using 'git push' ?	14:40
krh	xinming_: no, it's how you would fetch remote branches into a bare git repo	14:41
	→ tvachon joined	14:42
	← tvachon left	14:42
	→ iulian joined	14:43
	← ijcd left	14:44
	→ earcar joined	14:44
	← mbuf left	14:45
	← arun left	14:46
xinming_	krh: How could I also fetch the same effect as doing 'git push' using git fetch please?	14:47
krh	xinming_: you should be able to say GIT_DIR=local-repo.git git fetch git://remote.org/repo.git master:master	14:49
	to fetch the master branch from the remote repo into you local bare repot	14:49
xinming_	krh: thanks really	14:51
krh	xinming_: no problem	14:51
	alb → albertito	14:51
	→ deskin joined	14:55
	→ stouset joined	14:57
	← earcar left	14:58
	← johan-s left	14:59
	← Beket left	14:59
	→ bdiego joined	15:02
	→ camwest joined	15:03
	← camwest left	15:04
	→ bobmcw joined	15:05
	← priidu left	15:06
	→ jackdempsey joined	15:06
	→ jpeeler joined	15:07
	→ ramontayag_ joined	15:07
	← schmir left	15:07
	→ sveajobb joined	15:08
	← jackdempsey left	15:08
sveajobb	is there a way to store a view in gitk ? I'm interested in looking at all available branches	15:08
	→ jackdempsey joined	15:08
	← uggabooh left	15:08
charon	you could make an alias that calls 'gitk --branches'	15:10
	← ramontayag left	15:11
	ramontayag_ → ramontayag	15:11
	→ ijcd joined	15:12
	→ selckin joined	15:13
	← mediogre left	15:13
sveajobb	charon, aah, excellent ! thanks :)	15:13
jogla	When i update a remote branch, i can see with what changes have been made in that branch "since i last updated it"?	15:15
	s/i can see/how can i see/	15:15
deskin	jogla: git-fetch tells you the commit ids from..to for all the remote branches; e.g. if it prints '0123456..9abcdef foo -> bar', just do 'git-log 0123456..9abcdef'	15:18
	← flaguy left	15:18
deskin	or git-diff, git-shortlog, depending on what you want to know	15:18
	← ramontayag left	15:19
	→ advany joined	15:21
	← bobmcw left	15:22
jogla	deskin: Thank you	15:25
	← Hobbes left	15:27
	→ Hobbes joined	15:27
	→ svend joined	15:27
	← ijcd left	15:29
	→ ijcd joined	15:30
	→ hurikhan77 joined	15:30
	← koke left	15:31
	→ kumbayo joined	15:32
	← kanru left	15:33
	→ kowey joined	15:38
	→ LostInTheWorld joined	15:38
	→ jaalto joined	15:39
jaalto	I did commit. Now I want to move it to separate branch, and remove it form current. What's the procedure?	15:39
	I could take diff HEAD~1, save it, branch, patch. Then go back and reset --hard to HEAD~1. But is there anything better?	15:40
vmiklos	git branch foo HEAD; git reset --hard HEAD~1	15:41
	← neurodamage left	15:41
	← hurikhan\|Work left	15:42
jaalto	vmiklos: Uhm, of course. How simple.	15:42
vmiklos	actually you can use HEAD^ instead of HEAD~1 as well :)	15:42
jaalto	vmiklos: thanks	15:42
	→ LiamH joined	15:42
jaalto	vmiklos: It's stylistic issue, I hate ^, whereas HEAD~N is logical, when N can be increased as will	15:43
	← ijcd left	15:43
jaalto	at will	15:43
vmiklos	k	15:43
	→ ijcd joined	15:43
jonkristian	If i set rules in my remote .zshrc regarding umask, will those rules be used when i do a git push to the remote location?	15:46
	→ tsb joined	15:47
vuf	jaalto: maybe HEAD~ for HEAD~1 is not so bad?	15:47
tsb	I have a rep in /etc. I want to add X11/xorg.conf, but nothing else in X11, how would I do that? git add X11/xorg.conf seems to work, but when I do git commit -a there's nothing to commit	15:47
	← deavid left	15:48
	→ TML joined	15:49
	→ mjansen joined	15:50
TML	Shouldn't "git prune" clean up all unreachable objects?	15:50
jaalto	tsb: git status?	15:50
	tsb: ... after add. Maybe file is ignored by .gitignore	15:51
mjansen	I just fetched with "git svn fetch" and now try to rebase with "git rebase git-svn". That fails with "could not detach HEAD"	15:51
	→ danten joined	15:51
mjansen	but it won't tell me why.	15:51
vuf	TML: seems like it	15:51
mjansen	git -v rebase ... doesn't help either	15:51
TML	mjansen: "git svn rebase"?	15:52
mjansen	sorry ... forget that svn part.	15:52
	same result	15:53
doener	mjansen: probably an untracked file would get overwritten or something like that	15:53
	mjansen: IIRC git 1.6.0 and newer have a more verbose output	15:54
mjansen	when trying to checkout git-svn i get a warning for such a untracked file	15:54
doener	mjansen: try "git clean -nx" to see all untracked and ignored files	15:54
	that's it then	15:54
	get rid of the offending file	15:54
	→ Voker57 joined	15:55
mjansen	so a newer version will tell me what is wrong?	15:55
	i have 1.5.6	15:55
	← sgrimm left	15:55
doener	the "checkout" part of "git svn rebase" is silenced in older versions	15:55
Voker57	can i do something like `svn export`, to copy all tracked files somewhere?	15:55
doener	it's the file that the checkout command complained about	15:55
TML	vuf: That's sort of what I thought, too; but after a "git prune" and "git repack -a -d" I'm still seeing unreachable trees, blobs, and commits when I "git fsck --unreachable HEAD $(cat .git/refs/heads/*)"	15:55
doener	Voker57: git archive	15:55
Voker57	doener: thanks! :)	15:56
vuf	TML: are they perhaps reachable from your reflog?	15:56
TML	vuf: I don't even know how to find that out, but it's something for me to read up on at least. Thanks. :)	15:57
doener	TML: try "git repack -a -d -f" _or_ "git repack -A -d -f" and then "git prune"	15:57
	← archis left	15:57
doener	TML: one of those should work if the objects are truly unreachable through any ref or the reflog	15:57
mjansen	doener, thx	15:58
TML	doener: Ah - "repack -A -d -f" seems to have cleaned it, thank you.	15:58
doener	TML: I think the "-a -d -f" version should be enough, but I've seen at least one case where the other, together with git prune was required	15:58
	← deskin left	15:59
TML	I only wanted to make sure those were all really removed because somewhere between git and svn, we lost an entire svn commit last night - repo changes from 11269 to 11271, and we can't see the changes that were supposed to be in 11270	16:00
	Just trying to remove all variables :)	16:00
	← kef left	16:03
	→ kef joined	16:05
	→ bts- joined	16:05
	→ perezd88 joined	16:06
	→ ShadeHawk joined	16:06
ShadeHawk	Is logo used on http://git.or.cz/gitwiki/EclipsePlugin official JGit/EGit logo?	16:07
	→ bobmcw joined	16:07
ShadeHawk	Hmmm... it looks like it is, as http://www.jgit.org/ uses it too	16:08
	Does StGit have logo or icon?	16:11
	Or Guilt, or TopGit?	16:12
	← nud_ left	16:13
	← ijcd left	16:15
	← advany left	16:18
	← mjansen left	16:20
	→ capisce_ joined	16:22
	→ fanthomas joined	16:22
	← statim left	16:22
	→ kbingham joined	16:22
	→ statim joined	16:22
	Hobbes → Guest58712	16:23
	→ Hobbes`` joined	16:23
	→ koke joined	16:24
	→ sgrimm joined	16:25
	→ carrus85 joined	16:26
	→ Ryback_ joined	16:27
	← tritz left	16:30
	← svend left	16:31
	ph^_ → ph^	16:36
	← jmesnil left	16:37
	→ jmesnil joined	16:37
	→ pt joined	16:38
	← Guest58712 left	16:38
	→ ptah joined	16:38
	← ptah left	16:38
	← ebel left	16:39
	Hobbes`` → Hobbes	16:39
	← ToxicFrog left	16:40
	→ ToxicFrog joined	16:40
	← ToxicFrog left	16:40
ShadeHawk	Hmmmm... it is silent here...	16:41
	→ ToxicFrog joined	16:42
	← capisce left	16:42
	← rmh3093 left	16:42
	vuf bangs on stuff to make so noise	16:42
	stouset → stouset\|away	16:44
	← gbucher left	16:46
	→ schacon joined	16:47
	← koke left	16:48
	← hurikhan77 left	16:48
	→ kroimon joined	16:51
	← christophsturm left	16:51
kroimon	hey all	16:51
	i just wanted to know if there's a git version i can use on suse 9.3 since i have a production server with this distro and dont want to update all requirements to install the latest version...	16:52
drizzd	kroimon: I'm not aware of any new requirements introduced by the latest version	16:54
jonkristian	is there anything i need to enable to get the post-update hook going, i did a chmod +x on it, but it still won't do anything.	16:55
Tv	jonkristian: what do you think the default post-update hook does?	16:57
	that is, are you sure it's not working ;)	16:57
	crab → justacrab	16:57
	← MaddHatter left	16:58
jonkristian	it was working, but then i went ahead and created a new repository and just copied over the post-update i was using.	16:58
	→ ijcd joined	16:59
	← olauzon left	16:59
	→ nbrosnahan joined	17:01
drizzd	maybe it doesn't run if all refs are already up-to-date?	17:01
jonkristian	yeah, im changing a file locally, doing a commit and then a push	17:02
drizzd	and you're aware of the fact that the post-update hook has to reside on the remote side	17:02
jonkristian	the change isn't updated in that folder, im using this, http://www.linuxworld.com/community/?q=node/3057	17:03
	drizzd: yes, it's in my repository hooks folder on the server	17:03
	← schacon left	17:03
jonkristian	but it was working before, except that i threw out some permission errors which i have fixed since	17:04
Tv	jonkristian: "it's not working" is not very useful for debugging	17:04
vuf	jonkristian: you can stuff in "echo test" to make sure that it is running at all	17:04
drizzd	and you also updated WORKDIR, of course	17:04
	→ amerine_ joined	17:04
	→ mithraic joined	17:05
	← amerine_ left	17:05
	→ amerine joined	17:06
drizzd	I wonder if there's a better way to get the working directory	17:06
jonkristian	yeah, sorry, ill try that	17:06
	→ mw-home joined	17:08
	→ MaddHatter joined	17:08
jonkristian	test is echo'ed during push, so it is running	17:08
vuf	drizzd: "unset GIT_DIR ; cd repo" is maybe cleaner	17:08
	← kroimon left	17:09
	→ kroimon joined	17:09
mw-home	When I clone a branch with git "checkout -b newbranch", does my newbranch get the same origin as its parent?	17:09
drizzd	vuf: so there's no way to tell where "repo" actually is I gather?	17:09
	(from inside repo)	17:09
vuf	drizzd: or not, I didn't notice the update-server-info	17:09
	drizzd: I think the hook tries to update one repo in the hook of a different repo	17:10
drizzd	I think it's for pushing into a non-bare repo	17:10
Tv	mw-home: branches have no parents	17:10
jonkristian	ah	17:11
Tv	mw-home: "git checkout -b newbranch" creates a new branch pointing to current commit (aka HEAD)	17:11
vuf	drizzd: in that case, I use "cd .. ; unset GIT_DIR" with success ... the hook starts out in .git with GIT_DIR="."	17:11
ShadeHawk	mw-home: "git checkout -b newbranch" starts 'newbranch' at current HEAD	17:11
Tv	mw-home: http://eagain.net/articles/git-for-computer-scientists/	17:11
jonkristian	drizzd: my old repo had, project.git/.git/ but new one is bare, project.git/ could that be it then?	17:11
drizzd	vuf: I see. jonkristian: maybe you should use what vuf suggested instead of hardcoding the working directory path	17:12
vuf	jonkristian: if it is bare, there is no work area to update	17:12
drizzd	jonkristian: if it's bare you don't need the hook at all	17:12
ShadeHawk	git rev-parse --show-prefix (or something like that)?	17:12
	→ qrush joined	17:12
drizzd	ShadeHawk: doesn't work from inside .git	17:12
mw-home	OK, after making my branch, I did "git push origin newbranch". Then I did git branch -r. I don't see newbranch in the origin though.	17:12
	never mind	17:13
jonkristian	drizzd: don't need the hook? how can i else be updating the branch in th webfolder on push?	17:13
mw-home	I do see it	17:13
	now, when somebody else also already cloned my origin, how do I say, hey, now retrieve in the newbranch on origin?	17:13
Tv	mw-home: git fetch	17:14
ShadeHawk	drizzd: git rev-parse --git-dir (or man git-rev-parse in general)	17:14
Gitbot	drizzd: the git-rev-parse manpage can be found at http://git.or.cz/man/git-rev-parse	17:14
drizzd	jonkristian: git reset only works from within a git directory	17:14
	jonkristian: s/directory/repository	17:14
mw-home	Tv: ok, I also did git pull origin newbranch. That works too, right?	17:14
	← namenlos left	17:14
	→ ijcd_ joined	17:14
	mw needs to configure his client to not match "mw-home" :S	17:14
	mw-home → notmw	17:15
Tv	mw-home: for some definition of works; it does something else, which might or might not match what you wanted	17:15
	mw: i usually advise people to avoid "git pull foo bar"	17:15
	→ schacon joined	17:15
Tv	mw: because many people don't realize that syntax does a merge, too	17:15
notmw	Tv: it seems to have retrieved the edits I made while in my newbranch , and it applied them in there.	17:15
jonkristian	ok, so maybe i should go back to bare none bare	17:15
drizzd	jonkristian: but you can specify the working directory somehow, not sure if that works with a bare repo though	17:16
mw	Tv: i realize it, but maybe notmw doesn't!	17:16
Tv	mw: exactly. use git fetch	17:16
notmw	Tv: you like git fetch instead? That downloads the code, but keeps it somewhere else?	17:16
Tv	puts it in remote branches	17:16
notmw	Tv: I changed my nick a minute ago from mw-home to notmw.	17:16
Tv	sounds like somebody needs to (re)read the tutorial	17:17
notmw	Tv: yeah, I've read it, but I think I need to re-read it. thanks for the hel.	17:17
	p	17:17
	.	17:17
	→ jatt joined	17:17
Tv	i can give hell too	17:17
jonkristian	drizzd: yeah, though it's a bit over my head:D	17:18
	→ bobesponja joined	17:18
	→ rmh3093 joined	17:18
drizzd	jonkristian: same here, sorry	17:18
jatt	hi, I do have a big git repository and now I want to clone (part of it) in a machine of my university which has a pretty small quota. How do I clone for example a subdirectory from my master repository?	17:18
drizzd	jonkristian: you could simply export the repo using 'git archive', but that's probably slower than git reset	17:19
	→ narflekw joined	17:19
	→ gittish joined	17:19
	→ lolage0 joined	17:19
	← jmesnil left	17:20
jonkristian	maybe GIT_WORK_TREE	17:21
	→ eikonos joined	17:21
	← ijcd left	17:21
drizzd	jonkristian: you can try it. If git barks about the fact that it's bare, change the config	17:22
vuf	jonkristian: I think the bottom line is that the linuxworld.com blog is wrongish, which is confusing you	17:22
	jonkristian: the eikke.com blog that it references includes a git-pull, which makes more sense	17:23
jonkristian	ill look at it:)	17:23
	that's actually what i tried to do myself, before i came across that article	17:24
	← powr-toc left	17:24
	← harinath left	17:24
	← gittish left	17:24
charon	jatt: you can't. to save space on the object storage, try clone --depth. if you really must have a subtree checkout (not clone!), you could try duy's patches on the list, search for "narrow/partial/sparse checkout", but you'll have to compile git yourself.	17:24
	→ rmh3093_ joined	17:24
	← hiffy left	17:25
vuf	jonkristian: you need a non-bare repo somehow, though the article suggests two repos, one bare and one not	17:25
drizzd	jonkristian: yes, vuf is right. git pull is a much better solution	17:25
Tv	jatt: git really is built on the idea of having everything always handy, you may need to find a better box to work on	17:26
drizzd	you can even use clone -s if you're concerned about space	17:26
	→ Fullmoon joined	17:26
jonkristian	vuf: drizzd thanks guys, i'll take a deeeeep look into it, and hopefully get it working:)	17:27
vuf	jonkristian: my full script is "cd .. ; unset GIT_DIR ; git reset --hard HEAD" (for a non-bare repo that I push to)	17:28
drizzd	actually, what's the downside of using clone -l vs. clone -s?	17:28
narflekw	can anyone confirm or deny that hooks work on remote git repos accessed over http?	17:28
	i can't seem to get mine to run after a `git push`	17:29
drizzd	narflekw: hooks do not work over http (currently)	17:29
narflekw	ahhh ok	17:29
charon	drizzd: afaics from the descriptions, -s saves a few inodes and can cross FS boundaries, but is dangerous	17:30
narflekw	so i need to set git up to run as a daemon yes?	17:30
	→ deavid joined	17:30
jonkristian	YEES, it's working:D	17:30
	thanks guys, i felt so lost before.	17:30
	stouset\|away → stouset	17:30
	→ pmf joined	17:31
vuf	drizzd: I think that "git gc" in one repo could break the hardlinks, rigth?	17:31
drizzd	charon: right. forgot about that. I guess it's also useful in the context of --reference.	17:31
	well, hardlinks are safe, alternates are not	17:32
	a hardlinked file will note be deleted until all links are gone	17:32
ShadeHawk	charon, jatt: partial checkout would save only space taken by checkout; use shallow clone, or alternates with reference repository outside quota (local disk, or USB key, or CD-ROM,...)	17:32
vuf	also, I am not sure whether fetch will hardlink ?	17:32
narflekw	drizzd: thx for the heads up, that was frustrating	17:32
RandalSchwartz	and open references	17:33
ShadeHawk	there is git-relink (man git-relink)	17:33
Gitbot	ShadeHawk: the git-relink manpage can be found at http://git.or.cz/man/git-relink	17:33
drizzd	narflekw: yeah, a new http protocol which supports hooks is in the works	17:33
	← ShadeHawk left	17:33
jatt	thanks for your suggestions. It seems it needs a bit of work... If it gets too complicated for me then I will need to reduce the size of my repository.	17:33
	← pmf left	17:34
drizzd	narflekw: over WebDAV it's not possible to execute hooks, obviously	17:34
	← joevandyk left	17:35
drizzd	vuf: good question	17:35
	← mcella left	17:36
drizzd	vuf: OTOH, it probably uses the same transport code 'git clone' does	17:36
	→ tritz joined	17:36
	→ priidu joined	17:36
	← kroimon left	17:37
	← rmh3093 left	17:37
narflekw	drizzd: yeah i was wondering how that would work, but i couldn't find anything documented about how it doesn't work yet	17:37
	← kbingham left	17:37
drizzd	I don't want to start a "how secure are sha1s" discussion. But I've been wondering how git would actually deal with a sha1 collision. It would detect different object sizes at least, right?	17:38
ThaDon	Allo, is it possible for git to clone off an http location protected by basic auth (i.e. .htaccess)? I'm getting a 401 (rightly) but no prompts for username/password. Is it supported?	17:38
	→ niki joined	17:38
	→ segher joined	17:38
Tv	drizzd: don't think so	17:38
jatt	If I create a shallow clone (with --depth 1) will I be able to pull changes from the master repository?	17:38
drizzd	ThaDon: man netrc	17:39
	← charon left	17:39
	← sveajobb left	17:39
drizzd	Tv: true... it would have to read the object for that, which would be slow.	17:39
raz	hmm i'm still confused by "git mergetool". why on earth does it give me 3 files: file.LOCAL, file and file.REMOTE. i'm fine with file.LOCAL and file.REMOTE, but the middle on ("file") is a combination of the two, garbled with merge markers. how is that helpful?	17:41
	→ prof_freud joined	17:41
Tv	raz: because only some of the things conflicted; anything that managed to automerge is automerged	17:41
vuf	drizzd: if a collision happens by accident, that would be during an add ... my guess is that git would then just reuse the wrong file, seeing that it knows the sha-1 already	17:42
Tv	raz: besides, the middle file is the output, so you need that there	17:42
	→ kbingham joined	17:42
drizzd	vuf: makes sense	17:44
raz	Tv, yes but why is the middle file not blank by default? at least xxdiff and meld seem to be pretty clueless about merge-markers and give me a less-than-helpful view like this: http://mbox.bz/s/ss34.jpg	17:44
	← sgrimm left	17:44
	→ sgrimm joined	17:44
Tv	raz: if it was blank, you'd need to merge everything manually	17:44
	→ hiffy joined	17:45
Tv	raz: in the real world, only a small fraction of the changes cause conflicts	17:45
xinming_	anyone here can tell me why in 1.6.1 series git, the git-shell will be removed in /usr/bin/ please?	17:45
	→ KevBurnsJr joined	17:45
KevBurnsJr	Best git client for mac?	17:45
Tv	xinming_: mistake	17:45
xinming_	for now, If we want to setup the restricted shell, we have to use /usr/libexec/git-core/git-shell thing	17:45
	which isn't a natural thing to me	17:46
	Tv: do you mean it'll be corrected in newer release?	17:46
raz	Tv, yes but when large parts of the file are identical, isn't it just a matter of choosing this large hunk from either left or right? in contrast with the merge-marker file in the middle there is sometimes no straightforward way to replace the garbled parts with the left/right version in the correct order	17:46
	← ph^ left	17:46
Tv	xinming_: was already	17:46
ThaDon	drizzd: thanks dude	17:46
	→ joydivider joined	17:46
ThaDon	drizzd: got me past the 401, now it complains with "info/refs not found: did you run git update-server-info on the server?" however I did run that command	17:47
raz	Tv, i would understand git mergetool put the common ancestor in the middle. but this garbled file in the middle makes no sense to me.	17:47
	+if	17:47
drizzd	ThaDon: man git-update-server-info	17:47
ThaDon	drizzd: might be an apache setting	17:47
Gitbot	ThaDon: the git-update-server-info manpage can be found at http://git.or.cz/man/git-update-server-info	17:47
Tv	raz: if you want that, erasing the middle file is a few keypresses away, and you can even automate it if you want	17:47
ThaDon	drizzd: yeah, i ran it, I read it, etc...	17:47
drizzd	ThaDon: maybe you got the url wrong?	17:47
xinming_	ha, thanks, After tried and googled, It always not worked as expected. and at last find we lacks git-shell. >_<	17:47
Tv	raz: just saying that doesn't sound helpful in non-pathological cases, to me	17:47
drizzd	does the file info/refs exist in your repo?	17:48
vuf	raz: you can define a mergetool that uses $BASE for the middle, I guess	17:48
raz	Tv, well yes, figured that out. just scratching my head about the default. why is the default not to put the ancestor in the middle? that seems much more helpful, especially in non-pathological cases?	17:48
	← loiseau left	17:48
vuf	Tv: one could imagine a mergetool that also has auto-merge capability	17:48
drizzd	ThaDon: you can also test by entering the url in your browser	17:48
raz	Tv, or, in other words, how is putting the mergemarker-version in the middle helpful at all?	17:48
Tv	raz: shrug current works very well for me, in the real world	17:48
raz	hmm well ok	17:49
	maybe i'll grasp it when i do some more real-world merges	17:49
	→ ph^ joined	17:49
drizzd	raz: I don't think mergetool is very popular	17:50
ThaDon	drizzd: yeah, tested with the browser and see everything fine :(	17:50
	ThaDon confused	17:50
vuf	raz: I like a view like this better, http://www.scootersoftware.com/moreinfo.php?zz=screenshot&shot=TextMerge	17:50
	← kowey left	17:50
drizzd	ThaDon: can you try the same url (to info/refs) with curl?	17:50
raz	drizzd, what do you recommend instead?	17:50
ThaDon	drizzd: Im supplying git clone with the url to the parent directory of .git	17:51
drizzd	ThaDon: I think that's fine	17:51
	raz: I use git diff and an editor	17:51
raz	vuf, hmm yea that looks like what i would have expected from merge-tool (ancestor in the middle or left, right?)	17:51
	drizzd, yea well, i guess that becomes cumbersome with more complicated merges	17:53
	→ ponch84 joined	17:53
vuf	raz: ancestor in the middle is what it says in the caption (I didn't get it set up myself, yet)	17:53
drizzd	ThaDon: actually, you may have to append .git. Could you try that?	17:53
	Tv huggs wiggle for quick kludging annoying rejects... wiggle does the editing, diff & unit tests ensure sanity	17:54
ThaDon	drizzd: interesting, yes that was it	17:54
raz	vuf, well, git-mergetool is just a shell script. i guess some patching is due. baffles me a bit since git kinda advertises its superior merge-facilities everywhere :\	17:54
Tv	that scooter thing looks way too visually busy to be desirable to me	17:54
	looks like a typical windows ide thingie ;)	17:54
drizzd	ThaDon: hmm, that's a little stupid. I guess you're the first to try cloning a non-bare repo over http ;-)	17:54
ThaDon	drizzd: so do you only have to do that when you clone from http? I don't seem to have to do that when cloning from ssh or a mounted filesystem	17:54
drizzd	ThaDon: exactly	17:54
ThaDon	drizzd: lol, no no there has to be more idiots out there than just me	17:55
vuf	Tv: it is, but it beats any unix compare tool by miles	17:55
ponch84	hi all, anybody using hg-fast-export knows how to track hg HEAD on a specified branch and not on master?	17:55
raz	vuf, doesn't seem to work on my 64bit gentoo :(	17:55
drizzd	ThaDon: I'll see if that easy to fix and send a patch	17:56
raz	i generally like meld a lot but it has some annoying bugs (undo doesn't always really undo)	17:56
vuf	raz: I think the advertised superiority is in auto-merging more than the manual conflict resolution	17:56
raz	that's the last thing you want in a merge tool	17:56
	→ catphish joined	17:56
ThaDon	drizzd++	17:56
	→ markkalderon joined	17:56
glommer	guys, I have this repo, say, kvm-userspace.git. For those who don't know, it has a local copy of qemu inside a qemu/ directory. Now, I'd like to merge a branch of mine of qemu into it. Problem is that for that to happen, I'd need to have the "qemu/" path prepended to everything that comes to that repo. How can I do that ?	17:56
vuf	raz: I got it running in 64 bit Ubuntu, now I forgot the details :-(	17:57
	← capisce_ left	17:58
MaddHatter	glommer: git filter-branch, I think	17:59
	→ capisce joined	17:59
glommer	MaddHatter, will take a look. Thank you.	17:59
catphish	quick question: when using git over ssh, how do you inject plaintext errors back to the client? ie. "you do not have write access to this repo"	17:59
Tv	catphish: stderr	17:59
	catphish: btw sounds like you might want gitosis	17:59
catphish	ooh of course, thanks tv	17:59
vuf	raz: it took a bunch of 32 bit libraries, obviously ...	17:59
catphish	thanks but i've just written a gitosis clone as i wanted user access to come from mysql	18:00
	→ abenea joined	18:00
Tv	shrug	18:00
	← catphish left	18:01
	← floppyears left	18:02
	→ WALoeIII joined	18:02
	← jackdempsey left	18:03
vuf	the best thing about that scooter compare is that you can have the diff pushed up or down by marking which lines align in the left and right side	18:06
	→ floppyears joined	18:07
Tv	yeah sometimes meld gets pretty badly confused about that	18:08
abenea	suppose I touch a bunch of files, some of wich I want to commit on one branch and the others on the current branch. Is there a painles way of doing this?	18:08
Tv	abenea: commit some, checkout branch, commit rest	18:08
	→ krawek joined	18:08
vuf	what are the exact rules for when a dirty work tree will prevent you from changing branch?	18:09
	→ nickh joined	18:10
abenea	Tv: can't checkout another branch on dirty working dir	18:10
Eridius	vuf: when changing the branch would normally modify the dirty file	18:10
	abenea: you can if the file is unchanged in that other branch	18:10
	← floppyears left	18:10
Eridius	basically, it errors out when it goes to write out the new version of the file and realizes the file's dirty	18:10
	although that's not really true, because it errors out before writing any files	18:10
Tv	it refuses only when the alternative would be losing your edits	18:11
	and then you can say "oh you'll probably get it right" with -m	18:11
	→ floppyears joined	18:11
vuf	okay, so no merging is done	18:11
abenea	Eridius: thanks, I know what I was doing wrong :)	18:11
	rmh3093_ → rmh3093	18:12
notmw	Question about git pull. When I do git pull origin branch99, does that pull the remote origin's branch 99 into what ever branch I currently have checked out?	18:12
Tv	notmw: merges, yes	18:13
	→ advany joined	18:14
	← advany left	18:14
	→ advany joined	18:14
	← advany left	18:16
	← sgrimm left	18:21
	→ Beket joined	18:22
	← prof_freud left	18:23
Voker57	can i push via git:// protocol? Can't find any docs...	18:23
thiago_home	yes	18:23
	unauthenticated and unencrypted	18:24
	→ advany joined	18:24
thiago_home	i.e., anonymous pushing	18:24
	→ tongueroo joined	18:24
Voker57	ugh :(	18:24
	→ charon joined	18:25
	→ david_koontz joined	18:25
	→ deskin joined	18:26
Voker57	SVN is better in that way, i can push to repo without shell access	18:28
Tv	Voker57: git can do that perfectly fine	18:28
thiago_home	you don't need shell access to push	18:28
	(even with ssh)	18:28
Voker57	Tv: unauthenticated and unencrypted?	18:28
thiago_home	i.e., ssh != shell	18:28
Tv	Voker57: no, that would be silly	18:28
demq	could someone remind me of the command to uncommit the last commit?	18:29
thiago_home	demq: just the last one? Remove it from history completely?	18:29
demq	no i want the changes, i just dont want them to be in one commit	18:29
Voker57	i have only ftp access and ability to run CGI scripts, can i push to git with those?	18:29
demq	i meant to commit only one file, but out of habit include -a	18:29
thiago_home	so you're going to commit them again?	18:29
	ah, I see	18:29
	git reset HEAD~	18:30
glguy	HEAD^	18:30
thiago_home	same	18:30
demq	yes, but as three commits not 1	18:30
glguy	~3	18:30
Voker57	(i can launch daemons via php-shell too :P)	18:30
thiago_home	HEAD~ works on Windows and zsh, HEAD^ doesn't	18:30
glguy	err	18:30
	thiago_home: doesn't ~ signify things not to match?	18:30
	hmm, you're right, of course	18:31
thiago_home	yes, with extendedglob, sure. But you need something not to match	18:31
glguy	good tip	18:31
	demq: if you reset one commit	18:31
thiago_home	you can also set nonomatch, which I do	18:31
glguy	you can then commit it back in parts	18:31
	← BabelO left	18:31
Tv	Voker57: you could kludge something, but if you can't figure it out, you probably don't want to even try	18:31
	Voker57: decent virtual machines are $20/month, gitorious and github are free	18:32
	or repo.or.cz	18:32
demq	seemed to work as i wanted.	18:33
	thanks a lot folks.	18:33
thiago_home	Voker57: someone was working on cgi receiving	18:33
	maybe that was spearce	18:33
Voker57	Tv: yes... looks like i have to find better hosting. I'm on github meanwhile, but i don't like services which i can't change	18:33
thiago_home	but I don't think it's anywhere near complete	18:33
Tv	Voker57: $20/mo, slicehost has been good to me	18:34
demq	btw, would it have made any difference if i wanted to go back further? like three commits? i guess not but....	18:34
	← tsb left	18:34
thiago_home	demq: nope. HEAD~3	18:34
	→ tsb joined	18:34
demq	cool.	18:34
	thanks a lot. much obliged	18:34
thiago_home	just remember that "reset" is equivalent of removing that commit from history	18:34
demq	yeah.	18:34
thiago_home	so, if you've pushed it and other people are using, they'll have a hard time following up	18:34
demq	yeah	18:35
	no this was unpushed.	18:35
thiago_home	good	18:35
	← ciaran_ left	18:35
demq	ive used reset a lot, but with --hard mostly. for some reason it didnt "click" that it was relevent here tho. silly me.	18:35
	→ neurodamage joined	18:38
neurodamage	what's a good way to send email notifications when source code changes are commit?	18:39
vuf	Tv: so, can I install an existing server image to slicehost?	18:40
Tv	vuf: hmmm, not just like that i think	18:40
	vuf: though they might do it if you ask them	18:40
	vuf: you get a selection of most popular OS preinstalls	18:40
	vuf: and naturally you can overwrite it, if you're careful	18:40
	vuf: i know people switched from rh to debian remotely ;)	18:41
	vuf: (last line not related to slicehost)	18:41
	← WarrenK left	18:42
	→ j-b joined	18:43
j-b	hello	18:43
vuf	Tv: I guess my question really is whether you can modify the bootloader at will, to point to a different root	18:43
	(anyway, off topic)	18:43
j-b	How can I split a file into many while keeping the history ? doing cp oldfile $newfiles; git add; git commit doesn't detect the copies ?	18:44
Tv	vuf: hmm they use xen so you're stuck with the stuff that's in dom0	18:44
	vuf: or at least, my box there uses xen; it predates kvm	18:45
vuf	j-b: it will, with some options	18:45
j-b	vuf: options in cp or in add ?	18:45
Tv	j-b: write operations never detect anything	18:45
vuf	j-b: copies/renames are detected when reading out of the repo, not when storing in it	18:46
	Tv: thanks. Is eagain on slicehost?	18:46
Tv	vuf: yes	18:46
j-b	Tv: when I do one file to another one, it seems to work	18:46
	but not in one to many	18:47
	← spool left	18:47
charon	j-b: try 'git blame -C $file' and see if it detects the right source	18:47
vuf	j-b: for example, "git show -C -C" will work	18:48
	→ johan-s joined	18:49
j-b	charon: no, it doesn't	18:49
	vuf: yes, it does.	18:51
	→ sgrimm joined	18:51
j-b	vuf: but then git blame -C doesn't	18:51
charon	apparently your slices are too small to be detected	18:51
	give a numeric arg to -C	18:51
j-b	too small ?	18:52
	235Ko is too small ?	18:52
charon	hmm	18:52
	"<num> is optional but it is the lower bound on the number of alphanumeric characters that git must detect as moving between files for it to associate those lines with the parent commit."	18:52
vuf	j-b: try some more -C	18:52
j-b	vuf: in blame ?	18:53
vuf	yeah	18:53
tmm1	how do i undo a merge.. branch X was a featured merged into master, but it had a bug	18:53
j-b	vuf: trying it now	18:53
	→ loiseau joined	18:53
vuf	tmm1: reset to before the merge	18:53
charon	default copy score is 40 apparently	18:54
tmm1	what if i have commits since then	18:54
	← joshbuddy left	18:54
vuf	tmm1: fix the bug	18:55
	← giom left	18:55
j-b	charon: git blame -C -C takes forever, but I am still hoping	18:55
	← sgrimm left	18:56
	→ sgrimm joined	18:56
	→ andreaja joined	18:57
tmm1	vuf: ...	18:57
	git revert -m looks like what i need	18:57
vuf	tmm1: right	18:57
Tv	tmm1: in general, you can't undo anything you've published; you can make the reverse change, and that's what revert does	18:57
	tmm1: but it's not the same as undo; the merge is done for good	18:57
j-b	btw, we received a mail for the GitSurvey... and our answer is : "Git is great" 'make it faster on windows' :D	18:57
	vuf: charon: blame -C -C works	18:58
	→ WarrenK joined	18:58
charon	nice... unfortunately it's so expensive that iirc linus recommended it as a stress test	18:58
	← whee left	18:58
j-b	charon: so, how do I commit that now ?	18:59
tmm1	Tv: that's fine	18:59
	→ whee joined	18:59
charon	j-b: you've already committed, haven't you? blame is strictly inspection	18:59
j-b	yes	18:59
	charon: sorry, push	18:59
charon	all copy and move detection in git is after the fact	18:59
j-b	yes	18:59
tmm1	Tv: i don't get what argument -m takes though, how are the merge parents numbered?	19:00
charon	j-b: how to push depends on where to, but you just push out the branch like usual	19:00
	i.e. 'git push <remote> <branch>', or given sufficient tracking setup, 'git push'	19:00
vuf	charon: it may be expensive, but do other tools even have the operation?	19:00
	→ gbucher joined	19:00
charon	vuf: i'm not complaining :)	19:01
j-b	charon: yes, but other people will need to git blame -C -C to get it, right ?	19:01
vuf	charon: that was not a rhetorical question :)	19:01
charon	j-b: to _inspect_ it	19:01
	→ renan_saddam joined	19:01
renan_saddam	well, I do "git pull" and git returns a list of files like: "app/views/tasks/_form.rhtml: needs update"	19:02
	What I need to do to update it?	19:02
Tv	tmm1: branch things are merged to is always #1	19:02
charon	j-b: to _get_ the change, they just proceed like always. to inspect in a way that shows where the lines "really" came from, they'll have to blame, but even if they just look at 'git log -- $file', the last commit in their history should be a pretty strong hint, shouldn't it	19:02
vuf	renan_saddam: commit it, it means the work area is dirty	19:02
Tv	tmm1: so depends on where you run merge/pull	19:02
charon	s/last/earliest/	19:02
Tv	tmm1: git show to see	19:02
j-b	charon: hmmm... ok... but not really optimal, but I understand. thanks	19:03
renan_saddam	vuf: commit? commit means send my version to server?	19:03
charon	vuf: i don't really know that many vcs but i recall a blog post by junio where he claimed this was unique to git and invented by him	19:03
renan_saddam	vuf: I wanna update my repo, like in svn	19:03
	→ boto joined	19:04
vuf	renan_saddam: you have local, uncommitted changes	19:04
renan_saddam	vuf: wow! I think I got it!	19:04
	vuf: just config files	19:05
j-b	charon: vuf: Tv: thanks you very very very much. Keep the good work.	19:05
	← univate left	19:05
renan_saddam	thanks vuf!	19:05
	← j-b left	19:05
	charon thinks renan_saddam should read a tutorial on the D in DVCS	19:06
tmm1	Tv: do merge commits only happen on conflicts?	19:07
Tv	tmm1: http://eagain.net/articles/git-for-computer-scientists/	19:07
renan_saddam	charon: you'r right! I get this tip to me, thanks	19:07
	← Voker57 left	19:08
	→ foca_ joined	19:09
	→ cryo_ joined	19:11
	← cryo_ left	19:13
	→ cryo_ joined	19:14
	← qrush left	19:16
	← ponch84 left	19:18
	← renan_saddam left	19:19
	← Beket left	19:19
	→ cilly joined	19:21
	→ litage joined	19:25
	← Lash\| left	19:27
	← irotas left	19:28
	← foca left	19:29
	→ ponch84 joined	19:30
	→ Beket joined	19:31
	→ timcharper joined	19:31
	← loiseau left	19:32
	← iulian left	19:33
	← ponch84 left	19:34
jatt	I compressed all pdf files with gzip and did a git-rm ./\.pdf followed by git-add ./\.pdf.gz and a commit. The repository is somewhat smaller now, but the .git directory size hasn't changed much. Is there a way to prune the deleted files (in such a way that .git gets smaller?)	19:37
Mikachu	if you still have the pdf$ files in the history, no	19:37
jatt	I don't need the history for them anymore, maybe is there a way to prune the files as if they never existed?	19:38
Mikachu	you need to write new history then	19:39
jatt	how?	19:39
	→ irotas joined	19:39
	← notmw left	19:39
Mikachu	check git-filter-branch and git-rebase manpgaes	19:39
charon	jatt: the history is compressed anyway, why do you want to compress the pdfs yourself?	19:39
jatt	charon: I want to pull my repository in a machine from the university where I do have a small quota. So I was thinking in making my repository smaller in order to be able to pull it from the university.	19:40
	andres → andrezZz	19:40
	andrezZz → andres	19:40
charon	oh	19:40
	→ botanicus joined	19:41
charon	the only thing that might get smaller through such a move is the working copy. .git definitely keeps its size	19:41
Mikachu	of course .git will be larger if you have both .pdf and .pdf.gz in the history though	19:42
jatt	charon: you are right. That's exactly what happened. Nevertheless I will check Mikachu advice for git-filter-branch and git-rebase.	19:42
charon	indeed	19:42
	→ ph^_ joined	19:42
	← ph^_ left	19:42
	→ qrush joined	19:45
	→ djwonk joined	19:46
ThaDon	general git-svn q here, have a local git repo, client wants all my stuff up on their svn repo. How do I tell my local repo that it should track a remote svn repo? Such that I can use svn dcommit/rebase. I've only ever used git svn in scenarios where I've git svn clone'd a remote svn repo at project start.	19:50
djwonk	I'd like to test the outcome of 'git pull' for "Already up-to-date.". However, bash return value indicates success for that case. Suggestions?	19:50
ThaDon	now I'm doing the opposite, I have an existing repo and I want to dcommit it all to the remote svn repo.	19:51
jatt	$ git filter-branch --index-filter 'git rm --cached ./\*.pdf' HEAD	19:51
	Cannot rewrite branch(es) with a dirty working directory.	19:51
	How do I clean a dirty working directory?	19:51
Mikachu	jatt: git status	19:51
vuf	I want to stuff my whole project into a subdirectory. Can filter-branch do that with --index-filter ?	19:52
Mikachu	vuf: why?	19:52
vuf	why I want to arrange it like that?	19:53
jatt	Mikachu: thanks that worked	19:53
Mikachu	vuf: yes	19:54
jatt	$ git filter-branch --index-filter 'git rm --cached ./\*.pdf' HEAD	19:54
	Rewrite 1fef263b9bb90b3883a2ef78557663314da21b56 (1/263)fatal: pathspec '*.pdf' did not match any files	19:54
	hmm.. so I guess I need to give the name of the files	19:54
	← ph^ left	19:54
jatt	the full path	19:54
charon	why not just 'git rm --cached ./*.pdf'	19:55
Mikachu	jatt: --ignore-unmatched	19:55
	→ brendon_work joined	19:55
charon	the argument is evaluated in the shell which will expand the * for you	19:55
jatt	charon: I am not sure if that will purge the history too.	19:55
Mikachu	jatt: s/ed$/	19:55
charon	jatt: with gfb of course	19:55
	git filter-branch --index-filter 'git rm --cached *.pdf'	19:55
	of course, you need a different approach if you also have pdfs in subdirectories	19:56
	← cilly left	19:56
charon	(with ls-files or so)	19:56
jatt	oh I see... hmmm is that recursive? If yes, of course 'git rm --cached ./*.pdf' is the right argument	19:56
charon	oh, sorry, i forgot that you're in index-filter	19:56
jatt	charon: yes I compressed all pdf in the subdirectories too	19:56
vuf	Mikachu: rearranging a bit during a cvs conversion, and I would like the history not to have a rename of all files	19:56
charon	jatt: forget what i said, i'll check what the real problem is	19:57
Mikachu	vuf: it just seems weird to have a repo where all files are in a subdir	19:57
charon	jatt: mikachu was right... almost. the option is --ignore-unmatch	19:58
vuf	Mikachu: there are other parts of the project, but they have not been kept in cvs ... I want those in the root, too (eventually)	19:58
Mikachu	charon: i was right, as i corrected myself	19:58
charon	oh, right.	19:58
Mikachu	vuf: ah	19:58
	charon hides	19:58
jatt	I see. So --ignore-unmatch will go ahead even if the files were previously removed?	19:58
Mikachu	or didn't exist yet	19:59
johnw	given a modified file, can I make every modification in that file appear as if it were a merge conflict, with <<< == >>> markers to show the two?	19:59
charon	no, it will go ahead even though there is no file called (literally) *.pdf	19:59
	instead of causing gfb to abort	19:59
	by the way, to really rewrite all of history you probably need to use '-- --all' instead of HEAD	19:59
Mikachu	vuf: git mv --cached something something?	20:00
johnw	i guess what I want generally is: giving a patch, change the subject file so that every patched region is now looks like a conflict marked region, so that both the patch and the original are now sitting side-by-side. Is there a tool that does this?	20:01
Mikachu	oh that option doesn't exist, bad advice :)	20:01
Tv	git add and rm --cached	20:01
vuf	Mikachu: well, something like it, anyway	20:01
jatt	nice	20:01
	$ git filter-branch --index-filter 'git rm --ignore-unmatch --cached ./\*.pdf' HEAD	20:01
	worked recursively.	20:01
	Ref 'refs/heads/master' was rewritten	20:01
	let's see if .git is smaller after commit :)	20:02
Mikachu	jatt: it's not	20:02
	jatt: and what do you mean "after commit"?	20:02
	jatt: you just changed every single commit in your history	20:02
jatt	hm. it is larger	20:02
	oh! I see	20:02
	I was thinking about doing git commit after the filter-branch.	20:02
Mikachu	you still have the old ones in refs/original and your reflog	20:02
jatt	(don't know why)	20:03
	hm. Is there a way to get rid of the old ones? (I'm going in the opposite direction my repository is getting larger and not smaller :-))	20:03
Mikachu	yes, rm -rf .git/refs/original; man git-reflog	20:04
Gitbot	yes: the git-reflog manpage can be found at http://git.or.cz/man/git-reflog	20:04
Mikachu	heh	20:04
	jatt: you realize that if the pdfs have changed in history and you only added the latest as gzips, you lose actual history?	20:05
jatt	Mikachu: I think so. I don't need the history of those pdfs at all. They will not change and I only need the current version of them.	20:07
Mikachu	then why are you using git? :P	20:07
jatt	Mikachu: In the repository I do have some pdfs but also text files (code, etc.). I need version control for the text files.	20:08
vuf	jatt: you realize that compressing the pdf files will not make .git smaller?	20:08
jatt	vuf: yes, I'm working in doing it using git-reflog and git-filter-branch, I'm still reading the git-reflog man page :)	20:09
	→ Guest45175 joined	20:10
vuf	jatt: I mean, git compresses files anyway	20:10
	it might actually be smaller if you don't compress, as zipped files cannot be delta compressed	20:11
Mikachu	if you want to make .git smaller, make core.compression 9 or so	20:12
	and repack -adf	20:12
vuf	if you want the checked out files to be small, though, pdf.gz is a fine idea :)	20:13
Mikachu	you can just delete them from the checkout when you don't need them too	20:13
	but i guess most pdf viewers handle .gz transparently?	20:14
Tv	vuf: pdf doesn't compress all that well	20:14
	back in the day of postscript, .ps.gz was common; no longer with pdf	20:14
jatt	vuf: I want to clone my repository in a machine at the university which has a ridiculously small quota, so I want that the cloned repository fits there	20:14
vuf	Tv: that depends, but it can have gzip/png/jpeg internally, so it might not compress well	20:14
Tv	vuf: sure, i mean on the average	20:15
jatt	right some debian packages contain compressed pdfs, but they are not so frequently used (I think acrobat reader doesn't uncompress them automagically as evince does).	20:15
vuf	is LaTeX .pdf output compressed by default?	20:16
jatt	pdflatex doesn't compress by default AFAIK	20:16
johnw	cool, this did it: git show branch:file > /tmp/file ; diff -D NAME /tmp/file file	20:17
vuf	jatt: okay, so compressing to pdf.gz makes sense ... proper hardware in a university would make even more sense, though :-)	20:19
jatt	I'm lost which option for git-reflog should I use after rm -rf .git/refs/original Mikachu	20:19
Mikachu	jatt: you want to expire the reflog entries from before the rewrite so that a repack will delete the old commits	20:20
	← rmh3093 left	20:20
jatt	I will go for higher compression rate after I see what the filter-branch option did for a smaller .git	20:20
	→ Tammy-- joined	20:20
vuf	I am thinking that cloning the repo should get rid of the old cruft, no?	20:20
	→ timprouty joined	20:22
jatt	ok so I need to give a sound value to the --expire option? What if I choose --expire=0? will this affect all commits or only those for the files affected with filter-branch in the previous step?	20:22
drizzd	vuf: I don't know what happens if you hardlink a pack file though	20:22
	→ rdavila joined	20:23
Mikachu	vuf: yeah, if you use file://	20:23
	← botanicus left	20:24
jatt	and the last thing: I'm not sure if git-reflog will create a new .git/refs/original directory?	20:24
Mikachu	no, only filter-branch does that	20:24
	and the other question doesn't matter, since you haven't made any commits after the rewrite	20:25
	→ gbucher_ joined	20:26
jatt	I see, I won't do rm -rf .git/refs/original after I understand better all involved commands :)	20:26
Mikachu	you mean before?	20:26
	you can as vuf suggested do a git clone file://your-repo somewhere else on the filesystem, and it should have the 'new' size	20:27
	← deskin left	20:27
jatt	anyway it looks like this route will not compress my .git directory too much. I will try now with a higher compression rate (core.compression 9)	20:27
	Mikachu: I see just cloning it, ok will try that now	20:28
	← tongueroo left	20:29
	→ tongueroo joined	20:29
	→ thannoy joined	20:30
jatt	du --max-depth=1 says 261120 for the clone and 261256 for the master.	20:30
	no much difference	20:31
narflekw	how do i see the message that was associated with a tag? i.e. `git tag` lists just the tag names	20:32
Tv	narflekw: git show	20:32
narflekw	ahh thanks	20:32
	← warlock_mza left	20:33
	→ timcharper_ joined	20:33
	→ Arrowmaster joined	20:33
Mikachu	you can try one of	20:34
	git for-each-ref --format='%(tag) %(subject)' refs/tags	20:34
	git for-each-ref --format='%(refname) %(contents)' refs/tags	20:34
	too	20:34
	or either combination	20:34
	← Tammy-- left	20:35
Tv	if he went that far, surely cat-file would be better	20:35
	← Weaselweb left	20:38
	→ w00t_ joined	20:38
	→ deskin joined	20:40
	← w00t left	20:42
	← cryo_ left	20:45
	← Bryson left	20:46
jogla	if i have different repositories, can i make something like a tag that is visible in more than one?	20:46
	← Guest84351 left	20:47
selckin	submodules maybe	20:47
	← gbucher left	20:47
	← gbucher_ left	20:48
	← vmiklos left	20:48
	← ToxicFrog left	20:49
	← timcharper left	20:49
	← JimmyB_ left	20:50
	→ mcella joined	20:53
	← mcella left	20:53
vuf	jogla: like different repositories of the same project?	20:54
	→ vmiklos joined	20:54
	→ ToxicFrog joined	20:54
	→ drewr joined	20:56
	→ wagle joined	20:56
drewr	Will git check ~/.git/hooks for hooks to run?	20:57
	← bobmcw left	20:58
jogla	vuf: Seems to be something like that, they are at least related	20:58
	But have different releases and different versions	20:58
	I think submodules are the solution	20:58
drizzd	drewr: no	20:58
	→ mcr joined	20:59
drewr	So I need to keep a separate post-receive in each repository?	20:59
drizzd	yes	20:59
Mikachu	drizzd: if you put it in the template dir, every new repo will get it	20:59
drewr	k, thanks.	20:59
Mikachu	drewr even	20:59
mcr	under what conditions would: "git fetch --tags origin" do less than "git fetch origin"? This is with Rails/Capistrano's GIT SCM module.	20:59
	→ Bryson joined	21:00
	← DrFrasierCrane left	21:00
Mikachu	mcr: when origin has updated branches, --tags will not fetch them	21:00
mcr	it won't fetch the branches, it won't fetch the tags, or it won't fetch even origin/master ?	21:01
	← lorandi left	21:01
Mikachu	mcr: it will only fetch tags	21:01
mcr	hmm. so why would capistrano do that. That doesn't make sense.	21:02
	I thought maybe it was something weird about our repo that was confusing things.	21:02
Mikachu	maybe i am confusing fetch with push	21:02
	the manpage is inconclusive	21:02
	→ yofel joined	21:03
	← cannonball left	21:03
	← jatt left	21:03
Mikachu	but thinking about it, i'm probably wrong, and --tags downloads both branches and tags, while push --tags will only push tags	21:03
mcr	that's what I would have concluded from the man page.	21:03
	capistrano wants to run git fetch --tags origin to update the repo on the production server.	21:03
	that does nothing, and so the sha1 hash that it wants to branch against is not present, and things break.	21:04
	but running "git fetch" or "git fetch origin" pulls the missing piecs.	21:04
	cap 2.5.0 does: git fetch -q origin	21:05
	problem solved.	21:05
	← priidu left	21:05
	← eternaleye left	21:05
	→ mkdek joined	21:08
	ToxicFrog ponders writing git integration for nedit	21:09
mkdek	hi, how can I undo local changes non added to the index?	21:09
ToxicFrog	mkdek: as in, you have some modified files, you want to return to the unmodified state recorded in HEAD?	21:10
johnw	git checkout -- file	21:10
ToxicFrog	git checkout -- filenames	21:10
deskin	or git reset --hard HEAD	21:10
ToxicFrog	If you want to revert the entire working copy.	21:10
deskin	if you want to undo all your changes	21:10
mkdek	thanks, checkout worked fine	21:10
ToxicFrog	What gestures would I need... init, clone from, stage/unstage/revert this/all, commit, push/pull/fetch/merge...others?	21:10
johnw	or even git checkout -- .	21:10
	← kumbayo left	21:11
johnw	which is arguably more consistent	21:11
	→ ShadeHawk joined	21:11
	→ dato joined	21:13
	→ etgryphon_ joined	21:13
dato	hello, if I had some commits in an old branch that I would like to keep around, but I would not like to see the branch in `git branch`, what are my options?	21:14
	→ fixUp joined	21:14
johnw	create a tag	21:14
	then delete the branch	21:14
dato	create a fake remote "trash" so that they only show with branch -a?, or there is a better way?	21:14
	johnw: but then the tag will get pushed with push --tags, which I wouldn't like either	21:14
fixUp	Am I right in assuming that to back up a git repository you simply copy the folder to the backup media ?	21:14
johnw	i woudn't mind having "invisible tags" myself	21:14
	← timcharper_ left	21:15
johnw	you could always use a wrapper around "git branch"	21:15
	just filter it out	21:15
dato	heh, I'm not going to do that :)	21:15
ToxicFrog	fixUp: yep, it's entirely self-contained	21:15
dato	but thanks for the suggestion	21:15
johnw	you could create your own ref	21:15
fixUp	ToxicFrog: so It's ok to tar it onto tape ?	21:15
	← mkdek left	21:16
johnw	echo $commit > .git/refs/hidden/foo	21:16
ShadeHawk	dato: move it for example to refs/Attic	21:16
	(copy as a file)	21:16
ToxicFrog	fixUp: indeed, and if you s/tape/USB/ I've done that myself quite often	21:16
johnw	just mkdir .git/refs/hidden first	21:16
	that will keep the commit from being gc'd	21:16
	→ timcharper joined	21:16
	→ eternaleye joined	21:16
Mikachu	fixUp: and it's best if nobody is writing to the repo while you're copying it	21:16
fixUp	ToxicFrog: USB ? an external hdd ?	21:16
johnw	you can then reference it as "git checkout refs/hidden/foo"	21:16
dato	johnw: right, that's what I proposed in the first place	21:16
johnw	ah, I missed that	21:16
dato	hm	21:16
ToxicFrog	No, flash memory - I have a few repositories I keep copies of on a 4GB pendrive for working on on the go	21:17
dato	well, I said a "fake remote"	21:17
ToxicFrog	In cases when I won't reliably have my laptop or a net connection	21:17
dato	because for all I know the "hidden" directory will still show in `git branch`	21:17
	(I have already a "kde3" and "kde4" directories)	21:17
ToxicFrog	Anyways, we digress; all information about a repository is stored in the repository itself (in the .git directory for non-bare repos, or just lying around for bare ones), including history, tags, etc	21:18
	Plus the working copy itself, of course	21:18
	So tar it up and you can untar it later and it's just as you left it	21:18
dato	though maybe I could go for "refs/.hidden"	21:18
johnw	doesn't git branch only show refs/heads	21:18
	?	21:18
dato	though maybe that'll gc it	21:18
fixUp	ToxicFrog: ok, thats what I thought it's good to know	21:18
dato	johnw: all files under refs/heads, including those in subdirectories of it (which is very nice)	21:19
fixUp	ToxicFrog: actually that ability is one of the best features of git	21:19
	ToxicFrog: cheap offline hosting in comparison to svn	21:19
	→ cilly joined	21:19
dato	meh	21:20
	johnw: I'm sorry	21:20
	22:16 <johnw> you can then reference it as "git checkout refs/hidden/foo"	21:20
	I read "refs/heads/hidden/foo" there, go figure	21:20
	johnw: so indeed thanks, I'm going to do that, is a very good suggestion	21:20
	← stouset left	21:21
	→ fade-in joined	21:21
johnw	i thought of that because it's what git-filter-branch does with your old HEAD	21:21
ToxicFrog	fixUp: that's one of the reasons I initially looked at it - I was looking to put my stuff under version control, and git and svn were both high on the list	21:21
johnw	refs/original/etc.	21:21
ShadeHawk	ToxicFrog: ever heard of git-bundle?	21:21
ToxicFrog	So I looked at the install instructions...svn started with "install apache". Then I looked at git..."yum install git; git init"	21:21
	ShadeHawk: nope	21:21
	→ eternaleye_ joined	21:21
ToxicFrog	Hmm. That's cool.	21:22
	→ bobmcw joined	21:22
ShadeHawk	instead of tarring repostiroy on USB drive, you can use git-bundle which is meant for off-line transport (think of it as transport for sneaker-net)	21:22
	man git-bundle	21:22
Gitbot	ShadeHawk: the git-bundle manpage can be found at http://git.or.cz/man/git-bundle	21:22
ToxicFrog	Yeah, I just looked at the man page	21:22
	Hence "that's cool"	21:22
	I will definitely keep this in mind next time I need to do that	21:23
jogla	Do third party dependencies, like jar files, belong to the repository	21:23
ToxicFrog	At the moment I'm not doing it so much, since classes have resumed and I'm always either at home or within campus wireless coverage.	21:23
	← dato left	21:23
	← EmilMedve left	21:24
	← Ryback_ left	21:24
	← langenberg left	21:25
	→ Man-erg joined	21:26
	→ jatt joined	21:27
jatt	hi, I'm trying to clone a repository. I'm using sshfs to mount the partition where I want to clone it (the partition is on a remote machine).	21:28
	$ git clone foo	21:28
	shows:	21:28
	Cannot commit config file!	21:28
	fatal: Not a valid object name HEAD	21:28
cbreak	can you clone it to local?	21:28
	and then just copy it over?	21:28
jatt	if I use a local partition the clone works. What can be the problem?	21:28
	cbreak: I can try that sure	21:29
	← advany left	21:29
ShadeHawk	let me guess: challenged filesystem, with case folding (so it uses .git/head instead of .git/HEAD)	21:29
jatt	cbreak: but actually I wanted to use sshfs to avoid the copying.	21:29
cbreak	I have no idea what SSHFS is :)	21:29
jatt	by me	21:29
	manually	21:29
	← Beket left	21:29
ShadeHawk	why sshfs and not simply ssh? no git on other end?	21:29
jatt	ShadeHawk: the remote filesystem is a linux one (I hope).	21:29
cbreak	I think that you are in theory supposed to be able to rsync git repositories without problem	21:30
ShadeHawk	the question is does sshfs do case folding (vfat, anyone?)	21:30
	← etgryphon left	21:30
johnw	cbreak: as long as they aren't being written to at the time, sure	21:30
ShadeHawk	so I guessed wrong...	21:30
jatt	until now I've been using git to synchronize things locally (as a replacement from rsync (with the advantage git gives me history))	21:31
DrNick	sshfs is just a FUSE interface to sftp/scp, so it does whatever the remote end does	21:31
	← etgryphon_ left	21:31
jatt	afs is a linux filesystem, isn't it?	21:31
	← Guest45175 left	21:31
jatt	the remote side is running afs	21:31
	← tritz left	21:31
cbreak	afs is cross platform network file system	21:31
ShadeHawk	Hmmm...	21:31
jatt	hmm probably my student account files are living in a vfat partition then :(	21:32
cbreak	I think it's intended to be used as replacement for NFS	21:32
	← zachinglis left	21:32
jatt	well that's the one used by my university. I didn't expect the git error only because I use a mounted partition through sshfs.	21:33
	very strange.	21:33
cbreak	can you try to create some files?	21:33
jatt	in fact the git clone created the .git file but no files where transfered	21:33
cbreak	"TEST", "test" and then try if they are different	21:33
jatt	cbreak: yes, I tested touch foo; cat foo > foo; sshfs seems to work fine, even the owner is correct	21:34
cbreak	also with FOO?	21:34
	I think the theory is that it does not work because the file system ignores case and does not preserve case	21:34
jatt	$ touch foo FOO; ls -altr foo FOO	21:34
	-rw-r--r-- 1 376403 10000 0 2008-09-18 22:34 FOO	21:34
	-rw-r--r-- 1 376403 10000 0 2008-09-18 22:34 foo	21:35
cbreak	hmm.... weird.	21:35
jatt	looks alright	21:35
	but, what does this	21:35
	Cannot commit config file!	21:35
	message mean.	21:35
	in fact the .git directory was created	21:36
	← eternaleye left	21:36
jatt	but no files where transfered to the remote side	21:36
	no other files	21:36
	maybe is a FUSE bug?	21:38
charon	that message comes from a failed rename() system call	21:38
	so maybe sshfs doesn't like renames	21:38
jatt	hmm that's vfs I guess	21:38
	so probably is the fuse kernel module the cause of the problem?	21:38
ShadeHawk	git uses "create temporary file, then rename to final destination' trick for atomic change in many places	21:39
jatt	rename aka. mv?	21:40
	or something else?	21:40
charon	mv fails with "Operation not permitted". -- Use -o workaround=rename (requires sshfs version >= 1.3).	21:40
ShadeHawk	well, from C it is probably rename(...)	21:40
charon	from the google cache version of the sshfs faq, no idea why the original is dead.	21:40
	← vbabiy left	21:41
ShadeHawk	but some of git are shell scripts, and they use probably 'mv a+ a'	21:41
jatt	you are right... man 2 rename	21:41
charon	mv is rename() unless across fs boundaries iirc	21:41
	← sergiodj left	21:41
	→ sergiodj joined	21:42
charon	that might even make sense	21:42
jatt	charon: so passing -o workaround=rename to sshfs should help to fix my problem?	21:42
charon	no idea, that was just quoting the faq, you'll have to try it yourself	21:42
	← wagle left	21:42
charon	perphaps sshfs decides not to support rename(), instead of half-supporting it via the remote system's 'mv' where it cannot easily guarantee atomicit	21:43
	+y	21:43
	← fanthomas left	21:44
jatt	wow... is working!	21:45
	Checking out files: 3% (29/940)	21:45
	charon: thanks, -o workaround=rename fixed it.	21:46
charon	thank the google cache ;)	21:46
jatt	he :)	21:46
	→ flaguy joined	21:50
jatt	hm. and now after git pull:	21:51
	error: unable to write sha1 filename .git/objects/27/17293ce2bb0ccf3e061c8b5e32b902a80be440: Operation not permitted	21:51
	fatal: failed to write object	21:51
	again, this is only in the partition mounted with sshfs	21:51
	→ capisce_ joined	21:54
	→ jds_ joined	21:55
	← deskin left	21:55
charon	so apparently it doesn't like link & unlink either huh	21:55
	← kbingham left	21:56
charon	in case anyone is interested, http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=318078#10 explains where the problem originates	21:56
Gitbot	[git 318078]: http://tinyurl.com/4w3ozx [blob]	21:56
Mikachu	heh	21:56
jatt	hm. I will need to try nfs or some other distributed filesystem different than FUSE then	21:56
charon	it's an sshfs/sftp problem, not a fuse problem	21:57
cbreak	afs? :)	21:57
jatt	I see	21:57
	← bobmcw left	21:58
	← eternaleye_ left	21:58
	→ eternaleye_ joined	21:58
jatt	hm that's a pretty old bug	21:58
cbreak	doesn't look like a bug	21:59
	more like a lacking feature	21:59
jatt	oh! and cvs update doesn't work with it either. Probably svn won't do.	21:59
	git pull doesn't.	21:59
	→ rmh3093 joined	21:59
	→ loinclot_ joined	22:00
	← SRabbelier left	22:00
	→ jesselucas joined	22:01
	← capisce left	22:02
	→ doener_ joined	22:02
charon	i don't get this. the code should resort to rename() which you seemed to have fixed	22:02
	← Catfish_ left	22:02
charon	unless you _also_ have a write permissions issue. dunno.	22:03
jesselucas	I'm switching from subversion to git and having a bit of trouble wrapping my head around submodules. My current file structure has a directory called Projects with new projects as children. I'd like to make each project it's own submodule and make the parent directoy (Projects) the super project. Of course this doesn't work because submodules can't be direct children (from what I've read). What is the best practice for creating the file structure for multipl	22:04
charon	fae046bd around line 2165	22:04
Gitbot	[git fae046bd]: http://tinyurl.com/3qy6pk [blob]	22:04
jatt	I can create files and it works. In the bug report someone tries	22:04
	$ touch x y; mv -f x y	22:04
	and it fails for him	22:04
	but for me it works	22:04
	so I'm not sure I'm hitting the same lacking feature	22:04
ShadeHawk	jesselucas: did you use svn:externals with Subversion?	22:05
jesselucas	ShadeHawk: No just one big nasty repo. Trying to fix that though	22:06
ShadeHawk	jesselucas: if not, perhaps you don't need submodules, only good split to modules (repositories)	22:06
charon	jatt: can you try 'touch x y; ln x y'?	22:06
	and if that succeeds, 'rm -f x'	22:06
ShadeHawk	jesselucas: but I have't use submodules myself, so I cannot help more than pointing to manpages, User's Manual, Community Book and Git Wiki	22:06
jatt	I see that some files where written (probably by the git clone), e.g.:	22:07
	-r--r--r-- 1 376403 10000 199 2008-06-22 12:47 .git/objects/27/238f0cca02a7b0c010781783ee99a3622fdfa6	22:07
	why .git/objects/27/17293ce2bb0ccf3e061c8b5e32b902a80be440: Operation not permited hm.	22:07
	sure	22:07
	← capisce_ left	22:07
jatt	$ ln x y	22:07
	ln: creating hard link `y': File exists	22:07
charon	oh wow	22:07
	int ret = link(tmpfile, filename);	22:07
	← loincloth left	22:07
charon	but man 2 link says: On success, zero is returned. On error, -1 is returned, and errno is set appropriately.	22:08
jatt	I get the same error in my local partition (not in the sshfs one)	22:08
charon	jatt: you're on newer than 1.5.6-rc3 right?	22:08
	→ loiseau joined	22:08
jogla	Hello, if i understood it right, the commit id is a sha1 hash over the diff	22:09
charon	(file exists is fine)	22:09
jogla	How can you be sure that there are no collisions	22:09
	→ capisce joined	22:09
Eridius	jogla: actually, it's not over the diff	22:09
cbreak	I think it's more complicated	22:09
Eridius	it's a hash of the commit object	22:09
	which includes a reference to the tree object (referenced via the hash)	22:10
jatt	charon: no, I'm using git version 1.5.4.3. In my university is even older git version 1.5.5.1	22:10
charon	hmmm	22:10
jatt	I mean newer	22:10
	ops	22:10
Eridius	and the tree's hash depends on the hashes of every single file reachable from that tree, and the tree structure	22:10
cbreak	it also contains the hash of the previous commit object	22:10
charon	ok i didn't read into that old code	22:10
Eridius	and the author and message and other metadata	22:10
charon	but the newest one is definitely broken	22:10
cbreak	(the commit object of a merge commit has several parent hashes)	22:10
jatt	I see	22:10
	So upgrading won't help for me	22:10
Eridius	in other words, the hash of the commit object is effectively a hash of the entire history and contents of the repo	22:11
cbreak	I think you can get collisions	22:11
	but it's quite unlikely	22:11
Eridius	extremely unlikely	22:11
jatt	so all this problems I'm having are due lack of features (or maybe bugs) in git _and_ sshfs?	22:11
Eridius	you're more likely to have a cosmic ray blow up your hard drive	22:11
cbreak	what problem in git?	22:11
charon	jatt: well it's too late to fix bugs in pre-1.6.0.x anyway, so let's just focus on the newer ones ;)	22:12
jatt	I wonder if using anohter remote filesystem (e.g. nfs) would help.	22:13
	instead of sshfs	22:13
jogla	Eridius: Good to know :) Thank you	22:13
loiseau	Eridius: What would be the effect if there was a collision?	22:13
Eridius	loiseau: git would probably complain about corrupted objects	22:13
	unless the collisions were of two of the exact same object. Then it would probably just get confused	22:13
	but again, you're never ever going to get a collision	22:14
jesselucas	ShadeHawk: Do you know of a tutorial on using git for managing multiple projects?	22:14
loiseau	ah interesting	22:14
	yeah I know ;)	22:14
Eridius	s/exact same object/exact same object type/	22:14
	← doener left	22:14
charon	jatt: if you want to build 1.6.x and try with that, i can cc you the patch i'm going to send	22:15
	→ gbucher joined	22:15
loiseau	Just some kind of academic interest	22:15
jatt	charon: sure I will install and test with 1.6.0.2 now.	22:16
	← jesselucas left	22:19
charon	jatt: http://rafb.net/p/VumbRM50.html	22:20
	save it somewhere and feed it to 'git am'	22:20
	→ ijcd__ joined	22:21
	← danten left	22:21
charon	(running the testsuite, but since it was broken all the time i don't expect anything bad to happen)	22:22
jatt	charon: ok, will do, I'm building git (without the patch) right now	22:22
	← gbucher left	22:23
	→ gbucher joined	22:23
	← gbucher left	22:23
	→ archis joined	22:23
charon	that went through fine... yay for tmpfs	22:24
ThaDon	Hey guys, any clue why when I execute a git clone using ssh, I'm getting: bash: git-upload-pack: command not found. I mean, I login to the server using the same user that git is using and I can execute git-upload-pack no problems	22:26
jatt	charon: without the patch (only with git-1.6.0.2) the git pull worked.	22:26
	makes sense?	22:27
charon	ok, then my logic wasn't all that flawed earlier	22:27
	ThaDon is perplexed	22:27
	← lamont left	22:28
charon	jatt: i think you can forget about the patch, the bug only triggers on EEXIST which means a hash collision or a broken link()	22:28
	so, glad it worked.	22:28
	→ dsyzling joined	22:28
	→ ijcd joined	22:29
	→ jackdempsey joined	22:29
ThaDon	oh... git expects git to be /usr/bin....	22:29
	← ijcd_ left	22:30
jatt	charon: ok, thank you very much for your help :), I will keep using sshfs to synchronize my stuff	22:31
	sshfs+git	22:31
charon	but if i get this right, you still need -oworkaround=rename?	22:31
ThaDon	jatt, yeah I'll have to get sshfs going fosho	22:31
	→ wagle joined	22:32
jatt	charon: yes the partition is still mounted with -workaround=rename, will test without it	22:32
jonkristian	this is driving me crazy, in my webfolder i have setgid to let people in the group work on it, and i have a hook script that pulls updates to that dir, but git reverts the setgid bit to normal, 755/644 , any ideas how i can fix this?	22:32
charon	jatt: thanks. if you want to do everyone a favour, you could write down on http://git.or.cz/gitwiki/GitFaq that sshfs needs -o workaround=rename and a relatively new git	22:33
jatt	sure I will create an entry for this	22:34
cbreak	jonkristian: can you do it in a hook?	22:34
vuf	jonkristian: setup a webfolder user, and have people push through ssh	22:34
	jonkristian: wait, do you have people working directly in the webfolder?	22:35
	← Ademan left	22:36
adamm	jonkristian: maybe the umask is not set right?	22:36
	← boto left	22:36
charon	jatt: great, thanks. i'm off to bed, happy hacking :)	22:37
	← charon left	22:37
	← radarek left	22:37
	→ zjason joined	22:38
jonkristian	vuf: yes, well, git is updating that webfolder directly on a push from each user	22:38
	vuf: each user has umask 0002 in their .zshrc on that directory	22:38
	← dsyzling left	22:39
	→ codeshot joined	22:39
vuf	jonkristian: I think the ssh pushing should work	22:39
	← juanjoc left	22:39
vuf	then all files are created as the webfolder user, and you don't need setgid. personally, I like that better	22:40
	← jackdempsey left	22:40
	← ShadeHawk left	22:40
jonkristian	vuf: im not sure i understand	22:41
vuf	hrm	22:41
	→ lamont joined	22:41
jonkristian	can i get git to pull as a certain user for that dir?	22:41
	ideally, it would be awesome if it could pull files with web user/group since it needs either one	22:42
	people are already pushing through ssh, that's how we've set up git	22:43
	ssh://user@domain.com/path/to/repo/project.git	22:43
vuf	oh	22:43
	← bentob0x left	22:43
vuf	then my proposal is to have a common [email@hidden.address]	22:44
	if you have authorization concerns, that may not work	22:44
jonkristian	ah	22:44
	i could create a git user and let everybody use that, but if there's some other way	22:45
	← BarryCarlyon left	22:45
jonkristian	and i'd really like an overview on who's done what	22:45
vuf	I always hated the setgid since it seems to break every now and then - maybe that is just me	22:46
jonkristian	yeah, i hate that stuff like that doesn't work	22:46
	← ijcd__ left	22:46
vuf	you should have the overview in the author field of commits, though :)	22:46
jonkristian	but having all users in a group should really be enough, i don't know why this happens	22:47
	yeah	22:47
vuf	well, I guess you should be able to fix the permissions in the hook, as cbreak suggested 15 minutes ago :)	22:48
	→ k-s joined	22:49
k-s	guys, what to do when git-rebase fails with whitespace errors?	22:49
	git-rebase --continue fails with "nothing added"	22:49
	no files are modified	22:49
	← fixUp left	22:49
vuf	k-s: that is two questions?	22:49
Mikachu	up to three now	22:50
	k-s: if the rebased commit is no longer needed, you can skip it with git rebase --skip	22:50
	→ madewokherd joined	22:50
k-s	Mikachu: i guess it's still needed	22:50
	← zjason_ left	22:50
k-s	the patch still sits under .dotest/patch	22:50
	should I manually git-apply it?	22:51
vuf	k-s: you should add the changes	22:51
k-s	vuf: yep, but git-apply .dotest/patch ?	22:52
	← thannoy left	22:52
	→ perezd88_ joined	22:52
	← david_koontz left	22:53
	→ ajonat joined	22:53
	→ ciaran joined	22:54
jonkristian	vuf: someone was talking earlier about a group setting in git, is that right?	22:54
vuf	hrm, they were, but I don't know it	22:55
jonkristian	ok	22:55
	← ciaran left	22:56
	→ ciaran joined	22:56
vuf	jonkristian: I have planned that same setup, so I hope you find a solution for me :)	22:56
k-s	ok, git-apply .dotest/patch + git-add files worked, thanks	22:56
jonkristian	--shared	22:57
	git init [-q \| --quiet] [--bare] [--template=<template_directory>] [--shared[=<permissions>]]	22:57
	djwonk → djwonk_afk	22:57
jonkristian	but that can only be set up when first creating a repo, or can i change the one i have	22:57
	← djwonk_afk left	22:58
	← loiseau left	23:00
	→ grumpy joined	23:01
grumpy	lo	23:02
	→ rtomayko_ joined	23:02
	← aroben left	23:04
tongueroo	hey guys	23:04
grumpy	yop	23:04
jonkristian	vuf: git init --shared=group will setgid on the repositories, don't know how it acts on a pull yet though	23:04
tongueroo	how do i delete a remote tag? i just pushed all my tags "git push --tags"	23:04
	and i dont want some of those tags on the remote repos :P	23:05
	← softdrink left	23:06
vuf	jonkristian: great, thanks	23:06
tongueroo	is it similar to the weird syntax for deleting branches :) git push origin :branch_name ?	23:06
jonkristian	vuf: but, do you know if it's possible to apply that to an already existing repo?	23:07
grumpy	Linus was talking about tracking a function even after moving it from one file to another, what command does that?	23:08
tongueroo	:) yup, i guess that does the trick	23:08
vuf	jonkristian: it seems you should set core.sharedRepository(?)	23:09
	← Toranin1 left	23:10
vuf	grumpy: git blame -C -C destinationfile	23:10
	← LiamH left	23:11
	← perezd88 left	23:13
	← ciaran left	23:13
drizzd	how can I fetch the full history with a shallow repository? git fetch --depth=0? It does not remove .git/shallow, however.	23:14
	→ ciaran joined	23:14
sgrimm	I have a bit of corruption in a git repo, a commit whose tree object is missing. But it's a leaf node commit from an old topic branch I don't care about. How do I remove it from the object database? (It is making git repack fail.)	23:18
	← ciaran left	23:18
grumpy	vuf: ???	23:19
	→ ciaran joined	23:19
grumpy	vuf: assuming I have a function named 'foo' which is currently in finalfoo.c, this function change files and also prototypes...	23:19
	vuf: what would be the command line to type?	23:20
	← narflekw left	23:20
	← ciaran left	23:21
	→ ciaran joined	23:21
grumpy	git blame -C -C destfile seems to be showing who, when, changelist#, ... for each line of a file	23:22
jonkristian	vuf: good news, it's working now, i can push changes and they will get pulled in the webdir, and my co-worker can do the same, with --shared=group it seems to preserve setgid even on the webdir	23:30
	so basically i can just add whomever to that group, and they're ready to go	23:30
	← foca_ left	23:32
	← jpeeler left	23:33
	→ softdrink joined	23:34
	→ Beket joined	23:35
	← rtomayko_ left	23:38
	← bronson left	23:39
	← ciaran left	23:39
	→ pcc1 joined	23:39
	→ ciaran joined	23:39
mutex	lets say I have a file in some state that is in the git repository in that state as well	23:40
	but I don't know what commit ID that is	23:40
	is there a way to somehow find the commit id ?	23:40
mugwump	hmm	23:41
	what are you trying to achieve?	23:41
	are you looking for a commit that had a file in state X ?	23:41
mutex	I have a file somebody gave me from a revision I don't have a tag for	23:41
mugwump	ah	23:41
mutex	so I'd like to figure out exactly what commit-ID the file in this state is	23:41
mugwump	ok, so first find the blob id of the file you have	23:41
	git-hash-object -t blob filename	23:41
	then, you can search for that sha1 in the output of 'git-whatchanged --all'	23:42
	or, if you know what the filename is going to be called, 'git-whatchanged --all -- filename'	23:42
mutex	ah, and I can sha1 the file that I currently have	23:42
	← FunkeeMonk left	23:43
mugwump	right. just not with `sha1sum` :)	23:43
mutex	oh...	23:43
	→ Big_Lebowski joined	23:44
pcc1	how can I extract only those commits that exist within a branch but not in the mainline, so that I can back them up? I am working privately on a modification to a large open source project and would rather not back up the whole history	23:44
	→ ciaran_ joined	23:44
mutex	mugwump: well it did narrow everything down to two changes, so that's good ;-)	23:44
mugwump	pcc1: see man git-rev-list. basically, git-rev-list branch1 branch2 --not origin/branch1 origin/branch2	23:45
	← ciaran left	23:45
Gitbot	mugwump: the git-rev-list manpage can be found at http://git.or.cz/man/git-rev-list	23:45
grumpy	does anyone know how to track the changes made to a function "foo" across the history (including file changes...)?	23:46
	← ciaran_ left	23:46
	→ ciaran joined	23:46
	← Big_Lebowski left	23:46
mugwump	grumpy: see git-blame, and git-log -Sfoo	23:46
	you need to set options to git-blame to get it to follow functions moving around files	23:47
	← ereslibre left	23:47
mugwump	and it's not perfect, esp. wrt to changing indents etc	23:47
	→ djanowski joined	23:49
djanowski	hi folks. is there a way to have git ignore "newlines at end of file"?	23:49
	← carrus85 left	23:50
	← archis left	23:51
	← Yuuhi left	23:52
pcc1	mugwump: thanks. how do I get git-rev-list to show the diffs. the man page talks about formqtting options but not about actually showing the diffs	23:54
grumpy	stupid question, if 2 users pushes their changes at the same time, what happened? will one be accepted and the other one rejected?	23:54
mugwump	pcc1: well, git-rev-list options are common to other commands that do that	23:55
	ie, git-log -p	23:55
	or you can use git-show with individual commits	23:55
	grumpy: correct	23:55
grumpy	also, is it possible to do a commit ammend in a remote repository? (I setup my home server with git, and I pushed a commit realizing later that I messed up the comment... To fix it, I had to recreate the whole database...)	23:57
	← lolage0 left	23:57
	← schacon left	23:58
	← krawek left	23:58
	← ciaran left	23:59
	→ ciaran joined	23:59
	← djanowski left	23:59

Logs Search ←Prev date Next date→

Channels Documentation