IRCloggy #git 2008-10-19

Provider of IRC logs since 2005.

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

	2008-10-19
doener	johnw: so I'd drop the "to commits" part there, and just say that they're fancy refs	00:00
jnareb	actually, everything except heads detaches HEAD	00:00
johnw	doener: noted	00:00
offby1	johnw: the diagram on page 11 has commits of different colors, but I can't tell what those colors signify	00:00
doener	offby1: maybe read the pastebin about HEAD that I made earlier ;-)	00:00
	→ Jacolyte joined	00:00
jnareb	I think when talking about tags (and tag objects are more than 'fancy refs', or 'refs in tags namespace) it is important to tell _why_ we need tags (to tag releases of course)	00:00
offby1	doener: huh, ok	00:01
jnareb	see you later	00:01
johnw	that's true, the colors shouldn't be different there	00:01
jnareb	Can someone remove "\| http://tinyurl.com/gitsurvey08" from channel description? TIA	00:01
	← jnareb left	00:01
vuf	hm, which document are you discussing?	00:01
Garoth	Hey, how do I see what commit introduced some file, or possibly which commits have edited some file?	00:01
doener	offby1: that (in some way) explains why branches are advanced and tags are not (checking out a tag causes you to get a detached HEAD)	00:01
	Garoth: git log --diff-filter=A -- file	00:02
johnw	vuf: http://www.newartisans.com/blog_files/git.from.bottom.up.php	00:02
Garoth	doener: thanks, that'll do	00:02
doener	Garoth: --diff-filter=M for modifications, AM for both	00:02
Garoth	Oh, very cool	00:02
doener	see the manpage for the other available flags	00:02
Garoth	Will do.	00:03
vuf	johnw: okay, thanks ... I read that at one point :)	00:03
doener	offby1: paste was here http://git.pastebin.com/m6258402f	00:03
offby1	yeah, I'm reading it, but I don't see how it relates to the colors in his diagram	00:05
	← cryolab left	00:05
doener	offby1: uhm, I meant the "difference between branches and tags" thing	00:05
	offby1: the difference is not that the tag doesn't advance, but that you simply can't checkout the tag, but just the commit it references. And then the "detached HEAD" stuff kicks in	00:06
	offby1: didn't notice the graph question yet when I wrote that	00:07
	← RaceCondition left	00:10
johnw	offby1: you're right that the colors don't add anything, they just represent "unmerged branches", hence yellow+blue=green	00:10
	offby1 <- color-blind	00:12
johnw	ahhh	00:12
offby1	either explain 'em or nix 'em	00:12
johnw	then you missed the colored text that said "This diagram exists to haunt the comprehension of offby1"	00:12
	→ vintrepid joined	00:13
offby1	yeah, it was in pink text which I can't see on a white background :)	00:13
	← \ask-_ left	00:13
	← drobbins left	00:13
	← vintrepid left	00:14
xenoterracide	I do git log filename and it shows me the last log entry about how I moved the file... is there anyway to get all the log history of the file before I moved it?	00:16
Garoth	Heh.	00:17
doener	xenoterracide: --follow	00:17
xenoterracide	thanks	00:17
	seems I resolved this ones...	00:17
	ESC[33m but my git log shows a bunch of those	00:18
doener	probably you have the LESS environment variable set, but without the R flag	00:18
	← reals left	00:18
Garoth	Is it possible to have a remote that is set up to push to one location, but pull from another?	00:19
xenoterracide	so it should be -MiR ? :P	00:19
doener	xenoterracide: if you want that ;-) IIRC FRSX is quite popular, if you care ;-)	00:19
xenoterracide	I should really spend time looking up what that does	00:20
johnw	also tree less -r	00:21
	try	00:21
	i often use -Mr	00:21
doener	yeah, -r if you don't want colors, regardless of the git settings	00:21
	oh wait, -r was "pass through all raw characters"	00:22
	so forget the color thing...	00:22
	← yofel left	00:22
	→ yofel joined	00:22
xenoterracide	ok now one last question... is there any way to pull the entire history of just this one file into it's own repo? I want share it seperately as sort of it's own project	00:22
	→ rdavila joined	00:23
doener	xenoterracide: clone and filter-branch	00:23
Garoth	Magical filter-branch :)	00:23
doener	but it really becomes it's own project then. There won't be common commits between that repo and the original repo anymore	00:23
xenoterracide	that's fine...	00:24
doener	s/it's/its/	00:24
Garoth	Of course, you could then add it as a sub-module, right?	00:24
	Though I think that forces it to have a folder around it.	00:24
	→ kef joined	00:25
xenoterracide	in theory... might be the thing to do... not sure... I store all my dot files in git repos	00:25
	← SRabbelier left	00:25
	→ \ask-_ joined	00:25
xenoterracide	but I've been making changes to some .vim scripts that appear unmaintained and I figure it's better not to propagate my whole dot_usr direoctory to people who want to know what I've done to 1 script	00:26
Garoth	Good idea.	00:26
	Submodules, when I used them, were a bit combersome, but worked well.	00:26
	If your vim scripts have folders, you could totally do that.	00:26
	Of course, you could get _really_ hacky and have them have folders and do symbolic links to them.	00:27
	Anyway, I could be wrong about the folders thing, but I'm pretty sure.	00:28
	→ nkallen joined	00:28
	← Gitzilla left	00:28
xenoterracide	how do you think I make the repo work in the first case?	00:28
Garoth	Hmm?	00:28
	Lots of sym links?	00:28
xenoterracide	I'm sure not going to edit /etc/bashrc to look for .bashrc in .etc/bashrc	00:28
	← nkallen left	00:28
	→ Gitzilla joined	00:28
xenoterracide	so I symlink it	00:28
	→ nkallen joined	00:28
Garoth	Well, idk, you could have put your home dir under git control	00:29
	And ignored all the non dot files	00:29
	← kumbayo left	00:29
Garoth	But I think your way is cleaner.	00:29
	← nkallen left	00:29
xenoterracide	yeah think I tried that... it was a pita	00:29
Garoth	I once put my home dir under git control... <3	00:30
	It worked great, though I had to commit it in chunks due to memory constraints in with git add	00:30
xenoterracide	the problem is all those dot files I don't want to track	00:30
Garoth	Or maybe git commit.	00:30
	xenoterracide: more entries in .gitignore :)	00:30
xenoterracide	yeah... like I said... pita	00:30
	I did it	00:30
	but I like the .etc and .usr better	00:31
Garoth	Yeah, it's pretty nifty.	00:31
	Though... well...	00:31
	Garoth thinks about it	00:31
Garoth	ls \| grep -v "files you don't want ignored" >> .gitignore	00:31
	→ toofishes joined	00:32
xenoterracide	I don't recall all the issues I had to be honest	00:32
	repressed memories	00:32
Garoth	:}	00:32
xenoterracide	and too many files I want to ignore	00:32
Garoth	Well, that's not really a problem, but I'm sure you had some legitimate problems	00:33
xenoterracide	well the problem is figureing out what I want to ignore	00:33
	it's not a technical issue it's a time/patience one	00:33
Garoth	"everything except these few files," right?	00:33
	It's the same as selecting a few files and moving them into an .etc	00:33
xenoterracide	in a sense	00:34
Garoth	Anyway, I'm arguing with you about something I shouldn't be arguing with you about ;)	00:34
	→ alb joined	00:34
xenoterracide	lol	00:34
Garoth	gotta be careful, or I'll become one of those annoying trolls that always tells you to do things better	00:34
	← Yuuhi left	00:35
Garoth	Oh god, IRC client for Emacs.	00:35
	Garoth shudders	00:35
	← tvw left	00:36
xenoterracide	heh	00:36
	xenoterracide tries to wrap head around filter-branch	00:39
Garoth	Emacs, like, totally breaks the Unix Way.	00:39
j_engelh	Emacs is not Unix.	00:40
Garoth	Which, I guess, is why it does that ;)	00:40
	← JeFeDe left	00:40
Garoth	But what I'm referring to is the "do one thing and do it well" guideline that I'm rather fond of	00:40
xenoterracide	emacs seeks to replace unix ;)	00:41
	it IS the GNU operating system	00:41
Garoth	You know, rather than "Emacs is my OS and Linux is my device driver"	00:41
	Yeah.	00:41
	← Eridius left	00:41
xenoterracide	emacs should only be run on the HURD :P	00:41
Garoth	GNU has done some good stuff, though.	00:42
xenoterracide	I hear duke nukem forever runs on the hurd and is written in perl 6	00:42
Garoth	They're a little misguided, but I'm happy with them overall.	00:42
	;)	00:42
	Haha, Perl 6.	00:42
	For the sake of sanity, I pray that Perl 6 happens and is good.	00:42
	I have to write a lot of hardcore Perl5 code for work.	00:42
	It does a lot of things really awkwardly.	00:43
	But, well, most of those problems are solved with "moar evals"	00:43
	Garoth feels uncomfortable not talking about git	00:43
xenoterracide	so does anyone have any idea how I would remove all the stuff but this one file and it's history? (note file was moved)	00:44
	Garoth shuts up.	00:44
toofishes	xenoterracide: so you want only one file	00:44
xenoterracide	because filter-branch is confusing my brain	00:44
	yep	00:44
	just the one	00:44
	→ maestrolinux joined	00:45
maestrolinux	http://s2.ar.bitefight.org/c.php?uid=19732 hello --!!! hola alguien me ayuda con esto	00:45
xenoterracide	http://github.com/xenoterracide/dot_usr/tree/master/vim/ftplugin/sql/sql_iabbr.vim	00:45
Garoth	... is that spam?	00:45
xenoterracide	that file :P	00:45
	kinda looks like it but you could redirect him to a spanish channel	00:45
Garoth	God, if there were spambots for IRC, I'd be up in arms.	00:46
xenoterracide	um....	00:46
Garoth	With all the admin possibilities at my disposal, and all the code I could think of.	00:47
toofishes	xenoterracide: filter-branch definitley takes work	00:47
	but it is easy to keep retrying until you get it right...	00:47
xenoterracide	heh	00:47
	← maestrolinux left	00:47
	← lirit left	00:47
xenoterracide	well --subdirectory would work... except wouldn't that cause problems because I've moved the file?	00:48
	on the other hand... I haven't had any commits since the move... so why do I care about that	00:48
	← albertito left	00:49
	→ Eridius joined	00:49
	← johnw left	00:53
	← jmatthews-out left	00:54
	→ jmatthews-out joined	00:54
	← pcapriotti left	00:55
	← reallyidle left	00:55
	→ reallyidle joined	00:55
xenoterracide	%$#@!	00:57
Garoth	xenoterracide: ? :)	00:57
	→ albertito joined	00:57
xenoterracide	it's annoying me	00:57
	ok... so what I want to do to hopefully make my life easier is check out the checkout before the move...	00:58
	then rewrite it to make that directory the root	00:58
	that should work	00:58
	problem is when I check it out it tells me I'm on a remote branch	00:59
	so how do I make that remote branch the local master?	01:00
Garoth	git branch -M?	01:01
offby1	check it out!	01:01
	offby1 dances to the Tom-Tom club	01:01
offby1	git checkout -b master remotes/whatever	01:01
	's what I'd do	01:01
Garoth	I'm not entirely sure how you can be on a remote branch.	01:01
xenoterracide	me either	01:01
Garoth	I'm guessing it's the local version of a remote branch.	01:01
	In which case, you can just handle it like any local branch.	01:02
	and git branch -M it (for move)	01:02
	and offby1's answer will make sure that it's a local copy of a remote branch	01:02
	But... as far as I know, you can't _be_ on a remote branch.	01:02
	But, well, with git... it has so many options, I can never be sure.	01:03
xenoterracide	well maybe not so much as remote as not in a branch I guess	01:05
Garoth	are you on "(no branch)"?	01:05
xenoterracide	yeah	01:05
Garoth	I think you can still move that, but I haven't tried it.	01:05
	(git branch -M)	01:06
xenoterracide	tried git branch -M master and that doesn't work	01:06
	error: refname refs/heads/HEAD not found	01:06
	→ paltman joined	01:06
Garoth	Why not?	01:06
	Oh, I see.	01:06
	xenoterracide: and you got here by checking out some file or whatever?	01:07
	Do what offby1 said -- instead of whatever you did, use git checkout -b <branchname>	01:07
xenoterracide	I checked out the last commit before I moved that file	01:07
Garoth	That'll make a new branch out of whatever, and switch into that rather than (no branch)	01:07
	^^	01:07
	git checkout -b <branch> <commit-hash>	01:08
	if I recall correctly, that works great	01:08
xenoterracide	yeah got that far... and then I got the rewrite right	01:08
	but when I tried to move it back to master it screwed up	01:08
Garoth	say what?	01:08
	You made a branch, changed it, saved your changes, and then couldn't make it master	01:08
	← p0w3r3d left	01:09
	← paltman left	01:09
xenoterracide	at least I think I did... hmm... maybe I'm missing a step	01:09
	→ alb_ joined	01:10
xenoterracide	it seems it rewrote master not the branch I'm in	01:11
	wtf	01:11
	Garoth hasn't used filter-branch before	01:11
Garoth	Guess it's the destructive sort because after a filter-branch, the history now has little in common with the parent project	01:12
	Which I guess, is why you were told to clone first.	01:12
	But I don't really know.	01:12
toofishes	so who's knee-deep in git internals here? i know right now the only place pthreads are used is in pack-objects	01:13
xenoterracide	yeah... I figured that	01:13
toofishes	and i was curious if there was any other places that might give a big speedup	01:13
Garoth	Lol, a big speedup for git :D	01:13
	→ G_SabinoMullane joined	01:13
xenoterracide	now I'm getting this... Namespace refs/original/ not empty	01:14
	how do I fix that?	01:14
	it's preventing this...	01:14
	git filter-branch --subdirectory-filter vim/syntax/sql/ -- --all	01:14
	which should do what I want	01:14
Garoth	Well, again, no idea, but you could try deleting everthing in refs/original/ :P	01:14
	← marze left	01:15
Garoth	Mind you, I'm not the right person to be trusting here ;)	01:15
xenoterracide	well that's not right...	01:16
	wtf	01:16
	somehow it took me back so the only history was of the file being added	01:16
	← alb left	01:17
toofishes	Garoth: when that pthreads code went in a year ago, it was a 200% speedup when using multiple cores	01:18
	so i'd say it was worth it	01:18
	*when using 2 cores	01:18
Garoth	I'm just having a chuckle because Git's already so damn fast. I don't really know anything about anything, as usual :)	01:19
	→ alb joined	01:20
	← lolage0 left	01:21
xenoterracide	I've seen times where it's not	01:22
	at least on certain functions	01:22
	I added an image once that really screwed up the performance of git gc	01:22
doener	xenoterracide: that might have been caused by unreferenced objects as well. It unpacks these, which isn't _that_ fast, and there was a memory leak in that code path, which could get quite bad (fix will be in 1.6.0.3 it seems)	01:24
	← Zimm left	01:24
	→ paltman joined	01:27
Garoth	So, is there a way to set up a remote to pull from one place and push to another?	01:28
xenoterracide	how do I see more info about which files changed in git log?	01:28
Garoth	What constitutes "more info"?	01:29
	← albertito left	01:29
xenoterracide	I just need to know which files were actually changed... in addition to what it shows now	01:29
Garoth	git log --stat	01:29
doener	--name-status	01:29
	--name-only, ...	01:30
Garoth	doener's ways are better, but, yay mine works too :)	01:30
xenoterracide	heh	01:30
doener	just giving additional possibilities ;-) maybe he even likes --stat better ;-)	01:30
xenoterracide	I was trying --verbose... and that doesn't work :P	01:30
Garoth	I think git's killer feature is --color :D	01:30
	git log --stat --color	01:31
	very pretty	01:31
doener	try "git log -p --color-words" ;-)	01:31
Garoth	How... strange	01:32
	Oh, i get it!	01:32
	Wow, that's crazy	01:32
doener	sometimes it's really useful, but sometimes it just sucks	01:32
Garoth	doener: 0_0	01:32
	doener: can git generate one of those side-by-side diffs?	01:33
	If you know what I mean.	01:34
j_engelh	yeah	01:34
doener	not on its own (or I just don't know how), but there's eg. a script that feeds stuff to vimdiff. And I think someone had a similar script for meld	01:34
	(or some other tool, I actually don't even know if meld does side-by-side diffs hides)	01:35
Garoth	Yes, meld does.	01:36
	← rmh3093 left	01:37
Garoth	I personally don't care, but I had a guy complain to me that git doesn't do that for its web interface	01:37
	(whereas svn does)	01:37
	← parasti left	01:37
	→ rmh3093 joined	01:37
j_engelh	no one needed it so far :p	01:39
xenoterracide	vimdiff ftw	01:39
doener	I usually prefer unified diffs, side-by-side diffs just take up so much horizontal space and force more eye movement	01:40
xenoterracide	he could write it himself :P	01:40
	← alb_ left	01:40
xenoterracide	ok so I think... that I've properly filtered the repository... how should I check? and the logs don't appear correct so what do I use to fix that?	01:41
	→ naoshige joined	01:44
naoshige	anybody annoyed by vim's .swp files?	01:45
xenoterracide	lol	01:45
naoshige	can i just add *.swp to .gitignore?	01:45
xenoterracide	I think that works	01:45
naoshige	word	01:45
	thanks man	01:45
Garoth	Yes, I add that all the time.	01:46
	xenoterracide is very annoyed	01:51
xenoterracide	I'm thinking of just copying the text of the log and say screw the history	01:53
	it would be easier than filter-branch	01:53
	→ jhelwig joined	01:54
	→ bwwww joined	01:54
Garoth	Well, yes, of course :)	01:55
	→ vintrepid joined	01:56
xenoterracide	doesn't help that I'm dead tired now	01:57
	although if you ask me this documentation could be better and the command could be much more user friendly	01:58
	← jhelwig left	01:58
	← rmh3093 left	02:00
	→ deskin joined	02:00
jast	naoshige, you can configure vim to store swapfiles in a dedicated directory	02:00
	('directory' setting)	02:01
naoshige	well	02:01
	nah	02:01
	or maybe some day	02:02
	i dunno, maybe tomorrow	02:02
	→ gimmulf joined	02:05
	→ selckin joined	02:06
	→ rmh3093 joined	02:06
	← G_SabinoMullane left	02:06
	← gimmulf left	02:12
	← rmh3093 left	02:12
	→ rmh3093 joined	02:13
	← naoshige left	02:15
	← rmh3093 left	02:16
	→ rmh3093 joined	02:17
	→ neurodamage joined	02:18
neurodamage	so I was looking to make a branch again :)	02:18
	but take an exisiting branch and then upsh the new changes for that branch?	02:19
	→ albertito joined	02:20
	→ Arafangion joined	02:21
Garoth	neurodamage: please explain.	02:22
Arafangion	Looks like git-p4 doesn't use the p4 libraries at all, preferring instead to use the command line applications. (For both git, and p4) - which explains the nature of some of the issues it has.	02:23
	→ p0w3r3d joined	02:26
	← alb left	02:35
	← jmrodri-gone left	02:36
	→ alb joined	02:36
	← limux left	02:36
	→ FunkeeMonk joined	02:38
	← deskin left	02:43
toofishes	so please tell me i'm not crazy, and i've always found the git-format-patch rev specifiers odd, but how can i turn a single commit (that isn't the head revision) into a formatted patch?	02:45
	e.g. i know the sha1 sum	02:45
	is this really the easiest way? $ git format-patch 0344fc303204~1..0344fc303204	02:46
doener	git format-patch 0344fc3^!	02:47
	c^! means c but none of its parents	02:50
	← eMBee left	02:50
	→ alb_ joined	02:50
	→ eMBee joined	02:50
	→ cedricv joined	02:50
	← albertito left	02:55
	→ albertito joined	02:56
	← alb left	02:57
toofishes	doener: well that makes sense.	02:57
	buried pretty deep in the rev-parse manpage, thats for sure.	02:58
doener	format-patch either takes a starting point or a range. So "format-patch foo" == "format-patch foo..HEAD", that's probably where most confusion about it comes from	02:58
toofishes	yeah, that wasn't the part i didn'tk now	02:59
	i just had no idea how to easily give a single-commit range	02:59
	← elliottcable left	02:59
	→ jackdempsey joined	02:59
toofishes	1 of 2 things would make sense here- either add some sort of '--single' option to format-patch (seems too much), or just note the '^!' syntax in the format-patch manpage	03:00
	you think the second idea makes sense?	03:00
	← elmex left	03:00
	← eternaleye left	03:01
	→ elmex joined	03:02
	← madewokherd left	03:02
	→ madewokherd joined	03:02
	→ Marmouset joined	03:12
	← rdavila left	03:14
	← alb_ left	03:15
	→ alb joined	03:22
	← jackdempsey left	03:27
	→ jackdempsey joined	03:29
	→ brendon joined	03:30
	→ deskin joined	03:33
	→ stouset joined	03:35
	← stouset left	03:36
	← albertito left	03:39
	← deskin left	03:39
	← vintrepid left	03:40
	← bwwww left	03:41
	← brendon left	03:43
	→ deskin joined	03:43
	→ jesselucas_ joined	03:47
	← jackdempsey left	03:47
	← jesselucas_ left	03:48
	→ elliottcable joined	03:53
	→ mankind_tweezer joined	03:54
	← ereslibre left	03:55
	→ EiNZTEiN joined	03:57
	← deskin left	04:01
	→ jesselucas_ joined	04:05
	← EiNZTEiN left	04:06
	→ paltman_ joined	04:06
	→ jackdempsey joined	04:08
	→ tjafk2 joined	04:10
MaddHatter	I think I've used C~1..C to specify a single-commit range before	04:10
j_engelh	C^..C is shorter!	04:10
	(and C^! even more)	04:10
	← paltman left	04:13
MaddHatter	I avoid ^ as much as possible because I use git sometimes on Winows	04:15
j_engelh	and? ^ is really inert on windows	04:15
offby1	it's especially pointy there	04:16
	→ EiNZTEiN joined	04:16
MaddHatter	^ is to Windows what \ is to *nix	04:16
j_engelh	orly	04:16
	no wai	04:16
MaddHatter	so you end up doing stuff like C^^^^ to get C^	04:16
j_engelh	why if you use bash it will be fine	04:17
	→ clairvy joined	04:22
	→ cedric_ joined	04:25
	← tjafk1 left	04:26
	→ paltman joined	04:31
	→ tk81 joined	04:31
jesselucas_	if I'm trying to setup a post-update hook to push to a mirrored remote. Would it just be exec git push NAME_OF_MIRROR_REPO ?	04:32
tk81	hey all, i was wondering if revisions in git are treated more like SVN or CVS?	04:32
	is it by file, or by directory?	04:32
	err.. by repository*	04:32
j_engelh	tk81: neither.	04:32
jesselucas_	I believe it's by content	04:32
tk81	what is by content?	04:33
j_engelh	well, what's "by file"/"by directory"?	04:33
	← cedricv left	04:34
MaddHatter	the commit id ("revision" if you like) is the sha1 of the commit itself	04:34
	← kukks left	04:34
MaddHatter	git is not centralized like CVS and SVN, so you can't use a monotonically increasing serial number	04:34
tk81	maddhatter: ok I understand. Im assuming though its easy to see the id's in a descending order in the master copy though, correct?	04:35
MaddHatter	I think you mean like "git log" or "git whatchanged" so try those	04:36
offby1	define "descending" :)	04:36
Arafangion	MaddHatter: There was discussion on the mailing list about using one anyway, iff you centralise it, utilize various hooks, and otherwise restrict it, etc.	04:36
offby1	seriously: I'm pretty sure you can sort 'em by date	04:36
tk81	well by date in respect to when changes were made in the master copy. but i bet thats what git log or git whatechanged does	04:37
MaddHatter	ah. I sorta gave up on the mailing list. I just don't have enough time to read it all. But I'm surprised people like Linus wouldn't be strenously opposed to that idea.	04:37
	← ned left	04:38
MaddHatter	tk81: git doesn't have a "master copy"	04:38
	← paltman_ left	04:38
tk81	eh. origin i think its called	04:38
MaddHatter	but you can see the changes in order for any repository	04:38
	origin is just where you cloned from	04:38
tk81	first origin ?	04:38
MaddHatter	which could be a "central" repo, a coworker, etc.	04:38
tk81	Dad?	04:38
	central. got it	04:38
Arafangion	MaddHatter: The issue is that having such numbering is actually a checkbox requirement in some environments.	04:39
MaddHatter	It would only be central because you made it that way though, not because git did	04:39
offby1	silliness	04:39
	those environments deserve to fail	04:39
tk81	yeah i understand that	04:39
	thanks guys	04:39
MaddHatter	You could use tags + a hook to achieve that functionality	04:41
tk81	One more question. When you have, lets say, customer apps and company apps, would you be better off splitting that into two repositories or one repository?	04:41
	I am asking this because I would imagine you need to clone the entire repository in order to check something out on a customer server, giviing them the company data as well	04:42
MaddHatter	is one repository per app not possible?	04:42
	← rajiv left	04:42
tk81	it is possible since i havent started anything	04:42
	so yes	04:42
	ill do that	04:42
MaddHatter	that's sort of a shot in the dark. Naturally what's best in your environment depends on lots of details we don't know.	04:43
Arafangion	tk81: For git, I'd use one repo per application/library, unless you want to track movement of code from one project to another.	04:43
tk81	and does git have some kind of 3rd party support for nicely colored diff emails?	04:43
Arafangion	tk81: ie, suppose you have three distinct repositories, for projects A, B, and libC, using separate repos doesn't really let you track movement of code from B into libC.	04:44
	tk81: But arguably having separate repos entirely would help clarify the project boundaries.	04:44
	← madewokherd left	04:44
tk81	Arafangion: I see what you are saying, and thats usaully why I like to use one repository, but the other features such as the branching and merging beat SVN in entirety so ill think ill have to make that sacrafice	04:45
	is their a way to export history in some way to move code into another repository?	04:45
Arafangion	tk81: I'm still very new to git, but I've noticed that it's much, much easier to create new repos compared to, eg, subversion.	04:46
	(Especially if you use svn-dav)	04:46
tk81	nope, never got that working nicely so i stuck with svn+ssh. but thats off topic	04:47
Arafangion	tk81: I think you could make each repo contain remote tracking branches to the other repos, but I'm not sure.	04:48
tk81	hmm.. i guess you cant lock files in git because of its distrubted nature	04:48
Arafangion	Why would you want to lock files, anyway?	04:49
MaddHatter	why would you need to lock files in git?	04:49
tk81	but i guess you dont need to either because of its distrubuted nature.	04:49
Arafangion	Why would you even want to lock files in any other RCS system.	04:49
tk81	eh. i never used it.. just knew it was there	04:49
j_engelh	see	04:49
MaddHatter	If you're storing, e.g. Word documents, in the repository, it can be nice to serialize the changes	04:49
	because it's so hard to merge them after-the-fact	04:49
j_engelh	well word is stupid	04:50
	use something, eh well, xml/tex based	04:50
MaddHatter	it's the same for any other binary type of file	04:50
j_engelh	well it is serialized actually.	04:50
tk81	MaddHatter: is a serialized process flow supported in git?	04:50
Arafangion	So, you'd see that there is a conflict, so you send of an email to the conflictor to ask them what the heck they're doing?	04:50
j_engelh	no, you'd just resolve it	04:50
tk81	is there*	04:51
	← bobmcw left	04:51
MaddHatter	the idea would be to avoid doing a bunch of work (before you've already done the work) that will result in a conflict	04:51
j_engelh	the idea is not to keep the file open needlessy!	04:51
	because that's lost productivity for every one else...	04:52
	← jackdempsey left	04:52
MaddHatter	it's lost productivity if you lock it; it's lost productivity if you don't	04:52
	toss a coin: which is worse	04:52
	tk81: no. But I don't think it's supported in any other VCS either (except RCS and CVS)	04:53
Arafangion	Files can be locked in VSS, P4, SVN...	04:54
MaddHatter	didn't realize svn supported that	04:55
Arafangion	It's a new feature, and allows people to force the lock open.	04:55
	I don't think anyone uses it, though.	04:58
offby1	you'd be surprised	05:04
	I hear svn is big in Japan^W^W^W big in "corporate environments"	05:04
	Arafangion is surprised.	05:04
Arafangion	I thought perforce would've been preferred over svn.	05:05
MaddHatter	that doesn't surprise me... corporate environments are scared of stuff that's new (i.e. dvcs) and want centralization for backups and such	05:05
tk81	isnt SVN big in corporate environments here? the git shift is the new thing	05:05
MaddHatter	and perforce costs money	05:05
j_engelh	perforce is also ridiculously fragile	05:05
MaddHatter	In my world, if you're gonna pay for a VCS, Clearcase seems to be the popular choice	05:06
j_engelh	the db needs to be rebuild every weekend to keep it in shape	05:06
Arafangion	j_engelh: I haven't found that.	05:06
j_engelh	well maybe you did not do as many commits	05:06
Arafangion	j_engelh: We're on commit 101000+ now, in one of the repos.	05:07
	→ bobesponja joined	05:07
j_engelh	i'm talking about 80 million.	05:07
Arafangion	Nice. :)	05:07
j_engelh	no, it really sucks	05:07
Arafangion	Yeah, that's what I meant (text really is low-bandwidth)	05:08
j_engelh	it's like it's battling SVN in being the slower one	05:08
Arafangion	From what I've seen it's damn fast, but maybe the performance does slow down as you get to high numbers of repos (and branches)	05:08
j_engelh	well, everything that is network-based, and P4 is like that, is slow by definition	05:10
Arafangion	p4 does like to have a very powerful server.	05:10
tk81	when you push a repository, im assuming it still only sends diff's correct?	05:10
Arafangion	tk81: Perforce, or git?	05:11
tk81	git	05:11
	hence this is git ;)	05:11
Arafangion	tk81: It only sends compressed changesets, I believe. :)	05:12
j_engelh	git sends packs	05:13
tk81	compressed? nice	05:13
j_engelh	compressed and deltified	05:13
	the server only really needs to copy the received pack to disk and update the branch pointer	05:14
	→ eno__ joined	05:20
	← eno left	05:21
cedric_	j_engelh: deltified against what?	05:33
	→ jackdempsey joined	05:33
j_engelh	itself	05:33
cedric_	uh?	05:33
Arafangion	cedric_: The previous version it applies against.	05:33
j_engelh	no no	05:34
cedric_	Arafangion: that would be the plain changeset then	05:34
j_engelh	{the first tree is raw, further commits are deltified.} <- and the whole thing is compressed	05:34
cedric_	so I wonder what the changeset is deltified ageinst?	05:34
j_engelh	a single changeset is not	05:34
	the pack which contains the changesets uses deltas	05:34
cedric_	j_engelh: to rebuild the commits in the changeset? right	05:35
j_engelh	rebuild huh?	05:35
	commit=changeset, in case you use the BK term "changeset"	05:35
	git only knows commits and packs..	05:35
	← jackdempsey left	05:37
cedric_	hmm yeah, so if i understand correctly it does: <full diff between a and c>, then deltaitifation for a->b b->c, then conpression?	05:37
j_engelh	i do not think git uses diffs	05:38
	→ tonguero_ joined	05:38
cedric_	when i say diff, i mean the changes between a and c, whatever the format used is	05:39
toofishes	cedric_: the changes are deltified against any other object available in the repository	05:39
	it really doesn't matter which object that is, whether it be one from the previous commit, or one from 5 years previous	05:39
j_engelh	usually this is all hidden behind the pack mechanism. period :P	05:39
toofishes	and then the "tree" or list of files that make up that commit is just another thing that can be deltified, and it would tend to be recorded as a delta from the previous tree.	05:40
j_engelh	well with normal packs, deltas are within a pack only	05:40
	only thin packs may have deltas that delta against an object from another pack	05:40
toofishes	j_engelh: baby steps here :)	05:40
cedric_	toofishes: interesting :)	05:41
	← statim left	05:42
	→ statim joined	05:42
	← litage left	05:45
cedric_	on a completely different subject, do you know how i can change tab width displayed by git diff/log -p ? i've found no related option in git config, nor google	05:45
j_engelh	cedric_: huh? that would be a feature of your pager	05:46
	or even your terminal	05:46
	← tk81 left	05:46
cedric_	oops, yay, so simple i did not think about it :p	05:49
	← tongueroo left	05:53
	← Jacolyte left	05:55
	→ Jacolyte joined	05:56
	→ timcharper joined	06:00
	← bgerber left	06:02
	→ shweazel joined	06:03
	← paltman left	06:03
	→ bgerber joined	06:04
	← bgerber left	06:04
	← shweazel left	06:06
	← kef left	06:09
	→ kef joined	06:11
	→ malahal_ joined	06:16
Arafangion	Hey, is it possible to have multiple working directories sharing a single .git repo?	06:17
	→ rubydiamond joined	06:17
MaddHatter	yes... see there's a script for that in contrib, I think	06:21
Arafangion	Nice.	06:23
	→ GNUix joined	06:23
	← malahal left	06:32
	← rubydiamond left	06:41
Arafangion	Actually, would that mean that one could use a git repo without a local .git?	06:43
	→ rubydiamond joined	06:45
MaddHatter	well, the .git has to be accessible somehow	06:46
	and I hear weirdness happens if you check out the same branch in both working dirs	06:46
Arafangion	How so?	06:47
	Wierd.	06:47
	That doesn't make sense, I thought that what would happen is that git would get confused /which/ user is accessing it.	06:47
	Alternatively, how would one setup a git repo such that the 'master server' is /not/ cloned, but rather uses only the head revision.	06:48
MaddHatter	git doesn't care about users and I don't understand "uses only the head revision"	06:49
Arafangion	Well, where does git store user.name and user.email?	06:49
MaddHatter	under ~ somewhere usually	06:50
Arafangion	Nice.	06:50
thiago_home	use git-new-workdir	06:50
Arafangion	'new is not a command'. Where is it?	06:51
	→ tk81 joined	06:52
MaddHatter	Arafangion: that's the thing under contrib I mentioned earlier	06:52
Arafangion	MaddHatter: By "only the head revision", I mean only the most recent repository.	06:52
	only the most recent version, I meant.	06:53
thiago_home	git new-workdir	06:53
tk81	would a xen virutal machine image work well inside a repository?	06:53
Arafangion	tk81: You're kidding, right?	06:53
tk81	or is it not going to diff well?	06:53
thiago_home	tk81: Git doesn't interfere or help Xen.	06:53
	but versioning a file that is bound to be gigabytes in size... not a good idea	06:53
MaddHatter	you can set up a commit hook to reset --hard HEAD so there's always an up-to-date working dir on the server if you wanted	06:54
Arafangion	MaddHatter: Say, a repo is 1TB, but a particular version of a particular branch is only 45GB if you just grab all of it's files. How would one just check that out, do work against it, and then commit changes back to the server.	06:54
tk81	whats the best way to version an entire server	06:55
	rsync?	06:55
thiago_home	no	06:55
	you'd use git on the particular files you want to version	06:55
	you don't want to version /var/log, for instance	06:55
tk81	ok thats doable	06:55
MaddHatter	Arafangion: you can't do a "shallow" checkout without any history and then commit from it	06:55
tk81	thx	06:56
MaddHatter	or even "shallow" as in only one branch, I think	06:56
Arafangion	MaddHatter: So, the only possible way would be to have a fast-import tool that does it, sending commits to a staging server that would then make the real commit. Unrealistically complex.	06:56
	← selckin left	06:56
MaddHatter	I think a 1TB repository is already unrealistically complex	06:57
Arafangion	MaddHatter: I'm just having a discussion with someone who works in game development, with a 2TB repo.	06:57
thiago_home	and I think my 900 MB is big	06:58
Arafangion	MaddHatter: Which to me does scream "Badly organised repo", but anyway. A large amount of that would be binary graphics and sounds, I would imagine.	06:58
	thiago_home: At work, that's already a single branch in my checkout. ;)	06:58
MaddHatter	No VCS is going to do "well" with binary files, especially large ones	06:58
thiago_home	the checkout is 300 MB	06:59
	the problem isn't the nature of binary	06:59
	Git will do well with binaries	06:59
	the problem is whether they are compressible or not	06:59
Arafangion	MaddHatter: Indeed, but VCS's other than git would at least allow you to only check out the branch you're interested in.	06:59
thiago_home	Arafangion: with Git you can check out only the branch you want	06:59
	and clone/pull only that branch too	06:59
	→ alreves joined	06:59
MaddHatter	thiago_home: no, the problem with binaries is that $vcs doesn't know what changed	06:59
	so a merge is impossible	07:00
Arafangion	thiago_home: Unlike git, you'd /also/ be able to commit changes, though.	07:00
thiago_home	MaddHatter: Git doesn't care about what changed, even in text files	07:00
	Arafangion: and you can commit changes too	07:00
Arafangion	thiago_home: How?	07:00
thiago_home	MaddHatter: well, you don't want to do a merge in binaries.	07:00
	Arafangion: the normal, usual way.	07:00
	Arafangion: nothing special.	07:00
Arafangion	thiago_home: Ahh, the way that requires a full clone of the repo?	07:00
MaddHatter	not being able to merge eliminates most of the usefulness of a vcs	07:01
thiago_home	Arafangion: no, you can git init, then fetch/pull the branch you want.	07:01
MaddHatter	this is why people hate CVS and to a lesser extent, SVN. The merging sucks.	07:01
thiago_home	MaddHatter: you can merge. But you need a tool to do merging.	07:01
	MaddHatter: Git and other VCS provide one for text files.	07:01
	MaddHatter: if you want to do it for other types of files, use another tool.	07:01
Arafangion	perforce (apparently) provides ones for images as well.	07:01
MaddHatter	so I guess I should say the problem with binaries is that $vcs doesn't provide a merge tool	07:02
Arafangion	thiago_home: And fetch/pull doesn't pull all the changes in the branches's history?	07:02
thiago_home	Arafangion: it does	07:02
Arafangion	thiago_home: Suppose for some reason you had .iso's versioned in the repo. (Yes, this is an EXAMPLE), and you had 100 versions in the repo in a single branch.	07:03
	thiago_home: How would you grab just hte latest ISO, and commit a new ISO as a change?	07:03
MaddHatter	the very nature of what you're asking requires a centralized VCS	07:04
	which git is not	07:04
Arafangion	MaddHatter: I realise that now, but thiago_home hasn't realised the nature of the question I'm asking.	07:05
MaddHatter	it is theoretically possible to commit from a no-history checkout when the commit would be a fast-forward on the remote, but I don't believe git offers any support for that	07:05
	but a merge is not possible without knowing history	07:05
thiago_home	Arafangion: you wouldn't	07:05
	you can use --depth when cloning, but that's for all branches	07:06
	so you want a one-branch repository for which you can clone with --depth=1	07:06
Arafangion	And that wouldn't allow commits, then, would it?	07:06
thiago_home	it would	07:06
	I don't know about pushes	07:06
Arafangion	MaddHatter: Why couldn't it provide a merge? (Obviously, it wouldn't provide the most efficient delta, though).	07:07
thiago_home	you can use bundle for pushing, though	07:07
MaddHatter	if you only have two endpoints, you don't know what was changed along each pathway	07:08
	and so you don't know what was changed by one side or the other	07:08
Arafangion	So this is an algorithmic issue?	07:09
MaddHatter	as I understand it, yes	07:09
	→ vintrepid joined	07:09
MaddHatter	thiago_home: are you saying you can pull a branch out of a repository into a new repo, clone it with --depth=1, commit onto the new repo, then rejoin the new repo as a branch back into the old?	07:10
thiago_home	MaddHatter: I've just checked	07:10
	you can't push from a shallow clone	07:11
	but you may be able to use bundle	07:11
MaddHatter	interesting	07:11
	← tk81 left	07:14
	→ cedricv joined	07:16
	← cedric_ left	07:19
	← jm left	07:21
	→ cko joined	07:21
	← Jacolyte left	07:23
Vortex35	git fsck reports several dangling commits and blobs. Is this critical? I suppose not because i've seen it regularly. But where does it come from?	07:23
Arafangion	Vortex35: Do you use rebase?	07:25
Vortex35	Arafangion: yes, sometimes!	07:25
thiago_home	or amend?	07:25
	if so, it's normal by-product of the use of the repository	07:26
Vortex35	thiago: i use amend also sometimes.	07:26
	thiago_home: ok, so it can be safely ignored?	07:26
	→ priidu joined	07:26
thiago_home	yes	07:32
Ilari	Vortex35: Dangling blobs tend to pop up very often. Dangling commits are sightly less common (but still quite common). And ocassionally one sees dangling trees or tags...	07:33
	Vortex35: Easiest way to get dangling blob is to stage the same file twice (with some modifications in between) in the same commit cycle...	07:34
	← dramsay left	07:35
Ilari	Vortex35: 'Dangling object' means object that has no other object or ref pointing into it...	07:36
	→ Weaselweb joined	07:38
doener	Vortex35: if you use "stash" and pop entries or clear the stash, that's probably the fastest way to get dangling commits	07:39
jaalto	How do I cancel commit SHA1? I removed some files in SHA, and now I want the commit "undone".	07:40
doener	the other "common" way requires reflog entries to expire and that takes quite a while (90 days?). Though if you regularly rebase/amend, you'll regularly see dangling commits either way	07:40
	→ snitko joined	07:40
Ilari	jaalto: 'git revert <SHA1>' (record new commit which undoes effects of specified commit)?	07:41
jaalto	Ilari: thanks	07:41
doener	that adds a revert commit on top though (which is the right thing if you already published the original commit)	07:42
	other possibilities include commit --amend and reset	07:42
	quite frankly, your question made no sense to me :-/	07:42
Ilari	jaalto: If you haven't pushed that commit yet, and there are no merges after it, you could also try 'git rebase -i <SHA>^'.	07:42
	→ Weasel[DK] joined	07:44
	→ bgerber joined	07:53
	← bobesponja left	07:54
	← timcharper left	07:55
	→ advany joined	08:00
	→ ph^ joined	08:00
	← robin left	08:08
	→ robin joined	08:08
	← clairvy left	08:08
	→ tk8 joined	08:18
	← tk8 left	08:19
	→ owen1 joined	08:21
	← snitko left	08:22
owen1	i started working with git. i commit localy. how do i create a public repository that will serve as backup and allow second dev to clone it?	08:23
thiago_home	clone	08:24
doener	either "git clone --bare" or "git init --bare". Then you can start to push stuff to the repo	08:24
owen1	thiago_home: doener: don't i need to put my .git folder on a remote server?	08:25
	← CodeOfficer left	08:25
thiago_home	no	08:26
doener	owen1: you just create a second, bare repo. That just consists of the .git directory	08:26
	→ pcapriotti joined	08:26
doener	well, the contents of it	08:26
	→ kumbayo joined	08:26
doener	with the "clone" approach, it will be, well, a clone, so there's already content in it. With "init" it will be empty and you need to push there to fill it	08:27
	"clone" is fine if the public box can connect to your local box and you want it to have all you local branches	08:28
	if you can't get a connection in that direction, or you want only some of your branches to be in the public repo (eg. just master, but none of your WIP topic branches), you'll need to init the bare repo and then push whatever you want	08:29
owen1	doener: i'll try the init approach first. i assume i can also try it on my local machine instead of remote server.	08:29
doener	sure	08:29
	← Weaselweb left	08:30
	← priidu left	08:30
	→ Fullmoon joined	08:31
	→ priidu joined	08:31
owen1	doener: git init --bare	08:33
	doener: usage: git-init [-q \| --quiet] [--template=<template-directory>] [--shared]	08:33
drizzd	git --bare init	08:33
doener	which git version?	08:33
owen1	doener: 1.5.4.3	08:33
	doener: (ubuntu 8.04)	08:34
doener	ah, ok, the "init --bare" version only exists since 1.5.6-rc1. So do as drizzd said	08:34
	→ deb08 joined	08:40
owen1	doener: git push ~/Desktop/git-test-server/ and i get this: No refs in common and none specified; doing nothing.	08:41
deb08	hi, I'm trying to merge without commiting but I don't know how to do that	08:41
drizzd	git merge --no-commit maybe?	08:42
deb08	drizzd: I found that in the doc, but in the last version (which I have) the option is not available	08:42
doener	owen1: without a refspec, push update only branches that have the same name in both repos	08:42
	owen1: your bare repo doesn't have any branches yet, so there are no matches	08:43
	owen1: try "git push <path> <branchname>"	08:43
drizzd	deb08: the latest version is 1.6.0.2	08:43
doener	owen1: or if that fails, "git push <path> refs/heads/<branchname>:refs/heads/<branchname>", not sure when the "just a branchname" DWIM feature was added to push	08:44
owen1	doener: looks good	08:44
deb08	drizzd: indeed	08:44
owen1	doener: now i am going to test pulling form my 'server'	08:44
doener	owen1: maybe start with trying to clone ;-)	08:45
drizzd	actually git merge --no-commit has been in git since version 1.0	08:45
doener	owen1: there won't be anything to fetch/pull yet, as it's all in your "local" repo as well	08:45
	← Fullmoon left	08:45
	← Marmouset left	08:45
owen1	doener: ok, i'll try to clone..	08:46
deb08	drizzd: http://paste.debian.net/19545/	08:47
doener	deb08: well, that seems more like a CLI quick help inaccuracy	08:47
	deb08: did you actually just try to use --no-commit?	08:47
deb08	doener: yes of course. I did git merge --no-commit mybranch	08:48
	and then did git log	08:48
	and all the commits of the other branch were in my master, commited	08:48
doener	then it was a fast-forward	08:48
deb08	yes	08:48
	it displayed fast forward	08:48
doener	there simply is nothing to commit in that case	08:48
deb08	oO	08:48
doener	fast-forward means that the branch head is simply updated to reference a newer commit	08:49
drizzd	there is --no-ff though	08:49
owen1	doener: cool. worked. so first i clone, and than pull (a la check out)?	08:49
doener	say you have this history: A-B-C-D-E-F. "master" is at C, "foo" is at F	08:49
drizzd	not sure how that's useful though	08:50
deb08	I don't really understant (I'm new to git) but what I needed, is to merge the branch into my master, no commit, because I need to do merge diff to see the difference AND THEN commit if all is ok	08:50
doener	then to integrate the changes from foo into master, you can simply say that master now points at F	08:50
thiago_home	deb08: you can look at the diff after the commit	08:50
doener	deb08: you could just do "git diff master...foo" beforehand	08:50
	thiago_home: fast-forward case	08:51
thiago_home	yep	08:51
	→ RaceCondition joined	08:51
thiago_home	git diff ORIG_HEAD	08:51
drizzd	git log -p foo should be more useful than the plain diff anyways	08:51
deb08	doener: this is exactly my case	08:51
	so I do the diff before, then merge.	08:52
doener	thiago_home: does merge set that? Anyway, I just tripped over your "after the commit", as there is no commit happening ;-)	08:52
	→ charon joined	08:52
doener	owen1: pull is fetch+merge, has nothing to do with checkout	08:52
deb08	there is no way to have the merge into my master without commit, as if I edited the files by hand ?	08:52
	it's the --no-ff ?	08:52
thiago_home	doener: it sets that, yes	08:53
drizzd	--no-ff --no-commit might do it	08:53
thiago_home	"after the marge"	08:53
doener	deb08: --no-ff will avoid the fast-forward and instead create a (in general) pretty useless merge commit	08:53
	← arelius left	08:53
deb08	ok	08:53
thiago_home	deb08: squash	08:53
drizzd	but a git diff then will show exactly the same as the commands already suggested	08:53
deb08	yeah	08:53
owen1	doener: so what is checkout in the git world?	08:54
thiago_home	what are you trying to do anyways?	08:54
	→ d0k joined	08:54
deb08	I'm switching from svn to git, that's why I need those extra (unecessary) steps :)	08:54
	I'll try what you said guys, thanks for the tips	08:54
	→ gimmulf joined	08:54
thiago_home	forget your svn bad habits	08:55
	what is your goal? Your objective?	08:55
doener	owen1: with checkout you, well, checkout a branch (or a commit)	08:55
reallyidle	owen1: checking out a branch is, in essence, switching to that branch, it does _not_ involve any network	08:56
	reallyidle → idletask	08:56
doener	owen1: "pull" is fetch+merge. It fetches new stuff from a repo and merges it into your current branch.	08:56
	owen1: so for "pull" to do anything, you need some new commits in the repo from which you pull, ie. some that aren't in your current branch already	08:57
	owen1: you can eg. add a new commit in your original repo, push that to the bare one, and then pull in your new clone	08:57
deb08	owen1: drizzd: git merge --no-ff --no-commit did exactly what I need, thanks :)	08:58
	← gimmulf left	08:59
owen1	doener: got it. so pull is similar to checkout + merge in svn terminology.	09:00
Ilari	deb08: Except that it does likely unwanted stuff to history...	09:00
deb08	Ilari: what do you mean ?	09:01
doener	owen1: no, "svn checkout" is like "git clone". pull is more like "svn update"	09:02
Ilari	deb08: It records next commit to have old master as direct parent (which is very likely unwanted).	09:03
idletask	And git clone will "svn checkout" ALL branches	09:03
deb08	Ilari: I understand	09:03
	→ arelius joined	09:04
	→ Beket joined	09:04
Ilari	deb08: There should not be redundant parent links. That is, dropping any parent link should always cause commit past-future-elsewhere relationships to change.	09:05
deb08	ok	09:06
Ilari	deb08: The --no-ff was introduced mainly for use with git-svn...	09:06
deb08	hehe	09:06
	← arelius left	09:07
	→ gimmulf joined	09:07
thiago_home	but you shouldn't use it	09:07
Ilari	deb08: Now, maybe examine the diff with 'git diff HEAD' and if it looks OK, do 'git reset <other branch>'. Or something like that.	09:08
deb08	I'll try	09:09
Ilari	deb08: Usually if you want to inspect the changes, use something like 'git log -p master..foo' or 'git diff master...foo' (yes, one has two dots, the other three) and if it looks OK, try the merge.	09:12
deb08	I see, I check BEFORE the merge and then do the merge	09:12
	← gimmulf left	09:12
	→ gimmulf joined	09:13
deb08	why 3 dots ?	09:13
Ilari	deb08: Compare base of master and foo to foo.	09:14
deb08	ok	09:14
Ilari	master..foo would be "compare master to foo".	09:14
deb08	ok	09:14
	thanks	09:14
drizzd	the two coincide in the ff case	09:14
	→ DrFrasierCrane joined	09:14
doener	"diff master..foo" would show the differences betwene master and foo as they currently are (thus including the removal of changes that happened only on master)	09:15
	"diff master...foo" shows a diff of the changes that happened on foo but not on master	09:15
deb08	ok, so two dots are good for a global view	09:17
Ilari	deb08: Actually, when examining changes some branch does, two dots are better for commands like log, but for diff three dots are better. Diff is special as it can only compare two revisions.	09:19
doener	"cross branch" the two dot version is rarely useful. It's good for getting, for example, a diff between version 1.0 and 1.1 of a project or so	09:19
deb08	yeah	09:21
Ilari	deb08: 'git log -p master...foo' would also be valid. It shows all changesets that are only on one of master and foo (but not both).	09:22
deb08	I get it	09:23
Ilari	Yes, git merge really sets ORIG_HEAD if it moves current branch pointer.	09:26
deb08	what is the difference between orig_head and head ?	09:27
	head is the last commit and orig_head is ?	09:28
Ilari	deb08: All the information git really stores about the branch is pointer to latest commit on branch (and likely past vaules of this pointer).	09:28
	deb08: HEAD (all caps!) is last commit on current branch, and ORIG_HEAD is previous value of HEAD. Nowadays, HEAD@{1} would likely be better way to refer to last value of HEAD before this one.	09:29
	deb08: See 'git reflog'.	09:29
deb08	ok I'll take a look at reflog	09:29
	thanks	09:29
	→ pantsman joined	09:29
Ilari	deb08: ORIG_HEAD is very old thing. HEAD@{1} is much newer...	09:30
deb08	ok, HEAD@{1} = last commit HEAD@{2} = commit -1, etc ?	09:31
drizzd	I have to admit that I still don't understand the rationale behind the two dot/three dot notation	09:31
	← cko left	09:31
drizzd	two dots mean difference	09:31
	three dots mean symmetric difference	09:31
	for revisions	09:31
	but for diff it's the other way around, IMHO	09:31
	→ bentob0x joined	09:33
Ilari	drizzd: Its mostly random effect of rewriting those dot notations in terms of positive and negative commit notation...	09:33
drizzd	I'm happy to accept that there is no rationale behind it.	09:34
	But I remember a post by linus saying that it made sense in some way.	09:34
	← deb08 left	09:34
drizzd	Ilari: how would I rewrite three dots? two dots I think is a..b == b not a	09:35
	→ ereslibre joined	09:35
drizzd	a...b == a b not merge-base(a, b) ?	09:35
Ilari	drizzd: Yup.	09:35
drizzd	okay, but for diff it's a..b == a b	09:36
	a...b == merge-base(a, b) b	09:36
thiago_home	git diff takes exactly two branches	09:36
	(if you give it less than two, it infers the missing ones)	09:36
	so it has technically abused the three-dot notation for the very-useful case of "what are the changes made to that branch that aren't in this"	09:37
	→ alexross joined	09:37
drizzd	I think for diff, the current two-dot notation is redundant, and the three dots should be two dots	09:37
	thiago_home uses the two-dot notation often	09:38
Ilari	Earliest mention of ORIG_HEAD I can find from git history is from April 2005 (and first commit is from the same month)...	09:38
thiago_home	to find out if there's anything in other branches that need merging	09:38
drizzd	thiago_home: but it's the same as having no dots	09:38
thiago_home	no, it's not	09:38
	git diff otherbranch..	09:38
	it's not the same as git diff otherbranch	09:38
	sorry, dots in the wrong side	09:39
	git diff ..otherbranch	09:39
	← ToxicFrog left	09:39
drizzd	that's equivalent	09:40
	except for the fact that + and - are reversed	09:40
thiago_home	yep	09:40
	note that there's no way in the two-branch notation to say "the index" or "the working tree"	09:40
drizzd	well, it's probably too late to change it anyways	09:40
thiago_home	if there are no changes in the index and in the working tree, then git diff ..otherbranch is the same as git diff HEAD otherbranch	09:41
	← advany left	09:41
drizzd	thiago_home: I still think that the notation is confusing because it clearly means something different than in the revisions case	09:41
thiago_home	but it is different from revisions anyways	09:42
drizzd	it would even make more sense to exchange the meaning of the two notations	09:42
thiago_home	you're getting a diff of many changes together	09:42
	not a diff of individual changes	09:42
drizzd	sure, but	09:42
	the two dots in the revisions case is very similar to the three dots in the diff case	09:43
	and vice versa	09:43
	→ Zimm joined	09:43
	← rubydiamond left	09:43
thiago_home	I don't agree	09:43
	well, but that's because I know the technical reasons	09:43
	I agree that behaviour-wise, it is similar	09:43
drizzd	rev a..b == b not merge-base(a, b) [almost], diff a...b == merge-base(a, b) b	09:45
Ilari	Digging the history, HEAD@{1} notation appeared in v1.4.0, when the ORIG_HEAD was alredy in v0.99.	09:46
drizzd	but diff a..b == a b, which is completely unrelated	09:46
	and what are the technical reasons? rev-list shouldn't even be invoked in the diff a..b case	09:47
	so it's parsed differently anyways	09:47
Ilari	drizzd: The notation parsing was done using rev-parse...	09:47
	← gimmulf left	09:48
drizzd	so git diff a..b actually results in git diff b ^a ?	09:49
thiago_home	git diff b ^a shows nothing here	09:49
jaalto	The "git push --tags" pushes all tags, but is is possible to push only specific tags?	09:49
drizzd	it does work here	09:49
thiago_home	jaalto: yes, name them	09:50
jaalto	thiago_home: Ah, tag push is like branch push?	09:50
drizzd	with sha1s, at least	09:50
thiago_home	weird	09:50
	$ git diff $(git rev-parse HEAD~..HEAD) \| wc -l	09:51
	43	09:51
	$ git diff 6802a1dbf227d5e3835cfcc0f3017344455b4c05 ^651ebe808e5d750cf3d7a5fad17770c29d4db861 \| wc -l	09:51
	0	09:51
drizzd	I tried with current git.git master and next	09:51
thiago_home	oh, shell escaping	09:52
	$ git diff 6802a1dbf227d5e3835cfcc0f3017344455b4c05 \^651ebe808e5d750cf3d7a5fad17770c29d4db861 \| wc -l	09:52
	43	09:52
xenoterracide	how can I use filter-branch to remove a commit by reference (sha1)?	09:53
thiago_home	xenoterracide: skip the committing of it	09:54
	→ _graham_ joined	09:55
doener	if there are no merges following that commit and it appears on only one branch, "git rebase --onto <commit>^ <commit>" is probably easier	09:55
drizzd	if there are not too many branches based on it and it's not too old, maybe you can also use rebase	09:55
thiago_home	note that there's a difference in behaviour between those two options	09:56
	you're probably looking for rebase's behaviour	09:56
	filter-branch will remove the commit, but not its effects	09:56
xenoterracide	no some of these commits are older than some of the history I need. I'm basically trying to strip the repo down to one file...	09:56
thiago_home	that means the next commit will probably include the changes	09:56
xenoterracide	and it's history	09:56
	I haven't had much luck	09:57
thiago_home	use --directory-filter	09:57
idletask	Argh	09:59
	The git-cvsserver won't allow the CVS clients to branch and tag :(	09:59
xenoterracide	you mean --subdirectory-filter?	09:59
thiago_home	yes	10:01
	sorry	10:01
drizzd	idletask: you have git and you still want to use CVS to branch?	10:02
xenoterracide	I did unfortunately other stuff in that directory... but	10:02
	← ereslibre left	10:02
idletask	drizzd: that's not for me, that's for my work... And the devs use CVS at work	10:02
drizzd	I just cannot imagine anyone using CVS _with branching_ and not wanting to switch to git (or anything else that actually supports branching)	10:03
idletask	drizzd: oh yes you can	10:04
	Spell "transition"	10:04
	I do want them to use git, eventually	10:05
thiago_home	good	10:05
	let them use CVS for now, without branches	10:05
	once they need branches, they switch to Git	10:05
	trust me, switching is easy	10:06
idletask	thiago_home, the problem is, they create branches with CVS right now (they're Eclipse users, heh)	10:06
	I want the transition to be smooth	10:06
thiago_home	you get a hectic first week, but then things settle down	10:06
idletask	We cannot afford that :/	10:06
thiago_home	we could	10:07
	how big is your development team?	10:07
idletask	5 people, plus me	10:07
	I already use CVS to git to CVS	10:07
thiago_home	we switched 50 people	10:08
	→ not-xjjk joined	10:08
xenoterracide	so I'm thinking I want to do something like... git filter-branch --commit-filter skip_commit <sha1> but I'm not sure that syntax is right	10:08
thiago_home	the transition from Perforce to Git was much better than the transition of CVS to Perforce 12 years ago	10:08
	when the company was the size of yours	10:09
drizzd	idletask: I think trying to _use_ both at the same time makes things more complicated	10:09
	setting up a git mirror to play with is a different matter	10:10
thiago_home	xenoterracide: you did read what I said about removing a commit not changing the contents, right?	10:10
	← Weasel[DK] left	10:10
drizzd	but eventually I think you should make a hard switch	10:10
xenoterracide	maybe I missed that sorry...	10:10
	→ Sho_ joined	10:12
	→ ereslibre joined	10:13
idletask	drizzd: well, I'm looking at the source right now, I'll see if I can make up a patch	10:15
	Who knows	10:15
	Also, since we have 52 modules, I wonder whether I should set up a superproject...	10:15
xenoterracide	well then here's the problem if I use filter-branch to remove the files the commits are still in the log... (and maybe more survives?)	10:17
Arafangion	thiago_home: How did you do the transition?	10:20
	← xjjk left	10:20
	← MarcWeber left	10:24
	→ Juice10 joined	10:27
	→ JensB joined	10:28
JensB	Hi everybody	10:28
	← Thumper_ left	10:29
MadCoder	vmiklos: you need to check that argc == 0 if --stdin was specified	10:31
	though like junio said, there is no need for it to be at the end anymore	10:31
	that's all	10:31
JensB	I have just started using Git for my Ruby on Rails project and for testing, I created a branch (thing_test) and within this branch, I pulled a plugin from a different Git repo which gets treated as a submodule. Now when I change to a different branch, the plugin dir is still there. -> Is this intentional? I would have expected the submodule only to exist in the branch it was added.	10:31
MadCoder	git submodules UI still needs polishing	10:32
JensB	hm	10:32
MadCoder	so what happens is not what you would like to happen	10:32
	but it's what is coded for now	10:33
	you must do a lot of things by hand with submodules for now	10:33
JensB	ok	10:33
	so basically the easiest thing would probably be to remove the ".git" in the subdirectories and have them treated like normal files for the time being?	10:34
Ilari	JensB: Existence of submodule repository doesn't imply existence of submodule reference...	10:34
JensB	I don't really need to stay current for the plugins	10:34
MadCoder	JensB: submodule are a PITA if you switch between branches that have them and have them not	10:35
	else it's bareable	10:35
	(IOW you need to git-submodule update way to often to my taste, but it works)	10:35
	*too often	10:35
JensB	Ilari: when I do something like git init; (add files); then mkdir -p vendor/plugins; cd vendor/plugins; git clone $url; doesn't this create a submodule?	10:37
MadCoder	no	10:37
	a submodule is git submodule add $url	10:38
	a submodule is git submodule add $url vendore/plugins	10:38
JensB	ok	10:38
MadCoder	actually	10:38
	and when you commit the super module, it remembers the exact sha1 at which vencor/plugins is	10:38
	← owen1 left	10:38
MadCoder	it's not the same as svn:externals or recursive CVS update is	10:38
JensB	strange (ok, I'm a total greenhorn here ... my bad)	10:38
MadCoder	np	10:38
JensB	the problem is that when I 'git checkout master' the vendor/plugins// stays there - I don't want that	10:39
MadCoder	submodule is meant to track a specific version of the submodule, so that when you commit, you push something that is meant to work	10:39
	yes, that's still a problem with submodules as well	10:39
	like I said before, submodules are rough on the edges right now	10:39
	← alexross left	10:39
MadCoder	what it does wrt versionning is exactly what it should, and it's perfect in that sense	10:40
	the UI to deal with it is a bit cumbersome	10:40
	← CIA-15 left	10:40
JensB	so can/should I delete those files while working on the master branch? or ignore them? 'git diff' doesn't find them	10:40
MadCoder	yes you should delete them, though the issue is that if you had work in progress or not pushed stuff, you'll lose them	10:41
	that's why git checkout doesn't remove them atm, because no proper solution has been decided for that	10:41
JensB	ah ok	10:42
	→ alexross joined	10:43
JensB	work is only lost if it was in the submodules, right? I don't push them back anyway	10:44
	→ ToxicFrog joined	10:46
	Arafangion reckons that having submodules are a bad solution to crap build systems.	10:46
	← charon left	10:47
MadCoder	JensB: yes	10:48
	→ woeye joined	10:48
	→ CIA-15 joined	10:49
ereslibre	is gitweb part of git ? (aka. can I ask about gitweb here ?)	10:49
Pieter	madcoder: did you think about packaging gitg for debian?	10:49
MadCoder	what is gitg ?	10:49
	ereslibre: yes	10:49
	and you can also ask any question related to git stuff you know	10:49
ereslibre	do you guys know where can I find extra css styles for gitweb ?	10:49
MadCoder	though on non core-git stuff you may have no answer that's all	10:49
Pieter	MadCoder: http://github.com/jessevdk/gitg/tree/master a gitk-clone for gtk	10:49
MadCoder	I don't know if that exists	10:49
ereslibre	MadCoder: hehe, ok, thanks =)	10:50
MadCoder	ereslibre: you may use google and have a look to gitwebs that are customized	10:50
	and steal theirs ;)	10:50
ereslibre	haha =) ok, thanks	10:50
MadCoder	but I know of no css repositories for gitweb	10:50
	ereslibre: is it fast ?	10:50
ereslibre	MadCoder: well... reasonably	10:50
	i have it on my own pc for my own projects...	10:50
MadCoder	err I meant Pieter actually	10:50
ereslibre	ah	10:50
	hehe	10:50
MadCoder	I know about gitweb	10:51
	I was asking about gitg	10:51
Pieter	MadCoder: I think it was made because giggle is so slow	10:51
ereslibre	yep... ;) thanks for the info anyway ! ;)	10:51
MadCoder	Pieter: hmm but it's a gitx clone not gitk	10:51
	and I dislike gitx a lot, it's only pretty, not functional ot me :P	10:51
Pieter	MadCoder: ok, well I just wanted to show it to you :)	10:52
vmiklos	MadCoder: ah, thanks.	10:52
	← woeye left	10:54
bentob0x	there was a qgit update over the week-end (Kubuntu KDE 4), where can I find what got updated?	10:57
thiago_home	Arafangion: we were using git-p4 initially	10:58
Arafangion	thiago_home: I note the 'were'.	10:58
thiago_home	Arafangion: then, on Sep 26, we turned perforce read-only	10:58
	everyone had to install Git and clone from the new server	10:58
Arafangion	Ahh, cool. :)	10:59
	thiago_home: How big is your perforce repo?	10:59
thiago_home	320k submits	10:59
Arafangion	With many branches?	11:00
thiago_home	yes, lots of them	11:00
Arafangion	How'd you manage those?	11:00
thiago_home	they were all imported	11:00
Arafangion	I'm guessing, from your success, that you must've converted the repo from the beginning.	11:01
	Rather than at commit number 320k?	11:01
thiago_home	yes, we converted all the history	11:01
	→ Sigma joined	11:01
Arafangion	And you left all the perforce branches in the single git branch?	11:02
thiago_home	no	11:02
	each perforce branch is a git branch	11:02
	but most of them are old branches no longer used, so they are in a repository people don't usually clone	11:02
Arafangion	Oh? I thought that git-p4 didn't really support branches?	11:03
thiago_home	it does, just fine	11:03
	← ToxicFrog left	11:03
Arafangion	Nice.	11:03
thiago_home	git-p4 is like git-svn: it allows you to interact with a p4 repository	11:04
	but more than half of our developers had not even tried it before they had to switch to git	11:04
Arafangion	thiago_home: Yeah, I've been using it, but for various reasons, it didn't really work for me. I had ot modify it to get it to work at all.	11:05
Pieter	thiago_home: didn't they find that hard?	11:05
Arafangion	I didn't import the entire repo, mind you - just the head, and that was essentially the biggest issue.	11:05
thiago_home	Pieter: we made it very simple for them	11:06
	git clone at the beginning	11:06
	work is: git pull; work; git add file; git commit; git push	11:06
	if it fails to push, git pull again	11:06
	Arafangion thinks he should give git-p4 another go.	11:07
thiago_home	rebasing is considered an advanced feature	11:08
	we taught them git diff, git diff --cached, etc.	11:08
	and our Gitorious main repositories are single-branch repos	11:08
	← CIA-15 left	11:09
	← vintrepid left	11:11
	bcarlyon\|laptop → BarryCarlyon	11:11
	→ JeFeDe joined	11:14
	tmz_ → tmz	11:16
	→ CIA-15 joined	11:17
	→ imexil joined	11:17
	→ JensB_ joined	11:19
	→ AreliusFreeNode joined	11:19
	→ pgokeeffe_ joined	11:20
Arafangion	thiago_home: Did you have purged files in the perforce repo?	11:23
	← robin left	11:23
	→ robin joined	11:23
imexil	I found a patch on the mailing list http://is.gd/4lUg that would allow git svn to set autoprops on upstream svn repos. Does anybody know if this is going to be included in git svn in the near future?	11:23
	→ Yuuhi joined	11:25
	← kef left	11:27
	← JensB left	11:32
xenoterracide	the top entry in git log would be ~1? or...	11:34
Ilari	xenoterracide: The top entry is one of postive refs specified...	11:35
	xenoterracide: Unless reverse output is in effect...	11:35
xenoterracide	huh	11:35
	→ zachinglis joined	11:35
Ilari	xenoterracide: So in 'git log', the top entry would be HEAD.	11:36
xenoterracide	somehow this isn't really helping me figure out what I'm trying to do...	11:38
	which is to remove the 5th commit down from history...	11:40
MaddHatter	have you published your history?	11:41
	→ dramsay joined	11:42
vmiklos	Pieter: am i right about you are a bzr fast-import contibutor? :)	11:42
xenoterracide	MaddHatter: irrelevant because I'm making a new repository. basically taking 1 file and making a new project	11:43
	← p0w3r3d left	11:43
Vortex35	Ilari: thanks for the info about dangling objects!	11:43
	← bgerber left	11:43
Arafangion	Given that people seem to think that git-p4 works well, I have decided that perhaps there's something idiotic in how I'm using it, so I tried to use it again on my linux machine (At work I'm using vista), and use the perforce public repo as my test project, however I can't get git-p4 to even import that. :(	11:43
	http://rafb.net/p/nhyCdg13.html	11:44
tokkee	xenoterracide: git rebase --onto HEAD~6 HEAD~5 HEAD should do it.	11:44
MaddHatter	I think that means no, in which case do what tokkee said	11:45
Pieter	vmiklos: yes	11:45
MaddHatter	or git rebase -i works too	11:45
vmiklos	Pieter: is there a mailing list for it, to ask questions?	11:45
MaddHatter	^git rebase -i HEAD~6	11:45
Pieter	vmiklos: I think your best shot is the bazaar mailinglist	11:46
	vmiklos: or the #bazaar channel	11:46
vmiklos	https://lists.ubuntu.com/mailman/listinfo/bazaar this one?	11:47
Pieter	yes	11:47
vmiklos	ok, thx.	11:47
	← Beket left	11:48
	← robin left	11:49
	→ robin_ joined	11:49
xenoterracide	ok... think I got it... does rebasing like that actually remove everything about those commits from the repo?	11:50
MaddHatter	they will disappear eventually	11:50
	in the meantime, they hang around in the reflog	11:50
xenoterracide	should I clean that up before publishing?	11:51
	→ parasti joined	11:51
MaddHatter	no need	11:52
	the reflog isn't published, and unreachable commits aren't cloned	11:52
xenoterracide	ok	11:53
	ty	11:53
	← zachinglis left	11:53
	Arafangion found the issue he had with the jam repo.	12:00
Arafangion	Nevermind.	12:00
thiago_home	Arafangion: no	12:01
Arafangion	thiago_home: The other complication I have is that there is very liberal use of spaces in directory names in the repo.	12:03
thiago_home	that should work without issues	12:05
Arafangion	Should, except that git-p4 doesn't use git or perforce libraries, but instead runs the respective binaries, and parses the results in a simplistic fashion.	12:05
thiago_home	well, for starters, there aren't any libraries	12:06
Arafangion	Yes, there are.	12:06
thiago_home	second, it uses the python interface of p4, so spaces shouldn't be an issue	12:06
Arafangion	Are you sure?	12:06
thiago_home	and git-fast-import doesn't have space issues either	12:06
Arafangion	I'm looking at the source for git-p4, it definetly uses popen, it does not use any p4 library.	12:07
thiago_home	because there aren't any	12:08
	but it uses the Python interface	12:08
Arafangion	There are two python libraries.	12:09
	There's also a C++ library, all available from perforce.	12:09
thiago_home	look at p4CmdList	12:09
	what's the first argument it passes to p4?	12:09
Arafangion	-G	12:10
	→ clairvy joined	12:10
thiago_home	and what does that do?	12:10
Arafangion	That explains why there was so little parsing code.	12:12
thiago_home	yep :-)	12:12
Arafangion	return read_pipe("git name-rev HEAD").split(" ")[1].strip()	12:12
	I'm not properly investigating the code yet, but wouldn't that incorrectly determine the branch name?	12:13
thiago_home	why would it?	12:13
Arafangion	What if a branch had a space in it?	12:14
	→ zachinglis joined	12:14
thiago_home	that would be a bad idea	12:15
Arafangion	Some branches in perforce have spaces in it. :(	12:15
	→ Beket joined	12:15
thiago_home	Git won't like that	12:15
	not that it isn't allowed, but it is a bad idea	12:15
	← mithro left	12:17
Arafangion	I'll take another look into getting git-p4, and maybe I'll start contributing patches. I gotta start somewhere. :)	12:18
	Thanks. :)	12:19
	→ deskin joined	12:21
	→ jmd joined	12:22
	→ malahal__ joined	12:23
jmd	How can I make my working directory use a different repo?	12:23
	→ ben_h joined	12:24
thiago_home	set GIT_DIR and GIT_WORK_TREE	12:24
Pieter	or use the script in contrib	12:25
	→ alreves_ joined	12:25
	← alreves left	12:26
	← clairvy left	12:31
	← malahal_ left	12:36
	← CelticSoul left	12:39
	→ ToxicFrog joined	12:39
	→ d0k_ joined	12:45
	← imexil left	12:46
	← d0k left	12:53
	→ jm joined	12:54
	← pantsman left	12:54
	← d0k_ left	12:57
	← bentob0x left	12:58
	→ cedric_ joined	13:07
	← Arafangion left	13:08
	← zachinglis left	13:09
Pieter	http://pastie.org/295693	13:14
	hmm	13:14
	← JeFeDe left	13:16
	→ CelticSoul joined	13:18
tango_	Pieter, maybe hardlinks to the local repo cause that effect?	13:20
Pieter	tango_: there are no hardlinks	13:20
	that's why I used the file:// protocol	13:20
tango_	when you clone from a local repo you don't get hardlinks?	13:20
Pieter	not when you use the file:// protocol	13:22
	← cedricv left	13:25
	← priidu left	13:25
	← BarryCarlyon left	13:26
	→ bcarlyon\|laptop joined	13:27
	bcarlyon\|laptop → BarryCarlyon	13:27
	→ abbe joined	13:29
	→ nessundorma joined	13:36
	← jmd left	13:39
	← Sho_ left	13:45
	→ snitko joined	13:46
	→ advany joined	13:49
	→ priidu joined	13:50
	xenoterracide → xenoterracide_	13:52
	xenoterracide_ → xenoterracide	13:58
	alb → albertito	14:01
	→ rubydiamond joined	14:05
	← jesselucas_ left	14:08
	← Beket left	14:11
	→ jackdempsey joined	14:11
	← jackdempsey left	14:12
	→ prophile joined	14:15
	← zoke left	14:15
	→ vinnl joined	14:19
	← Juice10 left	14:20
vinnl	Hi, I'm having a little trouble rewriting history with Git and I couldn't find out how to do this with the Git manual, so I hope somebody will be able to help me here...	14:23
	My problem is this: there's a file (an image) in my history that has incorrect license information that I thus cannot publish online. For this reason, I'd like to edit the respective commit to set the correct license before I push this repo online. However, the file is also edited in later commits. How could I go about doing this?	14:23
	Alternatively, if I could completely remove the file from the history and re-add it now that'd be fine with me as well.	14:23
MadCoder	Pieter: I had a look at gitg, okay it's decently fast, but it's mostly unusable right now	14:24
	I mean the UI it's promising, but it's not near being finished yet.	14:24
	I miss the affected files gitk thing, there is no way to set a fixed font for the source view widget (this is completely idiotic btw), it doesn't take args on the command like like gitk does, ...	14:25
	IOW it's a pretty poor gitk replacement yet	14:25
	so no I won't package it, I usually don't package things I won't use	14:25
	→ jesselucas_ joined	14:29
	→ ShadeHawk joined	14:30
ShadeHawk	pasky: Did you have time to check and test my "gitweb: Better processing format string in custom links in navbar" (including % -> %)?	14:30
toofishes	vinnl: git-format-branch	14:32
deskin	vinnl: have you taken a look at git filter-branch? That seems the easiest way to me; also, the second option (removing it from history, and adding a correct version on top) is probably the easier/safer way to go	14:32
toofishes	err	14:32
	git-filter-branch	14:32
ShadeHawk	tango_: what do you think about the way cgit does URL for snapshots?	14:32
	vinnl reads up on git-filter-branch, thanks toofishes and deskin	14:32
deskin	vinnl: the manpage has an example of exactly what you want, that is, removing a file from all history	14:33
vinnl	Great :)	14:33
	→ jackdempsey joined	14:33
tango_	ShadeHawk, haven't the slightest idea, looking it up now. btw, I just sent a v2 for the snapshot patches	14:33
ShadeHawk	Ah	14:33
	→ rdavila joined	14:34
jesselucas_	I'm trying to setup post-update to push to a mirror after it's updated. Here is my file: http://www.pastie.org/295720. This is a bare repo on server (B) being pushed to from local computer (A). I'd like server (B) to then push to server (C) the mirror. It doesn't seem to be working though.	14:34
ShadeHawk	tango_: well, the path/to/project/snapshot/next.tgz resulting in project-next.tgz looks a bit funny.	14:34
tango_	ShadeHawk, I know, in fact I was thinking about having gitwbe support project-head.tgz as synonym for project/snapshot/head?sf=tgz too	14:35
	← Flathead left	14:35
tango_	btw I can't find how cgit does snapshots	14:35
ShadeHawk	tango_: the problem is with prject names like on kernel.org, with deep directory hierarchies	14:36
tango_	well, there's only so much we can do about it, can we 8-P	14:36
	→ loiseau joined	14:37
tango_	ShadeHawk, we could just go by trying to replace the first - with /snapshot/ and see if it works, then try the second -, then try the third, etc	14:37
	not really ideal, so I desisted	14:37
	← rdavila left	14:38
tango_	also, if you have branches that go like blah/branchname, the file WILL be named branchname	14:38
ShadeHawk	tango_: http://permalink.gmane.org/gmane.comp.version-control.git/97985	14:38
	This is a minor feature-release which adds support for extracting	14:38
	snapshot revision (i.e. tag name) from the snapshot name instead of	14:38
	relying on querystring parameters, i.e. the following urls will	14:38
	download the expected revisons:	14:38
	http://hjemli.net/git/cgit/snapshot/cgit-0.8.1.tar.gz	14:38
	http://hjemli.net/git/cgit/snapshot/cgit-0.8.tar.gz	14:38
tango_	well, that would make snapshot behave VERY differently from the other actions	14:39
	also, I'm not sure how it does it, since the cgit project itself is tagged as 'v0.8.1' ... I don't want to impose conventions on the tagger 8-P	14:40
	→ ttt-- joined	14:41
ShadeHawk	well, one thing that can be done (and it is in my gitweb TODO list), is to use <project>-<tag>.<suffix> for snapshots of tags, and not <project>-<hash>.<suffix>	14:41
tango_	ShadeHawk, yeah I was thinking about that too	14:42
ShadeHawk	tango_: that is certain DWIM-mery (the way cgit does it)	14:42
tango_	(of course it's not easy in case multiple heads refer to the same commit)	14:42
	so what would cgit do with rbot, where tags are already named rbot-version and not just version?	14:42
	eh	14:42
	lines 139 and following: http://hjemli.net/git/cgit/tree/ui-snapshot.c	14:43
ttt--	hi, how do checkout the latest commit? when i do "git checkout master" it just lists files with "D", but it load them. what am i doing wrong?	14:43
	→ bobmcw joined	14:43
ttt--	do i*	14:43
ShadeHawk	tango_: the problem with <project>-<tag>.<suffix> is that I think we cannot use 'f' parameter (we could want to snapshot some subdirectory), and using 'fp' is abuse; new parameter? or just 'h' parameter?	14:44
ttt--	it doesnt load them*	14:44
	← deskin left	14:44
tango_	oh snapshot works for files too?	14:44
jesselucas_	ttt--: How you trying to switch branches or actually fetch the new content?	14:44
tango_	uhm	14:44
	I'll have to reconsider my patches	14:44
jesselucas_	*are you	14:45
ttt--	jesselucas, i deleted files since the last commit. i'd like to go back to the last commit (when the files were still there)	14:45
	→ charon joined	14:45
ttt--	there is only 1 branch	14:45
tango_	btw ShadeHawk the cgit way of doing stuff looks really a little too guessy for me ... I suspect it'd fail on comples projects	14:46
	(projects with complex versioning)	14:46
tokkee	ttt--: Do you want to throw away _all_ local changes?	14:46
ttt--	yes	14:46
tokkee	ttt--: git reset --hard	14:46
	→ alb joined	14:46
jesselucas_	I think git-revert does that	14:46
tango_	ShadeHawk, are you sure snapshots work with file_name too?	14:46
ttt--	thanks, that did it	14:47
ShadeHawk	tango_: well, DWIMmery could look like this: try <ver>, try v<ver>, try <project><ver>, try <project>-<ver>, fail	14:47
tokkee	ttt--: To "reset" a single file, do git checkout HEAD -- <file>	14:47
tango_	ShadeHawk, and what about generation?	14:47
ShadeHawk	tango_: I'm not sure, but I think they should... "$hash_base:$file_name" for $hash/<tree-ish>	14:48
tokkee	jesselucas_: git-revert reverts an existing commit by introducing a new one.	14:48
tango_	ShadeHawk, git_snapshot only checks $hash, not $hash_base or $file_name	14:48
jesselucas_	tokkee: gotcha, I guess I need to do some more research to see when the best time to use git-revert is.	14:50
	← snitko left	14:50
ShadeHawk	tango_: hmmmm... that looks incorrect... but $file_name should be used only for "$hash_base:$file_name" (or get_hash_by_path($hash, $file_name)), and for --prefix=$project/$file_name/	14:51
jesselucas_	tokkee: By chance, would you know why this post-update file isn't working: http://www.pastie.org/295720 ?	14:52
	← jackdempsey left	14:52
	→ trochala joined	14:52
	← GNUix left	14:52
ShadeHawk	jesselucas: on server? is git in PATH on server? what is configuration on server?	14:54
tokkee	jesselucas_: The first call to exec replaces the running shell with git, so you never reach the second exec.	14:55
tango_	ShadeHawk, well, I'm only adding path_info stuff, not reworking the internals at the moment, so I guess you should mark that for a TODO 8-)	14:55
tokkee	jesselucas_: s/shell/shell process/	14:56
jesselucas_	ShadeHawk: It's a ubuntu server. Using gitosis to manage the repos. Not sure about PATH (first time setting up everything)	14:56
vinnl	toofishes, excellent, git-filter-branch did it, thanks again :)	14:56
jesselucas_	tokkee: Do you happen to know of a tutorial that explains this?	14:56
tokkee	jesselucas_: Explains what?	14:57
jesselucas_	tokkee: I guess it's probably more of a linux admin thing. Basically what language is the post-update written in? So I can search for how to modify it?	14:58
tokkee	jesselucas_: That's plain shell script (in that case).	15:00
jesselucas_	tokkee: awesome, thanks for the help. Now to figure out shell scripts :)	15:01
	← albertito left	15:05
	→ deskin joined	15:06
	→ CodeOfficer joined	15:08
	→ diask joined	15:10
	← Gitzilla left	15:13
	← vinnl left	15:13
JensB_	how do I commit only parts of the changes in one file using git?	15:19
	along with other files	15:19
Pieter	git add -p	15:19
JensB_	thank	15:20
	s	15:20
ShadeHawk	tango_: I just had a weird ida, of having project/snapshot/ be a tree-like list of possible snapshots of tags in <project>-<tag>.<suffix> format (like for example real http://www.kernel.org/pub/software/scm/git/)	15:20
	JensB_: or git-gui	15:20
	s/ida/idea/	15:20
tango_	ShadeHawk, that's a possibility, but it's not cheap to build	15:21
	→ woeye joined	15:21
	→ truebosko joined	15:22
truebosko	Hi there, I'm dealing with a remote repo and after pushing some changes to master I want to pull them all back to another server which has already cloned the repo. If I go into the root directory of the project and do `git pull` it seems to only grab the changes in that root dir and not the ones under it .. how can I grab all changes and apply them?	15:23
drizzd	it's impossible to get only a part of the changes with git	15:24
truebosko	I dont want just a part, I want all	15:24
drizzd	you are clearly mistaken somewhere	15:24
truebosko	Ok	15:24
	I have /dev/cake	15:24
	Cake is my project directory. I go into cake, git pull	15:24
	and only the files in that dir are updated, every subfolder has nothing updated	15:24
	← diask left	15:24
truebosko	But when I look at my commits, all other files in sub dirs are being updated	15:25
drizzd	is it a public repo?	15:25
truebosko	No, it's a private repo on github	15:25
	If I clone it, it works fine but I dont want to constantly clone .. :)	15:26
drizzd	ok, how do you "look at your commits"?	15:26
truebosko	Either gitk or going on github.com	15:26
drizzd	so you can see the changes to the subdirectories in gitk	15:26
truebosko	To the files in the sub directories, yes	15:26
drizzd	in the branch you're on (the one shown with bold face font in gitk)	15:26
	or the one with the asterisk in front of it in `git branch`	15:27
	← eno__ left	15:27
truebosko	oh you know what	15:28
	one of the files was edited on remote without any git-relation .. as I wasnt at my main pc	15:28
	→ bdiego joined	15:28
truebosko	Is it possible to ignore the not uptodate error on a file?	15:28
	→ shodges joined	15:29
ShadeHawk	tango_: well, not that much expensive than 'tags' view	15:29
truebosko	Because I dont want to really commit it to the repo anymore	15:29
drizzd	what's a "not uptodate error"?	15:29
truebosko	settings.py: needs update	15:29
	fatal: Entry 'settings.py' not uptodate. Cannot merge.	15:29
	→ madewokherd joined	15:29
drizzd	you don't have to commit changes	15:29
	you can undo changes	15:30
	you can ignore changes	15:30
truebosko	That's on my remote. . which I am pulling to	15:30
drizzd	you can remove files from git and not track them any more	15:30
truebosko	I edited it on remote, but I want to pull the latest from my master	15:30
	No no	15:30
	I know that	15:30
	→ bentob0x joined	15:30
truebosko	The process was like so:	15:30
tango_	ShadeHawk, I guess that could be implemented, in the form http://gitweb.example.com/snapshot/somestuff	15:31
	ShadeHawk, however, that assumes no project named snapshot exists	15:31
truebosko	I got my working copy on HOME, I pushed it to my repo. Then I cloned that repo onto REMOTE. The next day I needed a quick change and modified settings.py on REMOTE. Then, I went back HOME and made more changes, blah blah the usual development cycle. Pushed those changes to repo, but now when I try to PULL the latest changes to REMOTE it crashes because I modified settings.py on REMOTE	15:32
tango_	or were you suggestiong http://gitweb.example.com/project/snapshot/project-tag ?	15:32
truebosko	Seems just deleting settings.py on remote does the trick though :P	15:32
ShadeHawk	tango_: no, I thought about http://gitweb.example.com/project/snapshot/ (only wothout either hash or filename)	15:32
	→ stouset joined	15:32
drizzd	truebosko: first of all, you seem a little confused about push and pull	15:33
tango_	ShadeHawk, and then have snapshot behave totally different from the other actions?	15:33
ShadeHawk	tango_: so the snapshow would be http://gitweb.example.com/project/snapshot/project-tag.tgz or something like that	15:33
drizzd	truebosko: push and pull has nothing to do with changes	15:33
ShadeHawk	tango_: that is just a wild idea	15:33
drizzd	those commands are simply used to download and upload data from one repository to another	15:33
tango_	could be a fallback	15:33
truebosko	drizzd: I know, and that's what Im using them for :?	15:34
	Sorry just having a hard time explaining t he situation but no worries I figured it out	15:34
ShadeHawk	tango_: special case: snapshot without hash or hash-ish (hash_base + file_name) lists "available" snapshots	15:34
	→ mw joined	15:34
drizzd	truebosko: I'm just saying, your error has nothing to do with where or how you made your change.	15:34
tango_	ShadeHawk, that's surely an interesting option. of course the next matter is what should the snapshots link to 8-D	15:35
drizzd	truebosko: everything is local with git	15:35
	← bdiego left	15:35
	→ pehlert joined	15:37
	← bobmcw left	15:40
	← nessundorma left	15:40
lack	I've cloned a repository and set up my own branch, where I've deleted one of the directories in the original repository. Now every time I pull and upstream has changed the contents of that repository, my merge fails and I have to manually delete that directory again. Is there any way I can automate this procedure? Or tell git not to merge things in the deleted directory?	15:40
tokkee	lack: Did you have a look at git-rerere? That _might_ help ...	15:42
Pieter	I don't think it will	15:42
	← deskin left	15:45
	→ nuncanada joined	15:48
	← stouset left	15:49
Ilari	truebosko: "fatal: Entry 'settings.py' not uptodate. Cannot merge." means that 'settings.py' has local uncommitted modifications and it is also modified on remote side (and Git can't merge files in that state).	15:51
	→ zachinglis joined	15:53
lack	tokkee: Nope, I have rerere enabled, but it doesn't seem to do anything special.	15:54
	← woeye left	15:55
	→ Weasel[DK] joined	15:57
JensB_	newbie question #2934513: how do I remove a "git mv" from the current commit? git update-index --force-remove $file will mark as removed, not "undo" the "git mv".	15:58
ciaran	reset	15:58
JensB_	reset seems also to mark as "deleted" (in the commit message comment).	15:59
	← jesselucas_ left	15:59
	← ph^ left	16:00
Ilari	JensB_: 'git mv' the file back?	16:01
JensB_	will try	16:01
	← ShadeHawk left	16:01
JensB_	thanks	16:02
	← JensB_ left	16:02
	→ ph^ joined	16:02
	← elmex left	16:04
	→ elmex joined	16:04
	← ndim left	16:07
	tonguero_ → tongueroo	16:08
	← RaceCondition left	16:11
	→ bobesponja joined	16:12
	← EmilMedve left	16:19
idletask	Grr	16:23
	I can't make the git eclipse plugin to work _at all_	16:24
	→ t_ joined	16:24
	→ kukks joined	16:26
	→ p0w3r3d joined	16:27
	← alexross left	16:28
	→ alexross joined	16:32
	← t_ left	16:39
	→ blacky joined	16:40
	← shenie left	16:41
	← ereslibre left	16:43
	→ ndim joined	16:43
	→ snitko joined	16:43
	→ spearce joined	16:44
	← truebosko left	16:44
	→ bieneff joined	16:47
	← prophile left	16:47
	→ diask joined	16:52
	← KiBi left	16:57
	← p0w3r3d left	17:00
	→ p0w3r3d joined	17:02
	→ koke joined	17:03
	← vuf left	17:05
	← alreves_ left	17:12
	← doener left	17:13
	→ doener joined	17:13
	← rubydiamond left	17:14
	→ rubydiam_ joined	17:14
	→ litage joined	17:15
	→ chris2 joined	17:20
	← koke left	17:20
	→ Beket joined	17:22
	← CIA-15 left	17:29
	← toofishes left	17:30
	← Vortex35 left	17:32
	→ Vortex35 joined	17:32
	← Sigma left	17:33
	← nano- left	17:33
	→ radarek joined	17:37
	← mankind_tweezer left	17:39
	offby1 → plotz	17:40
	→ CIA-15 joined	17:40
	plotz → frotz`	17:40
	frotz` → offby1	17:41
	alb → albertito	17:41
	← statim left	17:42
	→ statim joined	17:42
	rubydiam_ → rubydiamond	17:50
	→ cko joined	17:51
	← CIA-15 left	17:52
	→ harryjr joined	17:52
	→ Jacolyte joined	17:52
harryjr	is it possible to specify multiple remotes, so a git push will be done two multiple locations?	17:52
drizzd	I don't think so	17:54
spb	i don't know that 'git push' on its own can be made to push to two places	17:58
	but you can certainly configure two remotes and do git push remote1 ; git push remote2	17:59
	← ia left	17:59
	← cko left	17:59
	→ ia joined	17:59
	← ia left	18:00
	→ CIA-15 joined	18:04
	→ marze joined	18:06
	→ Sigma joined	18:08
	← harryjr left	18:09
	not-xjjk → xjjk	18:09
	→ kef joined	18:10
	→ bytefield joined	18:16
	← tongueroo left	18:17
	← bentob0x left	18:18
	← ttt-- left	18:20
	→ lamont` joined	18:20
	← iulian left	18:26
	← pgokeeffe_ left	18:28
	← bytefield left	18:28
	← lamont left	18:29
	→ pantsman joined	18:29
	→ koke joined	18:30
	→ dotsintacks_ joined	18:33
	→ reallyidle joined	18:33
	← ben_h left	18:36
	← kef left	18:38
	← FunkeeMonk left	18:39
	→ eternaleye joined	18:42
	← dotsintacks left	18:45
	← idletask left	18:46
	← snitko left	18:48
	← shodges left	18:48
	→ tiglionabbit_ joined	18:49
	← tiglionabbit left	18:49
	← zachinglis left	18:52
	← pygi left	18:56
	→ offby1` joined	19:00
	← bryanray left	19:00
	← offby1 left	19:06
	offby1` → offby1	19:06
	→ bentob0x joined	19:08
	→ shenie joined	19:08
	← dotsintacks_ left	19:11
	← BarryCarlyon left	19:12
	→ dotsintacks joined	19:12
rubydiamond	How do I remove a tag	19:15
cehteh	trag -d	19:16
	s/r//	19:16
	← dthomas left	19:21
MadCoder	rubydiamond: git tag -d to remove it locally	19:21
	git push <remote> refs/tag/<tag>:	19:22
	to remote it from the remote	19:22
rubydiamond	MadCoder: Can I use same tag name again and agan	19:23
	suppose I tag it now.	19:23
MadCoder	yes you can	19:23
rubydiamond	and later I tag my different commit again with same name again	19:23
MadCoder	git tag -f will allow you to reuse the same tag name	19:23
	though you should make it a lightweight tag and not push it	19:23
rubydiamond	okay..	19:24
MadCoder	(IOW do not use git tag -a for tag you intend to move)	19:24
rubydiamond	what is git tag -a	19:24
MadCoder	an annotated tag	19:24
	IOW a tag with a comment	19:24
	it's usually what you mean to use for releases and stuff like that	19:25
	you can sign that (with gpg) and a couple of such features	19:25
	it's frowned upon to move such tags as those are supposed to be what they are forever	19:25
	← charon left	19:28
rubydiamond	okay	19:29
	→ Beowulf_ joined	19:30
Beowulf_	for sure this is plain easy but, how do I copy a file in a branch to another?	19:30
	→ paltman joined	19:31
	→ jesselucas_ joined	19:31
MadCoder	Beowulf_: go in the branch where you want to "copy" the file	19:33
	← CIA-15 left	19:33
MadCoder	git show otherbranch:path/to/file.c > path/to/file.c	19:33
Beowulf_	uhm... interesting	19:34
	thanks! :D	19:34
Ilari	Beowulf_: Or 'git checkout otherbranch path/to/file.c' (or something like that).	19:37
	← p0w3r3d left	19:37
robin_	Gitster has to change his mind on those five (now four) years. We need a fix for non-ascii filenames in git.	19:37
	→ rtomayko joined	19:38
MadCoder	for what ?	19:38
robin_	http://code.google.com/p/msysgit/issues/detail?id=159 for example	19:39
	there was another thread a day ago too	19:39
MadCoder	I hardly understand how this is a git problem	19:40
	or his file system uses latin1 file names encoding whereas git likes it utf8	19:41
	I dunno, it looks like a file system issue though	19:41
DrNick	it's a Windows thing	19:42
parasti	does git even care about the encoding of file names? it just stores bytes	19:42
DrNick	you either use UTF-16, or you use the 8-bit locale charset	19:42
	← \ask- left	19:42
robin_	parasti: and that is the problem	19:42
MadCoder	robin_: I hardly see why.	19:43
	but oh well, my system is full utf-8, and systems that are not full unicode nowadays should die an horrible death	19:43
	→ sabooky joined	19:43
DrNick	git is writing UTF8 filenames using the locale APIs, which means they get converted to the native UTF-16 using the wrong conversion tables	19:43
robin_	MadCoder: Windows is full unicode	19:43
	→ RaceCondition joined	19:44
	← jesselucas_ left	19:44
MadCoder	robin_: it maintains legacy horrible code which does that many FS you create are not using full unicode file names	19:44
	→ d0k joined	19:44
MadCoder	and it's a pita	19:44
sabooky	Can someone point me to a tutorial or documentation explaining the "proper" workflow when using git. Like the whole cheap branches/rebasing and how that's supposed to be used.	19:45
MadCoder	there is no proper workflow	19:45
	reallyidle → idletask	19:45
MadCoder	there are always a couple of bad things not to do with git	19:45
	→ fhobia joined	19:45
robin_	MadCoder: but git really shouldn't use the legacy encoding. Only legacy programs should.	19:46
MadCoder	but there is not one single worflow, git is versatile enough	19:46
	← paltman left	19:46
MadCoder	robin_: I dunno, I don't use windows, I only know git works for me here	19:46
	and I seldomly version files with 8bit chars in'em anyway	19:46
robin_	so then nobody else should..	19:46
sabooky	MadCoder: fair enough.. can you point me to a good documentation on git in general? Been reading into "Git Internals" from peepcast, but trying to expand on that.	19:46
	→ ereslibre joined	19:47
DrNick	the problem is that windows has something like four different file access APIs, and none of them resemble the POSIX APIs that git is using	19:47
MadCoder	sabooky: git has documentation, with tutorials and lots of things in it	19:47
	like the everyday's git and docs like that	19:47
	→ paltman joined	19:47
robin_	Nobody uses Windows posix API	19:47
idletask	sabooky: try the docs at git.or.cz	19:47
DrNick	which means if you were to do it right, you'd have to insert a filesystem abstraction layer, and nobody who cares has bothered	19:47
MadCoder	DrNick: maybe	19:47
	and nobody should	19:48
	→ bryanray joined	19:48
	→ KiBi joined	19:48
robin_	yiises	19:48
Ilari	sabooky: Basically, you are supposed to fork a branch for each feature and then (assuming it isn't so large change that you need to publish it) keep them up to date by rebasing.	19:48
DrNick	MadCoder: maybe? there's the ANSI and Wide versions of the native functions, and then the different versions of the C library functions	19:48
MadCoder	I said maybe because I really don't care	19:48
	windows APIs are a mess	19:49
Ilari	sabooky: Finally, when the changes are ready, merge them back into master. Keep master free for quick and obivious changes.	19:49
MadCoder	and I'm glad I don't have to fiddle with'em	19:49
	→ proxie joined	19:49
	→ awarde joined	19:49
sabooky	Ilari: ah, thats what I read in "Git Internals" and was looking for more examples of this. Since its a very different approach than cvs/svn.	19:49
DrNick	although, I would have though that msysgit took care of this crap for you	19:50
Ilari	sabooky: Its different there for two reasons: 1) Merging in CVS/SVN is a pain. 2) CVS/SVN can't rebase...	19:50
sabooky	Ilari: Trying to see if git is a good tool for work or not. Still trying to wrap my head around the concept differences. Anyways, I'll check out the official documentation(s).	19:51
	Ilari: yea.. I've been using git wrong, or atleast not taking advantage of that at all (for personal use)	19:51
	→ carllerche joined	19:52
robin_	DrNick: I tried setting the utf-8 codepage, but that didn't work out.	19:52
Ilari	sabooky: Alternatively, one could just update topic branches by merging, but this leads to "messier" history.	19:52
	→ Marmouset joined	19:52
robin_	seems msys (and cygwin) lacks support for it	19:52
Ilari	sabooky: Update-by-merge is regarded lot less bad than not using branches at all...	19:53
robin_	DrNick: It was actually solved a year ago, but junio didn't want it	19:54
DrNick	url?	19:54
	→ CIA-2 joined	19:54
	→ alb joined	19:54
sabooky	Ilari: I'm trying to figure out how my workplace layout would be converted to git. There's Trunk (bleeding edge, devel, not stable), then branches for each major version (7.x.x-branch, 8.x-branch, 8.x.x-branch, etc..)	19:56
Ilari	sabooky: CVS/SVN quickly convince developers to avoid branches... And then they fell lost when they come across systems where branching is important part of standard workflows...	19:56
	→ k776 joined	19:57
DrNick	Ilari: that's both true and largely irrelevant to his question, I think	19:57
sabooky	Ilari: Yea, branches are considered a "necessary evil" at my place. I'm actually very new to the field btw, but I'm the cvs admin currently (yes we use CVS :\)	19:57
Ilari	sabooky: The standard way to do changes is to make the change on most stable codebase it appiles to and then merge towards least stable one.	19:58
robin_	DrNick: I'll see if I can find it	19:58
	← Beket left	19:59
robin_	DrNick: http://kerneltrap.org/mailarchive/git/2008/1/18/578936	19:59
Ilari	sabooky: Git.git does it that way. It has 3 main branches, maint (most stable), master and next (least stable). Additionally there is 'pu', but thats special case.	20:00
	← albertito left	20:02
	→ tongueroo joined	20:03
sabooky	Ilari: Thanks for all the info so far, it was definitely helpful. Gonna start reading the official docs and see where that gets me.	20:05
offby1	Ilari: you've just mentioned some good ideas for using git ... and I'd never heard them before :-\| (1: fork a branch for each feature, keep it up to date by rebasing, when the changes are ready, merge them back into master; 2: make the change on most stable codebase then merge towards least stable one.)	20:05
	← paltman left	20:05
parasti	hmm	20:06
	→ JensB joined	20:10
JensB	Hi everybody	20:10
	← bryanray left	20:10
	← bobesponja left	20:11
JensB	I'm using Git for a Ruby on Rails project (new user). Say I've made a couple commits (all local) in my master branch and now I see I should have branched off the master five commits ago, because my idea didn't work out. how would I do that?	20:11
MadCoder	say you're in branch bar and would have liked to branch off foo	20:12
	be in bar	20:12
	git branch foo bar	20:12
	git checkout bar	20:12
	git reset --hard HEAD~5	20:12
	→ johnw joined	20:12
MadCoder	(provided you have no local modifications, and you never pushed bar, if you did you're screwed)	20:13
JensB	:)	20:13
Ilari	offby1: Those ideas have been known for long time (first is called "topic branches").	20:13
JensB	thanks!	20:14
offby1	Ilari: perhaps, but I didn't know them :-\|	20:14
Ilari	MadCoder: Eh... If you are on 'bar', is that checkout going to do anything?	20:15
MadCoder	no	20:15
offby1	I'd certainly heard the phrase "topic branch", but didn't know that it implied that specific workflow (keep up to date by rebasing; merge when done)	20:15
MadCoder	unless you have a file named bar	20:15
	there it should error out if your git is recent enough	20:15
	else it will revert changes to that file	20:15
	← blacky left	20:15
	← wanders left	20:15
	← Soliton left	20:15
MadCoder	(that's a bug that I fixed in 1.6.sth)	20:15
	← abbe left	20:15
	→ bcarlyon\|laptop joined	20:22
	← radarek left	20:22
	→ pygi joined	20:22
	→ wanders joined	20:23
	← bentob0x left	20:23
	→ johnw_ joined	20:25
	← diask left	20:25
	← loiseau left	20:25
	→ diask joined	20:26
	→ Gitzilla joined	20:27
	bcarlyon\|laptop → BarryCarlyon	20:29
	← johnw left	20:30
	← tongueroo left	20:34
	→ loiseau joined	20:35
	→ tongueroo joined	20:35
	proxie → bryanray	20:37
	→ Soliton joined	20:38
	→ diiask joined	20:40
	← diask left	20:40
	← Sigma left	20:42
	→ nikolasco joined	20:44
	→ jesselucas_ joined	20:45
	→ rtomayko_ joined	20:48
comp	hi, I made some changes to my local repository (which is clone from somone else's repo) and then merged that remote, which resulted in conflict. That's common thing for me, but I always wondered if it's possible to determine which commit(s) changed my modified lines and thus threw my branch into conflict(s).	20:49
drizzd	you can find out which commit last modified a line using git blame	20:51
	That probably doesn't work on unmerged files directly though	20:51
offby1	don't the conflict markers include a description of the commit that caused them?	20:53
Ilari	offby1: Nope, they only can refer to versions being merged...	20:54
drizzd	seems git blame outputs some funny stuff with unmerged files	20:54
	00000000 (Not Committed Yet 2008-10-19 22:54:47 +0200 10) <<<<<<< HEAD:file	20:55
	c3395968 (Clemens Buchacher 2008-10-19 22:53:25 +0200 11) 10 a	20:55
	^fe2c981 (Clemens Buchacher 2008-10-19 22:53:09 +0200 12) 11	20:55
	00000000 (Not Committed Yet 2008-10-19 22:54:47 +0200 13) =======	20:55
	00000000 (Not Committed Yet 2008-10-19 22:54:47 +0200 14) 10	20:55
	00000000 (Not Committed Yet 2008-10-19 22:54:47 +0200 15) 11 b	20:55
	00000000 (Not Committed Yet 2008-10-19 22:54:47 +0200 16) >>>>>>> b:file	20:55
	I'm not sure what the caret means here	20:55
	hmm, maybe because it's the root commit...	20:57
comp	so git pickaxe is now git blame?	20:57
	← carllerche left	20:57
	→ lydgate joined	20:57
comp	I thought about some easier way, but that will work too	20:58
Ilari	comp: Pickaxe is something different.	20:58
Tv	comp: well the thing is, git doesn't merge your changes commit by commit, so the conflict isn't really "caused" by any single commit	20:58
comp	Tv: I know, just wanted to know if there's some "automated" tool which uses git-blame ..	20:59
	tool or way	20:59
Tv	comp: so what you want to ask is, what commits made these lines that end up being in conflict, and blame is pretty much that	20:59
	comp: you can give blame ranges etc	20:59
	something like -L '/^<<<</,/^====/'	21:00
comp	well in the end, I'd like to cherry-pick only those from the main repository so I don't have to merge like once a week ..	21:00
Tv	huh	21:00
	→ teolicy_ joined	21:00
comp	it's for single-feature branches	21:01
Tv	comp: don't cherry-pick unrelated commits into topic branches	21:01
	i don't see why you'd need to keep redoing any merges	21:01
teolicy_	Hello. I'm new to git, and am wondering whether I should bother with gitosis (or another git server... is there?) for my 1-2 user home open-source project.	21:02
	I mean, would just using git from my two desktops using NFS be enough?	21:02
	(what would I miss out on)	21:02
Ilari	teolicy_: Git has shared flag you can set on repository so it overrides overly restrictive umasks...	21:03
Tv	teolicy_: what about the people who aren't on your home nfs setup?	21:03
teolicy_	Tv: That's an excellent question and the reason I started with gitosis in the first place, I thought that would be the quickest way to share over SSH (my server is 'net accessible using SSH).	21:04
	Tv: What would you recommend?	21:04
comp	Tv: well how can I then stay in sync with the origin/master? It's a public "fork", so I can't afford rebases even if I wanted to ..	21:04
Tv	teolicy_: umm, please read the gitosis README.rst and think ;)	21:04
	← rtomayko left	21:04
mugwump	you don't need to do anything to push via ssh etc	21:04
	just make sure git is installed on the remote end	21:05
Tv	comp: well if you merge, then you merge, and git remembers	21:05
comp	teolicy_: there are few gitosis tutorials on the google ..	21:05
teolicy_	Tv: I tried going through two gitosis tutorials, but had trouble installing it on my server (it's a reasonably recent OpenSolaris).	21:05
Tv	mugwump: you're assuming 1) full shell users or 2) user is willing to set up and understand security implications of git-shell etc	21:05
	← RaceCondition left	21:05
mugwump	Tv: right, but in a home situation that's	21:06
teolicy_	I seem to have trouble authenticating SSH with RSA keys on the gitosis user.	21:06
mugwump	usually the case	21:06
teolicy_	I'm not sure why, I humbly think I'm fairly proficient with SSH.	21:06
Ilari	teolicy_: But if you have multiple multi-user projects with differing access rules, then adminstering using standard unix tools and shared flag can get painful...	21:06
	teolicy_: What SSHD are you using?	21:06
comp	Tv: that's right, just wanted to keep the branch without things like 5 merges in a row ..	21:06
	→ deskin joined	21:06
	→ advany_ joined	21:06
teolicy_	Ilari: OpenSSH_5.1p1	21:07
Tv	comp: man git-rerere?	21:07
Gitbot	comp: the git-rerere manpage can be found at http://git.or.cz/man/git-rerere	21:07
Tv	comp: merges even to throwaway branches can remember their conflict resolution; that mechanism is just a cache, not a part of the full history	21:07
mugwump	teolicy_: if in doubt, use sshd -d -t	21:07
	er, or is it -d -v, I forget	21:07
	→ carllerche joined	21:08
Tv	teolicy_: something like AllowUsers set?	21:08
mugwump	the one that makes it tell you why the auth failed :)	21:08
Ilari	teolicy_: You don't happen to have AllowUsers in SSHD config by any chance? That caused problems for one user day or two ago...	21:08
teolicy_	Just to zoom out of gitosis a bit, the reason I came to the channel was that I'd rather not focus on git, but rather focus on my project. Unless I have to use gitosis for SSH based remote access to my repositories, I'd rather not.	21:08
	Tv, Ilari: No.	21:08
	← d0k left	21:09
mugwump	you don't. in fact if it's firewalled you can even make git:// push-able	21:09
Ilari	teolicy_: How do those problems show up anyway? It asks for password?	21:09
teolicy_	mugwump: I tried keeping sshd in the foreground and seeing what's wrong, I couldn't discern something obvious in a quarter of an hour and decided before I dig deep into sshd to make gitosis run so I can use git so I can work on my project, maybe I should see if I /really/ need gitosis.	21:09
	Ilari: Yes.	21:09
	→ RaceCondition joined	21:10
teolicy_	Ilari, Tv, mugwump: Give me a minute, I'll get it back to a state I can copy from sshd -d	21:10
Tv	teolicy_: gitosis is absolutely and fully a convenience + extra security wrapper	21:10
Ilari	teolicy_: Look at .ssh/id_rsa.pub. It should have long block of letters. Look at .ssh/authorized_keys of git user. It too should have long block of letters. Are those blocks the same?	21:10
offby1	ssh-copy-id is a handy way to get one's public key into an .ssh/authorized_keys file	21:11
Ilari	teolicy_: Well, letters, numbers and some other characters...	21:11
teolicy_	Ilari: Yes they are.	21:11
Ilari	teolicy_: Try 'ssh -vvv git@host git-upload-pack foobar'. Does it offer id_rsa for authentication (and then get rejected)?	21:12
comp	Tv: well I had in mind something little different, but this tool is perfect, thanks	21:12
teolicy_	Ilari: Aye, sec. Though the line in my authorized_keys has the following options:	21:13
	→ blacky joined	21:13
teolicy_	command="gitosis-serve [email@hidden.address] ssh-rsa <... key here ...>	21:13
	← nuncanada left	21:13
jesselucas_	If I'm using gitosis, and I setup a hooks/post-update hook to git push to a remote. Is git the user that's pushing it to the remote server. So I'd need to give the git@serverA access to serverB in order for the hook to work?	21:13
teolicy_	I thought no command but 'gitosis-serve teo...' would work. No?	21:13
Ilari	teolicy_: What are permissions of that authorized_keys file and _ALL_ directories up to root directory (yes, SSHD really does check those) and it won't work if they are too lax...	21:15
	→ markelikalderon joined	21:15
teolicy_	Ilari: On the client (ssh -vvv ...), I see "debug1: Offering public key: /Users/teolicy/.ssh/id_rsa", following which it /looks/ like the key is OK (for example, I can see the forced command the server dictates, etc, But the next thing that happens is my client gives up on that and continues to the next key and then keyboard auth.	21:15
	Ilari: All permissions are OK. I often use ssh keys for authentication, I humbly think I'm comfortable with it.	21:16
Ilari	teolicy_: The server sends the forced command back to client?	21:17
	teolicy_: What SSH client?	21:18
teolicy_	OpenSSH_5.1p1	21:18
	→ paltman joined	21:18
	→ dave_h_d joined	21:18
teolicy_	This is from the client: "debug1: Remote: Forced command: gitosis-serve [email@hidden.address] I understand the server sent it, otherwise how the client knew.	21:19
	← paltman left	21:19
Ilari	teolicy_: So the server accepts the key... But why doesn't the client proceed?	21:19
teolicy_	This is from the server (when ran with sshd -D -d): "debug1: matching key found: file /home/git/.ssh/authorized_keys, line 2".	21:19
	Ilari: So it seems.	21:20
	→ ia joined	21:21
teolicy_	Ilari: Check this out, this is from the server: http://pastebin.com/m66b9b29a	21:22
	← markelikalderon left	21:23
	→ markelikalderon joined	21:23
Ilari	teolicy_: If you remove the commmand part (THIS BREAKS GITOSIS!), does the key allow getting shell? After test, put the command=... stuff back...	21:24
	← advany left	21:24
	→ bdiego joined	21:25
	AreliusFreeNode → Arelius	21:26
	← dave_h_d left	21:27
teolicy_	I tried that (I've done all steps we did so far including removing the command, and then turned to the channel). I admit this doesn't look very gitosis related, but I got frustrated and came here to hear if there are alternatives. Seems not, as for all I understood so far.	21:28
	← carllerche left	21:28
	← JensB left	21:28
	→ Innovia joined	21:29
Ilari	teolicy_: Create unix user group for developers, and set 'core.sharedrepo' to 'world'... Or something like that...	21:29
	← jaalto left	21:29
Innovia	hi why my push doesnt update the code on the server?	21:29
cehteh	because you want to push to a bare repo	21:30
teolicy_	Yes, that sounds sucky. I'll focus on fixing getting shell using the git user, then come back.	21:30
	I think I need gitosis.	21:30
Innovia	cehteh: what do you mean	21:30
Ilari	teolicy_: Are you really sure you don't have AllowUsers or any kind of simililar setting in SSHD config?	21:30
cehteh	repositories on a server where one only pushes to should be created as --bare	21:31
	that is no checked out sources	21:31
	read the docs about	21:31
	if you really need some checked out copy then you have to do some little advanced stunts	21:31
teolicy_	Ilari: Yep.	21:31
	You know what, I'll do a magic trick.	21:32
Innovia	i got the peepcode screencast which explain to create the git repo on my local then scp it to the server then clone it back to my local	21:32
teolicy_	I'll add a user and authenticate to /that user/. No gitosis at all.	21:32
	(again, once we removed command=..., I quit believing gitosis is to blame, this has to be me)	21:32
Innovia	i want to push and pull	21:32
	← Weasel[DK] left	21:32
cehteh	Innovia: instead scp'ing you may rather git init --bare a empty repo there and right push to it	21:33
Ilari	teolicy_: These also sound potentially troublesome: AllowGroups, DenyUsers, DenyGroups...	21:33
Innovia	can i recreate the git init on the server now with --bare then clone it to my machine?	21:34
	will i then be able to push and pull freely?	21:34
	← blacky left	21:34
Ilari	Innovia: Easier way to convert it is just to rename the foo/.git to foo.git...	21:35
teolicy_	Ilari: cat /etc/ssh/sshd_config \| gegrep -v \# \| gegrep 'Allow\|Deny' yields nothing but AllowTcpForwarding, which isn't related.	21:35
Ilari	teolicy_: What's gegrep. And is that search case-sensitive (it should perhaps be case-insensitive)?	21:35
	← bryanray left	21:35
	→ charon joined	21:36
	→ drizzd_ joined	21:36
	→ bryanray joined	21:36
Ilari	teolicy_: Ah, GNU Grep.	21:36
	← felipec left	21:36
Ilari	teolicy_: Does the same public key work for some other account on same server?	21:38
Innovia	Ilari: i rename the directory but how do i make my local changes push (since now the path is broken)	21:39
	do i need to rename it on both ends?	21:39
Ilari	Innovia: You can edit the URL in '.git/config'.	21:40
Innovia	ok	21:40
	do i need to rename the .git in my local as well?	21:40
	→ HopsNBarley joined	21:40
Ilari	Innovia: No. Doing that rename would in fact break stuff...	21:41
Innovia	ok	21:41
	thanks	21:41
Ilari	Innovia: You probably do want to keep working tree for local repository...	21:42
Innovia	yes	21:42
teolicy_	Ilari: Yes.	21:42
	(same results with -i on the grep, btw)	21:43
Ilari	teolicy_: Do logs SSHD saves contain anything relevant?	21:43
teolicy_	Anyway, let me try my suggested magic trick (creating another account, very similar to git).	21:43
Innovia	Ilari: i changed the url, but now when i do push it reply with "No refs in common and none specified; doing nothing.	21:43
	oops ignor	21:45
	its a permission denied thing now... i can find where...	21:45
teolicy_	Ilari: I think I got it.	21:46
Ilari	Innovia: Permission denied errors come from standard Unix permissions...	21:46
Innovia	755?	21:46
Ilari	teolicy_: What was the issue?	21:46
	Innovia: That's OK if users match, but not OK if they don't...	21:46
Innovia	i saw somewhere that git can only take 2 persmissions 755 640	21:46
	← pantsman left	21:46
teolicy_	I /think/ (still need to test) PAM is blocking this because the /etc/shadow entry has LK for a password for git (and not for the other user authenticating ok). The LK must mean Locked out in Solaris.	21:47
	I'm checking this now.	21:47
	Yep. Ugh, how sucky.	21:47
	Sorry about that. I'll keep going through the git tutorial now... :)	21:47
	s/:\)/:-\//	21:48
Ilari	Innovia: Huh... Git also uses (at least used to) 444 a lot...	21:48
Innovia	my user match but for somereason i get permission denied cant create temp file	21:49
Ilari	Innovia: Are you pusing over SSH://, GIT:// (yuck) or locally?	21:49
Innovia	Ilari: that's what i got drwxr-xr-x on mor.git	21:49
	ssh	21:50
nadim_	why yuck?	21:50
Tv	teolicy_: yeah there's a difference between "no password will ever match" and "locked out", even on linux	21:50
Ilari	Innovia: Which user owns mor.git? What user are you trying to use on remote side?	21:50
Innovia	git	21:51
	git	21:51
Ilari	Innovia: Is the url something like ssh://git@host...?	21:51
Innovia	url = [email@hidden.address]	21:51
teolicy_	Tv: Yes, I knew that, overlooked it when I copied an old entry from Solaris. They could have just used "locked-out" rather than "LK", but what the heck, I guess not everyone imported this.	21:51
Innovia	i guess i thought since i scp it, the push will remain ssh but i guess i'm way wrong there ha?	21:52
Ilari	Innovia: You could try 'chown -R git' that repo...	21:52
Innovia	i did chown -R git:git mor/	21:53
	is my url is ok? can you tell if its not using ssh?	21:53
	← drizzd left	21:54
Ilari	Innovia: Perhaps some directory on path to root has no +x for git (through group or world access bits)?	21:54
	← idletask left	21:55
Ilari	Innovia: Can you get shell as git user on remote side?	21:56
Innovia	Ilari: when im connected to the remote through ssh i can create files (when su git) inside mor	21:56
Ilari	Innovia: In practicular inside 'objects/' ?	21:57
Innovia	you mean connect ssh git@myserver...	21:57
nadim_	Any mailing list administrators around?	21:57
	git has one of the best irc support around and the worst mailing list ever	21:58
Ilari	Innovia: Can you change directories to / and back inside mor as git user?	21:58
nadim_	pasky: around?	21:59
Innovia	Ilari: i think i know what's the problem, I have mor directory sits in monqiboy user folder, git home folder is empty	21:59
	is this a problem>	21:59
Ilari	Innovia: Ah, that permissions-on-path probably can't be the issue as it recognizes git repo on remote side...	22:00
Innovia	the monqiboy user folder is where my public_html (my site is)	22:00
Ilari	Innovia: Invalid URLs get very different error message...	22:00
	← kumbayo left	22:00
Ilari	Innovia: What 'ls -ld objects' shows (inside that repo)?	22:01
Innovia	yeah the url is absoulete path /home/monqiboy/public_html/mor	22:01
	let me check	22:01
	→ henux joined	22:02
Innovia	drwxr-xr-x 258 git git 4096 Oct 19 19:42 objects	22:02
henux	I accidentally committed a change to a local repo, how do I get that back?	22:02
tango_	git reset HEAD^	22:02
henux	I run that command verbatim?	22:02
tango_	yes	22:03
Innovia	Ilari: this is what i get drwxr-xr-x 258 git git 4096 Oct 19 19:42 objects	22:03
Ilari	Innovia: It should have writing permitted for git user...	22:03
henux	That didn't do anything	22:04
Innovia	Ilari: i just created a file inside mor.git from the remote side	22:04
henux	It just gave me a list of files with local changes	22:05
	No files were modified	22:05
Ilari	henux: But it uncommitted last change... See 'git diff HEAD'...	22:05
Innovia	Ilari: here's my git config file http://www.pastie.org/295911	22:06
	maybe you can spot something?	22:06
henux	git is very confusing	22:06
Ilari	Innovia: Can you paste what push attempt shows?	22:07
Innovia	yes	22:07
	check it now just refresh the page	22:08
henux	Okay how do I see the list of commit messages for the past N commits?	22:09
Innovia	henux: git log	22:09
	then just copy the commit # and do git reset --hard and the commit #	22:09
	@least that's how i do it	22:10
	Ilari: you can refresh the page i gave you	22:10
Ilari	Ouch... The error message in earlier versions was sightly more confusing looking, but more informative...	22:11
Innovia	i have 1.6 installed	22:11
Ilari	Innovia: Hmm and 'find /home/monqiboy/public_html/mor.git ! -user git' gives nothing?	22:13
henux	Okay thanks	22:13
Innovia	Ilari: it shows the content of the repo	22:14
Ilari	Innovia: Hmm... "pack-objects died with strange error"...	22:14
	Innovia: AFAIK, pack-objects is not invoked on remote side, it is invoked on LOCAL side.	22:14
	Innovia: On push anyway...	22:14
Innovia	should i try to git add . and commit again?	22:15
Ilari	Innovia: Maybe check permissions of repository you are trying to push from.	22:16
	Innovia: You should probably have write permissions there...	22:16
Innovia	i have innovia staff maybe thats my issue	22:17
	Ilari: drwxr-xr-x 14 Innovia staff 476 Oct 19 18:15 .git	22:18
Ilari	Innovia: Including stuff like .git/objects and .git/objects/pack?	22:18
	← charon left	22:18
Ilari	Innovia: That seems correct...	22:19
Innovia	Ilari: check this http://www.pastie.org/295911	22:19
	← lydgate left	22:20
Innovia	my packed ref has rw r r	22:20
	Ilari: if it would be easier i will ftp the mor directory and create the git --bare on the server then clone it back	22:20
	← rubydiamond left	22:21
Ilari	Innovia: Try creating another repo locally (mkdir ../testrepo; cd ../testrepo; git init --bare) and pushing to it (git push ../testrepo master)... Does that work?	22:22
Innovia	yes	22:23
	← tongueroo left	22:23
Innovia	Ilari: if that was a local push then yes i just did what you said and it worked	22:24
Ilari	Innovia: Well, that test repo is FTPable if you want to FTP it to server...	22:24
	← Gitzilla left	22:24
	← _graham_ left	22:27
	← pehlert left	22:28
comp	does git-rerere remember multiple merge resolutions which can be applied in one merge from "topic" to "master" ?	22:28
	(I guess it works that way)	22:29
Innovia	Ilari: i am trying to create a new git init --bare inside my mor and it just dumps all the git folders in there and give me a out of memory error how do i create the --bare	22:29
	do i have to create a folder for it?	22:30
	→ mithro joined	22:30
Ilari	Innovia: Huh, 'git init --bare' gives out of memory error?	22:31
Innovia	yes	22:31
	Ilari: why does it have to be a bare repo?	22:32
	this bare is so complicated, i keep getting weird errors like: fatal: This operation must be run in a work tree	22:33
mutex	what ?	22:33
	that is pretty unusual	22:33
Ilari	Innovia: Well, strictly it doesn't, but pushing to checked out branch either errors out or screws the repository...	22:33
Innovia	all i need is to be able to update my code on the website so that my client can see the changes	22:34
parasti	what's unusual? that a bare repo doesn't have a working tree?	22:34
Ilari	parasti: Probably OOM from 'git init'...	22:35
	← DrFrasierCrane left	22:35
	← ereslibre left	22:36
Innovia	Ilari: i thought that this would be the best way to work with changes to a website	22:36
Ilari	Innovia: Well, don't you have that bare test repo. You could FTP that to initilize the repo on server...	22:36
Innovia	Ilari: o.k how do i go about this	22:37
Ilari	Innovia: Tar that testrepo directory, copy the tar to server and extract...	22:38
	← chris2 left	22:38
Innovia	here's where i lost you, i have my code in the mor directory, should i just mkdir testrepo, git init --bare inisde that folder then ftp an empty repo to the server?	22:39
	Ilari:	22:39
	→ albertito joined	22:39
Ilari	Innovia: Nope, push to that repo as well (since that push works)...	22:40
henux	When I try to set my user.name by `git config --global user.name "My Name"` git gives me "warning: user.name has multiple values" and only my first name is shown in commit logs. How to fix this?	22:40
Ilari	Innovia: And then FTP the populated repo...	22:40
	Innovia: There are three pitfalls in copying git repos as files: 1) Doing so invalidates working tree cache data (it needs to be refreshed after copy) 2) Corruption is propagated 3) Copying stuff with remote tracking branches may give unexpected results (if repo is used for cloning)...	22:43
Innovia	Ilari: http://www.pastie.org/295911, i can google this if you are tired	22:43
	well i think im gonna use git only as a local thing, if it can push the changes and apply them to the files on the server (not just the ref)	22:44
	if it can not	22:44
Ilari	Innovia: You created bare repository inside another repository?	22:45
Innovia	???	22:45
	no	22:45
	→ dantrell joined	22:45
Innovia	mor is not a repo	22:45
	mor is the code for the website	22:45
dantrell	Is there a way to change a name on a commit? Like from unknown to <actual name>?	22:45
	← ph^ left	22:46
Ilari	Innovia: Ah, both the source and target of push must be valid repositories for it work...	22:46
parasti	dantrell: commits don't have names	22:46
	← trochala left	22:46
Innovia	so i should create a regular repo inside more and the a bare inside of mor_repo	22:46
dantrell	parasti: But there's a name (of the person who made the commit) assoicated with every commit?	22:47
parasti	dantrell: ah, indeed :)	22:47
dantrell	parasti: How do I change that?	22:47
parasti	I believe you can use the environment variables documented in man git-commit-tree	22:48
Gitbot	parasti: the git-commit-tree manpage can be found at http://git.or.cz/man/git-commit-tree	22:48
Ilari	Innovia: Don't you have local git repo for mor already?	22:48
Innovia	wiped it out , i just recreated it and push that into mor_repo (bare)	22:48
Ilari	Careless and destructive recovery attempts are a major cause of data loss...	22:49
parasti	dantrell: if you don't have your name/e-mail set in config, have a look at man gittutorial as well	22:50
Gitbot	dantrell: the gittutorial manpage can be found at http://git.or.cz/man/gittutorial	22:50
dantrell	parasti: It isn't for me, working with a new dev who is using git.. on windows... So his commits say unknown. He has it set now but I'd like to change the previous ones.	22:51
	← bdiego left	22:51
dantrell	So I'll at commit-tree. Thanks.	22:51
	look at*	22:51
	→ jackdempsey joined	22:53
	← Vortex35 left	22:54
	→ Vortex35 joined	22:54
parasti	dantrell: if there's a lot of history, git filter-branch --env-filter might be preferable... but it's a bit complicated	22:55
	← alb left	22:55
	← robin_ left	22:56
	→ robin_ joined	22:56
	→ dgrazier joined	22:56
	→ WALoeIII joined	22:57
	→ abbe joined	22:57
abbe	hi all	22:57
	I'm getting an error, while doing 'git pull' or 'git fetch'	22:58
	fatal: cannot pread pack file: No such file or directory	22:58
	← rtomayko_ left	22:58
abbe	I'm running git-1.6.0.2 on freebsd 8-current	22:58
	→ rtomayko joined	22:58
offby1	sounds like the remote repository is messed up	22:58
	→ proxie joined	22:59
Innovia	Ilari: o.k i have the mor_repo(bare) up and extracted now i should just clone it?	22:59
abbe	offby1, thats gnulib	22:59
	← bryanray left	22:59
Ilari	Innovia: You can try cloning it if you want to test cloning...	22:59
Innovia	i'd like to push to it	22:59
abbe	offby1, so no problem at my end, right ?	23:00
offby1	doubt it, but I'm not sure	23:00
	abbe: why don't you gimme the URL and I'll try it	23:00
	unless it's an ssh:// URL, in which case I won't have a login	23:00
Innovia	Ilari: what about the code that is already in the server should i just do commit there	23:00
Ilari	Innovia: You can't commit to bare repo (using the standard methods anyway)...	23:01
abbe	offby1, git://git.savannah.gnu.org/gnulib.git	23:01
	← koke left	23:01
abbe	offby1, thanks in advance :)	23:01
Innovia	how does it change the code on the server then?	23:01
	← qrush left	23:01
Ilari	Innovia: Bare repos are usually updated by pushing...	23:01
	← Beowulf_ left	23:02
Ilari	Innovia: There are other ways, like using bundles...	23:02
Innovia	hmm, a small missing peice for me here is that i still have files in my mor dir on the server	23:02
	how are those files knows of mor_repo (the bare one)	23:03
offby1	abbe: seems to be working OK	23:03
	worked fine :-\|	23:03
teolicy_	I'm getting this ("hooks/post-update: gitosis-run-hook: not found") when using gitosis, after pushes. Is this necessarily an error message, or just a warning that no hooks are defined?	23:03
	→ quoin joined	23:03
	← Yuuhi left	23:04
Ilari	teolicy_: Pushing to gitosis-admin.git?	23:04
teolicy_	Yes.	23:04
	(uhm, to gitosis-admin, I don't know where the '.git$' is)	23:05
abbe	offby1, hmm..., okay	23:05
Ilari	teolicy_: Is gitosis-run-hook in the same place as gitosis-serve?	23:05
abbe	offby1, I'll try again	23:05
teolicy_	Yep.	23:06
Innovia	Ilari: thanks for you help man but i i' giving up, this is not how i thought git works, so git is not for a maintaining live website with changes on it from your local machine I see now that git is just a collaboration tool	23:06
	→ bobmcw joined	23:07
teolicy_	Though I know I have path issues on that gitosis installation, I had to edit my authorized_keys file to change 'gitosis-serve' into '/usr/bin/gitosis-serve'.	23:07
Ilari	Innovia: Git used for maintaining live website? Ugh....	23:07
Innovia	among local repo	23:07
abbe	any ideas anyone ?	23:08
Innovia	i thoought i can push my changes to the remote repo , and then the fles will be updated on the server as well	23:08
	it is just a repo, local or remote, i appreciate your help man	23:09
Ilari	Innovia: That's very nontrivial to do. It requires some tricky stuff in hook...	23:09
teolicy_	Again, I'm not sure why git's shell has no path. When I 'sudo -i -u git' I have a proper PATH.	23:09
Ilari	teolicy_: What's '-i' to sudo? You probably do need login mode...	23:10
	← Marmouset left	23:10
Innovia	but generally speaking i have a remote repo and i push from my local , cloning on a different machine (my desktop) would be in sync with what i had on the laptop ( untill the next commit and pull) right?	23:10
abbe	offby1, git index-pack -v --stdin <.git/objects/tmp_pack_&X8a3H	23:11
Ilari	teolicy_: Ah, -i is login mode...	23:11
abbe	offby1, resulted in fatal: pack is corrupted (SHA1 mismatch)	23:11
Ilari	teolicy_: But interactive logins don't execute .bash_profile!	23:11
offby1	abbe: ?	23:11
Ilari	noninteractive, that is.	23:11
offby1	abbe: so the remote had an evil file in objects ?	23:11
	← xaiki left	23:11
Ilari	Its temporary pack, so it might not be complete...	23:12
	← pygi left	23:12
	→ pygi joined	23:12
abbe	Ilari, okay so how to figure out if it is client side's problem or server side	23:13
	→ qrush joined	23:13
	← dgrazier left	23:15
	← bobmcw left	23:15
abbe	on my gentoo gnu/linux with git-1.5.6.4, I got fatal: Out of memory, malloc failed	23:16
	← qrush left	23:16
abbe	I simpled 'rsync -a'ed gnulib repo from my FreeBSD 8-CURRENT installation to Gentoo GNU/Linux installation.	23:17
	← WALoeIII left	23:19
	← fujin left	23:20
	→ fujin joined	23:20
abbe	git index-pack -v .git/objects/pack/pack-ee521e12c9e959eab07e67db3d9629c5aa0dd21d.pack	23:25
	also didn't resulted in any error	23:25
	→ xaiki joined	23:27
abbe	oops, power cut, see you later	23:27
	← abbe left	23:27
	→ kef joined	23:28
	← Innovia left	23:30
	← alexross left	23:31
	← markelikalderon left	23:32
	→ qrush joined	23:33
teolicy_	Ilari: I tried both with and without -i, both have proper paths. Never mind, I appreciate your help, but I think I should go to sleep and read more of gitosis' source.	23:33
	I don't like it sneaking up behind my back like this, I need to know what its doing.	23:33
	offby1 sneaks up behind teolicy_	23:35
	→ markelikalderon joined	23:39
	→ bryanray joined	23:39
	← qrush left	23:40
Tv	hehe, go to sleep and read more source	23:40
	← proxie left	23:43
	← robbyonrails left	23:47
	← henux left	23:48
	← johnw_ left	23:50
	→ proxie joined	23:54
	← bryanray left	23:55
	← jesselucas_ left	23:56

Logs Search ←Prev date Next date→

Channels Documentation