IRCloggy #git 2008-05-26

Provider of IRC logs since 2005.

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

	2008-05-26
	→ schacon_ joined	00:01
	← pflanze left	00:02
	← cypher23 left	00:08
	→ blix_ joined	00:09
	← BabelO left	00:12
	→ fhobia joined	00:14
	→ johnw joined	00:17
	← devogon left	00:21
	→ ironfroggy joined	00:26
	→ sintaxi joined	00:33
	← larsl left	00:35
sintaxi	Im using git on a rails project and I just installed a plugin that was hosted on a git repo. now that plugin does not seem to be a part of my main apps repo. anyone know how to resolve this issue?	00:35
	← felipec left	00:45
	← mugwump left	00:50
	← sbeyer left	00:54
	→ peeja joined	00:54
	← sintaxi left	00:56
	→ chendo[work] joined	00:59
chendo[work]	New to git here; I want to revert a file to the previous commit	00:59
	git revert seems to revert a whole commit	00:59
	← schacon_ left	00:59
chendo[work]	git reset -- <file> did what I wanted	01:01
	... or not	01:01
jmworx	is it possible to do a git-svn dcommit if I merged branches in the trunk? If so, what does it do? (both to svn and to my git repo)	01:01
	→ doener joined	01:06
	← tcoppi left	01:07
	→ impulze joined	01:08
	← kanru left	01:09
	← IRSeekBot left	01:09
impulze	hi there. i used git-svn clone somerepos -r startref:HEAD and after that i wanted to use git-bisect start which shows "fatal: ref HEAD is not a symbolic ref" did i screw up or did i just forget to invoke something before bisecting?	01:10
	→ Mekham1 joined	01:17
	← doener_ left	01:17
Mekham1	http://enigmaengine.wordpress.com/2008/05/25/bring-back-the-fairness-doctrine/	01:17
	← Mekham1 left	01:17
	→ spearce joined	01:20
	→ tcoppi joined	01:23
	← Ken69267 left	01:28
	← AndyP left	01:31
	→ ajonat_ joined	01:39
	← ajonat left	01:39
	→ Ademan joined	01:40
	← johnw left	01:41
	← goibhniu left	01:43
	→ ggeecko joined	01:45
	← ironfroggy left	01:54
	← kukks left	01:55
	← radarek left	01:58
	→ bdiego joined	01:59
	← FunkeeMonk left	02:01
	← bdiego left	02:07
	→ schacon_ joined	02:08
	→ krawek joined	02:09
	→ kewark joined	02:16
	→ imyousuf joined	02:17
	→ rwanderley joined	02:22
	→ lirit joined	02:23
	→ bdiego joined	02:29
	← lirit left	02:31
	← krawek left	02:32
	→ _krawek_ joined	02:33
	→ imyousu1 joined	02:41
	← imyousuf left	02:43
	imyousu1 → imyousuf	02:43
	← kewark left	02:51
	→ spearce` joined	02:53
	← EiNZTEiN left	02:56
	← nDuff left	02:59
	← rwanderley left	03:05
	← csc` left	03:06
	→ spearce`_ joined	03:07
	← spearce left	03:10
	← eternaleye left	03:10
	← agib left	03:15
	← spearce` left	03:24
	→ csc` joined	03:27
	spearce`_ → spearce	03:32
	← schacon_ left	03:34
	→ vbabiy-laptop joined	04:04
	← cmarcelo left	04:04
	← chjunior left	04:08
	→ tjafk2 joined	04:09
	← serfurj left	04:13
	→ serfurj joined	04:13
	→ eternaleye joined	04:17
	← hacim left	04:24
	← bobesponja left	04:26
	← tjafk1 left	04:26
	→ ajonat__ joined	04:32
	ajonat__ → ajonat	04:32
	← _krawek_ left	04:34
	← vbabiy-laptop left	04:35
	← madewokherd left	04:39
	→ MrDarkUser joined	04:39
	→ mugwump joined	04:41
MrDarkUser	I'm looking for a better tutorial for git than the ones directly listed on the home page	04:44
	step by step for making a new project... I think I've seen all the commands at the tutorials, but I'm not quite good enough with cvs or svn to have helpfull experience.	04:45
	→ mediogre joined	04:45
	→ clsdaniel joined	04:51
	← ricky left	04:51
	→ ricky joined	04:51
	← ajonat_ left	04:55
angasule	MrDarkUser: the lack of knowledge of cvs and svn is more likely to help than hinder. I recommend you create a test repository and play with it, to see exactly what each command does	05:00
	→ angasule_ joined	05:10
	← brendon left	05:22
	← angasule left	05:25
	→ robin joined	05:30
	→ schacon joined	05:34
jdavis	how do I export a git repo, similar to "svn export", so that it doesn't give me the ".git" directory?	05:36
spearce	jdavis: `git archive --format=tar HEAD \| (cd ../otherdir; tar xf -)`	05:38
jdavis	spearce: thanks	05:38
imyousuf	spearce: just wanted to thank you in person about the jgit licensing :)	05:38
	now there is a possibility of us using it in our commercial product, thanks	05:39
spearce	imyousuf: given that irc is as in-person as i'll get with most folks i meet on online, i guess you are welcome. :)	05:39
imyousuf	:)	05:39
spearce	if you guys fix bugs in jgit, or add interesting features, we'd certainly appreciate you feeding them back to the project. :)	05:39
imyousuf	sure	05:42
	I will ensure that :)	05:42
	I have been advocating for GIT from a product of ours for some time	05:42
	← nkallen left	05:42
imyousuf	the only obstacle was a IDE integration	05:43
spearce	well, jgit/egit isn't a full solution. yet.	05:43
imyousuf	I am working on one for NetBeans with JGit	05:44
	→ kixx joined	05:44
imyousuf	it was the basic features we need	05:44
	with NetBeans UI it will be perfect :)	05:44
	→ priidu joined	05:45
spearce	fortunately we've been trying to keep the bulk of our code down in jgit, where netbeans could use it.	05:46
	→ nkallen joined	05:46
imyousuf	yeah, I am grateful about it :)	05:46
	→ nud joined	05:47
	← bdiego left	05:49
	← blix_ left	05:50
	← robin left	05:59
	→ spearce` joined	06:06
	← schacon left	06:07
Ademan	if i pushed a change to a local repository owned by another user, and one of the commit hooks fired, what user would it be running as?	06:11
	→ deavid joined	06:13
Ilari	Ademan: It would be the user pusher logs as to push destination in ssh:// or whatever user git-daemon runs as in push over git:// (not recommended). HTTP push will not run hooks.	06:16
	→ IRSeekBot joined	06:17
Ilari	Ademan: For local pushes, the hooks in destination repository are run by the pusher...	06:19
Ademan	ah thanks	06:20
	kinda unfortunate, i hoped they were run by the owner, but i suppose it's no big deal...	06:20
	← spearce left	06:22
	→ namenlos joined	06:24
MrDarkUser	angasule_ trying that ;) Oh I so want IDE integration. It looks like Git isn't very mature. (duh..)	06:25
	→ thiago_home joined	06:25
Ilari	Ademan: IIRC, Spearce made a hacked version of git-receive-pack that used suid and setuid() to run hooks and do writes as repository owner... Don't know if patches are available though...	06:26
	→ Xyri0 joined	06:26
MrDarkUser	Git Gui keeps asking for Revision Expression and stuff like that, I'm of course looking it up, but I don't have any idea why/what that would be.	06:26
	← johan-s_ left	06:27
	← igorgue left	06:28
Ilari	MrDarkUser: revision expression can e.g. be name of branch, name of tag, fully qualified refname (refs/heads/foo, etc...) or raw commit SHA-1 (or its abbreviation).	06:29
MrDarkUser	LOL, I tried just giving it a few random characters (I have nothing defined) and it wouldn't take em.	06:30
	→ anderbubble joined	06:30
	→ adante_ joined	06:30
MrDarkUser	I thought it was just a sort of identifier that I was specifying... and it seemed like it was being super selective	06:30
	← nkallen left	06:31
MrDarkUser	so, when i was making a new branch, and it doesn't take a name that I'm giving it I get confused	06:31
Ilari	MrDarkUser: And then there can be nth parent (^n) and n iterated first parents (~n) operators appended to the end...	06:31
	MrDarkUser: Doesn't it ask for new branch name in seperate textbox?	06:32
gitster	"SPECIFYING REVISIONS" in git-rev-parse(1)?	06:32
Ilari	MrDarkUser: And one can select the base branch from the list of branches (there are three radio boxes that change what is show)...	06:34
	→ mjc joined	06:35
MrDarkUser	urm.. /var/git/t1test is my repository.. and I think I only have the main branch and no files	06:35
	← mjc left	06:35
MrDarkUser	I'm trying to use the gui... I got a bit farther with the comand line on my other repository	06:36
	← anderbubble left	06:36
	← ben_h left	06:36
Ilari	MrDarkUser: You pushed to repository with working tree?	06:36
	← Xyrio left	06:37
	Xyri0 → Xyrio	06:37
MrDarkUser	I think I need to shut up now... It's pretty easy to clone the git home page repository	06:41
	with the command line, but the gui seems nuts.	06:41
	← nud left	06:44
	→ advany joined	06:45
	← adante left	06:46
	adante_ → adante	06:46
Auris-	guis are for nuts.	06:46
	→ BabelO joined	06:48
	→ bentob0x joined	06:54
	→ charon joined	06:55
	← clsdaniel left	06:55
	→ FunkeeMonk joined	06:57
	→ johan-s_ joined	06:57
	→ sbodo_w joined	07:00
	→ ph^ joined	07:01
	← statim left	07:04
	→ redondos joined	07:07
	← rphillips left	07:09
	← nipuL left	07:10
	← chendo[work] left	07:11
	← pete__c left	07:12
	→ ph^_ joined	07:15
mstormo	pasky: ping	07:18
	Anyone else which has admin access to repo.or.cz?	07:19
	→ aarapov joined	07:20
	→ nud joined	07:20
	← johan-s_ left	07:21
	← ph^_ left	07:24
	→ ph^_ joined	07:24
	← ph^_ left	07:25
	→ mcella joined	07:30
	← mcella left	07:32
	→ mcella joined	07:32
	→ CrawfordComeaux joined	07:51
	→ statim joined	07:51
	← csc` left	07:52
	← thiago_home left	07:54
floppyears	hi guys	07:59
	what are some best practices when it comes to committing ?	07:59
	← TuxCM- left	08:00
	← sbodo_w left	08:00
jengelh	Don't commit crimes.	08:01
	← aarapov left	08:02
frsk	You know it's early in the morning when you find yourself laughing of that :)	08:02
CrawfordComeaux	why git and not mercurial?	08:05
thiago	why mercurial and not git?	08:05
	I can ask the same question.	08:05
mstormo	lol	08:05
	→ cypher23 joined	08:06
eMBee	good afternoon	08:06
imyousuf	floppyears: in git branches, cherry-pick, filter-branch patches make it easy to make your own convention and tolerable to mistakes	08:06
thiago	good morning	08:06
jengelh	CrawfordComeaux: because git is better :p	08:06
imyousuf	so as long as the commit is representing a unit change I think it would be ok :)	08:06
eMBee	is there a tutorial for using git commit --interactive? i thought it would ask me if i want to commit one hunk at a time, but i can't see how to do that	08:07
Ademan	is there any git - mercurial interoperability around?	08:07
CrawfordComeaux	what's with coders always needing to be so damned sarcastic? it's 3am...I just would like to know why some people choose git over mercurial, or vice-versa...	08:07
mstormo	CrawfordComeaux: Topic branches are more light-weight in Git than in Mercurial. Though the patch queue concept is a neat idea..	08:07
thiago	Ademan: hg convert and hg-fast-export	08:07
Ademan	thiago: hrm, alright, thanks, i'll have to write that down lol	08:08
thiago	mstormo: I've been considering writing patch queue support in Git	08:08
	by using refs/patchqueue/name...	08:08
	it would work better than the solution that the hg developers came up with	08:08
mstormo	thiago: That's be very nice! Please do! ;-)	08:08
	thiago: I'll give you a cookie	08:08
thiago	my only problem is that I need to mark the initial commit somehow	08:08
eMBee	thiago: why would it work better? some inherent feature of git that makes it better?	08:09
thiago	or, to put it another way, I need to know which branch a patchqueue is meant to be on top of	08:09
	eMBee: yes, the git repository design	08:09
	← redondos left	08:09
eMBee	ah	08:09
thiago	on hg that's easy: there's only one branch in each repo (in usual setups)	08:09
eMBee	doesn't stgit do something similar?	08:09
	← charon left	08:10
thiago	never used it	08:10
floppyears	jengelh: hehe	08:10
jengelh	stgit is cool	08:10
thiago	a patch queue is nothing more than another branch in Git	08:10
	in fact, it is the recommended working procedure for Git	08:10
eMBee	how can i select parts of my current changes for a commit?	08:11
thiago	I'd be just making that a set of commands that match the Hg way	08:11
mstormo	eMBee: Parts of one file (hunks), or certain files?	08:11
thiago	in git we recommend that you do: checkout master, pull, checkout your branch, rebase	08:11
eMBee	mstormo: hunks	08:11
floppyears	imyousuf: I'm just wondering if it's best to have a bunch of small commits or only commit when I have a finished editing the file	08:11
thiago	that's exactly what the patchqueue concept does	08:11
jengelh	but the git porcelain itself is rather hard to work with in a stack fashion	08:11
mstormo	eMBee: If it's just the files, just git add <files> && git commit	08:11
thiago	eMBee: git add --patch	08:12
mstormo	eMBee: ^^^ :-)	08:12
jengelh	like, when amending a patch three commits downwards..	08:12
	you would use rebase with the normal git porcelain, which is a bit more work	08:12
imyousuf	floppyears: as I said, I am in favor of making each commit unit and not effeting any other task	08:12
	for this having a unit task is also important :)	08:12
	→ redondos joined	08:13
eMBee	thiago: thank you!	08:13
imyousuf	floppyears: concept of editing a file and making a commit are different.	08:13
Arrowmaster	when using 'git-archive tag' with .gitattributes export-subst, is there any $Format:$ that will give the tagname?	08:14
mstormo	floppyears: It all depends on you. Normally, it's nice that one commit is self-contained. Meaning that a commit doesn't obviously break something. However, you can still do many smaller commits locally, before you 'git rebase -i', and squash, reorder and change the commit msgs when you are done, and then finally push them into the public	08:14
	← ricky left	08:14
jengelh	Arrowmaster: UGH.	08:15
	Arrowmaster: Oh, git-describe will.	08:15
	→ ricky joined	08:19
	← spearce` left	08:22
	→ blondandy joined	08:27
	→ Eludias joined	08:28
	← mcella left	08:28
	← kixx left	08:28
	→ mcella joined	08:29
floppyears	thanks guys	08:29
mstormo	sure, np	08:31
	← blondandy left	08:31
jengelh	pnp.	08:31
	npn?	08:32
	;p	08:32
	→ radarek joined	08:33
thiago	fet are cooler	08:33
jengelh	hmk gates are	08:35
	→ ben_h joined	08:35
	→ hein2 joined	08:35
	← Sho_ left	08:35
	hein2 → Sho_	08:35
	← CrawfordComeaux left	08:36
hliusv561	thiago: MOS, MIS or MES? :-)	08:36
	thiago: Oh, and of course J... :-)	08:36
	→ yesudeep joined	08:37
thiago	MOS	08:38
jengelh	moosefet :)	08:38
thiago	much easier to calculate than PNP and NPN since there is no current in the base. There is no base :-)	08:38
jengelh	that's because all your base are belong to us	08:39
	→ ebel joined	08:39
	← ben_h left	08:40
	→ ben_h joined	08:41
hliusv561	Not to mention the quadratic (as opposed to exponential) dependency of current to voltage... Mixing exponential to linear terms easily creates equations with no solution in terms of elementary functions...	08:42
	← ben_h left	08:47
	→ putrum joined	08:53
	→ koke joined	08:54
	← CodeOfficer left	08:56
hliusv561	As rough rule: MOSFET => Silicon, MISFET => Carbon, MESFET => Gallium Arsenide. :-)	08:57
Ademan	the man page for git-svn mentions /pub is that in the fhs? i didn'	09:02
	t think it was	09:02
	yeah i don't see it	09:03
	→ wgaffa joined	09:04
mstormo	grmbl..	09:06
	Anyone know how I can update a remote HEAD with a symbolic ref?	09:06
	→ Kbyte joined	09:07
hliusv561	mstormo: And presumably you can't get shell on that machine with account allowed to write to it?	09:08
mstormo	nope, it's on repo.or.cz :-)	09:08
	← Kbyte left	09:08
	→ Kbyte joined	09:09
mstormo	For some odd reason, the HEAD that I get from the repo I just pushed _is_ a symref, and points to a non-existing branch	09:09
	it points to refs/heads/devel	09:09
	→ devogon joined	09:09
	← Kbyte left	09:09
	→ Kbyte joined	09:10
	← radarek left	09:11
hliusv561	mstormo: And presumably 'git push <repo> HEAD' does something wonky...	09:11
Ademan	any git-svn users know if this is a good way to go about things: git-svn clone the remote svn repos, then have myself and a few other developers interact with this git repository as though it were our collective public git repository and if need be, commit upstream using git-svn rebase and git-svn dcommit	09:13
thiago	no, it isn't.	09:14
	you should commit to svn, not push to git	09:14
hliusv561	Ademan: Git-svn is not suitable for distributed development. SVN is just too limited.	09:14
	→ radarek joined	09:16
	→ gitte joined	09:16
mstormo	hliusv561: Well, my local repo shows the correct HEAD, and result on 'git branch', but 'git push --force HEAD' says "Everything up-to-date". However, a fresh clone from that remote repo ends up with a different symref (non-existing) in HEAD, and says "Warning: Remote HEAD refers to nonexistent ref, unable to checkout."	09:16
Ademan	thiago: doesn't that mostly defeat the purpose of using git-svn at all?	09:19
thiago	Ademan: not at all	09:19
	→ guardian joined	09:20
	→ johnw joined	09:20
thiago	Ademan: git-svn is very useful for branching, merge-tracking, annotating locally, etc.	09:20
	and not having to remember the svn commands	09:20
Ademan	haha	09:20
	thiago uses Git for his Git, Subversion and Perforce projects	09:20
thiago	the point being: only one person should be using git-svn in each git-svn import	09:22
	if you want more people to use the repository, they can, but they mustn't use git-svn. They should use git only.	09:22
	and they should know that they have to rebase against the SVN imports, never merge.	09:22
Ademan	thiago: well that's sorta what i'd saying, they'd push and pull to that single git-svn repos, which would communicate with upstream if need be	09:23
	well, that may not be what you're saying though...	09:23
thiago	that repos might be updating automatically from SVN	09:23
	but committing to SVN automatically is a bad idea	09:23
jengelh	svn is a bad idea.	09:23
thiago	one person should do the committing if necessary	09:24
Ademan	thiago: well it wouldn't be automatic, in fact that's why it's important that my development group has it's own internal 'master' copy	09:24
thiago	Ademan: http://code.staikos.net/cgi-bin/gitweb.cgi?p=webkit;a=summary is a good example	09:24
	Git-based development on top of a Subversion source, using git-svn	09:25
	as you can see from that webpage, the "svn/master" head is moving, but master isn't	09:26
Ademan	indeed	09:26
	i don't even see a master lol	09:27
thiago	right, I meant the main branch (called "qtwebkit" there)	09:27
	the one that gitweb shows by default	09:27
Ademan	ah	09:28
	yeah	09:28
	→ pergesu joined	09:28
thiago	what happens there is an interesting case: to submit to WebKit's mainline branches, you need the patch to be approved	09:29
	or you need to be a known committer	09:29
	so only a few people can cherry-pick and submit to SVN	09:29
	→ chris2_ joined	09:30
	chris2_ → chris2	09:32
Ademan	ergh, why does svn have to complicate things... lol	09:32
	→ pflanze joined	09:33
thiago	it doesn't.	09:33
	svn works very well if you're content with one-branch development	09:33
	and centralised	09:33
	← johnw left	09:36
	→ sbeyer joined	09:37
	← fhobia left	09:37
	→ jamessharpe joined	09:39
Ademan	thiago: well, my friend and i are working on a website using turbogears 2.0 which is under heavy development. I expect things will turn up broken very often, and we'll want to fix it right away, not waiting for upstream to approve patches and all that. I figured that would work with a single git-svn repository that spits out a working copy when needed, and two satellite normal git repositories.	09:40
	← Sho_ left	09:45
	← priidu left	09:45
	← imyousuf left	09:47
	→ priidu joined	09:48
mugwump	Ademan: probably bzr works better in cases like that	09:49
	if you want to use git you will have to have a single git-svn bridge	09:49
	← priidu left	09:51
	→ Sho_ joined	09:53
gitte	pasky: is it possible that you installed a submodule-unaware git on repo.or.cz?	09:53
	→ priidu joined	09:53
	→ Ademan_ joined	09:55
	← Ademan left	09:56
hliusv561	gitte: That would have to be 1.5.1.X or older than that (altough there are some submodule-related bugs fixed in later versions)... I don't think that's likely. Gitweb doesn't support submodules very well however...	09:56
	Ademan_ → Ademan	09:56
gitte	hliusv561: just try to clone msysgit.git, and weep when you do not get 1bdd3b0f84f2ffeb9f08fce80c8ddc3e76c17210.	09:56
	hliusv561: I finally found out that this is a commit of Junio's 'html' branch.	09:57
	pasky: it could be that it is only git-pack-objects (possibly git-upload-pack, too).	09:57
	← priidu left	09:58
Ademan	thiago: i dunno if my last message before i was disconnected made it through	09:59
	→ _graham_ joined	10:00
hliusv561	Ademan: The longish message about turbogears 2.0 at 0940Z?	10:01
	→ priidu joined	10:01
Ademan	yep	10:01
	so i guess it did make it through	10:01
	was anything said in response?	10:02
pasky	gitte: in ssh chroot or outside (gitweb, git-daemon)?	10:02
gitte	pasky: "git clone git://repo.or.cz/msysgit.git"	10:02
	ChanServ set mode: +o	10:02
	Mikachu changed the topic to: 1.5.5.2 \| Homepage http://git.or.cz/ \| GSoC 2008 Is On! \| Everyone asleep or clueless? Try [email@hidden.address] \| Channel log http://colabti.org/irclogger/irclogger_log/git \| Mailing list archives: http://marc.info/?l=git \| Gits on git: http://tinyurl.com/2xq3ke \| You want $ID?: http://tinyurl.com/yqpgv9 \| Need to paste something?: http://git.pastebin.com	10:02
	Mikachu set mode: -o	10:02
thiago	Ademan: "my friend and I are working..."	10:05
	that's the last	10:05
Ademan	thiago: indeed, did you have anything to say regarding that? because it seems to be the most logical solution, but it sounds as though... i can't or shouldn't	10:06
thiago	no	10:06
pasky	gitte: that's curious... but... hmm	10:07
	gitte: can you remind me about it later plz?	10:07
	pasky also hoped to finally install the ssh keychange patch but turns out he has no time for anything but the thesis for the last few days :/	10:08
gitte	pasky: when's your thesis over?	10:09
hliusv561	pasky: Hopefully you have systems in place for monitoring and dealing with bruteforcing attempts on SSH keys? :-)	10:10
gitte	hliusv561: he just said that he does not have time, so give him a break, will ya? ;-)	10:10
pasky	gitte: end of this week, but I'd like to finish the content part today so that I can send it around to few people for review	10:11
	Ademan brute forces access on pasky's machine	10:11
Mikachu	if you're worried about your project you can log into the admin page and add new users and remove the old ones	10:11
gitte	Ademan: no luck: only gits.	10:11
	Mikachu: I already removed all my endangered users.	10:11
Pieter	hmm	10:24
	perhaps we can include the "git describe" output in "git log", for the people that really want some sort of revision number	10:24
	→ Xyri0 joined	10:26
tokkee	Does anybody know the reason for the commas used in the Subject-line generated by post-receive-email? Imho it would be correct to leave them out...	10:27
	Pieter: tig does that for tags and heads.	10:28
	→ goibhniu joined	10:29
Pieter	post-receive-mail?	10:30
	← ricky left	10:30
Pieter	oh, the hook	10:30
	→ ricky joined	10:31
	← jas4711 left	10:35
Arrowmaster	being able to get 'git-describe' output in a $Format:$ replacement for git-archive would be great	10:35
	→ entropy joined	10:35
entropy	i just realized that several commits ago a file that was meant to be renamed ended up being cleared (zero byte size) in the destination. how can I (re)copy the deleted file from before it was deleted to overwrite the cleared copy I now have?	10:37
	← Xyrio left	10:37
	Xyri0 → Xyrio	10:37
gitte	Arrowmaster: is it not?	10:40
Arrowmaster	from the docs it doesnt seem like it	10:41
Pieter	I think you might be able to do that with gitattributes?	10:41
	← entropy left	10:41
Arrowmaster	you can do the replacement when using git-archive with gitattributes	10:42
	but afaik there isnt a replacement that gives the git-describe output	10:42
gitte	It's not hard to code. AFAIR I even posted a patch to the list, once upon a time.	10:43
	Arrowmaster: search for it, clean it up, and push for inclusion.	10:43
mstormo	Hmm, entropy fell out without an answer?	10:44
	heh, impatient guy	10:44
gitte	maybe he goes back to CVS.	10:45
mstormo	lol	10:45
Pieter	there's already	10:45
	E.g. the string $Format:%H$	10:45
	will be replaced by the commit hash.	10:45
tokkee	Pieter: Yes, the hook ;-)	10:45
gitte	Pieter: that is not exactly the output of "git describe".	10:46
Pieter	oh	10:46
	doh, I forgort what we were talking about	10:46
	forgort :)	10:46
	might be nice to add a git-describe option in --pretty for git log	10:47
	→ janm joined	10:48
	→ heidaro joined	10:50
	← lu_zero left	10:54
	→ lu_zero joined	10:54
gitte	Pieter: that is exactly what we were talking about.	10:57
	Pieter: hopefully, Arrowmaster searches the mailing list for my patch right now, brushes it up and reposts it.	10:57
Arrowmaster	i dont think thats possible	10:58
gitte	Arrowmaster: then I have nothing more to tell you.	10:58
	→ imyousuf joined	10:58
Arrowmaster	my c is very very rusty	10:58
	gitte 's manners, too	10:58
	→ qiyong joined	10:58
qiyong	How I put a website under SVN	10:59
	http://www.advogato.org/person/coywolf/diary.html?start=13	10:59
	any comments are welcome.	10:59
Pieter	gitte: do you know around when you posted that?	10:59
gitte	Pieter: no.	10:59
	Pieter: but I could imagine that searching the mailing list for "pretty" and "describe" _could_ turn out something useful.	11:00
Arrowmaster	http://article.gmane.org/gmane.comp.version-control.git/63392 ?	11:02
	→ charon joined	11:03
Pieter	it'd be nice if gmane could export a mbox of search results	11:03
	← ajonat left	11:05
gitte	Arrowmaster: I think that is too old.	11:07
Arrowmaster	november 07	11:08
	→ dsyzling joined	11:10
gitte	Arrowmaster: Given that there were massive restructurings in the code in February by Marco, I think this is a bit too old. You can try to start from there, though.	11:13
	← dsyzling left	11:15
	← advany left	11:16
	→ dsyzling joined	11:19
	→ ben_h joined	11:21
	→ koke_ joined	11:22
	← Fullmoon left	11:24
	← imyousuf left	11:27
	→ Fullmoon joined	11:34
	→ ironfroggy joined	11:35
	← ironfroggy left	11:37
	← koke left	11:38
	→ pclouds joined	11:43
	← bentob0x left	11:44
	← chris2 left	11:47
	→ mcella_ joined	11:55
	→ tobstarr joined	11:58
tobstarr	hi	11:58
	is there a command to repair a git repository?	11:58
	when I try to clone I get an error like that	11:58
	remote: error: remote: unable to find 888a2927b65889465ce7a1a71e87d37640a2b41bremote:	11:58
	remote: fatal: remote: unable to get type of object 888a2927b65889465ce7a1a71e87d37640a2b41bremote:	11:58
	← mcella left	12:00
thiago	git fsck	12:00
tobstarr	that command is not available on my server	12:01
	do you know the debian package for that?	12:01
	→ ironfroggy joined	12:02
	← deavid left	12:02
	→ icry0000 joined	12:03
tobstarr	missing blob 888a2927b65889465ce7a1a71e87d37640a2b41b	12:03
	broken link from tree f24e68956d05713b67a96730d2a44c8f97139239	12:03
	to blob 888a2927b65889465ce7a1a71e87d37640a2b41b	12:03
	but how can I fix that?!?	12:03
	Mutru_ → Mutru	12:04
	← ironfroggy left	12:04
	← flaguy left	12:05
	← koke_ left	12:05
Pieter	find the missing blob	12:06
	look for other copies of the repository if they have that object	12:06
thiago	run "git show 888a2927b65889465ce7a1a71e87d37640a2b41b" in other copies of that repo	12:07
	← ben_h left	12:08
tobstarr	git show says bad object	12:09
thiago	try another repo	12:09
Pieter	tobstarr: http://thread.gmane.org/gmane.comp.version-control.git/40893/focus=40896	12:09
	tobstarr: http://thread.gmane.org/gmane.comp.version-control.git/64091/focus=64213	12:10
	git ls-tree f24e68956d05713b67a96730d2a44c8f97139239 will show you what file is missing	12:11
	that might help with replacing it	12:11
	← pclouds left	12:12
	→ pclouds joined	12:12
tobstarr	thanx a lot	12:14
	I do have a copy of the missing file	12:14
	no I have to find out how to replace it	12:15
Pieter	git hash-object -w <file>	12:15
	that should be enough	12:15
	← namenlos left	12:16
	→ namenlos joined	12:16
	→ kukks joined	12:17
tobstarr	hmm	12:18
	Unable to add ... to database?	12:18
gitte	write-protected?	12:18
	← pclouds left	12:19
tobstarr	nope	12:19
gitte	(I mean the database)	12:20
tobstarr	where is it?	12:21
gitte	In .git/, where else?	12:21
tobstarr	the hole git folder is writable	12:21
jengelh	disk full?	12:21
	folders are for windows..	12:21
tobstarr	sorry	12:23
	disk is not full	12:23
gitte	Can you read <file>?	12:24
	(I.e. you did not really call "git hash-object -w <file>" _verbatim_, did you?)	12:25
	→ lcapitulino joined	12:29
gitte	Hey lcapitulino!	12:30
lcapitulino	gitte: hi!	12:32
gitte	lcapitulino: do you have a GSoC project this year, too?	12:32
	← fearoffish left	12:32
lcapitulino	no, I don't. I wouldn't have enough time to work on it :(	12:33
gitte	Are you employed by now?	12:33
lcapitulino	yeah	12:33
	how is git this year? Lots of students? :)	12:34
gitte	Yes, more than last year.	12:34
	→ redondos_ joined	12:34
lcapitulino	great	12:34
Pieter	I wonder if gittorrent's going to work out this year	12:38
gitte	Probably. We have a very capable mentor, and a very capable student.	12:38
	(Actually, we had 2 candidates, but one of them found another project he preferred to work on)	12:38
	→ deavid joined	12:41
Pieter	if all the gsoc projects work out, and we have builtin merge and rebase, the builtinification will be almost complete	12:42
	→ d0k joined	12:44
	→ ramontayag_ joined	12:44
gitte	Yes.	12:44
	gitte is happy.	12:44
	ramontayag_ → ramontayag	12:44
	← redondos left	12:45
	redondos_ → redondos	12:45
	gitte wonders if tobstarr is still there.	12:47
Pieter	repack.sh and pull.sh seem to be mostly option parsers :) so the big ones are am, bisect, stash, submodule and send-email	12:47
gitte	send-email will never be builtin.	12:47
	(Just a guess, but an educated one.)	12:47
	You missed mergetool and instaweb, BTW.	12:49
Pieter	yeah, it would be no fun and not very useful	12:49
	I didn't really consider instaweb as part of core git	12:49
ramontayag	hey everyone. i'm new to git but i know subversion. i started to try to move today (trying to host my repo at assembla.com.. but i think i entered the wrong passphrase. how do you change the passphrase? now, all i get is, when i try to upload code, is http://pastie.caboo.se/203401. is this a git thing or a linux thing?	12:52
halorgium	ramontayag: do you have a local branch called master ?	12:53
ramontayag	halorgium - eep, entering "git branch" doesn't show anything	12:55
Mikachu	is your head detached? did you make commits on it?	12:56
halorgium	Mikachu: he should still have a branch though, right?	12:56
Mikachu	unless he deleted it somehow? but it's weird	12:57
Pieter	halorgium: not if he never committed anything	12:57
Mikachu	ah right, it isn't a clone of an existing repo then?	12:57
halorgium	Pieter: you have an empty branch if you do 'git init'	12:57
	← nud left	12:57
	→ nud joined	12:57
Mikachu	halorgium: you can't push it until you make a commit	12:57
ramontayag	i can try to restart the whole thing. i guess i followed the assembla instruction w/o making sure what it was. i did "git init" then i tried to push it.	12:58
	→ hsuh joined	12:58
Tv	halorgium: refs/heads/master won't exist after git init	12:58
	→ koke joined	12:58
Mikachu	ramontayag: you have to commit before pushing	12:58
ramontayag	when pushing, i was asked to enter my passphrase, which it said it was wrong. then i started getting that error everytime i'd try to push.	12:58
	ah	12:58
Tv	halorgium: HEAD may point to it, but it doesn't exist	12:58
Pieter	halorgium: and thus git branch won't show anything	12:58
hsuh	hi guys.. is it normal for git to change file modes ? like i just cloned to a pendrive repo, then i git status and all files appear modified.. and then i git diff and the modifications are that the old mode was 644 and new mode 755	12:59
halorgium	Tv: ah, right	12:59
	i have the bash prompt which shows the current branch	12:59
	it was showing 'master'	12:59
Mikachu	hsuh: fat doesn't support executable permissions	12:59
ramontayag	oh	12:59
	takes getting used to :o	12:59
	hehe.. so, now that i'd added and committed	12:59
hsuh	Mikachu: whats the usual solution to things like this?	13:00
ramontayag	i can push it? :) will get back to you guys	13:00
Mikachu	hsuh: use another filesystem? :P	13:00
hsuh	i'm copying to windows at work...	13:00
Tv	hsuh: only put bare repos on the usb stick?	13:00
ramontayag	cool thanks halorgium, Mikachu	13:00
hsuh	Tv: what do you mean? just cp -a my repo?	13:01
	→ qubodup joined	13:01
Tv	hsuh: zip/tar/whatever the files?	13:01
	← qubodup left	13:01
hsuh	how unpractical is that? :)	13:01
Tv	hsuh: ehh, definitely no	13:01
	hell, tar/zip might be a lot faster than pure fat	13:01
	→ oliver3 joined	13:01
hsuh	what about git-dont-you-touch-file-modes?	13:01
Tv	fat directory operations are hellishly slow	13:01
	hsuh: it's not git doing it	13:01
hsuh	oh no?	13:01
Tv	hsuh: it's your mount command default options	13:02
hsuh	oh, i see..	13:02
Mikachu	either all files are -x or all files are +x	13:02
Tv	umask=	13:02
	most often -x is nicer for usb sticks	13:02
hsuh	hm	13:02
Tv	or fmask=	13:02
oliver3	Hi, me and another dev have been working on the same code (without realising), now we need to decide who's code to keep, and who's to drop, or whether we can merge the two branches together. His code is on his machine, mine on mine, how could I pull in his code without clobbering mine?	13:03
Tv	oliver3: git fetch	13:03
oliver3	Tv, thanks!	13:03
	:D	13:03
Mikachu	oliver3: that's the intended mode of operation with git	13:04
oliver3	Tv, that doesn't seem to have done anything at all...	13:04
	→ orion_ joined	13:05
hsuh	maybe ill try to format it as ntfs? would everyone be happy then ?	13:05
Tv	oliver3: if you want us to help, give details	13:05
	oliver3: (and read the man page)	13:06
	→ SRabbelier joined	13:07
oliver3	Tv, there are two of us currently working on this program. We have a central public git repository which we push to every so often. We have our own local clones of the central repo. We have both edited the same set of files, and need to decide who's copy to keep. I need to get a snapshot of his repo in a separate branch of my local copy.	13:07
Pieter	oliver3: push his changes to separate branch on the central repo	13:08
	then you can fetch that	13:08
oliver3	Hrm yes, thanks.	13:08
	:D	13:08
	← deavid left	13:08
	→ deavid joined	13:09
	→ EmilMedve joined	13:11
orion_	has anyone written a programming API around git (besides bash?)	13:11
Tv	orion_: yes	13:12
Pieter	there are like 3 ruby api's	13:12
	and some python projects have created their own git api	13:12
orion_	anything strongly typed like java?	13:13
Pieter	I don't think so, but there's jgit	13:13
Tv	wouldn't the eclipse plugin need something like that?	13:13
Pieter	which is a reimplementation of git in java, but in alpha stage	13:13
Tv	i think the intersection of java and git is still pretty slim	13:13
Pieter	yeah, egit uses jgit	13:13
	gitte: False is capitalized in python :)	13:17
gitte	Okay.	13:17
	→ etnt` joined	13:17
	← putrum left	13:18
	← etnt left	13:18
	→ yesudeep1 joined	13:20
	← shenson_not_here left	13:21
	← icry0000 left	13:22
	← mcella_ left	13:31
	← yesudeep left	13:32
oliver3	How would I completely ditch my local changes, and replace with the contents of another repo?	13:33
	oliver3 bashes his head against the wall in despair	13:33
	→ glommer joined	13:34
Mikachu	oliver3: it would be more productive to read the documentation while you wait	13:38
oliver3	Mikachu, I _am_.	13:38
	I can't find anything like "git undo", to undo my accidental git pull	13:38
	I now have an incorrectly merged repo	13:38
	Half of my changes, and half of someone elses.	13:38
	Which are incompatible.	13:38
	I'm tearing my hair out here.	13:38
Mikachu	depending on if the merge succeded you want git reset --hard HEAD or git reset --hard HEAD^	13:39
oliver3	Mikachu, none of those commands did what was needed.	13:40
	I have a file which both me and another dev have edited.	13:40
	I need his copy, not mine.	13:40
	I have my copy, his copy is in the central repo	13:40
	git fetch does _nothing_...	13:40
Mikachu	so you want to throw away all your work and use his?	13:40
	just to make sure	13:41
oliver3	Yes	13:41
	Pretty much	13:41
Mikachu	if you've run git fetch on his repo, his branch should be stored in the temporary FETCH_HEAD ref	13:42
	so you can git reset --hard FETCH_HEAD	13:42
oliver3	Didn't work	13:42
Mikachu	usually one would set up a remote with git remote add foo repo-url, and then git fetch foo, then git reset --hard foo/master	13:42
oliver3	Now I don't have his copy of the file, or mine, now I have an old version from before we both began work on the file.	13:42
Mikachu	then you probably didn't fetch the correct branch from his repo	13:43
oliver3	Mikachu, we have a remote repo, he has already pushed his changes.	13:43
	I'm trying to get those changes.	13:43
	And at the same time, remove mine.	13:43
Mikachu	well, if you fetch the correct branch, it should work out	13:44
oliver3	I am...	13:44
Mikachu	is this repo publically available?	13:44
oliver3	There is only one branch on the central repo, master.	13:44
	← Aeternus_ left	13:44
Tv	oliver3: the thing is, we don't trust you when you say that	13:44
	oliver3: and you aren't providing any details	13:44
oliver3	Mikachu, yes, http://git.savannah.gnu.org/gitweb/?p=kaiser.git	13:44
	Tv, what more details do you need?	13:45
	I want to clobber my changes, and get his.	13:45
	→ Aeternus_ joined	13:45
Mikachu	the super easiest way is to throw away your repo and reclone	13:45
	you can't fail then	13:45
Tv	oliver3: well things that "i'm now at commit <sha>, his head is <sha>, i ran this exact command, output is on this pastebin, but my head is still <sha> -- why?"	13:45
	s/that/like/	13:46
	← SRabbelier left	13:46
	→ SRabbelier1 joined	13:46
	SRabbelier1 → SRabbelier	13:46
	← hsuh left	13:47
oliver3	Tv, oh, my god. I _do_ have the master of the central repo. He hasn't added the damn file to his commit. brb going to stab him in the face.	13:48
Tv	git push other-developer off-his-chair	13:48
oliver3	Tv, Mikachu thanks for your help.	13:48
	Lol Tv	13:48
Mikachu	haha	13:48
	i was confused earlier when git said master was an ambiguous ref, but i only had one in git branch -a, it turned out i had a .git/master for some reason that i had to find by myself	13:49
	→ holizz joined	13:50
	→ qknight joined	13:51
qknight	i've accidentialy added a .svn dir to my git repo (didn't commit yet) how do i remove that?	13:51
	→ bentob0x joined	13:51
qknight	i don't want to delete the file as git rm does (or am i wrong?)	13:52
dato	qknight: git rm --cached .svn	13:52
holizz	git rm --cached .svn	13:52
qknight	thanks	13:52
	oh well	13:54
	git status said:	13:54
	# Changes to be committed:	13:54
	# (use "git rm --cached <file>..." to unstage)	13:54
Mikachu	git reset .svn would also work	13:54
qknight	sorry for not reading this .. ;P	13:54
ruphy	Tv: loool :D	13:56
qknight	Mikachu: even before i do a commit?	13:59
	← serfurj left	14:01
Mikachu	qknight: only then	14:03
	→ krawek joined	14:03
	→ mcella joined	14:05
qknight	Mikachu: ok	14:06
	thanks for the help!	14:06
	→ kewark joined	14:07
	← jstrom left	14:08
qknight	i've done: echo ".svn" > .gitignore in my project dir (where .git can be found as well) and git status tells me still that .svn is not tracked altough it shoul be ignored now	14:08
	oh no	14:08
	all right	14:08
	i moved the file into the wrong directlry	14:09
	how nice git is!	14:10
thiago	I don't think git and svn in the same checkout is a very good idea...	14:13
	wait, no, I'm thinking of a different problem	14:13
	it should be ok	14:13
	← deavid left	14:17
	→ schacon_ joined	14:23
	← krawek left	14:24
	← namenlos left	14:26
	→ cmarcelo joined	14:28
	→ deavid joined	14:31
Pieter	Beginning with the referenced commit, the git project cannot be checked out on Cygwin (and I assume cannot be checked out on Windows using msysgit, though I have not verified this) as this commit introduces the file "t/5100/nul." On Windows, the file name "nul" is reserved, regardless of path, and cannot be created or deleted. It serves essentially the same function as /dev/null.	14:32
	that's just genious	14:32
	kewark → krawek	14:33
thiago	yeah	14:33
	there are some reserved paths that Windows doesn't allow	14:33
	"aux" is another	14:33
Mikachu	you could crash windows 98 just by typing nul/nul in the run dialog	14:33
	or something along those lines anyway	14:33
	cygwin should have workarounds for things like that	14:35
	it's supposed to provide a posix environment, isn't it?	14:35
thiago	I don't think it handles filenames with characters not accepted	14:36
dionoea	Hi. Is it possible to checkout a branch from a clone in a seperate directory? (If that makes any sense)	14:36
thiago	you can't have a + in a filename either	14:36
	dionoea: git-new-workdir	14:36
Tv	Pieter: oh even better is that access to con: on windows used to hang	14:36
	← impulze left	14:36
Tv	Pieter: like, <img src="con:"/>	14:36
Mikachu	ah, maybe it was con, not nul	14:36
Pieter	:D	14:36
Tv	or file:///con$, something	14:37
dionoea	thiago: hum ... I don't have that command here (debian sid)	14:37
Pieter	dionoea: it's in contrib/ and not included by defaul	14:37
	t	14:37
dionoea	ah ok :/	14:37
	← bentob0x left	14:37
thiago	and you should use git 1.5	14:37
dionoea	git version 1.5.5.1	14:38
Pieter	dionoea: http://git.kernel.org/?p=git/git.git;a=blob;f=contrib/workdir/git-new-workdir;h=7959eab902d28bb3307c542514ca4c5f49deee0f;hb=2d3922dc611e4615bc2a8a151f2e382d44971aed	14:38
dionoea	thanks	14:38
	→ bobesponja joined	14:40
dionoea	awesome! that's exactly what I wanted. Thanks a lot.	14:42
Mikachu	just be sure not to check out the same branch in two workdirs	14:42
	→ Pistos joined	14:42
	→ madewokherd joined	14:42
Pistos	I know this is discussed extensively and in vary degrees of detail on the web, but I'm trying to wrap my head around some best practices for merging as considered from the old school angle of cvs/svn.	14:43
	← tobstarr left	14:43
Pistos	I've made some changes on a branch which I'd like to now pull into master. Should I use fetch, merge, or pull, or what combination?	14:43
Pieter	hmm	14:43
	I don't get git fast export and renames	14:43
Mikachu	pull = fetch + merge	14:43
Pistos	Mikachu: Okay, that's what the websites seem to say.	14:43
Mikachu	it doesn't matter if you use pull or merge	14:44
Pistos	So if I want to "play" the diff patch of the branch "onto" my master, but not commit yet, what do I execute?	14:44
dionoea	Mikachu: ok.	14:44
Mikachu	Pistos: it doesn't matter if you commit yet or not, but you can do git merge --no-commit thebranch	14:44
	Pistos: you can always amend the commit afterwards, or undo it	14:44
Pistos	Mikachu: I see.	14:45
	So then it's more "proper" to merge than to pull?	14:45
	→ unreal_ joined	14:45
Mikachu	generally you use pull when you want to merge something from a remote repo but not set up any tracking branches for it	14:45
	for merging local branches, "merge" and "pull ." are pretty much identical	14:45
	← unreal left	14:46
	unreal_ → unreal	14:46
Pistos	Mikachu: Okay, thanks, that clarifies things.	14:46
Pieter	blegh	14:46
Pistos	I appreciate your time.	14:46
Pieter	this is so broken	14:46
Mikachu	if you like to be roundabout, you can git fetch foo; git merge FETCH_HEAD	14:47
	← wgaffa left	14:54
	← ggeecko left	14:55
eMBee	good evening	14:55
	mugwump: remember our discussion about adding a different branch handling to git-svn? i am reading the code now and i can't really figure out where to start. are you familiar with git-svn at all?	14:57
	→ brosner joined	14:58
	→ bdiego joined	15:02
	← gitte left	15:06
	→ Ace_NoOne joined	15:08
Ace_NoOne	hi there - does Git use some kinda cache? after rebooting, the first git-status (or similar command) always takes a while (500-800 MB SVN repo)	15:09
broonie	The OS caches file accesses; git doesn't do any of its own, it relies on the OS.	15:10
Mikachu	after rebooting you sort of have to check the files to be sure they weren't modified	15:12
	even if you were to have some sort of cache + daemon	15:12
Pistos	Is there a sort of "git ls" I can do in refs/* to see what tags and branches and things I have?	15:14
Mikachu	'git branch (-a)' and 'git tag'	15:14
	or /bin/ls	15:15
Pistos	heh	15:15
	Alright. I only ask because I was trying refs/branches/branchname, but that was failing so I googled to learn that it's refs/heads/branchname	15:15
comp	greetings, will git modify files during merge if newly merged files are matched by .gitignore? or will it resolve this as a conflict?	15:15
	or just "ignore them" :)	15:16
Mikachu	.gitignore only affects untracked files when doing git status	15:16
	or git add	15:16
	→ Razerblader joined	15:17
	← Razerblader left	15:17
comp	my problem is that I want to pull from the upstream but I have already only part of that repo (usable as submodule) ... however I need to deal with things like modifying files in that submodule during pull in that "superproject"	15:18
Mikachu	i think that's a case of "you get to keep the pieces"	15:19
	→ kumbayo joined	15:20
comp	the problem isn't in updating the submodule itself, but in superproject possibly writing into the submodule	15:20
	i think the "better way" would be to merge it into the superproject and maintain only one repo, but how to do that? how to change those paths?	15:21
	→ ggeecko joined	15:21
	← Pistos left	15:22
comp	(from the submodule view, it's ./file1 , from the superproject it's some/long/path/file1)	15:22
	so cherry-pick probably won't work here	15:23
	← ramontayag left	15:24
	← SRabbelier left	15:24
	← schacon_ left	15:24
	→ SRabbelier1 joined	15:24
	SRabbelier1 → SRabbelier	15:25
	← EmilMedve left	15:25
Pieter	Hmm	15:31
	when I give "checkpoint" to git-fast-import, it's not really exporting marks	15:32
	→ igorgue joined	15:33
Ace_NoOne	thanks broonie and Mikachu (sorry, was called away there)	15:34
	→ schacon joined	15:39
comp	... and can git share object files between remotes? for example I want to have "clean" repo as master and add another remote (which is for example a fork) that has most of its blob objects identical to mine ... but it seems that git download everything for each remote separately ...	15:40
Mikachu	it only compares commit objects	15:41
comp	ah	15:41
	so if there are some old commits, any newly merged (in that remote) will have different SHA-1s, so ...	15:41
Mikachu	as long as what you fetch and what you have, have a common history, things should work fine	15:42
	for separate histories with the same files, probably not so much	15:42
ToxicFrog	Ick.	15:43
	Aren't blobs identified by sha1 as well?	15:44
Mikachu	yes, but there are very many blobs	15:44
comp	the situation was a problem in privilegies .. I decided to let others to create their own forks themselves rather than giving them privilegies on my own repo (can't restrict them to one branch)	15:44
	→ sbodo_w joined	15:44
comp	so in this case, I would have to duplicate many blobs and many MBs ...	15:44
Mikachu	i started running git rev-list --objects --all in linux-2.6.git one minute ago, it's still going	15:45
ToxicFrog	Yeah, but checking to see if you already have them isn't very expensive - it would be nice to at least have an option to save bandwidth by doing the iteration+stat() necessary to see what actually needs to be transferred.	15:45
comp	...and stored	15:45
Mikachu	% git rev-list --objects --all\|wc -l	15:46
	822167	15:46
	git rev-list --objects --all 39.05s user 1.54s system 46% cpu 1:26.38 total	15:46
ToxicFrog	So around 20k objects/second.	15:46
	← schacon left	15:46
ToxicFrog	Even if we assume checking existence is one tenth the speed of listing them - which seems very pessimistic - that's still way faster than network copy speed in most cases.	15:47
	→ pete__c joined	15:48
Mikachu	i think most people would be bored if every push / fetch of linux code took 1.5 minutes + network transfer	15:49
ToxicFrog	So make it an off-by-default option.	15:49
comp	I'm doing a small test	15:50
	the problem for me isn't in bandwidth ...	15:50
	it's 50*5MB anyway ... and only once	15:50
ToxicFrog	For that matter, it already has to iterate over all the referenced objects in the commits to build the list of stuff to send	15:50
comp	but 50*5 instead of 50MB of disk storeage .. that's something	15:51
	storage	15:51
Mikachu	ToxicFrog: you want to iterate over all objects in every commit in history on both ends though	15:51
ToxicFrog	Oh, right, as it is only the sender needs to do it...	15:51
	comp: ok, that strikes me as very weird	15:51
	All the objects are stored in $GIT_DIR/objects/ regardless of branch, right?	15:51
Mikachu	comp: it won't actually store the objects twice	15:52
ToxicFrog	And they're stored by sha1sum?	15:52
Mikachu	comp: you might need a repack -ad to combine the packs	15:52
comp	Mikachu: just testing it	15:52
jast	only if they're not stored in pack files	15:52
ToxicFrog	So if the same object is used by multiple commits it should be stored only once...	15:52
jast	that's right	15:52
	→ putrum joined	15:52
comp	(rebasing 1200 commits in a test repo so they will get different checksums)	15:52
Mikachu	that won't test anything like what happens in your fetch	15:53
	you'll just get 1200 new (unique) commit objects	15:53
	← putrum left	15:53
Mikachu	or you're going to fetch it to somewhere else? that might work	15:54
	→ putrum joined	15:54
comp	I'm gonna fetch	15:54
	and see if the space gets duplicated :)	15:54
	hope it won't	15:54
jast	identical objects don't get stored twice	15:54
comp	that's the way it should work	15:55
jast	that's the way it does work	15:55
comp	so if the fork has 99% of blob objects identical to mine, only tons of commit objects with different sha1s, fetch will download everything but store only changed (new) blobs (and trees).. ?	15:57
	← charon left	15:57
jast	that's the way i understand it	15:57
	← redondos left	16:00
	← Fullmoon left	16:01
comp	there's some magic in it	16:01
	it downloaded only 4MB	16:01
	and the repo is 4MB larger	16:01
	so it perhaps does some calculations about blob objects	16:02
	(whole repo is 65MB large)	16:02
	→ csc` joined	16:04
	← orion_ left	16:04
	← loops left	16:05
	→ flaguy joined	16:06
	→ p4tux joined	16:07
	← ggeecko left	16:09
	← lcapitulino left	16:10
	→ _zack joined	16:10
	→ johan-s_ joined	16:12
	→ ggeecko joined	16:14
	johan-s_ → johan-s	16:14
comp	... so if there's no way how to merge a submodule into superproject (including commits) I will have to create them by hand I guess..	16:15
	← ggeecko left	16:16
	← sbodo_w left	16:17
	→ ejs joined	16:18
	→ lcapitulino joined	16:19
	→ ggeecko joined	16:21
	← ggeecko left	16:23
	→ rphillips joined	16:24
	← Kbyte left	16:25
	← Ace_NoOne left	16:25
	→ wferi joined	16:26
	→ blix_ joined	16:27
	→ ggeecko joined	16:27
vmiklos	comp: subtree merge?	16:28
comp	exactly	16:28
	got that idea right now	16:28
	:)	16:28
	I used it instead of submodules before	16:28
	ToxicFrog ponders submodules	16:30
ToxicFrog	This could be useufl.	16:30
comp	but I don't need it in fact ... as I'm moving to new master (another svn-rebased one) soon so I will have to manually re-create the history (conflicts, ....)	16:30
	using dangling object as features ...	16:31
	s/object/objects/	16:31
	← ejs left	16:32
	→ gregf joined	16:36
	→ namenlos joined	16:37
	← mcella left	16:40
	← nud left	16:43
	← guardian left	16:44
	← oliver3 left	16:46
	← ckknight left	16:46
	← blix_ left	16:46
	← ggeecko left	16:47
	→ ggeecko joined	16:48
wferi	Hi! I'm in a git clone, how can I find out what it's a clone of?	16:50
Mikachu	git remote show origin	16:50
	or cat .git/config	16:50
	if it's a bare clone, you can't	16:50
wferi	thx	16:50
	→ CodeOfficer joined	16:50
wferi	If I do a git-push --dry-run --all, it complains about some branches being non-fast forward.	16:51
	How can I find out what happened?	16:51
jengelh	well obviously you replaced them	16:52
Mikachu	or someone else did a commit, if it's a shared repo	16:52
wferi	Sure, it is.	16:52
jengelh	and no one told you heh?	16:52
wferi	no. why should have?	16:52
jengelh	contradiction.	16:53
Mikachu	you have to fetch and then either rebase or merge before you can push	16:53
wferi	branch by branch?	16:53
Mikachu	yes?	16:53
wferi	?	16:53
Mikachu	what was the other option?	16:54
wferi	I'm confused about the branch stuff.	16:54
jast	in general you can't merge several branches at the same time because conflicts may arise	16:54
wferi	If I fetch, do I fetch all the branches?	16:54
Mikachu	yes, usually	16:54
wferi	ok	16:54
Mikachu	unless you changed the config	16:55
wferi	So I have to fetch, then see the log of each non-fast forward origin/* branches	16:55
	if I want to get a picture what causes the problems, right?	16:55
jast	you can do that, yeah	16:56
jengelh	well it's obvious	16:56
Mikachu	you can gitk --all but it might be messy	16:56
jast	but you can also assume that people pushed new commits to these branches	16:56
jengelh	what you are trying to push does not have the remote's tip as an ancestor	16:56
jast	because the only other possibility is that they did evil things	16:56
wferi	No, they aren't evil, just doing some work.	16:57
jast	my point exactly :)	16:57
wferi	I'm simply interested in what they did, before I rebase.	16:57
	← namenlos left	16:57
jengelh	git log -p HEAD...origin/master	16:57
wferi	after a fetch?	16:58
jengelh	yes	16:58
jast	something else you can do is look at git cherry	16:58
Mikachu	i like the new log --graph	16:58
	→ niki joined	16:58
jast	no wait	16:58
	that was wrong. you can do that but it's not useful for this situation. :}	16:58
wferi	They git-log -p trick is neat, though.	16:59
	→ loops joined	16:59
Mikachu	i would probably use .. and not ... since you already know what changes you did	16:59
wferi	What's the difference?	16:59
	between .. and ...?	17:00
	how can I find the common ancestor of HEAD and origin/master?	17:00
Mikachu	... shows both yours and their commits since the common ancestor	17:00
	while .. only shows the rhs	17:00
	← p4tux left	17:00
Mikachu	git merge-base HEAD origin/master	17:00
	→ eikonos joined	17:01
wferi	Ah, changes from HEAD to origin/master doesn't make sense after all.	17:01
jengelh	git log --graph is still no match for forest	17:01
	artistic wise :)	17:01
wferi	How can I change the origin URL of a clone?	17:02
	→ charon joined	17:02
Mikachu	wferi: git remote add nickname URL	17:02
wferi	nickname=origin?	17:02
Mikachu	if you want, you can delete the remote "origin" and use origin as the nickname	17:02
	→ jeckel joined	17:02
jast	see also: git remote rm	17:03
	← koke left	17:03
wferi	ok, so I have to delete the current origin.	17:03
jengelh	just edit it in the config, no need for rm	17:03
Mikachu	if you want it to be called origin, yes	17:03
wferi	great.	17:03
jast	you ca also edit .git/config	17:03
	→ koke joined	17:03
wferi	thanks for the help, guys, no evaluating your answers...	17:03
	s/no/now/	17:03
	← SRabbelier left	17:05
	→ redondos joined	17:05
	→ SRabbelier1 joined	17:05
	SRabbelier1 → SRabbelier	17:05
	← Sho_ left	17:07
	→ kanru joined	17:10
	→ Sho_ joined	17:15
	→ alley_cat joined	17:20
	← jeckel left	17:23
	→ jeckel joined	17:23
	→ nud joined	17:24
	← koke left	17:24
	→ d0k_ joined	17:27
	→ guardian joined	17:28
	← jeckel left	17:28
	→ jeckel joined	17:28
	← d0k_ left	17:29
	← deavid left	17:33
	← d0k left	17:34
	→ dave_h_d joined	17:35
	← igorgue left	17:36
	→ shaftyy joined	17:36
	→ robin joined	17:37
	← ph^ left	17:39
	→ ph^ joined	17:39
	→ brosner_ joined	17:43
	← brosner left	17:43
	brosner_ → brosner	17:44
	→ threeve joined	17:44
	→ igorgue joined	17:49
	→ alb joined	17:49
	→ sohail joined	17:50
sohail	good day	17:50
	I was wondering if it was possible to use git-svn to import a revisionr ange	17:50
	← putrum left	17:51
Ilari	sohail: AFAIK, at least importing from specified version forward is possible...	17:51
sohail	Ilari, well I was thinking the easy way to do it is to do ane xport of the repo in question and import that into git	17:52
	but that doesn't sound like anyone but me knows whats going on which is never good	17:52
	angasule_ → angasule	17:54
sohail	so Ilari I tried git-svn clone -r N http://foo/bar .	17:55
	and it doesn't seem to do anything	17:55
	← alley_cat left	17:56
Mikachu	is the . part of your command?	17:57
sohail	Mikachu, yes	17:57
floppyears	quick question	17:57
Mikachu	sohail: remove it	17:58
sohail	Mikachu, ok!	17:58
floppyears	I use a library, and the devs use svn	17:58
Mikachu	the man page says you can use I:J ranges for the revision	17:58
jengelh	please please don't use svn	17:58
sohail	Mikachu, still didn't seem to download anything	17:58
	jengelh, a project I'm tracking does	17:58
floppyears	can I use git-svn to create a copy of that svn library	17:58
	and in my project use git-submodule to keep that library to the latest nightly	17:59
Ilari	floppyears: If you do that, you should also publish a copy of that library repo with your main repo...	17:59
	← holizz left	18:01
floppyears	thanks Ilari	18:01
	→ peeja_ joined	18:01
	→ Xyri0 joined	18:02
nud	hello	18:06
	I've got some pain to get git mktree work as expected	18:06
	here is what I get: http://pastebin.com/m4866ff2b	18:07
	← albertito left	18:07
nud	in fact, it writes into the tree that the commit is a 'blob', despite I tell it it's a 'tree'	18:07
	yesudeep1 → yesudeep	18:07
jast	what commit?	18:08
Pieter	you'r missing an e	18:08
	hah, me too	18:08
nud	hrm	18:08
	→ alley_cat joined	18:09
nud	updated link: http://pastebin.com/m7d17ae47	18:09
	jast: there is no commit in the game	18:10
jast	oh, then that was a typo in what you said :)	18:10
	← igorgue left	18:10
jast	afaics from mktree.c, it verifies the type of the objects listed in your tree based on the existing object with that ID	18:10
	in fact it should fail if the object doesn't exist or has a type mismatch	18:11
	(that's git 1.5.5.1)	18:11
nud	I have that version	18:11
doener	nud: trees seem to have mode 040000	18:11
nud	hm	18:12
	it works with that mode	18:12
	funny, does the tree format only rely on the mode?	18:13
	← Xyrio left	18:13
	Xyri0 → Xyrio	18:13
doener	the word "tree" in the output is probably just for human consumption	18:13
nud	is there some documentation lying somewhere about the modes?	18:13
	← ec left	18:14
	alb → albertito	18:14
	→ holizz joined	18:14
	← dsyzling left	18:15
	← wferi left	18:15
	→ dirk1296 joined	18:16
	→ thiago_home joined	18:17
	← peeja left	18:17
nud	this denotes some incoherence on how git-mktree works	18:20
	it looks like it checks the object type against the "human readable" type but writes the information according to the mode	18:20
	→ rubydiamond joined	18:21
	← FunkeeMonk left	18:22
	→ blix_ joined	18:23
	→ deavid joined	18:24
	← alley_cat left	18:25
	← threeve left	18:25
	→ d2dchat_ joined	18:29
	← d2dchat left	18:30
	d2dchat_ → d2dchat	18:30
	→ alley_cat joined	18:33
pergesu	if someone sends me a patch and I apply it with "git apply", but the patch fails, how do I modify it and do a clean merge? I know that if I merge two branches and have a conflict, but resolve them, then git knows how to resolve that for others. So I'd like to have that same behavior when applying a patch	18:34
	← djwonk_ left	18:36
vmiklos	git am -3?	18:36
	← yesudeep left	18:36
jast	apply is a rather low-level command	18:36
	→ drizzd joined	18:37
	→ thannoy joined	18:39
pergesu	vmiklos: this patch was made with "git diff sha > blah.diff", not git format-patch. So I don't know if I can use git am? At least it complains that there's no email address	18:40
Pieter	pergesu: no way to merge anything then	18:42
Mikachu	yeah, i would expect git-apply to have more of the patch specific functions that git-am has, while git-am only adds the ability to apply several patches from an mbox	18:42
vmiklos	you can just add a fake header	18:42
	→ tglx joined	18:43
	← alley_cat left	18:44
	← _zack left	18:47
floppyears	what protocol does git use when it pulls from another repo? ssh, rsyn or http ?	18:48
Mikachu	or git	18:48
	← ggeecko left	18:48
	→ etnt joined	18:48
	→ ggeecko joined	18:48
frsk	Depends on the repo's hosting	18:49
	← pergesu left	18:49
floppyears	ah	18:49
Mikachu	if you write git:// it uses git, if you write ssh:// it uses ssh, etc	18:50
sbeyer	.o( or "filesystem", if the other repo is on the harddisk )	18:50
	→ felipec joined	18:53
	→ kristoffer joined	18:54
	← ebel left	18:57
	→ EiNZTEiN joined	18:58
	→ Xyri0 joined	18:59
	→ alley_cat joined	19:00
	← mediogre left	19:01
	→ dsyzling joined	19:01
pflanze	Is there a tool to output the path to the git repository directory currently being used?	19:02
Pieter	git rev-parse --show-cdup	19:02
	or --git-dir	19:03
pflanze	ah the latter, yep, thanks	19:03
	funny that this is in the revision parsing tool :)	19:03
	← Xyrio left	19:10
	Xyri0 → Xyrio	19:10
Pieter	anyone know of a OSS project with a git repo of around 30MB? :)	19:16
jengelh	Git.	19:16
Pieter	yeah, something that doesn't use octopus merges	19:17
	:)	19:17
jengelh	what you got against octopussies?	19:17
felipec	Pieter: I just created a clone repo of the pidgin's one	19:17
	Pieter: it's 29 MB and no merges	19:17
Pieter	jengelh: most importers crash on / ignore them	19:17
jengelh	mine does not	19:17
Pieter	felipec: I'd like a real git repo, not an import	19:17
	jengelh: what does it import to? :)	19:18
jengelh	a local directory?	19:18
felipec	Pieter: I did it myself, well, my script, commit by commit	19:19
Pieter	felipec: yeah, but I'd like a repo that uses merges and so	19:19
nud	I have a 28M copy of a git-svn project import (gedit)	19:20
	probably doesn't qualify since it's a svn import	19:21
	← EiNZTEiN left	19:21
Pieter	yeah :)	19:21
	SRabbelier: impressive git-stats repo you got there ;)	19:21
SRabbelier	Pieter: awesome huh? :D	19:21
	Pieter: will probably switch to a python script	19:22
	I wanted to get more done today but I still have a project for school to finish this week	19:22
	← d2dchat left	19:25
	→ d2dchat joined	19:26
	← alley_cat left	19:28
Pieter	I'll use the coreutils repo	19:29
	→ alley_cat joined	19:29
	→ ekidd joined	19:29
	→ advany joined	19:31
	→ ajc2008 joined	19:31
	← gregf left	19:32
ajc2008	is there a way to compile git without the git-gui/gitk requirements? I don't want to have to install lots of X libraries on my server but the standard git library doesn't seem to compile without Tk installed	19:33
nud	ajc2008: you can pass options to make	19:34
ajc2008	I just don't know what options ;)	19:34
nud	vi Makefile ;-)	19:34
	for instalce make NO_TCLTK=1	19:34
	instance*	19:34
doener	you can also put that into config.mak	19:35
ajc2008	Thanks :)	19:35
	→ Xyri0 joined	19:37
	← d2dchat left	19:41
	→ d2dchat joined	19:42
	← robin left	19:42
	→ robin joined	19:42
	← robin left	19:42
	→ robin joined	19:42
	robin → gittish	19:43
	← csc` left	19:48
	← ajc2008 left	19:48
	← rubydiamond left	19:49
	→ ajc2008 joined	19:49
	← Ingmar left	19:50
	→ Ingmar joined	19:51
	← Xyrio left	19:55
	Xyri0 → Xyrio	19:55
	← glommer left	19:57
	→ koke joined	19:59
	← vbgunz left	20:01
	← advany left	20:01
	→ csc` joined	20:02
	→ xl0 joined	20:04
xl0	Hi. I thought it was possible to remove some top-level commits in a public repo using git-push, but can't find how exactly this is done..	20:05
	← kristoffer left	20:05
	← eald left	20:06
Ilari	xl0: Remove them from local repo (possibly creating backup branch) and 'push -f' it?	20:06
xl0	Ilari: tried it, and it says Everything up-to-date.	20:07
Ilari	xl0: Are those commits you want removed from remote side removed from local side?	20:08
xl0	Ah, I has to specify the local branch name too. ;)	20:09
	Thanks a lot. ;)	20:09
	← sohail left	20:10
	→ radarsat1 joined	20:11
	← dirk1296 left	20:12
	← radarsat1 left	20:12
	→ radarsat1 joined	20:14
	← janm left	20:15
	← shaftyy left	20:16
	→ daaku joined	20:18
	← radarsat1 left	20:18
	← alley_cat left	20:20
daaku	if i'm currently in my-branch, and my-branch is a fully-rebased branch of origin, is there a way to say replace origin with my-branch and dont change any of the files (since they dont need to anyways)?	20:21
	they only merge i know is to switch to origin first, but i dont want the files to go back to that state (running apps are using these files)	20:22
	← priidu left	20:23
	→ fhobia joined	20:24
	← ggeecko left	20:25
jast	'switch to origin'? do you have a branch called origin? typically origin is the name of the default remote created by clone	20:26
	→ mario_ joined	20:26
	← xl0 left	20:27
	← pygi left	20:27
	mario_ → pygi	20:27
	→ alley_cat joined	20:30
	→ ggeecko joined	20:30
	→ chris2 joined	20:31
Ilari	daaku: Origin is not a branch (unless repo has some really obsolete configuration).	20:31
	→ intripoon joined	20:32
intripoon	hi	20:32
	where can I set core.autocrlf = true on windows ?	20:32
	or check what its current setting is	20:32
Ilari	intripoon: 'git config core.autocrlf'? 'git config core.autocrlf true'?	20:33
intripoon	since when does this variable exist?	20:34
	I have one user using git 1.5.4.2.1161... and one using 1.5.4.960... and the later one doesn't seem to have that setting	20:35
	git config core.autocrlf prints nothing there	20:35
Mikachu	Author: Linus Torvalds [email@hidden.address]	20:36
	Date: Tue Feb 13 11:07:23 2007 -0800	20:36
	Lazy man's auto-CRLF	20:36
jast	afaics it was introduced in 1.5.1	20:36
intripoon	and pulling from that repository doesn't merge the files efficiently meaning it creates a file with one conflict where the first part is the complete one file and the 2nd part is the complete other file	20:36
	← tagx left	20:37
jast	it prints nothing because no value is set by default	20:37
	try Ilari's second command. :)	20:37
Arrowmaster	on msysGit its in git-config --system core.autocrlf by default	20:37
intripoon	ok	20:38
	so how would I go about merging our work now?	20:38
	the pre 13th feb rep exist and my exist	20:38
	if I pull the pre 13th feb version in my, the merge screws up	20:39
	→ lolage0 joined	20:41
daaku	sorry, i meant master	20:43
	→ Ramune joined	20:44
shauber	so, has anyone here managed to get git (any version) compiled on Solaris 9/SPARC? Since I can't seem to find any binary downloads...	20:45
	i've been fighting with it for a week now	20:45
BabelO	ph^: pH5 you ? have a problem with git track, i did not see mainline change :(	20:46
	oups sorry, wrong channel	20:47
	← Ademan left	20:49
	→ thesaint4444 joined	20:51
BabelO	hi, that s time it is the right channel, i ve used git --trackto create a branch which track mainline linux kernel, when i update my branch with git fetch && git pull origin .... , i did not get mainline change ? is that normal ? what i miss ? thanks	20:51
thesaint4444	hi guys, I have a problem with a git, capistrano setup. Do I have to have passwordless ssh setup? thanks.	20:52
	to deploy the git repo....	20:53
jast	BabelO, you did clone from the linux repo, i assume? in what way? how exactly did you create the branch?	20:53
	← thannoy left	20:53
Ilari	daaku: 'git update-ref refs/heads/master my-branch' would update branch 'master' with what 'my-branch' currently points to...	20:55
	daaku: Note, the first name (the ref to update) really needs to be fully qualified.	20:55
Mikachu	hm, maybe that's how i created .git/master in my git.git repo	20:58
	→ nkallen joined	20:58
Ilari	Mikachu: Yes, 'git update-ref master ...' does unwanted stuff... :-)	20:58
jast	probably better to use push in order to avoid spreading the use of plumbing	20:58
BabelO	jast: i ve used git branch --track htc-omap origin/mainline to create branch htc-omap whcihc is tracking mainline	20:58
jast	BabelO, try git pull without arguments in that branch (that's what --track is for, after all)	20:59
BabelO	jast: so only solution is git merge origin/mainline ?	20:59
	← koke left	21:00
BabelO	jast: i ve tried, but still old ... i use merge then, thanks .	21:00
jast	because when passing arguments to pull you need to use the correct refspec	21:00
	→ JRV joined	21:01
daaku	Ilari: sweet! exactly what i needed	21:01
JRV	Hi, does any of the graphical git tools allow splitting hunks into smaller ones yet?	21:02
BabelO	jast: thanks	21:02
Mikachu	git-gui has for a long time	21:02
JRV	Mikachu: hm, maybe my version is too old	21:02
	that would be "git citool", right?	21:03
RandalSchwartz	what work flow results in asking that question?	21:03
	← lcapitulino left	21:03
RandalSchwartz	I mean, when would you want a commit to have parts?	21:03
	→ Ademan joined	21:03
JRV	When reworking patches along a different dimension, I might want to separate two logical changes in a hunk?	21:04
jast	when you're doing large-scale surgery but later decide you want to split it into smaller commits	21:04
	i've had it happen myself	21:04
RandalSchwartz	yeah, ok, haven't had it yet.	21:05
shauber	i take it that would be a "no" then?	21:06
Mikachu	RandalSchwartz: it happened to me a lot right after i switched to git and i had some svn checkouts with lots of unrelated changes in them	21:06
	not of my projects, that is, so i couldn't commit them with svn, but with git(-svn) i can have local commits, so splitting hunks with git-gui / git add -p was very helpful	21:07
JRV	hm, so in my version (0.8.4) of git-gui, it doesn't seem to be supported yet	21:07
	maybe I should get a newer one	21:07
RandalSchwartz	ahh - once again, SVN is part of the equation	21:07
Mikachu	JRV: right click and choose increase context / decrease context	21:07
	RandalSchwartz averts his eyes	21:07
Mikachu	RandalSchwartz: or cvs of course	21:07
JRV	Mikachu: yeah, but that doesn't always help	21:07
	→ cliffstah joined	21:07
JRV	What about lines that are adjacent or have only one line between them? Then it says the hunk doesn't apply anymore... Or is that a problem I will always have?	21:08
	→ eald joined	21:08
	← Ademan left	21:09
JRV	What I mean is that I have to shrink the context to such an extent that the patch can't be applied anymore because it doesn't know where to put it.	21:09
	← Eludias left	21:09
	← alley_cat left	21:10
BOOKZ_	if I have two branches, can I remove a directory from one of them, and ignore subsequent changes when merging from the other branch	21:12
gittish	JRV: Micachu: git gui doesn't split hunks. It does selective staging of individual hunks	21:12
BOOKZ_	typically, a dev branch with a test dir and a live branch without it	21:12
JRV	gittish: yep, but you can make the hunks somewhat smaller by decreasing their context sometimes	21:13
	gittish: so in effect, you can split hunks to an extent as a side-effect of that	21:14
gittish	I did not count that as hunk splitting	21:14
JRV	it isn't really, in the sense of 's' in git-commit --interactive	21:14
gittish	It would have to handle the one line context in between to count	21:14
JRV	yeah, that is what I'm missing	21:14
gittish	I'd like that too. Patches welcome :	21:14
	→ TuxCM joined	21:15
JRV	Well, I saw a thread already about that	21:15
BOOKZ_	currently, I get messages like: CONFLICT (delete/modify): test/zlonk deleted in HEAD and modified in master. Version master of test/zlonk left in tree.	21:15
	→ gitte joined	21:15
gittish	for deleted replaced lines one could to with no context. Somewhat dangerous though.	21:16
	BOOKZ_: that means you have to decide what to do. Git doesn't know.	21:17
	← bdiego left	21:17
gittish	it isn't fatal per se	21:17
BOOKZ_	yes, I know, and how can I tell git what my decision is?	21:18
	← thesaint4444 left	21:18
BOOKZ_	"ignore the test directory in this branch, until I change my mind"	21:19
JRV	gittish: http://lists-archives.org/git/640447-teach-git-gui-to-split-hunks.html	21:19
Ilari	BOOKZ_: To declare conflicts in file resolved: 'git add <file>'.	21:19
	→ igorgue joined	21:19
BOOKZ_	Ilari: except the conflict resolved by removing the directoy and its content	21:22
	→ Ademan joined	21:22
BOOKZ_	is resolved	21:22
Ilari	BOOKZ_: Then 'git rm <file>'.	21:23
	← holizz left	21:23
BOOKZ_	$ git rm -r test	21:23
	fatal: pathspec 'test' did not match any files	21:23
	I removed these files in an earlier commit	21:24
	(on that branch)	21:24
RandalSchwartz	on both sides of the merege?	21:24
	or just one side?	21:24
BOOKZ_	on one side only	21:24
RandalSchwartz	then rm -rf test	21:24
	and see what git-status says	21:24
BOOKZ_	nothing to commit	21:25
RandalSchwartz	good to go then	21:25
	the merge commit will remove those files then	21:25
BOOKZ_	how do I commit, it it tells me there's nothing to commit?	21:26
RandalSchwartz	are you still in a merge?	21:26
BOOKZ_	yes	21:26
RandalSchwartz	git commit :)	21:26
gittish	git add -u to update the index	21:27
RandalSchwartz	that too	21:27
gittish	i.e. remove the conflict marks (non stage 0 entries)	21:27
BOOKZ_	ah that might be what I missed	21:29
	← nud left	21:29
gittish	was that the only conflict?	21:29
BOOKZ_	$ git add -u	21:29
	fatal: test/zlonk: unable to stat (No such file or directory)	21:29
	yes	21:30
gittish	hmm	21:30
BOOKZ_	maybe I should have git rm'ed ? rather than just rm ?	21:30
	← daaku left	21:30
gittish	I'd like to think add -u would do that	21:30
	→ tzafrir_home joined	21:31
Ilari	BOOKZ_: Maybe git rm it...	21:31
gittish	git rm --cached <dir>	21:31
BOOKZ_	ok, git rm + git add -u + git commit did it	21:31
Mikachu	have we established what version of git this is?	21:31
BOOKZ_	1.5.5	21:31
Mikachu	not that then	21:32
BOOKZ_	and I guess everytime something changes in the unwanted directory, git will bring the changes in	21:32
	which seems fair	21:32
tzafrir_home	any idea how I can set certificate for authentication for usage with git-svn?	21:32
BOOKZ_	I wondered if their was a way to say "ignore this directory from now on"	21:32
	gitignore doesn't seem to be it	21:33
RandalSchwartz	edit .gitignore?	21:33
Mikachu	gitignore is about untracked files	21:33
RandalSchwartz	it won't ignore tracked files	21:33
Ilari	tzafrir_home: Client certificate?	21:33
tzafrir_home	Ilari, yes	21:33
	← dave_h_d left	21:34
tzafrir_home	I can't seem to figure this from docs, source and some web search	21:34
	RandalSchwartz averts his eyes at the mention of svn	21:34
Mikachu	i think git-svn will just use whatever configs you've set up for svn	21:34
Ilari	RandalSchwartz: Even worse: Client certificates are involved. :-)	21:35
tzafrir_home	Mikachu, I was optimistic as well	21:35
RandalSchwartz	there's no relative rings of hell	21:35
	it's just Hell	21:35
tzafrir_home	Sadly it doesn't	21:35
Mikachu	hope they weren't generated on debian :)	21:35
Ilari	Or Ubuntu.	21:35
BOOKZ_	oh, I added a file, it was merged automatically	21:36
tzafrir_home	On the same system I have pooled the same directory with SVN and it works just fine	21:36
BOOKZ_	mmm, this proves a little trickier than I thought... it's not impossible, but it requires somes manual fiddling every time	21:36
Ilari	BOOKZ_: Generally that sort of stuff is done the other way: The side branch adds something.	21:37
BOOKZ_	ok	21:37
	→ markkalderon joined	21:38
BOOKZ_	we're currently trying to decide if we want our code and our tests in the same repo or not, and I thought, deployement could be done from a live branch, which could not include the tests	21:38
	BOOKZ_ → BooK	21:39
gittish	Cherry-picking is also a way of solving these kinds of problems, sometimes	21:39
Mikachu	i would maybe just fetch && reset --hard origin/master && rm -rf tests	21:39
gittish	BooK: why don't you just skip the tests in the deployment then?	21:40
BooK	that's the other options, yes :-)	21:40
	just evaluating the possibilities, at the moment	21:40
gittish	automating the rm's seems like an option too	21:41
BooK	deployment implies copying files on a bunch of servers, so skipping one directory isn't so hard	21:41
	the fact that the test files are here doesn't force anyone to actually run them ;-)	21:41
	I wonder if I would have the same problem with unwanted files	21:42
RandalSchwartz	you could even git-archive a subset	21:42
	and then unpack that .tgz multiple places	21:42
BooK	ok	21:43
	← felipec left	21:43
	→ EmilMedve joined	21:47
BooK	thanks everyone	21:52
	← johnreilly_ left	21:55
	→ joelr1 joined	21:55
joelr1	good evening! has anyone set up access with git-shell on solaris (joyent perhaps?)	21:57
	→ dsyzlin1 joined	21:58
	← dsyzlin1 left	21:58
	← kumbayo left	21:58
context	joelr1. ask your real question now	21:59
joelr1	context: well... i set up a user with git-shell but cannot pull now	22:00
d2dchat	grrr	22:00
joelr1	i can ssh to the box as the user and get a reply from git-shell: fatal: What do you think I am? A shell?	22:00
	but i cannot clone and cannot pull	22:00
	fatal: The remote end hung up unexpectedly	22:00
	context: end of real question	22:00
	← MrDarkUser left	22:00
	→ ebel joined	22:04
	← dsyzling left	22:05
	→ sabooky joined	22:05
Ilari	joelr1: Maybe the URL is wrong? Git is not very good at reporting bad URLs (if they point to valid servers)...	22:06
joelr1	Ilari: i can pull using my primary account and the same path to repo	22:08
	of course that account does not use git-shell	22:08
Ilari	joelr1: Does that other user have sufficient file-level permissions?	22:09
	← ph^ left	22:09
joelr1	Ilari: they belong to my group and the repo has group read/write permissions	22:10
	Ilari: the other user has the same group id	22:11
	Ilari: the repo does not live in their home directory so i give them the full path to mine	22:11
Ilari	joelr1: And you gave that full path also when you tested with your account?	22:12
joelr1	Ilari: yes, i just replaced the user name in the command	22:12
Ilari	joelr1: Can SSH set environment variables (could be useful for debugging this)?	22:14
joelr1	Ilari: on the server? i have root access. what do i need to do?	22:14
Ilari	joelr1: There is environment variable GIT_TRACE, that when set to 1 causes git commands to dump exec tracing info...	22:15
joelr1	Ilari: where do i place it?	22:15
Ilari	joelr1: No idea. git-shell does not have run control files, so you need to force that setting from SSH level...	22:17
	← deavid left	22:17
joelr1	Ilari: thanks! googling for env vars with ssh	22:17
	→ sergio joined	22:18
	→ pergesu joined	22:19
jast	man ssh_config, see SendEnv	22:19
joelr1	jast: thanks. no SendEnv in my man page	22:21
	→ threeve joined	22:21
jast	hmm. i don't know when it was added	22:21
	← ekidd left	22:21
jast	joelr1, here's an idea for additional tests: give that user a normal shell and try various git operations directly on the server	22:22
	← threeve left	22:23
joelr1	jast: ok	22:23
	jast: i set up .ssh/environment in the users home dir	22:24
	trace: exec: 'git-upload-pack' '/home/fjj67yaa/git/bridge'	22:24
	trace: exec failed: No such file or directory	22:24
	is there a special syntax for absolute paths?	22:25
Ilari	joelr1: It doesn't find 'git-upload-pack'.	22:25
joelr1	Ilari: what's that?	22:25
jast	one of git's binaries	22:25
	← devogon left	22:25
Ilari	joelr1: The repo it is trying to access is '/home/fjj67yaa/git/bridge', which looks reasonable.	22:25
	← saintdev left	22:25
gittish	there is an option to specifiy the path to the binaries	22:26
Mikachu	i believe the man page for git-push says how to configure that	22:26
joelr1	ls -la /opt/local/bin/git-upload-pack	22:26
	-r-xr-xr-x 1 root root 343244 Mar 19 22:37 /opt/local/bin/git-upload-pack	22:26
	it's the same directory where git-shell lives	22:26
	← JRV left	22:26
comp	is there any way how to "merge" two branches with "almost" same tree and blob objects, but different commit hashes? ... I need to "synchronize" my local git-svned branch into our "official git-svned branch" and don't want to go through resolving conflicts ...	22:26
Ilari	joelr1: How was that git installed? From source? What put those files in that location?	22:27
	← brosner left	22:27
joelr1	it's not clear from the error whether it's not finding git-upload-pack or the repo itself	22:27
	Ilari: it's pkg-src	22:27
jast	you could try setting PATH in ~/.ssh/environment	22:27
Mikachu	it's pretty clear since it says "exec failed"	22:28
	← _graham_ left	22:28
joelr1	fair enough	22:28
jast	comp, only manually (e.g. format-patch)	22:28
joelr1	setting path ...	22:28
	path was it, i think	22:30
comp	jast: so reset to the oldest and cherry-pick my oldest commit, then pull new things up to my next commit and so on ...	22:30
jast	comp, that sounds a lot more complicated than it should	22:30
	← tglx left	22:30
jast	comp, what kind of differences are there between the two repos?	22:31
joelr1	thanks you guys!	22:31
Ilari	Hmm... git-shell does set the path for the CVS case, but not the Git pack transfer case... Why it doesn't set the path for the other case?	22:31
comp	jast: my custom changes that I would like to keep on my branch there ... with some resolved conflicts ..	22:31
	and, ... commit sha1s	22:32
	← joelr1 left	22:32
comp	as git-svn generates new ones everytime (as it's intended for one-man-usage I suppose)	22:32
Mikachu	git-svn imported commits are usually identical	22:32
jast	that's my experience too. i had no problems syncing.	22:33
comp	well, my problem isn't in this	22:33
	but I git-svned only from certain point	22:33
	and they want full history from 1	22:33
jast	hey, that's actually possible?	22:33
comp	yep ..	22:34
	-r	22:34
jast	ah	22:34
comp	so different parents ...	22:34
jast	well, as i said, git format-patch and git am (on the other side) is probably easiest	22:35
comp	will the time and date of creating a commit get over cherry-pick ?	22:35
	jast: I can do it on my local machine ...	22:36
	mtodd__ → mtodd	22:36
jast	do what? cherry-pick is a lot more work than format-patch and am, anyway	22:36
comp	(just pull the "oficial git-svned repo"	22:36
	jast: I used git-format-patch and git-am only for some easy tasks, but I'll try it	22:38
jast	i would suggest something like the following: 1) clone official git repo, 2) git format-patch your own commits (with --stdout and redirect to file), 3) git am them in the clone of the official repo, 4) push from the clone	22:38
comp	but what's the difference between git-format-patch and git-cherry-pick anyway?	22:38
	(I mean the final result)	22:39
jast	format-patch can handle any number of commits, while cherry-pick only does a single commit	22:39
	← peeja_ left	22:40
jast	i believe there is no notable difference in the end result	22:40
	← pergesu left	22:40
comp	the problem is that my commits aren't in sequential order (like in git-cherry) .. I have lots of merges from upstream there as well	22:40
	→ pergesu joined	22:40
jast	merges from svn? svn doesn't have merges	22:41
comp	I'm making custom changes to some project that uses svn	22:41
RandalSchwartz	why aren't you rebasing your changes on top of the current upstream head?	22:42
jast	using merge when you're dealing with svn stuff is usually not a good idea	22:42
comp	one day, they decided to provide a git-svned mirror for those, who can't commit to the oficial svn, but wants to pull easily between each other	22:42
	well, yes, there's a rebase solution ... but I wanted to keep the history	22:43
RandalSchwartz	rebase keeps history	22:44
	← ebel left	22:44
comp	interactive rebase?	22:44
jast	do you really need your commits to stay strewn all over the place? all rebase does is pull them all to the top of the branch, keeping their internal order	22:44
comp	jast: yes .. and I need to resolve conflicts between my oldes commit and the newest clone	22:45
	→ glommer joined	22:45
jast	i don't see any problem. feel free to use cherry-pick on every single commit if you think that's the best solution, though.	22:46
	→ brosner joined	22:47
comp	well ... so at least I can filter out many commits which I reverted afterwards and so on ...	22:47
	← etnt left	22:49
	→ spearce joined	22:50
	→ dsaxena joined	22:50
	← dsaxena left	22:50
	→ cedric_ joined	22:54
	← sergio left	22:54
	→ dsaxena joined	22:55
	← cedricv left	22:55
	→ MrDarkUser joined	22:56
	← BabelO left	22:57
	→ MrDarkUserX joined	22:57
comp	I can't believe it	22:59
	just did git rebase oldbranch ...	22:59
	it printed few errors	22:59
	but it doesn't look like rebase result	22:59
	(in tig)	22:59
	→ daaku joined	23:01
comp	http://rafb.net/p/2pm5OU66.html	23:02
	but the standard merge had many conflicts	23:02
	→ _zack joined	23:04
	← thiago_home left	23:05
	← _zack left	23:06
	← intripoon left	23:08
comp	anyway, git rebase has --preserve-merges which can help ...	23:09
	→ spearce` joined	23:12
	← comp left	23:15
	← MrDarkUser left	23:16
	← dsaxena left	23:16
pasky	scarily-illuminated maniacal face within darkness grooow my thesis grooow	23:18
RandalSchwartz	you forgot "bottom-lit"	23:18
jengelh	and perhaps stencil-shadowed	23:19
pasky	right right right >:)	23:21
	← chris2 left	23:21
pasky	mhm, I wish bzr repository format would be better documented though	23:21
	← igorgue left	23:22
spearce`	what, its possible to document a format less than git is documented?!?!?	23:23
	← spearce left	23:23
	spearce` → spearce	23:23
pasky	spearce`: git is documented pretty well	23:23
	though the documentation is slightly hard to read in this particular case ;)	23:23
	pasky ponders about a leading quote for his thesis	23:23
cehteh	i once read extensive docs about the bzr format	23:23
pasky	like...	23:23
	Always give a person as much rope as possible. This is so when you do pull the lever, he swings real good.	23:23
	-- S. Short	23:23
	cehteh: oh oh, pray tell, where to get it? :)	23:24
cehteh	but they changed the underlying format serveral times since then in extreme ways -- prolly they got bored about maintaining the docs :)	23:24
pasky	ah	23:24
cehteh	pasky: was when bzr began :)	23:24
	→ comp joined	23:25
pasky	hmm i like the Short quote more and more ;))	23:25
	← pergesu left	23:25
	→ igorgue joined	23:27
pasky	If you steal from one author it's plagiarism; if you steal from	23:27
	many it's research.	23:27
	-- Wilson Mizner	23:27
RandalSchwartz	there's actually a table of rope lengths based on weight	23:27
pasky	I think this one would be too true ;)	23:27
RandalSchwartz	too little, doesn't snap the neck	23:27
	too much, the rope breaks.	23:27
pasky	RandalSchwartz: heh :))	23:27
RandalSchwartz	so there's an official table of "length of rope per body weight" that they used	23:27
	can't remember where I read that, but it seems obvious now.	23:28
pasky	well, yeah...	23:28
RandalSchwartz	oddly enough, they accumulated the data from many military hangings, apparently	23:28
jengelh	pasky: Those who only cite will never get cited. --Anon.	23:28
Ilari	RandalSchwartz: Depends on nature of the rope how much it takes before it breaks. :-)	23:28
RandalSchwartz	Yes - it was based on type of rope as well	23:28
	oh - too much rope and it could also cut clean through a neck. ouch.	23:29
	as in, behead.	23:29
Mikachu	i had to think for a while to realize it's because of the falling distance :P bit tired here	23:29
	→ mcella_ joined	23:29
RandalSchwartz	heh	23:29
jengelh	not the distance is important, the force is!	23:30
mugwump	pasky: did you get the message about the v6 routing to repo.or.cz2A	23:30
	?	23:30
pasky	i think so	23:30
Mikachu	jengelh: the gravity is fairly constant :)	23:30
RandalSchwartz	... http://www.ehow.com/how_2146065_calculate-necessary-rope-length-hanging.html	23:30
pasky	i need to get rid of v6 in the dns again	23:30
RandalSchwartz	... http://www.capitalpunishmentuk.org/hanging1.html	23:30
Ilari	BTW: Some earlier cars had glove compartment doors that could open in accident and cut through neck of small-sized passager in front seat (while the keys sliced kneecap of the driver). :-/	23:30
pasky	i'll do it all just when i finish the first content-complete draft of the thesis :)	23:30
mugwump	ok, it's just a pita for v6-enabled sites with that record there and no routing	23:31
	but I guess git seems to recover ok	23:31
RandalSchwartz	.. "The long drop method was designed to break the prisoner’s neck by allowing them to fall a pre-determined distance and then be brought up with a sharp jerk by the rope. At the end of the drop, the body is still accelerating under the force of gravity but the head is constrained by the noose which delivers a massive blow to the back and one side of the neck, which combined with the downward >>	23:31
	momentum of the body, breaks the neck and ruptures the spinal cord causing instant deep unconsciousness and rapid death."	23:31
jengelh	Mikachu: the less gravitational pull, the longer you could survive :)	23:32
RandalSchwartz	you can't get hanged in 0 g!	23:32
Mikachu	maybe with some magnets	23:32
	← jamessharpe left	23:32
RandalSchwartz	yeah, after swallowing buckshot. :)	23:33
	← charon left	23:34
	← redondos left	23:36
Ilari	Hmm... Wonder how much magnetic field it takes to make human levitate... With frogs, about 16T is enough... And in weightless conditions, that sort of magnetic field should approximately simulate normal gravity... Too big magnets required though...	23:38
Mikachu	i would probably attach some magnets to the subject	23:39
pygi	Ilari, you're thinking in todays technology	23:39
	think in the inventions of tomorrow	23:39
	← cmarcelo left	23:39
cehteh	Ilari: and hope that there is no small particle of iron in your body	23:44
Mikachu	depending on how small it is and how powerful the magnet is, you'll just end up with a very small hole in your body :)	23:44
cehteh	16T is quite a lot	23:45
	→ dsaxena joined	23:45
Ilari	cehteh: Yea. It is a lot... IIRC even the most magnetic-field-resistant superconducting materials max out at 22T...	23:46
	← gittish left	23:48
	→ Cerebon joined	23:57
	GyrosGeier points out that this would probably be lethal if you had a tattoo	23:59

Logs Search ←Prev date Next date→

Channels Documentation