IRCloggy #git 2008-06-18

Provider of IRC logs since 2005.

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

	2008-06-18
aeruder	you've just created a multi-parent commit	00:00
	← johnw left	00:01
claymccoy_	wow, so that is the difference in ~ and ^	00:01
Mikachu	~3 is ^^^	00:02
	← ckknight left	00:02
	→ ph^ joined	00:02
	← eventualbuddha left	00:03
	← Ryback_ left	00:03
	→ ezyang joined	00:04
claymccoy_	so ^2 will give me the 2nd parent, can I do this with ~ too. Something like ~1:2	00:04
	← ciaran left	00:04
	← eternaleye_ left	00:05
	→ eternaleye_ joined	00:05
	→ ciaran joined	00:05
ezyang	What happened to all the msysgit developers? :-( There's a patch I'd like to get merged into the main repos.	00:05
Mikachu	claymccoy_: ?	00:05
	claymccoy_: if you want the second parent, use ^2	00:05
	→ pombreda joined	00:06
	→ tsoiland joined	00:07
claymccoy_	Yeah, that is the second parent of the last commit, right. And ^^^2 would be the second parent of 3 commits ago. so if I wan to use ~ to specify how many commits ago, but I want the second parent of it.	00:07
Mikachu	~2^2 then	00:08
doener	claymccoy_: ~3^2	00:08
	← ciaran left	00:08
Mikachu	it's easy to do that wrong ;)	00:08
	→ ciaran joined	00:08
doener	I read the text, you parsed the ^^^2	00:08
claymccoy_	cool, thanks! this is a new concept for me.	00:08
Mikachu	yeah, let's say i said that to claymccoy_ :)	00:09
doener	:-)	00:09
	← ciaran left	00:09
	→ ciaran joined	00:09
	← timocratic left	00:09
	← ciaran left	00:10
	→ ciaran joined	00:10
	← claymccoy_ left	00:10
	← ciaran left	00:11
tsoiland	question: is it possible to have multiple public repositories under one project on repo.or.cz? The reason is that we want to do this: http://www.kernel.org/pub/software/scm/git/docs/user-manual.html#public-repositories instead of everyone pushing to the same repo.	00:11
	→ ciaran joined	00:11
	→ ckknight joined	00:12
	← markkalderon left	00:12
vmiklos	sure.	00:12
	the maintainer registers the project and then others can create forks	00:12
	← ciaran left	00:13
	← radarek left	00:13
	→ ciaran joined	00:13
	→ markkalderon joined	00:13
tsoiland	is that what forking means on that site? not actual forking? if that makes any sense.	00:14
Mikachu	forking means they use less space on the server which makes pasky happy	00:14
jast	tsoiland, i guess the question is: what do you actually want to do?	00:15
	you can always have everyone use a separate project. forking just means they can reuse the same hard disk space for common history.	00:17
	→ zawir joined	00:17
	← pdherbemont left	00:18
	← ciaran left	00:18
tsoiland	I guess that would work. But will it look like the project split up in n projects. Where n is the number of devs (since everyone will have their own public repo)	00:18
	→ ciaran joined	00:18
tsoiland	I guess there should be a question mark in there somewhere.	00:18
	← ciaran left	00:19
	→ ciaran joined	00:19
	← ciaran left	00:19
jast	tsoiland, well, suppose you use the forking mechanism. you create one 'main' public repo where things can occasionally be merged together into an 'official' representation. then each fork is for a specific developer. in the end, it will look like this: http://repo.or.cz/w/git.git (scroll down to "forks")	00:22
	→ pdherbemont joined	00:22
Mikachu	hm, changing the sort order on the forks page doesn't work	00:23
	http://repo.or.cz/w/git.git?a=forks -> http://repo.or.cz/w?o=age	00:23
tsoiland	okey.. so its a common way of doing it?	00:23
	blowery → blowery`afk	00:24
jast	yeah	00:24
	at least on repo :)	00:24
	well, sorting works manually	00:24
	not that that's much of a solution	00:24
tsoiland	great! forking it is.	00:24
	thanks	00:24
	→ mithro joined	00:28
	blowery`afk → blowery	00:29
	← markkalderon left	00:29
	→ ijcd joined	00:30
	← Slashbun left	00:32
	← robinr left	00:35
	← cmarcelo left	00:40
	→ DrNick joined	00:41
	← chrisb left	00:46
	blowery → blowery`afk	00:49
	← branstrom left	00:50
	← tsoiland left	00:56
	← k0001 left	00:57
	→ vbabiy joined	00:57
	← thannoy__ left	00:58
	→ VitaRara joined	01:03
	← david_koontz left	01:03
	← Tsaghen left	01:06
	← csc` left	01:07
	→ doener_ joined	01:09
	→ brosner joined	01:17
	← patchwork left	01:18
	← vbabiy left	01:20
	→ louipc joined	01:21
	→ vbabiy joined	01:21
	→ csc` joined	01:23
louipc	csc`: nooooo	01:23
	← doener left	01:24
csc`	what louipc	01:24
louipc	csc`: you're following me aren't you	01:24
csc`	i was here before you	01:24
	← tvachon left	01:25
Mikachu	so you came from the future too	01:25
louipc	!	01:25
ezyang	and now you must be destroyed (creepy alien noises commence)	01:25
	louipc pulls out the theremin.	01:26
	kelvie__ → kelvie	01:28
	← cnk left	01:28
	← pombreda left	01:28
	→ erh__ joined	01:36
erh__	is there any good way to permanently delete a file? from every revision? i tried filter-branch - but the pack was still huge	01:37
	someone added a 500mb tgz file by accident and pushed it...	01:37
	← brosner left	01:37
Mikachu	you have to delete it from several places before git will throw away the data	01:38
	reflogs, remote refs, the refs/original namespace, then run gc	01:39
	possibly some more	01:39
ezyang	erh__: owwch	01:39
erh__	so there's no easy way, right? or a script someone knows about?	01:40
ezyang	erh__: Do you have any repo backups? That might be faster, and then you can merge in recent changes	01:40
erh__	yeah - thats what i was thinking	01:40
	→ EmilMedve joined	01:40
erh__	if i take a backup - and people have local repos that were cloned from a previous point - will they be able to use them	01:40
	→ claymccoy joined	01:40
erh__	should be as far as i can tell	01:40
ezyang	erh__: IIRC, you'll need to do a bare clone, and swap that in	01:41
erh__	anyone know if this is a planned feature?	01:42
ezyang	Of course, there will still be the poor saps who git pulled the 500 mb tz	01:42
	erh__: No.	01:42
	→ igorgue joined	01:43
ezyang	Deleting commits goes completely against what git is about	01:43
erh__	yeah - pulling that 500mb file for eternity doesn't seem like fun	01:43
	→ Sake joined	01:43
Sake	is there a good cvs/svn hosting project where I can put a project, make it private and only give certain people access?	01:43
ezyang	Sake: This is the #git chan\	01:43
	Sake: Ask at #svn or #cvs	01:44
erh__	I imagine this comes up sometimes though - if this happens, or if someone accidentally commits a file with a password	01:44
	← l0verb0y left	01:44
	← SuttoL_ left	01:45
vmiklos	you can use git rebase -i to remove the pass from the commit.	01:45
louipc	damn	01:45
ezyang	Hopefully, the realize before pushing	01:45
shd	Sake: seriously, why ask at this channel?-)	01:45
louipc	get them to send a patch next time	01:45
ezyang	once you push, the cat's out of the bag	01:45
louipc	then it'll be freakin obvious	01:45
erh__	:) 500mb email attachment...	01:46
ezyang	hehe, the mailing list will probably reject it	01:46
louipc	that's the way I work	01:46
	all devs send patches for review to the ml	01:46
	even if they have push access	01:47
	← Ramune left	01:47
	→ jmspeex joined	01:49
erh__	vmiklos: not quite sure i get how i would you rebase	01:49
	use rebase that is..	01:49
Sake	I meant to add Git to that	01:51
	sorry...	01:51
vmiklos	the -i switch is important :)	01:51
ezyang	git rebase is one of the things that is difficult for newbies, since there's no equivalent in most other svms	01:51
Sake	I just found out about github though, looks cool	01:51
ezyang	Sake: You have to pay them for private access	01:51
	→ Leefmc joined	01:51
	← ajonat left	01:51
ezyang	Sake: Do you have a server you have SSH access to? That's probably your best bet	01:51
louipc	yea	01:52
ezyang	I don't understand it myself, actually	01:52
	→ ajonat joined	01:52
erh__	vmiklos: could i rebase master -> master	01:52
louipc	I did a rebase once	01:52
	it was awesome	01:52
ezyang	</sarcasm>?	01:52
louipc	well I mean i did a -serious- rebase	01:52
	conflicts through 20 commits	01:52
ezyang	Sake: The bells and whistles github has are pretty nice, though	01:53
vmiklos	erh__: just 'git rebase -i HEAD~X' (if the problematic ommit was 5 commits ago, then X=5, for example), change the problematic commit's 'pick' to 'edit' and that's it. see also the interactive mode part of man git-rebase	01:53
	commit*	01:53
erh__	vmiklos: ok - i did the base. the commit is not in the log. but .git/objects is still huge	01:56
	so i probably need a gc or a prune	01:57
louipc	yeah	01:57
	← VitaRara left	01:57
erh__	what's the best way to get rid of it forever?	01:57
Mikachu	cd ..; git clone file://$PWD/oldrepo newrepo	01:57
erh__	still there - (i'm just looking at the size of .git/objects)	01:58
louipc	hmm	01:58
vmiklos	or just git reflog --expire=now; git gc	01:59
	← unreal left	02:00
erh__	now .git/objects/pack is huge	02:02
	← digitalhobbit_ left	02:04
	← saintdev left	02:05
	← vbgunz left	02:05
	→ saintdev joined	02:07
louipc	erh__: is the change in your commit history at all?	02:07
erh__	nope - gone	02:07
	→ stouset joined	02:10
	→ unreal joined	02:11
	→ jjuran joined	02:12
	← pflanze left	02:14
	→ destruct_ joined	02:15
	← claymccoy left	02:17
	← Leefmc left	02:17
	→ claymccoy joined	02:18
	← saintdev left	02:21
ezyang	erh__: What happens when you do `git log --all -n 5`?	02:21
	erm -n5	02:21
	← ajonat left	02:22
	→ Leefmc joined	02:22
	→ ajonat joined	02:23
erh__	its not here	02:23
	well - i guess there is a merge there	02:24
	→ agib joined	02:24
	← aroben left	02:26
	← destruct left	02:26
	← claymccoy left	02:26
agib	does anyone know how to deploy a git branch with capistrano? cap deploy -s branch=dev only seems to work if the cached_copy is empty...	02:27
DrNick	you should probably ask in a capistrano support channel, not a git channel	02:28
	→ saintdev joined	02:28
	← krawek left	02:32
	← ngng left	02:32
agib	DrNick: true... just trying my luck, sorry	02:33
	stouset → stouset\|away	02:36
	destruct_ → destruct	02:36
	→ krawek joined	02:36
erh__	i tried git-rebase -ip but that just seemed to modify the working directory	02:38
agib	is there a way to make it so that my deployment server fetches the dev branch reference during a git fetch origin ?	02:39
ezyang	agib: Doesn't it already?	02:40
	→ jordi_bunster joined	02:41
agib	ezyang: hmm doesn't seem like it: "git fetch -q origin && git reset -q --hard d212518d499757d41ec19ae41c74bc9fc05c5f2f" seems to be giving me "Usage: /usr/local/bin/git-reset [--mixed \| --soft \| --hard] [<commit-ish>] [ [--] <paths>...]"	02:42
	← jerbear left	02:43
agib	old version of git or something?	02:43
Eridius\|work	agib: sounds like it doesn't like -q.	02:43
	which would imply old version of git	02:44
agib	1.5.3.7...	02:44
	ahh...	02:44
	gotta upgrade	02:44
Eridius\|work	-q was in 1.5.4	02:44
agib	ahh that explains it	02:44
	thank you	02:44
	what's the latest recommend version now?	02:45
Eridius\|work	is 1.5.6 out yet?	02:45
	if not, then 1.5.5.4	02:45
	no, looks like we're on 1.5.6 rc 3	02:46
agib	1.5.5.4 it seems from gitorcz -- lol sorry for all the questions	02:46
	how do get git with git? git clone git://git.kernel.org/pub/scm/git/git.git, check out the desired tag, then install as usual?	02:50
ezyang	agib: That should be sufficient, yes	02:52
Eridius\|work	agib: as long as you're doing it that way, you may as well install HEAD	02:53
	Eridius\|work sits on next	02:53
agib	HEAD?	02:53
	→ digitalhobbit joined	02:53
agib	are you saying that's overkill what I'm doing?	02:53
	:)	02:53
Eridius\|work	I'm saying why bother checking out a tag when you could just install the tip of master	02:53
	master is pretty darn stable	02:53
	next is also pretty stable, but not quite as stable as master	02:54
agib	ah	02:54
	ok	02:54
	← jordi_bunster left	02:55
agib	thank you	02:58
	← Eridius\|work left	03:02
agib	arg... make configure gives me /bin/sh: autoconf: not found on my ubuntu server...	03:03
	any thoughts?	03:03
	← zawir left	03:03
ezyang	sudo apt-get install autoconf	03:03
	← warthog9 left	03:04
	→ Geck1 joined	03:04
	← ketralnis left	03:04
agib	ezyang: I wonder how I got around that before... maybe older versions of git didn't require?	03:05
ezyang	agib: Well, maybe you're trying to run ./configure when you don't need to	03:05
	there's already a Makefile...	03:05
	→ imyousuf joined	03:06
agib	which git tells me I installed it in /usr/local/bin... so I figure I must have done some configure at some point	03:07
	← rtomayko left	03:07
	← erh__ left	03:07
ezyang	agib: there's a Makefile in the repos	03:08
	git's funny like that :-)	03:08
agib	:)	03:09
	← Geck1 left	03:13
	stouset\|away → stouset	03:15
	← vbabiy left	03:19
	← ph^ left	03:20
	→ Eredicat1rX joined	03:21
	→ ph^ joined	03:23
	→ pombreda joined	03:26
	stouset → stouset\|away	03:30
	→ digm joined	03:34
digm	When trying to clone my remote repository, I keep getting an errors. "git pack has bad object" and "fatal: index-pack failed". Anyone know what I can do to fix this?	03:35
ezyang	digm: git fsck?	03:36
digm	ezyang: I was looking at that, but I'm sorta new and wasn't quite sure what parameters to use with that.	03:36
aeruder	digm: i'd just go to your remote repo and do a git fsck --full	03:37
ezyang	just run it and see what happens	03:37
digm	aeruder: Getting an error like this "error: Packfile ./objects/pack/pack-9a91240fffee73d4cbb2d01d215473e43defecc9.pack SHA1 mismatch with itself	03:37
	error: 094c382bd7fa8eaec98aea799cae9717f5d1f727: object corrupt or missing"	03:37
aeruder	that's not good	03:38
	sounds like your repo is hosed	03:38
digm	aeruder: My local repository is fine, but i'd like to know how to fix the issue without having to redo the remote repo.	03:38
	is there a way?	03:38
aeruder	digm: not sure, although i'd be asking how it got that way	03:38
	i've seen corruption issues crop up on the mailing list from time to time, and they seem to pretty reliably not be git's fault	03:39
digm	aeruder: Not sure either :( Googling isn't that helpful.	03:39
aeruder	i.e. memtest failed or filesystem weirdnesses (especially nfs, etc.)	03:39
digm	aeruder: Eh, nothing special about my server. Just running ubuntu hardy, although I had a strange problem that when cloning, the cpu would lock while trying to do git-index-repack or some command like that.	03:41
	aeruder: Alright, redid repo, that took a whopping 10 seconds :D Any recommendations on backing up git repos? Scripts, tools, etc?	03:42
aeruder	digm: git clone usually serves well enough as a backup	03:43
	and keeping your other clones up to date	03:43
	but really, keep an eye out for other weirdnesses on your server	03:43
digm	will do, thanks for the pointers	03:43
aeruder	i'm not trying to say git doesn't mess up, but there are multiple cases on the mailing list where people have had things corrupted and it ended up being flaky hardware	03:43
	← CodeOfficer left	03:44
digm	aeruder: In this case, it wouldn't surprise me. I have weirdness from time to time. I was hoping there was some way to recover this issue though, like deleting the pack file (whatever that is) and then repacking or something.	03:44
ezyang	aeruder: How would you automate checks for "weirdness"?	03:46
aeruder	digm: well, you might have been able to rm -f the pack in question and repush some commit to force it to update more objects, but i don't know if that'd work	03:46
	ezyang: usually? you download a memtest iso :)	03:46
digm	yeah, in this case it was just easier to redo the remote repo. Good thing i have multiple backups unlike with my svn setup.	03:47
aeruder	digm: yea, on a public repo, there could literally be thousands of full backups :)	03:47
ezyang	aeruder: Oh, I mean remotely. I actually don't have access to the server :-/	03:47
aeruder	ezyang: i'm just saying that by weirdnesses, i meant memory corruption, hard drive corruption, etc..	03:48
ezyang	oh, ok	03:48
	Hmm, while we're on the subject of backup	03:48
	→ spearce joined	03:48
aeruder	if other programs are randomly shutting down	03:48
	etc..	03:49
ezyang	How do you make sure you git clone all your git repos to somewhere safe?	03:49
	I've been using git for everything, so I use git init willy nilly	03:49
	and some times I forget I have git repos lying around	03:49
	so I'm mortally terrified one day my hd will fail, and I'll lose all my repos	03:50
aeruder	i don't know, i backup my entire Project/ hierarchy everynight incrementally, so that works outside the realm of git..	03:50
ezyang	aeruder: good old fashioned rsync, or what?	03:50
aeruder	rdiff-backup	03:50
ezyang	will definitely look into that	03:50
aeruder	yea, i'd recommend it, not too bad to get going, you can grab files from old days (i.e. i want this file 12 days ago)	03:51
	and the latest backup is available just as normal files	03:51
	a little trickier to get to the old ones	03:51
	yea, i've lost repos before due to lack of backups, and it sucks, entire cvs histories of some of my projects are gone, and i had to start over with the latest tarball	03:52
spearce	harrrg. github just sent camino off the deep end with a long running script	03:52
aeruder	i'm thinking i'm going to go back to firefox (from camino)	03:53
spearce	i generally like camino, but this github script choked on something in a jgit fork i was poking around at	03:54
ezyang	aeruder: well, on the plus side, most people never look at the history :-/ losing history sucks though	03:54
	→ mbuf joined	03:55
	← Leefmc left	03:59
	stouset\|away → stouset	04:00
	spearce curses people who fork a git repository without taking the git history, and then publish it on github anyway.	04:01
digm	aeruder: So i'm trying rdiff-backup now. Question though. Let's say I backup for 20 days. And I run the --remove-older-than 10 days switch, but have made tons of changes over those 20 days. Will that remove changes that put the .git repository out of sync at all?	04:03
	→ ciaran joined	04:03
	← louipc left	04:03
ezyang	g'night all	04:06
	← ezyang left	04:06
	→ jerbear joined	04:07
jerbear	i ran a git-gc, and now refs/heads/master is gone... what do i do?	04:08
	possibly a better question is.. what did i do wrong?	04:08
	hello?	04:10
	→ tjafk1 joined	04:11
	← kukks left	04:12
	stouset → stouset\|away	04:15
jerbear	i ran a git-gc, and now refs/heads/master is gone... what do i do?	04:16
	← jerbear left	04:17
	→ CodeOfficer joined	04:18
	← digm left	04:20
	← mbuf left	04:23
	← tjafk left	04:27
	→ priidu joined	04:29
	→ mbuf joined	04:47
	← madewokherd left	04:49
	→ capisce joined	04:54
	← Ariella left	04:57
	→ ketralnis joined	04:58
	← capisce_ left	05:08
	→ Tsaghen joined	05:09
	→ garrin joined	05:18
	→ spearce` joined	05:19
	← spearce left	05:19
	→ ketralnis_ joined	05:23
	← ketralnis left	05:24
	← igorgue left	05:24
	→ cedricv joined	05:27
	→ Ariella joined	05:30
	→ nud joined	05:37
	→ fhobia joined	05:42
lea_w	How come I have two packs in my .git/objects/packs dir after running git-repack -a -d?	05:46
	← bobesponja left	05:47
lea_w	One of them has a .keep a file, the other one doesn't.	05:48
spearce`	probably from when you cloned; the .keep file was created by git-clone.	05:48
	if the pack is under 200M i'd delete the .keep file and repack -a -d again to consolidate the two	05:49
lea_w	so why is it being kept?	05:49
spearce`	because git-clone though tit should always keep the first pack you get from the project.	05:49
lea_w	iow why are they not always consolidated?	05:49
spearce`	that rule works for _huge_ projects like say all of Mozilla or KDE.	05:49
	but its less useful for smaller projects like linux-2.6 or git.git	05:49
lea_w	hm. what's the advantage of that? mildly-confused	05:50
spearce`	well, some of the really big projects can be upwards of 1.5 GB in that initial pack file.	05:50
	copying that data over to a new pack file every time you repack can take a while.	05:50
	disks are not that fast. :)	05:50
lea_w	oh, that makes sense.	05:51
	well, thanks!	05:51
spearce`	but smaller projects under 300M or so can be repacked pretty quickly into a single pack on most systems. so its generally better to have just one pack with everything in it, and thus the .keep is actually counter-productive	05:52
	← Sake left	05:54
	→ anathematic joined	05:56
	→ kewark joined	05:58
	← kewark left	05:58
	→ Ramune joined	06:03
	ketralnis_ → ketralnis	06:05
	← krawek left	06:12
	← ketralnis left	06:15
	← csc` left	06:17
	→ deavid joined	06:17
	→ charon joined	06:25
	← spearce` left	06:27
	→ ketralnis joined	06:29
	← digitalhobbit left	06:36
	→ bentob0x joined	06:36
Bombe	Is there a way to disable the pager for the output of only certain commands?	06:38
	→ reel_ joined	06:38
	→ capisce_ joined	06:41
	← capisce left	06:43
	← nud left	06:44
	← pygi left	06:46
	→ robinr joined	06:47
	← ciaran left	06:51
	→ warthog9 joined	06:51
	→ naeu joined	06:51
	← naeu left	06:54
	← CodeOfficer left	07:00
	→ CodeOfficer joined	07:00
	→ progress0r joined	07:00
progress0r	if i have created a branch locally and commit then push to a github repo, will it be kept seperate (as a branch) or auto merge with master?	07:01
	← ajonat left	07:06
aeruder	a push does nothing more than upload your current commit objects/update remote refs	07:08
	no new commits can or will be created during a push	07:08
	so if it looks right on your side, and you push it successfully, it will look the same there	07:08
	→ naeu joined	07:08
	← robinr left	07:09
	← ThunderChicken left	07:09
progress0r	aeruder: thanks - thats what i wanted, keep it seperate	07:10
	→ lirit joined	07:11
lirit	does git support ssh1 or ssh2?	07:12
aeruder	lirit: git uses the system ssh	07:13
lirit	aeruder: thanks	07:13
aeruder	lirit: i.e. it actually calls out to the ssh executable	07:14
lirit	got it.	07:15
	→ ph^_ joined	07:16
	→ nud joined	07:19
	← naeu left	07:20
	← CodeOfficer left	07:21
	→ ThunderChicken joined	07:25
	← imyousuf left	07:32
	→ imyousuf joined	07:34
	→ ajonat joined	07:34
	← ketralnis left	07:36
	← Ademan left	07:36
	→ JonathanT joined	07:36
	← ajonat left	07:37
	→ ajonat joined	07:38
	← lirit left	07:38
	← jmspeex left	07:45
	← imyousuf left	07:47
	→ imyousuf joined	07:52
	→ Fullmoon joined	07:53
	→ Ademan joined	07:55
	dwmw2_gone → dwmw2	07:58
	→ naeu joined	07:58
	← lu_zero left	08:01
	← tenpaiyomi left	08:02
	← Fullmoon left	08:03
	→ _zack joined	08:05
	→ radarek joined	08:09
dmq	if I have a file's sha1 can i find all the commits that it was part of? i thought git-show would do this, but it doesnt.	08:13
	→ Fullmoon joined	08:14
	← ThunderChicken left	08:14
dmq	or it appears it doesnt do it anyway	08:14
	→ ThunderChicken joined	08:14
Tv	dmq: only by walking through every tree of every commit	08:15
dmq	ah, actually the problem was that git doesnt use sha1sum to calculate the SHA1 of the file, and thus has a different digest.	08:17
Tv	it's the sha of the compressed version of the file, as an optimization	08:17
	plus many objects have headers	08:17
Bombe	Is there a way to disable the pager for the output of only certain commands?	08:17
mnemoc	PAGER=cat git-foo .... should work	08:18
dmq	the question i am trying to address is that someone wants to know "how can they find out the 'version' of a uncontrolled copy of a file that is ultimately controlled by git"	08:19
Arjen	I remember seeing a script for that on the wiki, trying to find it again	08:19
	→ mcella joined	08:20
Tv	dmq: your problem is that "uncontrolled copy"	08:21
dmq	this stems from the issue that RCS tags like $Revision$ dont make a lot of sense in a distributed environment. so the question comes up "well then how else do you identify a file"	08:21
	→ paulproteus joined	08:21
dmq	yes I know.	08:21
	→ redondos joined	08:21
	← _zack left	08:21
Pieter	dmq: git hash-object gives you the sha	08:21
Tv	dmq: i'd suggest you embed git describe output in the tarball/file when exporting	08:21
paulproteus	Is there a way to explicitly ask git-filter-branch to filter every revision?	08:21
dmq	I get the idea that you Just Shouldnt Have Uncontolled Copies, but my colleagues who are all too used to CVS dont.	08:21
	→ _zack joined	08:22
Tv	Pieter: but even then, the answer would be a multiple commits	08:22
Pieter	yes	08:22
paulproteus	Oh, hi Tv (-:	08:22
dmq	yes i got it.	08:22
Tv	paulproteus: hi. what do you mean every revision?	08:22
Arjen	dmq: http://git.or.cz/gitwiki/ExampleScripts (second script)	08:22
Tv	paulproteus: --all?	08:22
dmq	sorry, yes i figured out git-hash-object. and im happy to see every commit its part of.	08:22
Tv	dmq: and then they modify it a bit and you get nothing out	08:23
paulproteus	Tv, That's not listed in the man page, fwiw..	08:23
Tv	paulproteus: "<rev-list options>"	08:23
Arjen	Hm, that script does too much, I think	08:23
paulproteus	Yes, I was about to say I don't know what sort of stuff one is allowed to jam in there. (-:	08:23
Tv	paulproteus: i have no clue if it works	08:23
paulproteus	I'll tell you in a sec!	08:23
	Tv, No, it spits back "Usage:" of git-filter-branch.	08:24
Tv	paulproteus: yeah i sort of expected that	08:25
	paulproteus: as it needs refs to update, and rev-list usually outputs commit shas	08:25
	← paakku left	08:25
Pieter	git filter-branch `git-for-each-ref` :P	08:25
Tv	paulproteus: i mean, for-each-ref will list ass of them	08:25
Pieter	wouldn't do anything like that though	08:25
paulproteus	lol! Why not?	08:25
Tv	paulproteus: i don't know if filter-branch is smart enough to not screw up the parent relations	08:25
paulproteus	It discusses in its man page how it has tried to be smart about parent relationships.	08:26
	That discussion only applies to when you remove commits. I'm "just" modifying the trees.	08:26
Tv	which changes parent pointers	08:26
	i don't see any example updating any ref except via HEAD	08:27
paulproteus	Modifying the trees without removing commits shouldn't change parent pointers, and that's all I'm trying to do.	08:27
Tv	which is slightly discouraging	08:27
	oh sure it does	08:27
	modifying the tree changes the commit sha, so anything pointing to old commit needs to be adjusted to point to new commit	08:27
paulproteus	Sure, okay.	08:27
	I'm vaguely under the impression it should handle that properly.	08:27
	It handles linear history properly, at least, changing parents to the new, filtered version.	08:28
	← Fullmoon left	08:28
Tv	# The refs should be updated if their heads were rewritten	08:28
	git rev-parse --no-flags --revs-only --symbolic-full-name --default HEAD "$@" \|	08:28
	sed -e '/^^/d' >"$tempdir"/heads	08:28
	hmmmm	08:28
	→ Fullmoon joined	08:29
Tv	i think that makes `git for-each-ref` safe	08:29
statim	if anyone has a second, i wrote up some basic instructions for one of the developers im working with on how to use git. since im also really new to git if anyone has comments on what im telling him that would be great. http://pastie.org/private/ho2tiswwhvfnsrytuxgew	08:29
	← mbuf left	08:30
	← Fullmoon left	08:30
paulproteus	Tv, Passing in $(git-for-each-ref) as the list of refs causes git-filter-branch to tell me back, "Which ref do you want to rewrite?"	08:31
	Oh, I see it just passes that to git-rev-list.	08:32
Tv	paulproteus: oh probably something about output format	08:32
paulproteus	Hah! That's why you guessed --all.	08:32
Tv	for-each-ref --format='%(refname)'	08:32
	← pdherbemont left	08:33
paulproteus	And so it churns.	08:33
	← sverrej left	08:36
Bombe	mnemoc, I meant more as a config option, so that git-status is never shown with a pager but git-diff is.	08:37
mnemoc	Bombe: no clue	08:38
	→ capisce joined	08:39
	→ munga joined	08:40
	← fhobia left	08:45
	← reel_ left	08:47
	→ reel_ joined	08:47
bartocc	hi	08:52
	I have a local branch that mirrors a branch on github	08:52
	I've made changes and would like to push them to github branch	08:53
	git push does nothing	08:53
	what am I missing ?	08:53
Arjen	Have you committed the changes?	08:53
context	ive said it before and ill say it again	08:53
	← capisce_ left	08:53
context	man i love bein a turtle	08:53
bartocc	Arjen: on my local branch, yeah	08:53
	I've also pulled from origin, to make sure I am up to date	08:54
	git push => Everything up-to-date	08:54
	but	08:54
	git diff mb origin/mont_blanc => show me the differences	08:54
	← mithro left	08:55
Tv	context: but are you a ninja, err i mean a hero turtle?	08:57
	bartocc: different branch names -> git doesn't know what's supposed to be pushed where	08:57
	bartocc: git push origin mb:mont_blanc	08:57
	← cedricv left	08:58
bartocc	Tv: that worked, thx	08:59
	but I don't get it	08:59
	I have this in my .git/config	08:59
	[branch "mb"]	08:59
	remote = origin	08:59
	merge = refs/heads/mont_blanc	08:59
	should't git be able to figure that for me ?	09:00
Tv	bartocc: that's one-way	09:00
Bombe	Hmm. I unset core.pager everywhere but git-status is still shown in /usr/bin/less?!	09:00
Pieter	Bombe: that's the default	09:01
	set the pager on cat or so	09:01
Bombe	Unsetting PAGER does the trick.	09:01
	Hmm. That's not optimal, though.	09:01
Pieter	cat is a special keyword that means 'don't page'	09:01
Bombe	Actually, cat is a binary that simply writes stdin to stdout. :)	09:02
Pieter	yeah	09:02
	I'm takling about git's pager thin	09:02
	g	09:02
	it has cat as a name to turn paging off	09:02
	similar to 'none' or so	09:02
Bombe	Small optimizations. :)	09:02
Tv	git config --global core.pager 'cat\|cat\|cat\|cat'	09:03
Bombe	And the "useless use of cat award" goes to: Tv! \o/	09:04
charon	i prefer 'tac\|tac'	09:07
Pieter	especially with streams	09:07
Bombe	:)	09:07
	→ ebel joined	09:09
	→ sverrej joined	09:09
	← mtkd left	09:14
	→ pygi joined	09:16
	← WildPikachu left	09:17
	← _zack left	09:18
	→ _zack joined	09:19
Tv	btw nice screenshot titled "SVN is hard": http://www.flickr.com/photos/davidreid/2588038044/	09:32
mnemoc	ROTFL	09:33
	← das-q left	09:34
progress0r	ugh i just had a network fail pushing my local branch and now git status says nothing to commit.	09:37
	looking at github i see nothing came through	09:38
	wait do i need to remote add my local branch to github? i didnt do that	09:38
	← progress0r left	09:40
Tv	well that was useful	09:40
	← petteri left	09:45
	→ janm joined	09:46
	→ petteri joined	09:47
	→ rofrol joined	09:53
	→ chris2 joined	09:56
	→ mbuf joined	10:00
	← lea_w left	10:02
	← petteri left	10:05
	→ petteri joined	10:07
	→ ijcd____ joined	10:14
	→ visik7 joined	10:22
visik7	hi	10:22
	is it possible to use git.kernel.org over http ?	10:22
	→ lea_w joined	10:23
	← ijcd left	10:25
	→ Fullmoon joined	10:25
	→ mithro joined	10:26
	→ zorglu__ joined	10:33
zorglu__	q. when launching qgit, i always get a popup window "range select" and i always choose the default. is it possible to avoid this popup all the time ?	10:35
	ŝorry im a morron. just found it. it is a radio on the bottom of the popup	10:36
	moving along with shame in the heart	10:37
	← zorglu__ left	10:37
	→ Xyri0 joined	10:43
	→ lu_zero joined	10:47
	← Xyrio left	10:54
	Xyri0 → Xyrio	10:54
	← EmilMedve left	10:55
	→ mtkd joined	10:58
	← ruphy left	10:59
	← petteri left	10:59
	→ petteri joined	11:00
	→ adante joined	11:03
	← MarcWeber left	11:07
	← anathematic left	11:08
	→ VitaRara joined	11:08
	← imyousuf left	11:09
	pilif\|away → pilif	11:10
	→ ruphy joined	11:17
	→ Ace_NoOne joined	11:17
Ace_NoOne	hi - I'm using git for working with an SVN repo - however, when creating patches the path isn't entirely compatible with the way SVN does this	11:18
	e.g. a/foo/bar instead of Trunk/foo/bar	11:18
	is there a workaround for this?	11:18
Pieter	I think everybody just uses patch -p1 in the trunk dir	11:21
	← kanru left	11:24
pilif	Ace_NoOne: there's always sed if you need to hide the fact that you are using git when creating those patches. Aside of that, -p1 is probably needed in SVN too.	11:25
Ace_NoOne	I'll just have to tell our team to use -p1 then - thanks	11:26
	← mithro left	11:28
mnemoc	Ace_NoOne: you can also `sed` the patches converting the a/ to ./	11:28
Ace_NoOne	mnemoc: yeah, but having to remember that would be a hassle	11:29
pilif	well... either you remember to sed or your team remembers to -p1 - your choice :p	11:30
mnemoc	:)	11:30
	→ kanru joined	11:30
pilif	also, you could always create a shell function that automatically does the sed for you	11:31
Ace_NoOne	I know, but I can't be bothered - using p1 is simpler	11:31
pilif	for you - but not necessarily for your team	11:32
Pieter	who cares :)	11:32
pilif	if you want to convert them to git in the long term, it's probably better to to everything possible now to show them that they won't have to worry about git - even if it's just one team member using it	11:32
broonie	GNU patch often manages to figure this out anyway...	11:33
pilif	heck - I'm even leading a team here and I'm doing everything I can to make those stuck with svn not notice me using git.	11:33
broonie	Also note that half the reason git uses the format it does is that it's always been the standard diff format in the community anyway.	11:33
pilif	yepp. and it's IMHO very clear on what it's diffing. I MUCH prefer the git diffs	11:33
	→ icwiener joined	11:40
	→ csc` joined	11:44
	→ SuttoL joined	11:45
ska-fan_	Hi. I did git fetch and now I need to move my master marker to the origin master commit - how do I do that?	11:45
	→ Kbyte joined	11:45
mnemoc	ska-fan_: rebase	11:46
ska-fan_	thanks	11:49
	← deavid left	11:54
	→ git joined	11:57
_zack	MadCoder: you are using vim/git plugin on debian, aren't you? by default when I commit vim split a buffer with the diff vertically	11:57
	MadCoder: do you know if it is possible to have it split horizontally>	11:57
	?	11:57
	→ madewokherd joined	11:59
	← chris2 left	12:03
	← Sho_ left	12:04
	→ Sho_ joined	12:04
doener_	_zack: the header in mine says that you should set g:git_diff_spawn_mode	12:05
_zack	doener_: uhm, let me see, I've probably overlooked it :)	12:06
	doener_: which .vim file are you looking at?	12:06
	I've looked at ftplugin/{gitcommit,git}.vim and I don't see any useful header	12:07
	→ mislav joined	12:07
doener_	git.vim	12:07
	_zack: which version are you using? mine's dated 26 Nov 2007	12:07
_zack	git.vim where? ftplugin/git.vim?	12:08
doener_	yep	12:08
_zack	mine is 2008 Feb 27	12:08
doener_	hm, weird, I'd expect newer ones to have more info, not less	12:08
_zack	i'll grep for spawn_mode	12:08
	nothing indeed	12:09
	bah	12:09
	→ sbeyer joined	12:09
	← kanru left	12:10
doener_	_zack: hm, on vim.org, it's called git-commit, so maybe it ftplugin/gitcommit.vim for you. I got it via debian's vim-plugins package	12:11
_zack	doener_: same here in fact	12:11
	debian/unstable	12:11
doener_	hm, me too	12:11
	why do you have a different file then?!	12:11
_zack	and I guess you mean vim-scripts	12:12
doener_	yeah, probably	12:12
_zack	$ sha1sum /usr/share/vim/vimcurrent/ftplugin/gitcommit.vim	12:12
	c3f8be57a0eb29e4bf52452c56f3fc56a8d8d491 /usr/share/vim/vimcurrent/ftplugin/gitcommit.vim	12:12
	→ lorandi joined	12:13
_zack	vim-scripts version 7.1.7	12:13
doener_	that's not from vim-scripts, that's from vim itself	12:13
_zack	ah	12:13
	right	12:13
	so I'm looking at the wrong file for the header	12:14
doener_	and it's not authored by madcoder, so that explains the confusion :-)	12:14
dato	_zack: if you open a vim and run ":pedit /", is the new window horizontal or vertical?	12:15
	_zack: and if you open vim with vim -U NONE -u NONE and do the same, is it now horizontal?	12:16
	_zack: and hi, btw :)	12:17
_zack	dato: that's exactly what puzzles me, I've tried pedit and by default it does that horizontally, so I don't get yet why pedit inside git support does that vertically	12:17
	dato: hi :)	12:17
	dato: the following wincmd do not seem to change window orientation either ...	12:18
dato	wincmd P wouldn't change the orientation..	12:18
_zack	dato: indeed	12:18
dato	do you run DiffGitCmd by hand?	12:19
_zack	dato: you using it? you got the vertical split too?	12:19
dato	er, DiffGitCached	12:19
_zack	dato: nope, just do "git commit" from the shell	12:19
jettero	I like that graph that git pull shows when it does the merge... is there a way to show it other than when you pull?	12:19
dato	I use it, and it gets opened horzontally	12:20
	_zack: hm, and what's opening it? afaik, vim-git does not call DiffGitCached automatically	12:20
	is it possibly you have two sets of git stuff installed?	12:20
Pieter	what graph?	12:21
_zack	dato: interesting idea, it is possible, let me check	12:21
	→ kanru joined	12:21
_zack	dato: yup, it was the git-commit plugin	12:21
	dato: so, how do you obtain the splitted window then? do you invoke DiffGitCached by hand?	12:22
dato	I have this:	12:22
	au FileType gitcommit setlocal et sw=2 pvh=30 \| DiffGitCached \| wincmd r \| wincmd k	12:22
_zack	dato: uhm, should probably be added to vim's global configuration then ...	12:23
dato	well, I don't think everybody wants a diff window automatically opened	12:23
_zack	dato: then it should be configured via a variable, having to add an autocommand by hand is basically defeating the usefulness of filetypeplugins :)	12:25
dato	_zack: to that I can agree	12:25
	dato points to tpope	12:25
	← Fullmoon left	12:25
	→ kukks joined	12:26
_zack	tpope: by the way, the default splitting imo should be horizontal, has it is rare to have terminals/consoles large enough to hold 160 characters	12:27
dato	_zack: mm, I thought we just agreed that it is?	12:27
Bombe	~/workspace/limiter# echo $COLUMNS	12:28
	237	12:28
_zack	dato: I haven't tried, but if it is why in your autocmd you have wincmd k	12:28
	ah, no, my fault	12:28
	I misread "k" for "K"	12:28
dato	ah :)	12:28
_zack	I was thinking you were moving the window around	12:29
dato	I'm swapping the diff window to be at the bottom, and not focused	12:29
	→ msh joined	12:29
msh	Hi!	12:29
	I just did a commit	12:29
	how can I revert that	12:29
_zack	msh: revert or improve?	12:30
	you can git commit --amend to fix it	12:30
	or you can add a new commit "undoing" the last one with git revert	12:30
	← lnostdal left	12:30
msh	basically i did a commit but that has to be totally not there	12:30
	possible?	12:30
_zack	or you can git reset --hard HEAD^	12:30
	stouset\|away → stouset	12:30
_zack	which is what implements your last request :)	12:30
aeruder	msh: that command _zack just said will wipe all local modifications too, so be careful	12:31
_zack	big fat warning: I'm assuming the commit you want to thread away is the last one	12:31
	→ deavid joined	12:31
msh	yes its the last one	12:31
	actually, I did not do an add before doing a commit	12:31
	I mean not added all the files	12:32
_zack	msh: you will loose the not added changes	12:32
	ec → elliottcable	12:32
aeruder	msh: you may be better off just doing git reset HEAD^	12:32
	(without the --hard)	12:32
	that'll leave the working directory as it is, and undo the last commit	12:32
	if there are files you want to reset the changes in, then you can just do git checkout -- <file>	12:33
msh	as far as the documents say, git reset HEAD will unstage the files	12:33
	→ stouset_ joined	12:33
aeruder	msh: yes, but you want to reset the last commit too	12:33
	so git reset HEAD^	12:34
ska-fan_	Is there a command that has output similar to what gitk shows, but for the console?	12:35
	← stouset_ left	12:35
ska-fan_	with ASCII art	12:35
	→ stouset_ joined	12:35
	← stouset_ left	12:35
Bombe	ska-fan_, git-forest.	12:35
	→ stouset_ joined	12:35
Mikachu	git log --graph in 1.5.6	12:36
_zack	msh: HEAD^ is the (first) parent of the HEAD commit (i.e. the last commit)	12:36
ska-fan_	debian seems to have neither, ok, thanks	12:36
	← stouset left	12:36
	stouset_ → stouset	12:36
Bombe	ska-fan_, Google does. :)	12:36
aeruder	tig can too	12:36
	← stouset left	12:36
	→ stouset_ joined	12:37
	← stouset_ left	12:37
	→ stouset joined	12:37
	← Mr_Mirsal left	12:38
Mikachu	maybe you can run some sort of SDL X server under aalib	12:39
vmiklos	haha	12:40
	← lorandi left	12:43
	→ Ryback_ joined	12:43
	→ chrisb joined	12:44
	→ spearce joined	12:46
	→ noodl joined	12:48
noodl	hi. i've just made a commit on master but want to switch that commit over to a new branch and continue on master from the previous commit (HEAD^). how is that done? git branch new-feature HEAD^ would create the branch but master would still have this commit i don't want..	12:50
vmiklos	git checkout -b new-feature	12:51
dato	rather:	12:51
vmiklos	git checkout master; git reset --hard HEAD^	12:51
Mikachu	in this case i think checkout -b is a roundabout	12:51
	just do git branch new-feature; git reset --hard HEAD^	12:51
vmiklos	ah yes, you can use git branch directly instead of git checkout -b and then you need only one checkout	12:51
Mikachu	assuming you're on master already	12:52
	→ d0k joined	12:52
noodl	bah, irssi having trouble drawing this channel	12:52
	→ Fullmoon joined	12:52
noodl	thanks Mikachu that's the ticket	12:53
Mikachu	you _can_ do git checkout -b newbranch; git push . +HEAD^:master , but it's a bit ugly	12:55
	→ gitte joined	12:58
	→ zawir joined	12:59
redondos	Mikachu: can you help me parse that last argument?	12:59
Mikachu	redondos: the + means the same as -f	12:59
redondos	ahh	13:00
	yes, found it in git-push(1)	13:00
	thanks	13:00
	→ snowblink joined	13:02
	← spearce left	13:02
	→ MarcWeber joined	13:03
MarcWeber	When using git-svn is there a way to get the git-svn hash by revision after importing ?	13:04
redondos	git-svn find-rev	13:05
MarcWeber	Ah thanks. I haven't seen it.	13:05
	→ capisce_ joined	13:07
	← git left	13:10
MarcWeber	Strange did not work with that repo. With another it did	13:10
	→ lorandi joined	13:11
redondos	MarcWeber: you have to specify an existing revision in that branch	13:11
MarcWeber	Ah this means I'm at the wrong branch?	13:11
redondos	probably.	13:12
MarcWeber	Unfortunate.. So I need a for `git-branch -a`; do .. or such ok.	13:12
	Anyway it's fun on a 400MB repo	13:14
	→ lea_w1 joined	13:15
	← lea_w left	13:15
	lea_w1 → lea_w	13:15
MarcWeber	Is there a git-checkout --soft without touching the work dir?	13:16
Mikachu	git read-tree	13:16
	+ git reset i guess, if you want both the index and HEAD to move	13:17
	(reset --soft obviously)	13:17
redondos	he just wants the HEAD to move	13:17
MarcWeber	Propably the right way to go is grepping the svn repo from .git/config and ask that for the right branch :-)	13:18
Mikachu	then just reset --soft	13:18
redondos	yes	13:18
MarcWeber	boy. Obviously. I can pass any hash.. not only HEAD^^	13:18
	I should have know that already	13:18
redondos	or branch names	13:18
	← stouset left	13:20
	→ stouset joined	13:20
	→ lnostdal joined	13:21
	→ queso joined	13:21
	← capisce left	13:22
	→ capisce joined	13:23
	← Fullmoon left	13:25
	→ ast3 joined	13:26
	comp_ → comp	13:26
	← stouset left	13:27
	← weswh- left	13:28
	← VitaRara left	13:29
	→ vbabiy joined	13:32
	← capisce_ left	13:35
	→ pflanze joined	13:36
	→ Beket joined	13:38
	← SRabbelier left	13:39
	→ SRabbelier1 joined	13:40
	SRabbelier1 → SRabbelier	13:40
	← peeja left	13:40
	← SRabbelier left	13:41
	→ SRabbelier1 joined	13:42
	SRabbelier1 → SRabbelier	13:42
mcgrof	is there a way to somehow to cheat git to think it doesn't have a few files?	13:42
redondos	.gitignore?	13:43
mcgrof	hehe...	13:43
Pieter	update-index --assume-unchanged	13:43
	oh	13:43
	nevermind	13:43
mcgrof	basically I ran into an issue with git-p4	13:43
	and p4 no longer has the files, and git does though	13:44
	→ cedricv joined	13:44
mcgrof	now I have people who pull from me though so they still have the files too	13:44
	git-p4 works by keeeping a remotes in p4/master in sync with the depot stuff	13:45
	oh I think I got it...	13:45
	hmm	13:45
	lets see	13:45
	← ast3 left	13:50
mcgrof	nope..	13:50
	oh well, ne wtree	13:53
	s/ne/new	13:53
	→ stouset joined	13:54
	← Sho_ left	13:59
	← ckknight left	14:00
	→ peeja joined	14:00
	← fafhrd left	14:02
	← charon left	14:02
	← ph^_ left	14:04
	→ charon joined	14:05
	← mahoski left	14:08
	→ ckknight joined	14:09
	telmich wonders why so many people like the rebase idea... for me it just breaks the history	14:10
	← SRabbelier left	14:10
	→ SRabbelier1 joined	14:10
mislav	telmich: for your personal development it often makes a lot of sense	14:10
	SRabbelier1 → SRabbelier	14:11
telmich	mislav: must confess I never use rebase, only with svn, but that's another story	14:11
doener_	hm, you get your commits perfect on the first try? Then there's no need for rebase -i	14:12
	but I really love it for bringing a series of commits into shape before sending them out for review	14:12
telmich	yep, I almost always do git-status; git-diff; git-commmit	14:13
	-%	14:13
	-m	14:13
mcgrof	if people can just follow the rules to commit...	14:13
	I mean to rebase	14:13
	← SRabbelier left	14:13
	→ SRabbelier1 joined	14:13
telmich	the big problem I see is that I am forced todo git-push -f which may overwrite other commits	14:13
	SRabbelier1 → SRabbelier	14:14
doener_	telmich: well, you're not supposed to rebase after you published the commits	14:14
comp	telmich: well ... if you want to make several commits to "mainline" from your devel branch (squach 120 commits into 8 for example) then rebase it's better idea than git-diff and git-apply ...	14:14
doener_	(unless you're on a branch that is known to be rebased, like git.git's origin/pu)	14:14
	← pflanze left	14:15
mcgrof	yeah in those cases we use +branch for the push and it works fine it seems	14:15
comp	I personally use rebase as some sort of push preparation ...	14:15
	→ fafhrd joined	14:15
telmich	comp: ok, that maybe a reason I did not see, as I mainly receive, not send patches	14:15
	← fafhrd left	14:16
	← SRabbelier left	14:16
	→ fafhrd joined	14:16
	→ SRabbelier1 joined	14:17
	SRabbelier1 → SRabbelier	14:17
	← lea_w left	14:21
	→ Mr_Mirsal joined	14:23
	← SRabbelier left	14:24
	→ SRabbelier1 joined	14:24
	SRabbelier1 → SRabbelier	14:25
tpope	_zack: pong	14:25
	→ omygawshkenas joined	14:25
	← peeja left	14:25
tpope	fwiw, I think having a bunch of configuration options that could easily be replaced with a simple autocmd is superfluous	14:26
	→ Jeraimee\|Wrk joined	14:28
	← Jeraimee\|Wrk left	14:30
	→ powerfox joined	14:31
powerfox	Hello! What is the best way to know if the current dir in Git?	14:32
	→ lea_w joined	14:32
	← lorandi left	14:32
powerfox	ajonat: .git, b_dir. I'm in b_dir, how can I understand b_dir is g«itted»	14:32
	?	14:32
jast	run any command that requires being in a git directory (e.g. git status). it will error out if you're not in a repo.	14:33
powerfox	jast: and better ways? I need to do it from C++ application, and doing if (task-exec() ) is not pretty	14:35
mnemoc	powerfox: hardcoding fragile git logic in your code is less pretty	14:36
jast	well, you can always check if certain things in $foo/.git exist, but it will be difficult to detect corner cases the same way git does	14:37
	→ tenpaiyomi joined	14:38
powerfox	mnemoc: maybe you're right	14:38
	← tenpaiyomi left	14:38
	→ ketralnis joined	14:38
	← ketralnis left	14:38
	→ ketralnis joined	14:38
powerfox	jast: after repo was created some dirs/files may be added, so the project's root can't help	14:39
jast	what do you mean?	14:39
	the structure of .git doesn't change much	14:39
omygawshkenas	'morning #git	14:41
	← garrin left	14:45
	← rofrol left	14:46
mnemoc	powerfox: btw, git-rev-parse --is-inside-git-dir is what you want to exec() to know if you are in a git tree	14:46
	blowery`afk → blowery	14:47
	→ redinger joined	14:47
mnemoc	err --is-inside-work-tree	14:48
powerfox	mnemoc: thanks. I think I will use it. But it has one problem: git init &&... mkdir foo && cd foo && git-rev-parse --is-inside-work-tree gives true	14:50
mnemoc	sure	14:50
bremner	powerfox: why is that a problem	14:51
ciaran29d	why is that unexpected :/	14:51
powerfox	mnemoc: but I think this would be enough :) Thanks a lot. bremner: I have to know if the dir was added to the git repo	14:51
	← SuttoL left	14:52
powerfox	In some cases it can be a git subproject	14:52
omygawshkenas	When dealing with conflicts while trying to "git rebase" against a git master of an svn trunk (whew), what's the deal with the .dotest directory mucking things up?	14:52
Ilari	powerfox: IIRC, you can query if there are files with given path prefix using ls-files...	14:52
omygawshkenas	I resolve the conflict with git mergetool, and try to "git rebase --continue", but it throws a buncha:	14:53
	error: .dotest/0016: already exists in working directory	14:53
	error: Untracked working tree file '.dotest/0008' would be overwritten by merge.	14:53
	Are you supposed to do something to .dotest after fixing conflicts?	14:53
powerfox	Ilari: thanks a lot. Didn't think about it.	14:53
mnemoc	powerfox: if it's a git subproject it will be a git work-tree anyway	14:53
Ilari	omygawshkenas: You shouldn't have added those .dotest files...	14:53
omygawshkenas	I didn't	14:54
powerfox	mnemoc: yes :) I forgot that subproject contains .git too as I remember :)	14:54
	Thanks to all for help :)	14:54
	I have a big plan to finish some code and make sexy screenshots :)	14:55
mnemoc	:)	14:55
omygawshkenas	llari: oh shit, from before...	14:55
	If I know the commit where I mistakenly put them in, is it possible to go back and remove that commit from the history?	14:57
	← sverrej left	14:57
Ilari	omygawshkenas: Since it is dcommitted to SVN, nope...	14:58
	→ aroben joined	14:58
Ilari	omygawshkenas: Probably abort the rebase (rebase --abort), create new branch starting from trunk, delete those files commit, dcommit and then retry the rebase?	14:59
	→ bdiego joined	15:02
	blowery → blowery`afk	15:06
	→ EmilMedve joined	15:06
	blowery`afk → blowery	15:06
omygawshkenas	Nope, it hasn't been dcommitted to svn yet.	15:06
	← etnt left	15:06
omygawshkenas	The mistaken .dotest files are in the git feature branch.	15:06
	Thank god.	15:06
	→ rdavila joined	15:06
	← rdavila left	15:07
	← mbuf left	15:07
Ilari	omygawshkenas: Then it is possible to remove it using interactive rebase (mark the commit in question to be edited).	15:07
omygawshkenas	Can I just git revert sha_of_bad_commit ?	15:08
Ilari	omygawshkenas: For finding it, 'git log -- .dotest' might be useful.	15:08
	← msh left	15:08
Ilari	omygawshkenas: If it is just in feature branch, then editing it away would be better solution.	15:08
	← visik7 left	15:09
omygawshkenas	Thanks. I'll give it a shot.	15:10
	→ jpeeler joined	15:11
	→ hobbes006 joined	15:11
	← ketralnis left	15:11
hobbes006	hi guys: quick question. how do i go about creating a gitignore file and tell git to ignore specific .pyc and .mo files? i found this page (http://download.ikaaro.org/doc/packaging/sect0004.html) but it isnt very descriptive. could someone point me in the right direction? thks.	15:12
	→ krawek joined	15:13
Ilari	hobbes006: Basically, .gitignore contains shell glob patterns, one per line of files to ignore. If no '/' is present, then it is appiled to all directories below the directory .gitignore is in, otherwise only to that directory.	15:13
	hobbes006: See gitignore(5).	15:14
hobbes006	so i should write *.pyc/	15:14
	from my project root directory ?	15:14
Ilari	hobbes006: If you want to ignore all .pyc-files that would be '*.pyc'.	15:15
	hobbes006: And file would go to project root dir.	15:15
hobbes006	got it. thanks for the tip :)	15:15
	just needed clarification.	15:15
	← mcella left	15:17
Ilari	hobbes006: I don't quickly recall any use for trailing '/' in gitignore patterns...	15:17
	hobbes006: '/*.foo' could actually be useful (only ignore foos in current directory, not subdirectories)...	15:18
hobbes006	ah. ok. makes sense.	15:18
	thank you Ilari.	15:19
	← Mr_Mirsal left	15:19
	→ bobesponja_ joined	15:20
	← MarcWeber left	15:21
	← bobesponja_ left	15:21
	→ bobesponja joined	15:22
	→ l0verb0y joined	15:24
	→ digitalhobbit joined	15:25
omygawshkenas	illari: Thanks so much. The dotest files are gone.	15:26
jengelh	dotest is part of rebase.	15:28
	→ fhobia joined	15:29
jast	trailing slash, in recent versions of git, makes the pattern apply only to directories	15:29
	(this has been in the docs for much longer than it was actually supported, by the way)	15:30
	→ freelock joined	15:35
	→ superdump joined	15:42
superdump	hello	15:42
	→ CodeOfficer joined	15:43
superdump	i tried to push all commits on all branches to a remote	15:44
	i.e. git-push --all <remote name>	15:44
	now, i ha	15:44
	oops	15:44
	i had three branches in the tree	15:44
jengelh	--all also includes tags, does not it	15:44
	blowery → blowery`afk	15:44
	→ ketralnis joined	15:45
superdump	master, branched from the head of master i had channel_elements and branched from the head of that i had sbr	15:45
	i made a change to master, rebased channel_elements to the new head of master, then rebased sbr to the new head of channel_elements	15:46
	but when i ran the push it gave me this:	15:46
	! [rejected] channel_elements -> channel_elements (non-fast forward)	15:46
	! [rejected] sbr -> sbr (non-fast forward)	15:46
jengelh	go figure	15:46
	where was that link	15:46
superdump	why is that?	15:46
jengelh	http://eagain.net/articles/git-for-computer-scientists/	15:47
	because the new branch you are pushing does not have the old one as a parent	15:48
superdump	ok, so is there a good solution to this?	15:49
	i.e. a way i can push the tree with all such changes to the server	15:51
	superdump reads the linked page	15:54
superdump	from what you just said it sounds like i can't push these kinds of rebases	15:55
	in which case, how is one supposed to manage them when pushing to elsewhere?	15:56
jengelh	you can force push, but people which are pulling from your repo will get duplicate commits	15:56
superdump	are you supposed to duplicate your rebase operations at 'elsewhere'	15:56
	ok	15:56
	superdump waits for a mandatory star wars joke	15:57
	← omygawshkenas left	15:57
superdump	pulling from the remote repo or my local one?	15:57
jengelh	remote	15:57
superdump	the duplicates being the old commits that were branched from the earlier commit on the master branch?	15:58
	← priidu left	15:58
jengelh	try it	15:58
	→ pflanze joined	15:59
superdump	i suppose it takes two to duplicate :)	15:59
	→ arun joined	15:59
jengelh	you can act as two	15:59
	→ sverrej joined	15:59
	← charon left	16:00
	→ VitaRara joined	16:01
	→ ijcd joined	16:02
	blowery`afk → blowery	16:02
Beket	pasky, is it ok for one person to have > 1 project in repo.or.cz ?	16:03
	← ijcd____ left	16:03
pasky	sure	16:03
superdump	jengelh: how would the duplicates be visible?	16:03
	← EmilMedve left	16:04
Beket	pasky, thanks mate	16:04
	← pombreda left	16:04
jengelh	gitk --all	16:05
	superdump: in every branch that branched from the original commits	16:05
	← naeu left	16:09
	← stouset left	16:09
	← munga left	16:12
	← nud left	16:12
superdump	i can live with that	16:12
	this tree won't be alive for all that long	16:12
	← Kbyte left	16:12
superdump	thanks for your help jengelh	16:12
	← cedricv left	16:13
	← redondos left	16:15
	← l0verb0y left	16:17
	→ p4tux joined	16:17
	→ ebel_ joined	16:17
	→ Alex_Gaynor joined	16:18
	→ timocratic joined	16:19
Alex_Gaynor	Hi, I have an existing git repo(hosted on github) that I would like to also make available via svn, so I'd like to host it on google code as well, how would I go about doing that	16:19
	← ebel left	16:20
	ebel_ → ebel	16:20
	blowery → blowery`afk	16:20
Ilari	Alex_Gaynor: I say that is not a good idea. SVN can't represent nonlinear history.	16:20
pasky	is that still true for 1.5?	16:20
jengelh	Does Google use 1.5? (Probably not)	16:21
pasky	besides, it's about as good idea as git cvspserver ;)	16:21
Alex_Gaynor	1,5 is still beta, isn't it?	16:21
Mikachu	pasky: i saw that the sort headers don't work on the fork pages on repo orcz, if you're bored and looking for somethnig to fix :)	16:21
pasky	I'm not but I might take a patch ;)	16:22
Ilari	Yes, it is still true in 1.5. It just tracks merges between linear branches (which is not the same as true nonlinear branches).	16:23
Alex_Gaynor	Any suggestions for how to make it available to SVN users then, the specific use case is they would like to use svn:externals on it	16:23
Ilari	Well, actually SVN doesn't even have first-class branches. :-)	16:23
avar	should [gc] auto = 0 \n packrefs = false \n autopacklimit = 0 in my .git/config be sufficient if I don't ever want packs created, I only want loose objects	16:24
Ilari	Alex_Gaynor: There is git2svn. But be warned that such SVN project is not allowed to have committers.	16:25
jengelh	only loose is inefficient over time	16:25
	→ alley_cat joined	16:25
avar	Yes, but this repository seems to work much better without packing stuff since there's no compression advantage	16:26
	→ ceez joined	16:26
jengelh	but reducing the number of files is an fs advantage nevertheless	16:26
	→ rubydiamond joined	16:26
avar	I'm not going to get enough files for this to be a problem	16:26
pasky	Ilari: what's the difference?	16:27
Ilari	pasky: The sidebranches are anonymous in Git. Good luck trying to represent that kind of thing in SVN...	16:28
pasky	/branch/auto-37F4B/ ;)	16:30
	→ rtomayko joined	16:30
Tv	have fun with octopus merges, too..	16:30
	← Alex_Gaynor left	16:30
Tv	avar: btw you don't need packrefs, that's something else and mostly good for you	16:31
	← loops left	16:31
Mikachu	do merges in svn 1.5 require you to keep the branches/foo branch around forever?	16:31
pasky	that's problem when converting to other vcses too... you just create virtual commits	16:31
Mikachu	(if you want to see the history of both sides of the merge)	16:31
pasky	(virtual isn't the right word, yeah)	16:31
jengelh	just leave it as git.	16:31
Tv	use that viewing-git-as-cvs thingie x.org wanted	16:32
avar	Tv: Thanks	16:33
	← fhobia left	16:33
avar	Tv: I tried using git-unpack-objects to get stuff out of the pack yesterday and into individual objects but it wasn't working and I couldn't be arsed to figure out why, so I'm just rebuilding it and telling it not to pack stuff again:)	16:34
Tv	avar: I'd guess you missed "Objects that already exist in the repository will not be unpacked from the pack-file."	16:35
	← superdump left	16:35
Ilari	avar: You have some special usecase like huge files in repo?	16:35
Tv	Ilari: exactly that	16:36
avar	I didn't, I just hoped it would magically work anyway and when it didn't I took the easy route:)	16:36
	→ redondos joined	16:36
avar	Ilari: Huge, uncompressable files:)	16:36
Ilari	avar: IIRC, there is patch that adds option to disable packing for files above configrable size...	16:37
	← kanru left	16:37
avar	That might be nice since I actually have some metadata files which are XML and packable, but overall I'd just like to turn it off.	16:37
	Their total size is currently 4MB out of 23 GB so I don't see a reason to worry about it :)	16:38
	queso → queso\|away	16:38
Ilari	avar: How often new huge object appears in that repository?	16:41
	→ mtodd__ joined	16:41
Tv	Ilari: more like all blobs are huge	16:42
avar	Every few days, whenever I take new pictures	16:42
Tv	oh he has xml	16:42
avar	http://git.nix.is/?p=avar/cam;a=tree;h=f84ddc0ee10814be72786f49e913ef842985fbb7;hb=f84ddc0ee10814be72786f49e913ef842985fbb7 <- this is an older version of the repo	16:42
	each .cr2 file has a corresponding .ufraw file	16:43
	elliottcable → ec	16:44
	← reel_ left	16:44
Ilari	avar: Have those megablobs somehow predictable filenames?	16:45
	blowery`afk → blowery	16:45
avar	they're named img_$increasing_int.cr2	16:46
	img_5000,cr2, img_5001.cr2, ...	16:46
	→ mrichman joined	16:46
mrichman	When I do a git push, how can I see it in the log?	16:47
	← ketralnis left	16:47
	→ VVSiz joined	16:47
VVSiz	evening, guys	16:47
jengelh	mrichman: what would you "see"?	16:47
Tv	mrichman: log shows commits	16:47
	→ charon joined	16:47
mrichman	Tv: okay...so how do I see what was pushed and when?	16:47
jengelh	there might be a server-side reflog	16:47
Tv	mrichman: what you may want is git log origin/master	16:47
	mrichman: but most of the time, people just don't track what was pushed when; only what's current	16:48
	mrichman: you can also use the hooks server-side to gather any info you want	16:48
Ilari	avar: Maybe take latest version of git-repack.sh from git.git (at least the latest that didn't use --revs) and edit it to ignore those .cr2 files?	16:48
mrichman	git log origin/master matches git log, so i'm going to trust it means what it says ;)	16:48
	← mtodd left	16:48
Tv	mrichman: means you have nothing extra in master to push	16:49
	← ijcd left	16:49
	→ thannoy joined	16:50
	→ branstrom joined	16:50
rtyler	any way to specify the SSH user name to use on git pull, etc	16:50
avar	Ilari: Will I need to do that? I.e. is git-repack run automatically?	16:51
	← bentob0x left	16:51
Ilari	avar: Well, if you disable autopacking, then you can run it manually.	16:51
	← wolf^ left	16:52
avar	I've disabled autopacking, my strategy is to do that and refrain from running git-repack :)	16:52
Ilari	avar: Well, it could be good idea to name your version something like 'git-repack-nocr2'...	16:52
Tv	rtyler: in the url	16:52
	→ mtodd joined	16:52
Ilari	avar: Then 'git repack-nocr2' should become available.	16:53
	→ xkenneth joined	16:53
xkenneth	anyone have a link to a good "best practices" guide for tagging versions with git?	16:53
	→ sverrej_ joined	16:54
	← lu_zero left	16:54
VVSiz	guys, is there a way to delete the braches in the remote repository, without having the file access to that repo?	16:55
	← sverrej left	16:55
VVSiz	I have my repo on GitHub and managed to put lots of wrong branches in there	16:55
Tv	VVSiz: git push origin :branchname	16:55
	← statim left	16:55
	→ pombreda joined	16:55
Tv	VVSiz: man git-push and looking for the word "delete" would have helped	16:55
Ilari	avar: Apparently latest usable version for editing is 62e27f273d66afa996cb7aee6cdb25fbedc053f6 (dated september 2006, but should still work)...	16:56
	s/is/is in/	16:56
VVSiz	Tv: ah, thanks! :) Who knew that in order to delete a branch I need to push! :)	16:56
Tv	VVSiz: it's the only over-the-network write operation you know of..	16:57
avar	Ilari: Do I need to? I'd rather have a setup that works on a stock git install and doesn't pack these files rarther than installing a custom version on every machine I use.	16:57
VVSiz	Tv: indeed, that simple enough. For some reason I was looking for some magic command to do that.	16:57
Ilari	avar: Doing fetch or push will pack those things for transfer...	16:58
	← mrichman left	16:58
avar	damn	16:58
	← mtodd__ left	16:59
Ilari	avar: The best could probably be to hack git to disable attempts to delta too large files...	16:59
	→ vineet joined	16:59
	→ wolf^ joined	17:00
Ilari	avar: For more advanced performance enhancement hacks: Restore capability to write new loose format (and make it default), and add capability to reuse those in packs.	17:01
	← bartocc left	17:02
	← JonathanT left	17:02
	→ tvachon joined	17:04
	→ k0001 joined	17:05
Ilari	avar: Also make it possible to reuse packed undeltified objects in loose objects (if you go that path)...	17:05
	← _zack left	17:05
	blowery → blowery`afk	17:06
rtyler	0/win8	17:06
	oopsies	17:06
	blowery`afk → blowery	17:09
	→ kanru joined	17:09
avar	Ilari: Since I have to avoid packing on fetch/push I could also hack around it by using rsync to sync repositories	17:10
	→ Eridius\|work joined	17:11
pasky	gitte: I'm not watching closely now, how was the Git.pm dependency hell made worse recently?	17:11
gitte	IO::String?	17:11
Ilari	avar: Doing the four modifications I outlined should be enough... Reusing data is quite efficient...	17:11
pasky	isn't that getting fixed now?	17:11
gitte	pasky: It seems that dependencies are piled onto Git.pm instead of removed. Ugly.	17:12
	pasky: Actually, it has been _introduced_ now.	17:12
	pasky: but I should learn to shut up, because I lack the time to do anything about it.	17:12
	gitte shuts up because she is lacking the time to do anything on Git	17:13
pasky	it's just the Git.pm's testcase	17:13
gitte	pasky: So? How does _that_ make it any better?	17:13
pasky	it won't break git :)	17:13
	as far as I understand the patch, IO::String usage is getting removed	17:13
	it was recently itnroduced, trouble quickly spotted and it goes away again - isn't that the situation?	17:14
Ilari	avar: Using rsync to sync repositories is really good way to spread possible corruption (should one of repositories get it for some reason)...	17:14
pasky	I assume it didn't even hit #next	17:14
	so I kinda fail to see a problem	17:14
Beket	Hi people. When pushing, I get this one: error: unable to create temporary sha1 filename ./objects/tmp_obj_gLdm82: Permission denied	17:15
	dwmw2 → dwmw2_gone	17:15
Eridius\|work	Beket: sounds like a problem server-side	17:15
pasky	you don't have write access to the remote repository	17:15
Ilari	Beket: Permission problems on the remote side.	17:15
	← zawir left	17:16
Beket	Thanks Ilari	17:16
pasky	pfft ;)	17:16
Ilari	Beket: That is, either user is not authorized to push, or repository permissions are screwed up.	17:17
Eridius\|work	Ilari: if user wasn't authorized to push I would expect a different error message	17:17
pasky	like?	17:17
Eridius\|work	I dunno, but Beket's message clearly indicates the repo is trying to accept his pushed objects, but is prevented from doing so with a permissions error	17:18
pasky	which is different how from "not being authorized to push"?	17:18
Eridius\|work	unless your definition of "user is not authorized" is "user doesn't have write permissions for the repo"	17:18
pasky	well, it is :)	17:18
	→ ketralnis joined	17:18
Ilari	Eridius\|work: In case user login is valid, but that user is not authorized to push to that repo, that is exactly the error you are going to get.	17:18
Eridius\|work	then don't say "not authorized or permissions problem", because by your definition those are the same thing	17:18
pasky	I don't ;)	17:19
Eridius\|work	fine, by Ilari's def	17:19
	→ loops joined	17:19
janm	hi! Is it ok to undo a commit in git and git-svn will do the right thing in the svn repo?	17:19
Eridius\|work	janm: define "undo"	17:19
	if you use git-revert to revert a commit, then git-svn should work fine	17:19
janm	Eridius\|work: I mean revert a commit	17:19
Ilari	'Screwed up' means things like objects/xx dirs having 755 mode on multi-user repo...	17:19
Eridius\|work	if you destroy a commit with git-reset, git-svn will just re-create it	17:20
janm	Eridius\|work: cool, thanks	17:20
	← loops left	17:25
	queso\|away → queso	17:27
	← snowblink left	17:28
	→ loops joined	17:29
	← ebel left	17:31
	→ ketralnis_ joined	17:35
	→ ijcd joined	17:35
	→ luxid joined	17:37
luxid	oggi è il download day! aiuta firefox ad entrare nel guinnes dei primati con il massimo numero di download in 24 ore! http://www.spreadfirefox.com/it/worldrecord entro le 20 di oggi	17:37
Ilari	luxid: What?	17:38
	← luxid left	17:38
Ilari	(some firefox related spamming, I guess)...	17:39
doener_	isn't there a "stupido" missing in that sentence?	17:39
Tv	yes let's all waste bandwidth and download even if we use debs!	17:40
	→ naeu joined	17:41
redondos	now we have geek flashmobs, too... :-\	17:43
	ketralnis_ → ketralnis_____	17:43
	← loops left	17:47
	→ VitaRara_ joined	17:48
	← VitaRara left	17:49
avar	Ilari: If I create a master repository at /tmp/git-test/master with a lot of small files and then another at /tmp/git-test/copy, do git init git fetch ../master I get packs automatically, and if I place 'exit 0' at the top of git-repack it gets fetched regardless and packed, what actually does the packing that I should modify?	17:51
	← ketralnis left	17:52
	ketralnis_____ → ketralnis	17:52
Ilari	avar: The packing is done by git-pack-objects, (virtually) invoked by 'git-upload-pack'.	17:52
	→ icwiener_ joined	17:53
	→ Usiu joined	17:53
Tv	and the pack is written to disk by git-receive-pack	17:53
	avar: but the data is always transferred in pack format, so you already paid for packing cost; might as well keep the file	17:54
Usiu	how to setup remote git repo ?	17:55
	← ijcd left	17:56
	→ ngng joined	18:00
	→ ShadeHawk joined	18:01
avar	creating an empty repo in 'tmp' then, git-unpack-objects < ../master/.git/objects/pack/$pack then rsync -av --progress --delete tmp/.git/objects/ master/.git/objects/ seems to work to eliminate packs after transfer	18:02
ShadeHawk	or you can just fiddle with configuration	18:03
	(the number of objects / size when git would not unpack transferred pack is configurable)	18:04
	byt why not simply repack?	18:04
Ilari	ShadeHawk: Huge blobs...	18:05
ShadeHawk	fetch.unpackLimit	18:06
	→ statim joined	18:06
ShadeHawk	pack.packSizeLimit (it will affect repacking only)	18:09
avar	yay, that works	18:09
	← icwiener left	18:10
avar	So I can set [fetch] unpackLimit = 999999 in every cloned so that the clones don't pack everyhting	18:10
ShadeHawk	(or transfer.unpackLimit -- this affect also push)	18:10
avar	apperently fetch.unpackLimit will be used if transfer.unpackLimit is not set	18:11
	transfer.unpackLimit	18:11
	When fetch.unpackLimit or receive.unpackLimit are not set, the value of this variable is used instead. The default value is 100.	18:11
statim	can someone help me figure out how to do this: developer A made a bunch of changes, sent a patch to developer B. developer B applied the patch, made additional changes to the files, then commited the changes. developer B pushed the branch to remote. now developer A wants to get back up to date, so i told him to git pull on master to get the latest, then switch back to his branch and git rebase, but there are all kinds of conflicts. is there a better strate	18:13
	gy?	18:13
Mikachu	of course there are conflicts if there's a different version of the patch applied	18:14
	← Usiu left	18:14
Mikachu	just use rebase --skip when you get to it	18:14
	queso → queso\|away	18:15
ShadeHawk	for patch-based workflow using "git fetch" and "git rebase" (or "git pull --rebase") is better solution than "git pull" which would use merge.	18:15
	→ schwarzes\|Geist joined	18:15
	→ Usiu joined	18:15
schwarzes\|Geist	is git good for downloading source and source control on php?	18:15
statim	ShadeHawk: im having hit git pull in master, and im having him never edit master directly	18:15
Mikachu	why?	18:15
	he already has origin/master	18:16
schwarzes\|Geist	i have the need to learn a solution for doing both of these and seem to be caught between svn and git	18:16
	looking for pointers	18:16
Eridius\|work	schwarzes\|Geist: git is good for version control for any language	18:16
statim	Mikachu: not saying its a good idea. i figured having him keep a clean copy that just follows the remote branch would be easier.	18:17
schwarzes\|Geist	any idea how i can go about implementing this or a good site to teach using it	18:17
Eridius\|work	statim: origin/master is exactly that	18:17
schwarzes\|Geist	i'm not familiar with git or svn so i would be learning it from scratch	18:17
Eridius\|work	schwarzes\|Geist: implementing what?	18:17
	ok, so go through git tutorials	18:17
schwarzes\|Geist	i have an existing php project and i need to pull it into git	18:17
Eridius\|work	unless you have a pressing need for a centralized version control system with strong per-user access controls and a linear history model, ignore svn	18:18
	ok, so read git tutorials.	18:18
schwarzes\|Geist	any particular site or just w/e i find in google?	18:18
Eridius\|work	git.or.cz should have a few pointers, but google works too	18:18
	I'm sure there are good tutorials that google will show that aren't linked on git.or.cz	18:19
Mikachu	and many bad	18:19
schwarzes\|Geist	Mikachu, yeah that's what i want to avoid	18:19
doener_	and lots and lots of outdated ones (which is sometimes worse than the plain bad ones)	18:19
schwarzes\|Geist	figure this would be as good a place to start as any	18:19
	so git.or.cz should have the most up to date ones then?	18:19
	queso\|away → queso	18:20
ShadeHawk	schwarzes\|Geist: documentation on git homepage (git.or.cz), especially "Git User's Manual"	18:21
schwarzes\|Geist	ok thanks	18:21
	i'll check it out	18:21
doener_	http://www.newartisans.com/blog_files/git.from.bottom.up.php	18:21
ShadeHawk	schwarzes\|Geist: links on GitDocumentation and GitLinks pages on git wiki (http://git.or.cz/gitwiki/)	18:21
doener_	that got some praise, I've never read it myself though	18:21
ShadeHawk	And for Subversion of course svnbook	18:22
	← schwarzes\|Geist left	18:22
ShadeHawk	if you (for the time being) plan on version control only for yourself, git is simpler to use: no need to setup server, just "git init; git add .; git commit -m 'Initial commit'"	18:23
	if not... Google Code and Sourceforge supports Subversion, for git hosting take a look at repo.or.cz, gitorious, GitHub; Savannah has also git support	18:24
mnemoc	did tla died?	18:25
	(OT)	18:25
	→ Sho_ joined	18:26
ShadeHawk	I think it did	18:26
	GNU Arch 2.0 (revc) never really started. From the ashes of GNU Arch (tla) there risen Bazaar (bzr), formerly Bazaar-NG	18:27
	but I don't think it has nowadays much in common with tla	18:27
	tla had, wrom what I've read, horrible user interface, used hostile non-portable filename conventions, and lacked 3-way merge	18:28
	mnemoc: but I think it would be best to ask on neighbour channel, #revctrl	18:28
mnemoc	the oddest thing is that sourcecontrol.net (tla hosting) has canonical copyright footer	18:28
	i didn't know that one :)	18:29
ShadeHawk	BTW Bazaar is now "GNU Bazaar"	18:29
mnemoc	oh	18:30
	copyright transfered to FSF?	18:30
ShadeHawk	more, Bazaar is now GNU project	18:31
	(from what I remember)	18:31
mnemoc	oh^3	18:31
Ramune	shade: tla filesnames: {arch}/=spork	18:31
	shadehawk: did funny things w/ shell scripts :)	18:31
	→ drizzd joined	18:31
mnemoc	:D	18:32
ShadeHawk	that's what I meant with "hostile non-portable filename conventions" (some of those characters used IIRC were forbidden on some non-POSIX filesystems)	18:32
mnemoc	,,foo was funny too	18:32
	→ david_koontz joined	18:32
Ramune	{arch}/++pristine-trees was also annoying	18:33
iabervon	For that matter, some contemporary versions of bash couldn't tab-complete those names.	18:33
ShadeHawk	nowadays Git, Mercurial, Bazaar are the DVCS considered, with a little side of Monotone	18:34
	mnemoc: as I said, try to ask about tla on #revctrl channel	18:34
Ramune	personally, i like monotone for one reason -- the logo :)	18:35
	→ mtodd__ joined	18:35
	← warthog9 left	18:35
	← timocratic left	18:35
	→ timocratic joined	18:35
mnemoc	ShadeHawk: yes, sorry for the off-topic... i was just wondering where tom lord and tom from libtomcrypt went	18:35
ShadeHawk	don't be. I just think that they might (might!) know more on #revctrl	18:37
	← punto left	18:37
	← Usiu left	18:37
mnemoc	:)	18:37
ShadeHawk	Hmmmm... Savannah now supports Mercurial (I have took a look to check if there are some stats what VCS offered by Savannah are used by projects most)	18:39
	→ robinr joined	18:39
ShadeHawk	if they are, I cannot find them...	18:41
	← mtodd left	18:42
	→ WALoeIII joined	18:45
	← drizzd_ left	18:46
iabervon	Hmm... I bet that Tom Lord isn't either of the first two hits on google for Tom Lord.	18:48
	→ ijcd joined	18:50
	← ijcd left	18:50
	→ ijcd joined	18:51
	→ zawir joined	18:52
	→ JonathanT joined	18:53
	→ kdubois joined	18:53
kdubois	i heard something like being able to collapse a bunch of commits into one commit, how can i do that?	18:53
jengelh	any overdose of heat should make anyone collapse	18:54
	→ warthog9 joined	18:57
drizzd	you could git reset --soft HEAD~5 and the commit, I guess.	18:57
Jc2k	or git rebase --interactive HEAD~5, not sure if thats the right syntax..	18:58
drizzd	and then there is git merge --squash, if that's what you heard	18:58
	Jc2k: how can you squash commits with rebase?	18:58
ShadeHawk	git merge --squash, git rebase --interactive and 'squash' line	18:58
Jc2k	drizzd: i used rebase to combine 2 commits before, you just put something like sqush in the $EDITOR that rebase --interactive pops up	18:59
	drizzd: try it - it even tells you in the help blurb	19:00
drizzd	ah, ok. didn't know that	19:00
	I guess that's the more visual method then	19:00
	← rubydiamond left	19:01
	→ lu_zero joined	19:03
	→ BinkyTheClown joined	19:03
BinkyTheClown	hi	19:04
	is it possible to git-svn only a branch of a svn repo?	19:04
	→ lorandi joined	19:04
mnemoc	BinkyTheClown: yes	19:05
vineet	yes binky, you can just specify the branch url as the repository url	19:05
BinkyTheClown	vineet: ok, thanks :)	19:05
vineet	since there's no distinction on the svn side; directories are directories	19:05
	→ bronson_ joined	19:05
BinkyTheClown	mnemoc: thanks	19:05
	git is the bussiness	19:06
	thiago_h1me → thiago_home	19:12
	→ yofel joined	19:13
	← mjc left	19:13
	→ ciaran joined	19:14
	→ harryv joined	19:16
	← harryv left	19:17
	← powerfox left	19:18
	blowery → blowery`afk	19:19
	← lorandi left	19:24
	blowery`afk → blowery	19:28
	→ stouset joined	19:30
	→ Eludias joined	19:31
	← yofel left	19:34
	→ l0verb0y joined	19:35
	← yann left	19:36
	→ johan-s joined	19:40
	→ yann__ joined	19:42
mcgrof	so I forgot to git-stash before a pull	19:43
	:T	19:43
	and I got a merge, and silly me, I didn't care about the stash so I just tried git-checkout -f then	19:43
	and now I get issues	19:44
	mcgrof@nesi:~/wireless-testing$ git-fsck	19:44
	dangling blob 7f0ce88cf21d247e8d88c3884d4a3fb033404845	19:44
	dangling blob e1c9deaa229d5a52ae9eaf42533f52a37405ddf7	19:44
jengelh	that's not an issue	19:44
mcgrof	oh, how do I move forward?	19:44
	mcgrof@nesi:~/wireless-testing$ git-pull	19:44
	You are in the middle of a conflicted merge.	19:44
jengelh	resolve it or abort it	19:45
	git pull is not the ultimate solution; sometimes you just want fetch	19:45
mcgrof	how do I do that, I didn't see an abort on the git-merge page	19:46
	yeah I should've just used git-fetch	19:46
	→ jerbear joined	19:46
jerbear	i ran a git-gc on a bare repo recently and it removed refs/heads/master and created a packed-refs. this is normal right?	19:47
	← deavid left	19:48
jerbear	anyone?	19:49
Arjen	Is anything broken?	19:49
jerbear	Arjen: the permissions on packed-refs was set to 660, which caused gitweb to show "no commits"	19:51
	← lea_w left	19:51
	→ bhuey__ joined	19:54
ShadeHawk	jerbear: yes, by default git-gc packs refs, although I think you can disable this via config	19:54
	jerbear: what umask do you use? git-gc lefs packed-refs with 644 in my clone of git.git repository	19:55
jerbear	ShadeHawk: this behavior is fine with me.. i'm just trying to figure out why the permissions were set to 660	19:55
	← statim left	19:56
jerbear	ShadeHawk: i haven't set a umask... it shows as 0022	19:56
ShadeHawk	hmmm... same by me	19:56
	← VitaRara_ left	19:57
ShadeHawk	do you have jerbear perhaps `core.sharedRepository` set to 0002 or something?	19:57
jerbear	ShadeHawk: i'll check	19:57
	blowery → blowery`afk	19:58
ShadeHawk	it looks like it got umask instead of permissions	19:58
	← l0verb0y left	19:58
jerbear	ShadeHawk: core.sharedrepository is set to 1	19:58
bhuey__	how do I get the diffs from a git branch ?	19:59
jerbear	ShadeHawk: some of the other repos do not have a value for that	19:59
bhuey__	I have the URI but I have little experience with git in general	19:59
jerbear	ShadeHawk: this is strange... my repos have varying settings, yet i have never changed them from the defaults	20:01
ShadeHawk	jerbear: I cannot help you further, unfortunately	20:01
jerbear	the only thing i can think of is that they were created at different times, with different git versions	20:01
	ShadeHawk: how can i find documentation on config settings?	20:02
	blowery`afk → blowery	20:02
jerbear	ShadeHawk: thanks for your help, i do feel closer to solving it	20:02
ShadeHawk	man git-config	20:02
	man git-clone (about sharedRepository)	20:02
jerbear	ShadeHawk: thanks	20:03
ShadeHawk	bhuey__: could you elaborate a little?	20:03
	bhuey__: if you want diff between two branches in the same repository, "git diff <branch1> <branch2>" would suffice	20:04
	← d0k left	20:04
ShadeHawk	bhuey__: if you want diff between two different repositories, take a look at GitFaq (or GitTips) on Git Wiki, http://git.or.cz/gitwiki/	20:05
	bhuey__: if the project has git web interface, be it gitweb or cgit (or other), you can request commitdiff between branches by handcrafting (modifying) URL	20:05
	HTH	20:05
	→ ebel joined	20:08
	→ nud joined	20:09
	← mislav left	20:17
	← hobbes006 left	20:18
	→ bremac joined	20:19
	→ chris2 joined	20:20
bhuey__	ShadeHawk: I just have a link to a URI where it contains the changes, it's its own tree	20:23
	← mtodd__ left	20:24
ShadeHawk	I don't quite understand	20:24
bhuey__	maybe neither do I	20:25
	→ l0verb0y joined	20:25
bhuey__	git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-kgdb.git	20:25
	is the tree	20:25
ShadeHawk	bhuey__: is it URI to remote repository, and you want to check if there were some new commits, or how tip differs?	20:25
bhuey__	I want to get a patch from that URI that I can apply to my tree that's non-git	20:25
ShadeHawk	bhuey__: you want diff (patch) between your non-git working tree (filesystem), and tip of given git repository?	20:26
bhuey__	something like that, but i know that this tree is against another branch	20:26
	← kdubois left	20:26
bhuey__	so I want a diff between that URL and the base branch	20:27
	URI	20:27
ShadeHawk	bhuey__: and you don't have Linux kernel git repository somewhere?	20:27
bhuey__	yes, I do	20:27
	I can cd into it and do various things like pulling it	20:28
	← Eridius\|work left	20:29
	→ Eridius\|work joined	20:29
ShadeHawk	bhuey__: do you want to get (a) diff between two different branches in mingo/linux-2.6-kgdb.git repository, or (b) diff between branch in mingo/linux-2.6-kgdb.git and some branch in _your_ git repository, (c) diff between mingo/linux-2.6-kgdb.git and filesystem?	20:30
bhuey__	ShadeHawk: (c)	20:30
	really that tree has been updated to mainline so there might be a name symbolic name for it	20:31
	← Eridius\|work left	20:32
	→ Eridius\|work_ joined	20:32
	Eridius\|work_ → Eridius\|work	20:33
ShadeHawk	bhuey__: simplest would be to either go to your git repo, use "git remote add linux-2.6-kgdb <URL>", then "git fetch --depth=1 linux-2.6-kgdb", then use "git diff <path to workdir> linux-2.6-kgdb/master" (--depth=1 might not work, if not try without it)	20:34
	← bremac left	20:34
ShadeHawk	bhuey__: or just "git clone --depth=1 <URL>" and "git diff <path to workdir> master"	20:34
	--depth=1 means only tip of history is interesting	20:35
	→ hendry joined	20:38
	queso → queso\|away	20:41
	← janm left	20:41
	→ tfogal joined	20:42
	queso\|away → queso	20:42
tfogal	Is there a good solution for git-svn development when `master' gets updated as you work?	20:43
	The situation I run into is that I branch off of master, start doing some work, including a few commits	20:43
	and then somebody else commits to the svn repo ... so I switch back to master, git-svn rebase, and then merge that into my branch	20:43
Eridius\|work	don't merge into master	20:44
	rebase your work branch on top of master	20:44
tfogal	you mean from master?	20:44
Eridius\|work	sure	20:44
tfogal	heh	20:44
Eridius\|work	git rebase master work	20:44
tfogal	which is equivalent to `git checkout work; git rebase master', yes?	20:45
	→ statim joined	20:45
Eridius\|work	yes	20:45
tfogal	that leaves me with strange conflicts; it seems it has forgotten commits I did a day or two ago.	20:46
Sput	git rebase is not a good idea with a branch that has been published though, right?	20:46
Eridius\|work	Sput: that's why he's rebasing his work branch	20:46
tfogal	Sput: unpublished branch, for better or worse	20:46
Eridius\|work	tfogal: what do you mean, forgotten commits?	20:46
Sput	yeah I mean in general... sometimes I push branches into the public repo for collaboration	20:46
tfogal	Eridius\|work: I have conflicts which are `<<< my updated code === /* nothing */ >>>>'	20:47
Eridius\|work	Sput: in general, don't rebase published branches	20:47
	tfogal: not sure what you mean	20:47
	oh, you mean the conflict had you added code	20:47
Sput	ok, so I thought, so the git rebase master work is for local, private working branches only	20:47
Eridius\|work	I assume the problem is it doesn't know if you added code, or if the other side removed code. which is odd, it should be a 3-way merge	20:47
	← Beket left	20:47
Eridius\|work	Sput: or for branches which you've explicitly advertised as subject to rebasing	20:47
	such as next/pu on git.git	20:47
tfogal	well, my merges onto master before I push to the svn repo is squashed	20:47
	err, are squashed, close enough..	20:48
Sput	Eridius: thanks	20:48
	← X-Blade left	20:48
Eridius\|work	tfogal: ok, sure. svn dcommit will commit the tree after the merge and follow the first parent	20:49
	so if that's the behavior you want, then that's fine	20:49
	← ajonat left	20:50
tfogal	It all seems to work out, AFAICT, but my `git merge --stat work', right before I merge from work, is quite scary. It makes it look like its re-applying patches which came in after I branched work.	20:50
	→ ajonat joined	20:51
tfogal	If I'm careful not to merge into work except at those points, maybe that will make it a moot issue.	20:51
	→ x-oweb joined	20:53
	→ cnk joined	20:53
	← freelock left	20:58
	→ _graham_ joined	20:59
	→ fafhrd_ joined	20:59
	→ jtbandes joined	21:00
jtbandes	Does anyone know if it's possible to use git with Xcode on Mac OS X like you can with Subversion?	21:00
	→ netoman joined	21:01
Eridius\|work	jtbandes: nope, sorry	21:01
jtbandes	:(	21:01
Eridius\|work	I don't think Xcode's SCM support is worth using anyhow	21:01
jtbandes	No?	21:01
	Why not?	21:03
Eridius\|work	it's just never seemed right	21:04
jtbandes	What do you mean?	21:04
	(I haven't really used it, but I was thinking it might be a good idea.. maybe not)	21:04
	← fafhrd left	21:05
Eridius\|work	eh, try it if you want, I didn't like it	21:05
	→ hansengel joined	21:06
hansengel	Hi, I just ran this command:	21:06
	→ AlexanderS joined	21:06
	← AlexanderS left	21:06
hansengel	`git svn clone -t http://www.something.com/svn/`, but I got this error:	21:06
	`Can't locate SVN/Core.pm in @INC`	21:06
	← l0verb0y left	21:07
	stouset → stouset\|away	21:08
bhuey__	ShadeHawk: thanks	21:09
Eridius\|work	hansengel: sounds like you're missing the subversion perl bindings	21:09
hansengel	hm.. would I need to build SVN from source to get these?	21:09
	→ peeja joined	21:09
hansengel	I installed SVN from my package manager	21:09
	and I don't see anything about Perl bindings..	21:10
Eridius\|work	your package manager probably has a package for it	21:10
	look for subversion bindings	21:10
hansengel	all I see here is `subversion` and `subversion-tools`, and subversion-tools only contains svn2cl	21:10
	→ mario_ joined	21:11
tfogal	hansengel: try grepping your pkg manager for `svn'	21:11
jtbandes	Is the git-osx-installer anything different than just compiling from source?	21:11
	← pygi left	21:12
	mario_ → pygi	21:12
hansengel	ah, found it.. it was libsvn-perl	21:12
	thanks, I'll try running the command again once it's installed	21:12
	→ felipec joined	21:13
ShadeHawk	hansengel: it is in subversion-perl package (Fedora Core) at my computer	21:14
hansengel	ShadeHawk: mine was libsvn-perl, Ubuntu 8.04	21:15
	stouset\|away → stouset	21:22
	→ proxie joined	21:24
	← mikearr left	21:25
	→ kewark joined	21:25
	→ mikearr joined	21:28
	← jerbear left	21:28
	← Ace_NoOne left	21:30
	→ AlexanderS joined	21:38
	← AlexanderS left	21:38
	→ AlexanderS joined	21:39
	← nud left	21:39
	← grasshopp left	21:40
	← spuk- left	21:41
	→ Mr_Mirsal joined	21:41
	→ spuk- joined	21:42
	← chris2 left	21:42
	→ freelock joined	21:43
	→ thannoy_ joined	21:43
	← krawek left	21:43
	← ph^ left	21:44
hendry	there is some code i want in an old revision. currently I `git diff b9626fdbfeec02d694d499a5b9c9f33699a2467a .xsession` and then copy the bits in the diff and remove the dashes etc.	21:45
	is there some merge tool invocation for what i'm trying to do?	21:46
	blowery → blowery`afk	21:46
	← pygi left	21:46
	← AlexanderS left	21:46
Ilari	hendry: One possibilty would be to dump that version to another file (git show b9626fdbfeec92d:.xsession >.xsession.another) and then move snippets over in editor.	21:48
	← tfogal left	21:49
hendry	Ilari: that sounds reasonable	21:51
	← xkenneth left	21:51
	← CIA-2 left	21:52
	← ebel left	21:52
jtbandes	If the project I'm putting on my repository is normally completely contained by a single folder, would it be common to simply put the contents of the folder on the repository, or the folder itself as well, so I'd have /path/to/working/copy/Project/stuff.etc?	21:53
	→ xkenneth joined	21:54
Ilari	jtbandes: I'd say puts the contents of folder in repository...	21:55
	→ CIA-2 joined	21:56
	← thannoy left	21:56
	← JonathanT left	21:57
	← mnemoc left	21:57
	← netoman left	22:00
	← gitte left	22:02
	→ pygi joined	22:03
	→ emlprime joined	22:03
	→ gitte joined	22:03
	← xkenneth left	22:08
	← stouset left	22:13
	← vbabiy left	22:16
	← \|Ryan52 left	22:18
	← emlprime left	22:19
	← fafhrd_ left	22:21
	← naeu left	22:21
	← _graham_ left	22:23
bhuey__	ShadeHawk: I did do the check out, but I'm having difficulty doing the diff	22:24
	← hansengel left	22:24
ShadeHawk	what are the results (error message)?	22:26
	→ troy_ joined	22:27
troy_	I've got a problem with git-cvsimport; I've been using it with no issues for many months; however I've got a problem with tags not being added/updated.	22:30
	(ie. tags from CVS)	22:30
	→ punto joined	22:31
	← radarek left	22:34
	← kewark left	22:36
	← BinkyTheClown left	22:36
bhuey__	ShadeHawk: mostly just nothing	22:36
	blank really, like doesn't respect the --depth=1 specification	22:36
	when I got a git-branch it says "master"	22:37
	got=do	22:37
	sorry for the lag, lots of screens here	22:37
	← jpeeler left	22:37
	→ krawek joined	22:40
	→ yofel joined	22:41
	← ceez left	22:41
	← Gitzilla left	22:42
	← Eludias left	22:42
	→ claymccoy joined	22:43
	→ BinkyTheClown joined	22:44
claymccoy	I just pushed some commits to a remote bare repo, and the files I added were the wrong permission (644 rather than 755). Not sure why that happened. But now I need to remove the last 3 commits I just pushed to the bare repo. I can't find instruction on how to do that.	22:44
	← wagle left	22:44
	← alley_cat left	22:45
Eridius\|work	claymccoy: do you want to keep the local changes, just destroy the commits themselves?	22:45
BinkyTheClown	I have run git-svn to clone a svn repo, is it possible to do 'svn diff' to send the output to a file?	22:45
bhuey__	ShadeHawk: any hints or a URL for the procedure ?	22:45
Eridius\|work	BinkyTheClown: use git diff	22:45
Mikachu	you can push any commit to a remote, git push -f theremote thebranch~3:thebranch	22:45
Eridius\|work	Mikachu: I would assume he wants to blow away the local commits too, so he can re-do them	22:46
Mikachu	you can have svn give you diffs without having a checkout :)	22:46
claymccoy	I already redid the local commits	22:46
Mikachu	then you can just push -f	22:46
claymccoy	made more commits that edited the permissions, then used "git rebase -i" and squashed them into the appropriate original commits	22:47
	→ stouset joined	22:47
Eridius\|work	claymccoy: so just push -f to overwrite the remote branch	22:47
	← charon left	22:50
claymccoy	can I just do -f or do I need to do the ~3 thing to remove the 3 unwanted commits on the bare repo?	22:50
Mikachu	no you can ignore me, just push -f	22:51
BinkyTheClown	when I was in the subversion bussiness I had to run 'svn diff --diff-cmd diff --extensions "-u -p" > patch.prop'; how can I do this with 'git diff'?	22:51
claymccoy	cool, thanks!	22:51
Mikachu	BinkyTheClown: git diff > patch	22:51
	i don't know exactly what "--extensions" does though	22:52
Eridius\|work	BinkyTheClown: you may want git diff --no-prefix > patch	22:52
	Mikachu: presumably it passes flags to diff	22:52
	but git already does a unified diff	22:52
Mikachu	ah	22:52
	right	22:52
Eridius\|work	and, er it already shows the C function too	22:52
BinkyTheClown	Mikachu: Eridius\|work: yes, they are flags to unix diff	22:52
Mikachu	i know what -up does of course :)	22:52
Eridius\|work	BinkyTheClown: git diff already uses those flags, apparently	22:52
Mikachu	i just didn't figure out the --extensions thing was to pass them to diff	22:53
BinkyTheClown	Eridius\|work: ok, i will try, thanks all	22:53
	← thiago_home left	22:54
	← bronson_ left	22:55
claymccoy	somtimes when I merge it puts the exact commits, and sometimes it rolls them up as a "Merge branch 'banch'". I prefer the first, but it seems random to me. Can anyone explain this?	22:58
	→ strangy joined	22:58
	→ schwank joined	22:58
	← robinr left	22:58
	← zawir left	22:59
claymccoy	actually I guess it always puts the rolled up commit in as a place holder in addition to the regular commits. What I notice is random, not what the computer does. :P	23:01
	schwank → schwanksta	23:02
bhuey__	how do I view the diff of a tree with a given a somekind of ancestor versus HEAD of the repo ?	23:03
xyzzy42	anyone know if there is a git mirror of buysbox's SVN?	23:03
Eridius\|work	bhuey__: git diff ancestor HEAD	23:03
bhuey__	It's completely unclear in the documentation how to do this	23:04
	Eridius\|work: ok, I tried that and it complains about the ancestor argument	23:04
Eridius\|work	what does it say?	23:05
bhuey__	how can I move a repo back in time and do that ?	23:05
	← schwanksta left	23:05
Eridius\|work	bhuey__: the command I gave should diff the ancestor vs the HEAD	23:05
	what error did you get?	23:05
	→ tenpaiyomi joined	23:05
bhuey__	Eridius\|work: complains about being an unknown branch or revision	23:05
tenpaiyomi	Somewhere in the process of doing all my work, I've managed to get 4 separate stashes	23:05
	Is there a way to merge them together?	23:05
Eridius\|work	sounds like ancestor doesn't exist. what did you give as ancestor?	23:05
tenpaiyomi	Or, merge them all into my working copy?	23:05
Eridius\|work	tenpaiyomi: you could try applying each one on top of each other. are you sure the stashes don't contain each other, though?	23:05
bhuey__	I don't know what the ancestor is which is one part of the probem	23:05
	problem	23:06
	← Ryback_ left	23:06
Eridius\|work	bhuey__: what are you giving as your ancestor?	23:06
bhuey__	I know how to do this in monotone but not git	23:06
tenpaiyomi	Eridius\|work: It looks like 3 of them are identical, so ideally what I would want to do is drop 2, and merge the 2 unique ones into my current working copy	23:06
bhuey__	I'm not sure, I don't know about the ancestors	23:06
	→ xkenneth joined	23:06
bhuey__	I'd like a list of of them but I don't know how to get them in git	23:06
tenpaiyomi	Eridius\|work: Or, I guess, stash my current work, and merge the 3 unique copies into a single stash, and apply that stash	23:06
Eridius\|work	bhuey__: ok, so when I gave you the git diff command, did you literally type "ancestor"?	23:06
bhuey__	yes	23:07
Eridius\|work	tenpaiyomi: you can destroy stashes with `git stash drop <stash>`	23:07
	you can apply them with `git stash apply`, etc	23:07
	bhuey__: ok, that's clearly wrong	23:07
	you want to diff "some ancestor" with HEAD. what ancestor do you want to diff?	23:07
	describe it	23:07
bhuey__	well, it would be clear if I knew git	23:07
	but I don't	23:07
Eridius\|work	in english	23:07
	just tell me what you're trying to do	23:07
	queso → queso\|away	23:07
	queso\|away → queso	23:07
tenpaiyomi	Eridius\|work: Is there some way to do a diff between 2 stashes to ensure there are no differences between them I am not seeing?	23:08
Eridius\|work	sure, use git diff ;)	23:08
tenpaiyomi	something like git stash diff stash@{0} stash@{1}?	23:08
bhuey__	Eridius\|work: take a branch and diff it against a base	23:08
Eridius\|work	tenpaiyomi: that should work fine	23:08
bhuey__	Eridius\|work: git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-kgdb.git	23:08
	that's the repo	23:08
Eridius\|work	bhuey__: so you have HEAD, and a branch. you want to find out how that branch differs from.. what. HEAD? the branch point?	23:08
	← DrNick left	23:08
bhuey__	I've pulled it and I don't know what else to do after that	23:08
	Eridius\|work: I'd like to get the branch point from the base that it's against	23:09
Eridius\|work	when you say the "base" what do you mena?	23:09
	*mean	23:09
tenpaiyomi	Eridius\|work: Not working. Figured it out to be git diff stash@{0} stash@{1} (or whatever you so desire)	23:09
	→ DrNick joined	23:09
bhuey__	so that I can get the patch and apply it into another tree that's non-git	23:09
Eridius\|work	tenpaiyomi: if you get the sha1's for the stash it'd be easier	23:09
	bhuey__: you're missing the question here	23:09
	do you want differences from HEAD, or do you want differences from the commit where the branch was created?	23:10
	stouset → stouset\|away	23:10
bhuey__	Eridius\|work: I have the tree, I have the HEAD, But this should be a branch against some base revision. I would like the diff between the branch point and the head	23:10
Eridius\|work	ok, git diff HEAD...branch (note the 3 periods)	23:10
bhuey__	Eridius\|work: commits from when the branch was created	23:10
Eridius\|work	where branch is the name of the branhc	23:11
	*branhc	23:11
	*branch	23:11
bhuey__	literally branch ?	23:11
Eridius\|work	no, the name of the branch	23:11
bhuey__	I don't know the name of the branch, how can I get a list ?	23:11
Eridius\|work	`git branch`	23:11
	actually, you probably want `git branch -r`, since you probably only have master as a local branch	23:11
bhuey__	I get HEAD and master only	23:12
tenpaiyomi	Eridius\|work: Don't suppose you know how to show the sha1 of a stash, do you? If I just do a git stash list it truncates the sha1 after 7 digits	23:12
Eridius\|work	bhuey__: right, use `git branch -r`	23:12
	tenpaiyomi: that 7 digits should be fine	23:12
	assuming that 7-digit prefix is unique, git will find the full sha1 automatically	23:12
bhuey__	Eridius\|work: that's the only two things that come out from that comment	23:12
	command	23:12
Eridius\|work	bhuey__: even with -r?	23:13
tenpaiyomi	Eridius\|work: Well, the first 3 stashes all have the same 7 digit prefix, but 2 of the stashes are the same, and one is different	23:13
bhuey__	Eridius\|work: yes	23:13
Eridius\|work	tenpaiyomi: that sounds highly improbably	23:13
	*improbable	23:13
	bhuey__: ok, what version of git are you using?	23:13
bhuey__	the latest stuff from Ubuntu	23:13
Eridius\|work	which is what?	23:14
	type `git --version`	23:14
bhuey__	1.5.4.3	23:14
tenpaiyomi	Eridius\|work: I can do a git diff stash@{0} stash@{1} and it shows no differences, but a git diff stash@{0} stash@{2} brings up differences, yet the list has them with the same 7 digit prefix	23:14
bhuey__	Eridius\|work: I appreciate the help btw	23:14
Eridius\|work	tenpaiyomi: check `git rev-parse stash@{0}` and `git rev-parse stash@{2}`	23:14
	← xkenneth left	23:15
bhuey__	Eridius\|work: maybe the tree was created improperly	23:15
	→ xkenneth_ joined	23:15
	← peeja left	23:16
Eridius\|work	bhuey__: `git branch -r` should show more than just "master", even on 1.5.4.3	23:16
bhuey__	Eridius\|work: git log is short as well	23:16
	Eridius\|work: no, it just shows those two lines	23:16
	origin/HEAD origin/master	23:17
Eridius\|work	ok, you didn't say the origin/ part. that makes a difference	23:17
	in other words... there's no branches	23:17
bhuey__	oh	23:17
Eridius\|work	why did you expect there to be a branch?	23:17
bhuey__	because it should be pulled against the latest kernel sources	23:17
	→ Gitzilla joined	23:17
	Eridius\|work goes to clone the repo	23:18
Eridius\|work	how large is this repo, btw?	23:18
bhuey__	40M<	23:18
Eridius\|work	oh	23:19
bhuey__	it comes down relatively quickly if you have a cable modem	23:19
Eridius\|work	ok, I'm not going to clone that here	23:19
	← stouset\|away left	23:19
Eridius\|work	ok, what does `git remote show origin` says?	23:19
	*say	23:19
bhuey__	I use to do this on monotone all of the time relatively easily, but I don't know git at all other than the pull command	23:19
Eridius\|work	bhuey__: you might want to start going through git tutorials	23:19
bhuey__	Eridius\|work: nothing really special, I don't see anything new regarded as tags	23:20
Eridius\|work	bhuey__: it has the right URL?	23:21
	→ EiNZTEiN joined	23:21
bhuey__	yes	23:22
Eridius\|work	ok, try `git remote update`	23:22
	→ wagle joined	23:22
	→ cliffstah joined	23:22
bhuey__	done, nothing much there to print out	23:23
	← felipe left	23:23
	← xkenneth_ left	23:23
Eridius\|work	ok, I just checked using the HTTP transport, it really does appear that there are no branches	23:23
	plenty of tags... no branches	23:24
bhuey__	yeah, improperly created tree ?	23:24
Eridius\|work	no	23:24
	your clone is fine	23:24
bhuey__	or do we need to use the tags instead ?	23:24
Eridius\|work	if there's a problem, it's on the remote side	23:24
	what did you expect to find as a branch?	23:24
bhuey__	something much more substantial than the git log output	23:25
Eridius\|work	what's wrong with git log?	23:25
	what does `git rev-list HEAD \| wc -l` say?	23:25
bhuey__	2	23:25
Eridius\|work	....	23:25
	try this: `git rev-list origin/master \| wc -l`	23:26
bhuey__	2 again	23:26
	← ft left	23:26
Eridius\|work	...	23:26
	→ Toad__ joined	23:26
Eridius\|work	ok, what about `git rev-list v2.6.25 \| wc -l` ?	23:27
	← troy_ left	23:27
bhuey__	unknown revision or path	23:27
tenpaiyomi	Eridius\|work: Do you know if it's possible to apply all stashes, and not just one?	23:28
Eridius\|work	tenpaiyomi: I would just try applying each stash in turn	23:29
	bhuey__: odd, you should have picked up the tags	23:29
bhuey__	recheck it out ?	23:29
tenpaiyomi	Eridius\|work: Well, the problem with that is that after the first stash is applied, any stash application past that gives a message about cannot apply over a dirty index	23:29
Eridius\|work	bhuey__: maybe you should try re-cloning, that would definitely fix any local issues	23:29
	queso → queso\|away	23:29
bhuey__	I did try to use --depth=1 when I checked it out	23:30
Eridius\|work	bhuey__: .....	23:30
	that's the source of your problem with git log	23:30
Mikachu	vital information revealed late	23:30
	→ stouset joined	23:30
Eridius\|work	you told git "only grab the latest commit from history"	23:30
	no wonder your `git log` only has 2 commits	23:30
bhuey__	ok	23:30
Mikachu	you can't do anything with a shallow clone except look at the actual files	23:30
Eridius\|work	tenpaiyomi: you could just stash, commit, stash, commit, etc., then squash the commits	23:30
bhuey__	well, trying to follow the instructions of another person on the channel	23:30
Mikachu	bhuey__: do you currently have a local linux-2.6.git clone somewhere?	23:31
bhuey__	yes	23:31
Mikachu	you want to use the --reference switch for clone then	23:31
bhuey__	what does that do ?	23:31
	and how do I go about using it ?	23:31
	→ mithro joined	23:31
Mikachu	it is in fact documented in the manual page	23:31
	← eternaleye_ left	23:32
Eridius\|work	bhuey__: you really should read git tutorials instead of stumbling around blindly	23:32
	→ xkenneth joined	23:32
	→ eternaleye_ joined	23:32
bhuey__	Eridius\|work: the tutorial is rather unclear and basic	23:33
	unless I'm looking at the wrong tutorial	23:33
Eridius\|work	there's a plethora of tutorials	23:33
Pieter	yay	23:33
Eridius\|work	in the last month or two, it seems the Hip New Thing™ has been to write git tutorials	23:33
Pieter	/topic \| s/1.5.5.4/1.5.6/	23:33
	ChanServ set mode: +o	23:34
	→ anzalaya joined	23:34
Eridius\|work	woot!	23:34
	Mikachu changed the topic to: 1.5.6 \| Homepage: 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 $ID?: http://tinyurl.com/yqpgv9 \| Need to paste something?: http://git.pastebin.com \| http://www.gitcasts.com	23:34
Eridius\|work	when is git.or.cz going to be updated so it can actually list releases with no patchlevel?	23:34
	← redondos left	23:34
	→ RobotDeathSquad joined	23:34
Pieter	itcan't do that?	23:35
Mikachu	heh, good way to minimize impact of brown paper bag bugs	23:35
	Mikachu set mode: -oo	23:35
Eridius\|work	Pieter: IIRC, it didn't list 1.5.5 until 1.5.5.1, and that's the explanation I was given	23:35
Pieter	heh	23:35
	← digitalhobbit left	23:35
Pieter	that should be a trivial patch :)	23:36
Mikachu	i'm sure pasky will be happy to accept it too	23:36
	← bdiego left	23:36
Eridius\|work	where's the relevant code?	23:36
	← icwiener_ left	23:37
Pieter	http://repo.or.cz/w/git-homepage.git?a=blob;f=update.sh;h=ad23f989f1bfd94c1dd5ba0c46062e1065b25a47;hb=92214ba9ea9ebda9022969d1197726917d562ccc	23:37
	I think	23:37
Mikachu	Eridius\|work: last section on the frontpage links to git-homepage.git	23:37
Pieter	time=$(git cat-file tag $ver \| sed -n 's/^tagger.*> //p' \| cut -d ' ' -f 1)	23:38
	doesn't look very bad	23:38
	→ ft joined	23:38
Mikachu	cut -d \ would be one character less! :)	23:38
Eridius\|work	oh that's right, it's not a problem with the script	23:38
	it's because the maint branch doesn't get updated	23:38
	until a patchlevel	23:39
	← strangy left	23:39
Pieter	huh	23:39
Mikachu	ah right	23:39
	maint isn't forwarded to 1.5.6 yet	23:39
bhuey__	Eridius\|work: yeah, tons of lines now	23:39
Pieter	hmm	23:39
Eridius\|work	bhuey__: great	23:39
	Mikachu: why isn't maint forwarded until a patchlevel release?	23:40
Pieter	Vienna:git pieter$ git describe	23:40
	v1.5.6-158-g3b7f0af	23:40
	whoo	23:40
Mikachu	why does it look in the git repo to find the latest tarball though?	23:40
	← eikonos left	23:40
Eridius\|work	Pieter: do you have a local patch series, because you have 11 more patches than me	23:40
Mikachu	or does it in fact generate the tarball in that script? i didn't read it very carefully :)	23:40
Pieter	Eridius\|work: yes	23:41
Mikachu	Eridius\|work: don't we all? :)	23:41
Eridius\|work	bah	23:41
BinkyTheClown	chao a todosssssssssssssz	23:41
Pieter	some basic things like making builtin-mailinfo grok Mail.app mails	23:41
BinkyTheClown	bye all	23:41
	Eridius\|work would be working on a patch series to allow interactive editing of diffs in add, except I only got the idea because I heard someone else is doing that :/	23:41
bhuey__	Eridius\|work: what to do with it next is another matter	23:41
	← Toad left	23:41
	← BinkyTheClown left	23:41
Eridius\|work	Pieter: you mean you want to point it at Mail.app's cache?	23:41
	bhuey__: it's called read tutorials	23:41
	→ eikonos joined	23:42
Pieter	Eridius\|work: yeah, ~/Library/Mail	23:42
	→ carrus85 joined	23:42
Pieter	Eridius\|work: now I can drag patches to a folder and have an alias that applies them all	23:42
Eridius\|work	nice	23:42
bhuey__	Eridius\|work: nice, git log	23:42
Pieter	don't use that as much though, as I now have a script that applies selected messages	23:42
Eridius\|work	Pieter: you could use osascript to ask Mail to hand back the source to the selected messages... ;)	23:42
	Toad__ → Toad	23:42
Pieter	hah, you're too late! :P	23:42
Mikachu	Eridius\|work: that is not only doing, it's done	23:42
Eridius\|work	:'(	23:42
	Mikachu has used git add -e a couple of times	23:43
Eridius\|work	Mikachu: really? where? last I heard it wasn't even in pu	23:43
Pieter	I love git add -e	23:43
Mikachu	Eridius\|work: git://repo.or.cz/git/dscho.git	23:43
Eridius\|work	how come it's not in mainline next/pu yet?	23:43
Mikachu	it's among the top 10 commits in my-next i think	23:43
	i think there are two commits you need	23:43
	Allow git-apply to ignore the hunk headers (AKA recountdiff)	23:43
	git-add: introduce --edit (to edit the diff vs. the index)	23:43
Pieter	yeah	23:43
	are those the newest? that fix the tmpfile thing?	23:44
bhuey__	Eridius\|work: just you helping me was sufficient to generate the patch :)	23:44
	thanks	23:44
Eridius\|work	bhuey__: great	23:44
Mikachu	Pieter, Eridius\|work: yeah the commit titles might have changed since i grabbed them	23:44
	so by yeah i mean no	23:45
Pieter	I wonder if I can do sha1^ in rebase-i	23:45
Mikachu	in where?	23:45
Eridius\|work	hrm, dscho/my-next isn't up-to-date wrt. origin/next	23:45
Mikachu	Eridius\|work: he is only human :)	23:45
Pieter	hah	23:46
	that's nice	23:46
	← p4tux left	23:46
Eridius\|work	heh, but it sounds like all I need is dscho/my-next^^ and dscho/my-next^	23:46
Mikachu	yeah, i just cherry-picked those to a local topic branch	23:46
Pieter	I just did a "pick 7eb692a918b180d1fc40438ef83c5536fcc34242^\npick 7eb692a918b180d1fc40438ef83c5536fcc34242"	23:46
Mikachu	Pieter: oh you cheat and add commits in rebase -i instead of giving on the command line?	23:47
	s/you/you can/	23:47
Pieter	yeah	23:47
	well	23:47
	I did a rebase -i in my editor	23:47
	and replaced my old versions with the new ones	23:47
	← xkenneth left	23:47
Mikachu	ah	23:47
	← ShadeHawk left	23:47
Mikachu	is there some easy way to do interdiff =(git show sha1) =(git show sha2) ?	23:48
Pieter	http://ss.frim.nl/==806	23:48
	:)	23:48
	Mikachu doesn't like tilted lines	23:48
Mikachu	or at least not lines that change direction more than once	23:49
felipec	Mikachu: if you use zsh yes :)	23:49
Mikachu	felipec: hm?	23:49
	oh you need <() in bash	23:49
Pieter	Mikachu: that might be a nice script to make	23:49
Eridius\|work	huh, Mikachu, I never knew about <(). useful!	23:50
felipec	Mikachu: oh, that works on bash too! didn't know that	23:50
	← claymccoy left	23:51
Mikachu	cat <(echo hello) and echo hello > >(cat) :)	23:52
	those are unidirectional pipes sort of, so if a command wants to lseek it will fall over	23:52
Eridius\|work	they're fifos	23:53
Mikachu	that's what you call them, thanks	23:53
	but they're not actual fifos as you would get from mkfifo	23:53
	$ echo <(:)	23:53
	/dev/fd/63	23:53
	← stouset left	23:53
Eridius\|work	ok, bash says it will either use FIFOs or /dev/fd devices	23:53
Mikachu	in zsh, with =() you get a temporary file in /tmp that gets removed after the command line finishes	23:53
	ah okay	23:54
Eridius\|work	I don't know how it determines which, but the idea is it's not supposed to matter	23:54
Pieter	Mikachu: hah, I get the same fd	23:54
Mikachu	maybe it starts at 64	23:54
	3*	23:54
Eridius\|work	wow I get the same fd too	23:54
Pieter	:D	23:54
xyzzy42	can I clone an existing git-svn created repo, then update it with git-svn?	23:54
Mikachu	try opening 63 fds in bash and see what happens	23:54
	Eridius\|work: irssi	23:54
Pieter	Vienna:~ pieter$ echo <(:) <(:)	23:54
	/dev/fd/63 /dev/fd/62	23:54
	going down!	23:55
Eridius\|work	Mikachu: I'm trying to find out OS	23:55
xyzzy42	the git-svn man page has some commands that are supposed to do this, but it doesn't work.	23:55
Mikachu	Eridius\|work: linux ~2.6.19	23:55
Eridius\|work	ok	23:55
	→ stouset joined	23:55
Mikachu	to save you the nmap	23:55
Eridius\|work	xyzzy42: you can, with a few steps	23:55
Pieter	Mikachu: /dev/fd/4 /dev/fd/64 /dev/fd/65 /de	23:55
Mikachu	Pieter: damn i just finished that too :)	23:56
	(repeat 63 echo '<(:)')\|xargs	23:56
xyzzy42	I followed the example in the git-svn manpage, but it fails on the last command, "git-svn rebase"	23:56
Pieter	what does it say?	23:57
	← stouset left	23:57
	→ stouset joined	23:58
xyzzy42	I think it might be because there is no refs/heads/master	23:58
	→ nexus10 joined	23:59

Logs Search ←Prev date Next date→

Channels Documentation