IRCloggy #git 2008-05-30

Provider of IRC logs since 2005.

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

	2008-05-30
	← jeckel left	00:06
	← devogon left	00:06
	→ jeckel joined	00:09
	→ gmcinnes joined	00:10
gmcinnes	hi all. How do I add an empty directory to a git repo?	00:11
Eridius\|work	you need to put something in there. I recommend .gitignore	00:11
	git doesn't track dirs	00:11
	although, in theory it could just use an empty tree. I wonder why it doesn't support that	00:11
	?	00:11
	pasky stays silent ;)	00:11
gmcinnes	right. I thought that was the case. That's what I've done.	00:12
	thx.	00:12
vmiklos	it's in the faq	00:15
	→ eikonos joined	00:15
vmiklos	http://git.or.cz/gitwiki/GitFaq#head-1fbd4a018d45259c197b169e87dafce2a3c6b5f9	00:16
	Eridius\|work wonders how much work it would be to teach the index how to track empty dirs	00:17
	→ z1o joined	00:19
	→ norflork joined	00:20
pasky	it would be trivial	00:21
	(I believe)	00:21
	→ tim_abell_ joined	00:21
	tim_abell_ → tim_abell	00:21
	← tim_abell left	00:22
	→ krh joined	00:22
pasky	I think the FAQ is misleading - I believe the proposals (by others and me as well, historically) were getting rather close to patch form but the powers that be did not like the idea :)	00:22
	→ tim_abell joined	00:22
pasky	maybe there are caveats I don't realize	00:22
	← schacon_ left	00:23
	→ bdiego joined	00:23
	← Rameteric left	00:24
	← sbeyer left	00:24
	← aroben left	00:25
	← jamessharpe left	00:26
	← hipertracker left	00:28
norflork	How do I modify my default commit message so that it always starts with a specific custom string at the very top for a single repository? I think I need to modify prepare-commit-msg in my .git/hooks directory, but the best I've been able to do is get my custom message printed at the very bottom of my commit message which isn't what I want. (I don't know shell very well.)	00:35
	→ avdi_ joined	00:35
	→ mrix_ joined	00:35
mrix_	hey! Need help on my question: 1) assume i have a repo and 2) i generated few files there (many) 3) i changed my mind and want to get clean repo. How do i do that except manually deleting all those generated files i don't need anymore? Thanks!	00:37
Mikachu	norflork: does it apply after you write the message in the editor?	00:37
	norflork: otherwise the obvious answer is to write the message under the generated content :P	00:38
	→ context joined	00:38
z1o	mrix_: depends on if the repo is cloned or not.	00:38
mrix_	z1o: it's a local repo	00:38
doener_	norflork: you can create a template for the commit message and set commit.template in the config	00:38
norflork	I see the message appear in the editor itself. But it is under the generated content. I can move it manually, but it gets annoying for when git auto-creates merge messages.	00:38
	← context left	00:39
Pieter	hmm	00:39
	git log HEAD~50000 is kinda slow	00:39
Mikachu	norflork: ah, can you pastebin what you have?	00:39
z1o	mrix_: git checkout master should work	00:39
	→ Aughey joined	00:39
tim_abell	I'd love to know what you guys think of my git / flash drive sharing howto: http://timwise.blogspot.com/2008/05/sharing-work-between-computers-with-usb.html	00:39
norflork	just a sec. I think I broke things again. :P	00:39
mrix_	z1o: those useless files do not get removed...weird	00:40
	→ rubydiam_ joined	00:40
z1o	mrix_: or clone the repo	00:40
tim_abell	ps, Hi all.	00:40
z1o	mrix_: cd ../ git clone $repo ./newrepo	00:41
	← rubbish left	00:42
z1o	mrix_: best thing to do though in the future is to branch. then you can checkout a branch ie master to remove changes	00:42
mrix_	z1o: well what i did instead is $ rm -rf./*; git checkout master ./	00:46
	→ brosner joined	00:46
mrix_	z1o: don't like that though	00:46
z1o	mrix_: right so use branching	00:46
	mrix_: git branch testing . git checkout testing	00:46
	← datafirm left	00:46
mrix_	z1o: yup, thanks!	00:47
z1o	mrix_: then you can just git checkout master .. poof changes gone. checkout testing poof changes back	00:47
	rubydiam_ → rubydiamond__	00:47
norflork	Mikachu: sorry, I really broke it now and don't remember what I had. I think I modified the commented out perl statement though.	00:48
	doener_, can you tell me more about the template?	00:48
doener_	uhm, there isn't much to say. You write it, you add the config entry, and then git commit will use it as its template when you need to enter a commit message	00:49
	→ spearce joined	00:50
	← jlh left	00:50
	→ jlh_ joined	00:50
	→ datafirm joined	00:51
	← avdi left	00:51
	jlh_ → jlh	00:51
norflork	so in a .gitconfig file, I would have a section called [commit] and a line that looks like template = /path/to/template ?	00:52
	← bdiego left	00:52
Mikachu	you can use the git-config command	00:54
norflork	Is it possible to have a gitconfig per repository instead of globally? I want this custom commit message only for a specific repository, not all of mine.	00:55
Mikachu	yes	00:56
	that's the default mode for git-config	00:56
	→ cout joined	00:56
Mikachu	while with --global it goes to ~/.gitconfig	00:56
	→ Leefmc joined	00:56
Eridius\|work	norflork: .git/config is the default config. note that this does not get cloned	00:56
tim_abell	I promise it's not shameless self promotion, rather a labour of love. I only discovered git recently and having had it improve my life in a small way I wanted to share my new found knowledge for the benefit of the greater good. Apologies for rambling, it's late.	00:57
Mikachu	that might also be a reason for not many looking at it :)	00:57
	tim_abell shoots self in foot as usual	00:58
norflork	okay, good information here. So .git/config will not be cloned. Do the hooks like prepare-commit-msg get cloned?	00:58
Mikachu	norflork: no	00:58
Eridius\|work	no	00:58
Mikachu	norflork: echo rm -rf ~ >> .git/hooks/prepare-commit-msg, hey norflork clone my repo it's so great	00:58
tim_abell	I mentioned in my article removing the contents of .git/hook on fat drives due to lack of execute flag. Am I asking for trouble there or is that considered ok?	00:59
Mikachu	tim_abell: that's a good idea i should think	00:59
tim_abell	phew, thanks	01:00
Mikachu	tim_abell: they're definitely not needed for normal operation, especially as they're all disabled by default	01:00
tim_abell	I can't wait to impose git on work. Er I mean persuade them it's better for us than svn.	01:00
Eridius\|work	good luck	01:01
	tim_abell: there's a config flag you can set that indicates exec flags don't exist	01:01
tim_abell	thanks, shouldn't be to bad. they'll just need time to adjust (/etc/init.d/svnserve stop. oops!)	01:02
	oh cool, thanks Eridius, I'll look it up	01:02
ToxicFrog	Hmm. That's odd.	01:02
	→ soto joined	01:03
Eridius\|work	tim_abell: it's core.fileMode	01:03
ToxicFrog	I cloned my repo onto windows using cygwin git, and it lost most, but not all, of the execute flags.	01:03
Eridius\|work	git config core.fileMode false	01:03
	when false, it means ignore exec bit differences	01:03
soto	How do I amend a commit message?	01:03
Eridius\|work	soto: if it's the latest commit, git commit --amend	01:03
	that'll re-create the commit using the current index, and let you edit the message	01:03
soto	Eridius\|work: But if I only want to change the message I have nothing to commit	01:03
Mikachu	Eridius\|work: i don't know if that applies to hooks	01:04
Eridius\|work	soto: wrong	01:04
	you have to create a new commit object	01:04
	← rwanderley left	01:04
Eridius\|work	Mikachu: heh, you're right, I don't know either	01:04
soto	Eridius\|work: Do you mind elaborating? How do I create a new commit object?	01:04
ToxicFrog	soto: the message is an intrinsic part of the commit object.	01:04
	→ avdi joined	01:04
Mikachu	soto: just run git commit --amend	01:04
ToxicFrog	There is no way to edit a commit once it's created; the closest you can come is to create a new commit and replace the old one with it.	01:05
Eridius\|work	soto: are you trying to edit the last commit object? if so, use git commit --amend like I said before	01:05
ToxicFrog	Which is what --amend does.	01:05
soto	Eridius\|work: Alright, I guess I thought that wasn't working when I tried it. Thanks	01:05
	← avdi left	01:06
	→ doener joined	01:06
	→ avdi joined	01:07
	← datafirm left	01:08
tim_abell	thanks Eridius, do you know if there's anything more enlightening docs for fileMode than this? http://www.kernel.org/pub/software/scm/git/docs/git-config.html#EXAMPLES	01:08
	grammer failing. i should go to bed	01:08
Eridius\|work	tim_abell: I think you need to ask someone who actually uses FAT ;)	01:08
tim_abell	poor souls	01:09
	→ Cherife joined	01:09
tim_abell	thanks for your help, I'll investigate further another day. goodnight all	01:09
	← shoe left	01:10
	← tim_abell left	01:11
	← avdi left	01:11
norflork	Thanks everybody! I think I got the template working.	01:12
	Since the config doesn't clone, are there any recommended practices to remembering to copy this with a new clone?	01:14
	→ tzafrir joined	01:14
Eridius\|work	you could make a global alias that calls clone and prints a reminder, and use that instead of clone	01:14
tzafrir	any idea how I get git-svn to authenticate with a certificate?	01:14
Eridius\|work	tzafrir: it should just work the same way svn does	01:14
Mikachu	you can put some stuff in a branch	01:14
	or a tag	01:14
tzafrir	Eridius, sadly, it doesn't	01:15
	Specifically, on the same machine subversion 1.4 reads configuration from files in ~/.subversion	01:15
Eridius\|work	yes, so should git-svn	01:16
	my git-svn reads ~/.subversion no problem	01:16
tzafrir	And I can't persuade git-svn to use the certificate listed there	01:16
Mikachu	i think someone else had the same issue (didn't see a solution)	01:16
tzafrir	Eridius, client-side certificate?	01:16
Mikachu	if it wasn't you	01:16
	the only thing i know works is accepting certificates from the server	01:16
	← soto left	01:17
Eridius\|work	tzafrir: I'm not entirely sure if any svn server I've used with git-svn uses certificates, but I know it reads my non-cert auth info just fine, and I looked at the code once and it looked like it should read everything	01:17
	← doener_ left	01:18
tzafrir	it asks me for a password	01:18
	Over a year ago I made an ugly patch to some perl module to get svk working	01:19
	But I don't have it here	01:19
	← goibhniu left	01:19
Mikachu	the perl module comes with svn, but it's possible git-svn still has to call some routine in it to make it work, or it coudl be a bug in svns perl module	01:19
	i say this knowing next to nothing about both :)	01:20
	← avdi_ left	01:20
	→ ben_h joined	01:24
norflork	Can somebody set me straight on git push? (newbie question). If I create a new repo 'orig' and then clone it 'copy'. I make some changes to 'copy' and commit them. Then from 'copy', I git push to 'orig'. What am I supposed to do in 'orig' to see the new changes from 'copy'?	01:25
Mikachu	norflork: you are not supposed to push to a non-bare repo	01:26
	http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	01:26
	Mikachu hopes that is the right link, did /lastlog faq	01:26
norflork	I think that link was correct.	01:27
	← david_koontz left	01:29
	ToxicFrog winds windows's entrails out on a stick	01:29
tzafrir	the password prompt is in the function ssl_client_cert_pw	01:29
	strace shows that it did read my certificate file	01:30
	But no attempt made to decrypt it?	01:30
	→ brosner_ joined	01:30
	← stillLotR left	01:30
ToxicFrog	'git checkout: we do not like "World: The Universes of Legend" as a branch name'	01:32
	Sad.	01:32
	← krh left	01:32
Mikachu	the colon is a big nono	01:33
ToxicFrog	Oh, right. This is windows.	01:33
	and it's reserved by git.	01:33
Mikachu	well, think of push and fetch	01:33
	i don't know about the spaces either	01:33
ToxicFrog	It doesn't like the spaces either.	01:33
	That I can't see any reason for.	01:34
	→ shoe joined	01:34
ToxicFrog	Hmm. Man page says it needs to pass git-check-ref-format	01:34
	← brosner left	01:34
	brosner_ → brosner	01:35
Mikachu	i'm sure you can use ∶ and instead	01:35
ToxicFrog	Ok. No nonprintings. Nothing in [[ ~^:?*]. No "..". Cannot end in /. Cannot contain "/.".	01:35
	Yes, but those are a pain to type.	01:36
Mikachu	sure :)	01:36
ToxicFrog	I could use "World/The.Universes.of.Legend"	01:37
Mikachu	note that you can't have two branches "foo" and "foo/bar" at the same time	01:37
ToxicFrog	Yeah.	01:37
	That won't be an issue.	01:37
	Every branch except master starts with World/ or Fortress/	01:37
Mikachu	are these your save files? :)	01:38
ToxicFrog	Yep	01:38
Mikachu	heh	01:38
ToxicFrog	Well, my complete active games	01:38
	World/ branches contain generated worlds with no running games; Fortress/ branches, active games.	01:38
	When a fortress finishes, I merge it back into the world it spawned from.	01:38
Mikachu	i see	01:39
ToxicFrog	It's not a true merge, sadly - if I have multiple concurrent fortresses, only one of them will succeed.	01:39
Mikachu	i think that's a new usage of git :)	01:39
ToxicFrog	But it makes keeping track of things much easier.	01:39
	(also makes trying out mods and the like easier - just create a new branch. No need to taff around with backup game installs!)	01:39
jwasson	OOC, why allow the limitation on branch names? I understand the technical reason with regards to the implementation, but why an implementation that creates those limitations?	01:42
Eridius\|work	jwasson: because it would be more work to remove the limitations, and it hardly seems worht it	01:42
ToxicFrog	jwasson: the man page explains it, actually	01:43
	← Ryback_ left	01:45
jwasson	Hrmm....I guess some of that makes sense.	01:46
norflork	Commit-template (again): Drat, it looks like the auto-generated merge message that looks like: "Merge branch 'master' of /path/of/repo/" is not respecting my template format. Are there any other templates I need to override?	01:48
	← warthog9 left	01:50
	→ ajonat joined	01:50
	← ajonat left	01:51
	← mrix_ left	01:51
	→ ajonat joined	01:52
	→ bdiego joined	01:55
	← mithro left	01:59
	→ destruct_ joined	02:03
norflork	It looks like commits during a merge-conflict also bypass my template.	02:09
	→ jamesish joined	02:11
	← eikonos left	02:12
	← destruct left	02:13
jamesish	Hi folks. We have an interesting situation. We had a git repository, call it foo. And one of our developers, Bob, pulled from it and started doing large, good work. In the meantime, we moved to repository bar and other developers committed and updated, while Bob was working away solidly. Now he needs to transition his work to the new repository.	02:14
	I know git can handle this kind of situation, but I don't have enough experience to know how to deal with it.	02:15
jwasson	Bar wasn't cloned from foo?	02:16
jamesish	Bar wasn't cloned from foo.	02:17
jwasson	If it wasn't, have Bob generate patches from his repository. Find the closest commit in bar and checkout a new branch there. Apply the patches to that new branch and then merge with master.	02:17
jamesish	Okay.	02:18
jwasson	I'm not that knowledgable either, so there might be a better way, but that's how I'd probably do it.	02:18
jamesish	I just checked with a guy who knows better than me; bar was actually cloned from foo	02:18
	→ angasule_ joined	02:19
jwasson	In that case bob should be able to add bar as a remote (maybe replacing foo) and just behave normally.	02:19
	That is, pull from bar, merge to master (or whatever branch), then push to bar.	02:19
jamesish	Wow. That'll make things _very_ simple. And he has a backup of the repo, so if it goes wrong we'll be fine.	02:20
	This could be an ass saver for us, thanks jwasson.	02:20
jwasson	No problem.	02:20
	Have bob create a tag of his current HEAD then he can rewind to there easily.	02:21
	(if things go wrong)	02:21
	→ FunkeeMonk joined	02:21
jamesish	Good idea.	02:22
	→ zawir joined	02:22
	← Leefmc left	02:22
	→ eternaleye_ joined	02:24
	→ jerbear joined	02:25
	← tflsh left	02:30
	← norflork left	02:32
	← angasule left	02:34
	destruct_ → destruct	02:35
	→ igorgue joined	02:38
	← timocratic left	02:44
	← ekidd left	02:45
	→ Leefmc joined	02:47
	→ timocratic joined	02:48
	← timocratic left	02:49
	← rubydiamond__ left	02:50
	→ lirit joined	02:51
	→ ekidd joined	02:52
	← igorgue left	03:01
	← Leefmc left	03:02
	→ donspaulding joined	03:04
donspaulding	is there some way to have `git svn clone` recognize svn:externals properties?	03:05
mugwump	it recognizes them, see the unhandled.log	03:08
	it doesn't currently do anything sensible with them though	03:08
	← bdiego left	03:09
donspaulding	hrm, could I `git svn clone` each of the externals in the appropriate subdirs of the main repo?	03:09
mugwump	you could, and add them as submodules	03:10
	→ rsharma joined	03:10
donspaulding	would git fetch then update from all the appropriate svn repos?	03:10
mugwump	no	03:10
	rsharma → horos22	03:10
mugwump	nobody's worked on that yet	03:10
donspaulding	ok	03:10
mugwump	also, there is no integration with external SCMs in 'git fetch'	03:11
donspaulding	right, I meant git svn fetch	03:11
	← ekidd left	03:11
mugwump	however, I suspect that git-svn could be enhanced to allow specification of paths to externals	03:11
	← Cherife left	03:12
horos22	hey all.. a question about repositories - is there a way to 'look before I leap',	03:12
donspaulding	well it's an itch right in the middle of my back....I'm unable to scratch it myself, so I'll just have to sit and wait for someone double-jointed to get the same itch	03:12
horos22	ie: see the changes that are going to be made to my local repository before actually pulling the switch with git pull?	03:12
mugwump	donspaulding: ok. I just have to keep trying to entice users into implementing the features they request ;)	03:13
	→ Cherife joined	03:13
albertito	horos22: yes, if you are going to pull from a remote, you can do "git remote update <name>" and then use git log or gitk to view the remote branch you're interested in pulling	03:13
	horos22: if it's not a remote, you can fetch and then use FETCH_HEAD as the reference for what you've just fetched	03:14
horos22	git remote update actually pull the changes over?	03:14
	sorry .. does git remote ...	03:14
	suppose I want to see only one or two files?	03:15
albertito	horos22: it will get them, and update the remote branches, but it will not touch anything that is yours	03:15
horos22	so if you do a 'git pull' afterwards, it's a no-op?	03:15
albertito	horos22: so it gets the changes, but pretty much nothing else. It will not merge or alter your branches in any way	03:15
	horos22: it's not a no-op, because the git pull will try to fetch the changes (you will already have them, this will be a no-op) and then merge them in your branch	03:16
horos22	got it..	03:16
	now suppose you've got lots of changes, and a slow connection, and you want to limit what you are seeing only to a couple files..	03:17
albertito	horos22: the last part is the "op". Another way to think about it is to think of pull pretty much as fetch + merge (but read the docs for the details). git remote update does the fetch part, and git merge does (obviously) the merge part	03:17
	horos22: there is no way you can see what you would pull without getting the changes	03:18
horos22	so it's an all or nothing deal.. suppose I wanted to pull only a couple of the files..	03:18
albertito	horos22: you can get them and then decide to do nothing, but in order to see and decide you have to get them	03:18
	→ avdi joined	03:19
albertito	horos22: it's pretty much an all or nothing deal, yes	03:19
horos22	ok.. thanks..	03:19
	probably should switch to smaller branches then..	03:19
albertito	horos22: another way, more rudimentary, if the project has some gitweb interface, is to browse it manually and get the files you're interested in	03:20
	horos22: it really depends on your workflow	03:20
	→ warthog9 joined	03:21
horos22	right.. but right now it's taking a given update about an hour to pull over to my local repo..	03:21
	it would be nice if there was a hack.. err workaround for this very issue.	03:21
albertito	horos22: git is very clever about what to send over the network, but you can't get just the changes that affect a given file because of how git works	03:22
	← lirit left	03:23
horos22	yeah.. that's what I figured..	03:24
	one more thing.. suppose you make some edits and you want to revert them, before you've done a commit (ie: there were some mistakes)	03:26
	any way to do that easily?	03:26
	← jerbear left	03:27
horos22	hmm.. I take it that you can't do this easily? I would think that would be extremely simple to do..	03:32
	← kukks left	03:36
horos22	I take it that I need to check my mistakes in, and then do git reset and/or git revert? that seems a little silly	03:36
mugwump	you can force it	03:38
	in general no git commands will let you throw away any changed file without requiring you to force it first	03:38
horos22	so you are saying 'git reset --soft HEAD --force'?	03:38
	or sorry 'git reset --force HEAD'	03:39
mugwump	why not read the man page a bit	03:40
	--hard will overwrite the files	03:40
	also, you can use git checkout to revert a single file	03:41
	← donspaulding left	03:41
horos22	yeah.. I saw that in the docs..	03:42
	← xyzzy_ left	03:42
horos22	but I tend to approach such things with great caution..	03:42
	so I thought I'd ask just in case..	03:42
	→ ckoehler joined	03:45
	← markkalderon left	03:45
horos22	thanks..	03:45
	← avdi left	03:49
	← stargaming left	03:50
	→ xyzzy42 joined	03:54
	→ theperitus joined	03:55
lea_1	Terminology question: What are the arguments to git-rev-parse called? Iow, how do I call arguments like 98badfa721, HEAD, ...?	03:57
z1o	anyone know way when I push changes. the changes appear as deleted	04:00
	← ckoehler left	04:00
doener	lea_1: git-rev-parse(1) calls it "extended SHA1 syntax"	04:01
lea_1	doener: Thanks!	04:02
	That's kinda unwieldy... :)	04:02
	I think I'll go with extsha1 or so. :)	04:03
doener	lea_1: note that it also supports ranges. I usually come up with some term "on demand"	04:03
	sha1, rev range, ref name, symbolic ref, ... And for the ^, ^!, ~, @, : and so on, I don't even have a term I use (exactly because I find the "extended SHA1 syntax" unwieldy and often too generic)	04:04
	← jamesish left	04:04
	→ mithro joined	04:09
	← madewokherd left	04:09
	→ tjafk2 joined	04:10
	← daperitus left	04:11
	← FunkeeMonk left	04:13
	→ a-priori joined	04:18
	← a-priori left	04:18
	→ jerbear joined	04:24
doener	z1o: http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	04:26
	← tjafk1 left	04:26
	→ spearce` joined	04:31
	← spearce left	04:32
	spearce` → spearce	04:32
	← jwasson left	04:32
	→ jwasson joined	04:43
z1o	doener: great thanks, makes more sense now	04:47
	← spearce left	04:51
	← cedricv left	04:53
	← lea_1 left	04:53
	→ alterscape joined	04:54
alterscape	Is this an appropriate venue to ask a gitosis setup question? Hope so, 'cause here goes: I'm setting up gitosis on OS X 10.4.11. Got it set up to the point where I can clone the gitosis-admin repository, change gitosis.conf and the contents of keydir, and commit/push the changes. I've verified that the changes are being pushed by cloning another copy of gitosis-admin. Server-side, I've created a test_project group, given access to @gitosis-	04:58
doener	alterscape: message got truncated it seems, right after "access to @gitosis-"	04:58
alterscape	@gitosis-admin, and committed/pushed those changes. Locally, I've created a local repository called test_project and registered the remote, but when I attempt to push, I get "ERROR:gitosis.serve.main:Repository read access denied." Thoughts?	04:59
	doener: Thanks for the heads-up, there's the rest	04:59
doener	no problem. No clue about gitosis either though ;-)	05:00
	← jerbear left	05:02
	→ ggeecko_ joined	05:04
	→ cedricv joined	05:05
	← ggeecko_ left	05:07
	→ ggeecko_ joined	05:07
	← ggeecko_ left	05:07
	→ ggeecko joined	05:07
	→ Leefmc joined	05:08
	← zawir left	05:08
	← tvachon left	05:09
	→ panagos joined	05:10
panagos	anyone know how to delete projects from repo.git.cz?	05:10
	*** repo.or.cz	05:10
	→ rwanderley joined	05:17
	→ schacon joined	05:17
	→ imyousuf joined	05:21
	← schacon left	05:22
	→ devogon joined	05:24
	← rwanderley left	05:26
	← Tv left	05:27
	→ d2dchat joined	05:31
	→ datafirm joined	05:35
cehteh	panagos: by sending patches to add this functionality to pasky :)	05:36
	← jwasson left	05:36
	← Leefmc left	05:39
panagos	right... :\| (:P)	05:41
	→ tflsh joined	05:41
panagos	pasky: can i delete a repo i created in repo.or.cz, by sending you an email? :) i didn't do any work with it...	05:46
	← datafirm left	05:48
	← rmh3093 left	05:49
	→ Tv joined	05:50
	← bobesponja left	05:51
	← sgrimm left	05:51
	→ sgrimm joined	05:52
panagos	i want to start a private project with some friends, and we want to use git. but we don't want "outsiders" to be able to pull from us, so what is the simplest way to do that? so that pulling would require authentication	05:59
	jeckel → away	06:00
	→ kixx joined	06:00
	← krawek left	06:01
	→ priidu joined	06:02
Auris-	the answer probably has something to do with ssh.	06:03
alterscape	panagos, I'd look into gitosis. Granted, I'm having trouble setting it up, but it seems to be a nice option	06:03
	you could also set up individual user accounts with git-shell as the shell and then, yes, have them ssh in	06:03
	← chrisb left	06:06
	← d2dchat left	06:07
	← kixx left	06:09
panagos	alterscape: i noticed that gitosis doesn't have any good documentation under debian...	06:11
alterscape	there's very little documentation in general, but if you look at the scie.nti.st blog entry, there's a link to some more debian info	06:11
	← theperitus left	06:13
	← ggeecko left	06:13
panagos	there aren't even man pages! which is mandated by the debian guidelines, if i'm not wrong	06:14
	away → jeckel	06:14
	← spuk- left	06:15
	→ spuk- joined	06:16
	→ _zack joined	06:17
	→ gittish joined	06:18
	→ deavid joined	06:18
	← gmcinnes left	06:23
	← Eridius\|work left	06:27
comp	panagos: another option would be usage of Github private repo ...	06:31
	→ daperitus joined	06:32
	→ BabelO joined	06:34
	← johan-s left	06:34
panagos	yeah, but it needs $$$$ :) plus, not as distributed :D	06:34
tflsh	gitosis isn't that hard to set up, go for that. once you've got it set up, it's very rewarding to add new projects and permissions as needed so easily	06:36
	dwmw2_gone → dwmw2_HEL	06:36
Ilari	panagos: The only real pitfalls in setting up Gitosis are various configuration errors in SSHD configuration, editing '~git/.ssh/authorized_keys' by hand or using existing account (as opposed to creating new one).	06:42
	→ TuxCM- joined	06:43
	← ajonat left	06:44
Arrowmaster	i never ran into any of those setting up gitosis	06:44
	→ bentob0x joined	06:45
	→ drizzd joined	06:45
Ilari	Arrowmaster: Yea, because they tend all to be screwups themselves...	06:46
	Arrowmaster: But I have seen descriptions of all those three on this channel at various points (breaking the stuff completely).	06:46
Arrowmaster	only issue i ran into was that the debian package wants to use the username:group gitosis:gitosis when i wanted to use git:git	06:46
panagos	btw, when you people write code, do you use some personal repo, while the others pull from you the new changes, or do you use a central repo, with each one pushing to their branch or whatever?	06:46
	i guess i'll be using the default debian configuration for sshd...	06:47
Ilari	panagos: I have used personal (bare) repo where I push changes and others can pull from.	06:48
panagos	Ilari: for that to work, i guess you have to have your repo online 24/7?	06:49
Ilari	panagos: The SSHD configuration issues where things like some configuration option locking gitosis account from remote logins and that...	06:49
panagos	aha, i see	06:50
	→ lea_w joined	06:50
Ilari	panagos: Well, considering the environment I used it in back then, keeping it online virtually 24/7 was no problem (shared filesystem)... :-)	06:51
panagos	i have the feeling that true distributed development asks for everyone to have a personal repo online all the time, so that others can pull from them (or maybe use schedules?)	06:53
	← saintdev left	06:53
	→ reel_ joined	06:55
	→ EiNZTEiN joined	06:56
	→ ph^ joined	06:58
mugwump	panagos: sure, or connect to repo.or.cz, push to your mob branch	06:59
	← drizzd_ left	07:00
	← mithro left	07:01
panagos	if i want to have a shered repo, how can i only allow each user to push to a specific branch?	07:04
mugwump	with an 'update' hook	07:07
	if that hook returns an error code, the push is denied	07:07
	also, standard error gets relayed to the user	07:07
	→ namenlos joined	07:08
Ilari	panagos: Except that standard trick of identifying by UID does not work with Gitosis (since all users share Unix user account)... :-/	07:08
mugwump	gosh, that's a bit of a bug	07:09
panagos	Ilari: so how do you do that with gitosis?	07:09
mugwump	do they have seperate pwents?	07:09
Ilari	mugwump: They have seperate keypair entries.	07:09
	← pygi left	07:10
	← csc` left	07:10
panagos	mugwump: pwents?	07:10
mugwump	right. Well, I guess they could set an environment variable. I'm thinking of rigging the perl repository to accept push via git:// if the signature is OK	07:10
Ilari	mugwump: And each keypair entry contains forced command with different key id (Gitosis uses that piece to authenticate).	07:10
mugwump	panagos: password entry. see pwent(5)	07:10
	er	07:11
Ilari	mugwump: Except that git:// can't authenticate pulls at all.	07:11
	mugwump: And this was about setting private repo.	07:11
mugwump	getpwent(3) :)	07:11
	← shoe left	07:11
mugwump	Ilari: it can't authenticate at the protocol level, no. but you can reject the push	07:11
	so, I'll only let signed tags get pushed	07:11
	and certain branches can only be pushed by "core" team	07:12
panagos	mugwump: no such man page on my box... (also an apt search dones't find one)	07:12
mugwump	It just means an entry in /etc/passwd; you can have seperate users with the same UID	07:13
	of course, they have the same access	07:13
	to unix, anyway	07:13
panagos	i think that would best be solved if we made git aware of account issues	07:14
	← thiago_home left	07:14
mugwump	it is aware :)	07:16
	it has core.sharedRepository, for instance	07:17
	← panagos left	07:18
Arrowmaster	all that does is set the file premissions	07:18
	→ panagos joined	07:18
mugwump	sure	07:18
	you want more? use a hook	07:18
	→ wuxia joined	07:19
wuxia	is there a short cut for git checkout?	07:19
	neither git c	07:19
	nor git co work	07:19
mugwump	wuxia: git config alias.c checkout	07:19
	add --global to that to make it effective for all your repos on that machine'	07:19
	→ mcella joined	07:20
wuxia	cool, thanks :-)	07:23
Ilari	panagos: I think that making Gitosis export the authentication identity so it is usable by hooks would be one-line addition to Gitosis source...	07:23
	← brosner left	07:25
	→ nud joined	07:26
	→ saintdev joined	07:31
	→ ph^_ joined	07:33
	← cedricv left	07:35
	← gittish left	07:40
	→ hdu_hh joined	07:50
	→ sbeyer joined	07:53
	← DrNick left	07:53
	→ DrNick joined	07:53
	← gef left	07:54
	→ ajonat joined	07:56
	← ajonat left	07:57
	→ ajonat joined	07:57
	← wuxia left	07:58
	→ fhobia joined	07:59
	→ csc` joined	08:00
	→ koke joined	08:01
	← sverrej left	08:01
	← ph^ left	08:01
	→ holizz joined	08:01
	→ Surfer8 joined	08:02
	Surfer8 Visit http://www.FakeMagazineCover.com (upload pic make mag) - http://www.SillyWebcam.com (play with webcam online) - http://www.Is-A-Jerk.com (insulter/anon email) - http://www.ComedySearchEngine.com (fun) - http://www.BodySwitcher.com (put your face on funny body) - http://www.MedChecker.com (health) - http://www.Canuckster.com (Canada eh) - http://www.Nerdful.com (geeks)	08:02
	← Surfer8 left	08:02
	← horos22 left	08:02
	→ Matsimitsu joined	08:05
	→ mcella_ joined	08:06
	→ rgs joined	08:06
	← mcella left	08:06
	→ LotR joined	08:08
	← lolage0 left	08:09
	→ sverrej joined	08:15
	← Ted left	08:18
	→ ebel joined	08:19
	← _zack left	08:21
	→ _zack joined	08:21
	→ Ted joined	08:21
	→ gitte joined	08:22
	→ d2dchat joined	08:22
	→ robinr joined	08:25
	← nicholas_ left	08:29
etnt	a friend of mine try to clone from my gitosis repo, but get this: ERROR:gitosis.serve.main:Need SSH_ORIGINAL_COMMAND in environment.	08:33
	any ideas? I have several other friends that it works perfect for	08:34
	← _zack left	08:34
etnt	that the key is a ssh-dss key shouldn't matter, or ?	08:34
	← jns left	08:35
	ph^_ → ph^	08:38
	→ redondos joined	08:39
	→ hein2 joined	08:39
	← Sho_ left	08:39
	hein2 → Sho_	08:39
	pilif\|away → pilif	08:40
	← TuxCM- left	08:41
	← sverrej left	08:44
	→ sverrej joined	08:46
	← fhobia left	08:47
	→ Catfish joined	08:52
	← Catfish_ left	08:52
ben_h	hi all	08:53
	anyone seen an issue where a hook is run even though it's not executable?	08:53
	i'm getting a "fatal: exec [ ... ]/.git/hooks/prepare-commit-msg failed."	08:57
	but,	08:57
	-rw-r--r-- 1 ben ben 1.2K 28 May 22:27 .git/hooks/prepare-commit-msg	08:57
cehteh	outch	08:59
ben_h	yeah	08:59
	sense: this error makes none	08:59
	← Cherife left	08:59
cehteh	which git version .. and what filesystem?	09:00
ben_h	1.5.5.1, filesystem is HFS+ on OS X 10.5.3	09:00
	→ Beket joined	09:00
ben_h	the git repo is a submodule, i thought maybe it was going on the parent's hook but that's rw-r--r-- as well.	09:01
	ok wow	09:03
	i moved the file out of the directory, committing works	09:03
	moved it back in, still works	09:03
	:\|	09:03
	the prepare-commit-msg file that is	09:03
cehteh	maybe rather a HFS+ / OSX issue	09:04
	does that happen for other hooks as well?	09:04
	and is that vanilla git or patched for osx somehow?	09:04
ben_h	it's vanilla, i compiled it myself the other day	09:04
	it's never happened before. and now i moved the file away and back the error has disappeared	09:04
	very strange indeed.	09:05
	← jeckel left	09:05
Ilari	etnt: 'grep -v "no-pty" ~git/.ssh/authorized_keys' doesn't yield anything except empty or commented out (starting with '#') lines?	09:07
	→ ^Beket\| joined	09:09
	→ ksvladimir joined	09:10
ksvladimir	Is there any way to forget changes that are already in the git index ?	09:12
	In other words: to revert an effect of git add ?	09:13
Ilari	ksvladimir: 'git reset -- <file>'?	09:13
pasky	panagos: if you wrote the name of the repo in your message right away, your chances would be much higher :P	09:13
	panagos: but you can still try	09:13
cehteh	git reset --cached	09:13
pasky	to send me the mail	09:13
cehteh	(git gui comes handy too)	09:14
pasky	(tig too)	09:14
ksvladimir	Ilari: cehteh: thanks alot, it's exactly what I wanted	09:14
cehteh	pasky: you can try the xor approach and remove all repos and then add back the ones where people complain	09:14
	← manveru left	09:15
pasky	cehteh: heh :)	09:15
	then i couldn't brag about hosting so many repositories I suspect	09:15
	← frsk left	09:17
	→ frsk joined	09:19
	→ _zack joined	09:19
	← Beket left	09:22
	→ charon joined	09:22
	→ duncanmv joined	09:22
duncanmv	hi guys, I host a lot of repos, and git-daemon-export-ok dissapears from time to time, and I have to touch it again, any idea where to start looking?	09:23
pasky	fsck ;))	09:24
Pieter	heh	09:24
	some commits in the samba repository have an extra space after the email address	09:25
	which makes git-fast-export give borked output	09:25
	Ilari: with checkpoints at 500, it continued until it was out of memory :)	09:26
	← Ted left	09:34
	→ Ted joined	09:36
	← d2dchat left	09:45
	→ markkalderon joined	09:47
Mikachu	duncanmv: you could use the --export-all flag? :)	09:48
	← ^Beket\| left	09:59
	← Ademan left	10:00
	→ Ademan joined	10:01
	← namenlos left	10:01
	← ben_h left	10:02
	→ kasimon joined	10:04
	← pdherbemont left	10:06
kasimon	Hi! What's a good way to separate changes to a git tree since into different branches?	10:06
	I've worked on some features in parallel on my devel branch and would like to put some of them into the production (which is push/pulled on another server)	10:07
rgs	create a branch and cherry pick commits to there ?	10:08
kasimon	That sound's reasonable. I'll have a try.	10:10
	← AlexanderS left	10:14
	← gitte left	10:18
	→ Xyrio joined	10:19
	← duncanmv left	10:21
	→ Yuuhi joined	10:22
	← Sho_ left	10:28
	← Xyri0 left	10:30
	← Matsimitsu left	10:32
	→ Matsimitsu joined	10:35
	→ d0k joined	10:37
	→ AlexanderS joined	10:42
	→ Sho_ joined	10:44
	→ Tgmessi_ joined	10:44
	→ ForsakenSoul joined	10:50
ForsakenSoul	hello I'm trying to commit some file changes ... to git hub but first when I type commit I get the .swp file I should edit ... how do I exit it ? and how do I push it to my branch in github	10:51
	← Matsimitsu left	10:53
	← jlh left	10:57
	→ jlh_ joined	10:57
	← Tgmessi_ left	10:59
	→ thommey_ joined	11:05
	← thommey left	11:08
	thommey_ → thommey	11:09
	→ WildPikachu joined	11:13
WildPikachu	if i create a shared local git repository (--shared=group), which directories does the group need access to?	11:14
	*write access	11:14
Mikachu	i would guess .git itself, objects, and refs, maybe logs	11:14
	ie, almost all	11:15
Arrowmaster	if your using --shared wont it automaticly be a bare repo anyway?	11:16
	← rgs left	11:18
WildPikachu	hrmmm "cannot clone empty repository"	11:19
	→ kukks joined	11:19
WildPikachu	darn	11:19
	how would i check something into a bare repository?	11:21
	(i tried to clone it btw)	11:21
	→ jettero joined	11:23
jettero	I am vexed by the absence of cogito in ubuntu/debian... is anyone working on that?	11:24
jmspeex	people are still using cogito?	11:24
Mikachu	cogito is depracated	11:24
jettero	oh, I see that	11:25
	well, crap... that's all I've ever really used	11:25
WildPikachu	should i create a git repo first and just rename it myrepo.git? or is there a way to check a file into a bare repo so i can clone it?	11:26
Mikachu	WildPikachu: push	11:27
WildPikachu	hrmmm	11:27
	ah, so i create one in my home dir?	11:28
Mikachu	presumably you already have a repo you want to publish in the bare repo?	11:29
WildPikachu	yea	11:30
	i got up to this far ...	11:30
	git-push /srv/git/myrepo.git	11:30
	hrmmm, i guess i need a ref	11:30
Mikachu	--all	11:30
	← hdu_hh left	11:31
WildPikachu	yea!	11:31
Mikachu	also useful is 'git remote add foo /srv/git/myrepo.git'	11:32
	→ pdherbemont_ joined	11:35
	→ ekidd joined	11:41
	→ ben_h joined	11:42
	→ panagos_ joined	11:42
	← ebel left	11:44
Pieter	Hmm	11:49
	the repository is now 4GB	11:49
	I hope it'll pack	11:49
	← ben_h left	11:49
	← meyering left	11:57
	← DrNick left	11:57
	← panagos left	11:59
	→ Ace_NoOne joined	11:59
Ace_NoOne	hi there - I did a commit where I included two lines that should not be in there yet - how can I retroactively remove them	11:59
	without a new commit that is	11:59
Mikachu	is it the last commit?	11:59
Ace_NoOne	in this case, yes - in general, might not be	12:00
Mikachu	in this case, remove the lines, git add, git commit --amend	12:00
Ace_NoOne	I've thought about that, but that would remove them from the working copy	12:00
Mikachu	in general, git rebase -i HEAD~N, change pick to edit for the commit you want to change, remove the lines, git add, git commit, git rebase --continue, where N is how many commits ago it was	12:00
	it would, yes	12:01
Ace_NoOne	ahh interesting	12:01
	thanls	12:01
	*thanks	12:01
Mikachu	to keep them in the working copy:	12:01
	git branch old	12:01
	do the other stuff	12:01
	git checkout old -- .	12:01
	git branch -D old	12:01
	that'll add the changes in the index too, so you might want to run 'git reset' after (with no arguments)	12:02
Ace_NoOne	cool	12:02
	appreciate your help	12:02
Mikachu	np	12:02
	→ KwisatzHaderach joined	12:04
KwisatzHaderach	Hi, I recorded some git commits with a bad email address. I fixed my email in the global config, but how can I record again the commits? git-rebase -i can change the comments but not the email address.	12:05
	← koke left	12:05
	→ devogon_ joined	12:07
	→ cedricv joined	12:07
dato	hm, send-email --compose creates a new email. is there a way to have it add the comments directly in the first (and possibly single) patch?	12:08
	← panagos_ left	12:10
	→ panagos_ joined	12:11
Ace_NoOne	KwisatzHaderach: you should have seen that coming...	12:12
vmiklos	dato: just use format-patch, edit the patches, then use send-email without --compose?	12:12
dato	vmiklos: yeah, I guess I could do that	12:13
	← devogon left	12:18
	→ devogon joined	12:20
	← ekidd left	12:22
	→ EmilMedve joined	12:23
	→ te joined	12:24
te	gitweb is a pain in the arse to put on nginx	12:24
	anyone have any thoughts on gitweb alternatives?	12:24
KwisatzHaderach	It seems possible with "git-filter-branch --env-filter" but I can't find the right command	12:24
	→ ekidd joined	12:26
jettero	gah, moving from cogito to regular git is going to be painful (and good)	12:27
	cogito was hiding a LOT from me apparently	12:27
	→ ggeecko joined	12:28
	→ devogon__ joined	12:29
Ilari	Pieter: The 4GiB limit was 1) For single pack 2) no longer relevant 3) And for output size. 4) Git-fast-import packs are far from optimal.	12:31
	→ vbabiy-laptop joined	12:31
jettero	is there a way to get 'git status' to use the pager without pipping through less by hand?	12:32
	→ DeFirence joined	12:32
Ilari	jettero: IIRC, it does in newer versions...	12:32
lea_w	I'm getting a clone of the git repository, but none of the heads (pu, man, next, master, ...) show up under git-branch or git-branch -r. How do I access them, and am I wrong to expect them to be there?	12:32
	← devogon_ left	12:33
DeFirence	could someone please give me an example line that excludes a directory to add to /info/exclude	12:33
Ilari	lea_w: They should be there in 'git-branch -r' output...	12:33
	lea_w: Unless it is some whacky copy without them...	12:34
jettero	Ilari: how new... (I'm git version 1.5.4.3)	12:34
Ilari	jettero: I have git version 1.5.6.rc0.40.gd683 (and it does page git status output).	12:35
KwisatzHaderach	git-filter-branch --env-filter "GIT_AUTHOR_EMAIL=foo@example.com" master..branch1	12:35
	→ lcapitulino joined	12:35
KwisatzHaderach	my command works 1 time, but then it fails with "Namespace refs/original/ not empty" :-/	12:36
Ilari	KwisatzHaderach: 1) That would change author email in ALL commits, 2) It is missing 'export' of that variable.	12:36
	→ ironfroggy joined	12:36
DeFirence	Ilari: cant you help me please?	12:36
lea_w	Ilari: Thanks! My clone must have had some problem.	12:36
Ilari	lea_w: That kind of thing could happen if you get clone of clone (and the clone is carelessly done).	12:37
lea_w	Ilari: What does "carelessly" mean here?	12:37
KwisatzHaderach	adding "export" does not fix the failure on "Namespace refs/original/ not empty"	12:37
lea_w	Ilari: I think it might have been a clone of a clone indeed.	12:37
jast	KwisatzHaderach, filter-branch stores a copy of the old refs in refs/original and it will refuse to work again until you move it out of the way or delete it	12:38
Ilari	lea_w: 'git clone git://foo' alone is 'carelessly done' in this context.	12:38
	← Catfish left	12:38
Ilari	lea_w: Since clone only makes local branch to track master...	12:39
tzafrir_home	I need help with adding a command-line option to git-svn: http://paste.debian.net/4674/	12:41
	to allow me to use an SVN client-side certificate for authentication	12:41
	(any better suggestions would be welcomed. If they actually work)	12:41
	← devogon left	12:42
tzafrir_home	Specifically: I'm having some problems with just the right place to stick an extra command-line option	12:42
lea_w	Ilari: Thanks! But, after reading man git-clone, I still don't get it -- what would I have to do to keep the branches? cp?	12:43
Mikachu	tzafrir_home: might get a better response on the mailing list if no one says anything here	12:43
	→ Catfish joined	12:45
	→ deavidsedice joined	12:46
	→ \|mako\| joined	12:47
DeFirence	could someone please give me an example line that excludes a directory to add to /info/exclude?	12:47
	← te left	12:47
jast	DeFirence, /directory (see man gitignore for details; the "ends with slash" form documented only works in 1.5.5 and newer afaik, even if it was documented before that)	12:49
Ilari	lea_w: Basically, soemething like 'git --bare clone ...'...	12:50
\|mako\|	hi	12:51
	→ kristoffer joined	12:51
lea_w	Ilari: that's what i would have thought, but it doesn't work -- just gives me a single master branch.	12:52
\|mako\|	I have another question - git-status shows me a bunch of files that were physically removed from the disk and now I want to utrack them - git-rm --cached would do the trick but is there a way to git-rm them in one go?	12:52
	I mean ... the files are scattered around the project, there is many of them and I don't want to git-rm each of them manually	12:52
	→ Ryback_ joined	12:52
jast	\|mako\|, if you're fine with updating all files that are already tracked in the same go, try git add -u .	12:53
Ilari	\|mako\|: 'git ls-files --deleted -z \| xargs -0 git rm --cached'	12:53
	→ manveru joined	12:54
\|mako\|	thanks a bunch :) good timing	12:54
	← markkalderon left	12:56
	→ pygi joined	12:59
	→ devogon joined	13:01
	← nud left	13:07
jettero	is "git status" the usual way of seeing what was modified?	13:07
Ace_NoOne	jettero: yes	13:07
	← TGEN left	13:08
	→ TGEN joined	13:08
	→ ggeecko_ joined	13:12
	← ggeecko_ left	13:12
	← lea_w left	13:12
	→ lea_w joined	13:12
	→ ggeecko_ joined	13:13
	← ggeecko_ left	13:13
	→ gitte joined	13:13
	→ ggeecko_ joined	13:15
Pieter	Ilari: yeah, I was thinking about my own memory usage when packing :)	13:17
	← ggeecko_ left	13:17
	→ ggeecko_ joined	13:17
	→ mcella joined	13:17
	→ chris2 joined	13:18
	← ggeecko_ left	13:19
	→ robin_ joined	13:19
jettero	how can you tell if your local commits have been pushed to a remote repo?	13:20
	→ ggeecko_ joined	13:21
	→ cmarcelo joined	13:21
	← ggeecko_ left	13:21
	← ekidd left	13:21
Pieter	jettero: git log --left-right master..origin/master will show you what commits you have and remote don't, and also the other way around	13:21
	→ nud joined	13:22
	← ggeecko left	13:25
jettero	Pieter: I get various errors form that. "unknown revision or path not in the working tree." and "Use '--' to separate paths from revisions" ...	13:25
Pieter	oh, you'll have to use 3 dots (...)	13:26
	→ juanjoc joined	13:26
jettero	same	13:27
jast	this will obviously only work if your remote is called 'origin' and the branch is called 'master' on both ends	13:30
jettero	how can I tell what my remote is called? git push and git pull are getting and putting things from somewhere...	13:32
	this transition from cogito is painful so far	13:32
jast	git remote	13:32
	afaik cogito set up remotes a lot differently than git does	13:33
jettero	yeah... I'm getting that. git-remote isn't showing anything, but push/pull still work somehow	13:33
	← mcella_ left	13:34
jettero	ahh, ... I re-cloned and now git remote seems to work	13:35
DeFirence	ive tried adding all sorts of excludes to the exclude file, but git-gui still shows all the files as unstanged changes, anyone know why?	13:36
jast	DeFirence, perhaps you already added the file previously? exclude/.gitignore only apply to untracked files	13:36
	→ ggeecko joined	13:37
DeFirence	the files were not there when i cloned the repo	13:37
	but when i compiled the code it created /bin and /obj directories	13:38
	→ ekidd joined	13:38
DeFirence	now i want it to ignore those	13:38
	but i cant get it to	13:38
	← ToxicFrog left	13:38
jast	so what does your exclude file say right now? (http://git.pastebin.com/ might be helpful)	13:39
	→ wagle_ joined	13:39
	→ schacon_ joined	13:39
	← nud left	13:39
DeFirence	ive tried everything	13:39
	ive tried directories like /bin, bin/, /bin/	13:40
	ive tried the full path	13:40
	etc	13:40
	nothing works	13:40
	→ wurble joined	13:40
	← wurble left	13:41
jast	/bin works fine for me	13:41
	→ koke joined	13:41
jast	obligatory question at this point: which version of git are you using?	13:41
DeFirence	even if its /project/something/bin?	13:41
	→ nud joined	13:42
DeFirence	Git-1.5.5-preview20080413	13:42
jast	yes	13:42
DeFirence	well git status shows all the bin files as modified	13:43
	:/	13:43
	← kristoffer left	13:43
jast	to make sure the files didn't end up in the index by accident, try this: git ls-files \| grep bin	13:43
	← tzafrir_home left	13:43
DeFirence	im putting the paths into /info/exclude btw	13:43
	← wagle_ left	13:44
jast	.git/info/exclude, right?	13:44
DeFirence	git ls-files \| grep bin lists all bin files?	13:44
	yeah	13:44
	← wagle left	13:44
jast	that command should list all files in bin that ended up in the index accidentally	13:44
DeFirence	ending up in the index accidentally?	13:44
	:S	13:44
jast	well, just tell me if it outputs any of the filenames you want ignored	13:45
	← Tv left	13:45
DeFirence	all of those i want ignored? :P	13:46
	→ schapht joined	13:46
	← ekidd left	13:46
jast	yeah, or that. do they show up or not?	13:47
schapht	hi, I commited some large files to my repository by accident. I've been googling around a bit but can't find anything about how to totally remove objects from a repository's history. Can someone point me in the right direction?	13:47
jast	schapht, did you commit stuff you want to keep after that?	13:47
	→ brosner joined	13:48
schapht	no, but I commited some code in the same commit that I'd like to get back. But it wouldn't be the end of the world if I had to redo it.	13:48
jast	so it's the most recent commit, right?	13:48
schapht	one back	13:48
DeFirence	jast: i want all of the files listed there ignored	13:49
schapht	jast: most recent I deleted the files thinking maybe I could prune or something	13:49
	→ wurble joined	13:50
jast	DeFirence, so what probably happened is that you accidentally told git to start tracking (in other words, you put them into the index, for example by using git add) them before you set it up to ignore them, in which case the ignore patterns are basically useless.	13:50
wurble	How do I get a diffstat since a certain date?	13:50
DeFirence	how can i make it not track them then?	13:50
jast	DeFirence, you can remove all that stuff from the index like this: git rm --cached -r <directory to remove>	13:50
DeFirence	ok, thanks	13:50
alterscape	Gitosis question: git/gitosis set up ok on the server, can clone/edit/commit/push gitosis-admin, but when I try to push any other repository (yes, set up in gitosis.conf and pushed), I receive the "ERROR:gitosis.serve.main:Repository read access denied" error.	13:51
	jlh_ → jlh	13:51
Mikachu	wurble: open gitk and pick a good commit sha1 and say git diff --stat sha1	13:51
jast	schapht, well, if it's one back, i'd suggest you first keep yourself a copy of the most recent patch (i.e. the good one): git format-patch HEAD^	13:51
	← brosner left	13:52
jast	schapht, after that, git reset --hard HEAD^ to kill everything (including changes in the working tree) after the commit we want to clean up	13:52
	schapht, then, git rm --cached thefile && git commit --amend to recreate the commit without that file	13:52
	→ wagle_ joined	13:53
jast	schapht, finally, git am <the patch file created earlier>	13:53
	→ ggeecko_ joined	13:55
jast	schapht, the old object will still be kept for a while, though. if you want to do the pruning immediately (at the cost of losing a couple of ways of going back to old states), do git reflog expire --expire=0 --all and then git prune --expire 0. although if you did anything wrong in the previous steps, this will probably eliminate all chances of undoing that mistake.	13:55
	→ discomury joined	13:56
	→ jbunster joined	13:56
jast	alterscape, a common problem is that "writeable" is actually spelled "writable" and the former is ignored by all but the most recent (master) versions of gitosis	13:57
	→ SPAZZZ joined	13:57
	→ cannonball joined	13:57
SPAZZZ	http://www.FakeMagazineCover.com - Upload photo to fake magazine cover templates	13:57
	http://www.SillyWebcam.com - Play with webcam online (games, take pics, effects)	13:57
	http://www.Is-A-Jerk.com - Random Insults/Anon Email	13:57
alterscape	jast: I thought I'd checked that, but let me have another look and make sure. thanks for the response.	13:57
SPAZZZ	http://www.VirtualHolyBible.com - Holy Bible, online version	13:57
	http://www.ScrapbookGenerator.com - Make free online digital scrapbook pages	13:58
	http://www.WHAK.com - W>H>A.K. Comedy Radio	13:58
	← SPAZZZ left	13:58
alterscape	jast: writable is spelled correctly. Let me paste my gitosis.conf into pastebin..	13:59
	← pygi left	14:00
jast	in that case i actually think it's more likely to be a filesystem permissions problem	14:01
	→ janm joined	14:01
lea_w	pasky: So, regarding the rev_parse/parse_rev patch to Git.pm I posted:	14:02
	Is it stylistically ok for parse_rev to return undef for invalid rev names, and throw exceptions for pipe-read (or more general) errors?	14:02
alterscape	ok. I did a sudo chown -R git /Users/git and a sudo chgrp -R git /Users/git ..	14:02
	hm	14:02
pasky	lea_w: I'd say so	14:02
	← \|mako\| left	14:02
wurble	Mikachu: Is there no way to use --since like in git-log?	14:03
lea_w	pasky: Great... I believe you wrote the command (piping) functions in Git.pm, correct? command_oneline (and probably other command methods) don't allow me to check for non-zero return value -- any fundamental reason for that? So if it itches, could I change it (e.g. by adding an option to throw an error on non-zero close, and storing that option in the context)?	14:03
	→ pygi joined	14:03
	← eternaleye_ left	14:03
alterscape	jast: ok.. my permissions string for gitosis-admin.git looks like thus: drwxr-x--- 12 git git 408 May 29 23:21 gitosis-admin.git	14:03
	jast: my permissions string for the other repository looks like thus: drwxr-xr-x 2 git git 68 May 29 20:45 eventory-viz.git	14:04
schapht	jast: that seemed to get the files out of there but my .git is still huge	14:04
	even after the prune ops	14:04
alterscape	or actually, hold on.. hold that last	14:04
	I think I created that folder last night so it probably shouldn't be considered valid info	14:05
pasky	lea_w: in case of non-zeor return value, an exception will be thrown	14:05
alterscape	I'm trying to push a newly created project called test_project.git now, and it won't even get to the point of creating a directory	14:05
pasky	lea_w: so, you catch Git::Error::Command when calling them and can check for $E->value() in the catch block	14:06
lea_w	pasky: That doesn't seem to work for me... Let me check.	14:06
jast	schapht, it's sometimes a bit tricky to get things pruned all the way down. anyway, rest assured that push/pull won't transfer that object and git gc will eventually clean away the old object.	14:06
pasky	lea_w: actually, the config*() methods do exactly that ;)	14:06
	← etnt left	14:07
	← ggeecko left	14:08
Ace_NoOne	I created a new branch from another branch, rather than from master - how can I make this as if I had used the master as "template"?	14:08
	← tcoppi left	14:09
jast	Ace_NoOne, have you committed anything else yet?	14:09
	into that branch, that is	14:09
Ace_NoOne	jast: yes	14:09
	unfortunately s	14:10
	o	14:10
	I noticed when merging with the master	14:10
jast	okay, that calls for a bit of history rewriting	14:10
	→ tcoppi joined	14:10
Ace_NoOne	uh-oh	14:10
	← pygi left	14:11
schapht	jast: so there's one more hitch to this. I already pushed this stuff to my origin. I can just dump my history and start fresh if you think this'll get too complicated.	14:11
jast	schapht, if other people have access to that origin, they already have the 'bad' version so it's possibly a bit late for fixing things	14:11
	Ace_NoOne, i assume you haven't published the changes in any way yet?	14:12
schapht	just me actually, the origin is just for storage really	14:12
	→ Jerdent joined	14:12
jast	schapht, ah, okay. in that case, git push has a -f flag for forcing evil updates :)	14:13
schapht	fun :)	14:13
jast	fun of the git variety	14:13
lea_w	pasky: You're right, thanks. :-) I don't have any way to distinguish non-fatal ("invalid revision name") errors from other problems though. Mh, so it'll always be undef in case of errors, I guess. No exception throwing.	14:14
	pasky: Do you have any general strategy on what to do with STDERR? Discard it to avoid clutter, or allow for error message on STDERR if invalid revision names (yielding undef) are pased?	14:15
	← imyousuf left	14:15
schapht	jast: awesome, that did it	14:15
	thanks!	14:15
	← Ademan left	14:15
	Bombe sometimes misses git-finish-my-code.	14:15
jast	you're welcome	14:15
	← ggeecko_ left	14:15
alterscape	jast: what should I be looking for in terms of permissions? (thanks for the ongoing help btw)	14:16
pasky	Bombe: I bet you could easily do it as diffcore frontend ;)	14:16
Bombe	g	14:16
jast	alterscape, good question. things pretty much worked out of the box here. sure would be nice if gitosis was a bit more verbose, wouldn't it? ;)	14:16
alterscape	jast: yes! I'm writing a blog entry on my experiences, though, so hopefully I'll add to the documentation at least a bit. (it's an OS X 10.4.11 install which is completely unsupported anywhere.. bleh.)	14:17
	→ leachim6 joined	14:17
leachim6	is this old ? http://gitcasts.com	14:17
pasky	lea_w: that is tough question	14:17
Ace_NoOne	jast: sorry, was called awat	14:18
	no	14:18
	just local	14:18
pasky	lea_w: I suggest not to touch STDERR unless failing command is normal condition (e.g. when calling git-rev-parse --gitdir from Git::new, which can fail but that condition is perfectly valid and supported)	14:18
	→ ggeecko joined	14:19
pasky	lea_w: we might want to allow later for applications to globally affect default stderr processing but there's currently no need for that	14:19
jast	Ace_NoOne, git checkout -b newbranch master; git rebase --onto master wrongoldbase oldbranch; (delete oldbranch and rename newbranch)	14:19
	→ Tv joined	14:19
lea_w	pasky: okay, good	14:19
	← robin_ left	14:19
Jerdent	Hi ! I can't clone a remote git repository on my FreeBSD6 box using git 1.5.5. The rep weights about 190MB, is cloned ok on other boxes, but here I got a fatal: Out of memory, malloc failed - fatal: index-pack failed	14:19
Ace_NoOne	jast: lemme try that	14:19
	← robinr left	14:19
pasky	lea_w: (by "not to touch STDERR" I mean "keep it at default value, that is, stderr output left turn on")	14:20
dsop	how can i see the file FOO.c at the state of commit acc01f without traversing the commit -> tree -> blob object?	14:20
pasky	*turned	14:20
dsop	manually...	14:20
Tv	dsop: git show acc01f:FOO.c	14:20
lea_w	pasky: There's some stuff about when and where to merge I'd like to discuss, but gitster should probably around for that. Junio, are you there? If not, I'll move it to the list.	14:20
dsop	Tv: thanks	14:21
	← deavidsedice left	14:21
leachim6	Ok...time for thunderbird to test it's strength ... downloading all the headers from gmane.comp.version-control.git	14:21
	84k of them!	14:21
	← kasimon left	14:21
pasky	lea_w: yes, it's best to discuss this on the list i think	14:21
jast	sounds like fun	14:21
leachim6	do any of you guys use git for you .files ?	14:21
jengelh	dotfiles?	14:21
	Bombe does.	14:22
jast	leachim6, yes, but it's not recommended. git is designed for repositories of files that belong together.	14:22
	← bryanl left	14:22
	→ bryanl joined	14:22
	→ ekidd joined	14:23
pasky	why is it not recommended?	14:23
	there should be no problem per se	14:23
	and some of the dot-files can belong together	14:24
	like, you move some of your stuff from .profile to .bashrc or vice versaq	14:24
leachim6	jast, there was a tutorial somewhere	14:24
	you move all of your config files to .config_files or something like that	14:24
	put them all under git	14:24
jast	as i said, i'm actually doing it myself	14:24
leachim6	and then symlink them to your home folder	14:25
	sounds like a good idea to me	14:25
	also...could someone please explain the difference between git commit and git commit -a	14:25
Tv	i don't even symlink	14:25
	but i'm careful about not causing conflicted merges	14:25
jast	symlinks can break with some programs	14:25
leachim6	I've read documentation everywhere ...	14:25
Tv	jast: that'd be why ;)	14:25
jast	leachim6, do you know what the index (sometimes called staging area) is?	14:26
leachim6	nope	14:26
	I've read about it ... but I don't understand it	14:26
	I've been using -a the whole time	14:26
	→ ToxicFrog joined	14:26
jast	let's summarize the alternative workflow then	14:27
ToxicFrog	I so wish that git-clone and git-fetch had a "pass these arguments to rsync" option.	14:27
jast	you can, at any time, call git add on files to put snapshots of these files into the index. when you later call git commit, exactly these version of the files will be committed. all commit -a does is automatically update all files in the index that are already in there, prior to committing.	14:28
ToxicFrog	Or a backend-independent "please do not murder my net connetion" option.	14:28
jast	in short, the index is a great tool for more fine-grained control of which parts of your uncommitted changes you want to end up in the next commit. with git add -i, you can even add only parts of changes in a given file and keep the rest of the changes in the same file out of the index.	14:29
	and now i'll be away for a while.	14:30
leachim6	so you could skip the index with no ill effects ?	14:30
ToxicFrog	What do you mean by "skip the index"?	14:31
	→ vbgunz joined	14:31
jast	basically, yes, you can ignore that concept. you just lose a bit of git's power by doing so. and now i'm really gone.	14:32
leachim6	ToxicFrog, use -a	14:32
	git commit -a	14:32
	← schacon_ left	14:32
ToxicFrog	That doesn't "skip" the index, it just updates and commits the index in one step	14:32
jast	no i'm not. the final thing i'd like to point out is that you can do both. sometimes i use commit -a and sometimes i don't.	14:32
ToxicFrog	But yes, you can safely use that. Note that if you want formerly-untracked files to be added to git, or currently-tracked files to be removed, -a won't cut it; you'll need to do that explicitly.	14:33
	-a is just "add all the files I already know about to the commit, and then commit them"	14:34
	dwmw2_HEL → dwmw2_gone	14:36
	← ggeecko left	14:37
Ace_NoOne	jast: what did you mean by "wrongoldbase" up there?	14:43
	→ gmcinnes joined	14:44
	→ ggeecko joined	14:46
	→ jmalonzo joined	14:46
Pieter	How hard would it be to get git-clone to support url aliases?	14:47
	← ciaran29d left	14:48
	← holizz left	14:48
	← arun left	14:49
	→ ciaran29d joined	14:49
jettero	Pieter: I think I just saw that under git help config	14:51
Pieter	jettero: it doesn't work for git-clone	14:51
	→ roylee17 joined	14:51
jettero	oh	14:52
	→ arun joined	14:53
	dwmw2_gone → dwmw2_HEL	14:54
Mikachu	wurble: no	14:54
alterscape	jast: fixed my gitosis issue, if you want some knowledge for future reference. The only change I made was to reorder my gitosis.conf so that the members line came before the writable line for each group except gitosis-admin. I have no idea WHY it fixed it, and that might be a red herring for something I accidentally changed in the last hour, but there you go.	14:55
	→ kumbayo joined	14:56
	← janm left	14:59
	→ a-priori joined	14:59
	← ruphy left	14:59
Ilari	alterscape: The file is versioned, so maybe try 'git log -p'? :-)	15:00
jast	Ace_NoOne, the branch you accidentally based your branch on, i.e. the one that should have been master	15:00
Ace_NoOne	ah ok, thanks	15:01
	but ... what is oldbranch then	15:01
alterscape	Illari: The other changes I'm referring to would be access permissions for the git account, etc. But yes, for gitosis.conf changes ;)	15:02
jast	Ace_NoOne, the branch you want to transplant onto master now, i.e. the branch you based upon wrongoldbase	15:02
	say you did git checkout -b wrongbranch notmaster rather than git checkout -b wrongbranch master	15:03
	and that's what you want to fix now	15:03
	then you do git rebase --onto master notmaster wrongbranch	15:03
jengelh	one parameter too much it seems	15:03
jast	in fact the whole renaming and deleting business is optional afaics	15:04
	jengelh, it would be if not for --onto	15:04
jengelh	git checkout branchtorebase; rebase --onto stickithere cutitoffhere	15:04
	→ schacon_ joined	15:04
Ace_NoOne	jast: ahh I see	15:04
jast	yeah, i suggested a few gratuitous steps earlier	15:04
Ace_NoOne	thanks jast	15:05
jast	not that they do any damage	15:05
Ace_NoOne	:)	15:06
	← solars left	15:06
	→ solars joined	15:06
	→ ruphy joined	15:09
	→ lea_1 joined	15:10
	→ bdiego joined	15:11
	→ mithro joined	15:12
	← gmcinnes left	15:15
	← Sho_ left	15:16
	→ krawek joined	15:16
gebi	git-svn makes a really good svndump replacement for non-root users :)	15:17
	it just saved my day because after a bigger commit the svn repos was dead	15:17
	→ Sho_ joined	15:19
ToxicFrog	It seems like a disproportionaly large amount of the discussion in here is about git interaction with svn. I do wonder why.	15:21
	→ schacon__ joined	15:23
	← schacon_ left	15:23
jast	probably because svn is often forced on you and git-svn is the least painful way to deal with that	15:24
	← reel_ left	15:25
	← bdiego left	15:27
	← lea_w left	15:28
	→ brosner joined	15:29
	← ggeecko left	15:29
	→ holizz joined	15:32
	→ bobesponja joined	15:32
DeFirence	when you use 'git commit -a'	15:34
	→ jeckel joined	15:34
DeFirence	what do you do after you've typed the commit msg??	15:34
jast	save and exit the editor	15:34
DeFirence	how?	15:35
	on git bash windows	15:35
	:p	15:35
	← agib left	15:35
jast	well, can you see what editor you've got there?	15:35
	my first guess would be nano	15:36
DeFirence	i have no idea	15:36
	its just an editor	15:36
jast	if it's nano it says so in the topmost line	15:36
	← WildPikachu left	15:36
	← Ace_NoOne left	15:37
Bombe	Does it show "~" at the start of every line below the commit message?	15:37
DeFirence	says the path and the lines at the bottom	15:37
jast	heh :)	15:37
DeFirence	no	15:37
	and no	15:37
Bombe	(Remember: If your editor doesn't show these it's TOTALLY BROKEN. :)	15:37
	→ krh joined	15:37
jast	let's save the confusion for later... :P	15:37
DeFirence	what?	15:37
Bombe	:)	15:37
DeFirence	lol	15:37
	:P	15:37
Bombe	DeFirence, what happens when you press ":"?	15:38
ToxicFrog	DeFirence: try hitting escape several times, then :w	15:38
jast	so did you install git via cygwin or is that msysgit?	15:38
ToxicFrog	And then enter.	15:38
jast	but... but... no tildes! :(	15:38
DeFirence	msysgit	15:38
	: works :P	15:38
	→ gmcinnes joined	15:38
Bombe	DeFirence, that's nice but what does it do? :)	15:38
DeFirence	: lets me enter a command	15:39
ToxicFrog	DeFirence: "works" in what sense?	15:39
	Ok, so you're using a vi-oid.	15:39
jast	in that case, :wq (enter) saves and exits	15:39
Bombe	Try "wq" as a command. :)	15:39
DeFirence	yeah, ty	15:39
ToxicFrog	:w will save, :wq will save and exit and :q! will exit without saving.	15:39
Bombe	Hmm. Strange. No tildes!	15:39
jast	consider using a different editor if you don't want the learning curve of a vi-like one :}	15:39
ToxicFrog	Not all vi-oids use the tildes.	15:39
	And some terminal editors have weird color configurations that mean they don't show up even if vi uses them.	15:40
jast	though i wouldn't know what else msysgit ships with	15:40
ToxicFrog	Er, terminal emulators.	15:40
jast	probably nothing	15:40
ToxicFrog	Well, it's a windows machine, so you can always do something like env EDITOR=notepad.exe git commit	15:40
	And msys should come with nano	15:40
jast	ah	15:40
	but notepad forks into the background, doesn't it?	15:40
Bombe	I don't think notepad.exe will work.	15:41
	Yes, it does.	15:41
ToxicFrog	(assuming msysgit is msys+git, not git + minimal subset of msys)	15:41
charon	it only needs vi. the only other option would be emacs, but that contradicts the assumption that the user wants to run/keep windows as OS	15:41
ToxicFrog	It does?	15:41
	Hmm, been ages since I used it, so.	15:41
Bombe	Yes.	15:41
ToxicFrog	That's annoying.	15:41
Mikachu	what about nano?	15:41
jast	pretty much all win32 GUI apps go to background	15:41
	(iirc)	15:41
DeFirence	env EDITOR=notepad.exe git commit works, but i think wordpad would work better	15:41
	since notepad doesnt work nice with LF instead of CRLF i think	15:42
jast	you can always use a real editor instead ;)	15:42
	there's tons of decent ones, e.g. notepad++	15:42
ToxicFrog	Aah, yes	15:42
jast	though, of course, none can beat vim	15:43
Bombe	So true.	15:43
DeFirence	but then ill have to type env editor = w/e before every commit line :/	15:43
jast	the only editor i've got that can be scripted in four different languages!	15:43
	→ pygi joined	15:43
Bombe	git config core.editor notepad.exe should take care of that, DeFirence.	15:43
jast	no, you can change environment settings in the control panel (system -> something)	15:44
	or that	15:44
ToxicFrog	Bombe: assuming he has a recent version	15:44
jast	though git config --global is probably more useful	15:44
Bombe	ToxicFrog, I have no idea when core.editor was added. I started using git a week ago. :)	15:44
jast	i don't think particularly non-recent versions of git exist as msysgit packages	15:44
ToxicFrog	And sys it's msys, you could also put it in your .bash_profile	15:44
	← Jerdent left	15:44
ToxicFrog	I think it was in 1.5.3.0	15:44
	1.5.2.x doesn't have it	15:45
jast	of course we all know what version debian stable ships with :}	15:45
	← EiNZTEiN left	15:45
Bombe	What's this debian thingy I keep hearing people talking about?	15:45
rafl	is there a git command to find the .git/ directory from within a subdirectory in a repository?	15:46
jast	Bombe, it's a special fork of openssl	15:46
ToxicFrog	Well, ubuntu 7.10 (which, to my sorrow, I must use at work) is still using 1.5.2.5	15:46
	So debian stable is probably somewhere around 1.5.0	15:46
Bombe	jast, ah... weak cryptography to comply with US export restrictions, I assume?	15:46
jast	rafl, git rev-parse --git-dir	15:46
	ToxicFrog, actually it's 1.4.4	15:46
rafl	jast: thanks!	15:47
jast	Bombe, in a manner of speaking, yeah.	15:47
	→ Overshard joined	15:47
	← mithro left	15:47
jast	Bombe, actually it was phenomenally weak cryptography to make debugging easier, thanks to an overzealous patch not really discussed with upstream	15:47
	hint: what do you get when your PRNG produces at most 32768 different seeds?	15:48
Bombe	jast, I read the discussion and upstream itself wasn't really sure about what to do. :)	15:48
jast	upstream wasn't aware that the patch was supposed to go into a production package	15:48
Bombe	jast, which doesn't really matter... if upstream knew that this patch is bad it shouldn't be in a testing package, either.	15:49
jast	anyway, nothing that a liberal application of decapitations can't fix.	15:49
Bombe	Right.	15:49
	← jeckel left	15:49
jast	upstream assumed that the question was asked for local debugging purposes, i.e. not supposed to ever go into any sort of package at all.	15:49
Bombe	Communication FTW.	15:49
	Bombe goes outside to have a smoke.	15:49
jast	the maintainer didn't make his goal very clear at all and he didn't include a patch, just a description of what he wanted to change	15:50
ToxicFrog	jast: and their answer, IIRC, was "use -DPURIFY", not "sure, comment it out"	15:50
jast	there were two different answers afaik	15:51
	one was "if it helps debugging i'm all for commenting it out"	15:51
	→ d2dchat joined	15:51
gebi	ToxicFrog: there are far more svn repositories than git out there, and every git user i know of uses git-svn to deal with them ;)	15:52
jast	it wasn't really hard to convince two of my colleagues (out of three) to start using git-svn even if they had never tried git	15:53
	→ larsl joined	15:54
gebi	as soon as there is one active developer using git-svn it becomes quite stressfull for the svn devs g	15:54
	→ zachinglis joined	15:55
Mikachu	and for people on #git :)	15:56
DeFirence	does git-gui not care about excluded files and directories?	15:56
jast	yes it does	15:56
DeFirence	cause it shows all the untracked excluded files as unstaged changes	15:56
jast	well, in the sense that it respects the ignore patterns	15:56
	did you do what i suggested you do?	15:56
Mikachu	DeFirence: and git status doesn't list them?	15:56
DeFirence	yes jast	15:57
	→ aroben joined	15:57
jast	okay, Mikachu's question next :)	15:57
DeFirence	http://rafb.net/p/hKaLG492.html	15:59
	thats my git status	15:59
	so its not excluding	15:59
	← gmcinnes left	15:59
DeFirence	my excludes file has the following:	16:00
	Thumbs.db	16:00
	"/bin"	16:00
	"/obj"	16:00
	→ gmcinnes joined	16:00
jast	/bin will only work on the top level	16:00
	to ignore all files/directories called bin, leave out the /	16:00
	← _zack left	16:00
DeFirence	ahhh	16:01
	→ schacon_ joined	16:02
	← schacon__ left	16:02
	→ mvanholstyn joined	16:02
	← lea_1 left	16:02
	→ hipertracker joined	16:02
	→ psilo2_ joined	16:03
	→ acm joined	16:04
vbabiy-laptop	Hey if I mad a change to a file and I want get clear the change before I commit how can I reset it	16:07
	Like it CVS I would just rm the and do a update	16:08
larsl	git reset --hard	16:08
	Oh, just one file?	16:08
	Just don't git-add it before you git-commit.	16:08
Bombe	Hmm... if I have added a file to the index and changed it afterwards, how can I reset it to what's in the index?	16:09
jast	Bombe, git checkout -- file	16:09
Bombe	Ah.	16:09
	← nud left	16:09
Bombe	Thanks.	16:09
	→ jeckel joined	16:09
	→ EiNZTEiN joined	16:13
	← psilo2 left	16:13
	→ lea_w joined	16:14
	→ EiNZTEiN_ joined	16:14
	← d2dchat left	16:15
	← ruphy left	16:17
	→ Xyri0 joined	16:17
	← alterscape left	16:17
	← discomury left	16:19
	← redondos left	16:20
	→ VitaRara joined	16:23
	→ eternaleye_ joined	16:24
	← schacon_ left	16:26
	← Xyrio left	16:27
	Xyri0 → Xyrio	16:27
	← arun left	16:28
	← EiNZTEiN left	16:28
	EiNZTEiN_ → EiNZTEiN	16:28
	← zachinglis left	16:32
pasky	uhhm	16:36
	→ horos22 joined	16:36
pasky	please, make it clear to me	16:36
	if I use --pretty=format, there's really _no_ way to get rid of the totally stupid extra 'commit' lines?	16:36
	pasky hates git :)	16:37
Eridius	pasky: what's wrong with those commit lines?	16:37
pasky	Eridius: they get in my way! :)	16:37
Mikachu	how do you mean?	16:37
Eridius	if you hate them that much, you can grep them out	16:37
pasky	when I use --pretty=format, that means _I_ want to have _full_ control on the format of the output	16:37
Mikachu	if i write git log -1 --pretty=format:hello, it just prints hello and nothing else	16:37
pasky	not wanting rev-list get in my way trying to outsmart me	16:37
Mikachu	ah, in that case, use log instead of rev-list? :P	16:38
pasky	pasky@pixie:~/src/git> git rev-list -1 --pretty=format:hello HEAD~100..HEAD	16:38
	commit 1af8bca79766198c591d8db58f778a98e96627e2	16:38
	← lcapitulino left	16:38
pasky	hello	16:38
	hm	16:38
	← mcella left	16:39
pasky	my idea was that since I'm coding porcelain I'm going to use the plumbing	16:39
	but apparently there's still no clear distinction of that in git	16:39
	oh well	16:39
	still, having such a distinction is really obscure	16:39
	but thanks :)	16:39
	pasky will not elaborate on the fact that -1 is undocumented	16:39
Mikachu	it was a pure accident i found it, since you didn't mention what command you were using :)	16:40
pasky	heh :)	16:40
	← horos22 left	16:42
	→ lcapitulino joined	16:42
pasky	ok, -1 is actually documented	16:42
	at least for log	16:42
	← gitte left	16:44
Eridius	wow	16:45
	mischan	16:45
	← eternaleye_ left	16:45
	→ rubydiamond joined	16:49
	→ arun joined	16:50
	← ciaran29d left	16:52
	→ d2dchat joined	16:55
	← priidu left	16:56
	→ timocratic joined	16:57
	→ mithro joined	16:59
	← timocratic left	16:59
	→ priidu joined	16:59
	→ timocratic joined	17:00
	→ ckoehler joined	17:01
	← mvanholstyn left	17:02
ForsakenSoul	hello	17:03
	can someone guide me through the commit process please ...	17:03
ToxicFrog	In general, it's just:	17:04
Ilari	ForsakenSoul: Having trouble understanding the index (a.k.a. staging area)?	17:04
ToxicFrog	git add <list of files or directories you want included in the commit> # this adds them to the index	17:04
	git commit # this turns everything in the index into a commit, and stores it	17:04
	→ tenpaiyomi joined	17:06
	← charon left	17:06
tenpaiyomi	What do I do if I'm trying to commit my changes to the server, yet I'm getting "Cannot dcommit with a dirty index. Commit your changes first, or stash them with `git stash'.", however when I do git status, it states I have nothing to commit	17:06
ToxicFrog	What happens if you try git reset --mixed HEAD?	17:08
	→ tvachon joined	17:08
tenpaiyomi	That worked	17:09
	Awesome, thank you	17:09
ForsakenSoul	ilari ... I know the concepts of the thing and I know the command but when I type git-commit with the files already added	17:09
	I get the index in my console to edit it	17:09
	and don't know how to exit from here	17:09
	is this because I have some unmonitored files ?	17:09
Ilari	ForsakenSoul: It is asking for commit message.	17:09
	ForsakenSoul: The lines prefixed with '#' are just comments and ignored.	17:10
	← gmcinnes left	17:10
ToxicFrog	ForsakenSoul: the default editor is vi; hit insert a few times, type stuff in, then exit with <esc>:wq<enter>	17:11
ForsakenSoul	ok	17:11
ToxicFrog	Alternately, set the environment variable EDITOR to something else like nano, or git config core.editor	17:11
ForsakenSoul	I'll try that	17:12
ToxicFrog	alternately alternately, use "git commit -m <message>" if you're ok with a short commit message	17:12
Ilari	Isn't it also possible to write commit message to file and use '-F <filename>' to commit?	17:13
	← bentob0x left	17:13
ToxicFrog	Aah. Yes, that'll work	17:14
	→ ggeecko joined	17:14
ToxicFrog	Or even -F-, type it in and hit ^D to finish	17:14
	→ gmcinnes joined	17:14
	→ nud joined	17:14
ForsakenSoul	toxicfrog * no commit message? aborting commit.	17:14
	← ggeecko left	17:15
ForsakenSoul	would you give me an example for a commit message	17:15
ToxicFrog	Anything non-whitespace	17:15
	And non-comment	17:15
	One of my recent ones, say:	17:15
	← gmcinnes left	17:15
	→ jaf joined	17:15
ToxicFrog	Fixed configuration stuff	17:15
		17:15
	- chmod a-x config files	17:16
	- changed default tileset to "lite" version	17:16
ForsakenSoul	didn't you say non-whitespace ?	17:16
ToxicFrog	As in "contains stuff other than whitespace"	17:16
	If your commit message is entirely whitespace and/or comments, it'll treat it like no message	17:16
	So you can't just immediately save and quit when it asks you for a message, you have to type something, or use -m or -F	17:17
jaf	I am trying to do my first git commit, and keep getting the error "fatal: There was a problem with the editor vi." - nothing in the docs that I can find on this	17:17
	this is on os x leopard	17:17
ToxicFrog	jaf: by default, it tries to launch vi so that you can type in a commit message	17:18
	I suspect this is what happens if you can't	17:18
	try any of:	17:18
jaf	this happens when I try to exit vi	17:18
	it launches vi fine	17:18
ToxicFrog	Ok, this is also probably what happens if vi exits with a nonzero exit code	17:18
	$ env EDITOR=nano git commit # try a different editor, or	17:18
ForsakenSoul	toxixfrog how do I set up a different editor like gedit for instance ?	17:18
ToxicFrog	$ git commit -m message # enter message on command line, or	17:19
	$ git commit -F file_with_message_in_it # create the message seperately	17:19
ForsakenSoul	do I have to do the env editor command every time ?	17:19
ToxicFrog	ForsakenSoul: if you have a recent git (1.5.3.x+), you can set the core.editor configuration variable	17:19
jaf	ToxicFrog: no way to just use vi?	17:19
ToxicFrog	Eg, git config --global core.editor gedit	17:19
	jaf: if you can get vi to exit normally rather than exiting with an error code, it should work, I think	17:20
ForsakenSoul	ok .. I'll just do it in a different file	17:20
	it would be less effordless	17:20
ToxicFrog	But this might be some weird OSX-specific bug or something, I don't know	17:20
	ForsakenSoul: what?	17:20
ForsakenSoul	and I can do it more detailed	17:20
jaf	ToxicFrog: ok, thanks for the info. much obliged.	17:20
ForsakenSoul	$ git commit -F file_with_message_in_it # create the message seperately	17:20
ToxicFrog	jaf: one thing you might be able to do, if you can't fix vi itself, is try a script that just does "vi $@; exit 0"	17:21
	And then make that your editor	17:21
	→ mediogre joined	17:21
ToxicFrog	ForsakenSoul: that's easier than using env or setting core.editor?	17:21
ForsakenSoul	well maybe not ... but I'm still very very new to git	17:24
	and don't want to mess around with the settings just for now	17:24
	just to get a bit confortable	17:24
	and than I'll play with it	17:24
	← d2dchat left	17:25
ToxicFrog	Ok	17:25
	Well, for future reference, the man page for the settings is man git-config	17:26
	The setting in question is core.editor, if you have a sufficiently up to date editor	17:26
	And the easiest nonpersistent way to do this is probably env, as it doesn't require seperately creating and editing files.	17:26
	← psilo2_ left	17:27
ForsakenSoul	well I might have one all the time	17:27
	and just update it	17:27
	:D	17:27
	well at least one for every project	17:27
	:D	17:27
	→ Eridius\|work joined	17:28
ToxicFrog	I would find that terribly inconvenient, but if it works for you, go for it	17:28
	→ gmcinnes joined	17:28
	→ schacon_ joined	17:29
	← ekidd left	17:32
	← lea_w left	17:32
	→ edwardog joined	17:32
edwardog	Using git-svn, how do I create new remote tags or branches?	17:33
	← acm left	17:34
	← schacon_ left	17:35
	← jaf left	17:35
	→ ekidd joined	17:36
Eridius\|work	edwardog: git push remote localbranch:remotebranch will create a new remote branch	17:37
	as for tags, try passing --tags	17:37
	→ dbhumphreys joined	17:38
dbhumphreys	so, i've got a clueless newbie question.	17:38
Thumper_	Eridius\|work: in git-svn ? I didn't think there was a way to do that...	17:38
dbhumphreys	i clone a repository from machine a to machine b. works! ... i push the changes from b back to a. i go back to a.	17:39
Eridius\|work	oh, missed the -svn part	17:39
dbhumphreys	how do i merge the change from b?	17:39
ToxicFrog	dbhumphreys: let me guess - the repo on a isn't bare	17:39
Eridius\|work	dbhumphreys: if your working tree is unmodified, a simple `git reset --hard HEAD` will work	17:39
dbhumphreys	that's probably the case, since i'm not sure what you're talking about.	17:39
edwardog	Eridius\|work: yeah, it’s a little funky. I get errors like	17:39
	fatal: 'svn': unable to chdir or not a git archive	17:39
	fatal: The remote end hung up unexpectedly	17:39
ToxicFrog	I think "git reset --hard <commit>", where <commit> is probably HEAD, will do it	17:39
	But in general pushing to non-bare repos is discouraged	17:40
dbhumphreys	okay.	17:40
edwardog	(Checking .git/config, I see that git-svn has named 'origin' as 'svn')	17:40
ToxicFrog	dbhumphreys: a bare repo, as it says in the tutorial, is one with no working tree - only the repo itself	17:40
dbhumphreys	so, the right way to do that is ...?	17:40
edwardog	Eridius\|work: good instinct though	17:40
ToxicFrog	you can pull from and push to it, but can't work directly in it	17:40
dbhumphreys	okay.	17:40
Eridius\|work	edwardog: to create branches and tags, the easiest thing is to use svn directly using URLs as the arguments	17:40
ToxicFrog	You can create one with git --bare init, and clone to a bare repo with git clone --bare	17:40
dbhumphreys	so, make a repository on machine a. pull from that repository on machines a and b?	17:41
ToxicFrog	Yes.	17:41
Eridius\|work	if you really want to do it within git-svn, you can try creating git remote-tracking branches that correspond to the git-svn tags/branches format, but I have no idea if git-svn will actually obey them	17:41
dbhumphreys	then i can push from both.	17:41
edwardog	Eridius\|work: yeah, that’s what I was thinking. I guess the git bridging thing only takes you so far	17:41
ToxicFrog	You can do it without a bare repo, but in that case it's much preferred to do pull-only - ie, b pulls from a, a pulls from b	17:41
	→ christastrophe joined	17:41
dbhumphreys	hm. okay.	17:41
	rubydiamond → lol	17:41
edwardog	Eridius\|work: ah, yeah, that’s a good idea	17:42
	cool, thaks	17:42
ToxicFrog	Since if b pulls from a, and then pushes to a, and then you want to work in a, which has a working copy, things can get messy fast, especially if you have uncomitted or unmerged changes in a	17:42
dbhumphreys	right. okay.	17:42
	thanks!	17:42
christastrophe	does gitosis have its own IRC channel for asking quesions?	17:42
ToxicFrog	The other common idiom for this is that a has an "incoming" branch	17:42
Eridius\|work	edwardog: honestly, even when using svn I like to create branches/tags using URLs (since I generally don't have the root of the repo checked out)	17:42
	→ DrNick joined	17:43
ToxicFrog	so you pull from a to b, edit some stuff in b, and then push to a incoming rather than a master.	17:43
	And then next time you're on a, you merge from incoming into master.	17:43
dbhumphreys	tricky.	17:43
	okay.	17:43
ToxicFrog	Not usually.	17:43
	→ kanru joined	17:43
	← sverrej left	17:44
	→ shoe joined	17:44
ToxicFrog	But the underlying thing here is that pushing to a branch that corresponds to a working copy is more trouble than it's worth most of the time - you either push to a bare repo, or you push to a not-currently-checked-out branch on a normal repo and then merge.	17:44
edwardog	Eridius\|work: to do that with a url, do I just do a svn co https://svn.whatever.com/my_project/branches/my_new_branch ?	17:44
dbhumphreys	"bare repo." ... nothing in there? or just nobody working inside of it?	17:45
Eridius\|work	edwardogyeah	17:45
	→ samgranieri joined	17:45
edwardog	Eridius\|work: sweet. Thanks	17:45
ToxicFrog	The latter, sort of.	17:45
Eridius\|work	if you want to save on typing you can do svn co https://svn.whatever.com/my_project/{trunk,branches/new_branch}	17:45
	bash syntax ftw!	17:45
ToxicFrog	Like I said earlier, a bare repo is one with no working copy	17:45
dbhumphreys	cool.	17:45
	→ mvanholstyn joined	17:46
ToxicFrog	Normally, for a git repo foo, the working copy is in foo/ and the git files are in foo/.git/	17:46
	For a bare repo, the git files are in foo/ and there's no working copy.	17:46
edwardog	Eridius\|work: whu? What do the braces do there?	17:46
ToxicFrog	edwardog: argument expansion.	17:46
dbhumphreys	okay. that was helpful -- thanks!	17:47
Eridius\|work	edwardog: in bash syntax, the braces expand into multiple args	17:47
ToxicFrog	foo/{bar,baz} turns into foo/bar foo/baz	17:47
Eridius\|work	yeah	17:47
	← larsl left	17:47
ToxicFrog	v. useful	17:47
Eridius\|work	foo{,bar} becomes foo foobar	17:47
edwardog	when you tab-complete?	17:47
Eridius\|work	you can even nest	17:47
	edwardog: no, when you hit return	17:47
	foo{bar,baz,blah/{one,two}} becomes foobar foobaz fooblah/one fooblah/two	17:47
	→ _graham_ joined	17:47
	← cedricv left	17:47
ToxicFrog	edwardog: test it with echo, if you like	17:47
	bkelly@b-kelly:~/devel/tools/performance/cpu_mon$ echo foo{bar,baz,blah/{one,two}}	17:48
	foobar foobaz fooblah/one fooblah/two	17:48
Eridius\|work	it's incredibly useful	17:48
edwardog	umm... ok, so this thing is doing some sort of zip? I’m going to echo stuff and try to figure out why this is useful	17:48
Eridius\|work	it saves typing	17:48
	for example, in the URL, it meant the root of the project only had to be typed once	17:49
	https://svn.whatever.com/my_project/{trunk,branches/newbranch}	17:49
ToxicFrog	Rather than having to type out "https://svn.whatever.com/my_project/trunk https://svn.whatever.com/my_project/branches/newbranch" in its entirety.	17:49
edwardog	oh!	17:49
	ok	17:49
	sorry, my eyes were slipping or something	17:50
	I thought I saw an extra slash in there	17:50
	→ advany joined	17:50
ToxicFrog	There's no magic occuring here, this is just a useful shell construct.	17:50
edwardog	Yeah, that makes sense	17:50
	so it copies the string before it	17:50
	gotcha	17:50
	Thanks!	17:50
	→ charon joined	17:52
	lol → rubydiamond	17:52
	← z1o left	17:52
	glauber_ → glommer	17:52
	← Catfish left	17:57
	← EiNZTEiN left	17:59
	→ p4tux joined	18:00
	→ eikonos joined	18:02
	→ zachinglis joined	18:03
	→ felipec joined	18:05
	← koke left	18:06
	→ koke joined	18:06
	← Overshard left	18:06
	← koke left	18:07
	← Gitzilla left	18:09
	← deavid left	18:10
	→ lea_w joined	18:17
	← priidu left	18:18
	← manveru left	18:21
	→ threeve joined	18:22
jettero	is there a way to combine git remote git+ssh://host/dir/dir with git config branch.master.merge refs/heads/master so my push pull work?	18:24
	or is it just two commands?	18:24
	← edwardog left	18:26
	← zachinglis left	18:27
	← warthog9 left	18:27
DeFirence	why does git say theres still merge conflicts in file x if ive fixed them?	18:28
Eridius\|work	you have to git add the file	18:28
	the index still has it marked as unmerged	18:28
DeFirence	im using git gui	18:28
	on windows	18:28
	:p	18:28
	→ tahooie joined	18:29
DeFirence	when i try commit a file that ive fixed conflicts in	18:29
	it says file x still has conflicts	18:29
	→ erikveen joined	18:29
	Eridius\|work can't help you with windows, sorry	18:30
doener	DeFirence: should boil down to the same though. The "unstaged changes" window should still list your file. Stage it	18:31
	← rubydiamond left	18:32
DeFirence	hmm	18:32
	im trying	18:32
	but it gives a merging error	18:32
tahooie	My project compiles an executable. I'd like the executable to be in the git repository so that people who check out the repository will have it, but I would like git to -NOT- keep a revision history of it because of size issues. Can I do this?	18:33
doener	no	18:34
tahooie	ok. thanks	18:34
	← mvanholstyn left	18:34
DeFirence	if i use git bash	18:34
	i can git add file	18:34
	but in the gui i cant	18:34
	$ git merge ?	18:35
	You are in the middle of a conflicted merge.	18:35
doener	Hm, no clue then, I don't use git gui	18:35
DeFirence	ive never done merges before	18:35
	uhm	18:35
	← erikveen left	18:35
DeFirence	what would i do next if i was using command line git?	18:35
Eridius\|work	DeFirence: so in git gui you can't just stage the file?	18:36
	← gmcinnes left	18:36
DeFirence	oh wait	18:36
	i can now	18:36
Eridius\|work	tahooie: in theory you could wrap git with commands that re-write the latest commit to remove the binary when creating a new commit ;)	18:36
DeFirence	odd...	18:36
Eridius\|work	but that would really fuck with people doing a pull	18:37
tahooie	Eridius\|work, sounds like way more hassle then I'd like to deal with	18:37
ToxicFrog	DeFirence: with command line, typically you either merge manually or use git mergetool, then git add it all, then git commit	18:38
	resolve the conflicts in the merge, rather	18:38
DeFirence	ok	18:38
	angasule_ → angasule	18:38
jettero	is there a way to hook some extra things to do before a 'git status'?	18:39
	← kelvie_ left	18:41
	← tahooie left	18:42
	→ queso joined	18:42
	→ mvanholstyn joined	18:42
	→ janm joined	18:42
	→ com4 joined	18:43
com4	is there a problem using git over a samba share?	18:43
	→ gmcinnes joined	18:44
	→ kelvie_ joined	18:44
	← dbhumphreys left	18:44
	← mediogre left	18:44
drizzd	Probably. You'll want the new case insensitivity features, at least.	18:44
	← mvanholstyn left	18:45
com4	drizzd: how about just on ntfs filesystems?	18:46
	→ koke joined	18:46
drizzd	well, I'd think that's slow as hell. But I guess you'll just have to try.	18:47
com4	i don't care if it's slow :) i'm not the one insisting on using windows	18:47
	← lcapitulino left	18:48
	→ warthog9 joined	18:48
	→ d2dchat joined	18:49
ToxicFrog	com4: I've used it on ntfs under windows, with no problems in either functionality or performance. I haven't tried it on ntfs under linux.	18:49
drizzd	com4: well, if you use windows then checkout msysgit. And to synchronize your repository to a fat/ntfs filesystem, you can just use git-push.	18:49
ToxicFrog	(this is cygwin git, btw)	18:49
drizzd	s/checkout/check out/	18:49
	oh	18:49
ToxicFrog	I haven't tried msysgit, so can't comment on it.	18:49
drizzd	I guess cygwin git makes sense, since there is no full GUI support for git you'll need some kind of console anyways.	18:50
	→ zachinglis joined	18:50
ToxicFrog	That's why msysgit comes with bash, presumably.	18:53
	← jmalonzo left	18:54
com4	awesome. thanks	18:56
	← zachinglis left	18:57
	→ dirk1296 joined	19:00
	→ Eludias joined	19:01
	← leachim6 left	19:01
	← kanru left	19:04
	← sgrimm left	19:05
	→ kanru joined	19:08
drizzd	oh it does?	19:09
	→ warlock_mza joined	19:09
	→ thiago_home joined	19:11
	→ redondos joined	19:14
	← PerlJam left	19:14
	← queso left	19:15
	← Ramune left	19:15
ToxicFrog	drizzd: from conversations in here, yes - otherwise it would be largely unusable	19:16
	→ robin_ joined	19:17
	→ benstiglitz joined	19:17
	← benstiglitz left	19:17
	← krh left	19:17
ToxicFrog	The git GUI tools still being very immature, especially when compared to SVN, Perforce, etc.	19:17
drizzd	yeah, I just remember trying it a while a go. The part about the bash slipped my mind, but it's still unusable without proper integration with other development tools.	19:18
	→ capisce joined	19:20
tflsh	anyone know how i can do a git clone of a specific tag?	19:20
	e.g. to clone the v1.5.5.3 tag of the git srcc?	19:21
Ilari	tflsh: You can't do it with git-clone (it gets all branches and tags too). init+fetch+checkout would work...	19:23
jettero	I think git init a dir and then fetch ... what ilari said	19:23
ToxicFrog	drizzd: eh, I have no problem using it with no more integration other than nedit's shell menu	19:23
	(it being git in general, not msysgit in specific)	19:23
	← redondos left	19:23
tflsh	hmm ok, git init; git fetch git://git.kernel.org/pub/scm/git/git.git ?	19:23
Ilari	tflsh: OTOH, just 1.5.5.3 would probably be around 99% of the size of full clone...	19:24
	→ redondos joined	19:25
tflsh	ok, well so i understand, what would the fetch step be?	19:25
Ilari	tflsh: 'git fetch <URL> v1.5.5.3:v1.5.5.3' (or something)...	19:26
tflsh	ah	19:26
jettero	there's also git remote add -t something -m something name url	19:26
	you can make it work just like a clone, but only for specific tags ...	19:26
tflsh	fatal: Couldn't find remote ref v1.5.5.3	19:27
	hm	19:27
	that's from the git fetch	19:27
	← mithro left	19:29
	→ strangy joined	19:30
	→ EiNZTEiN joined	19:32
	→ ruphy joined	19:32
jettero	git when you git config alias.test !ls ... what shell does it use?	19:32
	→ lcapitulino joined	19:33
Ilari	tflsh: Try 'git fetch <URL> refs/tags/v1.5.5.3:refs/tags/v1.5.5.3'	19:33
ToxicFrog	jettero: I don't know offhand, but I'd guess either $SHELL or the shell used by system()	19:33
	← d0k left	19:34
	→ ceez joined	19:35
tflsh	Ilari: thanks,, i think this is working now	19:36
	Ilari: yep! thanks!	19:39
	← samgranieri left	19:39
	← gmcinnes left	19:40
	→ cedricv joined	19:42
	→ shaftyy joined	19:42
	→ nessundorma joined	19:44
	← d2dchat left	19:46
	→ jmalonzo joined	19:47
tflsh	is there a way to remove a commit that was already push'ed to a remote rpository?	19:50
Mikachu	3219 pts/30 S+ 0:00 \| \| \_ git fla	19:51
	3220 pts/30 S+ 0:00 \| \| \_ sh -c cat;sleep 60	19:51
	when i made the alias just cat, cat was a direct child of git	19:51
thiago_home	tflsh: yes, but you shouldn't do it.	19:51
	tflsh: has anyone got that commit already?	19:51
tflsh	thiago_home: i dont really need to as this is just a test repo i'm playing with, but i'm interested	19:51
	thiago_home: nope, it's just me	19:51
thiago_home	remove the commit locally, then do a forceful push	19:52
	git push -f <remotename>	19:52
tflsh	ahh	19:52
	thanks!	19:52
	→ shaftyy_ joined	19:53
	→ zachinglis joined	19:54
tflsh	yeah, im sure that would be messy if people had already pulled that commit	19:55
thiago_home	you can also set up git to always force	19:55
	it's useful if you're pushing to your backup :-)	19:55
Mikachu	i wouldn't want force push by default on my backup :)	19:56
Ilari	Well, if it has reflogs forced... :-)	19:57
	→ jmalonzo_ joined	19:58
	← janm left	19:59
	← zachinglis left	20:00
	← holizz left	20:02
pasky	are there any perl-based gitk-ish history browsers?	20:09
jettero	gitweb.cgi?	20:10
	oh, not really gitkish	20:10
pasky	that's not quite what I meant :)	20:10
	there's git-browser but it's more js-based than perl-based :)	20:10
	pasky is writing a specialized porcelain in perl and would like to embed some history browser there	20:10
	← jmalonzo left	20:11
	← shaftyy left	20:12
	← aroben left	20:13
jettero	what is a porcelain?	20:13
capisce	a song by red hot chili peppers	20:13
	← koke left	20:13
	→ aroben joined	20:14
Ilari	jettero: Refers to UI-grade interface to Git.	20:15
	jettero: As opposed to non-UI-grade stuff like rev-list, write-tree, update-index, commit-tree, update-ref, fetch-pack, etc...	20:16
	← drizzd left	20:17
	← loops left	20:19
	← aroben left	20:20
	→ aroben joined	20:20
	← dirk1296 left	20:21
	← warlock_mza left	20:24
	→ Ademan joined	20:24
	→ sgrimm joined	20:25
	← mattf left	20:29
	→ zachinglis joined	20:30
	← kumbayo left	20:31
GyrosGeier	hmm	20:32
	does git-imap-send support SSL these days?	20:32
	→ MrJag joined	20:33
	← zachinglis left	20:35
	← afton left	20:38
	→ david_koontz joined	20:38
Mikachu	that's a weird date	20:38
ToxicFrog	?	20:38
Mikachu	afton [n=chatzill@24.86.64.198] has quit ["ChatZilla 0.9.82.1 [Firefox 3.0b5/2008050509]"]	20:38
ToxicFrog	so it is.	20:39
Ilari	Maybe that last '09' is hour?	20:39
	→ zachinglis joined	20:40
	→ discomury joined	20:43
jast	very likely	20:43
	← MrJag left	20:47
	→ agib joined	20:48
	→ _zack joined	20:49
	→ madewokherd joined	20:50
	→ gmcinnes joined	20:51
	→ datafirm joined	20:52
datafirm	Hello!	20:52
jast	hi	20:52
datafirm	I had a folder in .gitignore which I've removed. Git still want let me add that folder, so I have to use -f. Is there a way to refresh what this local repo sees in my modified .gitignore?	20:53
jast	.gitignore is always used as-is	20:53
datafirm	hmm	20:54
jast	perhaps you've got another pattern that matches the same dir	20:55
datafirm	**/.DS_Store	20:55
	that the only wild glob in there.	20:55
	if one added data/projects/* to git ignore, would it all addition of data and projects, but not the files within projects?	20:56
	that is my goal.	20:56
jast	i don't think '**' is a valid glob pattern	20:56
	← aroben left	20:56
datafirm	jast: to recursively go into each dir for a file of that name ?	20:57
jast	that's a bashism, i believe	20:57
ToxicFrog	datafirm: I'm not sure what you mean by "projects, but not the files within projects"	20:57
	git doesn't track empty directories.	20:57
datafirm	I have folders with lots of app specific data that should not be stroed in the repo, but the folder structure should.	20:58
	I guess I'll leave it up to the app to create the folders if they dont exist then.	20:58
jast	well, put a .keep file in each of those dirs and add an ignore exception for it: !.keep	20:58
	← nessundorma left	20:59
	→ ggeecko joined	21:01
datafirm	jast: never heard of .keep.. i'll have to take a look in the docs.	21:01
jast	it's got no special meaning	21:02
	→ GilbertErik joined	21:02
datafirm	jast: ahh just to allow it to grab them	21:02
	in the repo	21:02
	← zachinglis left	21:02
ToxicFrog	Yeah	21:02
	so that it has a file in the dir, and thus versions it	21:02
jast	but you need at least one file inside a directory so git can track the directory, and .keep is a classic convention to indicate that an empty directory is actually needed for something	21:02
datafirm	ToxicFrog: good idea.. and that exclusion has to go in the directories .gitignore.. not the root?	21:02
	jast: great! thanks	21:03
jast	you can put the exclusion in the root .gitignore file	21:03
	patterns that contain no slashes are applied in all directories	21:03
datafirm	jast: Good to know, thanks	21:03
	jast: is the order important?	21:06
	→ wagle_home joined	21:06
jast	i don't know how that works with exclusions	21:06
	it may be	21:06
	← lcapitulino left	21:06
	← wagle_home left	21:06
	→ wagle_home joined	21:07
ftehw	Is there a simple method (maybe through branching) which will allow me to save uncomitted work temporarily while I work in another branch (e.g. need to fix a bug). I guess, something like a reset --hard, but which is later recoverable.	21:07
	← ksvladimir left	21:07
	← wagle_ left	21:08
ToxicFrog	ftehw: man git-stash	21:08
	It exists exactly for this purpose, and is tasty.	21:09
	← schapht left	21:09
	→ wagle joined	21:09
ftehw	ToxicFrog: perfect! thanks.	21:09
jast	with just a hint of pepper	21:09
	← shaftyy_ left	21:09
Eridius\|work	it saves working tree and index	21:09
	→ shaftyy joined	21:09
jast	and the world	21:10
ftehw	i was gonna say, the index constitues stuff i've added but not yet committed right? So if it was just the tree and index then it'd miss some thing i wanted it to grab.	21:10
jast	so, you know, alias stash='mplayer eurythmics-i-saved-the-world-today.ogg </dev/null &; git stash'	21:10
	ftehw l0lz	21:11
	← shaftyy left	21:11
jast	what's there to grab apart from the tree and the index?	21:12
	well, untracked files, but those are normally left alone when switching branches anyway	21:12
ftehw	so the stash interface is like a stack? stash; stash; stash; stash apply; stash apply; stash apply; works as you'd expect?	21:12
jast	no, apply always applies the topmost stash (but doesn't pop it)	21:12
RandalSchwartz	the stash is a guy thing, for people who fear commitment. :)	21:12
	if you want to do somethign complex, use real commits.	21:13
Eridius\|work	ftehw: replace "apply" with "pop"	21:13
ftehw	jast: echo "foo" > foo; git add foo; git commit; echo "bar" >> foo --- the mod "bar" isn't in the index, is it?	21:13
Eridius\|work	hey RandalSchwartz, what's it take to get a trivial documentation patch accepted?	21:13
ToxicFrog	RandalSchwartz: if you want nonworking commits lying all over the place, perhaps	21:13
RandalSchwartz	I've never made a trivial doc patch. :)	21:13
jast	yes, but it's in the tree, and that is stashed too	21:13
RandalSchwartz	who cares about nonworking commits?	21:14
ToxicFrog	ftehw: it's not in the index, but it's in the working copy, and stash will grab both	21:14
RandalSchwartz	commit means "made some edits, might want to come back to that"	21:14
	→ rubbish joined	21:14
RandalSchwartz	it doesn't mean "... and this works"	21:14
jast	Eridius\|work, sneak it into a patch that introduces the pack v3 format ;)	21:14
ftehw	ToxicFrog: w00t.	21:14
Eridius\|work	RandalSchwartz: I've sent http://marc.info/?l=git&m=120992103604463&w=2 twice, and had zero response. It's very annoying	21:14
	Eridius\|work smacks jast	21:14
RandalSchwartz	send it to junio	21:14
	in the form of a patch :)	21:14
Eridius\|work	I CC'd him in both	21:14
	and it is a patch	21:15
jast	without doubt	21:15
Mikachu	i wonder if filter-branch might consider children "ancestors" since it starts at branch tips and goes backwards? but maybe not	21:16
	i get a headache thinking about it :)	21:16
jast	no, it doesn't go backwards	21:16
	advany → zzzz	21:16
Mikachu	well, everything in git starts from children since only parent info is stored	21:16
jast	it couldn't, could it? modifying one commit means modifying all of its children	21:16
Eridius\|work	now that I think about it, "children" might be better than "descendants", but both are better than "ancestors"	21:16
jast	therefore you need to start with the oldest commit in the branch	21:17
Mikachu	hm, that's true also	21:17
Eridius\|work	it may follow tips back, but it starts rewriting at the parents	21:17
ToxicFrog	Eridius\|work: I don't know, "children" and "ancestors" make it clear which way time is flowing	21:17
Mikachu	even if ancestors is correct, "children" is much less confusing	21:17
	← zzzz left	21:17
Eridius\|work	except ancestors is not correct	21:17
	if it behaved as documented, it would create a cycle	21:17
jast	it starts at the tip and works backwards to find the first commit, that's true	21:17
ToxicFrog	Er, descendants and ancestors	21:17
jast	but the actual filtering process works forwards in time	21:17
ToxicFrog	children/parents is a symbol collision with the graph structure itself	21:18
Mikachu	the whole sentence is very confusing	21:18
	→ Crema10 joined	21:18
Eridius\|work	ToxicFrog: not really	21:18
ToxicFrog	At least for me	21:18
Mikachu	regardless of the number of ids you output, the original commits will not change at all	21:18
	only the newly written commits will be affected	21:18
jast	graph structure is what we're talking about here, anyway	21:18
ForsakenSoul	how can I make a branch of master?	21:18
jast	of course the original commits won't change. the branch will no longer point to them though.	21:18
	← nud left	21:18
RandalSchwartz	git checkout -b newbranch master	21:18
	← Crema10 left	21:19
Mikachu	doesn't it mean "the rewritten commit will have all of them as parents"?	21:20
	i don't see why changing the parents of a commit would change what parents its children has	21:20
Eridius\|work	I suppose you're right	21:21
	→ deavid joined	21:21
jast	i just read the sentence in its context and i have to agree with that	21:21
ftehw	confirmed: git stash rocks my mom --hard.	21:21
Mikachu	i really don't understand what it does very well	21:22
Eridius\|work	it saves the index into a new commit, and the working tree into another commit that has the index as parent. It points the stash ref at the working tree commit	21:22
	and any previous stashes get stored in the stash ref's reflog	21:22
Mikachu	sorry, i meant the git-filter-branch commit filter when you output more commit ids	21:23
Eridius\|work	oh	21:23
	it lets you create merges, basically	21:23
	huh, git-svn seems intent on re-creating my 'master' branch	21:24
Mikachu	oh.. so it does affect the children of the original commit then :)	21:24
jast	no, it affects the rewritten commit	21:24
	the children are affected in the usual indirect way	21:25
Mikachu	but it takes one commit as input and outputs several commits	21:25
	metze → metze_away	21:25
Eridius\|work	no	21:25
	the commit filter is basically just "call this tool instead of git-commit-tree"	21:25
Mikachu	but the sentence says specifically "may emit multiple commit ids"	21:25
Eridius\|work	oh wait, I'm still reading it wrong	21:26
	← ironfroggy left	21:26
Eridius\|work	yes, the tool may emit multiple ids. gimme a sec to make sure i have this straight ;)	21:26
jast	i don't really see the purpose of that extension	21:26
	i just re-read it too	21:26
Eridius\|work	I think it means the "rewritten children of the original commit will have all of them as parents"	21:26
ForsakenSoul	what was the command of merging two branches	21:27
Mikachu	"merge"	21:27
ForsakenSoul	one more thing ... the branch and the master should be in different dirs right ?	21:27
jast	how does that make sense, though? you don't even know which children you are monkey-patching into merge commits there.	21:28
	← chris2 left	21:28
ForsakenSoul	:D ... two very stupid questions	21:28
Eridius\|work	ForsakenSoul: no	21:28
	← datafirm left	21:28
jast	ForsakenSoul, what do you mean by different dirs? git can handle any number of branches in one repo and you can switch back and forth rapidly using git checkout.	21:28
ftehw	yes! /cut !	21:28
Eridius\|work	jast: well no, the tool won't know that, but presumably you as the invoker of git-filter-branch will know that	21:28
jast	but do i? i only have the contents of the current commit	21:28
Eridius\|work	ya know what, someone needs to check the source for --commit-filter	21:28
jast	i don't exactly know what children i'll get	21:29
	yeah	21:29
	i was hoping to avoid that :)	21:29
Eridius\|work	again, the tool won't know, but YOU the USER will know	21:29
	← Eludias left	21:29
	← felipec left	21:29
Mikachu	Eridius\|work: it's a shell script, so it's impossible to see what it does :)	21:29
Eridius\|work	haha	21:29
	Eridius\|work takes a gander	21:29
ToxicFrog	ForsakenSoul: typically, one of the branches is checked out and you know the name of the other	21:30
jettero	if I mkdir -p /something/ and GIT_DIR=/something/ git init, what else do I have to do to it to be able to push files to it	21:30
ToxicFrog	ForsakenSoul: So if, say, you want to merge "master" into "feature":	21:31
	git checkout feature	21:31
	git merge master	21:31
	and you're done.	21:31
	jettero: in the other repo you want to push from, "git remote add <name> <path to /something>"	21:32
ForsakenSoul	ok so how do I delete a branch	21:32
jast	ForsakenSoul, check out the manpage for git-branch. it's useful. :)	21:32
ToxicFrog	ForsakenSoul: locally: git branch -d name, and more generally, man git-branch	21:32
jast	git-branch always operates locally	21:33
ToxicFrog	Remotely: git push <remote> :<branch>	21:33
	jast: yes, I mean, "if you want to delete a local branch"	21:33
	As opposed to "if you want to delete a branch in some other repo you can push to"	21:33
jettero	ToxicFrog: yeah, I did that ... but when I try to push it says No refs in common and none specified; doing nothing.	21:33
jast	right	21:33
	← juanjoc left	21:34
Mikachu	jettero: so then you didn't get the idea to specify some refs?	21:34
jast	jettero, when you push for the first time you need to specify which branches to push (try 'master' or, if you like, '--all')	21:34
ToxicFrog	jettero: try "git push <name of remote> <name of branch>", eg, "git push origin master"	21:34
ForsakenSoul	ToxicFrog .. jast ... wow guys ... you're taking this way too much passionate :D	21:34
Eridius\|work	Mikachu: ok, it's what I thought. If you emit 2 commit ids, any children get both commit ids	21:34
Mikachu	Eridius\|work: as parents?	21:34
Eridius\|work	basically, any time you map the original commit, you get both emitted ids back	21:34
	yes	21:34
Mikachu	ah	21:34
jettero	ToxicFrog: oic, thanks	21:34
jast	that's... yucky.	21:34
Eridius\|work	this seems like esoteric functionality, but presumably someone wanted it	21:34
jast	perhaps your doc patch should note that it's probably not what the user wants :)	21:35
Mikachu	couldn't you do it with --parent-filter?	21:35
	→ datafirm joined	21:35
Eridius\|work	Mikachu: but what if you're already doing other stuff with the commit filter? parent-filter won't know about that stuff	21:35
ForsakenSoul	but thank you guys	21:35
Mikachu	or hm, how much fun things can you do turning one commits into two and then the next one is a merge commit again? i don't really see the point :)	21:36
jast	you can always change the list of parents of the current commit within the commit filter	21:36
datafirm	have any of you seen any post-commit hooks that will send an HTML diff as email? I had something similar for SVN which I liked!	21:36
Mikachu	datafirm: if you have one for svn and one that generates plain diffs for git, presumably you should be able to combine them :)	21:36
Eridius\|work	jast: sure, because you're creating the actual commit object, you can do whatever the hell you want	21:37
datafirm	Mikachu: Yeah, perhaps.. its all in perl and i've not wrote perl in 8+ years.. :-D	21:37
	With that said, is there a way to place files in the post commit for all new repos? Maybe a sym link?	21:37
Mikachu	datafirm: locally, /somewhere/git/templates/hooks/*	21:38
	git could be git-core too	21:38
jast	where /somewhere possibly means /usr/share	21:38
Mikachu	i think it depends on distro, /usr/share/git-core is a good first guess	21:38
datafirm	jast: in other words.. there is a place for these template git will look for :-D	21:38
jast	correct	21:38
ToxicFrog	locate 'git/templates/' should find it	21:39
jast	you can even configure where git will look for the templates, i think	21:39
	hmm, apparently not via git config	21:39
datafirm	FreeBSD /usr/local/share/git-core/templates	21:39
	← gmcinnes left	21:40
	← LotR left	21:43
	→ pdherbemont joined	21:43
	← pdherbemont_ left	21:44
Eridius\|work	ok, patch sent	21:45
	→ drizzd joined	21:46
	→ hipertracker_ joined	21:46
	→ gmcinnes joined	21:47
	← _zack left	21:49
	→ d2dchat_ joined	21:53
	→ mvanholstyn joined	21:53
	→ _zack joined	21:53
	→ FunkeeMonk joined	21:54
	← vbabiy-laptop left	21:54
	→ pdherbemont_ joined	21:54
	← d2dchat_ left	21:55
	→ d2dchat joined	21:57
	→ juanjoc joined	21:59
	→ gitte joined	22:00
	← threeve left	22:00
	← juanjoc left	22:00
	← hipertracker left	22:01
	← pdherbemont left	22:02
	← mvanholstyn left	22:07
	→ spearce joined	22:09
	← jmalonzo_ left	22:10
	← david_koontz left	22:12
	← ph^ left	22:12
	← cedricv left	22:12
	← EmilMedve left	22:13
	→ david_koontz joined	22:13
	← ceez left	22:15
	← jlh left	22:18
	→ jlh_ joined	22:18
	jlh_ → jlh	22:18
	← charon left	22:19
	→ cedricv joined	22:20
	→ charon joined	22:21
	← cedricv left	22:21
	→ d2dchat_ joined	22:22
	jlh → jlh_	22:23
	jlh_ → jlh	22:23
	← d2dchat left	22:23
	→ jamessharpe joined	22:28
	→ redondos_ joined	22:29
	→ d2dchat__ joined	22:29
	→ Michelangelo joined	22:30
	← d2dchat__ left	22:31
	→ d2dchat__ joined	22:31
	← cmarcelo left	22:32
	← _zack left	22:33
	→ koke joined	22:36
	← redondos left	22:41
	redondos_ → redondos	22:41
	← VitaRara left	22:41
	← d2dchat__ left	22:43
	→ capisce_ joined	22:45
	← FunkeeMonk left	22:47
	← ekidd left	22:47
	← d2dchat_ left	22:48
	← jbunster left	22:48
	→ radarek joined	22:49
	hipertracker_ → hipertracker	22:51
	← EiNZTEiN left	22:52
	← DrNick left	22:52
	→ DrNick joined	22:52
	→ manveru joined	22:54
	← GilbertErik left	22:54
	← josb left	22:54
	← capisce left	22:54
	← david_koontz left	22:56
	← cannonball left	23:01
	← Ryback_ left	23:04
	← lea_w left	23:04
agib	Can anyone help me out? I was working on a branch "new_design" in my rails project and I decided it was time to use submodules for my rspec plugins... I switched to my master branch and deleted my rspec and rspec_on_rails directories... then I did a git submodule add and checked in the new additions... however now when I try to switch back to my new_design branch I get: "fatal: Untracked...	23:05
	...working tree file 'vendor/plugins/rspec/CHANGES' would be overwritten by merge." what's that about?	23:05
	← koke left	23:07
	← radarek left	23:07
	← rubbish left	23:09
	← gitte left	23:09
	→ koke joined	23:09
	← strangy left	23:09
	→ SuttoL joined	23:10
	→ flaguy joined	23:10
	← gmcinnes left	23:14
	← thiago_home left	23:14
	→ pombreda joined	23:15
	← Michelangelo left	23:16
agib	anyone?	23:16
	→ strangy joined	23:16
	← koke left	23:16
vmiklos	that means a merge would create such a file but you already has an untracked file with the same name in your working directory	23:16
	→ joevandyk joined	23:17
	→ capisce joined	23:18
agib	vmiklos: hmm I don't know how that file got there...	23:18
	weird	23:18
	an untracked file... I just did a git submodule add git://github.com/dchelimsky/rspec-rails.git vendor/plugins/rspec_on_rails etc	23:18
	← ckoehler left	23:20
	→ lea_w joined	23:25
	→ ekidd joined	23:25
	← deavid left	23:27
	← _graham_ left	23:29
	← capisce_ left	23:31
	← jpeeler left	23:33
	→ capisce_ joined	23:33
	→ mattf joined	23:37
	← wagle_home left	23:43
	← lea_w left	23:45
	← wurble left	23:46
	→ tongueroo joined	23:47
tongueroo	how do you reset changes that are not in the index but keep the changes that are...	23:47
	← brosner left	23:48
	← capisce left	23:48
Mikachu	tongueroo: you want to make the working tree match the index, in other words?	23:49
tongueroo	yup	23:49
	:)	23:49
Mikachu	git checkout -- .	23:49
tongueroo	nice :)	23:49
Mikachu	not so easy to guess maybe :)	23:49
tongueroo	just what ive been looking for, faster than what ive been doing, git ci -m 'tmp to save index' && git co -f	23:49
	git reset HEAD^	23:50
	one step	23:50
	← charon left	23:51
	→ igorgue joined	23:53
	→ gmcinnes joined	23:54
	← gmcinnes left	23:54
	→ gmcinnes joined	23:55
	← ekidd left	23:58
	→ ben_h joined	23:58
	← p4tux left	23:58

Logs Search ←Prev date Next date→

Channels Documentation