IRCloggy #git 2009-02-13

Provider of IRC logs since 2005.

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

	2009-02-13
	→ mattenoble_ joined	00:00
	← LiamH left	00:01
	← mattenoble_ left	00:02
	← lu_zero left	00:03
	→ program247365 joined	00:04
	← program247365 left	00:04
	← manithree left	00:04
brink	If I use git-filter-branch to permanently remove a directory can I still pull from the original repository?	00:05
DrNick	you'll get that directory back	00:05
	and the merging will be unpleasant	00:06
	→ sporkmonger joined	00:06
	→ faithfulgeek joined	00:07
	→ icwiener joined	00:07
brink	I just notice that if I delete a directory with git-filter-branch I don't save any space even after git gc. Is there something else I need to do to remove all history of a directory?	00:08
	← jackdempsey left	00:08
	← plediii left	00:09
gitte	brink: it's all in the reflogs.	00:09
	brink: just make a (temporary) clone to see the size you'll get in due time.	00:09
	← jstad left	00:10
brink	gitte: The clone seems to be the same size. Do I need to repack it or something?	00:11
gitte	brink: try it.	00:12
brink	gitte: I just did. No luck. All those binaries must be hidden somewhere.	00:12
gitte	brink: generally, I find that many people are most concerned with the size of the repository after filter-branch, but I'd rather try to make sure that all worked out as expected.	00:13
	← scientes left	00:13
	← shad left	00:13
	← eddyp left	00:14
	→ eddyp joined	00:14
brink	gitte: The directories are gone, but the size remains.	00:14
gitte	brink: I was not talking about the working tree when I mentioned the desire for verification.	00:15
	→ shad joined	00:15
brink	gitte: Ok. what should I verify?	00:16
gitte	brink: you should know what you wanted to achieve with the filter-branch call yourself :-)	00:17
	→ dmlloyd_ joined	00:17
	← radarek left	00:17
brink	I wanted it to delete all history of a directory. That way all the binary data that shouldn't have been checked in in first place would be removed.	00:17
	gitte: Do I misunderstand filter-branch?	00:18
gitte	brink: no, that's what it should be able to do.	00:18
	brink: now it is your job to find out if there are references left to those blobs, e.g. by calling "git rev-list --all --objects" and looking for the filenames.	00:19
brink	gitte: I see. It looks like they used to have different names. So they'll be difficult to chase down I assume.	00:22
	← Sigma left	00:22
	← dmlloyd left	00:23
	dmlloyd_ → dmlloyd	00:23
gitte	brink: See? Much more sense to verify what was achieved than to look at the size... :-)	00:23
brink	gitte: Yes. Unfortunately, it looks like it will be difficult to track down all the unecessary binaries because they've been moved all over the place.	00:24
gitte	brink: maybe you want to think about better ways to identify what blobs to remove than the absolute path?	00:25
brink	gitte: Can they be identified by some other id? They never change their contents.	00:26
	← icwiener left	00:27
gitte	brink: if you know the exact contents, you might want to identify them by their hash.	00:28
	elg wonders why vimdiff is preferred over opendiff (osx) when vimdiff is so pitiful for 3-way merges	00:28
gitte	brink: other than that, you could make a tree-filter (expensive; you want to use a RAM disk for that) that searches for files larger than a certain threshold, and delete them.	00:29
	brink: you could try to find out what directory names they had, or extension, or if there is another tell tale.	00:29
brink	gitte: I suppose that wouldn't be hard for an individual file. There are a number of them though.	00:30
	gitte: Assuming I hashed them is there a way to remove all history of a file using its hash?	00:30
gitte	brink: I'd probably use an index filter for that.	00:30
	← RoPP left	00:33
	→ ideogon_ joined	00:33
	← sporkmonger left	00:33
	← boto_ left	00:34
	← Phlogi left	00:34
	→ ideogon__ joined	00:36
	dwmw2 → dwmw2_gone	00:39
	→ kennethlove joined	00:40
	→ stathis_ joined	00:42
	← shawn-p left	00:43
	← stathis_ left	00:44
	→ maali joined	00:44
	← ideogon left	00:47
	← cgardner left	00:48
	→ sporkmonger joined	00:49
	← bradly left	00:49
	← eean left	00:49
	← sporkmonger left	00:49
	→ eean joined	00:50
brink	gitte: I suppose I could look up the commits and skip those. Does that make sense?	00:50
	← bdiego left	00:50
gitte	brink: not really, as the commits are not the patches.	00:50
	brink: commits in Git are snapshots, so if you skip a few commits in filter-branch, but then take some later commit, the files come back almost magically.	00:51
	brink: IOW filter-branch != rebase --interactive	00:51
	← ideogon_ left	00:52
	← faithfulgeek left	00:54
	→ a-priori joined	00:57
	← elg left	00:58
	← a-priori left	01:00
	← shruggar left	01:00
mutex	so if I want to do a test integration between branch A and B	01:02
	can I do it on a new branch C, that is a child of either A or B ?	01:02
	→ a-priori joined	01:03
	→ killerchicken joined	01:03
killerchicken	hey, I just used cherry-pick to pick a patch. It wouldn't apply cleanly. How can I undo everything to be the way before I picked a patch?	01:04
parasti	killerchicken: assuming you didn't have any uncommitted changes, git reset --hard should do	01:05
killerchicken	no, I didn't	01:06
	I was wondering if there was anything specific like rebase --abort or something	01:06
gitte	mutex: you can use a throwaway branch C, or you can detach the HEAD, or you can do it on branch A and reset --hard to the pre-merge state afterwards.	01:06
	mutex: a matter of taste, really.	01:06
	killerchicken: git reset --hard	01:07
killerchicken	gitte: :)	01:07
	← jboyens left	01:07
nomis	is it possible to specify a fake date when committing stuff?	01:08
gitte	nomis: yep.	01:09
nomis	how?	01:09
gitte	nomis: set "GIT_AUTHOR_DATE" and/or "GIT_COMMITTER_DATE"	01:09
nomis	gitte: ah. Thanks.	01:09
gitte	nomis: or something similar... my memory could fail me...	01:09
mutex	gitte: then just merge my integrated branch C at the end to whatever parent I want ?	01:10
	nomis ponders adding ancient history to a project.	01:10
	→ ben_h joined	01:11
	← chris2 left	01:12
gitte	mutex: maybe, depends what you really want (i.e. when you want to have your real merge, and between what)	01:12
	← plediii_ left	01:12
	→ pointlessjon joined	01:12
mutex	maybe I'll just go for it	01:12
	see if it blows up	01:12
	mutex runs to go play with dynamite	01:12
	→ faithfulgeek joined	01:13
gitte	Boom	01:14
	← calin left	01:16
mutex	well that was painless	01:16
	1 line merge conflict	01:16
	→ sporkmonger joined	01:16
gitte	mutex: and the best thing: if you activate rerere before you get the conflict, its resolution will be recorded for next time...	01:17
mutex	yeah I haven't played with rerere yet...	01:17
gitte	I have.	01:17
	mutex gasps	01:18
mutex	I just hear about how neat it is	01:18
	stories really	01:18
	→ toxx_ joined	01:18
gitte	The only thing that is not so brilliant is that you sometimes ask yourself "where are the conflicts?" until you realize that they have been resolved already by rerere.	01:18
mutex	I guess you just run it, and everything happens automagically	01:18
benn_d	fatal: failed to write object ... error: unpack failed: unpacker exited with error code	01:19
	crap I'm getting a weird error:	01:19
gitte	Well, you still have to add the "unmerged" files after checking that everything went fine.	01:19
mutex	ah	01:19
gitte	benn_d: disk full? lack of write permission?	01:19
	In any case, my body just threw an "OutOfEnergy" exception.	01:20
	git night!	01:20
	← gitte left	01:20
	← ekidd left	01:20
benn_d	yeah ,was a permissions error.. odd	01:21
	← faithfulgeek left	01:22
	← Yuuhi left	01:23
mutex	problem with exceptions is sometimes they go uncaught....	01:24
	← schierbeck left	01:25
	← toxx left	01:26
	← noth left	01:27
	← toxx_ left	01:28
	→ noth joined	01:28
	→ EmilMedve joined	01:28
	← sdboyer-laptop left	01:28
	← dwave left	01:29
	← kennethlove left	01:29
	→ ajsharp joined	01:29
	← noth left	01:29
	→ noth joined	01:30
	← a-priori left	01:35
	→ hobodave_ joined	01:37
	← mcella left	01:37
	← hobodave left	01:40
	→ offcent joined	01:42
	← hobodave_ left	01:43
	← offcent left	01:43
fynn	git git git	01:44
	→ wallerdev joined	01:47
wallerdev	yo	01:48
	so i pushed a branch up to a server (C) from computer A, how do i pull that branch down from another computer (B)	01:48
	→ gotnone joined	01:49
wallerdev	i pushed it with [1,2,3,4].any?{ \|e\| puts e ; e }	01:49
	err	01:49
	i pushed it with git push origin beta:refs/head/beta	01:50
	→ elg joined	01:50
gotnone	I have been having a hard time searching for the commands which will revert my local git tree the remote server. Specifically, I want to remove any extra branches and files I may have inadvertently added. I must have poor searchfu.	01:51
	→ lu_zero joined	01:53
PerlJam	gotnone: git help branch	01:53
	→ Naib_ joined	01:56
	← ajsharp left	01:58
	→ TML joined	01:58
	← lois[pac] left	01:58
	← priidu left	01:59
	← joshdavey left	01:59
	→ faithfulgeek joined	01:59
TML	What can cause "! [remote rejected] master -> master (n/a (unpacker error))"? I've been Googling for a while now, and am not finding a lot of useful info.	01:59
gotnone	thanks PerlJam I think I figured out what I needed to do	02:00
	← Naib left	02:02
	← pointlessjon left	02:03
	← hardcorelamer left	02:04
	→ a-priori joined	02:04
gwoo	TML: you are not updated?	02:06
wallerdev	oh god	02:06
	lol my git repo is probably so messy now, but i got it	02:06
TML	gwoo: git pull	02:06
	Already up-to-date.	02:06
gwoo	TML: maybe the remote is corrupted?	02:07
	→ eletuchy joined	02:11
	← faithfulgeek left	02:14
	→ zeahok joined	02:14
	← wallerdev left	02:16
TML	gwoo: It's a bare repo, this is the first time I've checked anything out of it. I made a one line change, and am now trying to push that change back to the bare repo.	02:17
gwoo	oh weird	02:17
	→ plediii joined	02:18
	← edwardam left	02:19
TML	Just to check, I just nuked the repo and the clone, set them back up again, and get the same error	02:19
gwoo	try git-fsck --full	02:20
TML	http://pastebin.ca/1335604	02:20
	gwoo: On the bare?	02:20
gwoo	yeah if you can	02:21
TML	Tried it on both ends, to no avail	02:21
gwoo	hmm	02:21
TML	There was no output 'git fsck'	02:21
	→ tjafk2 joined	02:21
TML	s/output/output from/	02:21
gwoo	http://kerneltrap.org/mailarchive/git/2007/11/9/399929	02:22
	i found this thread	02:22
TML	Thanks, I'll take a look at thsoe	02:22
	those, even	02:22
gwoo	git repack remote/origin/master	02:23
	that might help too	02:23
	→ loincloth joined	02:27
TML	Ah, damnit. PERMISSIONS!	02:27
	That's TWICE today.	02:27
	gwoo: Thanks. When I was deleting the repo, I was forgetting to chown the files over to the user running git-daemon.	02:27
	And I just can't seem to find anywhere that it logs an error on that	02:27
	→ theCarpenter joined	02:28
	→ faithfulgeek joined	02:30
	← gaveen left	02:34
	← faithfulgeek left	02:34
	← tjafk1 left	02:34
	→ kapone joined	02:36
	← awarrington__ left	02:36
	← eddyp left	02:36
	→ awarrington__ joined	02:37
	← _slacker__ left	02:39
	→ faithfulgeek joined	02:39
	← amfwrk left	02:40
	→ loinclot_ joined	02:40
	→ amfwrk joined	02:40
	→ loinclo__ joined	02:43
	← kapone left	02:43
	← loincloth left	02:44
	→ hobodave joined	02:44
	← Tv left	02:45
	← ice799 left	02:46
	← kapone_ left	02:48
	← zeahok left	02:48
gwoo	TML: funny that was my first guess	02:48
	→ dj_ryan joined	02:48
	← loinclo__ left	02:48
gwoo	but i did not say it, damn	02:48
dj_ryan	so i have a question	02:49
	→ gaveen joined	02:49
dj_ryan	is there a way to reset a github remote-branch	02:49
	i had a bunch of erronious commits due to ... learning exactly how git push works	02:49
	→ plediii_ joined	02:52
	→ zaius joined	02:52
zaius	is there a way to split line by line on an interactive add?	02:52
	i can only get it down to paragraph sized blocks	02:53
	→ loincloth joined	02:53
	← EmilMedve left	02:54
	← loinclot_ left	02:57
	→ EmilMedve joined	02:58
	← Flathead left	02:59
mugwump	zaius: use git-gui	03:02
	→ loinclot_ joined	03:02
mugwump	it can add individual lines	03:02
	← servilio left	03:02
plediii_	what is better than using git to do backups with full history?	03:03
	→ tro joined	03:04
	← loincloth left	03:05
mugwump	plediii_: tape? zfs?	03:05
	mugwump scrams &	03:05
	→ icwiener joined	03:06
plediii_	mugwump: what software for doing tape backup?	03:06
mugwump	real sysadmins use dump	03:07
	← Cacheaway left	03:08
	→ bdiego joined	03:08
zaius	mugwump: i'm only on a console	03:09
plediii_	mugwump: can dump let me restore backup from any point in the history?	03:09
zaius	looks like i can use the editor mode if i upgrade to 1.6	03:09
	anyone know an easy way to upgrade to 1.6 on ubuntu 8.10?	03:09
	← plediii left	03:11
	← spearce left	03:11
plediii_	mugwump: the only problem I see with git is the pain involved if what you're backing up contains git repositories	03:11
	and rsync's style of renaming changed files is just lame	03:12
	→ jony joined	03:15
plediii_	of course other people have already thought of it :-\|	03:17
	← loinclot_ left	03:18
	← ideogon__ left	03:20
	→ cytrinox_ joined	03:20
plediii_	here's what I'm using to to backup my home directory on a remote host. http://rafb.net/p/TX3E6k90.html	03:22
	Anyone want to tell me why this is stupid?	03:23
	→ AngryElf1 joined	03:23
	← AngryElf1 left	03:24
	← elg left	03:24
	← AngryElf left	03:24
	← AngryElf_ left	03:25
	→ AngryElf joined	03:25
MarcWeber	1234475693 +0100 Is this a timestamp? (found in .git/logs/refs/heads/branch	03:25
	→ wallerdev joined	03:25
offby1	plediii_: seems complex ... I can't tell by looking if it'll actually work	03:25
plediii_	offby1: it seems to work	03:26
	although the business with find could be simplified somehow	03:26
	i'm sure	03:26
offby1	plediii_: I've got a ton of backup links at http://delicious.com/tucumcari/backup	03:26
plediii_	i feel like i must be missing something obvious though	03:26
wallerdev	okay i seem to have a problemo, i accidentally created a branch refs/head/beta instead of refs/heads/beta, and now i can't figure out how to remove it	03:27
	← Kbyte left	03:27
wallerdev	i tried this but got: git push origin refs/head/beta	03:27
	error: src refspec refs/head/beta does not match any.	03:27
offby1	I suspect it'd be ok to just delete that file	03:28
	offby1 glances around nervously.	03:28
offby1	how'd you create it?	03:28
plediii_	offby1: what do you actually use?	03:29
offby1	"tarsnap" and "duplicity".	03:29
ben_h	strange - i just cloned linus' linux-2.6 tree to poke around, and straight after the clone there are about 20 changed-but-not-updated files.	03:29
offby1	But people's backup needs vary greatly	03:29
	ben_h: are you using Windows?	03:29
wallerdev	git push origin refs/head/beta	03:29
ben_h	no, Mac OS X	03:29
wallerdev	err	03:29
ben_h	with nontrivial diffs in a bunch of files.	03:29
	sensible-looking diffs that is	03:29
offby1	ben_h: weird.	03:29
wallerdev	git push --force origin origin:refs/head/beta is how i created it	03:30
ben_h	obvious code updates - spelling fixes, cleanups, etc	03:30
	← gaveen left	03:30
offby1	wallerdev: I think you can nuke it with something like "git push origin :refs/head/beta"	03:30
	i.e., if your refspec has nothing before the :, that means "nuke this branch"	03:31
wallerdev	thanks offby1	03:32
	worked	03:32
	needed the colon i guess	03:32
offby1	ha!	03:32
	I will soon earn my Yellow Belt in git-answering	03:32
	doener is my mentor	03:32
dj_ryan	sweet	03:32
	is there a way to reset a github remote-branch	03:33
offby1	you mean, make it point someplace else?	03:33
dj_ryan	almost, iw ant it to point to a earlier time	03:33
	there is a pile of cruft commits that were accidently pushed there	03:33
	the commits also got pushed to a different branch	03:33
offby1	I would bet that you could delete it (using the recipe I just gave above to wallerdev) and then recreate it. There's probably a better way, but ...	03:33
dj_ryan	the "theory" was to branch from a point, then later diff to produce a patch for upstream	03:34
	→ utnick joined	03:34
plediii_	offby1: thank you	03:34
ben_h	here's a log of the unexpected local changes: http://gist.github.com/63027	03:34
offby1	ben_h: might your "linux" directory not have been empty before you cloned?	03:35
	only explanation I can think of ...	03:35
ben_h	nah - if you look higher in the history, i created the directory just beforehand	03:36
	got me stumped	03:36
	→ Flathead joined	03:36
ben_h	and check the 'git clone' output - ~/projects/linux/linux-2.6 was definitely created by git like normal, so it should be clean	03:37
	← cytrinox left	03:37
offby1	crazy	03:37
	sounds like that's worth a post to the mailing list.	03:37
	fo' sho'	03:37
ben_h	mm good idea	03:37
	→ lobolife joined	03:42
	→ bronson joined	03:43
	→ elg joined	03:44
	← kukks left	03:45
	← lobolife left	03:48
	→ mlh joined	03:48
mlh	I have a number of file heirachies around that re mostly the same. Can I git init and add one of them then copy the .git to all the others to find the differences?	03:49
	It would beat gdiff (doesn't network) and rsync -nvv (doesn't show diffs)	03:50
	I know there is _some_ metadata specific to a particular clone, but how important or mangleable is that metadata	03:51
MarcWeber	yes you can. the .git folder contains everything you nded. Hwoeve ryou can also use diff -r $pathA $pathB then	03:55
mlh	diff doesn't network	03:55
	MarcWeber: what about the metadata	03:55
MarcWeber	mlh I don't get the problem. You mean copying the .git folder is faster then mounting and comparing using sshfs or such?	03:56
	Which metadata.	03:56
	← a-priori left	03:57
mlh	.git/config will be different if you do a clone rather than do a copy (rsync)	03:57
	the url in particular	03:58
	← drewr left	03:58
MarcWeber	mlh: But all you are interested in is git diff , aren't you?	03:58
mlh	well ...	03:58
	at first	03:58
MarcWeber	You an tweak the .git/config manually as well if you konw what you're doing.	03:59
mlh	but I'd like to use them as real git repos if I could. with commits and merges and whatever	03:59
MarcWeber	You cann still add remote location create branches etc	03:59
mlh	yeah it's the if you konw what you're doing" bit that worries me	03:59
	Is there any docs on that stuff?	03:59
MarcWeber	Maybe don't diff but jsut use git checkout -b 'new branch'; git commit -am 'initial diff' then	03:59
	mlh: What kind of stuff. That's normal git usage. Which knowledge are you lacking?	04:00
mlh	the unknown unknowns :-)	04:00
	e.g. I wasn't sure that I could do a commit and have it make sense to git.	04:01
MarcWeber	I think man git-clone does contain the info that you can just copy a git repo instead of using git clone. everything but .git is a working copy. Thus you it doesn' matter wether you move .git or modify the working copy files	04:01
Gitbot	MarcWeber: the git-clone manpage can be found at http://git.or.cz/man/git-clone	04:01
mlh	BUt from what you've written about it looks like I can	04:01
	"it doesn't matter"	04:01
MarcWeber	mlh: If you're still unsure you can make small test. Try your use case with 2 files in /tmp	04:01
mlh	so why is the url different.	04:01
MarcWeber	the url in .git/config ?	04:01
mlh	e.g. if I clone it and then do a git ftech in the clone I guess it pulls from the cloned from repo	04:02
MarcWeber	the remote url of origin?	04:02
mlh	whereas if I copied it, the fetch would pull from the common upstream	04:02
	← aziz left	04:02
MarcWeber	Cause git clone always sets the remote url of default remote location called "origin" to teh repo you cloned from	04:02
mlh	yep. git clone does. and cp -a doesn't.	04:02
	so that's one difference.	04:03
	← elg left	04:03
MarcWeber	mlh: Just rewrite the config so that it fits your needs. There are more minor differences. git clone might create hard hardlinkes (see man page). But they shouldn't affect the final behaviour	04:03
mlh	ok thanks.	04:03
	← Bass10 left	04:03
	→ JohnFlux joined	04:04
MarcWeber	mlh: You can then just do git remote rm origin; git remote add origin $newurl and you're done. If you're really unsure start making a backup of those dirs. What can go wrong then?	04:04
mlh	I think it would be a nice enhancement to git so that it keeps such differing metadata outside of .git --- say in $HOME/.git-data or somesuch	04:04
JohnFlux	Are there any git developers in here?	04:04
mlh	(I know that would introduce other issues though)	04:04
JohnFlux	I want to try to persuade someone to make the error messages more helpful :)	04:04
MarcWeber	mlh: Which kind of meta data? If you have to put such kind of questions you've done above you should read more about git before putting such a request :-)	04:05
	JohnFlux: The mailinglist could be a good place to start. Which one is bothering you (I'm not a developer though)	04:05
JohnFlux	MarcWeber: pretty much all of them :-) I think it would help if they listed options to resolve the error that they give	04:06
mlh	MarcWeber: ok. I did look. But couldn't find a canonical categorical list with specificed semantics. Not that I was really expecting such.	04:06
	JUst a vague "should be ok" is good enough for now	04:06
MarcWeber	mlh: Thats what I was telling you all the time :-) I told you go on go on..	04:07
	cpg → cpg\|dinner\|brb	04:07
mlh	yep. is good.	04:07
MarcWeber	mlh: semantics of the git config? Try man git-config	04:07
Gitbot	mlh: the git-config manpage can be found at http://git.or.cz/man/git-config	04:07
JohnFlux	MarcWeber: 'git checkout master' 'You have local changes so cannot switch branches. Maybe commit your changes: git commit -a or Delete your changes: git rebase --hard"	04:07
	MarcWeber: an error message like that would be good :) Listing the options that I have to resolve each problem	04:08
MarcWeber	JohnFlux s/rebase/reset :-)	04:08
	JohnFlux: Do you know how to use grep ? Then you could replace those messages by better once and send the suggestions to the mailinglist.	04:08
JohnFlux	MarcWeber: see, even more evidence that the error messages should be useful :)	04:08
MarcWeber	JohnFlux: In this use case you should mention git stash as well	04:08
JohnFlux	MarcWeber: what does that do? :)	04:09
mlh	looks like git-config is about ~/.git-config not somerepo/.git/config	04:09
MarcWeber	git stash = save changes to the git internal memory place "stash" (which is kind of special commit) which you can restore by git apply later on any other commit	04:09
mlh	er cancel that	04:10
	← theCarpenter left	04:10
MarcWeber	mlh: Both are the same. you can set defaults int ~/.git-config and overwrite them in .git/config.	04:10
	done :)	04:10
	→ bdimcheff joined	04:13
	→ cemerick joined	04:13
mlh	MarcWeber: one thing I'd really like to know but can't really test for is whether .git/... would ever have a self-reference	04:14
	← markelikalderon left	04:14
MarcWeber	What do you mean by self-reference?	04:15
	← reprore__ left	04:15
	→ moconnor joined	04:15
	→ ekidd joined	04:17
TML	.git/hooks/post-update doesn't seem to execute when I do "git commit". I've checked that it has the execute bit set, what else do I need to do?	04:18
	→ jstad joined	04:19
	→ Tv joined	04:19
MarcWeber	TML try strace -o /tmp/out -f git commit ... Then you'll see that post-update isn't read at all. you should use the post-commit hook for that	04:21
halfline	TML: i once ran into a server where the public_html directory was on a home dir that was mounted noexec. the way I got around it was rm post-update; ln -s /usr/bin/git-update-server-info post-update	04:21
	← bdiego left	04:21
halfline	grr didn't see the "git commit" part	04:21
TML	MarcWeber: hah..right	04:21
	post-update needs to be on the bare repo	04:22
	Got it exactly backwards. Thanks.	04:22
	← wallerdev left	04:25
	← faithfulgeek left	04:28
johnw	i want to sync up a directory with files from my index. I know that checkout-index can update the changed and new files, but how do I delete the files in this mirror which have only been removed via rm --cache?	04:28
	git clean?	04:29
	→ h0nk joined	04:29
Tv	johnw: reset --hard?	04:29
johnw	oh, no, that needs a .git dir in that directory	04:29
Tv	nothing needs .git in a specific directory, --git-dir is for that	04:29
johnw	i'm trying to maintain an external DIR as an "image" of my cache	04:29
	but I don't want to rm -fr and recreate it every time	04:29
Tv	why not just have a clone, then?	04:30
johnw	i'm doing this to make a better pre-commit hook	04:30
Tv	local clones are cheap	04:30
johnw	because I can't push an index	04:30
	checkout-index won't remove files	04:31
	i guess i can parse the output from git clean -n to find out what should be removed	04:32
	→ samsoffes joined	04:33
	← softdrink left	04:33
	→ Hydrogen joined	04:33
	← warthog9 left	04:34
Tv	johnw: umm, why isn't reset --hard what you want?	04:34
johnw	because I have no .git directory in this copy	04:34
	→ warthog9 joined	04:34
Tv	johnw: --git-dir=	04:34
	→ a-priori joined	04:34
johnw	you're not understanding what I want to do	04:34
Tv	johnw: just make sure you use a different index file	04:34
johnw	i need a directory B to be an updated copy of my index, nothing else	04:35
	→ girishr joined	04:39
Tv	johnw: rm -rf a tmp && mkdir a tmp && ( cd a && git init && echo foo >bar && git add bar && git commit -m first; ) && ( cd tmp && GIT_INDEX_FILE=.index git --git-dir=../a/.git reset --hard; ls; ) && ( cd a && git mv bar baz && git commit -m second; ) && ( cd tmp && GIT_INDEX_FILE=.index git --git-dir=../a/.git reset --hard; ls; )	04:40
	→ muthu_ joined	04:40
johnw	doesn't reset --hard reset to HEAD?	04:41
Tv	yes?	04:41
johnw	i need the index	04:41
	not HEAD	04:41
Tv	then you need plumbing, not porcelain..	04:41
johnw	yeah	04:41
	← sporkmonger left	04:41
johnw	this is what I have so far: http://gist.github.com/63050	04:42
Tv	checkout-index	04:42
	not sure if that'll actually remove rules	04:42
johnw	rules?	04:42
Tv	sorry files	04:43
	brain fried	04:43
johnw	oh, oops	04:43
	it's going to remove them in my working tree!	04:43
	jstad → jstad\|away	04:43
Tv	johnw: your clean is misguided, at least	04:43
johnw	i just need to pipe to (cd $TMPDIR; xargs rm) instead	04:44
Tv	johnw: cd $TMPDIR && git --git-dir= might work better	04:44
	or --work-tree	04:44
	or just use ls-files \| xargs	04:44
	don't pipe clean output	04:44
johnw	why not?	04:44
Tv	don't use porcelain programmatically, go for plumbing instead	04:44
	plumbing is api, porcelain is human-friendly	04:44
	ls-files gives you same in guaranteed output format, with \0 termination and all	04:45
johnw	so, how do I list files registered for deletion in the index only?	04:46
	ls-files -d isn't it	04:46
	jstad\|away → jstad	04:46
johnw	oh, -t	04:46
Tv	there's no such thing as registered for deletion in index ;)	04:46
johnw	deleted in the index	04:46
	then	04:46
Tv	i think you want -o	04:47
	there's no deleted in index	04:47
johnw	-o shows all non-index files	04:47
Tv	that's just "no entry in index"	04:47
johnw	i did "git rm --cached foo"	04:47
	now I want a list that shows just "foo"	04:47
	cpg\|dinner\|brb → cpg	04:47
Tv	rm --cached removes it from index, afaik	04:47
johnw	yeah	04:47
	how do I find out that that happened?	04:47
Tv	so you want a list of files that are in dir but not in index	04:48
	aka -o	04:48
johnw	i.e., that between HEAD and index, a file was deleted	04:48
	i don't want files in dir but not in index, but rather in HEAD but not in index	04:48
Tv	diff-index	04:48
johnw	ah, great	04:48
Tv	--name-status --diff-filter=..	04:48
johnw	what does diff-filter do?	04:49
	ahh	04:49
Tv	git-diff-index - Compares content and mode of blobs between the index and repository	04:49
johnw	you're a wonder	04:49
	git diff-index --cached HEAD --name-status --diff-filter=D	04:49
	← greyface left	04:49
	→ psankar joined	04:49
johnw	if a file is renamed, does it show up only once as "R"	04:49
	?	04:49
	i'll try it out	04:49
	← cemerick left	04:50
johnw	ah, it shows D and A, not R	04:50
	git diff-index --cached HEAD --name-only --diff-filter=D -z \| xargs -0 rm is actually what I needed, thank you Tv!!	04:51
Tv	johnw: you need to ask to get rename detection	04:51
	rm -f --	04:52
	xargs --no-run-if-empty	04:52
	;)	04:52
johnw	ooh, good point	04:52
	you are a shell scripting yoda	04:52
Tv	been around that block a few times	04:52
johnw	xargs --no-run-if-empty -0 rm -f --	04:52
	perfect, i'll remember this	04:52
	← parasti left	04:53
	← ekidd left	04:53
johnw	http://gist.github.com/63050	04:53
	how's it look now?	04:54
	any other gotchas?	04:54
	ah, it doesn't exit 1 if myacprep fails	04:55
Tv	set -e	04:55
johnw	what does that do?	04:56
Tv	cd foo && ...	04:56
	set -e makes it fail if any command fails	04:56
	→ ekidd joined	04:56
Tv	anyway, i'm off to play Left 4 Dead ;)	04:56
johnw	ok, i edited the gist again	04:57
	have fun!	04:57
	i'd love to find 3 friends to play that with	04:57
	ah, xargs doesn't like rm -f --	04:57
	→ amitprakash joined	05:01
amitprakash	err.. can someone please tell me again how to ask git to ignore a file name Manifest in any folder in the repo?	05:02
offby1	I think just put "Manifest" in .gitignore	05:04
amitprakash	offby1, and .gitignore is in .git folder ?	05:04
jast	no, .gitignore belongs somewhere in the working tree	05:05
	→ SRabbelier joined	05:05
jast	if you want an untracked list of ignored patterns, use .git/info/exclude	05:05
	← girishr left	05:05
	→ tjafk1 joined	05:10
	← ekidd left	05:14
	→ a-priori_ joined	05:14
	← amitprakash left	05:16
	→ amitprakash joined	05:17
	→ sdboyer_ joined	05:18
	→ girishr joined	05:18
	← sverrej left	05:23
	→ releod joined	05:23
	→ peritus joined	05:24
	← ashleyw left	05:24
	← h0nk left	05:25
	← tjafk2 left	05:27
	← sdboyer left	05:27
	← a-priori left	05:28
	→ sverrej joined	05:29
	→ chalky joined	05:30
	→ sdboyer__ joined	05:31
	→ error404notfound joined	05:32
	→ ashleyw joined	05:33
	← krawek_ left	05:36
	← pigeon left	05:37
	→ pigeon joined	05:37
	→ plediii joined	05:37
	jstad → jstad\|away	05:39
	→ edwardam joined	05:39
	← edwardam left	05:40
	← a-priori_ left	05:41
	jstad\|away → jstad	05:42
	← jstad left	05:42
	← sdboyer_ left	05:43
Circuitsoft	Is there a way to tell fetch to only get heads with a name beginning with "my_namespace/"?	05:43
johnw	yes	05:43
	in your .git/config file	05:43
	you'll see it now says refs/heads/*	05:44
	just change it	05:44
Circuitsoft	Any way to do it on the command line?	05:44
	nevermind. I guess that helps well enough.	05:45
johnw	you can use the "git config" command	05:45
	to manipulate the same variable :)	05:45
	i just thought editing it would be easier	05:45
Circuitsoft	Next Q: A submodule in a commit points to the sha1 of the tree in it's place when the top tree was committed.	05:45
	← Hydrogen left	05:45
johnw	if I understand you, yes	05:46
Circuitsoft	Could that submodule point to a different head in my git repo?	05:46
johnw	sure	05:46
	i actually do that myself	05:46
Circuitsoft	Like a combination of submodule and git-new-workdir	05:46
johnw	the submodule's URL is just "."	05:46
Circuitsoft	cool.	05:46
	← giallu left	05:46
	← hobodave left	05:46
	← plediii_ left	05:47
	← peritus left	05:47
Circuitsoft	I'm working on a bug tracking system using git as a backing store.	05:47
	(Maybe backing store is the wrong phrase)	05:47
johnw	yeah, i've written one too	05:48
	→ karan joined	05:48
Circuitsoft	I planned on using one single git repo as the bug store for all projects in our company.	05:48
	→ Tuomas joined	05:48
Circuitsoft	First commit, in that repo, which is tagged, would be a template of an empty bug report.	05:48
	then refs/heads/projectname/<sha of original bug report> would point to each bug report.	05:49
	→ hobodave joined	05:49
Circuitsoft	Fetch refs/heads/projectname/* as a "bugs" remote on each workdir using it	05:50
	And I wanted a location where I could "check out" bugs so that they could have test cases in their subtrees, and those test cases could know exactly where, relative to the source, they are checked out to.	05:50
	← amitprakash left	05:51
mlh	MarcWeber: re self-reference; e.g. the name of the directory in which the repo sits be mentioned somewhere within .git/config	05:55
	I wouldn't expect such a thing; just giving an example of seomthing that might bugger up an rsync/cp -a of a repo	05:56
	→ Hydrogen joined	05:58
ben_h	is there a git equivalent to 'svn export'? i googled and found 'git archive', but it doesn't look like that does the same thing.	06:00
	← curvature left	06:01
	→ curvature joined	06:01
Circuitsoft	Is that basically "check out, using my working copy as a reference to speed it up"?	06:01
	← hobodave left	06:01
	→ zeahok joined	06:02
	← bdimcheff left	06:03
zeahok	With git, how can i modify without commit on more than one branches?	06:04
Circuitsoft	ben_h Exactly what result do you expect from this export-like-thing?	06:04
	zeahok - Easiest way is probably with stash.	06:04
	However, it's easy enough to amend a commit that you haven't pushed that you may as well commit, then come back later and reset the commit.	06:04
	→ hobodave_ joined	06:05
zeahok	oh, i see	06:05
	thx	06:05
Circuitsoft	There's also a git-new-workdir script, but it causes havok if both workdirs ever sit on the same branch.	06:05
ben_h	Circuitsoft: a checked out copy of just the given commit - i.e. without /.git and untracked files	06:06
Circuitsoft	Fastest probably would be to do "git archive <tree-ish> \| (cd /where/to/put/them; tar xv)"	06:07
	→ coffee1 joined	06:07
zeahok	and if i commit before switch to another branch, and later, i switch back and reset the first branch, this temp commit is just as if never exists, right?	06:07
ben_h	Circuitsoft: ahh, git archive outputs a tar?	06:07
	nice	06:07
Circuitsoft	zeahok: As long as you don't push.	06:07
	ben_h - unless you say --format=zip	06:07
	git archive is a very lightweight operation.	06:08
	Probably close to, if not as fast as, checkout.	06:08
coffee1	hi all.. new to git here, sort of stuck. doing git pull shows me You are in the middle of a conflicted merge.	06:09
Circuitsoft	coffee1: git mergetool	06:09
	Kind of a wizard - simple, but nice.	06:09
ben_h	Circuitsoft: that's good, thanks - i reckon i might wrap that up into a 'git-export' script	06:10
Circuitsoft	ben_h - you can also do	06:10
	GIT_DIR=/path/to/git/workdir/.git git checkout <tree-ish>	06:10
	← johnw left	06:11
Circuitsoft	or, wait...	06:11
	That would change your HEAD.	06:11
	Nevermind.	06:11
	→ reithi joined	06:11
coffee1	Circuitsoft: when im done wiht mergetool, and i chose left and rights.. do i just do "save merged"?	06:14
Circuitsoft	mergetool will add results that you confirm to the index for you.	06:14
	Then you just commit and you're done with the merge.	06:14
coffee1	err	06:14
	this is the first time im using mergetool, fairly new to commandline git (coming from tortoise svn)	06:15
Circuitsoft	oh.	06:15
coffee1	basically when im done choosing my left and rights...	06:15
	how do i close out of it	06:15
Circuitsoft	mergetool brought up some sort of graphical merge tool. It supports several. I likely haven't used the one you're using.	06:15
	Is it using tortoisemerge?	06:15
coffee1	ok i'll just click buttons	06:15
Circuitsoft	What program is it?	06:16
coffee1	filemerge	06:16
	launched when i ran git mergetool :)	06:16
	← hagabaka left	06:17
	← ben_h left	06:17
	← nick_h[litage] left	06:17
Circuitsoft	Are you on a mac?	06:17
	→ peritus joined	06:18
	→ hagabaka joined	06:18
coffee1	ya	06:18
Circuitsoft	You probably do want to do "save merged"	06:18
	that will return to mergetool, and it will continue with the rest of it's steps.	06:18
	← karan left	06:20
	→ ijcd_ joined	06:20
	→ kanru joined	06:21
	← ijcd_ left	06:22
zeahok	i accidently run <git reset HEAD^> more than once, when i want to cancel temp commit. is there anyway to cancel git reset?	06:23
coffee1	ar	06:23
	now i do git pull, and i get this weird msg: error: Entry 'app/controllers/about_controller.rb' would be overwritten by merge. Cannot merge.	06:23
	→ peritus__ joined	06:24
Circuitsoft	coffee1: app/controllers/about_controller.rb is an untracked file.	06:24
	Either add it, or, my favorite, do a "git stash" to save it for later then pull.	06:24
kevlarman	zeahok: find the most recent commit you want to keep in git reflog	06:24
	← moconnor left	06:24
coffee1	untracked.. so i have to manually set every new file to be understood by git?	06:24
kevlarman	and reset to it	06:24
Circuitsoft	coffee1 - yes.	06:25
	← johan-s left	06:25
coffee1	isn't that a bit annoying though?	06:25
	what if i accidentally create a new file, and forget to track it	06:25
Circuitsoft	That way, in the case of compiled languages, it's easy to avoid committing object and other temp files.	06:25
	In that case, do a commit --amend.	06:25
	← peritus left	06:25
Circuitsoft	Or, yet another option, use "git fetch" instead and manually merge the remotes in.	06:25
	gitk --all is quite handy.	06:25
coffee1	hmm lots of options here lol	06:26
	so gitk --all then.. that will track all my new files?	06:26
	← SLiMIT left	06:26
Circuitsoft	That will show you the entire history tree so you and browse around on it and merge with it more easily.	06:27
	I use it kinda like a GPS.	06:27
	or map.	06:27
	← releod left	06:27
	→ moconnor joined	06:27
zeahok	so git will track each commit or reset that i did, and i can set the current state to any of them ?	06:30
Tv	zeahok: resets aren't part of permanent history..	06:30
	zeahok: this may or may not help: http://eagain.net/articles/git-for-computer-scientists/	06:30
Circuitsoft	zeahok - The reflog is really more of an emergency facility that's there when you need it than something to rely upon every day.	06:31
zeahok	thx, but i just tend to make mistakes all the time ^_^	06:32
Circuitsoft	That's okay.	06:32
	It's good for reverting a mistake you just made.	06:32
	If you did a reset a few months ago, you can't undo it.	06:32
	← ZiGzAg_ left	06:32
	→ ZiGzAg__ joined	06:32
Circuitsoft	So, as long as you notice your mistakes quickly, you'll be fine.	06:33
	→ johnw joined	06:33
zeahok	so i should use reset when i want to forget it, and use revert when i want to remember it?	06:33
coffee1	ugh	06:34
	i can't git push... " ! [rejected] master -> master (non-fast forward)"	06:34
Circuitsoft	Did you reset master?	06:34
	As long as no one else is pulling from that repo, do a "git push -f"	06:35
	(force)	06:35
	← Beket left	06:39
	→ Eulex joined	06:42
	← Hydrogen left	06:43
	← sverrej left	06:43
	→ sverrej joined	06:44
	← chalky left	06:52
	← samsoffes left	06:53
	→ Bennid joined	06:53
	→ samsoffes joined	06:54
	← samsoffes left	06:54
	→ j416_pod joined	06:56
	← johnw left	07:00
	→ ludde joined	07:01
	← elmarco left	07:01
	→ elmarco joined	07:01
	← jmspeex left	07:02
	→ joshdavey joined	07:02
	→ Laibsch joined	07:05
	← coffee1 left	07:05
	← plediii left	07:06
	→ plediii joined	07:07
	← icwiener left	07:09
Laibsch	"git mv A dir/;git commit dir/" treats the change as if dir/A was added instead of moved. If I consolidated a couple of files into that directory how can I correctly commit this as a move operation without specifying each and every single file on the command line?	07:09
Circuitsoft	Laibsch - there's no difference.	07:10
doener	"git commit dir/" means "commit only the changes for file matching the path spec dir/"	07:10
Laibsch	Circuitsoft: I am sorry, I don't think you understood my question.	07:10
	Let me try again	07:10
RandalSchwartz	most moves are irrelevant to git	07:10
doener	so the removal of A is not committed	07:10
RandalSchwartz	if it's intereting to git, it'll just be noticed	07:10
Laibsch	doener: OK, understood.	07:10
RandalSchwartz	interesting	07:10
	the "git mv" command is completely irrelevant	07:10
Circuitsoft	yup	07:11
doener	Laibsch: you can do "git add dir; git commit" or "git commit -i dir"	07:11
RandalSchwartz	it adds nothing to the commit	07:11
	commits are just "the new tree"	07:11
Circuitsoft	It will, however, detect the move later if you do a diff and pass the proper filters to gitdiff-core	07:11
RandalSchwartz	they contain no info of renames, deletes, or adds	07:11
	i know it's hard to imagine that the extra meta info isn't being used, but in fact, it isn't. it's derived from contents when needed.	07:12
doener	Circuitsoft: well, what Laibsch did would never end up in git detecting a rename, only a copy, and only when using --find-copies-harder	07:12
Laibsch	doener: Thank, -i seems to do what I want	07:12
	← stephenlb left	07:12
	← agenteo left	07:13
	← j416_pod left	07:13
Laibsch	RandalSchwartz: I think you are mistaken (coming from an mtn background, so you are the expert). Think about git log	07:13
RandalSchwartz	git log is derived from commits	07:13
Laibsch	http://wiki.openembedded.net/index.php/Upgrading_Packages	07:14
doener	Laibsch: no, he's right in the fact that no rename meta data is stored	07:14
RandalSchwartz	and commits don't care about renames, or deletes, or adds.	07:14
doener	Laibsch: that's all derived from the content	07:14
Laibsch	OK, I'm still rather new to git	07:14
RandalSchwartz	a commit is a tree plus parents	07:14
Circuitsoft	Laibsch I just read a bunch about the git diff core today, and it was quite insightful on just this question.	07:14
RandalSchwartz	git log derives its data from comparing two commits	07:14
Circuitsoft	I'll see if I can find a link.	07:14
RandalSchwartz	nothing in there talks about rename	07:14
doener	Laibsch: he just missed the fact that your command sequence never committed the removal of A, so there's no way for git to detect a rename there	07:14
	← bobesponja left	07:14
Laibsch	ic	07:15
RandalSchwartz	"ic"?	07:15
Laibsch	Let me see if I understand this properly	07:15
RandalSchwartz	'ick'?	07:15
Laibsch	I see	07:15
RandalSchwartz	are you paying for IRC by the character?	07:15
Laibsch	No, I am not a Berliner ;-)	07:15
RandalSchwartz	if so, sucks to be "you"	07:15
Laibsch	;-)	07:15
RandalSchwartz	or even "u"	07:15
Laibsch	I am lazy	07:15
RandalSchwartz	please don't be	07:15
doener	RandalSchwartz: nah, sux2bu	07:15
RandalSchwartz	some people here aren't english speakers	07:16
	when you abbreviate, they may not follow you as well	07:16
Laibsch	I'm not	07:16
RandalSchwartz	so there you go	07:16
Laibsch	alright	07:16
RandalSchwartz	don't be skimpy	07:16
Laibsch	btt ;-) back to topic	07:16
Circuitsoft	I really don't get why txting is popular.	07:16
Laibsch	Let me see if I understand this properly	07:16
RandalSchwartz	Circuitsoft - blame sms	07:16
Circuitsoft	When I send SMS messages (160 char limit) I still use full words and sentences.	07:16
RandalSchwartz	140 char	07:16
	becuase twitter is the same	07:16
Circuitsoft	160 on CDMA in the us...	07:17
RandalSchwartz	odd	07:17
doener	160 over here (Germany), too	07:17
RandalSchwartz	sms is 140 in the us	07:17
Circuitsoft	AFAIK, SMS is always 160 chars	07:17
Laibsch	"cp a.txt b.txt ; git rm a.txt ; git add b.txt;git commit a.txt b.txt;git log b.txt" will show me the complete history of a.txt as well?	07:17
Circuitsoft	Twitter is 160 chars	07:17
	yes	07:18
	→ spooneybarger joined	07:18
Laibsch	Because if that is not the case a rename is not the same	07:18
doener	Laibsch: no, but "git log --follow b.txt" will	07:18
Tv	Circuitsoft: umm, twitter is 140 chars	07:18
RandalSchwartz	weird - why is twitter 140 chars then?	07:18
doener	Laibsch: but the same is true for "git mv a.txt b.txt"	07:18
Tv	they reserve some to say from who the tweet was, when they send SMS to your phone	07:18
	anyway, who uses sms anymore	07:18
	stop giving greedy telcos money	07:18
Laibsch	doener: OK, but then the "git mv;git commit -i" thing will automatically do the right thing and thus be "better"	07:18
Circuitsoft	Laibsch: git blame will also follow history back.	07:18
Arjen	SMS is 1120 bits, which is 160 or 140 characters, depending on 7 or 8 bit characters	07:19
RandalSchwartz	ahh	07:19
Circuitsoft	Okay.	07:19
	So SMS is 140 chars in languages that aren't 7-bit clean.	07:19
doener	Laibsch: hm? it's more convenient to use "git mv a b" than to use "mv a b; git add -u a; git add b", yes. But otherwise, there's no difference	07:19
RandalSchwartz	in fact... cp a /tmp/xx; git rm a; cp /tmp/xx b; git add b	07:20
	that'd do the same as "git mv a b"	07:20
	→ chalky joined	07:20
	← comebo left	07:21
	→ thiago_home joined	07:21
RandalSchwartz	the actual "mv" isn't recorded in the mv	07:21
	er... commit	07:21
	→ giallu joined	07:21
Tv	Circuitsoft: that's not really true either because then there's the bastard version that repurposes {\|}[\] for non-ascii characters ;)	07:22
	← rphillips left	07:22
	→ topinupa joined	07:22
Circuitsoft	Tv: I'm not aware of that.	07:23
Tv	Circuitsoft: that was used heavily in the days of non 8-bit safe transports on usenet etc	07:23
	recently i've seen some spam with that encoding, which means somebody fed old usenet archives to a markov chain ;)	07:24
	→ rphillips joined	07:24
	← ashleyw left	07:25
Tv	http://en.wikipedia.org/wiki/7-bit_ASCII#Incompatibility_vs_interoperability	07:25
	→ mikeando joined	07:25
	← robinr left	07:26
	→ gaveen joined	07:26
	→ samsoffes joined	07:27
	→ Weaselweb joined	07:29
	→ ph^ joined	07:29
	← samsoffes left	07:32
	→ tansell joined	07:34
	→ kate21de joined	07:35
	→ \ask- joined	07:40
	→ CardinalNumber joined	07:41
dj_ryan	hmm	07:44
	a question	07:44
	im trying git rebase	07:44
	dealing with the conflicts... once i hit 'git add file/been/resolved'	07:45
	it dissapears	07:45
	→ reprore joined	07:45
dj_ryan	and git rebase --continue dosnt work	07:45
JohnFlux	should I add the m4/ directory to git?	07:45
	in a c++ progarm	07:45
	→ scheng joined	07:47
	→ hdl2 joined	07:48
fynn	when I "git diff", it doesn't show the contents of added files.	07:48
	how can I get that?	07:48
Circuitsoft	fynn - do you want to diff the index against the working dir or against HEAD?	07:49
ZiGzAg__	anyone who uses githooks?	07:49
	← reprore left	07:49
doener	fynn: if you "git add" a file, it's added to the index, so the diff of the working tree against the index (that's what "git diff" does) is empty	07:49
	fynn: "git diff --cached" would show it, as that diffs the index against HEAD	07:50
	fynn: with recent git versions, you can do "git add -N file", that just records that you intent to add the file, so "git diff" will show it	07:50
	← mikeando left	07:51
fynn	doener: OK, how do I see both --cached and vanilla?	07:51
doener	"git diff HEAD" ==> working tree against HEAD	07:51
fynn	doener: thanks :)	07:52
	→ bentob0x joined	07:52
doener	dj_ryan: and "git rebase --continue" says what?	07:52
	← thiago_home left	07:54
	← ProperNoun left	07:55
ZiGzAg__	i cant find much documentation of the githooks, and creating a simple script like "touch /tmp/test" as commit-msg and chmod'ing to 755 doesnt do anything.. do i have to commit and push the files in the .git folder too?	07:55
	→ theCarpenter1 joined	07:55
	→ johnw joined	07:56
doener	ZiGzAg__: you can't even add those files.	07:57
	ZiGzAg__: how did you test that hook?	07:57
ZiGzAg__	i manually moved the file in the main repos	07:59
	← moconnor left	07:59
doener	"main repos"?	07:59
ZiGzAg__	on the server	08:00
doener	hooks work locally... the commit-msg hook is triggered by "git commit", so of course you need that hook in a repo where you actually run "git commit"	08:00
	← UnwashedMeme left	08:00
ZiGzAg__	ah, damn ofc.. im still transitioning from cvs	08:01
	→ ceyusa joined	08:05
	→ melvinram joined	08:05
	→ morph__ joined	08:06
	← morph_ left	08:06
	morph__ → morph	08:06
	← ZiGzAg__ left	08:06
	morph → morph_	08:06
	← theCarpenter1 left	08:07
	→ zobie joined	08:07
	→ UnwashedMeme joined	08:08
	← zobie left	08:09
	← girishr left	08:09
	← tansell left	08:10
	← ph^ left	08:10
telmich	moin moin	08:12
	what's the easiest way to find out, whether a commit was introduced before or after a specific commit or tag?	08:13
	→ UnwashedMeme1 joined	08:13
johnw	for the tag case, use git describe	08:14
	as in "git describe $commit"	08:14
	that will tell you which tag it's a part of	08:14
	for the general case, I would just say git log --pretty=oneline \| grep $commit	08:14
	to see if it's in the list	08:14
	→ ph^ joined	08:15
telmich	hmm	08:15
	← fynn left	08:15
johnw	ah, you wanted before or after	08:17
	one thing you can do is: git show-branch $commit1 $commit2	08:17
	then you'll see the relative history between the two	08:17
telmich	johnw: git describe $commit is pretty near to what I searched	08:17
	→ ph^_ joined	08:19
	→ mcella joined	08:19
telmich	johnw: git show-branch 89c581b3fb2986e303f1299e6458e3e9b115fa3f 2.6.29-rc4-1-gd5b5623 and git show-branch 2.6.29-rc4-1-gd5b5623 89c581b3fb2986e303f1299e6458e3e9b115fa3f both show many commits	08:21
	johnw: but this does not make clear for me, whether 89c581b3fb2986e303f1299e6458e3e9b115fa3f is already part of 2.6.29-rc4-1-gd5b5623 or not	08:22
johnw	they show the relative histories	08:22
	ah	08:22
telmich	johnw: the problem I'm solving is: i got a patch, that should fix my wlan problem; but I think the kernel I built already contains the patch (with a seperate commit)	08:23
	now I'm searching for the easiest way to tell the developer that I'm already running a kernel using that patch	08:23
	→ ProperNoun joined	08:24
	→ Niamor7 joined	08:24
johnw	do you know the commit id of the patch?	08:24
telmich	yep, it's 89c581b3fb2986e303f1299e6458e3e9b115fa3f	08:24
johnw	and you want to know if your current branch includes this commit?	08:24
telmich	and 2.6.29-rc4-1-gd5b5623 is the "kernel version" I'm running	08:25
	→ [1]metaperl_work joined	08:25
telmich	johnw: no, I want to know whether 2.6.29-rc4-1-gd5b5623 includes this commit	08:25
johnw	ah, ok	08:25
	now I get you	08:25
telmich	because currently I'm building latest linux-2.6, so HEAD is different	08:25
drizzd_	git rev-list d5b5623..89c581b3f	08:26
	→ dmlloyd_ joined	08:26
drizzd_	89c581b3f is included if the output of this command is empty	08:26
johnw	yes, drizzd_ is right	08:27
	was just getting the args right on the same command	08:27
	← UnwashedMeme left	08:27
	← ceyusa left	08:27
	← CardinalNumber left	08:27
	← gaveen left	08:27
	← pigeon left	08:27
	← Daniel_B left	08:27
	← Niamor left	08:27
	← Ratler left	08:27
	← MarcWeber left	08:27
	← schlort left	08:27
	← metaperl_work left	08:27
	← precision left	08:27
	← PugMajere left	08:27
	← johs left	08:27
	← Bracki left	08:27
	← jacobat left	08:27
	← milli left	08:27
	← Helios left	08:27
	← andres left	08:27
	← com4 left	08:27
	← mpr left	08:27
	← radii left	08:27
	← kadoban left	08:27
	← multani left	08:27
	← mike_m left	08:27
	← Pieter left	08:27
	← lamont left	08:27
	← KiBi left	08:27
	← ruphy left	08:27
	← cnu left	08:27
	← dotwaffle left	08:27
	← eMBee left	08:27
	← jpeeler left	08:27
	[1]metaperl_work → metaperl_work	08:27
telmich	ok, in that case it's not (got some commit ids)	08:27
	← jchris left	08:27
	→ ceyusa joined	08:28
	→ CardinalNumber joined	08:28
	→ gaveen joined	08:28
	→ pigeon joined	08:28
	→ Daniel_B joined	08:28
	→ Bracki joined	08:28
	→ Ratler joined	08:28
	→ MarcWeber joined	08:28
	→ schlort joined	08:28
	→ ruphy joined	08:28
	→ lamont joined	08:28
	→ mpr joined	08:28
	→ Helios joined	08:28
	→ jpeeler joined	08:28
	→ multani joined	08:28
	→ kadoban joined	08:28
	→ mike_m joined	08:28
	→ jacobat joined	08:28
	→ eMBee joined	08:28
	→ cnu joined	08:28
	→ precision joined	08:28
	→ johs joined	08:28
	→ milli joined	08:28
	→ radii joined	08:28
	→ com4 joined	08:28
	→ andres joined	08:28
	→ KiBi joined	08:28
	→ PugMajere joined	08:28
	→ dotwaffle joined	08:28
	→ Pieter joined	08:28
	← pigeon left	08:28
	→ pigeon joined	08:28
	← jpeeler left	08:29
	← gaveen left	08:30
	→ gaveen joined	08:31
	→ galderz joined	08:31
	→ charon joined	08:31
telmich	hmm	08:32
	drizzd_: are you sure?	08:32
	I just did this test: took one commit _after_ 89c581b3fb2986e303f1299e6458e3e9b115fa3f, which is 17c9d12e126cb0de8d535dc1908c4819d712bc68	08:33
	and then did git rev-list 17c9d12e126cb0de8d535dc1908c4819d712bc68..89c581b3fb2986e303f1299e6458e3e9b115fa3f	08:33
	which is not empty	08:33
	by the way: I'm using linux-2.6, if you want to test it yourself	08:34
drizzd_	telmich: git rev-list 89c581b3f..17c9d12e12	08:34
	\|wc -l gives 424	08:34
	→ girishr joined	08:35
telmich	yep, and the other way round it's wc -l = 11	08:35
Laibsch	http://wiki.openembedded.net/index.php/GitPhraseBook#Setup documents one way for users to access git and another one for those with write access. I have already checked out as a normal user. Can I switch to the other access method or do I need to make a completely new checkout?	08:36
	checkout = git clone	08:36
drizzd_	either way, the commit is clearly not "one commit after", or even "424 commits after"	08:36
	because 89c581b is not the only ancestor	08:36
	(otherwise one of the rev-lists would be empty)	08:36
telmich	drizzd_: I determined "after" by doing git log, /89c581b3fb2986e303f1299e6458e3e9b115fa3f, scrolled up to the next commit and selected its commit id	08:37
	← doener left	08:38
	→ doener joined	08:38
	→ msanchez joined	08:38
	← gaveen left	08:38
	thumper_laptop → thumper	08:39
	← dmlloyd left	08:40
drizzd_	telmich: git log has to display a DAG in a linear manner. it's not surprising that the commits don't actually have that linear order.	08:41
JohnFlux	how do I revert the changes on a single file?	08:41
drizzd_	telmich: see "Commit Ordering" in git-log(1)	08:41
	← CardinalNumber left	08:42
telmich	drizzd_: will do, thanks	08:42
drizzd_	JohnFlux: git checkout <revision> <file>	08:42
JohnFlux	drizzd_: thanks	08:43
telmich	drizzd_: I'm now using gitk for a better view	08:46
	→ j416 joined	08:47
telmich	dammit... I just got the commit, because the last commit is a merge of the branch containing it	08:49
	← ph^ left	08:50
	→ jmesnil joined	08:51
	→ tvw joined	08:51
	← benjo2 left	08:54
	→ temoto-wd joined	08:56
	← dreiss left	08:57
drizzd_	telmich: I don't know if there is a smarter way to do this, but to get the children of a commit you can do `git rev-list --children 89c581b3fb2986e303f1299e6458e3e9b115fa3f~..origin/master\|grep ^89c581\|cut -f2- -d' '`	08:57
	← lordpil left	08:58
	→ lordpil joined	08:58
	→ exon joined	08:58
exon	Anyone seen dscho around lately?	08:59
	→ jpeeler joined	08:59
	→ theCarpenter1 joined	08:59
doener	left about 9 hours ago	08:59
exon	I'll try email then. Thanks	09:00
	← killerchicken left	09:03
deavid	there is a standard way to add more apps to "git mergetool"; for example i want to add my own Qt3Designer for the .UI files.	09:05
	?	09:05
	← giallu left	09:05
	← ia left	09:07
doener	deavid: look for "mergetool" in git-config(1)	09:07
un\|ogic	or see http://www.kernel.org/pub/software/scm/git/docs/git-mergetool.html	09:08
	→ ia joined	09:08
deavid	<<Any other value is treated is custom merge tool and there must be a corresponding mergetool.<tool>.cmd option.>> good. Thanks!	09:09
	← elmex left	09:10
	→ elmex joined	09:10
	← tango_ left	09:14
	→ reprore joined	09:15
	→ giallu joined	09:15
	→ sitaram joined	09:15
	← sitaram left	09:18
	→ sitaram joined	09:19
	→ UnwashedMeme joined	09:19
	← bronson left	09:20
	← anjovis left	09:21
	→ offby1` joined	09:26
	Naib_ → Naib	09:27
	→ jony_ joined	09:27
	← jony left	09:28
	→ priidu joined	09:29
	← ph^_ left	09:29
	← orlando\|\|lando left	09:29
	→ ph^ joined	09:29
	← reprore left	09:30
	→ reprore joined	09:30
	→ rraasch joined	09:31
	← rraasch left	09:32
Circuitsoft	Is there any file I can write to that will be prepended to the blank commit message when an editor is brought up for commit?	09:32
	→ solydzajs joined	09:32
	→ rraasch joined	09:33
	← UnwashedMeme1 left	09:33
doener	Circuitsoft: commit.template	09:34
	Circuitsoft: set that to the template file	09:34
	← meyering left	09:35
	→ LotR joined	09:35
dato	uhm	09:37
	I wish log --abbrev-commit would not append trailing dots to the abbreviated hash	09:37
Circuitsoft	How will git gui handle commit.template and the file it points to changing relatively often?	09:38
	→ gaveen joined	09:39
	← kate21de left	09:40
	← offby1 left	09:40
	← iulian left	09:41
	→ moccuo joined	09:42
	→ naeu joined	09:43
sitaram	dato: try playing around with --format maybe?	09:44
moccuo	hi, so im trying to setup a remote git repo. i initialized a bare repo on the server, and then added a remote on my local repository adn then tries pushing to it and i get the whole "! [remote rejected] master -> master (n/a (unpacker error))" and "error: failed to push some refs" error.. any ideas?	09:44
sitaram	for my 1-line output, I've replaced the traditional 'log --abbrev-commit --pretty=oneline' with something using format that works very well	09:44
dato	sitaram: yeah, except that I'd like it for the default log output, and that's a bit more cumbersome	09:45
	and --abbrev=7 doesn't seem to work here	09:45
JohnFlux	how can I commit-and-merge-with-last-commit? :)	09:45
	I committed, but now realised that I made a mistake	09:45
doener	JohnFlux: --amend	09:46
kadoban	JohnFlux: commit --amend	09:46
JohnFlux	ah	09:46
dato	ah	09:46
	← tvw left	09:46
sitaram	dato: true :-(	09:46
JohnFlux	awesome :)	09:46
	← theCarpenter1 left	09:47
	→ meyering joined	09:47
dato	ah, --abbrev is only for diff --raw	09:47
	← meyering left	09:47
charon	dato: --abbrev-commit	09:47
	→ meyering joined	09:47
charon	oh, sorry, i should've read the backlog.	09:47
	charon hides again	09:47
	→ tango_ joined	09:50
	← drizzd_ left	09:51
	→ dwave joined	09:54
	→ lxsameer joined	09:55
	→ radarek joined	09:56
sitaram	dato: are you parsing it? meaning if you don't need the colors you could use a sed or something in between...?	09:56
lxsameer	if i clone from a big repository and in the middle of cloning i disconnct from internaet , the what's happend ? can i resume that later?	09:57
dato	sitaram: no, this is just for my consumption; I find the short version much more pleasing (particularly to copy&paste), even if it costs some computational power to calculate it	09:58
sitaram	dato: before I decided I like the one line summary a lot more for most day-to-day stuff, I used to use this: log --abbrev-commit --pretty=oneline	10:01
charon	dato: http://gist.github.com/63828	10:01
	i wonder if this could be turned into a patch without breaking too many things	10:01
	→ mithro joined	10:02
	→ yofel joined	10:02
	← j416 left	10:03
	→ JensB joined	10:04
JensB	Hi everybody	10:04
JohnFlux	Hi Dr. Jens!	10:05
	← \ask- left	10:05
JensB	ah. so that's where my thesis has been	10:06
	:)	10:06
dato	ah, at least now I understand why it adds dots	10:06
JensB	anyway .. I have a small problem with Git: heres what I did	10:06
	→ drizzd joined	10:07
JensB	situation: local Git repo connected to remove SVN repo, every now and then I do "git svn dcommit". Yesterday I dcommitted that way, today I discovered I wanted to --amend some files to the commit. Today, "git svn dcommit" tells me "Item already exists in filesystem: File already exists: filesystem '/srv/db', transaction '198-1', path '/dor/trunk/app/controllers/meeting_points_controller.rb' at /usr/local/git/libexec/git-core/git-svn line 469".	10:08
	How do I fix this?	10:08
	any way I can remove the --amended files from the commit and split the commit in two so it matches the SVN state again?	10:09
charon	JensB: find the old state in 'git log -g', then say 'git reset --soft $old_state' to undo the amending and 'git commit' to make a new commit for the second half	10:09
	and don't _ever_ rewrite (amend, rebase, filter-branch) commits that come from git-svn	10:10
JensB	ok, I'll tattoo that on my forehead	10:10
	thanks	10:10
	→ Phlogi joined	10:13
	→ _slacker_ joined	10:14
	← johnw left	10:17
	← Phlogi left	10:19
	← warthog9 left	10:19
	→ warthog9 joined	10:20
	← sverrej left	10:25
	← girishr left	10:25
	→ girishr joined	10:27
	→ cemerick joined	10:30
	→ markelikalderon joined	10:32
	← avtolfm left	10:32
	melvinram → melvinram\|away	10:33
	← drizzd left	10:33
	← gaveen left	10:34
	→ Zaphodbee joined	10:34
	→ bram joined	10:35
	← ndim left	10:36
	← Zaphodbee left	10:36
bram	hey guys, I have a simple question: I have some unstanged changes in master, and I decided it's best to switch to a branch and work on these changes instead of doing them in master. so, how can I make a branch, and have the uncomitted/unstaged changes in my master "moved" to that branch ?	10:36
doener	uncommitted changes are not on any branch	10:37
Circuitsoft	bram "git branch new_branch_name"	10:37
doener	just "git checkout -b bla"	10:37
snuxoll	doener's is correct	10:37
Circuitsoft	Don't those actually do the same thing?	10:37
snuxoll	Circuitsoft: no, git branch won't create a new branch	10:37
doener	snuxoll: hm?	10:37
Circuitsoft	I've done it.	10:37
	← lxsameer left	10:37
doener	"git branch" won't checkout	10:37
bram	well, tried that, but as doener says, the uncomitted changes aren't on a branch, so they aren't there in the branch	10:37
	→ ndim joined	10:38
snuxoll	doener: ah, yes	10:38
Circuitsoft	nothing is on a branch until you commit it	10:38
	"git branch new_branch" will make a new branch off the current head.	10:38
bram	so, my only option would be to create a patch from the diff, undo the changes in master, branch, apply the patch in the branch?	10:38
snuxoll	bram: when you checkout a new branch you'll fork a new branch based on head, any uncommited changes are corried between branches	10:38
	*carried	10:38
doener	bram: no, just checkout the other branch and commit	10:39
bram	snuxoll: umm, no, they aren't	10:39
	I just tried that	10:39
alus	bram: last name C.?	10:39
snuxoll	bram: if I modify a file and switch branches without commiting, the file won't be checked out from the switched branch	10:39
bram	alus: hehe, no, last name "de Jong"	10:39
charon	dato: i think this is cleaner: http://gist.github.com/63843	10:39
alus	bram: ok	10:39
charon	dato: running tests now, i'll send it out after lunch.	10:39
doener	bram: yes, uncommitted changes _are_ kept when you switch branches	10:40
bram	snuxoll: if I git checkout newbranch, the (non-comitted, nonstaged) changes are undone and a checkout is made from head	10:40
	snuxnull: git checkout geotag-refactor <- results in two modifications	10:40
	that reset my files to what they were before the changes	10:41
snuxoll	bram: no, it's stating that those files are modified	10:41
doener	bram: http://git.pastebin.com/m68453cb0	10:41
	→ merlin83 joined	10:41
bram	er, oh, errrrrrrrrrrr	10:41
	damn	10:41
dato	charon: cool	10:41
snuxoll	bram: unstaged files are not modified when you switch branches, that's what it's telling you	10:41
dato	charon: I didn't realize you were Thomas Rast	10:41
bram	I'm a stubornheaded pig, I am!	10:42
	:_	10:42
	:)	10:42
	sorry guys, you are completely right	10:42
dato	charon: thanks a lot for ca6ac7f ("add -p: prompt for single characters")	10:42
Gitbot	[git ca6ac7f]: http://tinyurl.com/aur9xw -- add -p: prompt for single characters	10:42
dato	ah, ok	10:42
	← snuxoll left	10:44
	→ snuxoll joined	10:44
	← cemerick left	10:47
	← exon left	10:48
	← eean left	10:49
	→ drizzd joined	10:49
	→ eean joined	10:49
	← yofel left	10:50
	→ yofel joined	10:51
	← _Vi left	10:52
	→ _Vi joined	10:54
	← peritus__ left	10:55
	→ j416 joined	11:02
	← markelikalderon left	11:03
	← j416 left	11:06
	← warthog9 left	11:06
Circuitsoft	On git-fetch, I can limit what refs it pulls down with the remote.remotename.fetch config variable.	11:06
	Can I limit tags?	11:06
	→ warthog9 joined	11:07
Circuitsoft	Or will it only fetch tags associated with the pulled refs?	11:07
	← scheng left	11:10
drizzd	--tags will download all tags	11:10
	← Fissure left	11:11
	← dirker left	11:11
	→ dirker joined	11:11
	→ Fissure joined	11:11
drizzd	tags "associated" with fetched refs are also fetched automatically (even without --tags)	11:11
Circuitsoft	ok	11:11
	cool. Thanks.	11:11
	← kimrhh left	11:12
	← rsd left	11:13
	→ \ask joined	11:15
	→ peritus joined	11:18
	← Bennid left	11:20
	→ markelikalderon joined	11:23
	← jony_ left	11:25
	→ ntoll joined	11:26
	← zeahok left	11:27
	→ lorandi joined	11:30
	→ cemerick joined	11:32
	→ gzy joined	11:34
gzy	hi, after merging back to trunk, and deleting the branch i noticed that i have it merged wrongly (i'm missing some functionality that was in the branch)	11:35
	is there a way to get the branch back, or to be exact it's changes?	11:35
	← peritus left	11:35
	→ johan-s joined	11:37
Eridius	gzy: yes there is, but I'm not in a mood to hold your hand through it. I'm only responding to let you know that your problem has a very simple solution, and hopefully someone else will step up and teach you	11:38
	← muthu_ left	11:39
	porcnick_ → porcnick	11:43
	→ exon joined	11:44
	→ mfilipe joined	11:51
	→ chrisdone joined	11:51
	← porcnick left	11:52
sitaram	gzy: if I understand you correctly, (and if I had to do this very simply), use gitk to point to the merge commit, look at the 2 parent SHA1, figure out which one was your old branch, and then 'git branch my_branch abcdef' (or whatever the sha1 was)	11:52
	→ Hendu joined	11:52
sitaram	other methods include doing stuff like trunk^2 (or trunk^; not sure which without experimenting)	11:53
drizzd	gzy: did you delete with -d or -D?	11:53
	→ eletuchy_ joined	11:53
gzy	hey, thanks, i deleted with -d, and i found the change removing my missing code in the log...	11:54
	becouse the individual logs from the branch dev show up in my log after merge right?	11:54
	anyways i fetched the changset for that commit and done the whole thing somewhat manually.. the important thing is the data was there.	11:54
drizzd	the "logs" being the commits, yes	11:55
gzy	is there a log parameter to output the branch they came from to? meaning is the commit - branch relation kept in history?	11:55
drizzd	no	11:56
	but it's in the merge message	11:56
gzy	ok, thanks	11:59
	← gzy left	11:59
	← yann left	12:03
	← eletuchy left	12:03
	→ jmspeex joined	12:06
	← mcella left	12:07
	← moccuo left	12:07
	→ moore33 joined	12:10
	← sitaram left	12:11
moore33	How do I list the direct children of a commit?	12:11
	← Laibsch left	12:12
	← wanders left	12:12
heipei	children?	12:12
	you mean parents	12:12
moore33	No...	12:12
	Those commits that have a given commit as a parent.	12:12
heipei	you mean "the commits that this commit is based on"	12:12
Circuitsoft	No.	12:12
moore33	No! "the commits that are based on this commit."	12:12
Circuitsoft	He means the commits that are based on this commit.	12:12
	heivei: Assume "this commit" is not a head.	12:13
charon	dato: heh. you're welcome... and tbh the '...' bothered me too, otherwise i wouldn't have jumped at the opportunity to write a patch	12:13
	→ wanders joined	12:13
moore33	gitk lists this directly for a commit, so I assume it's not too hard to dig out of git-log or something.	12:13
jnl	moore33: git rev-list --children?	12:13
moore33	jnl: I thought so, but I can't seem to make it do what I want.	12:15
heipei	well, looking for children isn't really how git works, so i'd assume its very inefficient	12:16
	because it basically needs to check _every_ object and compare if the parent-entry matches	12:16
jnl	moore33: in what way?	12:16
	← psankar left	12:17
moore33	heipei: Does that imply that git-branch --contains is very inefficient?	12:17
heipei	no, it just checks the commits reachable by the branch	12:18
	→ ekidd joined	12:18
moore33	jnl: In my (limited) screwing around with --children, it's not clear what it's doing.	12:18
heipei	moore33: im talking about all the objects in a repository	12:18
	moore33: well, that is an idea: you could do a "git branch --contains <parent>" for every branch, and get the branches containing children	12:19
	→ Sho_ joined	12:19
jnl	moore33: hm yes, not doing what i thought it would do ...	12:19
moore33	heipei: OK. If I assume that the children I'm interested in exist on some branch, then it seems like a feasible problem.	12:19
heipei	moore33: yeah, if not, sure, you'd have to search the whole object database	12:20
moore33	heipei: I guess I need to look at gitk and see how it gets that info. But gitk does look at the whole database :) hmm.	12:20
	← Flathead left	12:20
heipei	moore33: yeah, probably it looks at all the refs	12:21
azuk	it might be illuminating to add --all to git rev-list --children?	12:22
jnl	ah, yes. so without the --all it wont traverse the children, so they wont show up...?	12:24
	so... --all \| grep $sha :)	12:24
	→ peritus joined	12:26
	← awx left	12:26
	→ sporkmonger joined	12:26
	→ Laibsch joined	12:27
heipei	ah, now i understood how to read the output of rev-list with --children	12:27
	moore33: git rev-list --children --max-count=2 <commit>	12:29
	try that	12:29
azuk	jnl, um no, there are no children that it could traverse, think that the history is represented using a single linked list where you can get to the parent but not the other way, and see the description of --all parameter	12:29
	→ boto joined	12:29
	← countd left	12:31
jnl	azuk: why wouldnt it traverse the children with the --all? git rev-list --children --all \|grep $sha seems to work (based on sample size of 1 :) )	12:32
	→ countd joined	12:32
moore33	heipei: I see how it might work, but it's not actually finding the children of the commit I'm interested in...	12:33
heipei	moore33: it is for me. the commit is listed left, and right of it a child	12:33
	→ luis joined	12:33
azuk	jnl, actually, I misinterpreted what you said, forget it	12:33
heipei	i've only tried it with linear history	12:33
luis	Hello. What would be the most straight forward way to do checkout onto some temporary dir?	12:34
moore33	I think the command that is closest to what I want is git rev-list --parents --all \| grep $sha1	12:34
jnl	yeah, the max-count might not work for all cases	12:35
azuk	moore33, as a fellow Common Lisper, you might find it helpful to think of git's history as a DAG represented using conses (even the need to have gc should be obvious then)	12:35
luis	I want to do two separate builds of my project.	12:35
moore33	Then extract the parent/children info from that.	12:35
jnl	moore33: did you try --chlidren but without the max-count?	12:36
	→ rgr joined	12:36
moore33	azuk: Yeah, I get that :) and I realize that the parent-child relationship might be expensive to derive, but it's still useful.	12:36
rgr	sorry to ask obvious but need quick help. Just pulled a program from guit and it wont compile. How do I list last know "good" stage and then retrieve it?	12:37
moore33	Basically, I'm trying to answer the question: "What topic branches can be trivially merged to this branch?" Think maint in the git repo.	12:37
	← kbingham_ left	12:38
	→ drewr joined	12:39
azuk	moore33, you might want ^$sha1 instead of just $sha1	12:39
jnl	moore33: you mean fast-forward?	12:39
	← peritus left	12:39
moore33	jnl: Yes.	12:39
jnl	moore33: isnt it easier to do it the other way? check which other branches contain this branch?	12:40
	moore33: maybe merge-base is usefull	12:42
	← r0bby left	12:42
jnl	moore33: just a guess is that it should return the commit itself if the other branch is a ff	12:42
moore33	jnl: Yes, but they may contain other stuff too. To use the git repo example again, next merges frequently from maint. So a topic branch that was based from some commit in next may have the current maint head as an ancestor, but you can't merge it to maint without getting all the commits between maint and its ancestor on the next branch.	12:42
	jnl: Yes, that's a good idea.	12:43
	→ faithfulgeek joined	12:43
	→ gilimanjaro joined	12:43
gilimanjaro	hello folks...	12:44
	→ cannonball joined	12:44
rgr	How can I find an old compilable "label" in my git? Sorry if terminology is not correct. I need to try and rewind to a compilable code base as master does not compile.	12:44
	melvinram\|away → melvinram	12:44
	→ apassi joined	12:45
gilimanjaro	i was wondering if anyone here knows how to do a 'git pull' for a branch i have not got currently checked out	12:45
ekidd	rgr: checkout HEAD^, HEAD^^, etc.?	12:45
apassi	git pull origin branch ?	12:45
gilimanjaro	(for the purpose of updating my local version with any remote changes, in preparation for a merge)	12:45
	doesnt that pull it into my currently checked out branch?	12:46
	→ mcella joined	12:46
apassi	git pull origin remotebranch:localbranch	12:46
moore33	rgr: You might want to look at git bisect	12:46
	biab	12:46
gilimanjaro	apassi: aha! excellent	12:46
apassi	gilimanjaro: i am not sure, i havent use that much but i pretty much think it should do it	12:46
gilimanjaro	it seems to work! in our case it would be 'git pull origin trunk:trunk' followed by 'git merge trunk' so i can incorporate the latest changes in our central repo in my branch development	12:47
apassi	my 'git svn rebase' didnt pass because conflict, i fixed it and committed, but now when i try to call 'git rebase --continue', it just response 'No changes - did you forget to use 'git add'?', and git svn rebase complains something it cannot create rebase-apply directory. how to solve git svn rebase conflict?	12:49
	← kanru left	12:50
	← levifig left	12:50
apassi	ok, git rebase --skip fixed it	12:51
	← error404notfound left	12:51
	← solydzajs left	12:52
	→ samsm joined	12:58
	← sporkmonger left	12:58
deskin	apassi: you don't use commit to resolve conflicts while rebasing; you just git add after editing the conflicted files and then git rebase continue	12:59
	melvinram → melvinram\|away	13:00
deskin	er --continue; at any rate, if you're happy how things turned out, you can rm the rebase directory (or rebase-apply) from .git; or, you can git rebase --abort and redo it properly	13:00
apassi	deskin: ah, great to know =) thankx	13:01
deskin	be sure you aren't on a detached head now, which is the usual result of walking away in the middle of a rebase	13:01
	→ error404notfound joined	13:01
	← samsm left	13:01
apassi	it should be on head, i tried continue rebase after problems but there was no more left	13:03
	melvinram\|away → melvinram	13:03
deskin	ya, if you happened to be on the last commit, doing git commit and git rebase --skip is basically the same as doing git add && git rebase --continue	13:03
	← brink left	13:04
deskin	it might work earlier in the chain too, but it's still not the 'right' way :)	13:04
	just do git branch and make sure your current branch (with a *) is an actual branch, not (no branch)	13:05
apassi	yeah it is master ;)	13:06
	hmm.. what is "not (no branch)" ? =)	13:06
chrisdone	hey, before I go looking around, is it possible to disable git's tty colour output? I'm using it inside emacs's shell-mode, y'see	13:07
apassi	yeah, git-config	13:07
deskin	if you do git branch and see '* (no branch)' it means your HEAD is detached from a local branch- e.g. if you do git checkout origin/master or some other remote branch, this will happen	13:07
chrisdone	cheers	13:07
	← error404notfound left	13:08
deskin	you can still commit and do most things, but no branch head (like master) is getting updated to point to any new commits you make, so if you git checkout otherbranch, the commit has disappeared	13:08
apassi	chrisdone: there is lot's of properties for colors, i cant remember which you need, the manual specifies these.	13:08
deskin	you can still get it back via the HEAD reflog, but it's annoying	13:08
	← rgr left	13:09
	→ cmarcelo joined	13:09
apassi	deskin: ok, thanks for the hint, never seen this no branch :)	13:09
chrisdone	apassi: yeah, I think under git-status is what I'm looking for	13:09
	← muszek left	13:09
chrisdone	"color.status" :-)	13:09
	← svanlund left	13:09
deskin	if you do git branch in the middle of a rebase conflict, you'll see it	13:09
apassi	i must try that next time	13:10
	← mike_m left	13:10
deskin	which is why I was saying to make sure the rebase had finished, and your future dev work would go on an actual branch	13:10
	→ jackdempsey joined	13:10
	→ sosko joined	13:10
	→ mike_m joined	13:10
apassi	yeah, it would cause lots of gray for my hair.. =)	13:12
sunoano	what's the major difference in between 1.5 and 1.6? do have to read all the changelogs to figure that out ...	13:13
	→ error404notfound joined	13:13
chrisdone	hmm, nah colours isn't it. I really need to tell git to assume it's not in a tty at all. hrm	13:14
charon	sunoano: <HEAD:Documentation/RelNotes-1.6.0.txt> ?	13:14
Gitbot	[git RelNotes-1.6.0.txt]: http://tinyurl.com/6zwgca [blob]	13:14
	← Gitzilla left	13:14
sunoano	charon: ok, will just do that :)	13:15
	→ jrmuizel joined	13:15
deskin	chrisdone: perhaps --no-pager will also do something about colours? I seem to recall a message on the mailing list about this, let me look	13:15
	'this' = 'git from within emacs being strange'	13:15
	← luis left	13:17
	→ Ryback_ joined	13:17
deskin	hm, looks like that was about getting CR codes within emacs	13:17
sosko	hi all	13:19
	melvinram → melvinram\|away	13:19
deskin	chrisdone: not sure how useful it is, long list of messages, but http://thread.gmane.org/gmane.comp.version-control.git/106487/focus=108057 is what I was referring to	13:19
	→ vals_ joined	13:19
	← tango_ left	13:20
	→ awx joined	13:20
	← joshdavey left	13:20
	← ludde left	13:24
sgronblo	How do I interactively revert a file to only restore some of the modified sections?	13:24
	→ kbingham joined	13:25
deskin	sgronblo: git revert --no-commit && git reset HEAD && git add -p path/to/file	13:25
	→ Yuuhi joined	13:25
deskin	er, the revert needs a commit id to revert, but all the same	13:26
sgronblo	Ok I meant revert in the svn way	13:26
	I don't need to undo a commit	13:27
doener	"unstage" is the probably the right term then	13:27
sgronblo	Aha	13:27
doener	and IIRC there's no way to do that, except for using git gui	13:27
	vals_ → tango_	13:27
sosko	i'm about to start using git in production environment... i'm working on some php framework... and i'd like to be able to setup my repository in some sort of "layers".. first, i need to be able to update framework itself, which provides basic skeleton that i put my files in. next, i have my own modules that i'd like to maintain separately (i read that git submodules are useful for that), and those modules have to be placed in existing directory	13:27
sgronblo	Hmm, but isnt staging defining what will be included in the next commit?	13:27
doener	sgronblo: ah, oh, well, you meant changing the version in the working tree	13:28
	sgronblo: sorry, jumped to conclusions there	13:28
sgronblo	i did? :)	13:28
	i'm looking for something like git checkout --interactive filename	13:29
eikke	is there any way, next to git-status (which has no useful exit code) to figure out whether there are changes in a tree compared to index/last commit?	13:29
	→ kanru joined	13:29
doener	sgronblo: I guess so? There are only two things you could mean, the working tree and the index, and the index would be staging/unstaging ;-)	13:29
sosko	provided by framework.. then, i have my own code, that also goes "on top" of existing code, in some folders.. and on top of all that, i need to be able to mark some code state as "production" or "development" so i can deploy that code on dev/prod servers respectively...	13:29
deskin	sgronblo: a not-recommended-but-should-work way would be to git commit && git checkout HEAD~1 -- path/to/file && git reset HEAD && git add -p path/to/file, then git commit --amend (maybe?)	13:29
doener	sgronblo: you could do: git reset -- file; git add -i file; stage what you want to keep git checkout -- file	13:29
	sgronblo: the last checkout gets the version from the index into the working tree, so you keep exactly the stuff you staged	13:30
sgronblo	but... that looks so complicated for something that should be simple	13:30
	Niamor7 → Niamor	13:31
sgronblo	it should work similar to git add -p	13:31
	deskin ponders whether a switch-index-and-working-tree command would be useful	13:32
doener	sgronblo: hm, that add -p + checkout is easy enough for me, and for more complicated things, I have a script wrapping vimdiff	13:32
sosko	any hint about where to look for an answer to similar question?	13:33
deskin	actually I think something crazy could be done with stash --keep-index and reading some of the stash commits into the index, but it wouldn't be any less complicated; all it would give would be add -p in reverse	13:34
sgronblo	doener: so when you do the last git checkout -- file with the changes you want to keep staged, will that notice that you have staged changes for that file that are not in the file as it is in the working directory?	13:35
	← boto left	13:37
	← priidu left	13:37
petaflot	hello! I tried to nuke the master branch to "ignore" a few commits, and now I'm having a problem pulling from the repo. git's asking me to specify which branch it should merge	13:37
	← gilimanjaro left	13:37
doener	sgronblo: hm? That command replaces the file in the working tree with the version from the index	13:37
ekidd	petaflot: git checkout -b master origin/master may help if you've deleted master.	13:38
	petaflot: For more complicated problems, you may need to edit .git/config directly.	13:38
	→ priidu joined	13:38
ekidd	It rather depends on exactly how you "nuked" the branch.	13:39
sgronblo	doener: ah and you already wrote that once. sorry.	13:39
	→ sporkmonger joined	13:40
sgronblo	then i guess i understand how that solution works and it doesnt seem that complicated. thanks for the advice.	13:40
ndim	Is it on purpose that "git svn clone" (git version 1.6.1.3) adds all SVN tags as remote branches instead of git tags? git-svn(1) does not say anything useful about it, apparently.	13:41
	← cmarcelo left	13:41
doener	ndim: tags in svn are just by convention	13:41
	→ davisp joined	13:41
	← EmilMedve left	13:42
davisp	Is there a way to make gitosis not require a config change for every repository?	13:42
	err, every new repository that is	13:42
petaflot	ekidd: thanks! it worked like a charm	13:42
doener	ndim: and I've probably seen more "not so much of a tag"-tags in svn repos than such tags that you could sanely and directly convert to git tags	13:42
ndim	doener: OK, that makes sense. I knew the SVN-tags-as-convention part, but not the horrible state of real world SVN tags. Makes sense though. :)	13:44
	→ cmarcelo joined	13:44
	→ jolelion joined	13:45
jolelion	hello	13:45
	← cmarcelo left	13:45
doener	ndim: well, might be that I've just been exposed to the bad repos and got a wrong impression, but even a wrong "svn cp" invocation can cause svn to record metadata that makes it pratically impossible to figure out the right git commit to use for the tag	13:46
	ndim: because then svn doesn't record the whole directory as a copy of some other directory at revision X, but some mix of file A at revision foo and file B at revision bar and so on	13:47
ndim	Urgh.	13:47
	Sounds like something one does not really want to know. :)	13:47
doener	the most common reason probably being that the copy was done in a working copy or so, where files might be recorded as coming from different revisions, when you didn't do a "svn up" recently or whatever	13:48
	→ aisbaa joined	13:48
doener	and yeah, one really doesn't want to know that, makes you wanna scream if you think about it	13:49
	→ brough joined	13:49
jolelion	I'm going to use git for my development, me and my group of work are using openvpn to be connected together, as I can't use ssh for some reasons , is it possible to use NFS or SAMBA to access shared repositories ?	13:49
	both linux and windows clients	13:50
Grum	it seems i've stumbled onto a bug with git-svn; where the clone will not see a certain branch deleted; what to do now? :(	13:50
	← girishr left	13:50
doener	Grum: hm? you mean in "git branch -r" the deleted branch is still there? That's expected	13:51
	← eean left	13:51
Grum	how is that expected?	13:51
brough	is is necessary to compress very large text files before checking into git, if space is a problem? I understand git does have excellent compression...	13:51
doener	Grum: the branch is not gone from svn either. In fact, you can't easily really delete branches from svn	13:51
Grum	there is an entry in the svn history which purges the branch	13:52
	doener: svn rm branches/xxx	13:52
doener	Grum: and if you checkout the parent revision, the branch is still there	13:52
	Grum: it's still in the repo	13:52
	→ RaySl joined	13:52
doener	it's just no longer in the later revisions	13:52
Grum	euhhmz but shouldn't that be reflected in the import in some way?	13:52
	right now it just skips over the revision	13:53
	→ keystr0k joined	13:53
Grum	what if you would (in svn): make a mistake; delete the branch; recreate it and work on it	13:53
	it seems that right now you will have a mixed up history	13:53
keystr0k	Can you change the message in a previous commit? I forgot to mention something in my last commit.	13:53
Grum	commit --amend afaik	13:53
kadoban	keystr0k: git commit --amend	13:53
doener	Grum: branches are very different in git. SVN is just one large versioned directory, everything else is by convention	13:53
keystr0k	kadoban, dang!	13:54
jolelion	is it possible to use git over NFS or SAMBA?	13:54
doener	Grum: git has "real" branches	13:54
Grum	doener: yeah i know	13:54
	i want to get rid of svn asap	13:54
	→ a-priori joined	13:54
Grum	but while trying to migrate the data i stumbled onto this inconvenience	13:54
doener	Grum: if seeing that branch bothers you, you can just drop it, "git branch -r -d $whatever"	13:54
	← ekidd left	13:54
Grum	what happens if you remove the branch in git?	13:54
doener	Grum: or -D instead of -d if it wasn't merged	13:55
Grum	it also gets purged from history?	13:55
	→ esparkman joined	13:55
moore33	Is anybody here using git for a large tree of binary data e.g. art assets for a game?	13:55
doener	the reference is dropped, and once all reflog entries referencing the commits, and the purge delay has passed, git gc will kill the associated objects	13:55
keystr0k	kadoban, I amended my last commit message... tried a push and it failed... ?	13:55
Grum	doener: ah ok so in order to keep the branch alive in the history the delete gets skipped?	13:56
	→ scientes joined	13:56
kadoban	keystr0k: afaik you shouldn't change history (--amend, etc) on already pushed commits. there's a way to override it, but I don't know it	13:56
doener	Grum: yeah. Otherwise, you could argue that the import is incomplete, as some data that is in the svn repo is not in the git repo	13:56
Grum	yeah that makes sense :)	13:56
keystr0k	kadoban, I see... Mmmkay. thank you :)	13:56
kadoban	(someone smart here surely does, i'm not that great at git yet)	13:57
Grum	doener: would you happen to know a 'good way' to make a svn import a bit more 'git like' ?	13:57
doener	define "git like" ;-)	13:57
Grum	well erm ... branches end up in refs/remotes/xxx	13:57
	shouldn't they be in refs/xxx if you'd used git from the start?	13:58
jolelion	Does anybody tried to use git over NFS or SAMBA	13:58
	?	13:58
doener	in refs/heads. But git-svn is (also) intended as a bidirectional gateway to a svn repo, so it's treated as a remote	13:58
Grum	doener: yeah i understand that; but i want to use it as a one time import	13:59
doener	after a one-shot import there's usually quite a bunch of stuff you want to clean up	13:59
Grum	there dont seem to be flags to support that	13:59
	yeah indeed; that is what i am running into right now	13:59
doener	yeah, there's --no-metadata to avoid the git-svn-id crap, but that's about it	13:59
Grum	there are some config entries which need to be purged; git config --remove-section svn-remote.svn	14:00
	and the branches moved; tags to be looked at	14:00
	is there anything more i am forgetting ? :)	14:01
	→ ciskje joined	14:01
doener	unwanted branches deleted, eventually merges reconstructed (if you had any "real" merges in svn)	14:01
Grum	we actually used svk ..	14:01
	so yeah we had merges	14:01
	how do you mean 'reconstructed' ?	14:01
doener	so that you have actual merge commits in your git repo	14:02
	i.e. commits with more than one parent	14:02
Grum	commit with 2 parents	14:02
	ok that is something i need to test then how it is handled	14:02
doener	the easiest way to do that is probably to use grafts and then using filter-branch to make the changes permanent	14:02
Grum	grafts? (gitk?)	14:03
doener	grafts override the parents recorded in the commit object	14:03
Grum	mm	14:03
doener	let me get an example done real quick	14:03
Grum	git config --remove-section svn-remote.svn	14:03
	mmeh fail copy	14:03
	i found a page; http://community.livejournal.com/evan_tech/255341.html	14:03
	← ph^ left	14:04
	→ beket joined	14:05
Arjen	Hey Grum	14:05
Grum	ey Arjen :)	14:05
doener	yeah, but that's so non-graphical :-)	14:05
Grum	doener: if you have a better way .. at least to visually verify =)	14:06
	planning to script it in either case	14:06
	→ welly joined	14:07
JohnFlux	Hey all. I made some commits locally, and now just did a 'git checkout <revision>' but now I want to go back to where I committed my revisions	14:07
	the trouble is, I can't find my revisions!	14:07
	git log \| grep -i john shows nothing	14:07
doener	Grum: http://git.pastebin.com/m2a94a2fa	14:07
	Grum: that script creates a repo like one you might get from git-svn	14:08
	Grum: the last commit on master was actually a merge of side, but the commit doesn't represent that, as you can see in gitk	14:08
dato	charon: :)	14:08
	← davisp left	14:08
doener	Grum: then a graft is added to override the parents of that commit to turn it into a merge commit, and get the correct history	14:09
	→ ekidd joined	14:09
Grum	indeed; gitk sees the merge with 1 parent	14:09
ekidd	Grum: git-svn can import some kinds of deleted and recreated branches, but not always.	14:09
Grum	i assume it can import deleted and completely merged branches	14:10
ekidd	Grum: I've occasionally needed to fix up the messier corner cases using grafts and filter-branch	14:10
doener	Grum: the nice thing about grafts is that they are immediately effective, so you can check the results before doing the time-consuming filter-branch stuff	14:10
	→ krh joined	14:11
	→ willb joined	14:11
JohnFlux	I've lost my commits :(	14:11
Grum	yeah	14:11
	ok; how can i make this change permanent now?	14:11
JohnFlux	I committed locally, then 'git checkout <revision>' and now I can't find my commits!	14:11
Grum	(i haven't used filter-branch yet)	14:11
doener	JohnFlux: if you do "git checkout <revision>" you get a detached HEAD	14:12
	JohnFlux: if you had a branch checked out before that, just do "git checkout <branch>"	14:12
kadoban	JohnFlux: probably because you're no longer on the branch you made the commits at	14:12
JohnFlux	doener: how do I reattached it?	14:12
ekidd	JohnFlux: Try looking at gitk to see what's going on.	14:12
JohnFlux	doener: whatever that means :)	14:12
doener	JohnFlux: or you can lookup earlier states of HEAD using "git reflog"	14:12
ekidd	JohnFlux: git checkout branchname will take you back to a specific branch.	14:12
kadoban	JohnFlux: look at git reflog, and see what branch you came from. then do git checkout <branchname>	14:12
	→ obeattie joined	14:13
jolelion	Does anybody tried to use git over NFS or SAMBA ?	14:13
bremner	jolelion: people do from time to time. they don't get too much sympathy when it breaks	14:14
Grum	jolelion: i imagine you might have problems with nfs lockfiles but otherwise should (theoretically) be working?	14:14
	← error404notfound left	14:14
JohnFlux	doener: I can't find it in gitk, but I see it in git reflog	14:15
	0e26ef2... HEAD@{4}: My commit message	14:15
obeattie	Hi everyone, I have a question I hope will be fairly easy for someone. I have just merged from a branch to my master (successfully), but I now want to undo the merge. I imagine this is done with git rebase, but does anyone know how? :)	14:15
jolelion	Grum: thanks	14:15
doener	JohnFlux: just "gitk", or "gitk --all"?	14:15
jolelion	bremner: sorry but I don't understand what you say (my english is so bad)	14:15
doener	obeattie: not rebase, reset	14:15
JohnFlux	doener: I don't see it in gitk --all either	14:16
doener	obeattie: use "git reflog show master" to find the pre-merge state	14:16
	→ EmilMedve joined	14:16
	→ eean joined	14:16
doener	obeattie: the use "git reset --hard <entry_from_reflog>" to reset your branch back to that commit (you should have master checked out when you do that, of course)	14:16
obeattie	doener: Okay, thanks, so what do I look for in the reflog?	14:17
doener	obeattie: and the reset --hard drops uncommitted changes, so if you have any, you might want to stash them first	14:17
	→ ph^ joined	14:17
	→ kapone joined	14:17
	← hobodave_ left	14:17
obeattie	doener: Just so we're clear, I want the changes to remain on the original branch (which I haven't deleted)	14:17
	doener: I'm going to merge them back in pretty much straight away	14:17
doener	obeattie: you want to find the entry for the pre-merge state, as I said	14:18
obeattie	doener: Okay	14:18
doener	obeattie: if that last thing you did was the merge, the master@{0} entry will say "master@{0}: commit (merge): ..."	14:18
	→ roger_padactor joined	14:18
doener	obeattie: so master@{1} is the state right before the merge	14:18
	← mcella left	14:19
doener	if you did anything else in the meantime, the number might be different now	14:19
Grum	jipes; filter-branch needs more examples =)	14:19
obeattie	doener: Crap, the merge commit isn't on the branch (the one I called master) -- the branch I'm on is a remote (SVN) branch	14:19
roger_padactor	hello, i installed git on my shared hosting account. everythin installed fine. but when i try to run the git command it sais command not found. so where and how do i put it in the env vars so i can use git?	14:19
doener	obeattie: hm? I don't think I understand what you mean	14:20
obeattie	doener: The "master" branch is actually the remote SVN branch, and when I switch to it I'm in "no branch"	14:21
doener	obeattie: so you already dcommitted the merge?	14:21
obeattie	doener: I've switched to/from a local branch since I did the merge, so now my merge commit is... missing	14:21
	doener: No I haven't dcommitted it yet	14:21
doener	obeattie: ah, so you did the merge on a detached HEAD? Nothing to undo then...	14:21
obeattie	doener: Okay, so what would happen when I do dcommit?	14:22
	doener: Is the merge ignored since the head was detatched?	14:22
	doener: *detached	14:22
	doener: --dry-run doh	14:22
	→ ehamberg joined	14:22
doener	obeattie: dcommit looks at the history of HEAD to figure out what to push to svn	14:22
obeattie	doener: Oh, nice, yeah it's not doing anything	14:22
	doener: Wow that helps me a lot	14:23
doener	obeattie: doing that "git merge foo" on a detached HEAD doesn't affect any ref but HEAD	14:23
	obeattie: and checking out anything else of course changes HEAD again	14:23
	← reprore left	14:23
obeattie	doener: yeah that's fantastic (in this case)	14:23
doener	obeattie: so the result of the merge is only kept in HEAD's reflog	14:23
JohnFlux	doener: I looked at git reflog, and found my last commit. I've now switched to that, and 'git branches -a' says that I'm now on '(no branch)' - I think that's why I was having problems	14:23
doener	yeah, it's great for integration testing	14:23
	← plasticine left	14:24
obeattie	doener: So what happens to the detached head? Will it get "cleaned up" at some point?	14:24
JohnFlux	doener: should I move back into the 'master' branch or something?	14:24
doener	JohnFlux: depends on what you want to achieve	14:24
Grum	doener: what would the proper filter-branch syntax be to make the grafts permanent? i got it working with HEAD^2..HEAD but erm .. i doubt that is proper ^^	14:24
doener	Grum: for your case, probably just "git filter-branch -- --all", assuming that you have no (annotated) tags yet	14:25
ekidd	Grum: Usually something like 'git filter-branch --all'	14:25
JohnFlux	doener: at some point I want to push my changes back upstream	14:25
ekidd	Yeah, use the extra '--' in there.	14:25
doener	obeattie, JohnFlux: HEAD is a special reference that stores what you have checked out	14:25
Grum	and after that purging refs/original and info/grafts manually?	14:25
ekidd	JohnFlux: Once you've rewritten, your only real option is to nuke upstream, and all the existing checkouts.	14:25
doener	obeattie, JohnFlux: when you check out a branch head (i.e. one of the things that "git branch" shows), then HEAD will reference that branch head	14:26
ekidd	JohnFlux: Whoops, sorry. Got two people confused. :-/	14:26
JohnFlux	ekidd: I was wondering :)	14:26
obeattie	doener: Yeah I know that, sorry I'm not too good at explaining what I mean	14:26
	← faithfulgeek left	14:26
doener	obeattie, JohnFlux: and doing "git commit" or "git merge" or whatever, will affect that branch head	14:26
	obeattie, JohnFlux: when you checkout anything else, HEAD will reference a commit, and "git commit" etc. will update HEAD itself to reference the new commit	14:26
obeattie	doener: Uuh I mean... what happens to the (then) head that I merged into in the first place? I can't see I have any way of making that my current head again	14:27
	doener: I don't want to, I'm just curious :)	14:27
doener	obeattie, JohnFlux: if you then checkout something else again, HEAD is changed, and so there's no ref left referencing that commit	14:27
	obeattie, JohnFlux: then, only the reflog is still referencing it, and once that expires, git gc may prune it	14:27
	← jolelion left	14:27
doener	because the commits became unreachable	14:27
ekidd	Yay, reflog.	14:27
obeattie	doener: Right okay I see, that's what I meant :)	14:28
	← rraasch left	14:28
obeattie	doener: Thanks so much for your help anyway	14:28
doener	JohnFlux: so in your case, you might want to create a branch head from that commit, using "git branch <name> <commit-id>"	14:28
	JohnFlux: or just merge the thing into one of your branches, using "git checkout <your_branch>; git merge <commit-id>"	14:29
	→ levifig joined	14:29
Grum	doener: seems the cleaning of the few files will indeed clean it up; thanks for the very nice script to display how it worked ^^	14:29
doener	JohnFlux: or both of those, or just cherry-pick some commits you made while you were on a detached HEAD, or whatever	14:29
	→ levifig_ joined	14:29
JohnFlux	doener: so, at the moment my commits are in 'no branch' can I just do 'git merge master' ?	14:29
doener	JohnFlux: that would merge master into what you have checked out, not affecting master at all	14:30
	cbrake_away → cbrake	14:30
JohnFlux	doener: I keep getting the feeling that I'm doing this wrong or something. This is so complicated compared to what I was doing with mercurial	14:31
	← Hendu left	14:31
JohnFlux	doener: in mercurial, I could just check out, make commits, check out again to update, and so on	14:31
	but in git I seem to keep disconnecting from master.. but why?	14:31
	am i doing something wrong which causes me to disconnect from master?	14:32
doener	because you checkout a commit	14:32
JohnFlux	doener: right, but that commit is still on master	14:32
	so shouldn't I still be on master?	14:32
kadoban	JohnFlux: but you aren't anymore if you checkout a random commit	14:32
doener	"master" is just the branch head, it is a reference to a single commit	14:32
	← levifig_ left	14:32
	← jackdempsey left	14:32
	← levifig left	14:33
Grum	what would be the proper way to let head point at whatever was in the trunk after a 'git svn clone'?	14:33
	← obeattie left	14:33
doener	Grum: head == HEAD, or a the currently checked out branch head?	14:33
	← UnwashedMeme left	14:33
doener	Grum: or even some other branch head?	14:33
	← sosko left	14:34
	← keystr0k left	14:34
Grum	i mean HEAD :)	14:34
	← lorandi left	14:34
doener	git checkout svn/trunk # Or whatever the remote tracking branch is called	14:35
	but I suspect you actually mean the checked out branch head ;-)	14:35
	→ lorandi joined	14:35
doener	(though strictly, HEAD^0 ends up the same)	14:35
Grum	... i just want to have a repository which looks like as it had been made with git all along	14:35
	but currently i'm stuck with svn ^^	14:36
doener	Grum: e.g. you can use "git reset --hard svn/trunk" to move the checked out branch head to the same commit as svn/trunk	14:36
	Grum: so after "git svn clone" you have a branch head "master", that will reference the latest commit on the branch that was last changed in svn	14:36
Grum	yeah	14:36
	and that happens to be a branch's head	14:36
	→ LiamH joined	14:36
doener	Grum: and then you can do "git reset --hard svn/trunk" to make master reference the same commit as svn/trunk	14:37
	← priidu left	14:37
doener	Grum: because after the git svn clone, HEAD references refs/heads/master, and thus the reset changes refs/heads/master	14:37
Grum	k	14:37
	thanks once more :)	14:37
doener	Grum: changing HEAD to reference the latest commit from the trunk would mean "git checkout svn/trunk", and that detach HEAD and not affect refs/heads/master at all	14:38
	→ mncaudill joined	14:38
Grum	yeah that is the effect it had with my tries before	14:39
	← kbingham left	14:39
Grum	after that i just need to do a bare-clone so i can put it on some central place others can clone from right?	14:39
	→ kbingham joined	14:39
	→ bdiego joined	14:39
	→ mattenoble_ joined	14:39
	→ mcella joined	14:39
	← mattenoble_ left	14:40
doener	Grum: yeah, either that, or create a bare repo and push to it. For probably totally insane reasons, I prefer the latter ;-)	14:41
	← kapone left	14:41
	← tek0 left	14:41
Grum	well then i test the transport as well	14:41
	and i can impress my coworkers with the awesome speed	14:42
doener	Grum: oh, and you should do some purging/repacking	14:42
	→ nick_h[litage] joined	14:42
	→ tek0 joined	14:42
Grum	just call git gc; git repack ? or any more special stuff?	14:42
doener	http://n2.nabble.com/purging-unwanted-history-td1507638.html	14:43
	second posting	14:43
	filter-branch leaves backup references, for the case that anything goes wrong, so you need to get rid of them	14:43
	← warthog9 left	14:43
Arjen	The reflog stuff and a gc --aggressive	14:43
doener	and the reflogs likely also keep the old stuff around, so that also needs cleaning	14:44
	→ aziz joined	14:44
	dmlloyd_ → dmlloyd	14:44
Grum	doener: yeah i saw the 'refs/original'	14:44
doener	Grum: of course the filter-branch invocation itself in that posting is uninteresting for you, just from the "git for-each-ref" onwards	14:45
Grum	yeah i gathered that :)	14:45
doener	Grum: and using -adf _might_ produce somewhat better results than -ad in some cases	14:45
	→ qrush_ joined	14:45
	→ gaveen joined	14:45
	← jrmuizel left	14:46
	→ parasti joined	14:46
	→ boto joined	14:47
roger_padactor	so how do i get git the git command to work if it doesn't on install	14:48
doener	Arjen: using "git gc --aggressive" would do a "repack -Adf" there (unless you set gc.pruneexpire to "now") causing unreachable objects to be forced to be loose. That would cause quite some useless IO action	14:48
	→ priidu joined	14:49
doener	Arjen: as you'd want a "git prune" afterwards anyway, to get rid of the objects	14:49
Arjen	Ah, right	14:49
	→ Hink joined	14:50
	→ faithfulgeek joined	14:50
	→ reprore joined	14:54
qrush_	roger_padactor: um, define not work...perhaps it's just not on your PATH	14:54
	← reprore left	14:55
roger_padactor	when i type git , i get i get command not found	14:56
	what should my path look like and how do i set it?	14:56
	→ nerochiaro joined	14:56
roger_padactor	i added expost PATH=$PATH:/path/to/bin to my .bashrc file. didn't seem to do anything	14:57
	export*	14:58
nerochiaro	hi all, i have a newbie question: is there a way to run a git command from outside the git repo tree ? something like "git status --repo=/my/repo" ? apologies if this is the wrong place to ask	14:59
Grum	just chdir to the repo?	15:00
	→ warthog9 joined	15:00
nerochiaro	Grum: of course. but let's assume i can't do that, is there any other way ?	15:00
Grum	mmm you can set GIT_DIR i think (not sure what it is called)	15:01
	← reithi left	15:02
Grum	so you could probably do: GIT_DIR=/my/repo git status	15:02
	or not ;(	15:03
nerochiaro	yeah, that works	15:03
	thanks	15:03
	dwmw2_gone → dwmw2	15:04
	→ xuhao joined	15:04
Grum	mm didnt work here; but i do see that my git has [--git-dir=GIT_DIR]	15:04
	prolly a version difference?	15:04
	seems i have to point it a the .git dir (not in a bare repo here)	15:06
nerochiaro	yes, i was about to say that	15:07
	← awarde_ left	15:07
nerochiaro	and here --git-dir doesn't work	15:07
Grum	both work here; got 1.6.1.3	15:07
nerochiaro	i'm on 1.5.6.3, that may be why	15:08
doener	nerochiaro: you need "git --git-dir=... <command>" not "git <command> --git-dir=..."	15:08
	nerochiaro: and for things like "status" you'll also want --work-tree	15:08
nerochiaro	doener: excellent, that work for me	15:09
	→ cmarcelo joined	15:11
Grum	man this migration from svn will be quite some work :D	15:11
PerlJam	Grum: but worth it!	15:11
Arjen	At least it's doable	15:11
Grum	i wonder what kind of nasties i will find from the cvs->svn migration :(	15:12
	→ spearce joined	15:12
Grum	Arjen: doable but a lot of manual work still	15:12
	→ awarde joined	15:12
	← awarde left	15:13
nerochiaro	hmm, another question: is it possible for git-rev-list to list only the commits that touch a certain subdirectory ?	15:14
	that change files inside a certain subdirectory, i mean	15:15
doener	git rev-list ... -- directory	15:15
	→ levifig joined	15:16
	← countd left	15:17
WhatAHam	how do i undo the changes to my working copy and go back to the last commit?!?	15:18
nerochiaro	doener: excellent, thanks	15:18
	→ hobodave joined	15:19
	← hobodave left	15:19
doener	WhatAHam: "git reset --hard" if you want to drop all changes	15:19
	→ mikl joined	15:19
	→ hobodave joined	15:19
doener	WhatAHam: "git checkout -- file" to get the version from the index into the working tree	15:19
WhatAHam	ok so git checkout	15:19
	ok	15:19
doener	WhatAHam: "git checkout HEAD -- file" to get the version from HEAD into the index and working tree	15:19
WhatAHam	what's the index?	15:20
	haha	15:20
	WhatAHam ducks	15:20
	→ loincloth joined	15:20
	← ekidd left	15:21
	← charon left	15:21
	← hpa left	15:22
	→ eventualbuddha joined	15:22
roger_padactor	so putting PATH=$PATH:$HOME/bin in my .bash_profile doesn't get git command to work. any other ideas?	15:23
	→ brink joined	15:23
roger_padactor	so i have to restart something for it to take effect?	15:23
vmiklos	logout and login	15:23
un\|ogic	wait	15:24
PerlJam	re-run your .bash_profile	15:24
vmiklos	or just 'source ~/.bash_profile'	15:24
un\|ogic	just source your bash_profile	15:24
doener	or put that into .bashrc instead of .bash_profile	15:24
roger_padactor	its in both	15:24
un\|ogic	or do a setenv	15:24
	:P	15:24
	wait, bash.. export then	15:24
roger_padactor	its exported	15:25
brink	What's the best way to permanently remove a directory of large binaries so they don't take up space?	15:25
doener	then new bash processes should have that path... what does "echo $PATH" tell?	15:25
un\|ogic	echo $PATH shows it's there?	15:25
	→ cilly joined	15:25
	← beket left	15:26
roger_padactor	its not in the path	15:26
	so its not set	15:26
	damn thing	15:27
	um what is source ?	15:28
	→ beket joined	15:28
	→ jrmuizel joined	15:30
roger_padactor	ok that sourcing thing worked... so how do i get it to work on login	15:30
	← cedricv left	15:30
	← a-priori left	15:31
	→ charon joined	15:33
	← merlin83 left	15:34
jaalto	brink asked interesting question: What's the best way to permanently remove a directory of large binaries so they don't take up space?	15:36
	E.g if you decide to split a repository in two and you don't care about the old history (pdf files, images ...)	15:37
	← drizzd left	15:37
	→ ekidd joined	15:38
brink	jaalto: Yes, I have 8M of code and 84M of binaries. Splitting this into two repositories would be a good solution. I don't care about the binary history.	15:38
	jaalto: Of course, I don't know how to split it.	15:39
charon	jaalto, brink: man git-filter-branch	15:39
Gitbot	jaalto: the git-filter-branch manpage can be found at http://git.or.cz/man/git-filter-branch	15:39
brink	charon: Yes. I've tried that often. Maybe I'm missing something, but I've failed to find an incantation with git-filter branch that actually reduces the size of the repository.	15:40
jaalto	charon: Doesn't that also copy history? I'd like to remove the old binaries permanently from current repo	15:41
	← JohnFlux left	15:41
	← Sho_ left	15:41
PerlJam	brink: did you remember to remove the backup refs after you ran filter-branch ?	15:42
charon	you'll have to force-expire everything after the filtering, something like 'rm -rf .git/refs/original; git reflog expire --expire=now --all; git gc; git prune --expire=now'	15:42
	obviously that removes all safety nets so make sure you have a backup	15:42
	might be easier to just clone the repo however.	15:42
brink	charon: Not worried. I do this in a test clone.	15:43
charon	jaalto: all rewriting always copies; as Tv put it, you cannot change, only rewrite and forget. if you fail to forget, of course you'll have the old stuff around too	15:43
	→ error404notfound joined	15:43
	→ cedricv joined	15:43
jaalto	charon: so if you copy the binaries to complely fresh repo (no split; start at zero), and 'rm <files>' from old repo, is there any way to completely wipe the old history?	15:46
brink	charon: I just tried each of those incantation. After each du -h gave the same answer - a whopping 149M	15:46
	← ntoll left	15:47
charon	brink: make sure you've really got rid of all refs pointing to the old history	15:47
	← _Vi left	15:47
	→ girishr joined	15:47
charon	brink: for example say 'git log --all -- $some_large_file'; if that returns anything you still have some refs lying around	15:47
brink	charon: Is there method for doing that?	15:47
charon	after finding some sha1 you can use 'git branch --contains $sha1' and 'git tag --contains $sha1' to find the offenders (the latter is only in very recent git however)	15:48
	← beket left	15:48
brink	charon: I have git 1.5.4. Is that recent enough?	15:49
charon	uh... i doubt you'll even have 'branch --contains' at 1.5.4	15:49
	← chalky left	15:49
	→ sitaram joined	15:49
charon	oh, you do. branch --contains is in since v1.5.4-rc0 and tag --contains since v1.6.2-rc0 (i.e. not released yet)	15:50
brink	charon: That's the latest on mac ports. On Ubuntu I have 1.5.2.5, which seems to be the latest in my current set of universes.	15:50
charon	but in any case, you can check if any tags still have the file with 'git log --tags -- $file' and if so, you probably forgot --tag-name-filter cat and they _all_ still point to the old history.	15:51
	← johan-s left	15:51
brink	charon: I have no tags. It's all from a ghetto model SVN repository.	15:51
	← girishr left	15:51
charon	brink: so did you filter the remotes? what does 'git branch -a --contains $sha1' say?	15:52
	→ error404notfoun1 joined	15:53
	→ HenryHuang joined	15:54
brink	charon: I just took a subdirectory of the trunk. Unfortunately, that included a sub-sub directory of binaries.	15:54
	→ rostedt joined	15:54
brink	charon: Is there a quick way to find the sha1 of a file?	15:55
charon	brink: umm, you need a commit sha1 for branch --contains. that's why i said to log --all -- $file earlier. (the real answer is git hash-object)	15:55
	← aziz left	15:56
brink	charon: Ok. So I can just do git log filename to find the sha1. Is that correct?	15:57
	→ ashleyw joined	15:57
charon	you can use 'log -- filename' to find the sha1 of _some_ commit that touches the file, yes.	15:57
rostedt	hi all. I was wondering about a new little feature. I find that often after pushing to a public git repo, I receive acked-by and tested-by on a commit	15:57
charon	brink: as for your subdirectory, re-filter with '--index-filter "git rm -r --cached --ignore-unmatch $subsubdir"' if you want to remove it. the gfb manpage has more details.	15:58
	← kadoban left	15:58
rostedt	is there some kind of "amendment" feature to add to a change log of a commit after it has been pushed?	15:58
	not actually changing the change log, but to just add a comment to it that can be seen by a 'git log'	15:58
charon	rostedt: currently not without rewriting history. man git-notes could help you but it's probably not going to make it into 1.6.2	15:59
Gitbot	rostedt: the git-notes manpage can be found at http://git.or.cz/man/git-notes	15:59
brink	charon: Ok, the git branch -a --contains sha1 says it exists in origin/HEAD and origin/master, but in the original git repo it's also in master.	16:01
	← RaySl left	16:02
	offby1` → offby1	16:02
rostedt	charon: thanks	16:02
charon	brink: you probably didn't filter the remotes. you could just delete the remote to get rid of its origin/* branches.	16:02
	← error404notfound left	16:02
	← drewr left	16:04
	→ krawek joined	16:05
brink	charon: I just did git remote rm origin. Still the same size though.	16:05
charon	it won't magically throw away objects, you'll have to repeat my "incantation"	16:05
	→ moore_ joined	16:05
charon	(or maybe i just forgot some part of the forget-everything-dance)	16:05
brink	charon: git prune says this: error: refs/remotes/origin/HEAD points nowhere!	16:07
	← moore33 left	16:07
	→ greyface joined	16:07
charon	hrm. bug in git-remote? you can remove it manually with 'rm .git/refs/remotes/origin/HEAD'	16:09
doener	yeah, there was(?) a bug with deleting symbolic refs, the referenced ref was deleted instead	16:10
brink	charon: That fixed the error. After prune it's still all 149M.	16:11
	charon wonders what he forgot this time	16:11
charon	brink: can you check 'git log --all -- $file' for some big file again? it should come up empty if you got rid of all refs	16:11
	(refs pointing to the old history, that is)	16:11
brink	charon: Do I need the full path to the file?	16:12
	→ muthu_ joined	16:12
charon	yes.	16:12
doener	charon: which cleanup dance did you suggest?	16:12
charon	doener: [16:42] <charon> you'll have to force-expire everything after the filtering, something like 'rm -rf .git/refs/original; git reflog expire --expire=now --all; git gc; git prune --expire=now'	16:12
doener	should be fine, except that I use "git repack -ad" instead of "git gc" (stops the unreachable objects from being written as loose objects -> faster)	16:13
charon	ah, good idea	16:13
	i'm actually writing a section for man:gfb now. this comes up every week :(	16:14
doener	the most common reason for the repo still being big is that the files have been moved/renamed, and now you don't remember the old names	16:14
brink	charon: yes. That prints out the reference to the commit.	16:14
doener	yawn the irc log search is slow again :-/	16:15
brink	charon: It has a git-svn-id.	16:16
charon	brink: then go back to finding a ref that points to the commit. if 'git branch -a --contains $sha1' doesn't show any branch, it's probably a tag.	16:16
	← tim_jenssen left	16:16
brink	charon: Yes git branch -a doesn't find it. How do I find a ref that points to the commit?	16:17
	← mfilipe left	16:17
doener	hm, why bother with finding a ref? Just filtering with " -- --all" catches everything	16:18
	→ mfilipe joined	16:18
doener	brink: just in case that the files have been renamed or so:	16:19
	git rev-list --objects --all \| cut -f1 -d' ' \| git cat-file --batch-check \| grep blob \| sort -n -k 3 \| tail -n10 \| while read hash type size; do echo -n "$size "; git rev-list --all --objects \| grep $hash \| cut -d' ' -f2- ; done	16:19
	brink: that monstrosity tells you the filenames used for the ten biggest blobs in your repo	16:19
	→ reprore joined	16:20
Grum	how can i find the branchpoint of a branch that i just know the name off?	16:20
brink	doener: I'm getting this: usage: git-cat-file [-t\|-s\|-e\|-p\|<type>] <sha1>	16:20
doener	brink: old git? Might need 1.6 or so	16:21
charon	Grum: branches have no intrinsic branch point, they just have a merge-base w.r.t. some _other_ branch. you can find that with git merge-base $branch $other	16:22
brink	doener: On Mac I'm at 1.5.4 and on Ubuntu I'm at 1.5.2.5. Those are the latests via port install and apt-get. Should I bypass mac ports and Ubuntu Universes somehow? Should I build it myself?	16:22
	← bram left	16:23
	→ stephenlb joined	16:23
	→ GeertB joined	16:24
doener	brink: I'm usually running a self-built git from git.git's master. IIRC for Ubuntu there are newer packages available "somewhere"	16:25
	brink: at that ubuntu version is _definitely_ worth upgrading. 1.5.2 is almost 2 years old, and there have been just too many improvements to still use that old thing	16:26
	s/at/and/	16:26
	← reprore left	16:26
charon	on my debian box i just downloaded the git-core package from unstable and installed it manually, worked fine	16:27
	→ outoftime joined	16:27
charon	(you'll want to update gitk etc. too, but that one's headless and i only cared about the tools)	16:27
outoftime	is there a way to get a log-type summary of all of the commits that are in one branch, but not in another?	16:27
	(like a git-diff, but showing the commit logs rather than the actual diff)	16:27
	← rostedt left	16:28
	← mfilipe left	16:29
	→ drewr joined	16:30
	→ mfilipe joined	16:30
	→ Hendu joined	16:30
	← giallu left	16:30
	→ Bass10 joined	16:31
charon	doener: how about http://gist.github.com/63983 ?	16:32
	→ edwardam joined	16:32
	← eean left	16:32
	→ eean joined	16:33
	→ Voker57 joined	16:33
	→ chris2 joined	16:33
	→ qrush__ joined	16:35
	→ nothinghappens joined	16:35
	→ kadoban joined	16:37
	← Korren left	16:37
	→ ashleyw_ joined	16:38
	→ mansour joined	16:38
mansour	I have added "*.o" to my .gitignore. However, I couldn't find a way to do d this in all directories. Do I need to put this in every directory ?	16:39
	← qrush__ left	16:39
ekidd	mansour: Just once, at the top level of your project.	16:39
doener	mansour: no, unless there's a slash in there, the rule works recursively for all subdirectories	16:40
	← naeu left	16:40
Grum	hmm so what is .git/info/exclude used for?	16:40
doener	mansour: and "dir/" works also in all subdirs, as that just means that "dir" must be a directory, not a file, and does not anchor the pattern	16:40
	Grum: same as .gitignore, but for things you want to ignore locally	16:41
	Grum: .gitignore is the tracked version, that you can commit so everyone has it	16:41
Grum	hmm for example if you are using a different editor which leaves ~ files orso?	16:41
doener	yeah, or if you have any other funny tools, that leave e.g. directory index files, or something else that not the whole world will care about	16:42
Grum	is there any way to find the most recent branchpoint of a branch?	16:42
doener	you can find the merge base of two branches	16:42
Grum	yeah git mergebase does that but i dont know what it forked from	16:43
	→ reprore joined	16:43
	→ kapone joined	16:43
mansour	ok, so do I need to put a slash only? like this ?? *.o/ ??!!	16:43
	← qrush_ left	16:43
	→ qrush_ joined	16:44
doener	mansour: no... that would mean that only directories ending in .o would be ignored	16:44
	mansour: just *.o is already fine	16:44
	→ akitada_ joined	16:44
	← lresende left	16:45
	← akitada_ left	16:46
	→ solydzajs joined	16:46
mansour	doener: thank you. I just tried it.	16:46
	← ashleyw left	16:46
mansour	so, if I have multiple modules, and I want to ignore a dir called target in each of them, all I have ot do is "target/", right /	16:47
	right ?	16:47
	→ RaySl joined	16:48
doener	modules == submodules?	16:48
ekidd	Is there a good way to iterate over two lists in parallel when using shell scripts?	16:48
	ekidd is working on git patch	16:48
mansour	hum, no. I missed this part. I forgot git has submodules keyword. I have few subdirectories and each of them contains "target"	16:49
	I need to ignore this "target" dir always.	16:49
doener	yep, just "target/" in a top-level .gitignore should be fine	16:49
ekidd	Or even /target/, I think.	16:50
doener	charon: hm, maybe you should mention that "git clone /path/to/repo" won't do the trick, as that just hardlinks everything	16:50
mansour	cool, for some odd reason I had to put .gitignore in each module directory ... I will go back again sometime and double check this	16:50
	← cilly left	16:51
doener	ekidd: no, that anchors the pattern, thus it won't ignore the directory in any subdirectory	16:51
ekidd	Oh, sorry.	16:51
	I wasn't paying enough attention. :-(	16:51
	→ tor3 joined	16:51
	→ akitada joined	16:51
	← sitaram left	16:52
charon	doener: true, i forgot. using file:// should do it, right?	16:52
Grum	doener: you know of now way to see where a branch forked from if you dont know the branch it came from?	16:52
doener	charon: I think so, I always forget how to force the pack generation. And the "rm -rf .git/refs/original" will fail when the refs got packed for some reason (e.g. earlier packing attempts)	16:53
Grum	it seems like you have to find the closest parent which has >1 children	16:53
	→ jordibunster joined	16:53
doener	Grum: git doesn't really have a "branched from" concept in its object model, unlike svn	16:53
xenoterracide	does git format-patch have a -m switch? where could I find the docs on it? I'm wondering if I've been screwing up patches... also where the -M option documented?	16:54
Grum	yeah i know; but the 'parent with multiple children' is what i am looking for; is there any way to easily get that information?	16:54
	or hard ... ;D	16:54
jordibunster	My prompt makes use of git status to see if there's anything dirty, anything untracked, and anything staged. Any way to do it faster than git status?	16:55
Grum	or mmm i could just bruteforce a mergebase on all of the other refs i have; and pick the one with the least revisions between me and the mergebase found	16:55
qrush_	jordibunster: mind control	16:55
doener	Grum: that might as well give you a commit at which another branch was created from this branch	16:55
	→ kukks joined	16:56
Grum	doener: true .. mmm	16:56
	← alip left	16:56
jordibunster	qrush: I was thinking something that stops as soon as it finds one of each.	16:56
	← reprore left	16:56
charon	Grum: git rev-list --boundary HEAD --not $(git for-each-ref refs/heads/ --format='%(refname)' \| grep -v $(git symbolic-ref HEAD)) \| sed-ne 's/^-$.*$/\1/p'	16:56
	→ alip joined	16:57
charon	gives you the most recent commit common with any other branch	16:57
Grum	i so need to decypher that hehehe	16:57
bentob0x	how do you 'pull' single commits from an external project on your local machine?	16:57
charon	Grum: afterwards 'git branch --contains $sha1_from_above' is probably the most helpful	16:57
	→ orlandov joined	16:58
charon	note that it's more or less the same as brute-forcing the merge-base as you suggested, except it's much faster	16:58
	→ reprore joined	16:58
Grum	mmm i am only missing where to put the input-branch	16:58
	(or commit even)	16:59
	← galderz left	16:59
charon	where i wrote HEAD	16:59
	(so you could also just check it out)	16:59
Grum	i tried; it gives nothing for me	16:59
	→ awarde joined	17:00
	→ sdboyer-laptop joined	17:00
mansour	every while and then I get: # Changed but not updated: when I do git status. These files have been added already !!	17:00
	→ kinyobi joined	17:00
doener	Grum: did you replace both "HEAD"s?	17:00
charon	it won't work with a sha1 in place of the HEAD unfortunately	17:01
	it also won't work if you have any other branch containing the same commit	17:01
	melvinram\|away → melvinram	17:01
Grum	i'm getting the error that i'm not inputting a 'symbolic ref'	17:01
charon	oh, right. just replace the entire $(git symbolic-ref HEAD) by the branch name in the second case	17:02
doener	easily fixable with $(git rev-parse --branches \| grep -v ($git rev-parse HEAD))	17:02
	← outoftime left	17:02
doener	s/($/$(/	17:02
	← qrush_ left	17:03
Grum	k, i'll toy with this :)	17:03
	thanks ^^	17:04
	i should start reading those manpages ;)	17:04
	→ kapone_ joined	17:04
charon	a way to spell --all-except that doesn't involve grep might be nice	17:04
	→ bryanl joined	17:04
	← kapone left	17:05
	← vmiklos left	17:06
	→ vmiklos joined	17:06
charon	(damn, another nice idea and we're supposed to be in -rc freeze)	17:06
	← cmarcelo left	17:11
	→ allan_wind joined	17:12
allan_wind	git show 18b6ecb0c	17:18
	...	17:18
	index 2c62fb8..cab14f3 100644	17:18
	How do I find the commit hash for 2c62fb8? Or how do I find the parent for 18b6ecb0c for a particular file?	17:18
offby1	I have this dim idea that you can call "git ls-tree" on lots of trees, and grep the output for that	17:21
	→ _stack joined	17:21
charon	allan_wind: git rev-list -1 18b6ecb0c^ -- file	17:22
MarcWeber	allan_wind: 2c62... seems to be the start of a hash. You can use them instead of the full has has well. Just try git show 2c62fb8	17:22
charon	allan_wind: that will give you the _first_ candidate. it may not be unique	17:22
	→ jchris joined	17:23
_stack	hi,is it possible to do auto replication of a certain repository throught varius servers with git?	17:23
charon	allan_wind: oh, and if 18b6ecb0c was a merge you'll have to investigate all of its parents. you could use 'git rev-list -2 18b6ecb0c -- file' then.	17:24
	→ softdrink joined	17:24
	→ plediii_ joined	17:25
Daniel_B	How do I find any tags that contain a given commit? (1.6.1.2)	17:25
doener	charon: commit^@ == all parents of commit	17:26
charon	doener: damn. thanks. i even took a brief look in rev-parse's manpage :)	17:26
	doener: oh, and i sent out the gfb doc patch, but forgot to CC you.	17:26
doener	no problem, I'm subscribed anyway	17:27
	→ icwiener joined	17:27
	→ cgardner joined	17:28
	← exon left	17:28
	→ abditus joined	17:29
allan_wind	charon: git show 2c62fb8 does show me the blob, but I need the commit for use with bisect	17:29
	← HenryHuang left	17:30
	→ schierbeck joined	17:31
charon	allan_wind: that's why i'm suggesting rev-list. i didn't bring up show :)	17:31
	← reprore left	17:31
	← benn_d left	17:31
charon	allan_wind: taking doener's suggestion into account, you should use git rev-list -1 18b6ecb0c^@ -- file	17:31
	→ Spacecrack joined	17:31
offby1	what exactly does "git clone --mirror" do, and why would one use it?	17:32
allan_wind	charon: git rev-list -1 18b6ecb0c^ -- file gives me 0d41c924293bec930b9c226ab3744a10b7ef0d97 and when I show that it does not contain the before changes corresponding to my diff	17:32
Spacecrack	I wanna clone a complete part of a repository with a non-standard layout. There's not tags, and the branches are located in _two_ subdirectories. Can I use multi-globbing in the .git/config to import that?	17:33
	→ bobmcw joined	17:33
	→ d0k joined	17:34
Spacecrack	I.e., the directory structure looks like this: [common path]/production hosts the stable branches, [common path]/development the unstable ones, and [common path]/main/ is the trunk itself!	17:34
charon	allan_wind: is 18b6ecb0c a merge?	17:34
	→ patrikf joined	17:35
Spacecrack	For testing I only imported all production branches, but gitk strangely only showed one linear path of development.	17:35
kadoban	Spacecrack: gitk --all ?	17:36
offby1	curse kadoban and his fast fingers	17:36
Spacecrack	kadoban: thanks, that did it. Now I see funky arrows. :)	17:37
	→ chalky joined	17:37
charon	offby1: bc699afcce3c4	17:37
Gitbot	[git bc699afcc]: http://tinyurl.com/clu5hn -- clone: Add an option to set up a mirror	17:37
	← yofel left	17:37
Spacecrack	BTW, what does it mean in gitk when there's a long arrow which is vertically interrupted?	17:38
offby1	hmm	17:38
	charon: thanks ...	17:38
abditus	Git newbie question: I've cloned off a remote repository and created a branch I'm working on, and now I'd like to update that branch with the latest remote repository changes. Which command/options do I run?	17:38
	→ NSaibot joined	17:38
offby1	charon: that kinda begs the question, though: I don't know what "git remote --mirror" does.	17:39
	Call me lazy, but I don't feel like deciphering the source to find out	17:39
allan_wind	charon: merge... I don't think so, how do I tell?	17:39
offby1	abditus: "git pull" should do it.	17:39
charon	offby1: unlike clone --mirror, remote --mirror is actually documented "in full" including semantics :)	17:39
offby1	hmm	17:39
charon	allan_wind: 'git show 18b6ecb0c' will show a line 'Merge: parent1 parent2 ...'	17:40
	← Niamor left	17:40
	→ Sitaramcell joined	17:40
offby1	I wouldn't call that "in full": it describes the behavior, but doesn't tell me why that behavior would be useful. What problem does "git remote --mirror" solve?	17:40
	→ Niamor joined	17:40
lack	offby1: It just puts the heads in refs/heads/ and not refs/remotes, so the repository is an exact mirror of the upstream repository.	17:40
	→ iulian joined	17:41
offby1	lack: I gathered that from the docs. But my question still stands: what problem does that solve?	17:41
	Why would I want to do that?	17:41
lack	offby1: I use it to keep a local copy of the kernel.org man git tree on a local server, updated by nightly cron job, so we don't clog our thin pipe during the day.	17:41
	← plediii left	17:41
abditus	When I run "git pull" I get the message " Please name which branch you want to merge on..." It's a little unclear how I specify that. "git pull mybranchname" doesn't seem to work.	17:41
allan_wind	charon: yeah, I thought so. No it was not a merge	17:41
charon	offby1: if people then fetch from the mirror, they'll get the branches; fetch doesn't grab refs/remotes/* by default.	17:42
	← temoto-wd left	17:42
offby1	charon: so basically it's just for saving network bandwidth?	17:42
charon	huh?	17:43
allan_wind	charon: let me ask it more generic. I see a commit, how do i find out who made the defective check-in? git blame 18b6ecb0c^ -- file didn't work	17:43
offby1	charon: I'm trying to figure out what problem "git remote --mirror" solves.	17:43
	"saving network bandwidth" is my guess at the problem; apparently, it's the wrong guess.	17:43
charon	it's to make a mirror of a repo that people can fetch from as if it was the original	17:44
	one use could indeed be to save bandwidth. but maybe you just have cronjobs mirroring your repos around the world to back them up, who knows.	17:44
offby1	yes, yes; but that's not a _problem_. I still don't understand why people don't just fetch from the original.	17:44
	ah	17:44
	now we're talkin'.	17:44
	thanks.	17:44
	← ciskje left	17:45
offby1	so a related question: I did git clone --mirror remote:foo.git mirror, and saw (among other things) "error: refs/stash does not point to a valid object!"	17:46
charon	allan_wind: and how exactly did it not work? that's the right syntax to blame a file in the parent.	17:46
offby1	what's that mean? What should I do about it?	17:46
	The remote indeed has a stash, and my clone indeed doesn't have it.	17:46
allan_wind	charon: blame output does not contain the before change	17:47
abditus	N/M. Got it. git pull <remote branch name here>	17:47
	← dwave left	17:47
charon	allan_wind: that would be a bug, and you'd have to make a test case.	17:47
	offby1: stashes are implemented through reflogs, and those aren't cloned. though i wonder why it would point at an invalid object; it should point to a merge commit of two other commits, all three of which make up the stash state.	17:48
allan_wind	charon: hmm... I think the problem is that 18b6ecb0c^ doesn't do it for the right file	17:48
	bb	17:48
	→ thumper_laptop joined	17:50
roger_padactor	i set up git on my webserver. how do i change my username so i can ssh and clone that repo	17:50
	← thumper left	17:50
roger_padactor	currently its using my unix box username instead of me specifing one	17:51
	→ juanjoc joined	17:52
	→ joshdavey joined	17:52
	→ marvil07 joined	17:53
	← jordibunster left	17:54
	→ a-priori joined	17:55
Ilari	roger_padactor: user@host?	17:59
	← lorandi left	18:03
	→ lorandi joined	18:03
	← Hendu left	18:06
	← cedricv left	18:06
	→ cmarcelo joined	18:08
	← cmarcelo left	18:09
	→ cmarcelo joined	18:09
	→ name joined	18:11
	← marvil07 left	18:14
	→ dwave joined	18:14
	← kapone_ left	18:15
	→ countd joined	18:15
	← lorandi left	18:16
	← Sitaramcell left	18:16
	→ lorandi joined	18:16
	→ kapone joined	18:17
	→ RoPP joined	18:18
	← tango_ left	18:19
	← jmesnil left	18:20
	→ bobesponja joined	18:22
	→ jboyens joined	18:24
Grum	charon: git rev-parse $(git rev-list remotes/$BRANCH --not $(git for-each-ref refs/remotes/ --format='%(refname)' \| grep -v remotes/$BRANCH) \| tail -1)~	18:24
	that worked like charm for finding the point it branched of from	18:24
	there is probably a more elegant way of doing it .. but hell =)	18:25
	mm and for some reason it fails on some of the branches ...	18:27
	→ bronson joined	18:27
	← JensB left	18:28
charon	some of them will be reachable from another branch	18:28
	so the rev-list will be empty	18:29
	→ tanek joined	18:29
Grum	yeah but that only happens when someone did something nasty in svk so it appears	18:29
	→ aziz joined	18:29
	← eletuchy_ left	18:30
	→ eletuchy joined	18:30
Grum	how does gitk list the branches a commit is in?	18:30
	← tanek left	18:31
	← akitada left	18:31
charon	git branch --merged $commit	18:33
	(presumably)	18:33
	er, 'git branch --contains $commit' of course.	18:33
	charon goes to grab a coffee	18:33
	← lorandi left	18:34
	← nick_h[litage] left	18:34
	→ lorandi joined	18:34
	→ tanek joined	18:36
TML	Is there a flag to 'git format-patch' that can make it output them in a format that patch(1) can read, rather than the mbox-ish format? I looked on http://www.kernel.org/pub/software/scm/git/docs/git-format-patch.html, but if it's there, I'm not seeing it.	18:36
doener	patch usually reads that just fine	18:36
	→ FFighter joined	18:37
FFighter	hello	18:37
	→ dreiss joined	18:37
doener	unless your commit message looks like a diff	18:37
FFighter	is this the right place to ask about gitosis ?	18:37
TML	doener: Ah, maybe that's what I did. Thanks.	18:38
	→ ceej joined	18:38
doener	so far, nobody has been blamed for asking about gitosis in here	18:38
	→ da1l6 joined	18:38
	→ Sigma joined	18:39
doener	at least no as far as I'm aware	18:39
FFighter	hehe ok thanks	18:39
TML	doener: I can see how that was stupid of me. Thanks.	18:39
FFighter	how do I specify more than on repo per group?	18:39
	like, can I do: writable = repo1,repo2,repo3	18:39
da1l6	hi	18:40
FFighter	?	18:40
da1l6	how do i get an accidently deleted file back from head state?	18:40
krh	da1l6: git checkout <file>	18:40
doener	FFighter: http://eagain.net/gitweb/?p=gitosis.git;a=blob;f=example.conf;h=87bd822fc17ebfe0132f7060afa8036f84c3be99;hb=master	18:40
da1l6	krh: thanks :)	18:41
doener	krh, da1l6: that's from the index, actually (so it won't work after "git rm")	18:41
	to get the file from HEAD, it's "git checkout HEAD -- file"	18:41
	→ rigel joined	18:41
rigel	does git have support for hardware signing systems like the rsa dongle	18:42
doener	FFighter: see "writable" for group "quux"	18:42
krh	oh yeah	18:42
FFighter	doener, thank you	18:42
	:)	18:42
	← mcella left	18:44
	→ merlin83 joined	18:44
TML	Maybe this doesn't make sense, but can I 'pull' a bare repo? That is, given a remote repo, I want to have a local 'bare' repo that matches it.	18:47
doener	TML: clone --mirror	18:47
krh	TML: git clone --bare, I think	18:47
	oh	18:47
doener	TML: then you can _fetch_ in the bare mirror	18:48
	TML: pull is fetch + merge, and merge doesn't work in a bare repo	18:48
	(well, you can always fetch in a bare repo, not just when created with clone --mirror, but that sets it up as a mirror, obviously)	18:49
	← lorandi left	18:49
TML	doener: I actually want to have the remote repo 'push' to this bare repo in a post-commit hook.	18:50
	← _slacker_ left	18:50
	← ceyusa left	18:50
doener	TML: well, then just "git init --bare" does the trick, as usual...	18:50
	→ jeckel joined	18:50
TML	doener: And then setup the remote and push, rather than 'clone --mirror'?	18:51
	→ tango_ joined	18:52
doener	TML: yep. You might want to set remote.<name>.push to refs/:refs/ as well (in the repo from which you push)	18:52
	← ToxicFrog left	18:56
	→ benn_d joined	18:56
	→ merlin1983 joined	18:57
	← kapone left	18:58
	→ solydzajs_ joined	18:59
	← plediii_ left	19:00
	→ warlock_mza joined	19:00
TML	doener: Sorry, that last bit was mostly nonsense to me. Got something I can read that would explain it?	19:04
	← eventualbuddha left	19:05
doener	the git-push man page	19:05
	→ eventualbuddha joined	19:05
	→ ice799 joined	19:05
TML	k	19:05
	Transitioning to git is either going to make me brilliant or kill me :)	19:06
doener	TML: oh, actually, setting remote.<name>.mirror is smarter	19:06
TML	Is remote.<name>.mirror the same thing as "[remote "<name>"] mirror"?	19:07
	→ kapone joined	19:07
doener	yes. The former is what you give to "git config", the latter is how it looks in the config file	19:07
TML	ok	19:08
	← solydzajs_ left	19:08
	← solydzajs left	19:10
	← merlin83 left	19:10
TML	And that's a true/false?	19:11
	"git config remote.hub.mirror true"?	19:11
	← msanchez left	19:11
	← eventualbuddha left	19:11
doener	yep	19:13
	(see git-config(1))	19:13
	← merlin1983 left	19:13
	→ merlin83 joined	19:13
TML	Ah - I honestly didn't even expect there to be a comprehsive list of all of them.	19:15
doener	some things aren't in there (the send-email stuff, for example), but most are	19:17
	← kinyobi left	19:18
bentob0x	can you add a remote in your 'global' ~/.gitconfig file instead of the local /.git/config file?	19:19
kevlarman	i think so	19:19
	although that sounds a little silly	19:19
	→ ijcd joined	19:19
kevlarman	(if you clone an unrelated project you'll end up fetching a lot of unrelated stuff)	19:20
	← Spacecrack left	19:20
bentob0x	I just use a remote reference when I need it	19:20
	like : git pull myremote master	19:20
	→ drizzd joined	19:20
	→ naeu joined	19:21
	← bobmcw left	19:21
	← Laibsch left	19:22
	← eletuchy left	19:24
	← noth left	19:29
Grum	how can i toss a commit away? (and yes i know i am altering history but that is ok)	19:30
	← chrisdone left	19:30
	→ maelcum\|konv joined	19:31
bremner	git rebase --interactive	19:32
	← sgrimm left	19:33
Grum	but this only works nicely with commits relatively close to head i presume?	19:33
	→ lorandi joined	19:33
	→ PsychoTrahe joined	19:35
	← Yuuhi left	19:36
	← mncaudill left	19:37
	→ dmlloyd_ joined	19:37
	→ noth joined	19:37
	← dmlloyd left	19:37
	dmlloyd_ → dmlloyd	19:37
	qrush → qrush\|away	19:39
	← csgeek left	19:39
	→ dsal joined	19:39
	→ fynn joined	19:40
	→ ryates joined	19:42
	→ AngryElf_ joined	19:43
ryates	I may be missing something obvious, but is there a way to clone an svn repository (aka git svn clone) to a bare repository that can then act as a shared git repo? so I need to rebase the git shared repo against svn occasionally and also push changes to the git repo from other git clients	19:44
	← AngryElf_ left	19:44
Tv	ryates: "rebase" and "shared" don't fit nicely in one sentence :(	19:44
PerlJam	ryates: you can make it a bare repo by hand editing things.	19:45
allan_wind	I have a commit `git show 18b6ecb0c7` and see index 2c62fb8..cab14f3 100644. I am interested in the commit for the blob 2c62fb8. git show 2c62fb8 does give me the right thing, but how do I find the corresponding commit? I tried `git blame 18b6ecb0c7^ but it does not give me the right content. git blame 2c62fb8 -- $file fails with "non commit ...". Any ideas?	19:46
ryates	Tv: that might explain it :)	19:47
parasti	allan_wind: blame takes a treeish and a filename, but you're passing it a blob SHA1	19:47
ryates	my approach was to git svn clone to a temporary git repo, then set up my final repo by doing a clone --bare on my temp repo	19:48
	but then I can't rebase!	19:48
parasti	allan_wind: "git blame 18b6ecb0c7^ -- $file" should do the right thing	19:48
ryates	PerlJam: you mean change the git config files to make it think it's bare?	19:48
PerlJam	ryates: yep (it's a one line change). but you'll also want to get rid of the working files.	19:48
	and hoise the .git dir up to project.git	19:49
ryates	PerlJam: can you still rebase after doing that?	19:49
PerlJam	s/hoise/hoist/	19:49
	ryates: you can, but it's not generally a good thing.	19:49
	→ Hydrogen joined	19:49
ryates	thanks folks - perhaps I should abandon the effort. I was trying to "test out" git while still maintaining my working svn repo. Maybe I should just dive in with both feet.	19:50
PerlJam	But I guess with svn's linearity it's probably okay. I don't know.	19:50
	→ eventualbuddha joined	19:50
ryates	I'd rather deal with a learning curve with git than risk the integrity of my repo, me thinks	19:50
	cbrake → cbrake_away	19:50
PerlJam	ryates: copy the svn repo and turn it into a git repo for play.	19:51
allan_wind	parasti: ahh... it does, I mixed up two files. Why doesn' git show 18b6ecb0c776f08556bf2fb0f6aef2bbd9d9b497^ $file give me the same thing?	19:51
ryates	PerlJam: I'll do that. Thanks again for the help.	19:52
PerlJam	ryates: it's really easy to go from git -> svn and not screw something up, but it's much harder to go the other way.	19:52
	(I mean for working files)	19:52
parasti	allan_wind: "git show $rev $file" shows the differences in $file made in $rev	19:52
PerlJam	you do things in git-land that have no equivalent in svn-land so that just becomes a simple patch.	19:52
ryates	I'm really the only devloper at this time so this whole thing may be overkill - I was just going with a shared git repo to manage re-cloning between my work computer and my laptop	19:52
allan_wind	parasti: how content (blob) instead of diff?	19:53
	how do I get...	19:53
parasti	allan_wind: "git show $rev:$file"	19:53
	→ drewr` joined	19:55
	← eventualbuddha left	19:56
	← lorandi left	19:57
	→ sgrimm joined	19:58
	← willb left	19:59
	→ engla joined	20:01
	→ elyezer joined	20:01
	→ plediii joined	20:01
	→ Toad__ joined	20:01
	→ yann joined	20:02
	← Hydrogen left	20:06
	→ lollan joined	20:09
	← drewr left	20:11
	← lollan left	20:13
	← Toad left	20:13
allan_wind	parasti: subtle... perfect, thanks a ton	20:13
	← da1l6 left	20:13
	drewr` → drewr	20:13
	Toad__ → Toad	20:16
	→ charon_ joined	20:18
	← charon_ left	20:19
	→ archis joined	20:20
	← tango_ left	20:21
	→ porcnick joined	20:21
	← archis left	20:21
	← porcnick left	20:22
	→ porcnick joined	20:23
	→ tango_ joined	20:23
	← unreal left	20:25
	→ unreal joined	20:26
	← dreiss left	20:28
	← topinupa left	20:30
	→ TheRealKerni joined	20:31
	← melvinram left	20:32
	→ pointlessjon joined	20:35
	← bentob0x left	20:37
	← error404notfoun1 left	20:40
	→ Hydrogen joined	20:41
	edwardam → edwardam\|Lnch	20:43
	→ dramsay joined	20:44
	← mfilipe left	20:45
	← warlock_mza left	20:45
	← bryanl left	20:46
	→ moccuo joined	20:49
	→ bobthecow joined	20:49
	→ qixyxexi joined	20:50
	← priidu left	20:52
	→ plediii_ joined	20:52
	→ melvinram joined	20:52
	← benn_d left	20:53
bobthecow	i'm having trouble with git-svn. It's set up, and working, but gets confused every time i pull from another repo.	20:54
	I can't find much on debugging git svn. All my googling returns "how to set it up", not much on "how to fix it"	20:54
	→ charon_ joined	20:54
bobthecow	anyone have debugging tips and/or links for git svn?	20:55
	← d0k left	20:56
	→ error404notfound joined	20:59
	← radii left	21:00
	← plediii left	21:03
	← SRabbelier left	21:03
	→ SRabbelier1 joined	21:04
	SRabbelier1 → SRabbelier	21:04
	← plediii_ left	21:05
	← charon left	21:05
	charon_ → charon	21:05
	→ plediii joined	21:05
parasti	bobthecow: can you give us more info? "pull" isn't something you do in git-svn workflows, and "gets confused" and "another repo" doesn't say a lot about what's happening and what you're trying to do	21:05
	→ perezd88 joined	21:06
bobthecow	sorry. I'll try again :)	21:06
	← aisbaa left	21:06
	← qixyxexi left	21:07
bobthecow	i have three repos, call them: svn, A and B.	21:07
	A was created using git-svn from the original svn repo.	21:07
	B is a clone of A	21:08
	→ lywex joined	21:08
bobthecow	currently the code in A B and the original svn is identical.	21:08
	making a change in B then pulling it to A works great.	21:08
	→ tvw joined	21:09
bobthecow	git-svn fetch and rebase in A works great.	21:09
	← jonnylamb left	21:09
	→ johnw joined	21:09
bobthecow	dcommit works great.	21:09
	basically everything's great.	21:09
drobbins	hiya	21:09
bobthecow	but the next time i try to pull from B to A then dcommit, it wants to recommit all the previous changes too.	21:10
	git-svn rebasing fixes it (with a few warnings about merges)	21:10
drobbins	hey, if I'm applying patches using git am -3, and a patch fails and it says it's unable to fall back to a 3-way merge, is there any steps that could have been taken (either in patch generation or otherwise) that would allow me to fall back on a merge?	21:10
	s/is/are/	21:10
bobthecow	but next time, it'll try all the same things plus any new commits.	21:11
	it's getting absurd.	21:11
	and i'm seeing more merge warnings.	21:11
	→ krawek_ joined	21:11
	← abditus left	21:12
parasti	bobthecow: merging + git-svn don't go well together	21:12
	← elmarco left	21:12
	← krawek left	21:13
	← dramsay left	21:13
bobthecow	parasti: that's not the answer i wanted to hear :)	21:13
parasti	bobthecow: hang on	21:14
	→ peritus joined	21:14
	→ abditus joined	21:14
parasti	bobthecow: git pull is fetch + merge, so I'd suggest you use git fetch + git merge --squash or git format-patch \| git am	21:15
	→ marvil07 joined	21:15
drobbins	parasti: any suggestions for my question?	21:16
	← Niamor left	21:16
	← seb42 left	21:18
	→ offby1` joined	21:18
	→ seb42 joined	21:19
	← Tv left	21:19
parasti	drobbins: not really. all I know about git am merge support can be read in man git-am, the paragraph about -3 :)	21:19
Gitbot	drobbins: the git-am manpage can be found at http://git.or.cz/man/git-am	21:20
	→ Radtoo joined	21:20
	→ eletuchy joined	21:20
drobbins	parasti: ok, no problem :)	21:20
ryates	bobthecow: I just got a similar answer and decided to just convert to using git full time - you can too :)	21:20
jaalto	Is it possible to merge two git repositories? I have code in CVS. Long time ago I set up fresh git repository, moved all CVS head files there, started at ground zro and continued developing with git, forgetting about the CVS. No I would like to import the "old" code somehow. Is this possible?	21:21
	← error404notfound left	21:21
	→ \|Voker57\| joined	21:22
	← Voker57 left	21:22
	← ashleyw_ left	21:24
	→ thiago_home joined	21:24
	← lywex left	21:25
parasti	jaalto: sure. I don't know about Git's CVS support, but I've done this with old tarballs and a new SVN repo	21:25
Radtoo	How do I replace missing/corrupt objects in a local repository with objects from a remote repository?	21:25
bobthecow	ryates: unfortunately, I can't. I have to commit to an existing SVN repo, and I have no control over the decision to switch it to git :(	21:26
	→ bywe joined	21:26
	→ error404notfound joined	21:26
charon	bobthecow, parasti: actually merging works with git-svn as long as you make sure the svn history is the first-parent line	21:27
	← Hydrogen left	21:27
jaalto	parasti: How did you do it? I'm currently running 'git-cvsimport -p x -v -d <pserver-address> project' to convert my old CVS repo into git repo. But I don't know how would I merge this "old" with current devel repo.	21:27
	← orlandov left	21:28
parasti	jaalto: I fetched the "converted" repo into the Git repo and used grafts to glue the histories together	21:29
	jaalto: to make the connection permanent, you'd probably want to use git filter-branch, too (which I never did)	21:30
bobthecow	parasti: Thanks, squashing seems to work.	21:30
	← hobodave left	21:30
	→ mriou_ joined	21:31
jaalto	parasti: What is grafts?	21:31
	→ benn_d joined	21:31
bobthecow	parasti: it's prob'ly better that way anyway. That way the SVN commits can be more concise :)	21:31
parasti	charon: I suspect the problem was that a particular "pull" didn't record a merge commit, so git-svn rewrote all the commits, Git lost track of what was already merged, and chaos ensued	21:32
	→ earcar joined	21:32
parasti	charon: probably wouldn't have happened with a --no-ff	21:32
	← nerochiaro left	21:33
	← offby1 left	21:33
bobthecow	parasti: is there any way to tell git-svn to start over?	21:33
parasti	jaalto: "grafts" is fake ancestry information, stored in .git/info/grafts	21:33
charon	oh this is about two git repos. then chaos will always ensue, as 'git svn rebase' rewrites the merge, but the other repo still has the old one	21:33
bobthecow	charon, parasti: so that will always happen unless changes from the second git repo are merged in as a unique commit, like with --squash ?	21:34
parasti	jaalto: not much info in the docs about it, but it's simple: one line per commit, "$commit $parent1 $parent2 ..."	21:34
charon	bobthecow: or take other steps to fix the rebasing. see the section "recovering from upstream rebase" in man git-rebase. the situation is pretty much the same, except that you can't really blame the upstream, as git-svn has no choice but to rebase	21:35
Gitbot	bobthecow: the git-rebase manpage can be found at http://git.or.cz/man/git-rebase	21:35
bobthecow	charon: thx :)	21:35
	→ Niamor joined	21:36
	← Weaselweb left	21:37
	← drizzd left	21:39
jaalto	parasti: Would you have any examples?	21:39
	parasti: I'm not sure if I understand that "commit parent1 parent2"	21:41
	→ ashleyw joined	21:42
parasti	jaalto: each commit can have 0 or more parents, have a look at "git log --pretty=raw" output	21:42
	→ _Vi joined	21:42
	→ alexis__ joined	21:42
parasti	jaalto: so, for example, you'd see that $parent1 is a parent of $commit	21:42
	jaalto: if you added a line in .git/info/grafts that said "$commit $parent2", that would override the actual parent info	21:43
	jaalto: and the log output would reflect that	21:43
	→ dreiss joined	21:43
	← esparkman left	21:44
	← allan_wind left	21:45
parasti	jaalto: usually there's just one commit in a repo that has 0 parents, the root commit (first commit in the repo)	21:45
	← bywe left	21:45
	← melvinram left	21:46
parasti	jaalto: after fetching your converted-from-CVS repo, you'll have two root commits	21:46
	← ryates left	21:46
	→ mucuqi joined	21:46
	→ nick_h[litage] joined	21:47
parasti	jaalto: so you'll want to connect the first root commit which was already in the repo, to the appropriate commit from the fetched repo	21:47
	→ rovalent joined	21:47
	jaalto listening parasti	21:48
bobthecow	I'm starting to think having a git repo stacked on top of an svn checkout would be a whole lot less hassle than git-svn :-/	21:48
	→ skeeziks joined	21:48
Radtoo	bobthecow: uh... what?	21:48
bobthecow	Radtoo: just hating on the rebase right now.	21:49
skeeziks	How can I list all the branches in a remote git repository, not just the ones that I'm tracking?	21:49
parasti	jaalto: but do note that grafts are meant for temporary such "reparenting". as I said, you'll want to use git filter-branch to rewrite the commit objects and make the connection permanent	21:50
jaalto	parasti: I think I now understood the principle. So I have manually create the .git/info/grafts. It still has remove fetch for the "old". What's the next step to unify these two as single git repo?	21:50
	→ melvinram joined	21:50
	→ felipec joined	21:50
jast	skeeziks: all branches you've fetched? git branch -r	21:50
Grum	jaalto: filter-branch iirc; scary stuff =)	21:50
jaalto	parasti: You were faster. Okay, so I filter-branch everything to new location and that unifies all.	21:50
parasti	jaalto: first, check git log to see if the grafts work	21:51
	← Circuitsoft left	21:51
Radtoo	bobthecow: Rebase works fine and without bad caveats if there's only the central repository you commit against and you don't touch the part that's represented in SVN already.	21:51
skeeziks	jast: Sorry, perhaps tracking wasn't the right term. I'd like to get a list of the branches I haven't fetched - is that possible?	21:51
jaalto	parasti: clear.	21:51
jast	skeeziks: please see man git-ls-remote	21:51
Gitbot	skeeziks: the git-ls-remote manpage can be found at http://git.or.cz/man/git-ls-remote	21:51
	← ph^ left	21:51
Radtoo	bobthecow: That's not very hard to do though... so perhaps you're not using it right yet?	21:52
qrush\|away	skeeziks: git remote show <remote> ?	21:52
skeeziks	jast: Aaaaaah! Thanks. I've been searching and haven't run across that yet.	21:52
	qrush\|away: No, that just shows what I'm tracking I think.	21:52
jast	git remote show is slightly friendlier and does the same thing	21:53
	→ Bombela joined	21:53
	→ AlexanderS joined	21:53
qrush\|away	git ls-remote --heads origin	21:53
parasti	jaalto: oh, that's right, using git filter-branch on a published history is a really bad idea	21:53
qrush\|away	that should do it for branches right?	21:53
	← berniyh left	21:53
qrush\|away	this smells like a good tip :)	21:53
jast	git remote show calls ls-remote	21:53
skeeziks	jast: Perfect, thanks!	21:54
bobthecow	Radtoo: "don't touch the part thats represented in SVN already" ?	21:54
parasti	jaalto: I always forget about "minor details" like that because I'm mostly using git-svn :P	21:54
	← plediii left	21:55
jaalto	parasti: opps. I have already published the new devel repo. So I cannot merge it with old?	21:55
qrush\|away	jast: is there a way to untrack a branch?	21:55
jast	depends on what you mean by tracking	21:56
Radtoo	bobthecow: Yes... basically your local branch that you will push against SVN and pull from has to be arranged nicely from revision 0 to svn trunk - and then your own commits on top of svn trunk.	21:56
jast	if you mean the thing branch --track sets up, just edit the config file	21:56
alexis__	remove the remote in your config file	21:56
jast	it'll have a [branch "foo"] section. remove it; done.	21:57
Radtoo	bobthecow: You never perform any actions against that branch that would mess up that order (rebase it against something else than SVN and all that...) and you'll not have much trouble.	21:57
qrush\|away	Alright, so I'm definitely not tracking some branches yet they're showing in git remote show	21:57
	so it really just checks ls-remote	21:57
	← markelikalderon left	21:57
jast	yes, as mentioned in the manual :)	21:57
	manpage, actually	21:58
bobthecow	Radtoo: but i need to push changes from another git repo into SVN.	21:58
qrush\|away	I just want to make sure what i'm going to write is accurate :)	21:58
	→ berniyh joined	21:58
parasti	jaalto: I think you can keep the graft around and it won't hurt, but filter-branch will overwrite all the commits, causing a lot of pain for anyone who's based any work on your published commits	21:58
Radtoo	bobthecow: one possible workflow is to NOT touch the svn branch at all, and work on another branch. Then right before you want to commit something on SVN, update the svn branch (rebase it) and merge the other branch on top of the svn branch. Then push.	21:58
	bobthecow: (well, dcommit in svn's case)	21:59
	bobthecow: You loose changes that happened "in between" with that of course, but there's not going to be any crazy conflicts that happened because of reordered commits in the other git repos	22:00
bobthecow	Radtoo: right. i don't care about "in between" that much.	22:00
Radtoo	bobthecow: Well, then the above workflow should work. probably.	22:00
bobthecow	Radtoo: and i have an svn branch that i'm not making any changes on.	22:00
parasti	jaalto: you can read more about problems with rewriting published history in man git-rebase (section "RECOVERING FROM UPSTREAM REBASE")	22:00
Gitbot	jaalto: the git-rebase manpage can be found at http://git.or.cz/man/git-rebase	22:00
bobthecow	Radtoo: the problem is that every time i merge the other branch into the svn branch, it tries to apply all previous changes to svn on the next dcommit.	22:01
	→ crowbar joined	22:01
	← crowbar left	22:02
jaalto	parasti: Would you suggest that it would be better to 1) git remote add URL old 2) fetch old 3) push that branch to server (for other to see) 4) document the use of grafs and live with it	22:02
bobthecow	Radtoo: if i do a merge with --squash seems to work fine. it's just when it does ff that it breaks.	22:02
	→ Tv joined	22:03
	→ msanchez joined	22:04
Radtoo	bobthecow: ... I don't exactly get that problem. You mean this happens when you merge on a svn branch that hasn't been rebased right before the merge, or always?	22:04
	← hagabaka left	22:06
	← elyezer left	22:06
parasti	jaalto: I'd go with keeping a separate repo project-old.git and grafts (I actually did something similar with the converted-from-tarballs repo, which is why I never used filter-branch either)	22:07
jaalto	prasti: okay, thanks for the insight	22:07
	→ hagabaka joined	22:07
	← cannonball left	22:08
Radtoo	bobthecow: Because what the svn-tracking branch really should have after the merge is one single additional commit on top of whatever is on the server. And you just like, commit that to the svn server.	22:08
	← a-priori left	22:09
	→ a-priori joined	22:09
	→ r0bby joined	22:09
bobthecow	Radtoo: that's what it does if i use --squash. but if i let it do an FF merge, it tries to commit everything since the initial svn checkout.	22:09
	← Ryback_ left	22:10
	← sporkmonger left	22:11
bobthecow	Radtoo: hmm. i just merged an identical working copy with my svn branch and it did a recursive merge instead of fast forward, and it looks like it created an empty changeset to push to SVN.	22:11
	→ ThaDon joined	22:11
Radtoo	bobthecow: seems all right?	22:12
	← dwave left	22:12
bobthecow	Radtoo: I thought so. then i did svn rebase and it tried to rewind head and replay a bunch of awesome all over the top of it.	22:13
Radtoo	bobthecow: but thats what it should do...	22:14
	edwardam\|Lnch → edwardam	22:14
	→ plediii joined	22:14
Radtoo	bobthecow: I think you merged somewhere in the middle instead of on top of HEAD or something.	22:14
	bobthecow: Would make sense if you used --squash and thus didn't move the HEAD...	22:16
	← felipec left	22:16
bobthecow	Radtoo: right. is there a way to fix that?	22:16
	→ robinr joined	22:16
	→ kate21de joined	22:17
Radtoo	bobthecow: I'd fix it like this: Make sure you got the history and most recent commit in your other, non svn-tracking branch. Then hard-reset the svn branch by as many commits as it takes for it to rebase without interaction.	22:17
	bobthecow: That way you'll be properly "mirroring" the svn repos again, without any altered history (or at least none that matters anymore).	22:18
	← plediii left	22:18
	← qrush\|away left	22:19
	→ qrush joined	22:19
	← kate21de left	22:19
Radtoo	bobthecow: Perhaps it won't be unwise to use checkout on the branch just to make very sure it points to head after rebasing. (I can't remember if it always does). And then you can merge. :)	22:20
charon	where on earth does the libgit2 discussion hide out? shawn only seems to host the api docs, even the repo is gone from his site. google isn't very helpful, or i'm doing it wrong. there's no readme in the repo either.	22:22
	← LiamH left	22:22
	→ orlandov joined	22:22
Radtoo	bobthecow: The dcommit then really should be like, one revision committed, nothing else to report... (unless someone else committed something in the meantime but then the auto-merge / manual merge cycle just happens once or as many commits as are needed, not the insanity of possibly replaying half of the repository's history under the fun assumption that the revision in the middle was different)	22:22
	← cemerick left	22:23
	→ Hydrogen joined	22:24
bobthecow	Radtoo: okay. I've rebased it, it says my svn branch is up to date. I merged in my working branch (which is actually identical to the svn branch) and it created a recursive merge, which appears to be empty.	22:25
	Radtoo: dcommitted that and got 'No changes between current HEAD and refs/remotes/git-svn'	22:26
	→ dmlloyd_ joined	22:26
bobthecow	Radtoo: that's a good thing, yeah?	22:26
	← jrmuizel left	22:26
Radtoo	bobthecow: Perfect. Now as long as you keep it in sync with svn prior to merging its gonna be easy.	22:26
bobthecow	Radtoo: so the process is... changes in working branch > head over to svn branch > svn rebase > merge in working branch > svn dcommit?	22:27
	→ plediii joined	22:28
Grum	is there an easy way to recognize a completely empty commit? (no permission/data/whatever changes at all?)	22:28
thiago_home	the tree is still the same	22:29
	git commit refuses to commit like that, but you can create them with other tools	22:29
Radtoo	bobthecow: Well, I merge things on top of the SVN instead of the working branch but I think it matters not. :)	22:29
Grum	yeah git svn created them	22:29
Radtoo	bobthecow: Might actually be better as you say it... otherwise you could forget to get the changes from svn into your working branch	22:29
bobthecow	Radtoo: so what happens when svn changes, and i need to get things back? right. that's what i was asking.	22:29
Grum	so thiago_home: i should diff the git ls-tree $commit vs git ls-tree $commit^ ?	22:30
Radtoo	bobthecow: Yep. Its very likely better. :)	22:30
maelcum\|konv	!time	22:30
	→ kate21de joined	22:30
	→ eddyp joined	22:30
	→ antirobotrobot joined	22:32
	← joshdavey left	22:32
antirobotrobot	how can i have git-archive include submodules?	22:32
Radtoo	bobthecow: Well, I think you got both what happened and why it happened, right? :)	22:32
bobthecow	Radtoo: i sure hope so. If I didn't, I	22:33
	'll be back :)	22:33
charon	Grum: there's diff-tree that does that for you	22:33
Grum	ok; so if that output is nothing i can safely assume the commit is completely not needed?	22:34
Radtoo	bobthecow: Sure. Glad to have helped, have fun with git.	22:34
	← Bombela left	22:34
Grum	i'm trying to find out if i can safely purge a branch (for some reason i got a bunch of them without any changes)	22:34
	→ lolage joined	22:34
bobthecow	Radtoo: Thanks for all the help.	22:35
	Radtoo: now how do I keep it from making SVN commit messages like 'Merge commit 'origin/master'' ?	22:37
charon	Grum: yes	22:38
Grum	charon: thanks; also what is the best way to rename a branch without leaving a trace of me doing so?	22:38
charon	Grum: though you may not want to drop a merge even if it is the same as all its parents :)	22:38
Grum	moving it in .git/refs ?	22:38
Radtoo	bobthecow: well, you can specify a message when merging (-m or something, or perahaps you can leave it uncommited and just specify a commit message) OR use git rebase --interactive HEAD~1 to rewrite the last commit :P	22:38
Grum	charon: its mostly for branches that are just useless	22:38
	the svn import created a bunch of them and they are just empty commits	22:39
Radtoo	bobthecow: Oh and of course git commit --amend...	22:39
Grum	so i'm going to loop over all branches; see where they came from and then check if there is just 1 commit and if that one commit contains no data .. purge	22:39
	i only get one chance to clean it up properly .. at the migration itself ;)	22:40
charon	Grum: why would you want to move them with no trace? ... i think you'll have to trash the reflog as the old entries are bound to contain some hints of the old name	22:40
	← dmlloyd left	22:41
Grum	mmm	22:41
	dmlloyd_ → dmlloyd	22:41
	← RaySl left	22:42
Grum	hmm so i should just rename them using git branch? (i was planning to move all the refs/remote/xxx)	22:42
	← msanchez left	22:43
	← kapone left	22:43
Grum	they should be renamed to refs/heads/xxx right? (in order to appear as 'normally created branches)	22:43
charon	if you're not afraid of trashing any existing refs/heads/, saying 'git fetch . refs/remotes/:refs/heads/*' will do that in one go	22:44
	well, not rename but copy.	22:44
Grum	ok but then i still need to git branch -D -r for every remote/x right?	22:45
	← drewr left	22:45
	← faithfulgeek left	22:45
Grum	mmm i have to loop over them individually to inspect them (to see if they arent empty and thus worthy of saving) so i might as well just rename them	22:46
charon	i'm not even sure it lets you rename across hierarchies. but in any case you'll have to delete them individually unless you resort to 'rm .git/refs/remotes/*'	22:47
	(and similarly for the logs)	22:47
Grum	lets try ^^	22:48
	→ johan-s joined	22:48
	→ orlandov_ joined	22:49
Grum	error: refname refs/heads/remotes/$BRANCH not found ... hmmm	22:49
	← ashleyw left	22:50
	← nick_h[litage] left	22:54
	→ bdimcheff joined	22:54
	← a-priori left	22:54
	← cnu left	22:55
Grum	oi that wasn't planned; it seems there are packed-refs as well	22:55
	← dmlloyd left	22:56
charon	man git-update-ref	22:57
Gitbot	charon: the git-update-ref manpage can be found at http://git.or.cz/man/git-update-ref	22:57
charon	note that the first argument must be the full ref name, usually starting with refs/	22:57
Grum	that works for both packed and unpacked refs i presume?	22:58
charon	yes	22:58
Grum	and for the logs?	22:58
SRabbelier	Is there a way to list all files that were modified since a certain time?	22:59
	→ sporkmonger joined	22:59
charon	Grum: deleting also deletes the reflog	22:59
	→ SLiMIT joined	23:00
charon	SRabbelier: maybe git diff --name-only $(git rev-list --first-parent --before=yesterday -1 HEAD) HEAD	23:01
	← boto left	23:01
	← krh left	23:01
	← Modius left	23:01
charon	though "since <time>" generally isn't very well defined	23:02
	← maelcum\|konv left	23:02
	→ dmlloyd joined	23:02
SRabbelier	charon: awesome, that seems to work :)	23:02
	→ Modius joined	23:02
	← cmarcelo left	23:04
	← moore_ left	23:04
	→ moore_ joined	23:05
	← peritus left	23:06
	← Sigma left	23:07
	← Arrowmaster left	23:07
	← NSaibot left	23:07
	→ Arrowmaster joined	23:08
SRabbelier	charon: mhhh, now how do I list only the files that were created since a certain date? :d	23:08
charon	--diff-filter=A	23:09
	← perezd88 left	23:11
	← naeu left	23:11
	← orlandov_ left	23:12
SRabbelier	charon: you are made of win :)	23:13
	→ krh joined	23:13
Grum	charon: for ref in $(git for-each-ref refs/remotes --format='%(refname)'); do git update-ref $(echo $ref\|sed 's#remotes#heads#') $(git rev-parse $ref); done	23:13
	that seems to so what i meant; now to just clean up the old refs	23:13
	→ ToxicFrog joined	23:17
	← orlandov left	23:17
	← Hydrogen left	23:18
	→ Hydrogen joined	23:19
	← Hydrogen left	23:20
	← antirobotrobot left	23:20
Grum	and it seems i can only do that with rm -rf .git/refs/remotes .git/logs/refs/remotes -- oh well .. starting to look nice now ;)	23:22
	← schierbeck left	23:22
	→ a-priori joined	23:23
alexis__	hi guy, I'd like to know how to do a "push --mirror" from a non-bare repo to another without having to create a local refs per remote branches?	23:23
	← pointlessjon left	23:24
	← ProperNoun left	23:24
	← PuffTheMagic left	23:26
	→ PuffTheMagic joined	23:26
	← loincloth left	23:27
	→ loincloth joined	23:28
	→ ashleyw joined	23:30
	← barjunk left	23:32
	← a-priori left	23:32
	← earcar left	23:33
	← bobthecow left	23:34
	→ cnk joined	23:34
	← cnk left	23:34
	→ ProperNoun joined	23:34
charon	Grum: well, update-ref can also delete	23:34
	→ orlandov joined	23:35
Grum	mmm how?	23:38
	→ ciskje joined	23:39
	← kate21de left	23:39
	← edwardam left	23:39
	→ kate21de joined	23:39
	← kate21de left	23:43
	← sporkmonger left	23:43
	← nothinghappens left	23:43
charon	Grum: update-ref -d refs/foo	23:44
	← dsal left	23:44
	← welly left	23:45
	→ davik joined	23:46
	← TheRealKerni left	23:46
	← charon left	23:47
	← loincloth left	23:47
	→ welly joined	23:47
	← patrikf left	23:48
	← jboyens left	23:49
	← orlandov left	23:49
	← rovalent left	23:50
	← krh left	23:51
	dwmw2 → dwmw2_gone	23:54
	← levifig left	23:54
	→ moore__ joined	23:55
	← EmilMedve left	23:56
	← moore_ left	23:56
	← chris2 left	23:58

Logs Search ←Prev date Next date→

Channels Documentation