IRCloggy #git 2008-05-19

Provider of IRC logs since 2005.

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

	2008-05-19
halorgium	git push origin --all	00:00
Ilari	corporeal: Shared flag is for overriding too restrictive umasks. It also forces newly created directories to have setgid bit set...	00:00
halorgium	that will push all the local branches	00:00
Dreamer3	ah	00:00
	thanks	00:00
halorgium	also, --tags	00:00
Ilari	corporeal: You know how groups work in Unix and what setgid bit does to directories?	00:00
halorgium	setgid is the suck for trying to be awesome	00:01
	if only setfacl was more awesome	00:01
	← p28301016 left	00:02
corporeal	Ilari: um, well i know how groups work of course but im unaware of what setgid is	00:02
Ilari	corporeal: Basically, if directory has setgid bit set, all files created in it will inherit the group of directory they are created to.	00:03
corporeal	Ilari: well, that's useful...	00:04
Ilari	corporeal: Without it, most systems will use user's EGID (FSGID on Linux) as group for the newly created file.	00:04
corporeal	aye	00:04
felipec	if I have two totally unrelated repos, let's say "main" and "john", and I want to merge "john" into "main", but I already have the conflicts solved in a working directory	00:05
corporeal	been using and administrating linux for... god knows how long and i never came across that	00:05
felipec	is there any easy way to do that?	00:05
	→ gef joined	00:06
Ilari	corporeal: After using 'git --bare init --shared' to create the repo, do recursive chgrp on it to change it to proper group. I once tried to first set the directory group and then create the repo, but the result was pretty much a mess...	00:07
corporeal	Ilari: okiedoke. i typically do a recursive chgrp after doing any major changes to a directory like that anyyway	00:07
halorgium	felipec: look at subtree merge	00:09
Ilari	'pretty much a mess' meaning that some files had inherited the group, and some not...	00:09
	→ osamu2001 joined	00:10
	← fhobia left	00:10
	→ radarek joined	00:11
	← ruediger left	00:11
	→ fhobia joined	00:12
	← osamu2001 left	00:12
	← markkalderon left	00:12
felipec	I'm doing git fetch "john.git"; git merge FETCH_HEAD; rm ; cp -r merged_wd/ .	00:13
	→ mtodd joined	00:14
felipec	halorgium: I don't think I want a subtree	00:14
halorgium	felipec: that'll get the tree into the repo	00:14
	are you not concerned about the second repos history?	00:14
felipec	halorgium: yes	00:15
halorgium	oh, i see	00:15
	i like history :)	00:15
felipec	halorgium: I mean, I'm concerned	00:15
halorgium	oh, you are ...	00:16
	so you want to have 2 repos (their contents and history) merged into one repo?	00:16
	overlayed	00:17
felipec	halorgium: yes, after that I don't care about the second repo	00:17
	← radarek left	00:17
felipec	halorgium: there will be only one	00:17
halorgium	i believe it is "there can only be one" :)	00:17
felipec	halorgium: yes, but there will be no more merges	00:19
	→ markkalderon joined	00:20
halorgium	felipec: are the repos downloadable?	00:22
felipec	halorgium: what do you mean? you want the repos?	00:23
	halorgium: actually, I'm creating all the repo history manually (converting from monotone)	00:24
	but some branches don't have any parent	00:25
halorgium	felipec: haha, i see	00:26
	yer, i was gonna try it for ya	00:27
	i would get it all into git first	00:27
	then join the 2 repos	00:27
felipec	halorgium: like this? git fetch "john.git"; git merge FETCH_HEAD;	00:28
	→ musix joined	00:28
halorgium	you can't merge a commit which doesn't have a common ancestor ..	00:28
	← sbeyer left	00:29
Mikachu	sure you can	00:31
	← sgrimm left	00:31
felipec	yeah, I think I just did	00:31
Mikachu	you'll get a mix of the files	00:31
	← musix left	00:31
	→ sgrimm joined	00:31
Mikachu	there are even some of those merges in git.git (gitk and git-gui)	00:31
felipec	the problem is that if there are conflicts I don't want to manually fix them, since somebody already did	00:32
	← sgrimm left	00:32
Mikachu	5569bf9bbedd63a00780fc5c110e0cfab3aa97b9 is the first such merge	00:32
	felipec: if someone merged and resolved the conflicts you want to pull from them	00:33
	→ sgrimm joined	00:34
felipec	Mikachu: in monotone	00:34
Mikachu	then i don't know :)	00:34
	you could run the merge, then dump the files in the workdir and add everything and commit?	00:35
felipec	Mikachu: yeah, that's what I'm doing, but for some reason the body of the merge is null	00:35
	← kanru left	00:36
doener_	felipec: body of the merge?	00:36
	→ igorgue_ joined	00:37
felipec	doener_: the diff	00:37
Mikachu	the diff of what?	00:37
mugwump	does 'git-whatchanged -1 -m' on that merge commit show anything?	00:37
doener_	felipec: against which parent?	00:37
Mikachu	more specifically	00:37
	isn't -1 quite new?	00:38
mugwump	no	00:38
felipec	mugwump: yeap, lots of changes	00:38
Mikachu	oh, i only saw the commit for adding the bash completion recently	00:38
felipec	mugwump: but git show isn't showing anything	00:39
Mikachu	felipec: were there any conflicts?	00:39
	felipec: do any of same files exist in both branches before merging?	00:39
felipec	Mikachu: yes, lots	00:39
Mikachu	i see	00:39
	i don't know if diff --cc works when there's no common parent	00:41
	→ heythisisdave_ joined	00:42
	← heythisisdave left	00:42
	← goibhniu left	00:43
felipec	git show 5569bf9bbedd63a00780fc5c110e0cfab3aa97b9	00:43
	that is not showing any body (git repo)	00:44
Mikachu	that was a subtree merge so it didn't have any conflicts	00:44
felipec	ah, true	00:44
Mikachu	i guess even if there are conflicts, a --cc diff might be empty depending on the resolution	00:45
felipec	Mikachu: if you do a git show on the initial commit you still see contents	00:46
	a diff from /dev/null	00:47
	← bdiego left	00:47
mugwump	I think without specifying -m you don't get diffs	00:47
	ok, so, if git-whatchanged on the commit is showing changes, it's showing you the way that the merge was resolved	00:47
	→ saintdev joined	00:47
	← tomtt left	00:47
mugwump	in this case, files were cleanly picked from one side it looks like	00:48
felipec	mugwump: :100644 100644 824f7bb... da4d52d... M libpurple/debug.c	00:48
	mugwump: I've always used git show to see the diffs	00:48
	← jeffpc left	00:49
mugwump	well, a diff with two bases is a bit tricky. GNU diff never handled that	00:49
	you might get something with git-show -m for a merge	00:49
doener_	mugwump: depends, eg. eafa29b7cb27fa0e14d9629e66d6866292620113 gets a diff with "-p -c" but not with -m	00:49
	... and neither with -p -m	00:49
felipec	ah, yeah, -p -c works here too	00:50
mugwump	ok. well, at least gitk can display them pretty well :)	00:50
Mikachu	if you do a merge and the resolution requires you to change some surrounding code, you'll see the diff for that change since it wasn't in any of the parent branches	00:50
mugwump	or you can use diff and specify which parent you want to diff against, ^1 to mean the left one or target branch, and ^2 to mean the right one or merged in branch	00:50
Mikachu	or if you actually do something completely unrelated, but that's a bit silly to do in a merge commit	00:50
	→ ekidd joined	00:51
	← sgrimm left	00:52
felipec	a merge has two parents, but it's executed in only one of them, right?	00:52
	→ sgrimm joined	00:52
	doener_ wonders how many svn "merge foo and ..." commits exist	00:52
doener_	felipec: "executed"?	00:52
	← igorgue left	00:53
felipec	doener_: ah, nevermind, I was going to say that git show should display the diff from the main parent	00:53
Mikachu	git doesn't really care about "main" parents	00:54
doener_	except for --first-parent ;-)	00:54
	though that's pretty unrelated to diff display	00:55
	← BabelO left	00:57
Mikachu	i guess if you use git log -p --first-parent it could default to diffing against the first parent ;)	00:57
	→ jeffpc joined	00:59
	← retroj left	01:01
felipec	is it possible to add some custom field? I would like to add the sha1 from monotone	01:03
	→ bootie joined	01:05
	→ ekidd_ joined	01:06
	← bootie left	01:06
	← gitte left	01:09
	← holycow left	01:15
	← felipec left	01:16
	igorgue_ → igorgue	01:17
	← fujin left	01:20
	→ osamu2001 joined	01:22
	→ fujin joined	01:23
	← ekidd left	01:25
	← sgrimm left	01:25
	→ sgrimm joined	01:25
	→ jmworx joined	01:27
jmworx	is it possible to change the timestamp of a commit?	01:27
	← ekidd_ left	01:28
	← sgrimm left	01:28
	→ sgrimm joined	01:30
	← geezusfreeek left	01:32
	→ bdiego joined	01:35
	← jerbear left	01:46
	→ ekidd joined	01:48
	← ggeecko left	01:51
	→ ggeecko joined	01:51
	→ spearce joined	01:55
	← zdennis left	01:57
	← flaguy left	02:06
	← ajonat left	02:08
	→ ajonat joined	02:08
Zyclops	i've got two git repositories, is there an easy way to merge them? or do i push from one to the other?	02:19
spearce	pull from one into the other; the one you are pulling into should have a working directory so we can merge the files and you can fix conflicts (if any).	02:20
Zyclops	i'm going through git gui.. there does not seem to be a pull option	02:21
	is it command line only?	02:21
spearce	fetch, then use merge->local merge	02:21
Zyclops	right ok :)	02:21
	thankyou	02:21
spearce	pull on the command line is the same as those two gui actions, strung back to back. :)	02:22
Zyclops	there does not seam to be a fetch either.. i'm checking in the command line utisl	02:23
spearce	remote->fetch from->origin ?	02:23
Zyclops	not on my version.. doesn't matter i'll try using the command line	02:25
spearce	oh, you must not have a remote configured. or are using a pretty old git-gui and don't have a remote configured. yea, command line it is.	02:25
	← heythisisdave_ left	02:26
	→ geckosenator joined	02:27
	→ daaku joined	02:31
	← osamu2001 left	02:32
	← ec left	02:32
	← ajonat left	02:34
	→ krawek joined	02:37
	→ kewark joined	02:41
	← threeve left	02:41
	← doener_ left	02:42
	→ garrin joined	02:45
	← ekidd left	02:46
	→ ajonat joined	02:48
	→ ekidd joined	02:48
ajonat	hi! does anyone know if there are any plans to support resumable git-clone?	02:50
spearce	not really. its actually harder to implement then it sounds.	02:50
ajonat	shawn pearce?	02:51
spearce	yes	02:51
ajonat	I read your email and gsoc entry :)	02:51
	I was actually looking more info if you're doing it :)	02:52
	← madewokherd left	02:52
spearce	no, we aren't. we received project proposals for it for gsoc, but other ideas seemed to be higher priority and were pushed up higher in the stack, and we ran out of slots for students.	02:52
	i'm also not working on it myself; most of my "git time" goes into jgit these days.	02:53
ajonat	oh :(	02:53
jmworx	Anyone knows how to either set a date in a commit or change the date of a commit?	02:53
spearce	one reason it wasn't given a slot is gittorrent was given a slot. someone should be able to resume a clone easier through gittorrent, once that project is implemented. and resumable clone is mostly wanted on the really big projects, which are also ideal candidates for distribution by gittorrent. so we felt two-birds-one-stone approach was acceptable for this year given our limited slot allocation.	02:54
	← krawek left	02:55
igorgue	jmworx: trying to cheat at work?, but seriously, I believe that's the timestamp of the commit object itself	02:55
spearce	jmworx: you have to rewrite the commit to change the date; that changes its SHA-1 and all SHA-1s of all commits after that. but in general you can force a different date by setting the GIT_COMMITTER_DATE or GIT_AUTHOR_DATE environment variables when you invoke git-commit.	02:55
jmworx	igorgue: no, trying to copy data from another repository while preserving history the best I can	02:56
spearce	igorgue: no, its not the date of the object. its the date the object was created.	02:56
	spearce assumes igorgue was thinking modification date of the loose object file in .git/objects/??	02:56
igorgue	ahh I'm sorry I'm wrong	02:56
	spearce: yeah I was thinking on that	02:56
ajonat	spearce, gittorrent sounds nice for resuming clones.. thanks for the info!	02:56
jmworx	I don't mind if it means I need to undo the commit and do it again	02:56
spearce	jmworx: yea, you'll have to recreate the commit. set GIT_COMMITTER_DATE to change the committer date; GIT_AUTHOR_DATE to change the author date.	02:57
	kewark → krawek	02:57
jmworx	spearce: Those are env variables or fields in a file?	02:58
spearce	env vars	02:58
jmworx	also, what's the format?	02:58
spearce	email rfc blah blah blah. whatever the email date format is.	02:58
jmworx	I mean for the date	02:58
	ah, OK	02:58
	← mordaunt left	02:59
spearce	it'll accept a lot of weird crap, but in general we set GIT_AUTHOR_DATE using the "Date:" field of an email message during git-am, and internally its the same code to parse both _DATE vars.	02:59
fujin	rfc2822, isn't it?	03:06
	→ blix_ joined	03:07
blix_	in git-rebase, what are the reset and tag commands for?	03:08
	there doesnt seem to be any info in the man pages	03:09
	← LiamH left	03:10
fujin	blix_: I'd go so far to assume that they reset (to) or tag that commit	03:10
	naut but speculation, though;	03:11
blix_	well, yeah, but there isn't any documentation. should it be a hard, soft or mixed reset? what are the parameters to the tag command?	03:11
	can tag only do simple tags or can they get annotated/signed etc?	03:12
	if a chicken in a half laid an egg and a half in a day and a half, how long would it take a monkey with a wooden leg to kick all the seeds out of a dill pickle?	03:12
	there needs to be more documentation on this stuff	03:12
	:p	03:13
fujin	no way! just read the code!	03:14
	fujin smirks	03:14
	← brosner left	03:14
spearce	blix_: those reset and tag commands were only very recently added to git-rebase--interactive.	03:18
	the reset is _only_ a hard reset, as its meant to allow the rebase process to switch to another branch and rebuild that branch, prior to issuing a merge to connect two just-rebased branches together.	03:19
blix_	ok, cool	03:19
	fujin goes crosseyed	03:20
fujin	and how about the tag command, spearce ?	03:20
spearce	the tag is meant to allow recreating an annotated tag if the commit one points at is being rewritten. which is sort of nuts, but someone wanted it and added code for it.	03:20
blix_	that one's easy	03:20
fujin	ah	03:20
blix_	ok, i guess ill have to look up the syntax for that one, its not straight forward	03:21
	or maybe i just wont do it	03:21
	i have that option at this point	03:21
	thanks	03:22
	← kukks left	03:24
	← markkalderon left	03:24
	← Sonderblade left	03:26
	← saintdev left	03:26
	← csc` left	03:29
	→ brosner joined	03:36
	← schacon_ left	03:43
	→ mithro joined	03:43
	← cmarcelo left	03:43
	→ schacon joined	03:47
	← ekidd left	03:58
	← bdiego left	04:00
	→ p28301016 joined	04:00
	→ uhop joined	04:01
uhop	@Tv: after upgrading ubuntu from gutsy to hardy gitosis started to ask for a password, fresh install of gitosis didn't help --- how to troubleshoot? maybe you know the solution already...	04:04
	← brosner left	04:06
spearce	maybe the /etc/ssh/sshd_config file isn't allowing public/private key authentication and is forcing password authentication only?	04:08
halorgium	or the permissions on the .ssh directory in the git user is broken now	04:08
uhop	@spearce: good idea. let me check...	04:08
halorgium	or it decided to eradicate your authorized_keys file cause it had a whole bunch of weak keys	04:09
uhop	@halorgium: it is possible too --- some ssh security hole was found and fixed before hardy...	04:09
	→ tjafk2 joined	04:10
halorgium	uhop: yes, that is what i was referring to	04:10
spearce	is that a new feature? now that they have shipped broken key generators they now ship broken key eradicators? "since your key is broken due to our patch-everything process you now must drive 500 miles to your remote server to generate install a new admin account key. hope you have a nice drive."	04:10
uhop	:-)	04:10
fujin	ssh remoteserver 'ssh-keygen -t dsa'	04:11
	qq	04:11
halorgium	spearce: haha, probably	04:13
	→ heythisisdave joined	04:15
	← Judofyr left	04:21
	→ priidu joined	04:22
	→ brosner joined	04:24
	← tjafk1 left	04:26
	→ a-priori_ joined	04:27
	← statim left	04:27
uhop	yep, authorized_keys got hosed somehow...	04:27
	thx spearce and halorgium!	04:28
halorgium	uhop: no problem	04:28
	← p28301016 left	04:29
	→ csc` joined	04:30
	→ osamu2001 joined	04:36
	← osamu2001 left	04:36
jmworx	How do I manually fix merge conflicts instead of using git-mergetool?	04:40
	← a-priori left	04:43
manveru	jmworx: maybe using git-update-index ?	04:47
jmworx	ah	04:48
	← phillips left	04:48
tflsh	is there a way to see if there is anything to pull, without having to change the current working files?	04:50
DrNick	fetch and then diff	04:50
tflsh	DrNick: but what if i want to get rid of those fetched files (eg if i didnt really want to pull the files)	04:51
DrNick	then delete that branch and let garbage collection take care of the objects	04:51
	→ statim joined	04:52
tflsh	what is the name of a fetch'ed branch	04:53
	i thought it didnt get a name	04:53
	hm	04:53
DrNick	depends on how you invoke fetch	04:53
	→ kewark joined	04:54
pasky	295304 files to consider	04:54
	that many files in repo.or.cz's /srv/git :)	04:54
	← schacon left	04:55
	→ ajonat_ joined	04:55
	← krawek left	04:55
	kewark → krawek	04:55
pasky	root 16023 0.0 0.0 4764 820 ? Ss 2007 18:39 /sbin/sshd	05:00
	it hurts my heart to kill this now :)	05:00
	I mean, "2007", aaw ;)	05:00
	← spearce left	05:01
comp	pasky: well .. it can't run forever ...	05:02
	← igorgue left	05:04
pasky	yeah, such is the circle of life i guess ;)	05:05
	← broonie left	05:06
pasky	hmm i guess i can leave it still running for the first rsync pass though :)	05:06
	→ broonie joined	05:06
	→ mediogre joined	05:07
Tv	pasky: for atomicity? you could just kill -STOP it	05:07
	pasky: or firewall it	05:07
blix_	whats the proper syntax for a push line in the .git/config file (for a remote). is it "push = refs/heads/:refs/heads/"?	05:08
pasky	Tv: well once it's down it won't go up again, since i'm moving repo.or.cz to another machine	05:09
Tv	ah	05:09
	→ ph^ joined	05:09
	← ajonat left	05:11
blix_	when i push to this repo, i want one branch to be a force, and the rest to be FF only	05:12
pasky	okay	05:12
	it seems the new machine is kinda broken	05:12
	the rsync is stuck in sync_buffer(); the other processes stuck in D are kjournald and pdflush, but I don't know if that isn't okay since these are kernel threats, who knows what is going on...	05:13
	any tips on how to find out what's going on?	05:13
	← npmccallum left	05:15
nkallen	i have a particular revision that I want to make be HEAD... how do i do that?	05:16
Tv	nkallen: git checkout foo	05:16
nkallen	Tv: did that	05:16
	but i want that to be HEAD on a certain branch	05:16
pasky	(echo -n 'w' > /proc/sysrq-trigger works nice - looks like it's stuck in raid_release_stripe)	05:16
Tv	git checkout -b branchname foo	05:16
pasky	:/	05:16
	← ajonat_ left	05:17
pasky	i'll try to upgrade to newer kernel maybe	05:17
	→ nickh_ joined	05:17
nkallen	Tv the branch already exists	05:17
	basically.. i was mucking around on master. i checked out a previous revision and made commits on top of that	05:17
	i want these latest commits to be the new HEAD of master.	05:17
Tv	nkallen: many would say git checkout branchame && git reset --hard foo	05:18
	nkallen: i tend to say git branch -D branchname && git checkout -b branchname foo	05:18
nkallen	the reset worked. i actually kinda understand why that works	05:19
	what does -D do?	05:19
	oh you destroy master?	05:19
Tv	man git-branch	05:19
nkallen	i guess it doesn't matter they're all just pointers right?	05:19
	← brosner left	05:20
	← danlarkin left	05:25
	← jmalonzo left	05:25
	→ Overshard joined	05:26
	← ph^ left	05:27
	→ ph^ joined	05:28
	→ saintdev joined	05:31
	→ mstormo joined	05:33
	→ angasule_ joined	05:34
	→ freshtonic joined	05:39
	← uhop left	05:43
	← angasule left	05:51
	→ aarapov joined	05:52
	→ Modius__ joined	05:55
	→ nud_ joined	05:56
	← Tux32 left	05:59
Ilari	pasky: I see that a lot. Maybe some suboptimal settings?	05:59
	→ robinr joined	06:01
	→ spearce joined	06:02
pasky	well, by "stuck" i mean pretty much deadlocked	06:02
	unless it's supposed to get stuck there for 20 minutes :)	06:02
Ilari	pasky: How did you see the time since last run?	06:02
pasky	huh?	06:03
	oh.. i know i sent it SIGKILL 20 minutes before ;)	06:03
	← blix_ left	06:03
Ilari	pasky: You SIGKILL'ed rsync, and after 20 minutes it hasn't died?	06:04
pasky	yes	06:04
	→ blix_ joined	06:04
Ilari	pasky: Sound like some kernel bug then...	06:04
pasky	now it's back up with newer kernel	06:04
	well yeah, that's the point ;)	06:04
	rsync is running again now, hope it owrks realiably now	06:04
	and /me is off to the city	06:04
	← Modius left	06:06
	← heythisisdave left	06:08
	← chrisb left	06:13
	→ deavid joined	06:13
	← dwave left	06:16
	→ _zack joined	06:16
	→ kewark joined	06:19
	← krawek left	06:21
jones-	Hello. How can I have git-status be automatically piped to less just like git-diff and git-log?	06:22
	(preserving colors, as well)	06:22
	← spearce left	06:23
	← a-priori_ left	06:23
	← robinr left	06:32
	Modius__ → Modius	06:34
	→ etnt joined	06:35
	kewark → krawek	06:37
	← Ademan left	06:42
	→ bentob0x joined	06:45
	← statim left	06:45
	← thiago_home left	06:54
	← blix_ left	06:54
	→ vbgunz joined	06:58
	← krawek left	06:59
	← aarapov left	07:01
	← SuttoL left	07:03
	→ cypher23 joined	07:03
	→ devogon joined	07:06
	← mediogre left	07:07
	→ pasky joined	07:23
	→ alenoosh joined	07:23
	→ goibhniu joined	07:23
alenoosh	hi all , is there any git command to avoid tracking some system specific folders or files ( like config files or cache folders )	07:24
maio	alenoosh: man gitignore	07:26
alenoosh	for example , when I work with the system , some changes happen to cache folder of my repository but I don't need to commit them but I need the folder cache itself , how can I avoid tracking its contents ?	07:26
	maio, ok I'll check that	07:26
	maio, why do I get "command not found" by running gitignore ?	07:32
jengelh	folders are windows.	07:32
	← etnt left	07:33
hliusv561	alenoosh: It is not a command, it is a file...	07:33
maio	alenoosh: I said 'man gitignore'	07:34
alenoosh	hliusv561, where is	07:34
	maio, yes I did that but sorry I thought it's a command	07:34
	hliusv561, I don't have .gitignore file	07:37
	→ reel_ joined	07:38
	→ statim joined	07:39
mugwump	alenoosh, create it - see the examples on the man page you were referred to	07:40
	→ aarapov joined	07:40
	→ angasule__ joined	07:41
	→ redondos joined	07:41
	← TuxCM- left	07:42
	→ ruediger joined	07:42
	← _zack left	07:43
	→ etnt joined	07:48
	← angasule_ left	07:55
	← nominolo left	07:59
	→ lirit joined	08:02
	→ _zack joined	08:04
	← seanius left	08:04
	→ seanius joined	08:04
	→ koke joined	08:10
	← fhobia left	08:14
	→ angasule_ joined	08:17
	→ visik7 joined	08:17
	→ redondos_ joined	08:19
	→ Ace_NoOne joined	08:19
Ace_NoOne	is there something like "git svn status" - i.e. comparing local working copy with the SVN repo?	08:20
visik7	is git good to merge a part of a branch into another ? ppl on bzr told me that bzr is not really good in this kind of work	08:20
Ace_NoOne	because git status just compares to the latest local commit, right?	08:21
jmspeex_	Ace_NoOne: well, you can always "git svn fetch" first	08:21
Ace_NoOne	jmspeex_: not sure I follow	08:21
	← redondos left	08:24
	redondos_ → redondos	08:24
	→ namenlos joined	08:28
nud_	visik7: what do you mean with "a part of a branch" ?	08:29
visik7	nud_: for example I've some features developed on a branch and I want to port to the other branch only some of those	08:30
context	ace_noone. uhh it compares to local HEAD and local index	08:30
	← visik7 left	08:30
context	git-svn is simply a tool used on top of the local git repo, not some magical tool that makes the local git repo look exactly like svn	08:31
	→ visik7 joined	08:31
nud_	visik7: then you can	08:31
visik7	ops	08:31
context	even svn status only relates to the last update...	08:31
nud_	visik7: you can cherry-pick every commit you want from one branch to another	08:31
Ace_NoOne	context: true - but if I do a local git commit, that will be the head - will fetch use the SVN head?	08:32
context	ace_noone. git-svn will add to trunk, which is a 'remote tracking branch	08:33
	you will want to rebase your local master branch	08:33
Ace_NoOne	I don't understand	08:33
context	before dcommit'ing	08:33
	maybe learn/use git a little more before trying to use git-svn	08:33
Ace_NoOne	I guess I should	08:33
	→ Sho_ joined	08:33
	← angasule__ left	08:34
Ace_NoOne	problem is, I have no projects using Git - just used it for local versioning so far	08:34
context	ace_noone. when you git-svn fetch it does not touch your local repo what so ever	08:34
	err	08:34
	your local checkout	08:34
	just like when you git-fetch from another git repo, it does not touch your local checkout	08:35
Ace_NoOne	hmm	08:35
	← goibhniu left	08:36
	→ Judofyr joined	08:37
	→ markkalderon joined	08:38
	← hendry left	08:39
	← freshtonic left	08:43
	← koke left	08:44
nud_	visik7: does it answer yr question?	08:45
visik7	nud_: I dunno what "cherry-pick" means :)	08:46
	→ trochala joined	08:46
nud_	it means you can take a bunch of commits from a branch and apply them on another	08:47
visik7	nud_: seems to be my case	08:47
jones-	Hi, I get this error from git-svn: what can I do? http://rafb.net/p/OmwnQm66.html	08:48
	← broonie left	08:48
	→ broonie joined	08:49
nud_	visik7: cherry-pick works one commit at a time, but I think git-rebase --interactive allows you to do that with an interactive interface, saying "I want this one" or "I don't want that one"	08:49
	→ GyrosGeier joined	08:51
	→ foucist joined	08:52
	→ gitte joined	08:53
	← sverrej left	08:53
	→ anotherjesse joined	08:54
	← pasky left	08:54
	← nickh_ left	08:54
	← csc` left	08:54
	← CodeOfficer left	08:54
	← lydgate left	08:54
	← lwhacker left	08:54
	→ pasky joined	08:54
	→ nickh_ joined	08:54
	→ csc` joined	08:54
	→ CodeOfficer joined	08:54
	→ lydgate joined	08:54
	→ lwhacker joined	08:54
foucist	how do i revert a file to it's previous commit (not necessarily the last commit)	08:54
context	visik7. the man page explains it	08:55
nud_	foucist: git-checkout $file ?	08:55
	or git-reset $commit $file	08:56
	← tonyj left	08:56
context	jones-. i use git-svn but never touched git-svn rebase. maybe try normal rebase	08:56
	→ tonyj joined	08:56
jones-	context: I found out. There were two "fetch" lines in .git/config	08:56
	← lamont left	08:56
	→ lamont joined	08:57
foucist	nud_: git checkout wouldn't do it, as that only returns the current commit	08:57
	how about just doing a git cat or git diff on the file	08:58
	→ Kbyte joined	08:58
foucist	but seeing the previous revision of it?	08:58
nud_	foucist: git-reset HEAD^ $file should do it, wouldn't it ?	08:58
	→ nickh joined	08:59
madduck	nud_: git-reset does not work on files	08:59
	use git checkout HEAD^ -- $file	08:59
foucist	i mean, two revisions ago	08:59
madduck	use git checkout HEAD~2 -- $file	08:59
foucist	not necessairly two commits ago though	09:00
	two commits ago relative to that file ;)	09:00
	heh	09:00
madduck	well, find the commit, using packaxe	09:00
	pickage	09:00
	pickaxe	09:00
	wow. :)	09:00
	or blame or whatever	09:00
	→ KiBi_ joined	09:01
	→ Malesca joined	09:02
	← Slashbunny left	09:02
	← nickh_ left	09:03
	→ Slashbunny joined	09:03
pasky	repo.or.cz gitweb and git.or.cz back up	09:03
	git wiki and ssh jail still to go	09:03
	→ sbeyer joined	09:05
nud_	madduck: actually, git-reset HEAD^	09:06
	madduck: actually, git-reset HEAD^ $file seems to set the index to what was in $file in the last commit	09:07
	it just doesn't update the working tree accordingly	09:07
madduck	well, you need to pass --hard for that to happen	09:07
nud_	so using git-checkout is definitely better anyway ;-)	09:07
madduck	anyway, i wouldn't use reset --hard without being relaly careful	09:07
nud_	yes, indeed	09:08
	→ ebel joined	09:09
	← lamont left	09:11
	← lwhacker left	09:12
	← pasky left	09:12
	← lydgate left	09:12
	← csc` left	09:12
	← CodeOfficer left	09:12
	→ pasky joined	09:13
	→ csc` joined	09:13
	→ CodeOfficer joined	09:13
	→ lydgate joined	09:13
	→ lwhacker joined	09:13
	→ lamont joined	09:14
	→ radarek joined	09:16
	← ckknight left	09:19
	→ ckknight_ joined	09:20
	← lamont left	09:24
	→ grasshopp joined	09:26
	→ lamont joined	09:28
	← markkalderon left	09:29
	← KiBi left	09:33
	→ koke joined	09:34
foucist	how do i get that nice little report that 'git pull' gives with the ++--- stuff, but for the combined last 2-3 revisions	09:35
	← Dreamer3 left	09:36
nud_	foucist: see git-diff	09:37
	something like git-diff --stat HEAD^^..HEAD should do	09:37
	← mithro left	09:40
foucist	thanks	09:40
pasky	# apt-get install git-core	09:40
	The following NEW packages will be installed:	09:41
	git-core libcurl3-gnutls liberror-perl rcs	09:41
	amusing :)	09:41
shd	:p	09:41
	→ janm joined	09:44
	→ koke_ joined	09:44
pasky	ok, everything but git wiki should be back up now	09:47
	→ sverrej joined	09:49
AlexB	pasky: Heh, rcs?	09:53
	← foucist left	09:54
	← radarek left	09:55
	← lwhacker left	09:55
	← pasky left	09:55
	← lydgate left	09:55
	← csc` left	09:55
	← CodeOfficer left	09:55
	→ radarek joined	09:55
	→ pasky joined	09:55
	→ csc` joined	09:55
	→ CodeOfficer joined	09:55
	→ lydgate joined	09:55
	→ lwhacker joined	09:55
	→ markkalderon joined	09:55
	← lwhacker left	09:58
	← pasky left	09:58
	← lydgate left	09:58
	← csc` left	09:58
	← radarek left	09:58
	← CodeOfficer left	09:58
	→ radarek joined	09:59
	→ pasky joined	09:59
	→ csc` joined	09:59
	→ CodeOfficer joined	09:59
	→ lydgate joined	09:59
	→ lwhacker joined	09:59
	← koke left	10:00
pasky	even wiki should run now	10:06
	so if anything wrt. git.or.cz or repo.or.cz is broken please tell me - all should work now	10:06
	← visik7 left	10:08
	← Antiarc left	10:11
	← geckosenator left	10:11
madduck	pasky: only git 1.4.4.1 in debian depends on rcs. i'd suggest using backports.org, which has 1.5.5.1	10:15
nud_	pasky: out of curiosity, how much disk space does repo.or.cz take on the server?	10:16
	← lwhacker left	10:17
	← pasky left	10:17
	← lydgate left	10:17
	← csc` left	10:17
	← radarek left	10:17
	← CodeOfficer left	10:17
	→ radarek joined	10:17
	→ pasky joined	10:17
	→ csc` joined	10:17
	→ CodeOfficer joined	10:17
	→ lydgate joined	10:17
	→ lwhacker joined	10:17
	→ pkufranky joined	10:18
	← Modius left	10:20
	→ Modius joined	10:21
	← jelmer left	10:26
	→ jelmer joined	10:27
	← deavid left	10:34
	→ deavid joined	10:35
	← sverrej left	10:38
	→ sverrej joined	10:38
pasky	okaaaay	10:38
	now things broke a bit :)	10:39
	← meandtheshell left	10:40
telmich	good morning	10:40
	← alenoosh left	10:42
	→ alenoosh joined	10:43
	← Modius left	10:47
	→ Modius joined	10:48
	← Malesca left	10:49
hliusv561	pasky: Define "broke a bit"? :-)	10:53
	← pkufranky left	10:55
pasky	hliusv561: as in "default gw was not set on the server" ;)	10:56
	→ Yuuhi joined	10:56
	← CodeOfficer left	10:57
	← trochala left	10:57
	← Modius left	11:07
	→ Modius joined	11:07
	→ Modius__ joined	11:11
	→ Ademan joined	11:11
	← Modius__ left	11:12
	← ruediger left	11:12
	→ charon joined	11:13
	→ pkufranky joined	11:13
	→ Modius__ joined	11:13
	→ ruediger joined	11:13
	← GyrosGeier left	11:14
	→ pilif joined	11:18
	← pkufranky left	11:19
	→ godog joined	11:22
	← adante left	11:22
	→ adante joined	11:23
	← te left	11:23
comp	pasky: well, if you have a remote serial console ... it shouldn't be a problem ..	11:24
	traceroute ends at aorta.net	11:24
pasky	i don't have remote serial console :)	11:25
	but i fixed that	11:25
	now the whole thing went down and i have no idea why	11:25
comp	some "monkey" on a telephone line?	11:25
pasky	someone in the server room is looking at it right now	11:25
comp	ah	11:25
	→ gerhardlazu joined	11:25
	→ kukks joined	11:26
	← Overshard left	11:27
	← Modius left	11:27
	→ Modius joined	11:28
gerhardlazu	how do I update a submodule after changing its path in .git/config?	11:29
	→ pkufranky joined	11:29
	← Modius left	11:29
	→ Modius joined	11:29
	← Modius left	11:32
	dwmw2_gone → dwmw2	11:32
	→ Modius joined	11:32
	← anotherjesse left	11:34
	← koke_ left	11:40
	→ ekidd joined	11:42
redondos	I was on a detached head, I din't notice and switched to master. how can I go back to the previous state?	11:44
	(assuming I don't know the commit id)	11:44
hliusv561	redondos: 'git reflog'? It prints abbrevated commit IDs and symbolic names for recent states.	11:44
redondos	ah, right. thanks hliusv561	11:45
	← Modius__ left	11:45
pasky	great	11:48
	kernel panic	11:48
	:/	11:48
	← Modius left	11:50
hliusv561	pasky: Killing interrupt handler, killing init or something even more exotic? :-/	11:50
pasky	i unfortunately wasn't able to get much details from the person rebooting it :/	11:50
	I'll setup netconsole not	11:51
	*now	11:51
	→ Modius joined	11:51
comp	pasky: but if you can't connect to it ..	11:52
pasky	well it's rebooted now :)	11:55
	but i can't figure out how to setup the damn syslog-ng to accept log messages by udp	11:55
hliusv561	pasky: 'netcat -u -l -p <port>'? Or use plain 'syslog'?	11:56
	← lirit left	11:57
pasky	i have syslog-ng, not syslog	11:57
	and i don't want to waste a screen window just for this, who know when does it oops again	11:57
Mikachu	are screen windows expensive?	11:57
comp	(or make a C forking program)	11:57
hliusv561	Mikachu: I have hit a limit of 40 per session... :-)	11:58
pasky	Mikachu: I rent one for 30 cents per day	11:58
	→ Lemurnomicon joined	11:58
hliusv561	And running nested screen sessions is not fun...	11:58
pasky	i actually do that a lot :)	11:59
telmich	hliusv561: well, just ctrl-a-a-a-a it ;-)	11:59
	← charon left	11:59
	→ lcapitulino joined	11:59
hliusv561	telmich: Use higher branch factor? If one uses ten of sessions in master for secondaries, one gets 430 ptys...	12:00
	← pkufranky left	12:00
telmich	pasky: udp();	12:00
	→ madnificent joined	12:00
	→ lirit joined	12:01
telmich	hliusv561: well, I use nested screens only over different hosts	12:01
	pasky: I got that in source s_all { ... };	12:01
hliusv561	telmich: I have actually used nested screens on the same host... :-)	12:01
madnificent	what must I add to .gitignore to ignore all files that end with ~ ?	12:01
hliusv561	madnificent: '*~'?	12:01
	telmich is a big syslog-ng-fan, although metalog is nice, too	12:02
	→ wgaffa joined	12:02
pasky	telmich: i needed custom port, but now I figured out that I have to write it like udp(port(1234))	12:02
	and I firmly believe syslog-ng is infinitely stupid	12:02
telmich	pasky: jep	12:02
pasky	:)	12:02
	← Modius left	12:02
madnificent	hliusv561: would that ignore those files in all subdirectories too?	12:02
hliusv561	madnificent: Yes.	12:02
	→ Modius joined	12:03
madnificent	thanks! I must've done something stupid... didn't work last time I tried it	12:03
comp	I can't figure out what are you trying to do :)	12:03
	← Modius left	12:03
comp	remote logging is possible with syslog-ng ..	12:03
pasky	yes but the syntax is too ugly	12:05
hliusv561	comp: That was about figuring out how to do UDP remote logging with syslog-ng. With non-default port. And, no, using TCP was not an option.	12:05
telmich	comp: he already fiexd it	12:05
comp	ah .. :)	12:05
wgaffa	does gitk work in a cygwin/windows environment?	12:06
	← deavid left	12:07
comp	I thought that syslog-ng is using UDP for it by default ..	12:08
	→ doener joined	12:08
Mikachu	wgaffa: i think i heard it does	12:09
telmich	comp: no, not if not enabled	12:10
	← madnificent left	12:10
	→ pkufranky joined	12:10
	→ cyclist joined	12:11
	← bentob0x left	12:14
	→ Modius__ joined	12:16
	← pkufranky left	12:23
	← cyclist left	12:25
wgaffa	I've read a few pages on git and svn and git seems good, but as a single programmer for personal VCS is git better than svn or no difference?	12:27
cehteh	git is better :)	12:28
Mikachu	with svn i think you'd need to set up a server too	12:28
ehamberg	git is much easier to set up	12:28
nud_	wgaffa: git is better for a single reason: no need for a server ;-)	12:29
	git-init in your project directory and you're done	12:29
wgaffa	ah, git seems to get better and better for each moment :)	12:30
hliusv561	Also, Git is usually much more space-efficient..	12:30
telmich	if I want to merge all changes from origin/master, but want to omit merging of file_a, which I know is good, which conflicts with the merge, what is the best reason to resolve it?	12:31
wgaffa	yeah, read something about 30x space efficient	12:31
telmich	wgaffa: git is useful for offline working, compared to central systems (which I use alot)	12:32
wgaffa	telmich, I'm never offline so that isnt that big deal to me but a good point if I will go offline sometime	12:33
Mikachu	telmich: if it's just one file, just do the merge and then git checkout master -- file_a and commit (assuming other conflicts are resolved)	12:33
	→ pkufranky joined	12:33
telmich	Mikachu: argh, that was way too easy	12:33
Mikachu	telmich: (and that master is the branch you're merging into)	12:33
	i guess HEAD would be fine too actually	12:34
	→ redondos_ joined	12:34
telmich	not sure where HEAD points to in a merge, but maybe yes	12:35
Mikachu	since it can only point to commits and the merge isn't committed yet, it should be pointing to the commit before the merge that has the file you want	12:35
hliusv561	HEAD during merge points to the branch one is merging to (i.e. the one that was HEAD immediately before the merge command was issued).	12:36
Mikachu	in other words, it doesn't change at all until you commit the merge	12:37
	← alenoosh left	12:37
	→ BabelO joined	12:43
	→ mabs29 joined	12:43
	← sbeyer left	12:44
	→ pkufrank1 joined	12:44
	← redondos left	12:45
	redondos_ → redondos	12:45
	← pkufranky left	12:45
gitte	pasky: any progress on the pubkey front?	12:48
mabs29	Hello everyone. I'm just replacing my Mercurial system with Git and there are a few hurdles I'm tripping over. My biggest one is private access with minimum control: Gitorious user keys don't see to permit password-less login no matter what sshd_config says	12:48
	→ d0k joined	12:48
mabs29	Oh god, that sounds hauntingly familiar territory. :(	12:48
gitte	mabs29: probably a PAM problem; I had the same (but without Gitorious)	12:48
	→ Modius joined	12:49
mabs29	gitte: aha! That makes sense. How'd you fix it? Also, what's your set up for SSH access without Gitosis (whoops, getting a bit fried I think)	12:49
gitte	mabs29: in particular, I had to edit /etc/pam.d/common-auth, and change the "nullok_secure" to "nullok".	12:49
	mabs29: I have regular non-chrooted access to that machine.	12:50
	mabs29: my users are real non-chrooted users, too ;-)	12:50
	cbrake_away → cbrake	12:50
mabs29	gitte: I tried that at first but it never found it like hg seemed to	12:50
	→ mcella joined	12:50
mabs29	gitte: in fact, that's my setup already, so I'd prefer to use that	12:50
gitte	mabs29: did you study the system logs? (I.e. everything in /var/log/?)	12:51
mabs29	gitte: yeah, I managed to take care of most of the issues, until fixing Gitosis seemed like a quicker solution (wasn't). I'd just like to get to a working point. I'm just making the PAM change you suggested and see if that turns the lights on.	12:53
gitte	mabs29: could be that it has to be somewhere else, too...	12:53
	ckknight_ → ckknight	12:53
mabs29	gitte: nod - any other tips for attack around the /etc/ jungle?	12:54
gitte	Well, best hint for me was in the /var/log/messages, IIRC.	12:54
	→ cmarcelo joined	12:54
gitte	Or .../auth	12:54
	← cmarcelo left	12:59
mabs29	gitte: nod. Whole heaps of things in there, but still not getting to grab a Git repository using SSH. Hmm. It is possible, right? :)	13:00
gitte	mabs29: the thing is: you want password-less access, right?	13:01
	→ cmarcelo joined	13:01
mabs29	gitte: nod	13:02
	← Modius left	13:04
mabs29	gitte: from a remote computer anywhere (with priv/pub keys shared), this to work smoothly: git clone ssh://user@host/path/to/repo.git repo-cloned	13:04
gitte	Yes, yes, I understand.	13:04
	Exactly the thing I tried to.	13:04
	Try to find out what sshd says about your attempts.	13:04
	→ Modius joined	13:04
hliusv561	mabs29: Maybe just 'ssh user@host' could be useful for troubleshooting. To get it work or at least abort after authentication completes...	13:05
	← Modius__ left	13:06
	← pkufrank1 left	13:06
gitte	Oh, and you need an empty password in /etc/passwd.	13:06
mabs29	gitte: it's been operating at debug3 just to be sure. The main complaint now for real user access is perhaps the following isn't right: sshd[xxxx]: git-upload-pack '/git/repo.git	13:06
hliusv561	mabs29: For account with full-powered shell, 'ssh user@host git --version' is usually a good test...	13:07
mabs29	hliusv561: oh I can definitely login without a password, it's the same account	13:07
gitte	So you can connect interactively, and it says "Go away! Do you think I am a shell?"?	13:07
	mabs29: then it is not an issue with the password-less account.	13:07
	mabs29: it is an issue with your git installation.	13:07
hliusv561	gitte: Has it changed? IIRC, it was 'What you think I am? A shell?'...	13:08
telmich	hmm, using git-svn rebase and dcommit really screws up workflow, when using git-push/fetch to a normal repo, too	13:09
hliusv561	mabs29: And if command like that suceeds, it should ONLY print 'git version <something>', possibly preceded by password prompt. NOTHING ELSE.	13:09
mabs29	hliusv561: I get a version back	13:09
	hliusv561: yup, that's all it does. Git 1.5.5.1 by the way.	13:10
hliusv561	mabs29: On the server, is 'git-upload-pack' in the same directory as 'git'?	13:10
	mabs29: Most probably in something like /usr/local/bin, or /usr/bin...	13:12
	→ Inviz joined	13:12
Inviz	Hey guys. Is there a way to search local repo for the line? I mean, i want to know when the line was added	13:13
	← ekidd left	13:13
hliusv561	Inviz: -S to git log?	13:13
	→ pkufranky joined	13:14
mabs29	hliusv561: Right you are! Didn't have its location in /etc/profile when setting PATH. Now I get a different error and I think that's fixed now, actually. Thanks a lot!	13:14
Inviz	hliusv561: can you please be more specific? :)	13:14
hliusv561	Inviz: git log has -S option (pickaxe), which can be used to search when some string appeared in source. See git-log(1). Also, there is command called 'blame'.	13:16
telmich	Inviz: git-log '-Syour line'	13:16
	→ koke joined	13:16
Inviz	many thanks, will try	13:16
	→ dwave joined	13:18
	→ entropy joined	13:21
	→ WorkingGeier joined	13:21
entropy	hi everyone, how do I "uncommit" my most recent commit? I've not pushed it yet and I've changed my mind and want to obliterate it	13:22
WorkingGeier	remove completely, or change?	13:23
doener	you can amend it with "git commit --amend" (stage the desired changes before that)	13:23
	to totally get rid of it (as far as your current branch is concerned): git reset --hard HEAD^	13:23
hliusv561	entropy: Or just 'pop' it off: 'git reset --soft HEAD^'.	13:24
doener	or, if you prefer that over amending it, you can pretend that you did not do the commit yet, with "git reset --soft HEAD^"	13:24
entropy	i have many commits though, I don't want to lose all commits not yet pushed, only the most recent	13:24
doener	entropy: thus HEAD^, that resets your branch to the parent of HEAD	13:24
entropy	ah	13:25
	← pkufranky left	13:25
doener	IOW, one commit earlier	13:25
entropy	thanks doener, hliusv561	13:25
doener	to go five commits back, that would be HEAD^^^^^ or HEAD~5	13:25
entropy	oh, nice	13:25
	will that change my working directory source?	13:25
	will my working source also be "reset" or just the commit log	13:26
doener	only with --hard	13:26
entropy	awesome	13:26
	ok	13:26
doener	see git-reset(1)	13:26
entropy	thanks	13:26
	entropy nods	13:26
doener	and for an explanation of HEAD^, HEAD~23, foo^! and so on, see git-rev-parse(1)	13:26
	→ LiamH joined	13:29
	← eternaleye left	13:30
	→ devurandom joined	13:32
	→ charon joined	13:33
	→ pkufranky joined	13:33
	→ mdales joined	13:33
	→ threeve joined	13:34
	→ krh joined	13:35
	→ threeve_ joined	13:35
mdales	how do I find the object id for a tagged object if I know the tag?	13:35
Tv	git show	13:36
	but every command should just accept the tag name	13:36
	← Modius left	13:36
	→ Modius joined	13:37
mdales	well, I wanted to inspect back in the tree programmatically to find the last point I tagged, and then work out how many commits there are between me and it, as a sort of way of generating a build number	13:37
hliusv561	mdales: If you really need the object ID of object pointed by tag (recursively peeled): 'git rev-parse tag^{}'...	13:37
doener	mdales: git describe	13:37
mdales	thanks all	13:38
	→ schacon_ joined	13:39
	← mabs29 left	13:41
	← janm left	13:41
mdales	then I can sort of do a git-rev-list [tag object id]..HEAD to get a number to tag a particular build made from the central repo	13:42
	actually	13:42
	just noticed I can put the tag in	13:42
	doh	13:42
	git-rev-list "tag"..HEAD \| wc -l :)	13:42
	→ zeenix joined	13:42
telmich	apropros git-describe, anyone a good hint why it returns "fatal: cannot describe 'da5a9fadb6aa13e9d3f4e98aff1baefab35cb508'"?	13:42
hliusv561	Actually, 'tag..'. HEAD is impiled there.	13:43
telmich	there are annotated tags, but maybe it broke because of git-svn rebase	13:43
	→ janm joined	13:44
	→ heythisisdave joined	13:44
	← pkufranky left	13:45
	→ EmilMedve joined	13:45
hliusv561	telmich: It returns that kind of error (with default options) it there are no annotated/signed tags in past of current commit.	13:45
devurandom	Hello.	13:47
telmich	hliusv561: git tag -n shows them	13:47
devurandom	Can I do what git-format-patch does in email format with just plain patches? A patch of a range of commits, each one prefixed with the commit message, etc?	13:48
	← ph^ left	13:48
doener	mdales: what's wrong with just using what "git describe" gives you?	13:48
hliusv561	devurandom: Huh. If you don't specify anything 'exotic', like --binary, -M or -C, those patches should be patch-compatible.	13:48
devurandom	Yes, they are. But they have all the email metadata in them.	13:49
	Like Subject, and so on.	13:49
	→ jpeeler joined	13:49
hliusv561	devurandom: AFAIK, They still are patch-compatible.	13:50
	devurandom: Modulo misbehavior from 'patch'... IIRC, commit messages containing what looks like diff output can cause problems...	13:51
	→ Modius__ joined	13:51
	→ cannonball joined	13:51
	→ Ryback_ joined	13:51
	← threeve left	13:52
gerhardlazu	I have changed the path of a submodule in .git/config . How do I force my already initialized submodule to re-download from this new path?	13:52
	← Sho_ left	13:52
devurandom	git-show cannot output multiple patches? I find it looks better.	13:52
	← garrin left	13:53
doener	devurandom: git rev-list start..end \| xargs git show	13:53
	maybe with --topo-order	13:53
hliusv561	gerhardlazu: You also probably need to alter the URL in both .git/config files (of supermodule and submodule)... And perhaps even in .gitmodules of superrepo... As one can see, submodules utilities are still a bit work in progress...	13:54
doener	and --reverse	13:54
Mikachu	git log -p?	13:54
	← Modius__ left	13:54
doener	ehrm, yeah... for a consecutive row of commits that makes more sense...	13:54
	I've been testing with random commits	13:54
	→ Modius__ joined	13:54
gerhardlazu	hliusv561: I thought .gitmodules was for other devs on the project as well. I only want to use a different path for a submodule locally. Can it be done or is it just a "feature" at this time :D?	13:55
hliusv561	gerhardlazu: Then just editing both config files should be enough...	13:55
gerhardlazu	nope	13:56
	after editing both of them && git submodule update, nothing happens	13:56
	hliusv561: maybe because the submodule already exists...	13:57
	because if I go into that dir and do a git remote show origin, I get the initial path...	13:57
hliusv561	gerhardlazu: In supermodule, do 'git rev-parse HEAD:path/to/submodule'. In submodule, do 'git rev-parse HEAD'.	13:57
	gerhardlazu: That path comes from .git/config of submodule...	13:58
	gerhardlazu: Those two commands should dump 40 hex numbers each? Are those the same?	13:58
	← Judofyr left	13:58
gerhardlazu	yes they are	13:58
hliusv561	gerhardlazu: Then no wonder 'git submodule update' did nothing.	13:59
gerhardlazu	an origin on the remote in the submodule still points to the other URL	13:59
	→ Judofyr joined	13:59
gerhardlazu	ok, so how should I go about it?	13:59
	delete and then re-add & init?	13:59
	is it at all possible to have a certain path for a submodule locally and a different path in the checked in repo?	14:00
hliusv561	gerhardlazu: Checkout proper version of submodule, and add it to superproject (WARNING: do not put trailing '/' to the submodule path in add command).	14:00
gerhardlazu	hliusv561: yes, that caught me only 1, the / :D	14:00
doener	devurandom: if you only want the patches, without messages, this works as well: git rev-list --reverse --topo-order HEAD~5.. \| nl \| (while read num hash; do git diff-tree -p $hash^! > $num; done)	14:00
Mikachu	so git doesn't mind if the superproject has files in the same directory as the submodule?	14:01
	(as long as the submodule doesn't have a file there of course)	14:01
	→ pkufranky joined	14:01
	← Modius left	14:02
hliusv561	Mikachu: 'git add submodule' on submodule adds the submodule. 'git add submodule/' adds the files from submodule (which almost invariably is unwanted)...	14:02
	→ Modius joined	14:03
	→ Sho_ joined	14:06
	← lamont left	14:08
devurandom	Git manual says about git-send-email: "If a username has been specified (with --smtp-user or a configuration variable), but no password has been specified (with --smtp-pass or a configuration variable), then the user is prompted for a password while the input is masked for privacy.	14:08
	Is that correct?	14:08
	← Modius__ left	14:09
devurandom	I have a smtpuser specificed in ~/.gitconfig, but I am not prompted for any password...	14:09
	→ zeenix_ joined	14:10
	→ ekidd joined	14:11
	← zeenix left	14:11
comp	how can I print a commit message of a certain commit? let's say I want "medium" output (the git-log default) .. but after reading through manpages of git-rev-parse and git-rev-list, I still don't know how to do that	14:13
	git-log can display ranges	14:14
cedricv	git show <commit> ?	14:14
comp	that will print it with a diff	14:14
doener	comp: git log -1?	14:14
comp	doener: this prints only the latest one..	14:15
nud_	comp: depends on the --pretty you give it (for git-show)	14:15
doener	well, add the commit-ish after the -1... git log -1 HEAD~534	14:15
nud_	you can also git-log HEAD~5 --pretty=oneline -1	14:15
	← pkufranky left	14:15
Mikachu	comp: you can use either git log -1 sha1 or git show -s sha1, git log -1 is faster	14:16
	→ lamont joined	14:16
comp	doener: works :)	14:16
	yeah, I thought that git-log -num display only a range of HEAD..commit	14:16
	← zeenix_ left	14:18
	→ pkufranky joined	14:18
	→ zeenix joined	14:18
	← pdherbemont left	14:19
	→ pdherbemont joined	14:19
	← lamont left	14:19
	→ lamont joined	14:20
	← etnt left	14:20
	→ ceez joined	14:21
	← ruediger left	14:25
	→ graywh\|work joined	14:26
	← reel_ left	14:28
	← pkufranky left	14:28
	→ reel_ joined	14:30
	→ bdiego joined	14:32
	→ leitgebj joined	14:34
mdales	doener: perhaps my ignorance of git describe :)	14:36
	mdales looks	14:36
	→ deavid joined	14:37
	← lirit left	14:37
	← Kbyte left	14:38
	→ pkufranky joined	14:39
cehteh	how do i reset another branch to my current HEAD without checking it out? git fetch HEAD:other_branch ?	14:40
gitte	git push . +HEAD:other	14:41
	← devurandom left	14:41
cehteh	heh .. fetch with . did it	14:41
mdales	doener: even better - thanks for that pointer	14:42
cehteh	but ok :) push sounds more intuitive	14:42
	→ shoe joined	14:45
	← warthog9 left	14:46
	← peeja left	14:49
	→ devurandom joined	14:51
	→ ph^ joined	14:51
devurandom	Does git-send-email work with ssmtp?	14:52
	← tcoppi left	14:52
devurandom	It does not seem to authenticate to the server, no matter what I pass as --smtp-user, etc.	14:52
cehteh	devurandom: never tried, but i would expect that	14:52
	→ retybok joined	14:52
devurandom	Isn't ssmtp expected to be sendmail compatible?	14:52
retybok	is there a way to configure a repo so that git push pushes to multiple remote repos?	14:53
cehteh	well yes .. but with most features beeing nops	14:53
devurandom	cehteh: If I pass -au and -ap manually, it works though...	14:54
cehteh	devurandom: well i cant say, i use ssmtp often for servers without bigger MTA's .. but not for git	14:55
	← namenlos left	14:56
	→ lolage0 joined	14:56
	→ tcoppi joined	14:57
	← wgaffa left	14:59
devurandom	If I see this correctly then it is the fault of Net::SMTP and not Git's own?	14:59
	← entropy left	15:00
	→ balrok joined	15:03
	→ brosner joined	15:07
	→ EmilMedve1 joined	15:07
	← EmilMedve left	15:08
	EmilMedve1 → EmilMedve	15:08
	← ph^ left	15:09
balrok	my "git clone" sends nearly all its output as stderr.. but everything works fine.. is there a way to tell the program, that it should send only stderr-output if its a real fault?	15:11
	i've uploaded an example what comes to stderr: http://rafb.net/p/eGlPEB47.html	15:11
	← dwave left	15:13
	← ekidd left	15:16
	← aarapov left	15:19
	→ ekidd joined	15:21
	← ekidd left	15:21
	← nud_ left	15:24
	→ a-priori_ joined	15:27
	← schacon_ left	15:29
	← Inviz left	15:29
	← pkufranky left	15:29
	→ devogon_ joined	15:32
	→ mando joined	15:33
	→ ekidd joined	15:35
	→ stouset joined	15:36
	→ mmsc joined	15:40
mmsc	Hi. I just installed msysgit, and I used the option that let's me use git with the Windows command line. Problem is when I enter 'git' or any of its commands, I see some text flash across the screen and then the command prompt immediately shuts down	15:41
doener	start the command prompt with "cmd" and enter the commands there?	15:42
	s/start/when starting/	15:44
	← devogon left	15:44
	→ halfline joined	15:46
mmsc	doener: I must have missed that. Is there a reason it works that way and you cannot just open up a command prompt and have 'git' work?	15:46
Ilari	mmsc: Well, if you could make the command prompt wait explicit close...	15:47
doener	mmsc: by "open up a command prompt" you mean "Start->Run", right?	15:47
mmsc	doener: yeah	15:48
doener	mmsc: that won't even have the right working directory, so it's kinda pointless	15:48
	→ taq joined	15:48
	→ chris2 joined	15:48
	→ peeja joined	15:49
mmsc	I've always done it that way, didn't know any better	15:49
taq	Hey, is there an option to ignore symbolic links with git ?	15:49
	→ krawek joined	15:50
	← markkalderon left	15:53
	→ kewark joined	15:54
	← stouset left	15:54
	→ ph^ joined	15:55
	→ ph^_ joined	15:55
doener	mmsc: maybe you got confused by the install option. It most likely meant that you can run the commands from cmd or similar, not "Start -> Run"	15:55
	← peeja left	15:57
	← ph^ left	15:58
	→ pkufranky joined	15:58
	→ ph^ joined	15:58
cedricv	is there an option equivalent for git apply --whitespace=fix for git commit?	15:58
leitgebj	Tv, any idea why the authorized_keys file for the git user wouldn't be updated on a push of the gitosis-admin repository?	15:58
	← charon left	15:59
mdales	hmmm. I seem to be able to do a git-rev-list tag..HEAD on a branch where tag is on another fork. is this normal?	16:00
	→ devogon joined	16:01
mdales	in the other branch it is one after the fork	16:01
	← devogon left	16:02
	→ devogon joined	16:02
	← devogon left	16:03
mdales	indeed. if I do it to the head of the other fork I get the same result - it seems to take me down as far is the fork point and stop	16:03
Tv	leitgebj: first thing to check, are there extra lines	16:05
	leitgebj: if there are, it might be updated ok, but you're just not using the gitosis-managed lines	16:05
leitgebj	Tv, do you mean that there should be no extra lines in the authorized_keys file?	16:05
	→ peeja joined	16:06
taq	would be a good idea to include the symbolic links dirs on .gitignore so? :-)	16:06
Tv	leitgebj: not unless you know what you're doing	16:06
	leitgebj: if you have trouble, you don't ;)	16:06
leitgebj	The problem is that when I add a new key to gitosis, an entry isn't added to the authorized_keys file.	16:06
Tv	leitgebj: then the postupdate hooks isn't run, or is failing	16:07
	leitgebj: any output on the push?	16:07
leitgebj	Tv, no output on the push	16:07
	Tv, the postupdate hook wasn't running at first on a debian system, so I had to change a symlink	16:07
Tv	leitgebj: huh	16:08
	leitgebj: what symlink?	16:08
leitgebj	Tv, from the hooks dir for post-update to the correct script on my system... at first the symlink didn't point to a valid file	16:08
Tv	uh-huh	16:09
	you broke it	16:09
	you get to keep both halves	16:09
leitgebj	:-(	16:09
Tv	leitgebj: ls -l ~git/repositories/gitosis-admin.git/hooks/post-update	16:10
	← taq left	16:10
leitgebj	Tv, lrwxrwxrwx 1 git git 97 2008-05-19 10:56 /home/git/repositories/gitosis-admin.git/hooks/post-update -> /usr/share/python-support/gitosis/gitosis-0.2-py2.5.egg/gitosis/templates/admin/hooks/post-update	16:10
Tv	ls -l /usr/share/python-support/gitosis/gitosis-0.2-py2.5.egg/gitosis/templates/admin/hooks/post-update	16:11
leitgebj	ls: cannot access /usr/share/python-support/gitosis/gitosis-0.2-py2.5.egg/gitosis/templates/admin/hooks/post-update: No such file or directory	16:11
	← krawek left	16:11
Tv	leitgebj: what do you mean "at first"?	16:11
	it's currently broken	16:11
leitgebj	Tv, I see now...	16:11
	Don't know how I missed that. I thought that I fixed it...	16:12
	← devogon_ left	16:12
	← ruphy left	16:12
Tv	now, as to why it would be wrong in the first place -- either your python, distutils, or the debian "helper" libs are buggy	16:13
	i template=resource_filename('gitosis.templates', 'admin')	16:13
	the path i use comes straight from pkg_resources, aka setuptools/distutils	16:13
	gotta walk the dogs, eta 50 min	16:14
leitgebj	Thanks for the help Tv!	16:14
	→ fhobia joined	16:15
	→ kumbayo joined	16:16
	kewark → krawek	16:16
	→ johan-s joined	16:19
	← mmsc left	16:21
	← ph^_ left	16:27
	← peeja left	16:27
	← aporetic left	16:27
	→ aporetic joined	16:28
	← SRabbelier left	16:28
	→ SRabbelie1 joined	16:28
	← mcella left	16:28
	SRabbelie1 → SRabbelier	16:28
	← _zack left	16:29
	→ schacon joined	16:31
	→ plediii joined	16:31
	← gitte left	16:31
	← koke left	16:33
	→ koke_ joined	16:33
	← redondos left	16:33
	→ roylee17 joined	16:35
	← jacobat left	16:35
	→ redondos joined	16:37
	→ nud joined	16:37
	→ charon joined	16:39
	← koke_ left	16:43
	angasule_ → angasule	16:44
	← fhobia left	16:48
	← deavid left	16:48
	← statim left	16:50
	← WorkingGeier left	16:51
	→ Chairos joined	16:52
	→ kanru joined	16:52
	→ ajonat joined	16:55
	→ broheem joined	16:56
	→ tvachon joined	16:57
	→ cliffstah joined	17:00
eMBee	good evening	17:00
	if i clone a repo with git svn, but include only the trunk, how can i add branches later?	17:01
	→ ^Beket\| joined	17:05
	← retybok left	17:05
doener	eMBee: add a branches entry to the svn-remote section in the config	17:05
eMBee	ok	17:05
	→ eternaleye joined	17:06
	→ alley_cat joined	17:06
eMBee	then i have a different problem. the branches thing only handles branches in one directory level. unfortunately thos repo has got branches in almost random places	17:06
	→ chrisb` joined	17:07
Chairos	that's kind of stupid	17:07
eMBee	it certainly is	17:07
Ilari	"This is your brain on SVN" :->	17:07
doener	you can also just add the branches in separate svn-remote, "git svn init svn://url/to/branch	17:07
	chrisb` → chrisb_	17:07
eMBee	but that'swhat you get when your revision control interface looks like a filesystem and allows people to move things around randomly	17:07
doener	or multiple svn-remote, each using --branches	17:07
eMBee	ahh	17:08
doener	last time I tried, git-svn still correctly found parent/child relations	17:08
	chrisb_ → chrisb	17:08
doener	might depend on the fetch order though	17:08
	→ timocratic joined	17:09
eMBee	i think the fetch order should not matter, if it fetches the branch first and then the trunk at the worst case it will import the same commits multiple times	17:09
Chairos	I have a different kind of git-svn problem. There's a commit I made which predates my move to git-svn, so it exists in the 'master' aka remotes/trunk branch. I need to revert it. But I have a local branch that depends upon that commit. Can I revert it in master, while still somehow keeping it in that local branch?	17:09
	← ph^ left	17:09
eMBee	you want to remove the commit from the master and rewrite the whole branch?	17:10
doener	Chairos: hm? git revert will just add a new commit that reverts the introduced changes. Your other branch won't magically get that commit	17:10
eMBee	i'd just fix the change it introduces (ie the reverse of the commit) and put that on top of the master	17:11
	yeah, exactly what git-revert does	17:11
Chairos	yeah, but then I want to re-add the commit at the 'bottom'	17:11
	of the branch	17:11
Tv	Chairos: what you want is a rebase..	17:12
Chairos	that's what I thought, but I'm not sure how to do it.	17:12
	guess i should go reread the man page	17:12
eMBee	did others clone your repo?	17:13
Chairos	I'm the only one using git-svn; everyone else is using regular svn.	17:13
	so, yeah, other people have that commit I'm reverting.	17:13
eMBee	then don't rebase	17:13
Tv	oh it went to svn? then don't touch it	17:13
eMBee	because you won't be able to push commits back to svn after that	17:14
Chairos	great	17:14
Tv	Chairos: please explain "There's a commit I made which predates my move to git-svn, so it exists in the 'master' aka remotes/trunk branch."	17:14
	Chairos: commit was made with svn?	17:14
Chairos	yes.	17:14
Tv	and the local branch?	17:14
	→ ph^ joined	17:15
Chairos	that commit SHOULD have been in the local branch instead of 'trunk', but I wasn't using git-svn at the time.	17:15
eMBee	if you rebase your master branch then it is no longer the same branch as the trunk on svn	17:15
Tv	Chairos: just revert that commit in your master	17:15
Chairos	tv: yeah, but I have a local branch that needs that commit.	17:16
Tv	that can be fixed afterwards	17:16
Chairos	that's what I'm asking how to do.	17:16
cedricv	why not cherry-pick that commit then from your local branch that needs it?	17:16
	→ koke joined	17:16
Tv	Chairos: go ahead with the revert first	17:16
eMBee	so the local branch does not contain the commit yet?	17:17
Chairos	eMBee: no.	17:17
eMBee	ah, ok	17:17
Tv	Chairos: git checkout -b temp the_original_mistake_commit && git rebase --onto master HEAD~ && git checkout your_local_branch && git rebase --onto temp master && git branch -d temp	17:18
Chairos	tv: ah, thanks.	17:18
Tv	oh one more thing	17:18
	Chairos goes to read the manpage to figure out exactly what that will be doing.	17:18
eMBee	hehe	17:19
Tv	Chairos: git checkout -b temp the_original_mistake_commit && git rebase --onto master HEAD~ && git commit --amend # remove any git-svn id, if there	17:19
	then continue with the previous	17:19
eMBee	HEAD~ is the reverse of a commit? why not just use git revert?	17:21
Tv	eMBee: huh?	17:21
eMBee	what does git rebase --onto master HEAD~ do?	17:22
Tv	eMBee: man git-rebase	17:22
eMBee	wait	17:22
	eMBee needs sleep	17:22
eMBee	hmm	17:22
	→ kewark joined	17:23
	← balrok left	17:23
	← ekidd left	17:23
	← krawek left	17:24
	kewark → krawek	17:24
	eMBee is still confused: git rebase --onto master HEAD~; would rebase the_original_mistake_commit on top of master, but the_original_mistake_commit is already in master, why put it there again? does that cause the commit to be reverted?	17:25
Tv	eMBee: it was reverted	17:27
	eMBee: that undoes the revert in temp	17:28
	s/undoes the revert/redoes the original/	17:28
Chairos	Tv: I get a conflict in the git rebase --onto temp master stage	17:29
Tv	Chairos: well rebases are merges, merges can get conflicts. resolve it.	17:29
	→ arwed joined	17:31
	← mdales left	17:31
	← ebel left	17:31
Chairos	I just ran git checkout -b temp the_original_mistake_commit && git checkout your_local_branch && git rebase --onto temp master your_local_branch && git branch -D temp and it seemed to have worked.	17:31
	← koke left	17:33
	→ aarapov joined	17:33
	→ WorkingGeier joined	17:34
	← alley_cat left	17:35
Chairos	thanks guys	17:38
	← yann left	17:39
	→ deavid joined	17:39
	→ robinr joined	17:40
	← gerhardlazu left	17:45
	← arwed left	17:45
	→ ruediger joined	17:48
	→ hagna joined	17:48
	→ bugfux joined	17:48
bugfux	I am using submodules, and when I try to change branches I get a fatal: Untracked working tree file 'vendor/plugins/will_paginate/LICENSE' when will_paginate is a submodule	17:49
	(so all of that is the submodule, which ofcourse its is untracked)	17:49
hagna	in my svn repo I have a bunch of projects but when if I switch to git I'll make a git repo for each project since I don't ever want to branch the whole collection of projects	17:50
	is that correct thinking?	17:50
Tv	hagna: most likely yes	17:50
hagna	Tv: ok good	17:50
bugfux	anyone have any ideas?	17:51
hagna	bugfux: what's a submodule?	17:51
bugfux	git submodules	17:51
	→ statim joined	17:52
	→ ekidd joined	17:54
hagna	Tv: it's all because we really use svn for copying and backing up but not for branching or doing scm things	17:55
eMBee	doener: if i run git svn init again on a repo that is an svn import than i get an error	17:56
	→ threeve joined	18:02
	← Chairos left	18:02
	→ arwed joined	18:02
	← ajonat left	18:02
	→ threeve__ joined	18:03
	← cedricv left	18:05
	→ cedricv joined	18:06
	← cedricv left	18:07
	← EmilMedve left	18:07
doener	eMBee: ah, you need to pass a unique name for the new svn remote	18:08
	IIRC as the last parameter, after the url	18:08
	git-svn(1) probably knows	18:08
Ace_NoOne	are there any public SVN servers (anon write access) I could test my git-svn skills on?	18:10
doener	just create one locally?	18:10
Ace_NoOne	I've never set up an SVN server, and have no intention of learning obsolete stuff ;)	18:11
madduck	Ace_NoOne: svnadmin init or so	18:11
doener	svnadmin /path/to/repo;	18:11
hagna	Ace_NoOne: it's two lines	18:11
madduck	then just use the file path	18:11
doener	then just use the file:/// protocol	18:11
Ace_NoOne	oh cool - thanks	18:12
eMBee	doener: yeah, figured it out, thanks	18:12
	← aarapov left	18:13
doener	eMBee: if you use "git svn fetch" you might want to add --all now	18:13
	→ EmilMedve joined	18:13
	← ruediger left	18:13
eMBee	doener: yeah, got that too, even though --fetch-all is only hinted at in the manpage, it is not documented (neither is --all)	18:16
hagna	is anyone using git for their whole machine?	18:17
Mikachu	i sure hope not	18:18
	← threeve_ left	18:18
	← threeve left	18:18
	← arwed left	18:20
eMBee	if you want revision control for your whole machine, take a look at foresight/rpath/conary	18:20
	→ arwed joined	18:20
hagna	eMBee: ok I'll do that, and why not git?	18:20
Mikachu	you can use git if you want all your files to be 644 or 755	18:20
hagna	Mikachu: ahh	18:21
eMBee	git is for source managment. conary is for packages	18:21
Mikachu	there are probably some other issues too	18:21
	→ nud_ joined	18:21
	← nud left	18:22
hagna	git seems like high quality stuff is conary similar in that way?	18:24
	← ^Beket\| left	18:24
	← a-priori_ left	18:25
eMBee	not sure, but conary is the only package managment system that works like revision control that i am aware of	18:28
	→ cnk joined	18:28
graywh\|work	is there a reason gitk wouldn't show the files added in the first commit when "Patch" is selected?	18:28
	← devurandom left	18:29
	dwmw2 → dwmw2_gone	18:29
	→ drizzd joined	18:30
Ilari	graywh\|work: Probably it tries to invoke 'git diff' without the --root option, so taking diff won't work...	18:30
graywh\|work	just making sure I'm not crazy	18:31
Ilari	graywh\|work: It even seems to be case. Gitk source does not contain '--root'.	18:32
graywh\|work	is it possible to squash the second commit into the first?	18:34
	→ thannoy joined	18:35
	→ netoman joined	18:35
Ilari	graywh\|work: That might not be so trivial if the branch contains more than two commits. Third to second would have been easier...	18:35
	→ grahal joined	18:36
Ilari	graywh\|work: Normally that sort of squashes would be done via interactive rebase, but rebase doesn't seem to work very well when initial commit is involved...	18:37
graywh\|work	what about just changing the message on the first commit?	18:38
cnk	is there a way to see what has changed in my remote submodule without fetching it? Or is it ok if I fetch and then chicken out merging it because I can just do a variant of 'git reset'?	18:38
	→ david_koontz joined	18:38
Ilari	graywh\|work: One could use filter-branch... I don't know working good ways to mess with initial commit after more commets have been made on top of it...	18:39
	← arwed left	18:39
	← dave_h_d left	18:40
graywh\|work	what about interactive rebase from the first commit, edit second, do a reset and commit --amend, then re-commit the second?	18:40
	← drizzd_ left	18:40
Ilari	graywh\|work: Well, you can try. Maybe do backup of old branch first (git branch foo-backup foo)...	18:41
netoman	Hello! Git newbie here... How can I undo an 'accidental' rm * wich is followed by some commit? Example: edit file1.txt ; edit file2.txt ; rm *; save file2.txt (which recreates file2.txt), then commit.	18:41
	← david_koontz left	18:41
	→ david_koontz joined	18:42
netoman	I ended up with a commit that "removed" every file, except for file2.txt... Now I want to recover all my lost files..	18:42
Ilari	netoman: Modifications to file1.txt is probably lost (unless you added it, and then it is pretty nasty to recover).	18:43
	netoman: Maybe something like 'git rev-parse HEAD' (get ID for current commit). 'git reset --hard HEAD^', 'git checkout <ID from parse-rev> -- file2.txt'...	18:44
	← SRabbelier left	18:45
	→ SRabbelie1 joined	18:45
	SRabbelie1 → SRabbelier	18:45
netoman	Ilari: I've tried with 'git revert', but when i give it a 'commit name' (SHA1) it complains...	18:46
Ilari	netoman: Git revert? Did I say revert? To me that looks like 'reset'.	18:47
netoman	Mmmmh... maybe I'm not understanding well 'git-revert' and 'git-reset'... I'm going back to documentation...	18:48
	→ goibhniu joined	18:49
Ilari	netoman: That command sequence basically first grabs handle to current commit, then removes the current commit and finally restores that one file from the commit pointed by handle.	18:49
	→ tehPope joined	18:49
	→ shaftyy joined	18:50
Ilari	Actually, not just removes, but also reverts the working tree to state before that commit.	18:50
	→ markkalderon joined	18:52
	→ ftehw joined	18:53
netoman	Ilari: So supose I'm an idiot and accidentally deleted 100 files... so I haven't got their names anymore... Wores, i've made a commit... Is it possible to restore to the state of the commit previous to the deletion of the files?	18:53
	Worse*	18:53
	→ mjc joined	18:53
	→ alley_cat joined	18:53
netoman	(My english is horrible, sorry)	18:53
	mjc → mjc_	18:54
	mjc_ → mjc2	18:54
doener	Ilari: btw, instead of using the rev-parse there, is it ok to rely on ORIG_HEAD being set by reset --hard, or is that an implementation detail?	18:55
Ilari	netoman: 'Popping' the last commit and reverting working tree to commit-before-that state would be one reset with suitable arguments. But if you want to preserve changes to file2.txt...	18:56
	→ sbeyer joined	18:56
doener	oh, nvm, the manpage has an example with ORIG_HEAD, so I guess I can rely on that	18:56
	← lamont left	18:57
comp	it is possible to make a branch diff in gitweb .. I can't figure it out ...	18:58
	→ lamont joined	18:59
comp	?	18:59
Ilari	Yeah, 'git reset --hard HEAD^', 'git checkout ORIG_HEAD -- file2.txt' would probably be nicer...	18:59
comp	I mean a diff from one certain commit	18:59
	← sgrimm left	19:01
cnk	OK I am very confused about git submodules. I just tried a pull + commit for my rails submodule. And now, even though I see the listing for it in my .gitmodules file, when I do 'git submodule status', it isn't listed. What have I done?	19:04
comp	-- never mind, used URL editing :)	19:04
netoman	Ilari: Thanks! =)	19:04
	→ etnt joined	19:05
	← Lucine left	19:05
shoe	cnk: did you: git submodule init ?	19:06
	← alley_cat left	19:07
	← ph^ left	19:08
cnk	shoe: when? when I first added the submodule? or jsut now?	19:08
	I don't recall either	19:08
	but before I did the 'git pull', rails showed up in my submodule list	19:08
	now it doesn't	19:08
	← Sho_ left	19:09
	→ felipec joined	19:09
	→ ph^ joined	19:09
	← roylee17 left	19:10
	← Judofyr left	19:10
	→ Sho_ joined	19:10
	→ Judofyr joined	19:11
Ilari	cnk: You did pull of submodule and it has disappeared from submodule list of parent?	19:12
cnk	yes exactly	19:12
	I have a sneaking suspicion that it isn't a submodule any more but perhaps checked into my repository	19:13
Ilari	cnk: Does 'git ls-tree -r \| grep 160000" list it?	19:13
	→ cedricv joined	19:13
cnk	I apparenly need another arg for this	19:14
Ilari	cnk: Oops. 'git ls-files -s \| grep 160000"	19:14
	cnk: (mode of 160000 means submodule).	19:14
cnk	Ilari: no, the only things listed are the same things now appearing in my 'git submodule status' output	19:15
Ilari	cnk: Then it looks like the submodule entry got removed somehow. :-/	19:15
	cnk: Then do 'git ls-files -s \| grep <submodule-mountpoint>' (where <submodule-mountpoint> is where it should be). Does it list files?	19:16
	← FunkeeMonk left	19:17
	← bugfux left	19:17
cnk	Ilari: that command gives me everything that is below the mountpoint. This means my submodule isn't a submodule any more but is committed to my repository as if it were my code, right?	19:18
Ilari	cnk: You did 'git add submodule/'? That does not do the right thing...	19:19
cnk	oh. hmmm so what is the right thing?	19:19
shoe	Ilari: 'git add submodule' (no trailing slash)	19:20
	er: cnk ^^	19:20
	→ _zack joined	19:21
Ilari	cnk: If you committed the supermodule, perhaps first remove the last commit (git reset --soft HEAD^).	19:21
	→ markkald_ joined	19:21
cnk	I am pretty sure I did exactly what is on this blog post http://daniel.collectiveidea.com/blog/2008/4/9/git-submodules	19:21
	Ilari: good plan	19:21
	← markkalderon left	19:21
cnk	--soft leaves the file system as it was, yes? perhaps I should reset everything to before I did the pull, with git reset --hard HEAD^ ??	19:22
Ilari	cnk: It also leaves index as is.	19:23
	→ alley_cat joined	19:23
Ilari	cnk: May be handy if you also had other changes...	19:23
cnk	Ilari: I was just doing that one thing.	19:24
Ilari	cnk: Then maybe 'git reset HEAD^' (reset commit and index, but not working tree)?	19:24
cnk	So after reset --soft, I see deleted: vendor/rails and then lots of new file: vendor/rails/.gitignore, etc	19:24
Ilari	cnk: Does 'git add vendor/rails' complain?	19:25
	← priidu left	19:25
cnk	no it is fine - but changes the list of deleted + add to just "modified: vendor/rails"	19:26
	← hagna left	19:26
Ilari	cnk: Did you use 'git add' or 'git commit -a' (as contained in that blog post) to update the reference. If 'commit -a', then it is pretty worrying...	19:26
	cnk: Then probably 'git commit -c ORIG_HEAD'?	19:27
	→ ruediger joined	19:27
cnk	oh, I didn't read carefully. I just did git commit - no -a but there were already a ton of files added. How the heck did I get there	19:28
	ahhh there it is	19:28
Ilari	That 'submodule/' bug/misfeature is pretty nasty...	19:29
cnk	I did the pull - down in the vendor/rails directory. Then I moved up to the main part of my repository. Then did 'git add vendor/rails' then committed.	19:30
	So clearly git add vendor/rails wasn't what I wanted	19:30
Ilari	cnk: Note that if you hit tab to autocomplete the name, bash autocompletes that as 'vendor/rails/' (which does nasty things).	19:30
cnk	but how could I get the update recognized w/o the git add?	19:30
	ohhhhhh	19:30
	→ doener_ joined	19:30
Mikachu	(in zsh the / is removed if you press enter or type a space)	19:31
cnk	so it is the trailing slash. 'git add vendor/rails' would have been OK but 'git add vendor/rails/' is what killed me, yes?	19:31
Ilari	cnk: Exactly.	19:31
	→ EmilMedve1 joined	19:31
cnk	OK all better now. I used the correct git add and now I just have updated vendor rails as a submodule.	19:33
Ilari	cnk: Also, if the submodule is already there (and you don't need to update anything else), 'git commit vendor/rails' should also work (but watch out for that trailing slash).	19:35
cnk	So what is everyone's favorite set of git docs? I have the peepcode PDF (and screencast). Have tried the users manual on kernel.org (but kept falling asleep). I have been using git for a month now and am still struggling. I want the git equivalent of the Red Bean books for CVS and SVN.	19:35
	← EmilMedve left	19:36
cnk	Ilari: thanks for the help. I really appreciate it - but am frustrated that I need so much help after having used git for nearly a month	19:36
Ilari	cnk: That thing was pretty much a gigantic trap, really surprising and easy to trigger.	19:37
	cnk: And submodule UI is still somewhat work-in-progress (just try to remove submodule. Fun.).	19:39
	← lcapitulino left	19:39
	← tehPope left	19:39
	← _zack left	19:41
cnk	Ilari: is that possible? The last big rescue I got from the list (quite possibly from you again) was when I tried to remove a submodule that I thought was in the wrong place by rm'ing the files and removing the submodule from my .gitmodules. That was a lovely adventure in missing blob references and a trip through ls-tree, etc.	19:41
	→ _zack joined	19:41
	← doener left	19:41
	→ Beket joined	19:42
Ilari	cnk: It should be possible. There are multiple places to deal with...	19:43
	→ lcapitulino joined	19:44
	← alley_cat left	19:45
	→ MarcWeber joined	19:45
Ilari	cnk: Missing blob references? Broken links?	19:45
	→ redondos_ joined	19:45
	→ Chairos joined	19:45
cnk	something I did when I just whacked files gave me dangling references that I needed ? git fsck and some removing to clear up.	19:46
	mostly what I learned from the experience is that, while git has lots of files of metadata, I can't just edit them by hand like I do when I make a mistake in CVS	19:46
MarcWeber	Is it possible to define arbitrary svn directories to track when using git-svn ? In my case I have this layout: A/{trunk,server} B/{trunk,server} C/{trunk,server} where A,B,C are very similar versions of the the same project, trunk is the development version which is merged to server	19:47
Chairos	what do the follows/precedes lines in gitk mean?	19:48
	I can't find anything by googling.	19:49
robinr	cnk: sure you can, it's just not encouraged	19:49
Mikachu	Chairos: it means which tag is the next/previous closest one	19:49
cnk	robinr: well my fingers are still feeling a little burnt from the whole experience. It came out ok but...	19:49
robinr	editing CVS files can hurt too	19:49
Ilari	For some kinds of mistakes, handediting some files is pretty much the only way to recover without great data loss...	19:50
	Such as mistakes with symbolic-ref	19:50
Mikachu	i don't really understand why all git tools fall over saying it's not a git repo when .git/HEAD doesn't start with "ref: refs/heads/"	19:51
	especially git-symbolic-ref should allow you to write a new value	19:51
	→ alley_cat joined	19:53
	→ sgrimm joined	19:54
	→ Sonderblade joined	19:54
	← d0k left	19:55
	← mando left	19:59
	← threeve__ left	20:01
	← redondos left	20:02
	redondos_ → redondos	20:02
	EmilMedve1 → EmilMedve	20:03
	← markkald_ left	20:05
	→ ebel joined	20:05
	← netoman left	20:07
	→ peeja joined	20:09
plopix	how do I clone a repository more than a master branch to another dir on the same host?	20:10
	s.more.with more.	20:11
	I can use git-branch -r to seee the origin branches, but I want a clone of all branches...	20:12
	→ mando joined	20:12
	→ ph^_ joined	20:12
	→ shaftyyy joined	20:14
	→ LetoTo joined	20:15
	→ a-priori joined	20:15
jast	plopix, i don't know to do it in a fully automated way, but you can use git branch localname remotename for each additional branch you want set up for automatic merging etc.	20:16
	← janm left	20:17
jast	uh, you might have to sprinkle in a --track there	20:17
	unless you do "git config branch.autosetupmerge true" before	20:18
	→ robhudson joined	20:19
doener_	plopix: that is a full clone, you have all history locally now. Just no local branches (except for master) yet, on which you would do local work	20:20
	← alley_cat left	20:20
	→ markkalderon joined	20:26
	→ PNC joined	20:26
	← Beket left	20:30
	← shaftyy left	20:30
	← Chairos left	20:32
LetoTo	Is it wrong to do a 'git fetch' on a remote repo and then use git checkout -b branchname?	20:33
	eg git init ; git-remote add foo git+ssh://somewhere/something.git ; git fetch foo ; git checkout -b branchname	20:34
	→ thiago_home joined	20:34
	→ Beket joined	20:34
LetoTo	because running 'git pull' or 'git fetch' after that seems to give an error	20:35
doener_	LetoTo: AFAIK that should git ssh:// not git+ssh://	20:36
	LetoTo: and the checkout command misses a starting point for the new brenach	20:37
	s/brenach/branch/	20:37
LetoTo	well, it picked up all the branches on remote	20:37
	and i can 'git checkout' them fine.	20:38
	← _zack left	20:38
	→ _zack joined	20:38
doener_	s/should git/should be/	20:39
LetoTo	using git-remote add remote ssh:// instead of git+ssh:// throws an error	20:39
doener_	LetoTo: git+ssh gives me an error here	20:39
	LetoTo: which git version is that?	20:39
	→ madewokherd joined	20:39
LetoTo	git version 1.5.4.5	20:39
	(or rather git-1.5.4.5-3.fc9)	20:40
doener_	oh, wait, had three slashes in the git+ssh form	20:40
	← shaftyyy left	20:42
	← thannoy left	20:48
	← lcapitulino left	20:53
	→ johnreilly joined	20:55
	→ p4tux joined	20:57
	→ threeve_ joined	20:59
	→ devogon joined	21:01
	← SRabbelier left	21:04
	→ SRabbelie1 joined	21:04
	SRabbelie1 → SRabbelier	21:05
	← robinr left	21:08
	→ amrox joined	21:09
	→ aroben joined	21:10
	→ bwwww joined	21:12
	← ebel left	21:12
	→ ebel joined	21:12
	← Sonderblade left	21:14
	→ ajonat joined	21:15
vmiklos	is there something in the git C api to store a list of sha1s? path_list seem to be a no-go since it expects the strings to be null-terminated.	21:15
	← ajonat left	21:15
	← markkalderon left	21:16
	→ ajonat joined	21:16
	← krh left	21:16
	← tflsh left	21:16
	← sgrimm left	21:17
	← kumbayo left	21:17
	← radarek left	21:17
	→ sgrimm joined	21:17
	→ eikonos joined	21:19
	→ robinr joined	21:20
	← deavid left	21:23
	← ekidd left	21:23
	← ajonat left	21:26
	→ ZycZyclops joined	21:27
	← Zyclops left	21:27
	← Ace_NoOne left	21:28
	← mjc2 left	21:29
	← aroben left	21:31
	← johnreilly left	21:33
	→ aroben joined	21:34
	→ dave_h_d joined	21:35
	→ rolfb joined	21:39
rolfb	how does one read the url for git?	21:40
	i've started to read it as git or seize	21:40
	← broheem left	21:43
	→ niki joined	21:43
Mikachu	i have never read it out loud	21:43
	← dave_h_d left	21:43
Ramune	sieze?	21:43
	er, seize?	21:43
Mikachu	cz	21:43
nud_	http://kubasik.net/blog/2008/05/19/bazaar-and-its-rockage/#comment-32823 <-- what do you ppl think of this?	21:44
	← bdiego left	21:44
nud_	“Git is only faster if you don’t count the time it takes to GC and repack - tasks that the other tools do automatically. Once you factor those in it is substantially slower [than bzr].”	21:45
	← pygi left	21:45
	← LiamH left	21:45
	→ tflsh joined	21:46
rolfb	git or seize [to exist]	21:47
thiago_home	nud_: depends on your work pattern, I guess	21:47
nud_	thiago, could you elaborate?	21:48
Mikachu	nud_: but you can pack when you don't have to wait for git to finish doing it	21:50
nud_	Mikachu: that's a wise point indeed	21:50
thiago_home	depends on which operatiosn you do more	21:51
	and how often you repack	21:51
pilif	I wonder why everybody keeps constantly comparing the different systems. Each has its advantages. Why can't everyone just settle with a system and stop bitching about the others?	21:51
Ramune	pilif: because it's not as entertaining	21:51
pilif	entertaining? It feels annoying.	21:52
Ramune	scm flamewars are our outlet for aggression -- without it, we'd go postal	21:52
	and geeks with guns going postal isn't a pretty sight	21:52
nud_	eheh	21:52
pilif	guns? I don't actually have guns - despite calling myself quite geeky	21:52
nud_	pilif: actually, my point commenting that post was just that it was easy to take exactly the same arguments for the other system, thus making it useless to argue	21:53
Ramune	that's okay -- if you're a geek, you can build WMDs :)	21:53
pilif	just not geeky enough to participate in flamewars (only in flamewars about flamewars)	21:53
nud_	I was just interested by the gc argument	21:53
robinr	I almost never run gc	21:53
pilif	me neither	21:53
	← Beket left	21:53
pilif	though I'm not what you might call a heavy user	21:53
Mikachu	i usually repack "manually"	21:53
nud_	same here	21:54
	old habits, etc	21:54
robinr	I hardly even repack	21:54
pilif	still... I went from zero knowledge to using-git-daily in about 30 minutes, so I don't understand the fuss about it being difficult to use either	21:54
	and... boy... I love it	21:54
	it completely changed the way I do work	21:54
Mikachu	pilif: it depends on your intelligence level probably :)	21:54
Ramune	mikachu: and how long you've used other scms in the past	21:55
pilif	git-svn was a bit fiddly at first	21:55
comp	.... and tutorial you find :)	21:55
robinr	well mastering everything git can do is hard	21:55
Mikachu	Ramune: the less the better, possibly	21:55
pilif	but I adapted	21:55
robinr	fortunately one only needs ten per cent	21:55
	← cypher23 left	21:55
Ramune	having been using cvs/sccs professionally for over 10 years, it was a bit of a pain unlearning things	21:55
pilif	robinr, exactly. I'm very happy with my 10% for example	21:55
Ramune	took a while to "click"	21:55
pilif	working at multiple features in parallel, taking work from home to the office and back again, sharing patches, discussing ideas, cherry-pick bugfixes... I LOVE it	21:56
comp	unfortunately, not everyone share this kind of software development	21:57
	I tried to "apply" git to one open source project	21:57
pilif	well.. thankfully, me and my collegues do :-)	21:57
comp	made a fork and tried to show git in action	21:57
pilif	at least those not insisting on using windows :-)	21:57
comp	well .... no one showed interest	21:58
pilif	unbelievable	21:58
	it's so easy. so useful.	21:58
comp	I made a nice "banner"	21:58
	(on that forum)	21:58
	people came, people went	21:59
pilif	I guess you need an itch to scratch to get hooked	21:59
comp	someone complained about advanteges in merging	21:59
	as tortoisemerge can do almost the same as vimdiff/kdiff3 interface	21:59
pilif	huh? Complain about stuff working better? Sounds like my coworker :-)	21:59
robinr	if it's easy to merge people will loose themselves in all branches ?	22:00
	s/lose	22:00
	← _zack left	22:01
comp	it's like ... pushing democracy where is dictatorship in charge	22:01
pilif	not sure... everyone should know their branches. Then it's just a matter of communication ("hey... I have this cool stuff x in my branch y... what about having a look at it?")	22:01
	the problem is, that svn (especially TortoiseSVN) seems to work nicely	22:02
comp	yea	22:02
LetoTo	now imagine yourself in a train with no internet...... :)	22:02
pilif	as long as you don't run into the inherent problems of centralized version controlling, you'll never go ahead and try git	22:02
comp	and if they share patches with other via forum and most "larger" patch writers have their own svns to keep their patches updated	22:02
pilif	but one you do, it'll change your world. forever.	22:02
	→ SuttoL joined	22:02
	LetoTo still feels like on his second day of using vi when he's playing with git :P	22:03
	← sgrimm left	22:03
comp	LetoTo: hehe	22:03
	LetoTo: but you know how to shut it down, right?	22:03
LetoTo	there is no ESC-ESC-ESC equivalent. only rm -rf :)	22:04
pilif	since when does ESC-ESC-ESC quit vi?	22:04
comp	I was using ctrl-Z to stop the process	22:04
LetoTo	there is also no git-undo :)	22:04
comp	you're kidding?	22:04
pilif	you don't need any. undo is the core concept of git	22:05
	it's very, very, very hard to screw up in a way you can't undo very easily	22:05
	← sverrej left	22:05
	← rolfb left	22:05
	comp just wonder what can git-undo do ..	22:05
LetoTo	pilif: that's not true :P	22:05
comp	git-reset --hard HEAD@{1} ?	22:05
LetoTo	pushing something to a remote is pretty hard to undo	22:06
	and things like a completely failed 'git merge' :)	22:06
comp	merge can be reseted ..	22:06
robinr	why would merge be hard to undo?	22:06
Mikachu	if you had dirty changes it is quite hard	22:06
pilif	... and pushing to remote is something that's inherently not undo-able for you	22:06
Mikachu	undoing a reset --hard on dirty changes is also hard :)	22:06
LetoTo	if you've done it partially and fixed some conflicts	22:07
comp	Mikachu: ckeckouting them from index & git-clean does the work :)	22:07
LetoTo	anyway, for now, i still use the "cp -a" and "rm -rf" method of undo :P	22:07
robinr	LetoTo: git reset --hard and restart	22:07
comp	checkouting*	22:07
LetoTo	until i get a better grasp	22:07
pilif	I agree... reset --hard always helps	22:07
comp	... in most cases	22:08
pilif	if you are willing to accept some loss of uncommited changes	22:08
	← WorkingGeier left	22:08
	cbrake → cbrake_away	22:09
	→ sgrimm joined	22:09
LetoTo	also, you all assume you know when you did something bad. you might found unknown actions later :P	22:10
pilif	meh... usually I indeed tend to notice if something goes wrong... so far at least	22:10
LetoTo	though i guess you can still cut off a week's work with --hard :P	22:11
pilif	not if you commit here and then	22:11
	which is one of the big advantages of git - that you can commit whenever you feel like it	22:11
	→ ajonat joined	22:12
	→ gregf joined	22:12
	← ajonat left	22:12
doener_	LetoTo: as long as you committed your stuff, the reflog will save it for 90 days (by default), so you can go back to the old commits	22:13
	→ rhett joined	22:13
	→ ajonat joined	22:13
doener_	accidently losing committed work with git is pretty hard	22:14
rhett	hi, i'm using textmate now for my editor instead of vim, and when my commit message comes up, i save it, and close the window, and get a fatal "not commit mesag?" error	22:14
	with vim or emacs, I would just save the message and close	22:15
pilif	rhett, what is $EDITOR set to?	22:15
rhett	/usr/bin/mate	22:15
pilif	use "mate -w" to make the mate command block until you close the window	22:15
	export EDITOR="mate -w"	22:16
rhett	thanks pilif	22:16
pilif	always a pleasure. -w is there precisely for that (to be called from other commands shelling out to $EDITOR)	22:17
	→ BW^- joined	22:17
	→ elliottcable joined	22:17
elliottcable	I'm getting the following error	22:18
BW^-	why does my GIT PUSH give this result:? ! [rejected] master -> master (non-fast forward) error: failed to push some refs to 'ssh://s@domain/home/s/src-repository'	22:18
	?	22:18
elliottcable	even after trying uninstalling and re-installing Git-core:	22:18
	Can't locate Error.pm in @INC (@INC contains: /opt/local/lib/perl5/site_perl/5.8.8 /System/Library/Perl/5.8.8/darwin-thread-multi-2level /System/Library/Perl/5.8.8 /Library/Perl/5.8.8/darwin-thread-multi-2level /Library/Perl/5.8.8 /Library/Perl /Network/Library/Perl/5.8.8/darwin-thread-multi-2level /Network/Library/Perl/5.8.8 /Network/Library/Perl /System/Library/Perl/Extras/5.8.8/darwin-thread-multi-2level /System/Library/Perl/Extras/5.8.8 /Lib	22:18
	BW^-: The connection may have failed, I've gotten that before when it does	22:18
BW^-	hmm. i'm 95% it doesn.t	22:18
LetoTo	someone else made and pushed changes?	22:19
	do a git fetch first?	22:19
	LetoTo <--- git newbie, add grain of salt	22:19
	elliottcable hands salt grain	22:19
BW^-	hehe	22:19
pilif	Error.pm... isn't that part of the core perl distribution?	22:19
elliottcable	pilif: I'd think so, but I uninstalled and re-installed perl too	22:20
	pilif: ... just incase >,>	22:20
pilif	elliottcable, how did you install perl?	22:20
	your @INC looks like the Mac OS default	22:20
	← etnt left	22:20
elliottcable	pilif: MacPorts	22:20
pilif	ok... then Error.pm should be in /opt/local/lib/perl5/vendor_perl/5.8.8	22:21
elliottcable	pilif: but thanks, you inspired me to check - path is screwed up, and it's using built in instead of MP perl d-:	22:21
	Thanks for the help!	22:21
pilif	thought so...	22:21
	which perl	22:21
BW^-	letoto: tested, didn't work.	22:21
pilif	should tell you which one is being used	22:21
BW^-	how succeed with push?	22:22
	← Yuuhi left	22:22
	← ceez left	22:23
	← ph^_ left	22:23
	← ph^ left	22:27
	→ nate_ joined	22:28
nate_	after git push origin master how do I update the repo that I pushed to?	22:29
Mikachu	http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	22:30
	→ CodeOfficer joined	22:30
	← leitgebj left	22:32
ftehw	lets say I want to role back exactly 1 commit (undo everything I've done with my 2nd to last commit). what's the simplest way to do that?	22:33
nate_	Mikachu: thanks	22:34
	→ sverrej joined	22:35
charon	ftehw: if you've published it, git-revert. if you haven't, you can rebase --interactive HEAD~3 and delete that commit from the list that shows up and it will rewrite history to appear as if it was never there in the first place	22:37
	→ dbrummer_ joined	22:38
	← nate_ left	22:39
Ingmar	How do I the current git-dir in a git hook?	22:39
dbrummer_	hello	22:39
	← gregf left	22:40
	← chris2 left	22:40
dbrummer_	I was wondering if someone can help me with an issue I'm having with git	22:42
	← eternaleye left	22:42
ftehw	thanks charon	22:42
	→ dotsintacks joined	22:45
	← dbrummer_ left	22:45
	← nud_ left	22:46
dotsintacks	question: I've got a project I've been hacking on that came from a tarball on a website. the original author just released a new tarball with lots of changes, and I want to merge in his fixes. To help git make this easier on me, should I create a branch from the /past/, from the original tarball revision, and then apply the new tarball changes, and then merge those changes into my branch?	22:46
	→ mithro joined	22:48
	← ebel left	22:48
	→ warlock_mza joined	22:48
jast	dotsintacks, that's what i'd do at least	22:51
	if you want to submit back patches at some point, consider using rebase rather than merge	22:51
	← vbgunz left	22:52
ftehw	if there's only 1 commit (a toy project), why doesn't `git checkout -b old abc` work if abc06a0a90fada443cde0ddeeae3b931ae49f3c9 is the commit.	22:54
	seems to me that 'abc' uniquely identifies it, trivially.	22:54
Mikachu	and with abc0?	22:55
ftehw	abc0 does work.	22:55
Mikachu	it might be that you need 4 chars	22:55
jast	it might be that checkout is afraid of mistaking an all-chars abbreviated sha1 id for a ref name	22:56
	no, that's not it	22:56
	← corporeal left	22:57
jast	four chars is the minimum for identifying a sha1 id, it seems	22:57
ftehw	that seems pretty arbitrary	22:57
jast	abc likely won't work anywhere else either	22:57
ftehw	jast: u mean on other git commands?	22:58
jast	yeah	22:59
	i tried with git log	22:59
	← cedricv left	23:00
ftehw	Seems like an undocumented restrictions. man git-rev-parse search SPECIFYING, first bullet implies 'abc' should work, as it matches uniquely, trivially.	23:00
	← cannonball left	23:01
	→ leitgebj joined	23:01
jast	i suggest you complain on the mailing list ;)	23:02
ftehw	jast: i guess in regular use u rarely reference commits by their hash, right?	23:04
context	why would git say a file is modified but git-diff shows nothing	23:04
jast	depends. i sometimes do.	23:04
context	oh wait	23:04
	nm its cached	23:04
	← heythisisdave left	23:05
	← charon left	23:05
ftehw	OK, new question: what's the simplest way to just undo all changes since the commit?	23:07
	→ ekidd joined	23:07
jast	git reset --hard to undo including working tree, git reset otherwise	23:07
ftehw	jast: u mean excluding?	23:08
jast	no, --hard means: reset index as well as working tree to state of last commit	23:08
	without --hard only the index is reset	23:08
	untracked files are not affected in either case	23:09
ftehw	jast: hrm, i guess i need to read up a little more. :)	23:09
jast	it certainly can't hurt :)	23:10
	unless it confuses you, but then this channel will still be here :)	23:10
	← elliottcable left	23:11
	→ TuxCM joined	23:14
	← redondos left	23:22
dotsintacks	how do I tell git that I resolved a conflict?	23:23
	I even used git mergetool, but "git commit" is telling me that the files are still in conflict.	23:23
	→ humbolto joined	23:23
humbolto	how can I add something to the last commit description?	23:24
felipec	dotsintacks: git add	23:26
	→ threeve joined	23:26
felipec	humbolto: git commit --amend	23:26
humbolto	tnx	23:26
dotsintacks	ok...I saw mention of git-update-index but I remember being told to stay away from the three word incantations :P	23:27
	← threeve left	23:27
jast	git add does the same thing, but it's shorter to type and can do a wider range of things and abstracts away a couple of things	23:27
dotsintacks	jast: gotcha. thanks	23:28
gebi	how to pull to a branch without checking it out?	23:30
	← mando left	23:31
gebi	e.g only if it's fast-forward (which doesn't matter as it's a tracking branch)	23:31
	→ danlarkin joined	23:31
	← EmilMedve left	23:32
jast	you mean it's a remote branch? that's pretty much what git fetch does	23:33
	← ekidd left	23:33
gebi	it's a tracking branch of a remote branch	23:33
	← warlock_mza left	23:34
jast	what i mean is that it shows up in git remote -r but not in git remote. those are tracking branches.	23:34
	err, remote branches. or something. the terminology is a bit muddled in my head.	23:34
	though you can use fetch either way	23:34
gebi	fetch can update local branches?	23:35
	i've a lokal branch which is a tracking branch of a remote one, can i update this lokal branch fast-forward (otherwise error) without switching to it?	23:36
jast	you mean you have a tracking branch and want a local branch to fast-forward to its current state?	23:36
	← ruediger left	23:37
	← BW^- left	23:37
gebi	yea, i mean i want _the_ lokal tracking branch to just fast-forward to the remotes last changeset	23:37
jast	local is a bit ambiguous here	23:37
	← niki left	23:38
	← Ryback_ left	23:38
jast	tracking branches are typically called remote even if they are local copies of branches from other repositories, whereas local branches mean branches that you created yourself, typically based on one of the tracking branches	23:38
gebi	01:33 < jast> what i mean is that it shows up in git remote -r but not in git remote. those are tracking branches.	23:38
	← sbeyer left	23:38
gebi	01:33 < jast> err, remote branches. or something. the terminology is a bit muddled in my head.	23:38
	01:34 < jast> though you can use fetch either way	23:38
	doh sorry, my cat :/	23:38
doener_	gebi: git push . origin/foo:foo	23:39
	gebi: that might do	23:39
gebi	doener_: i need to update the tracking branch without switching to it	23:40
doener_	gebi: since when does push switch branches?	23:40
	← felipec left	23:40
	← robhudson left	23:41
	→ BW^- joined	23:41
doener_	gebi: that command just locally pushes whatever is in origin/foo into foo. Having checked out foo would even be rather bad in this case	23:41
	← threeve_ left	23:42
	← SuttoL left	23:45
gebi	ok	23:46
	→ vbgunz joined	23:50
	← robinr left	23:55
gebi	doener_: is there a way to make the git push . origin/master:master implizit on every fetch?	23:55
	← devogon left	23:56
Arrowmaster	use pull?	23:57
	← goibhniu left	23:57
doener_	maybe some hook. But what's the point? You can write just use origin/master instead	23:58
	s/write just use/just write/	23:58
	I wonder how I managed to do that...	23:58
gebi	Arrowmaster: without beeing on the branch? i've a few tracking branches i want to upgrade in one go	23:59
doener_	after all, pushing into master all the time, means that you won't work on master (otherwise the push will fail), so having that done automatically on fetch would mean that always master == origin/master	23:59

Logs Search ←Prev date Next date→

Channels Documentation