IRCloggy #git 2007-11-11

Provider of IRC logs since 2005.

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

	2007-11-11
iabervon	That's very odd.	00:00
	← dshep left	00:03
kelvie	can you add a fork as a ref to an existing repo?	00:05
diacritical	by fork i think you mean "branch" ? then its already a ref	00:07
kelvie	naw... say repo.git is a fork of git.git	00:07
diacritical	what is a fork?	00:07
kelvie	can I make repo.git a ref inside my existing git.git repo?	00:07
	well it's essentially a branch :P	00:07
vmiklos	add repo.git as a remote then you can fetch it	00:07
kelvie	but located elsewhere	00:07
diacritical	anyway with git-push in 1.5.3.5 right now i'm having to git-push each branch explicitly. --all doesnt seem to work	00:07
kelvie	ah.. how dyou add a remote?	00:07
	manually?	00:07
	(i.e. editing .git/config)	00:08
	→ ebel joined	00:09
kelvie	ah git-remote does that	00:10
diacritical	one bad move with git-remote usually means editing manually anway :p	00:11
kelvie	hehe.. yeah	00:14
	it seems it won't let me git-pull	00:14
	← meandtheshell left	00:18
	← metafollic left	00:20
	→ metafollic joined	00:21
iabervon	dia: you're not using --tags, are you?	00:22
diacritical	iabervon indeed i was. smacks head of all the combinations I tried, i missed dropping --tags	00:30
iabervon	Right. --tags counts as a refspec.	00:31
	→ jkimball4 joined	00:42
	→ robfitz joined	00:51
	→ ebel_ joined	00:54
	← ebel left	00:54
	ebel_ → ebel	00:54
	→ marcheu joined	00:57
	← marcheu left	00:58
	→ meandtheshell joined	00:58
	← krawek left	01:02
	← greendale left	01:02
	← robfitz_ left	01:06
	← ShadeHawk left	01:07
	→ ShadeHawk joined	01:08
	← Sonderblade left	01:11
	← vmiklos left	01:12
	→ vmiklos joined	01:12
	← ShadeHawk left	01:17
	← devogon left	01:21
	← mithro left	01:22
	← diacritical left	01:22
	→ FunkeeMonk joined	01:23
	← meandtheshell left	01:24
	→ alley_cat joined	01:25
	← djpig left	01:25
	→ djpig joined	01:30
bobesponja	hey	01:40
	→ DrNick joined	01:44
	← jlh left	01:45
	→ jlh joined	01:46
	→ dashepar joined	01:52
	← Ryback_ left	02:01
	← RobotDeathSquad left	02:02
	→ \ask joined	02:07
	← cmarcelo left	02:08
	→ mutex joined	02:10
	→ ebel_ joined	02:20
	← ebel left	02:22
	ebel_ → ebel	02:22
	← ferdy left	02:28
	→ RobotDeathSquad joined	02:36
	→ priidu joined	02:37
	← myrizio left	02:38
	← HackyKid_ left	02:43
	← orospakr left	02:44
	← djpig left	02:51
\ask	I know it's the FAQ of all FAQs, but take this as another vote for "git reset/revert really needs fixing". I can NEVER remember how to just revert a single darn file.	03:02
	← mutex left	03:04
	→ mwalling joined	03:04
iabervon	You mean "checkout"?	03:05
	→ mutex joined	03:06
	← bobesponja left	03:08
vmiklos	\ask did i say i never tried to use git revert for something it's not supposed to? ;)	03:09
	← ebel left	03:11
	← alley_cat left	03:11
	← context left	03:12
	→ context joined	03:12
iabervon	I understand people not finding the way to get one file back to the last committed state, but it should be easy to remember than it's neither of the obvious things...	03:12
Tene	iabervon: you mean checkout?	03:14
iabervon	Yeah. When people can't find that checkout does that, they think it's either reset or revert, but think they've forgotten which.	03:17
context	iabervon: git co filename ?	03:18
	oh nm	03:18
	iabervon: just takes a little time to stray from old habbits. ;)	03:19
iabervon	I still remembered "cvs checkout filename" from way back when, and I'd mostly used arch in between, and never figured out how to do it with arch, so I was not confused.	03:20
	I bet if we made it "git retrieve filename", it would be perfectly sensible, and yet drive new users totally mad.	03:21
	← dashepar left	03:23
	→ spearce joined	03:27
	→ jhmoore719 joined	03:31
	→ bdiego joined	03:36
	← fdr- left	03:39
	→ fdr- joined	03:40
	→ orospakr joined	03:43
	→ Sonderblade joined	03:48
	← fdr- left	03:52
	→ fdr- joined	03:53
mithraic	Maybe it should be "git oh oops give me the old one"	03:56
	← jkimball4 left	03:59
	→ ramza3 joined	04:10
ramza3	strange task ahead for me.	04:10
	I branched a project in subversion and they are several months apart. I was thinking I might use git to actually merge them as opposed to subversion. Does that question even make sense?	04:11
	← mutex left	04:12
arw	yes it does, i've heard of some people using svk or git as a merge replacement in svn, because svns merge sucks so much.	04:13
	but i've never really done that myself.	04:13
ramza3	arw: merge in svn sucks hard	04:13
iabervon	I think so, although I'm not sure how to do svn imports with branches.	04:14
ramza3	actually, I may just remove all the svn stuff and just create two new git projects	04:14
arw	my only experience in subversion was a completely hosed source tree when somebody forced me to work on a windows box with only tortoisesvn installed.	04:15
	i really never want to repeat that week, ever :)	04:15
ramza3	I could be wrong (and I have been using svn for 4 years), I consider subversion as a source repository not that good a source control system	04:16
	but then again, that applies to CVS and a lot of the other tools out there	04:16
iabervon	If you can get the history, it'll make the merge a whole lot better.	04:17
arw	having the history is always nice.	04:18
iabervon	Although maybe you should just make tar files of each of the latest commits in the branches, and one of the point where they forked, and import each of those into git, in the appropriate organization.	04:18
arw	i would really recommend against that. not for technical reasons, just because you never ever want to throw your development history away.	04:19
	G_SabinoMullane → Z_SabinoMullane	04:19
ramza3	arw: hmm, the git/svn tools are pretty good?	04:20
arw	having that history available opens a bunch of very precious possibilities like bisect you would otherwise not have.	04:20
	ramza3: mostly, yes. there are some corner cases, but just give it a try, mostly it will just work.	04:21
	→ zapnap joined	04:21
	← sgrimm left	04:21
\ask	iabervon: yeah, I always think it's reset (because I have learned what revert does), but that I just don't remember which option to use. :-)	04:22
ramza3	arw; hmm, I wonder if there is a svn to git export, keeping the svn information on the initial import into git and then removing it afterwards. I dont plan on using svn after it is into git	04:22
iabervon	After you import into git with "git svn", you can just discard the extra info for the translation.	04:24
	← fdr- left	04:26
arw	git svn should just import all your commits, the history should look the same as in svn.	04:26
tpope	may want to map those usernames to email addresses though	04:28
	→ fdr- joined	04:28
ramza3	all I can do is try it	04:32
	← Z_SabinoMullane left	04:42
	→ jp_n9 joined	04:44
	→ mutex joined	04:54
	← jhmoore719 left	05:00
ramza3	hmm, git-svn doesnt seem to work on local working directories	05:01
iabervon	No, it needs the actual repository, which is what has the history.	05:02
	← saintdev left	05:07
tchan	git-svn actually needs access to a svn server. A local svn repo has history in it. Most local users don't have their local repos setup to use a local svn server.	05:09
	← mithraic left	05:09
	→ mutex_ joined	05:11
	← mutex left	05:11
	← orospakr left	05:19
ramza3	well, I meant some applications can process the data from the .svn working directory. Some dont	05:21
	→ orospakr joined	05:22
	mutex_ → mutex	05:26
	← rkaway left	05:27
	← bdiego left	05:28
	→ rkaway joined	05:29
	→ sgrimm joined	05:35
	← defunkt left	05:38
ramza3	is it possible to run git-daemon without any args, is there a default repository directory git-daemon will look for "/pub/scm"?	05:42
	→ mithraic joined	05:45
tpope	it has a --verbose option, why not try it and find out	05:51
up_the_irons	ramza3: i think the default is to use the current directory, but i might be wrong	05:53
	→ jhmoore719 joined	05:53
	→ mithraic_ joined	05:54
iabervon	Default is to not limit based on path, it looks like.	05:55
	← kukks left	06:04
	← mithraic left	06:10
	← jhmoore719 left	06:11
	→ saintdev joined	06:15
	→ cehteh joined	06:20
	→ mutex_ joined	06:32
	← mutex left	06:33
	mutex_ → mutex	06:33
	← mutex left	06:40
	→ mutex joined	06:41
spearce	i hate how git makes me work harder at writing the commit message than at writing the code. its too easy to write a good commit message, and even easier to find one later to better understand the code.	07:01
	spearce just spent like an hour writing a commit message for a 1 line fix.	07:01
iabervon	More useful than writing a long cover letter for a 1-patch series...	07:01
	← glommer left	07:05
	← madewokherd left	07:24
	→ jerbear joined	07:29
	→ SuttoL joined	07:30
	← priidu left	07:34
	← Sho_ left	07:38
	→ aroben joined	07:49
	→ priidu_ joined	07:55
	→ male joined	08:01
	← FunkeeMonk left	08:04
bdowning	spearce: aroundp?	08:08
spearce	yup	08:08
bdowning	Is there any way to quote filenames for the rename and copy commands in fast-import so they can have spaces?	08:08
spearce	use "" with c-style quoting.	08:08
	e.g. "" with \", \\, \n to escape.	08:09
bdowning	Ah, excellent. Thanks.	08:09
spearce	at a minimum you need to escape those three (", \ and LF) but otherwise you can dump the raw data into the string. or escape more if you want. it understands a lot of escapes.	08:09
bdowning	Geez, it's right here in the man page too. How embaressing.	08:10
spearce	its also one of the biggest man pages in git. :)	08:10
bdowning	I'm toying with a monotone2git, mostly because I'm bored. It never ceases to amaze me just how slow monotone is.	08:11
spearce	they really haven't improved its speed? i know linus looked at it and rejected it for performance reasons.	08:12
bdowning	Not so much. (caveat: I'm using whatever's in Gutsy, which is two minor versions behind). Pulling the repo for monotone itself took about 15 minutes of cpu time (!)	08:13
	I think that is actually much better than when Linus saw it too.	08:13
spearce	i couldn't imagine running the git of today on a 486 with 16M of memory. but git is pretty quick. :)	08:14
bdowning	I think the main problem mtn has is that is was simply designed to have any part of it replaced and hacked on without affecting the other parts. So data goes through three or four translations between database and usage.	08:15
spearce	yea, flexiblity like that will always bite you on performance.	08:16
	good for tinkering though.	08:16
	← orospakr left	08:19
bdowning	I suppose, if you like Boost-heavy C++. I found hacking on Git extremely clean and refreshing, personally, but maybe that's just me. :)	08:19
spearce	no, me too. one of the reasons i hack on git is because its pretty clean and easy.	08:19
	→ jerbear_ joined	08:34
	← jerbear_ left	08:35
	← samsarin left	08:45
	→ tcoppi joined	08:47
	→ Odd_Bloke joined	08:48
Odd_Bloke	Hey guys, I accidentally committed to the wrong branch. How can I revert the branch to the state it was in before the commit?	08:48
thiago_home	first, have you already committed to the right branch?	08:49
cehteh	(cherry pick would do that)	08:49
Odd_Bloke	thiago_home: I forgot to create the branch before committing. I have now done so, so the correct branch has the revision.	08:50
	← tcoppi left	08:50
tpope	git-reset can do it	08:50
thiago_home	ok, so use git-reset to reset your branch to what it should have been	08:50
cehteh	Odd_Bloke: git checkout -b newbranch	08:50
tpope	unless you have great faith in your skills, I'd do git diff HEAD^ > /tmp/my_changes.patch first	08:50
cehteh	wil then create the new branch for you with that commit	08:50
	after that you switch back to the old branch with git checkout oldbranch	08:51
thiago_home	tpope: that's why I asked him if he had committed to the right branch first	08:51
cehteh	and do a git reset --hard HEAD^	08:51
	but be careful ..when done wrong you loose the commit	08:51
tpope	really, what would be wrong with git reset --soft HEAD^; git checkout target_branch	08:52
Odd_Bloke	'git reset --hard HEAD^' has done the trick.	08:52
	Thanks guys. :)	08:52
tpope	then you'd get your changes over to the right branch automatically	08:53
cehteh	tpope: maybe .. didnt thought about that	08:53
tpope	grep 'reset --hard' ~/irclogs/freenode/\#git.log\|wc -l # 144	08:54
	grep 'reset --soft' ~/irclogs/freenode/\#git.log\|wc -l # 8	08:54
	--soft strikes me as almost as useful as hard, I'm just wondering why the former comes up so much more often	08:55
thiago_home	because, I guess, people don't remember the difference between --soft and --mixed	08:56
	whereas --hard is easy to remember	08:56
	← jerbear left	08:58
tpope	I suppose	09:03
	git reset --mixed sounds a lot like git rm --cached	09:05
halorgium	tpope: cause --hard is used for removing the commits	09:05
tpope	hard reverts the working tree too though	09:06
	generally, if I want to remove the commit, it's because I want to modify it or apply it elsewhere. very seldom do I want to do "discard my work. no really, discard it"	09:07
	(soft removes commits too)	09:09
	← Ademan left	09:13
	→ kahmalo joined	09:17
kahmalo	I have PATH="/home/Kalle/prefix/i686-pc-linux-gnu/bin:/home/Kalle/any-arch/bin:/usr/local/bin:/usr/bin:/bin:/usr/bin/X11:/usr/games" and EDITOR="emacsclient --current-frame". /usr/bin/emacsclient does not understand --current-frame but /home/Kalle/prefix/i686-pc-linux-gnu/bin/emacsclient does. Now when I run "git commit" in Git 1.5.3.1, it prepends /usr/bin to $PATH and runs the wrong emacsclient. Is this a bug and has it been fixed	09:19
	in a later version?	09:19
	The git binaries are in /usr/bin.	09:20
spearce	its a feature. we always force gitexecdir (in your case apparently /usr/bin) to the front of PATH so we can find all of the git plumbing tools.	09:21
kahmalo	If I run git-commit directly it works OK.	09:22
	Couldn't you explicitly use $gitexecdir/whatever instead of relying on PATH?	09:23
	Or would that again break when running e.g. git-commit directly...	09:23
spearce	not without breaking a lot in the process. so we went with shoving $gitexecdir in the front and suggesting distros set that to a location that isn't /usr/bin, and only installing git into /usr/bin. apparently your distro didn't do that.	09:24
kahmalo	This is Debian but the package is a bit old.	09:24
spearce	what about setting $EDITOR to the right emacsclient path?	09:24
	the debian maintainer is relatively active. he's quite responsive to bugs filed against the git package.	09:25
kahmalo	Well I have to set $EDITOR from within ~/.emacs.d/init.el to get sane behaviour and I'd prefer not putting architecture-specific directories there.	09:26
	rather rely on $PATH set in .bash_profile.	09:26
	→ mithro joined	09:27
kahmalo	http://packages.debian.org/sid/git-core/i386/filelist shows it's being installed in /usr/bin in the latest Debian package too.	09:28
	Where is this recommendation about not installing into /usr/bin?	09:28
spearce	apparently the mailing list archives. its not exactly documented in the source tree to use gitexecdir to avoid installing into /usr/bin	09:29
kahmalo	In fact git-1.5.3.4/INSTALL suggests prefix=/usr.	09:31
spearce	well prefix isn't gitexecdir. we default gitexcdir=$(prefix)/bin so i guess we implicitly are suggesting /usr/bin. sigh	09:32
	← iabervon left	09:39
	← DrNick left	09:41
	→ tcoppi joined	09:43
	→ HackyKid joined	09:44
	→ smartbhai joined	09:45
smartbhai	hello people im new to git, how can u use git such that it doesnt hang when the network connection is lost	09:46
thiago_home	by using it	09:47
	since it doesn't use the network unless you're pushing or pulling, it couldn't care less about the network	09:47
	→ nessundorma joined	09:49
smartbhai	im using git-clone	09:49
	→ csc`` joined	09:49
thiago_home	ok, once it's cloned, the network isn't used	09:49
	→ aroben_ joined	09:51
	← aroben left	09:58
	aroben_ → aroben	09:58
	← csc` left	09:59
	→ Sho_ joined	10:02
	→ Oejet joined	10:02
	→ Ademan joined	10:06
	→ devogon joined	10:06
smartbhai	hi im totally new to git, im using this git-clone git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git when i start downloading files the git speed gets reduced ( n if i kill the instance and start again it downloads all the files againaturally) and it doesn respond after that	10:11
	hi im totally new to git, im using this git-clone git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git when i start downloading files the git speed gets reduced ( naturally) and it doesn respond after that.. if i kill the instance and start again it downloads all the files again	10:11
	→ gitte_ joined	10:16
	→ pennyk_ joined	10:16
gitte_	spearce: still awake?	10:16
spearce	yup	10:16
gitte_	wow	10:16
cehteh	:)	10:17
gitte_	spearce: so will this week also be too much work for you?	10:18
spearce	well, i got quickfetch out. again. :)	10:19
gitte_	Heh.	10:19
	I did not have time to review it, at all.	10:19
spearce	but maybe things will slow down a bit for me. at least i'm paid by the hour.	10:19
gitte_	Nice thing.	10:19
spearce	good catch on the free thing though. i totally missed taht.	10:19
gitte_	Don't remember.	10:19
spearce	"a = NULL; free(a)" is a really bonehead mistake.	10:20
gitte_	Ah, but that was just a swap.	10:20
	I do that all the time.	10:20
cehteh	mhm	10:20
	cehteh does that only when a is intended to be reused	10:21
gitte_	;-)	10:21
spearce	othewise you don't bother to free the value in a ? :)	10:22
gitte_	RAM is cheap ;-)	10:22
cehteh	meant the null'ing	10:22
spearce	tell that to sun, whose $#@!*@#$(@!#(! jvm won't start when you have 7G free and only need 64M.	10:22
gitte_	Hehe.	10:22
	I _know_ that problem.	10:22
cehteh	haha signed size crap?	10:22
gitte_	But didn't they solve that in Java6?	10:23
spearce	this is java 6.	10:23
	f'ing pile o' crap that jvm.	10:23
gitte_	cehteh: no, it is a totally bogus 64/32-bit problem.	10:23
cehteh	sure	10:23
gitte_	At least it's GPL.	10:23
	Which is something.	10:23
spearce	have you read the source to the vm? its uh, not pretty like git.	10:23
gitte_	I have read parts of them.	10:24
	But git did not exist by then, so I was not spoilt.	10:24
cehteh	mhm ..	10:24
	cehteh looks up the source of the yap prolog vm .. just for reference :)	10:24
cehteh	http://www.pipapo.org/gitweb?p=yap-cvs;a=blob;f=C/absmi.c;h=d0c50b6194a78a6f70685299f1b88e94f4aa1af3;hb=43cb20694d5c600f317eb2e2e5604034b1382dc7#l543	10:25
	... which line does that function end? :)	10:25
	cehteh runs	10:26
spearce	i want whatever they were smoking when that was written.	10:26
gitte_	Hehe.	10:26
	→ \|ODIN\| joined	10:26
cehteh	well .. if you know its history it has some reason	10:26
Ilari	I suppose it is worse than that 900-line switch statement I wrote once...	10:26
spearce	heh.	10:27
gitte_	What was it that Theo wrote the other day? "You must have been smoking something very mind altering, and you should share it."	10:27
spearce	i once worked with a guy who wrote a perl script to generate a b-tree in if/else statements.	10:27
	to match a value in a range 0-16384.	10:27
cehteh	it was once translated from asm to C .. in the mid 80's ... and it is the fastest prolog vm out there	10:27
spearce	he choked when i told him a switch would have been nicer.	10:27
cehteh	it needs some love and refactoring .. but well as all such projects its sparse of developer resources	10:27
gitte_	cehteh: probably automated translation...	10:27
cehteh	gitte_: nope .. by hand :)	10:27
gitte_	cehteh: oh. Not fun.	10:28
	spearce: you shoulda told him what is the only sane thing: lookup tables.	10:28
	spearce: in 16-bit you need half a segment for that ;-)	10:28
spearce	this was java. all bets were off anyway on memory/perforamnce. :)	10:28
gitte_	At that time, yeah.	10:29
spearce	he did this huge nasty if/else thing to avoid a virtual function call.	10:29
gitte_	Until kaffe kicked their but.	10:29
spearce	i was like "wtf!"	10:29
gitte_	But that was not mid 80s	10:29
HackyKid	i think even java has a switch that uses a jump table or something	10:29
spearce	yea, it does	10:29
	you should have seen him trying to debug the inner part of the output file. "column 19,890, line 200,019" ... :)	10:30
HackyKid	hehe :-)	10:30
gitte_	Aaargh. I'm almost ready to write a patch to builtin-add which automatically calls git-remote when the file "remote" does not exist and you say "git add remote"!	10:31
swetland	hacky: java provides two switch opcodes, one which is a sorted value/addr table, one which is a direct indexed table of addresses	10:31
gitte_	spearce: lol	10:31
spearce	gitte: i do that a lot and expect git to dwim...	10:31
HackyKid	swetland: yeah, i thought so :-)	10:31
gitte_	Is git.kernel.org down or what?	10:32
swetland	timing out for me	10:32
gitte_	Yeah, me too.	10:33
\|ODIN\|	Hi there, I need some help: Trying to commit with GIT, but it says that I have suspicious path lines, and then it lists me some just changed files saying that there are trailing white spaces, while actually there aren't. It refuses then to commit. What should I do in this case?	10:33
gitte_	Windows?	10:33
\|ODIN\|	no, Gentoo	10:33
swetland	I need to figure out how to get send-email to do something sane	10:33
gitte_	Seems it automatically activates your hook.	10:33
swetland	it is currently the bane of my existence ^^	10:33
gitte_	\|ODIN\|: try chmod a-x .git/hooks/*	10:33
\|ODIN\|	yes, it works!! :)	10:34
	Thanks.. could You explain me quickly what happened? I'm stil at the beginning with this	10:34
	→ greendale joined	10:34
gitte_	\|ODIN\|: there are quite a number of hooks in git.	10:35
	→ iXce` joined	10:35
gitte_	These are small scripts that get called whenever you do a certain action.	10:35
	Something like pre-commit.	10:35
\|ODIN\|	i see	10:35
gitte_	Those scripts live in .git/hooks/	10:35
	They are activated by chmod +x	10:36
	I have the problem on a network share, mounted via cifs. It has problems with keeping track of the +x bit, and so I regularly do "rm -rf .git/hooks" on that share.	10:36
\|ODIN\|	is a check on the trailing whitespaces so essential?	10:37
gitte_	It is convenient, since those things invariably tend to get fixed. In one monster commit.	10:38
\|ODIN\|	Would be bad to delete the script which chechs them? I think it happens quite often to forget a whitespace somewhere, without having troubles from theat	10:38
	I see	10:38
gitte_	I do not recall off the top of my head where the templates are stored, but you'd have to change it in .../templates/hooks/.	10:39
HackyKid	gitte_: btw, i removed generated stuff from the msys-perl thingy, and split up their changes into seperate patches:	10:39
	Import interesting stuff from perl-5.6.1-MSYS-1.0.11-src.tar.bz2. - 97 files changed, 2557 insertions(+), 748 deletions(-)	10:39
	Import .def files from perl-5.6.1-MSYS-1.0.11-src.tar.bz2 - 19 files changed, 3496 insertions(+), 0 deletions(-)	10:39
	Import test changes from perl-5.6.1-MSYS-1.0.11-src.tar.bz2 - 80 files changed, 1652 insertions(+), 361 deletions(-)	10:39
	Import CYGWIN-PATCHES/ from perl-5.6.1-MSYS-1.0.11-src.tar.bz2 - 166 files changed, 68875 insertions(+), 0 deletions(-)	10:39
	you only need the first one to compile	10:40
gitte_	HackyKid: oh darn, I meant to send you a remark on what I did Friday night...	10:40
HackyKid	not sure what the .def files are for.... they don't seem to be used	10:40
	oh? :O	10:40
gitte_	In CYGWIN-PATCHES, there is a patch called "perl-5.6.1-2.patch".	10:40
	Appears to me that the diff after applying that is pretty small.	10:41
	But I think it did not compile.	10:41
	So your work is definitely not in vain.	10:41
	spearce: I'll hold off that builtin-add patch until I completely built in git-remote.	10:42
HackyKid	i also fixed it so you don't need to overwrite your libc.a and libm.a	10:42
gitte_	Very good!	10:42
	How?	10:42
HackyKid	(well, actually that was just a cherry pick from my previous attempts)	10:42
	there's a "fix issues linking with -lm" there that prevents the need for this overwriting	10:43
gitte_	Ah.	10:43
	BTW it's a really cool thing that you set up something quilt-like.	10:44
HackyKid	:-) seemed easiest at the time	10:45
gitte_	And indeed it is, but not only that time.	10:46
HackyKid	yeah, indeed	10:46
	→ meandtheshell joined	10:47
swetland	is there some way to generate the mail bodies that git-format-patch + git-send-email would generate without actually sending the email?	10:47
spearce	you mean git-format-patch ? :)	10:47
gitte_	--mbox?	10:48
swetland	git-format-patch doesn't put the author's From: line in the "body" of the generated message	10:48
gitte_	That's news to me.	10:48
spearce	oh.	10:48
	→ ferdy joined	10:48
spearce	if the author isn't you it leaves the from in the mbox position.	10:48
	send-email must be moving it down into the body.	10:49
swetland	yeah	10:49
gitte_	Ah, I misread.	10:49
iXce`	pasky: ping	10:49
gitte_	swetland: but why do you need that?	10:49
swetland	because git-send-email is a little crazy with the cc'ing and mangles the encoding of the name of my coworker who I'm submitting patches from ^^ though the latter looks like there is a fix in the works for	10:50
	I need to roll up my sleeves and start poking around in the git sources one of these days	10:50
	→ kumbayo joined	10:53
swetland	I think part of the issue is just a workflow flaw on my side... I'm posting my patches from my work email address (since it's stuff done for work), but the address I use for the actual sign-off, patch authoring, etc is a neutral one (since this linux work will likely outlive my employment)	10:54
	← \ask left	10:54
gitte_	Maybe there is a --dry-run?	10:54
	→ Eludias joined	10:54
swetland	there is, and it has helped a bit figuring out what's going on	10:54
gitte_	spearce: http://git.pastebin.com/m1a2415a	10:55
cehteh	does git-cvsimport store some metadata somewhere? .. i try to reset a tree to a older commit	10:55
	and then rerun it with a fixed AUTHORS file .. one new got added	10:56
	just resetting master and rerun cvsimport breaks history	10:57
gitte_	You need to reset origin, not master.	10:58
cehteh	i have only master :P .. blame me but thats the way this repo was set up	10:59
gitte_	Umm. Sorry. Then I'm out of clues.	10:59
cehteh	heh well master/origin are just names	11:00
	$ git fsck --full	11:01
	error27: 118a4a6085f4fbf1d597d9248c9d86444f93826c: invalid sha1 pointer in cache-tree	11:01
	missing blob 28009dc12bc54f54b5917aa972d57849bb81aa7e	11:01
	...	11:01
	uhm	11:01
gitte_	Ouch.	11:02
	cache-tree should be either in sync with the objects in the index, or invalidated.	11:02
cehteh	.. and many many more	11:02
	→ \ask joined	11:02
cehteh	well i am working on a copy .. checking the original repo now	11:02
gitte_	Easiest "fix" is to "git reset --mixed HEAD".	11:02
	But I wonder how that could happen.	11:03
	cehteh too	11:03
cehteh	i just cp -a the repo	11:03
	← nessundorma left	11:04
	→ tzafrir joined	11:06
cehteh	ok main repo is ok	11:10
	→ redondos_ joined	11:10
	← daveinthesky left	11:11
	← spearce left	11:11
	← redondos left	11:13
	redondos_ → redondos	11:13
	← \ask left	11:14
	→ nud joined	11:14
	← pigeon left	11:16
cehteh	mhm	11:19
	somehow history got chopped	11:21
	shrug	11:21
	non important mirror .. but still	11:21
gitte_	AFAIR it's the committer date that cvsimport uses to determine what is still to be fetched.	11:24
HackyKid	yay, now compiling even works with spaces in your $PATH	11:27
gitte_	Cool.	11:28
HackyKid	now i can do 'MSYSTEM=MSYS && ./Configure -de && make' to compile	11:28
gitte_	Hey, another idea: how about moving /perl to /share/msysPerl, and have a single script there which checks if it still needs to download/unpack/patch/build?	11:29
	Maybe even more... install/update-msysgit?	11:30
cehteh	gitte_: well somehow the whole thing got borked	11:31
gitte_	:-(	11:31
tzafrir	Is there anything better than running: git-update-index <every file that has changed> #before commiting?	11:31
HackyKid	hmmm, we could do that	11:31
tzafrir	I use git-1.4.4.4-2 from Debian Etch	11:31
cehteh	tzafrir: update instantly :)	11:31
gitte_	HackyKid: and then... git-svn! Yeah!	11:32
tzafrir	cehteh, thanks for the hint. I guess I'll just dump git	11:32
cehteh	the debian etch version is stone age	11:32
Pieter	cehteh: backports has an up-to-date version	11:33
cehteh	there where tons of user interface and documentation improvements in 1.5.x	11:33
gitte_	tzafrir: if you update, there is "git add -u".	11:33
tzafrir	cehteh, I heard "upgrade" even on the version in Sid :-p	11:33
Pieter	tzafrir: http://packages.debian.org/etch-backports/git-core	11:33
tzafrir	I'd like to keep my system clean of backports	11:33
cehteh	tzafrir: dunno .. the backports version should suffice .. but i build it on my own, git is easy to build	11:33
Ilari	tzafrir: Lenny has 1.5.3.4...	11:33
cehteh	tzafrir: seriously 1.4.x is very different to 1.5.x from a users experience i really cant recommend to use 1.4.x	11:34
	that are not just a few cosmetic changes ...	11:35
tzafrir	anyway, thanks for your replies	11:37
	← tzafrir left	11:37
cehteh	mhm restoring backup, rinse, repeat	11:39
	← fultilt left	11:45
HackyKid	hmm, seems /git/perl should be in @INC so it can find Git.pm	11:45
gitte_	HackyKid: but does not every perl script set this itself?	11:48
	use lib (split(/:/, $ENV{GITPERLLIB} \|\| "/home/gitte/lib/perl5/site_perl/5.8.8"));	11:48
	(The second line in my git-remote script)	11:49
HackyKid	hmm, right	11:49
	ah, i see	11:49
	my perl install is now using different paths from what git thinks it uses	11:49
	→ LotR joined	11:50
halorgium	HackyKid: use PERL_PATH='/usr/bin/env perl' ./configure	11:50
gitte_	HackyKid: just set GITPERLLIB to the new path.	11:51
	(The environment variable, that is.)	11:51
HackyKid	hmm, but my Git.pm didn;t actually change location	11:53
cehteh	bah	11:56
	something in git-cvsimport is broken	11:56
telmich	good morning	12:02
gitte_	Good noon!	12:02
telmich	in the update hook, can I somehow get the name of the projectdir?	12:02
gitte_	You mean the working directory?	12:02
telmich	gitte_: yep	12:02
gitte_	Or the repository directory?	12:02
telmich	hmm..	12:02
	I'm not sure, where git is.	12:03
gitte_	In general, you cannot get the working directory.	12:03
	"which git"?	12:03
telmich	let me explain the problem: I've some projects, that mail their commits to an mailinglist: http://l.schottelius.org/pipermail/eof/2007-November/thread.html	12:03
	the problem is that the standard update hook, which I've on the server, does not say to which project	12:04
	I see the project=$(cat $GIT_DIR/description) variable at the beginning and I'm thinking about whether that makes sense to include in the subject	12:04
gitte_	I think what you want is more like $(basename "$GIT_DIR")	12:05
	(Probably followed by "test .git = "$name" && name = $(basename "$(dirname "$GIT_DIR")")")	12:05
	→ samsarin joined	12:06
telmich	gitte_: sounds pretty like what I want :-)	12:06
gitte_	telmich: denkbrett ;-)	12:06
telmich	gitte_: yep, the notebook I'm hacking on :-)	12:06
gitte_	I would never have guessed ;-)	12:07
telmich	well, I always give names, that match the computer I'm on	12:07
	→ drizzd joined	12:07
telmich	some information about that strange idea: http://tech.schottelius.org/history/README.html	12:07
gitte_	mine is called racer.	12:07
telmich	a fast one, I guess?	12:10
robinr	mine is lathund ("lazy dog")	12:11
	→ exon joined	12:14
gitte_	Nope, racer is a slow dog. Mainly because it only has 512MB.	12:16
	But I'll get an EEE, hopefully this week.	12:16
	telmich LOVES the shell!	12:20
gitte_	So you're happy as a clam? ;-)	12:20
	→ alley_cat joined	12:20
telmich	gitte_: the basename/dirname trick does not work 100% and I love to omit basename/dirname	12:21
	name="${GIT_DIR##*/}"	12:21
	test .git = "$name" && name="${${GIT_DIR%%/.git}##*/}"	12:21
	that's so lovely	12:21
LotR	gitte_: won't the eee be slower?	12:21
gitte_	LotR: probably. But it has more RAM.	12:21
telmich	gitte_: who's clam?	12:22
HackyKid	a ghost!	12:22
gitte_	telmich: it's an expression in English, and a clam is an animal, having a shell.	12:22
	LotR: RAM++ => swap-- => real_speed++	12:23
telmich	gitte_: ah, ok	12:23
	gitte_: hmm, is it possible, that GIT_DIR is not absolute, bute relative?	12:24
	s/bute/but/	12:24
LotR	gitte_: nod (and the swap on eee wouldn't be on harddisk either)	12:24
telmich	http://l.schottelius.org/pipermail/eof/2007-November/000021.html <==	12:24
gitte_	LotR: exactly!	12:24
LotR	gitte_: I'm not sure I'd like a 7" screen though..	12:25
gitte_	I do.	12:25
	telmich: GIT_DIR=. is possible. You'll have to replace it with $(pwd) in that case...	12:25
	← kahmalo left	12:26
telmich	gitte_: I'm doing pwd -P general now	12:26
	telmich has to read the documentation of the hooks	12:27
telmich	NARF. a search for git hooks on google.ch matches my own site as first hit...	12:28
robinr	what's -P?	12:28
gitte_	Resolve symlinks, AFAICT.	12:29
robinr	man pwd doesn't say anything	12:29
	nor info	12:29
telmich	gitte_: ack	12:30
	robinr: read posix, everything else is not thrustworthy	12:30
	robinr: http://www.opengroup.org/onlinepubs/009695399/utilities/pwd.html	12:30
segher__	robinr: pwd is a bash builtin, so try "help pwd"	12:31
	"man pwd" works fine on most of my systems, fwiw	12:31
	← meandtheshell left	12:37
telmich	narf, although [ "$name" ] \|\| name="${${$(cd "$GIT_DIR"; pwd -P)%%/.git}##*/}" seems not supported on posix shells	12:41
	← smartbhai left	12:51
gitte_	I think ${...} needs a variable name.	12:52
telmich	ack	12:52
gitte_	You need something like this: name="$(cd "$GIT_DIR"; pwd -P)" && name="${name%/.git}"; name="${name##*/}"	12:53
	Ugly.	12:53
telmich	actually, I've exactly that :-)	12:54
	yep, that's really ugly. the love to shell has been suspended now.	12:54
	yeeeees	12:55
	http://l.schottelius.org/pipermail/eof/2007-November/000023.html	12:55
	telmich will send a request to the austin group of posix, perhaps they change it	12:56
	← Sho_ left	12:58
	← aroben left	12:59
gitte_	I'm sure you could write something elegant with "expr".	13:00
telmich	gitte_: I guess more to sed	13:00
gitte_	sed is not builtin, expr is.	13:00
	At least in bash.	13:00
telmich	bash is evil [tm]	13:01
	and expr is an utility, says posix: http://www.opengroup.org/onlinepubs/009695399/utilities/expr.html	13:01
	← Tali left	13:01
telmich	that's the reason why I always use posix, as bash is so fscking unportable	13:02
gitte_	Well, POSIX does not get any love from this developer.	13:02
	telmich loves it, because scripts run on dsh, ksh, bash, zsh and even csh in sh mode	13:02
robinr	can't you run bash anywhere?	13:02
gitte_	I tend to dislike meetings, thus people who love meetings, thus POSIX.	13:02
robinr	"a camel is a horse designed by a committee"	13:03
gitte_	And things break by relying on POSIX, and meeting a shell which has a subtle "bug" with regards to POSIX.	13:03
	robinr: lol	13:03
telmich	robinr: I use zsh as user shell and dash as /bin/sh; on my freebsd systems bash is installed as /usr/local/bin/bash for the co-workers, but thus /bin/bash as shebang does not work - and nobody uses /usr/bin/env in their shebang for bash	13:03
gitte_	robinr: but actually, that is too much honour: a camel is a functional system.	13:03
telmich	robinr: ok, that's true	13:03
gitte_	telmich: dash gave me too much grief to like it.	13:04
robinr	I use #!/bin/bash. If anyone wants to change that, fine, but then that person has to change and verify as I cannot	13:05
	→ meandtheshell joined	13:05
robinr	would #!bash work?	13:05
	→ Sho_ joined	13:06
telmich	robinr: well, /bin/bash breaks on almost all non-linux systems	13:06
gitte_	It works quite well in MSys, thankyouverymuch.	13:07
robinr	as I said, I cannot verify that my scripts works outside linux/cygwin	13:07
telmich	robinr: sounds somehow arrogant to me, if people use /bin/bash, although I'm a deep hearted pinguin	13:07
gitte_	zsh is utterly ugly. So I never use it.	13:07
telmich	robinr: well, you can simply install ksh/dsh/mksh or whatever as /bin/sh and focus on posix	13:07
robinr	i fact, I'm, almost certain that they don't work without explicitly verifying it	13:07
telmich	gitte_: zsh is so powerful. that's the reason why it has no sense to put it to /bin/sh	13:07
robinr	hence #!/bin/bash is a "watchout!"	13:07
gitte_	I never test my scripts for non-bash-ability.	13:08
	But then, I follow the advice of a long time shell scripting expert.	13:08
	Junio.	13:08
telmich	robinr: sorry, cannot agree with you. I would recommend using /bin/sh and checking $SHELL if you really insist on bash	13:08
gitte_	So I should be safe.	13:08
	The nicest shell ever is busybox, anyway.	13:09
robinr	I don't insist on bash, I say it's not tested with nothing but bash	13:09
gitte_	_Everything_ is builtin.	13:09
robinr	first one to test and say it's portable, then we change	13:09
	← HackyKid left	13:14
	→ HackyKid joined	13:14
robinr	gitte_: yea, maybe we should standardize on busybox	13:21
	→ vin` joined	13:29
	→ G_SabinoMullane joined	13:32
pennyk_	hi, all	13:36
	It seems that git.kernel.org is down.	13:36
	Anybody can confirm this?	13:36
gitte_	pennyk_: scroll back.	13:36
pennyk_	gitte_: oh, I didn't notice that. Thx	13:37
gitte_	Hmm. I thought it was only a temporary glitch. Maybe somebody should notify hpa.	13:38
	→ rhalff joined	13:39
rhalff	hi I like the combination trac + svn, is there a similar setup with git ?	13:40
segher__	gitte_, telmich: $(...) isn't POSIX either	13:41
gitte_	Ha!	13:41
doener	segher__: it's not? http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html#tag_02_06_03	13:45
	rhalff: trachacks has a gitplugin, no idea how well that works, but OLPC seems to be using it (well, or some other plugin)	13:50
	→ FunkeeMonk joined	13:51
segher__	doener: that's the 2004 version, it's not what is generally refered to as "POSIX"	13:52
	(the 1996 edition is)	13:53
	more importantly, many systems don't yet implement it	13:53
rhalff	doener: thanks	13:54
exon	segher: $() is too damn nice to be without though. Could you name a few systems where it isn't available?	13:54
segher__	exon: solaris, hpux, aix. i bet one doesn't :-)	13:55
exon	segher: "I bet" doesn't really count. If there are reports of it breaking for people, there's a chance it gets changed. Otherwise, why bother?	13:56
	← SuttoL left	13:56
	← Philantrop left	13:56
segher__	exon: otoh, if you require all these newfangled features, you can as well just require base	13:56
	bash	13:57
exon	segher: why? Plenty of other shells support them too. It's much better to just say /bin/sh and hope the user has something sane. When it turns out to break, the people for whom it break can submit a few patches, or at least report the error	14:00
segher__	the pragmatic approach :-)	14:00
	it would make sense to decide what exactly is required, but this works too, sure :-)	14:01
	→ Philantrop joined	14:02
exon	segher: It's fairly decided already; A shell which handles $(command) and ${var##*delim}	14:06
	segher: check the mailing list archives for the past week for a more structured document (CodingGuidelines would be good to search for)	14:08
segher__	ah cool	14:10
gitte_	It was even discussed a little in this channel.	14:11
	← mithro left	14:14
corecode	who said that $() isn't yet [sic] supported by many systems	14:17
segher__	corecode: i did. and i said _some_ systems	14:20
	corecode: at least for GCC we still cannot use it	14:20
	i don't know the exact details	14:20
	→ semi_ joined	14:20
semi_	there's git-remote add, should there perhaps also be git-remote remove?	14:21
corecode	i guess this rumor is kind of worthless without documents proving it as a fact	14:22
	for the unix community and as a greater good	14:22
semi_	when some public repo you are tracking gets removed	14:22
corecode	for instance	14:22
gitte_	semi_: there is. Only it is called rm.	14:23
	semi_: and I think it is only in master as of now, not released.	14:23
semi_	oh, ok	14:23
	I'll upgrade then :)	14:24
	→ kukks joined	14:28
ttf	hi.. if I clone a svn repo from the net and want to do development in it (in head) would I open a new branch for that or would I develop directly in master?	14:28
semi_	if you do everything in master, you lose the benefit of topic branches altogether	14:29
	→ mithraic joined	14:30
	← mithraic_ left	14:30
drizzd	What's the difference between pull . <branch> and merge <branch>?	14:30
ttf	semi_: I have come across the term topic branches before but don't know exactly what it means. care to elaborate or a pointer to some documentation?	14:31
	→ mithraic_ joined	14:32
	← mithraic left	14:32
semi_	ttf: when ever you start working on a new task, you start by branching, no matter how small the task is	14:32
	ttf: and then you can work on different tasks easily even if some of them are totally broken	14:32
	because they have their own branches	14:33
	← mithraic_ left	14:33
exon	ttf: That branch you create is your topic branch, and the end result of it is then a topic which you merge into your main release branch. It's quite useful once you get used to it.	14:33
semi_	but since you are dealing with svn, branches need some specific care	14:34
	→ mithraic joined	14:34
ttf	"main release branch != master" I suppose but yet another branch?	14:34
	→ bdiego joined	14:36
ttf	semi_: ^	14:38
semi_	you choose your release policy	14:39
hliusv561	semi_: :-)	14:39
gitte_	drizzd: http://colabti.de/irclogger/irclogger_log/git?date=2007-11-07,Wed#l1690	14:41
ttf	semi_: s/I suppose but/I suppose or/	14:41
drizzd	gitte_: thanks	14:41
gitte_	drizzd: you're welcome.	14:41
ttf	it sounds smarter to me use a seperate branch for release or is it very common to have master being the release branch?	14:42
	means would it have any advantages (I'm not aware of)?	14:42
	← csc`` left	14:43
	→ csc`` joined	14:44
	← mithraic left	14:44
semi_	well you could have your bleeding edge in master and then merge things to some release branch after they pass enough testing	14:44
	→ mithraic joined	14:44
	← samsarin left	14:45
	← mithraic left	14:46
	→ mithraic joined	14:46
	→ kanru joined	14:48
	← gitte_ left	14:51
	→ mithraic_ joined	14:54
	← mithraic left	14:54
	→ Tali joined	15:03
exon	Has anyone in here managed to get git-svn to import tags properly?	15:03
drizzd	how do I 'undo' a git add, the way I can do it with git gui by clicking in the staged changes box?	15:09
	← mithraic_ left	15:09
	→ mithraic joined	15:10
drizzd	Ok, that was a silly question	15:13
	← FunkeeMonk left	15:14
drizzd	Does git actually track renames, or does it simply detect them in retrospective?	15:15
Tali	drizzd: it only detects them while you are interested in it	15:15
drizzd	ok, so there is no difference between the following command sequences: cp a b; rm a; git rm a; git add b;	15:17
	and git mv a b;	15:17
Tali	exactly	15:17
	→ bobesponja joined	15:17
Tali	that way it can taylor the detection routine on the information you want to know	15:18
	you can e.g. trace back not only files renames but also code moves from one file to another	15:19
drizzd	yeah, makes sense. I was just hesitant about unstaging renames and adding them again, because I wasn't sure if that could undermine git-mv	15:19
	← bdiego left	15:23
	→ bdiego joined	15:23
tsuna	Hey guys. I have a project in Git with 15 years of history and I'm trying to find which commit introduced a given comment. I can blame the file, look at that line, then look at the commit that last changed that line, find out that the line was moved from somewhere else, blame that older revision and start again the process until I find the real commit that introduced the comment. This is inefficient, what would you recommand?	15:25
	recommend, even :)	15:26
	→ mithraic_ joined	15:31
	← mithraic left	15:31
drizzd	tsuna: blame -M doesn't do what you want?	15:33
doener	tsuna: git log -p --reverse -S'comment goes here'	15:34
tsuna	drizzd: how to use it exactly?	15:34
	doener: yes I found the commit with -S indeed	15:35
	→ mithraic joined	15:35
	← mithraic_ left	15:35
doener	:-)	15:38
tsuna	It's odd, the commit is 2d5c3abead3f72c457d886b92b3fbd977d273191 in autoconf (http://git.savannah.gnu.org/gitweb/?p=autoconf.git) but gitweb doesn't seem to be able to show it (it's like if that commit didn't exist in gitweb)	15:38
	argh	15:38
	no	15:38
	it's in automake	15:38
	err	15:38
	looking at the wrong repo doesn't help, obviously	15:38
doener	heh	15:39
drizzd	tsuna: git blame -M filename should also show the original commit which introduced the line. But doener's solution is better.	15:39
doener	drizzd: hm, maybe with -C, too? In case that it moved from file to file?	15:40
	→ grahal joined	15:40
drizzd	yeah, actually -C implies -M I think	15:41
	doener: is git-log -S an undocumented features? I can't find it in man git-log	15:42
doener	drizzd: no, it's just well hidden. git-log(1) tells you that it accepts all options for rev-list and diff-tree. And diff-tree(1) has -S.	15:43
drizzd	oh, I see	15:44
	→ jartz joined	15:44
doener	someone already complained about that not too long ago. It's sub-optimal, but nobody cared enough to come up with a better solution than the cross-ref	15:45
	simply moving or duplicating -S seems bad, and showing all available options in git-log(1) seems scary	15:45
drizzd	yeah well, I don't have a better solution either	15:46
doener	:-) welcome to the club	15:46
	→ madewokherd joined	15:48
	← mithraic left	15:54
tsuna	git.kernel.org[0: 140.211.167.38]: errno=Operation timed out	15:58
	is it me or git.kernel.org is down?	15:59
robinr	from my horizon it is down	16:00
	but it might be that it is only down when we try to access it..	16:00
tsuna	http://cacti.kernel.org/graph.php?action=zoom&local_graph_id=215&rra_id=1&view_type=	16:01
	seems to confirm that it's down	16:01
	git.kernel.org is poseidon.kernel.org	16:01
	that's where git is great, I'll just fetch another copy of Git from my own computer ;o	16:02
Ilari	There's also repo.or.cz. It has mirrors of both git and Linux kernel.	16:05
	← alley_cat left	16:12
	→ alley_cat joined	16:13
	→ GyrosGeier joined	16:15
	→ ebel joined	16:16
	← greendale left	16:17
	→ mithraic joined	16:24
	→ KirinDav joined	16:32
yann	d'oh - there is no mirror of gitk.git on repo.or.cz ...	16:33
	← priidu_ left	16:39
	→ priidu joined	16:39
Ilari	yann: Fairly recent version can be extracted from git.git history...	16:40
yann	Ilari: yes, but I precisely wanted to know if there was anything not in git.git :)	16:41
	← pennyk_ left	16:42
Ilari	yann: I think the answer is negative.	16:43
	→ samsarin joined	16:46
Ilari	I read the repository reference files from www.kernel.org, and newest in them is 7388bcbc5431552718dde5c3259d861d2fa75a12, which can be found from git.git.	16:46
	← alley_cat left	16:46
	→ alley_cat joined	16:53
	→ fhobia joined	16:55
	← kumbayo left	17:11
	← bdiego left	17:14
	→ nessundorma joined	17:17
	← nessundorma left	17:23
	← alley_cat left	17:30
	→ raalkml joined	17:40
	→ mithraic_ joined	17:40
	→ mtg joined	17:41
	← priidu left	17:45
	→ alley_cat joined	17:45
	← mithraic left	17:57
raalkml	git.kernel.org[0: 140.211.167.38]: errno=Connection timed out	18:11
	fatal: unable to connect a socket (Connection timed out)	18:11
tsuna	What's the easiest / fastest way of knowing whether the index contains changes to be committed (that is, whether it's `dirty' or not)	18:11
raalkml	anyone knows what happening?	18:11
tsuna	raalkml: yeah, it seems to be down. You can use the mirror at repo.or.cz	18:11
raalkml	quite some time already...	18:12
	→ orospakr joined	18:12
Tv	tsuna: maybe git diff-index --quiet	18:13
	err, git diff-index --quiet HEAD	18:13
tsuna	--quiet doesn't seem to be supported by diff-index so I think I'll use git diff-index --name-only HEAD and discard its output	18:13
raalkml	git diff --quiet --cached	18:13
Tv	tsuna: you're git is old, then..	18:14
	s/'re/r/	18:14
	yeah --cached looks good too	18:14
tsuna	er, no, it's just that git diff-index -h doesn't mention --quiet :/	18:14
	csc`` → csc`	18:15
tsuna	thanks	18:15
	← mithraic_ left	18:17
	→ mithraic joined	18:18
	← mtg left	18:19
	← alley_cat left	18:20
	← zapnap left	18:20
	→ alley_cat joined	18:20
heipei	git.kernel.org[0: 130.239.17.7]: errno=Connection refused	18:22
	anyone else getting that, or did i bork something here?	18:22
raalkml	the errno is different	18:22
	but yes	18:22
	seem to be restarted now	18:23
	← RobotDeathSquad left	18:24
heipei	hmm, no	18:24
raalkml	worked here. git1.kernel.org	18:25
	git2 still down	18:25
heipei	hmm, i have the newest version anyway, so i can wait ;)	18:26
	→ z3om joined	18:31
	→ bdiego joined	18:32
	→ aroben joined	18:34
	→ gitte_ joined	18:36
	→ cmarcelo joined	18:37
	← alley_cat left	18:44
	→ alley_cat joined	18:47
	← gitte_ left	18:49
	← mithraic left	18:49
	← orospakr left	18:49
	← bobesponja left	18:49
	← Tali left	18:49
	← exon left	18:49
	← Eludias left	18:49
	← vmiklos left	18:49
	← metafollic left	18:49
	← FauxFaux left	18:49
	← ribas left	18:49
	← tarrybone left	18:49
	← masterdriverz left	18:49
	← swetland left	18:49
	→ bobesponja joined	18:50
	← zeom left	18:51
	z3om → zeom	18:51
	→ gitte_ joined	18:52
	→ mithraic joined	18:53
	→ orospakr joined	18:53
	→ Tali joined	18:53
	→ exon joined	18:53
	→ Eludias joined	18:53
	→ vmiklos joined	18:53
	→ FauxFaux joined	18:53
	→ swetland joined	18:53
	→ ribas joined	18:53
	→ tarrybone joined	18:53
	→ masterdriverz joined	18:53
	→ ead joined	18:53
	→ kupsch joined	18:53
	← FauxFaux left	18:53
	→ metafollic joined	18:53
	← ribas left	18:53
	→ ribas_ joined	18:53
	→ cupsh joined	18:53
	← swetland left	18:54
	← kupsch left	18:54
	← Tali left	18:55
	→ Tali joined	18:55
	← tcoppi left	18:55
	→ tcoppi joined	18:56
	← alley_cat left	18:56
	→ vovo joined	18:56
vovo	git clone git://git.kernel.org/pub/scm/git/git.git - what will this do?	18:56
gitte_	ATM nothing... git.kernel.org seems not to respond	18:56
raalkml	a copy of git.git with local name "git"	18:56
	→ asdx joined	18:57
raalkml	git1's ok	18:57
vovo	whre is the local name mentioned?	18:57
	git:// is which port number?	18:57
raalkml	9418	18:57
gitte_	See man page...	18:57
vovo	I have tried this command and it has got stuck here: Initialized empty Git repository in /home/humpty/lab/downloads/git/.git/	18:57
	it is not going beyond that	18:57
	→ FauxFaux joined	18:57
raalkml	well... gitte already told you	18:58
gitte_	vovo: have you even bothered to read my first reply?	18:58
	→ swetland joined	18:58
raalkml	try git1.kernel.org	18:58
vovo	oh! I missed the first one. sorry. trying.	18:58
	how can I browse the git repo? is there any tool?	18:59
raalkml	many tools	18:59
	see git.or.cz or wikipedia article. try gitk and git-gui	18:59
vovo	is git used for anything else other than linux kernel?	19:02
raalkml	yes	19:02
vovo	like?	19:02
raalkml	X?	19:02
	Wine?	19:02
	google	19:02
	(well, meant: "google for it", but the Google too)	19:03
vovo	why is 'git' called, stupid control tracker?	19:03
raalkml	why do you have to ask>	19:03
	?	19:03
	→ DrNick joined	19:03
vovo	curious	19:04
raalkml	find out than	19:04
doener	because is tries not to be smart, just like the guy in front of my window, who enjoys to continously use his signal-horn...	19:04
gitte_	vovo: git.or.cz/gitwiki has a nice page of projects using it, which is not really hard to find.	19:05
	telmich: I'd really be curious to find out about the other names of your computers. Particularly "ei".	19:09
	→ Ryback_ joined	19:10
yann	Ilari: ok, thanks for the info	19:11
	← raalkml left	19:19
	→ alley_cat joined	19:21
	→ Yuuhi joined	19:27
	→ myrizio joined	19:27
	→ kumbayo joined	19:27
	← asdx left	19:29
	← alley_cat left	19:31
	← kukks left	19:31
	← vmiklos left	19:32
	→ kukks joined	19:32
	→ vmiklos joined	19:36
	→ alley_cat joined	19:41
	← vovo left	19:43
	→ asdx joined	19:50
	← kumbayo left	20:03
	→ arwed joined	20:11
	← alley_cat left	20:11
	→ apollo13 joined	20:14
arekm	is there a way to find out when some branch was created via gitweb?	20:14
apollo13	hi, how can I discard all changes made to my sources?	20:14
	eg I have some modified files which I wish to revert to head	20:15
context	arekm: you mean when it was branched	20:15
arekm	context: yes	20:16
context	arekm: you can easily find out when it was last merged, finding when it was actually originally branched i dont think so	20:16
exon	apollo13: git checkout -- file1 file2 file3	20:16
context	actually	20:17
apollo13	exon: still showing the as modified (but not tracked yet)	20:17
context	arekm: try git merge-base --all your-branch <branch-you-branched-from>	20:17
	arekm: and might be the last one	20:17
	ive never used --all	20:17
	→ aroben_ joined	20:18
exon	apollo13: If you want to discard ALL changes to the tracked files in your work-tree, do "git reset --hard"	20:18
	← aroben left	20:18
	aroben_ → aroben	20:18
	← Ademan left	20:18
apollo13	exon: strange I tried it before but they are still displayed as modified (could this have something to do with git svn?)	20:19
	→ Ademan joined	20:19
context	apollo13: no	20:20
exon	apollo13: what context said	20:20
	apollo13: Do you have permission to overwrite the files?	20:20
apollo13	nope, I even don't want to check them in, it's just a checkout of django...	20:21
	where I want to stay up2date	20:21
context	git reset --hard master if you dont care about any of the changes	20:21
apollo13	context, believe me it doesn't (but don't ask me why :()	20:21
gitte_	Isn't that in the FAQs by now?	20:21
exon	apollo13: ok. "git reset --hard && git clean -d" should probably do what you want then. Be warned that it'll get rid of all local modifications though	20:21
context	apollo13: what says they are modified	20:22
apollo13	git staus	20:22
	status*	20:22
	exon: didn't help either strange	20:22
context	rm -rf * ; git co master	20:22
apollo13	marked as " Changed but not updated:"	20:22
context	checkout* srry	20:22
apollo13	context marks al files as deleted and I am already on master^^	20:23
context	git checkout -f master	20:23
apollo13	k, the changes are back (although I never modified these files)	20:24
exon	apollo13: What system are you using?	20:25
context	git diff then	20:25
	yeah no kidding	20:25
	are they locale files ?	20:25
exon	apollo13: Or rather, is your filesystem case sensitive?	20:25
context	with unicode characters in the filename ?	20:25
apollo13	context: seems to be an ordinary commit from one of the django developers	20:25
	nope	20:25
	still no idea why they are marked modified	20:25
context	apollo13: does diff actually show changes ?	20:25
apollo13	exon: debian	20:25
	context: yeah	20:26
	this is what I am curious about	20:26
context	then you didnt rm -rf * or you dont have permission to .	20:26
apollo13	I never made them	20:26
	I have	20:26
	I checked...	20:26
context	files dont get modified between git co -f and git status :/	20:26
exon	nopes	20:26
	Unless you have some weird checkout hook	20:26
apollo13	git svn ...	20:26
context	no.	20:27
	i use git-svn every day	20:27
apollo13	I try to get my copy from the server	20:27
	context: me too, but I can't help myself with this one...	20:27
context	do this for fun and paste somewhere	20:27
apollo13	paste what?	20:27
context	git-status ; rm -rf * ; ls -la ; git checkout -f master ; git status	20:28
exon	The output (paste at http://git.pastebin.com)	20:28
context	just for fun	20:28
	:/	20:28
apollo13	yeah you will get your fun....	20:28
context	err	20:28
	exon: how do you wipe the index, though from what he said its not that	20:29
	← KirinDav left	20:29
context	and git diff doesnt show indexed	20:29
apollo13	context, exon: http://git.pastebin.com/d6cd9653f just for you :=	20:29
	and to be honest I have no idea what could have caused this...	20:31
context	remove one of the files and see what it says :/	20:32
apollo13	other idea, I used git for django to be able to switch easily between to svn branches	20:32
	could this have something to do with it?	20:32
context	im confused as much as you are right now	20:32
exon	I should think crlf <-> lf conversion or some such. What does 'find -type f -name ".gitattributes"' tell you?	20:32
apollo13	context says when?	20:32
context	how big is the django history	20:32
apollo13	context: 6000 checkuns	20:32
	← arwed left	20:32
apollo13	checkins	20:32
context	oh :x	20:33
	is that a git repo i can replicate	20:33
apollo13	but I could upload a working archive (19mb)	20:33
	→ KirinDav joined	20:33
apollo13	sure it's open http://code.djangoproject.com/svn	20:33
	should I rm or git rm one of the files?	20:33
context	i ment a git repo	20:34
	git-svn'ing 6000 revisions may take a bit ..	20:34
	← KirinDav left	20:34
apollo13	exon: autocrlf = true in my .gitconfig is this bad?	20:34
context	ahh	20:34
exon	apollo13: Try without it	20:34
thiago_home	I've done 12000, so it's doable :-)	20:34
context	thiago_home: i know its doable, just time consuming	20:35
apollo13	context I could provide you with a working checkout of mine packed repacked .... 19 mb .tar.bzip2 archive	20:35
exon	remove it and then re-do the "rm -rf *; git checkout -f master" thing	20:35
apollo13	exon: what does it default to?	20:35
	(so, was it me how set this or is true the default)	20:35
exon	core.autocrlf	20:36
	If true, makes git convert CRLF at the end of lines in text	20:36
	files to LF when reading from the filesystem, and convert in	20:36
	reverse when writing to the filesystem. The variable can be set	20:36
	to input, in which case the conversion happens only while	20:36
	reading from the filesystem but files are written out with LF at	20:36
	the end of lines. Currently, which paths to consider "text"	20:36
	(i.e. be subjected to the autocrlf mechanism) is decided purely	20:36
	based on the contents.	20:36
context	apollo13: comment out autocrlf and git checkout -f	20:36
thiago_home	context: yep	20:36
	← Tali left	20:36
apollo13	k working now	20:36
context	that should be something they have set in autoprops for svn anyway	20:36
apollo13	Still I have no idea who set this setting gg	20:37
context	that was a headache at work when we switched to svn	20:37
apollo13	so is autocrlf defaulting to true or false?	20:37
context	im pretty sure false .	20:37
exon	righto, something's broken with autocrlf then, I guess. I think that setting was primarily intended for project-specific configs though	20:37
context	no scm in they're right mind modifies files my default ..	20:37
apollo13	so wth did I put ut in^^	20:37
exon	apollo13: I pasted the setting above	20:38
	Currently, which paths to consider "text"	20:38
	<exon> (i.e. be subjected to the autocrlf mechanism) is decided purely	20:38
	<exon> based on the contents.	20:38
apollo13	exon: yeah I read it before in the web, but it didn't tell me whether it's defaulting true or false...	20:38
	but it works on my server so it must be false	20:39
exon	apollo13: It's defaulting to a content-based file-by-file decision	20:39
apollo13	ah :)	20:39
exon	apollo13: neither true nor false, that is	20:39
apollo13	okay so as long it's working leave it as default otherwise set it to false...	20:39
	and context, yeah it took long to download the repo :)	20:40
exon	ye, that's my philosophy too :)	20:40
apollo13	I only have one setting in my gitconfig (apart from email, name) which is filemode=false (I am pretty sure I didn't set it) keep it or kick it out? according to the docs: "If false, the executable bit differences between the index and the working copy are ignored; useful on broken filesystems like FAT. See git-update-index(1). True by default." So I think kikcing it out is a good option...	20:42
context	i have nothing in core, but a lot of other stuff	20:45
apollo13	good :) thx for your help	20:45
exon	np.	20:49
apollo13	you got enough to laugh didn't you ;)	20:50
ttf	me and a friend of mine are tracking a public svn repo in two separate local git repos. let's say I am developing in a branch called foo and he is developing something different in his branch bar. we want to share those branches with each other but can't pull from each other directly. we'll setup a shared repo. now - how do we use this repo to share our branches (or at least our local commits we made in them)?	20:50
	gitte_ wonders if there is a longest-line contest going on.	20:51
	→ alley_cat joined	20:51
	← orospakr left	20:52
gitte_	ttf: you just push your branches into that shared repo (which should be bare)	20:52
exon	ttf: You do "git push shared-repo foo" and he does "git push shared-repo bar"	20:52
apollo13	gitte_ am I going to win?	20:52
gitte_	Of course, don't forget to pull...	20:52
exon	gitte_: There's no real need to make it bare.	20:52
gitte_	apollo13: your _line_ might be longer than mine...	20:53
	exon: oh yes, there is. New git users regularly get confused by pushing into non-bare repos.	20:53
	exon: it's just asking for trouble.	20:53
apollo13	exon: gitte_ is right, I got confused to, really confused :)	20:54
	too*	20:54
ttf	and then I just pull branch bar and he pulls branch foo respectively? do we have to do any mantainance on the server than or locally (rebase?) to not mess up the history of our branches?	20:54
	or will be local foo be identically with remote foo?	20:55
	← alley_cat left	20:55
exon	gitte_: ah right. No technical reason then ;-)	20:55
	ttf: That depends whether or not you commit on local foo or not	20:56
	ttf: After having pushed it, that is	20:56
	ttf: If you just push into the repo and that other bloke fetches it, he'll see exactly what was in your branch at the time when you pushed.	20:57
ttf	as foo would be my local devel branch I would commit, push, commit, push regularly so that my friend can be up to date - or is there a better way?	20:57
exon	ttf: No, that seems to cover it, although you'd probably be better off using topic-branches instead	20:58
	ttf getting confused..	20:58
exon	ttf: If nothing else, it'll be easier for you to check in on a particular feature the other is working on	20:58
	ttf thought foo was a topic branch	20:59
ttf	isn't a topic branch just a branch I use for local development?	20:59
exon	ttf: Yes and no. A topic-branch is a fork off of the main development line that holds a single feature, usually implemented as a series of commits	21:00
	→ Astro joined	21:00
Astro	hi	21:00
	I've got some files containing chinese text	21:00
	(ought to be utf-8)	21:00
apollo13	exon: what happens if I do git pull in the topic branch?	21:00
Astro	but git-diff tells me "Binary files ... ... differ"	21:01
	how do I view the diff?	21:01
	→ warren joined	21:01
exon	apollo13: You fetch and merge whatever your config states that that branch should be merged with.	21:01
ttf	exon: it's meant to be that the other person can see the work in progress - so I could stick to pushing foo I guess.	21:01
Astro	doh, git-diff --text	21:01
apollo13	so if my config states nothing (eg doing git checkout -b topic [from my local master])	21:01
warren	Is there a convenient way to find the last git changeset that touched a given line?	21:01
	→ RetroJ joined	21:02
exon	warren: git blame file	21:02
	warren: or "git log -S<line>"	21:02
warren	<line> is line number?	21:03
ttf	exon: isn't it smarter to only push my actual changes to the shared repo instead of the whole branch? if so how would I achieve that?	21:03
exon	warren: no, it's a string on the line	21:03
	warren: It will locate commits that include the text you specify as argument to -S.	21:03
	ttf: You would push your branch	21:04
ttf	okay	21:04
RetroJ	hi. I would like `git push' to cause the commits in my local repository to be put into the upstream repository. is this what is meant by "CVS workflow"?	21:04
exon	apollo13: You'd better do "git checkout -b topic origin/master" if you want it to merge the 'master' branch from your remote named 'origin'	21:05
	ttf: Or send your changes as patches to that other bloke, but that's more hassle	21:05
thiago_home	RetroJ: "cvs commit" is split into two commands in git: commit and push.	21:05
	RetroJ: well, one of the possible workflows.	21:05
apollo13	exon: so then git pull/push would operate on the remote master?	21:05
warren	exon, is there an easy way to step through changesets and watch a particular part of a file?	21:06
exon	apollo13: "git push" will only push the branches that the two repos have in common	21:06
RetroJ	I am a bit confused by the terminology "remote" since it seems to refer to "local cached" copies	21:06
warren	exon, (it isn't isn't to see exactly where a deletion happened?)	21:06
thiago_home	RetroJ: remote is whatever is not your local repository.	21:06
exon	warren: What is it you want to accomplish?	21:06
ttf	exon: you said "That depends whether or not you commit on local foo or not". What do I have to take care of (or do) when I commit regularly on the branch I push and the other person regularly pulls?	21:07
apollo13	exon: to clarify two repos? remote and my local branch?	21:07
exon	RetroJ: You probably mean the tracking branches. they track branches in remote repos	21:07
RetroJ	exon: yes, I think so	21:07
warren	exon, two branches of a software have fixed bugs in two different directions, I'm pulling fixes from both branches. With "git mergetool" with meld, it isn't entirely clear why an insertion or deletion happened, so trying to find exactly where a change happened and why.	21:08
exon	ttf: Nothing, really. What I meant was just that the branch in the upstream repo and the one in your local one will only be identical after you've pushed it but before you've committed to it	21:08
	← nud left	21:09
exon	warren: Use "git log -p -Sline_that_changed -- paths/that/changed"	21:09
ttf	okay - thanks again	21:09
apollo13	exon: how would you work if you were working on two (or more) patches. leave the master clean make a branch for both of them and finally merge them back into master and push them then?	21:09
	→ alley_cat joined	21:09
exon	apollo13: Something like that, yes.	21:09
	apollo13: I'd commit typo-fixes and clearly trivial stuff directly to master	21:10
warren	exon, what does the -- mean?	21:10
apollo13	so it is a good idea of keeping my local master clean (besides from trivial stuff), as it allows easy merging	21:10
	to keep*	21:10
exon	apollo13: When there's a fair chance the feature/fix/whatever will require more serious surgery I start a branch and hack up the patch as a commit-series	21:10
	apollo13: yes	21:10
apollo13	okay :)	21:10
warren	exon, so I guess no tools like meld allow you to see during the mergetool where a change happened?	21:10
exon	warren: They're just markers to help git disambiguate between branches and files	21:11
warren	ah	21:11
apollo13	exon: and is it a good idea to use git pull origin/master to get new revisions?	21:11
exon	warren: git will show you all the commits that changed the line you specify in -S	21:11
	apollo13: I usually do "git checkout master; git fetch; git diff origin/master; if(changes_look_ok) git merge origin/master"	21:12
gitte_	exon: actually, -S is slightly different from that: it shows you all the commits where the given string was either added or removed. At least AFAIK.	21:13
exon	apollo13: Actually, I more often do "git rebase origin/master" than "git merge", but that's just a minor nitpick	21:13
	gitte_: Yes. I wasn't very clear there.	21:13
gitte_	Warning: rebase rewrites history.	21:13
ttf	it possible to setup a shared (but not public) bare repository, right?	21:14
apollo13	okay thx, but how are you porting the changes into the branches, by merging them with master or let them pull themselves?	21:14
exon	ttf: yes, ofcourse	21:14
	apollo13: When a topic is deemed complete, it's merged to master.	21:15
apollo13	ups, I meant how do you get new revisions from the origin into a branch, do you merge them in from master? (so the other way round...)	21:16
exon	apollo13: I go through a couple of topic-branches per day, usually. Using topic-branches is just plain awesome for testing, especially when multiple people are working on different features.	21:16
	← RetroJ left	21:16
warren	exon, gitte_: ok thanks, exactly what I needed	21:16
exon	warren: happy hunting	21:17
	apollo13: I merge them in, or rebase my local changes onto the upstream branch	21:17
warren	It would be cool if one day, one of the mergetools allows you to easily see what changeset caused a particular change you are seeing.	21:18
	→ TOMA4ATO_MONSTAH joined	21:18
warren	You're trying to merge stuff totally without context.	21:18
TOMA4ATO_MONSTAH	WRAARRRR!!! I'm the Tomato Monstahhhhh! WRAARRRR!!!	21:18
warren	wtf	21:18
TOMA4ATO_MONSTAH	WRAARRRR!!! I has the Cookies Tooo! WRAARRRR!!!	21:18
	WRAARRRR!!! I'm the Tomato Monstahhhhh! WRAARRRR!!!	21:18
thiago_home	that moron has been spamming a few channels	21:18
	← TOMA4ATO_MONSTAH left	21:18
LotR	hehe	21:19
exon	apollo13: When adding stuff from upstream I usually rebase. Primarily because such merges are usually "useless" (ie, they just clutter up history, whereas merging a topic-branch means a feature has matured enough to graduate to near release-status)	21:19
warren	Just imagine that kind of behavior in real life.	21:19
apollo13	okay thx again, learning more of git everyday :)	21:19
exon	apollo13: but also because rebasing local stuff onto upstream means I'd have to work really hard to destroy already published history	21:20
	apollo13: yw.	21:20
warren	exon, I'm rather new to this kind of tool, do you recommend any document that describes the methodology of branching and merging?	21:20
	← Sho_ left	21:21
exon	warren: Just run through the tutorial once. It'll give you quite a jumpstart	21:21
	→ Sho_ joined	21:21
warren	exon, a little more background. mkinitrd has an "upstream" git repository. There is an unofficial branch of mkinitrd that does wacky things, last synced with upstream 2 months ago. I want to continue bug fixes against wacky-branch, but first I want to sync with upstream fixes. I'm uncertain whether I want upstream pull wacky-branch or the other way around.	21:22
exon	warren: Getting an experimental repository to just test things in would also be a good idea.	21:22
gitte_	warren: re ui tool: have you tried "git gui blame <file>" lately?	21:22
exon	warren: do "git merge-base master whacky". If that works ok, do "git checkout -b try_merge <output>", where <output> is the output from "git merge base"	21:24
	warren: do "git merge-base master whacky". If that works ok, do "git checkout -b try_merge <output>", where <output> is the output from "git merge-base"	21:24
	← grahal left	21:25
exon	warren: read only the second of those posts ;-)	21:25
	← alley_cat left	21:26
warren	git-merge-base [--all] <commit> <commit>	21:27
	what are those <commit> parameters?	21:27
exon	warren: The commits you want to see the merge-base of	21:27
	warren: In your case, it'd be the master branch and the whacky branch	21:27
warren	Maybe I'm getting my terminology mixed up	21:28
	← asdx left	21:28
warren	git clone http://somewhere/something	21:28
	that puts a copy of a "branch" locally?	21:28
exon	warren: Where's this git repo, btw? It might help if I can see what's going on	21:28
thiago_home	warren: the two commits whose base you want to find.	21:28
exon	warren: That puts a copy of a repository locally	21:28
	warren: With all its history and everything	21:29
warren	Using bzr, hg and git for different things, the names get all mixed up =)	21:29
	warren finding url's....	21:29
exon	warren: I can imagine ;-)	21:29
	→ alley_cat joined	21:30
warren	official upstream git://git.fedoraproject.org/hosted/mkinitrd	21:30
	I guess that would be "master"?	21:30
	→ asdx joined	21:30
	→ mithro joined	21:30
thiago_home	no	21:30
	that would be git://git.fedoraproject.org/hosted/mkinitrd	21:30
	that's a repository, not a branch	21:31
exon	warren: What thiago said. It has a 'master' branch, and the 'master' happens to be the default branch for that repo, but it's not necessarily the only 'master'	21:31
warren	oh	21:31
exon	warren: Was there a url for the whacky work as well?	21:31
warren	http://fedorapeople.org/~katzj/git/mkinitrd.git/	21:31
	OK, so two separate repositories.	21:32
	They have common ancestry	21:32
	→ Tali joined	21:32
warren	I want to continue doing fixes to http://fedorapeople.org/~katzj/git/mkinitrd.git/, but first I want to pull the bug fixes from git://git.fedoraproject.org/hosted/mkinitrd	21:33
thiago_home	merging in git is a symmetric operation (except for the subtree merge strategy)	21:34
	so, it doesn't matter which way you do it	21:34
	are you merging ALL commits?	21:34
	→ arwed joined	21:34
warren	If I want to continue along http://fedorapeople.org/~katzj/git/mkinitrd.git/ without touching git://git.fedoraproject.org/hosted/mkinitrd might it be more logical to start from http://fedorapeople.org/~katzj/git/mkinitrd.git/ and pull git://git.fedoraproject.org/hosted/mkinitrd?	21:35
	Is it normal for merges to maintain changeset history of stuff you merged from elsewhere?	21:35
	← apollo13 left	21:36
exon	warren: Righto, here's what I just did	21:36
warren	The goal is eventually (maybe months from now) to merge all changes back from wacky into upstream, with all history.	21:36
exon	git clone git://git.fedoraproject.org/hosted/mkinitrd	21:36
	git remote add whacky http://fedorapeople.org/~katzj/git/mkinitrd.git/	21:36
	git fetch whacky	21:37
	git merge-base whacky/master origin/master	21:37
	← Sho_ left	21:37
exon	which gave the output 74c5e19aa93ce29f0ff0f21df0a2cb64205bb7ce	21:37
ttf	setting up a shared repo like this doesn't seem to work: git-svn clone --bare lxoffice https://lx-office.linet-services.de/svn/	21:37
	→ Sho_ joined	21:37
ttf	do I first have to git-svn clone the repo completely and from that point make a bare repo?	21:38
exon	warren: So... 74c5e19aa93ce29f0ff0f21df0a2cb64205bb7ce is the commit where the whacky-fork parted ways with the upstream	21:38
	← mutex left	21:38
exon	warren: With me so far?	21:38
warren	exon, yes	21:38
	hah...	21:40
	[warren@newcaprica git-help]$ git remote add whacky http://fedorapeople.org/~katzj/git/mkinitrd.git/	21:40
	Can't locate Error.pm in @INC	21:40
ttf	hmm - I probably better use git-init and git-fetch seperately..	21:40
exon	warren: ouch. Get the latest git from git.kernel.org. This'll be awfully awkward if I use some features you don't have	21:41
warren	git-1.5.3.4-1.fc8	21:41
exon	warren: That should work. I've got git version 1.5.3.5.1527.g6161, but afaik there aren't that vast changes between them	21:42
warren	ahh	21:43
exon	warren: Good news for you though. "git diff --stat whacky/master origin/master \| tail -n 1" says: 9 files changed, 390 insertions(+), 411 deletions(-)	21:43
warren	I had to cd into the repository for that command to work.	21:43
exon	warren: ah, yes. :)	21:43
warren	exon, one more question	21:45
	exon, http://git.fedoraproject.org/?p=hosted/mkinitrd;a=summary	21:45
exon	warren: Righto, do this now: "git checkout -b try_merge origin/master; git merge whacky/master"	21:45
warren	exon, under http://git.fedoraproject.org/?p=hosted/mkinitrd;a=summary "heads" there is a bash-branch	21:46
	exon, when I cloned this repo, did I pull bash-branch to local as well?	21:46
exon	warren: yes. You can see it in your local repo as well, if you do "git branch -r"	21:46
warren	ah	21:47
exon	warren: Yes, but only master was created as a local "here you can commit" branch. The others are sort of hidden so you don't accidentally commit to a branch you've set up to track upstream	21:47
ttf	hmm.. running: git --bare init --shared; git-svn --bare fetch https://lx-office.linet-services.de/svn/ doesn't work either - do I have to create the bare repo from the existing git-svn cloned repo?	21:48
	means the second command seems to be unsupported	21:48
warren	exon, what is the actual name for "sort of hidden" modE?	21:48
	← Sho_ left	21:48
	→ Sho_ joined	21:48
exon	warren: "remote tracking branches". They're not really hidden, but git won't list them by default	21:49
	warren: sorry, I'm not very good at explaining (I digress too much)	21:49
warren	I'm not exactly sure what is going on, but I'll keep figuring...	21:50
	if I have an intelligent question I'll come back here.	21:50
	exon, thanks for all the help.	21:50
exon	warren: Have you done the "git fetch" thing after adding the whacky repository as remote?	21:50
	→ nud joined	21:50
warren	exon, yes	21:51
	← vin` left	21:51
exon	warren: Righto, do this now: "git checkout -b try_merge origin/master; git merge whacky/master"	21:51
warren	Auto-merged nash/nash.c	21:52
	Automatic merge failed; fix conflicts and then commit the result.	21:52
exon	warren: You're about to run into the git index soon. I'll explain how you can use it during merge conflicts and then you'll be just fine fiddling with git.	21:52
warren	ah	21:52
exon	warren: Ah, there we have it. Now, nash/nash.c has been merged in the index	21:52
	← Sho_ left	21:53
exon	warren: So you can see two types of diffs now. What's already resolved/merged cleanly (git diff --cached), or what's left for you to figure out how to handle in the working tree (git diff)	21:53
	→ Sho_ joined	21:53
warren	exon, wait a second, what does index mean?	21:54
	← Yuuhi left	21:54
exon	warren: As you work your way through the merge conflicts you should do "git add" on each file as it's fully resolved to add it to the final merge commit and get rid of it in the regular "git diff" output	21:55
warren	oh	21:55
exon	warren: The index is a staging area where git prepares the next commit	21:55
	← alley_cat left	21:55
Ilari	warren: Index stores the contents of files about to be committed, and it also stores the conflicting files and their ancestor during merge...	21:56
exon	warren: It's particularly powerful when handling a merge with conflicts, since everything that merges cleanly will end up in the index, while that which doesn't is left outside it	21:56
warren	exon, I've been using meld to aid in visualizing the conflicts, is there a better way to both visualize and see changesets that caused the conflicts?	21:56
exon	Ilari: No, merge conflicts aren't added to the index.	21:56
thiago_home	warren: use git-merge-tool	21:56
warren	thiago, git mergetool is what called meld	21:56
	meld is only one option of many	21:56
Ilari	exon: The files to be merged are stored there, as well as the ancestor.	21:56
	thiago_home used only kdiff3	21:57
thiago_home	that works fine for me	21:57
exon	warren: git-mergetool just makes sure to send the right options to the diff viewer you choose	21:58
warren	Does any git-merge-tool allow you to visually see the changeset and comments taking part of a conflict?	21:58
exon	Ilari: Conflicted merges aren't added to the index. Only that which merges cleanly.	21:58
	← sgrimm left	21:59
ttf	sorry for repeating - does anybody know an answer to my "bare repository" question above?	21:59
	→ sgrimm joined	21:59
	← Sho_ left	21:59
exon	warren: There are conflict markers in the files now though. With "git diff --color=auto" you'll be able to resolve most of it very quickly, me thinks.	21:59
	→ eMBee_ joined	22:00
warren	[warren@newcaprica mkinitrd]$ git diff --color=auto	22:00
	error: invalid option: --color=auto	22:00
	← jlh left	22:00
exon	warren: err, skip "=auto" :)	22:00
	warren: my bad. I have it in my config file.	22:01
warren	exon, this still doesn't show me the changesets that caused the conflicting changes.	22:01
	if a tool doesn't exist to do this yet, then OK.	22:01
	I can survive.	22:01
exon	warren: Ah, now I see what you mean.	22:01
	That's where "git log -S" comes in though	22:02
warren	exon, ok, if I merge the conflicts, then what?	22:02
	git commit (but where?)	22:02
exon	warren: Resolve the conflicts in each file as you go along. Then "git add file" when you're done with that file. When "git diff" has no output, but "git diff --cached" has plenty, you just type "git commit" and edit the commit message in your editor as usual	22:03
warren	hm	22:04
	← mithro left	22:04
Ilari	exon: There are normal 'stage 0' index entries, and there are also stage 1, 2 and 3 entries, which are used during merge. AFAIK, index can't be committed before all stage 1, 2 and 3 entries are gone.	22:04
warren	exon, if I want to maintain whacky in bash-branch in parallel with master	22:04
	exon, how do I commit to bash-branch explicitly?	22:05
	→ Sho_ joined	22:05
exon	warren: "git checkout -b bash-branch origin/bash-branch; (work, work, work); git commit"	22:06
warren	[warren@newcaprica mkinitrd]$ (work, work, work)	22:07
	bash: work,: command not found	22:07
	sorry, bad joke =)	22:07
exon	:P	22:07
warren	Why is git pull -t not default?	22:08
	oh, I guess the default is more intelligent thant his.	22:09
	this	22:09
exon	warren: yup	22:09
clee	heh. heya, warren	22:09
warren	OK, I should be able to figure this out from here.	22:09
	clee, hey clee	22:09
	← Eludias left	22:11
exon	warren: gl with your merges	22:11
warren	gotta reboot, brb	22:12
	→ aroben_ joined	22:12
telmich	good evening	22:16
	gitte_: 'ei' was my g3-ibook	22:16
gitte_	Good evening, telmich	22:16
	← arwed left	22:16
gitte_	You had one, too!	22:17
telmich	gitte_: yeah, pretty nice laptop, though replacing parts took me hours	22:17
gitte_	Oh yes.	22:18
	Well, I did not replace anything, as it was something on the motherboard which went dead.	22:18
	But my sister's husband was really happy to have so many parts.	22:19
telmich	segher__: btw, $() _is_ posix: http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html	22:19
	gitte_: hehe	22:19
doener	telmich: you're 8 hours late ;-)	22:19
	14:52:11 <segher__> doener: that's the 2004 version, it's not what is generally refered to as "POSIX"	22:19
telmich	segher__: you are the second one telling me that it's not posix that weekend	22:19
	doener: tzzz	22:20
	doener: interesting, like "posix was and may never change"	22:20
	← aroben left	22:20
gitte_	segher__: AFAIAC it could be in X.org, too, or even in the Magna Carta. As long as the common shells grok it, it's okay, if there is just one which does not, well, we need a solution.	22:20
	→ aroben joined	22:21
doener	telmich: at least he said that the 1996 version is currently POSIX, originally, it was 1988 AFAIK	22:22
telmich	doener: 1988 is correct, but things changes, perhaps segher__ will accept that, too (=> 1988 hint is on the faq, btw: Although originated to refer to the original IEEE Std 1003.1-1988, the name POSIX more correctly refers to a family of related standards: IEEE Std 1003.n )	22:23
	(from http://www.opengroup.org/austin/papers/posix_faq.html)	22:23
	← exon left	22:24
doener	telmich: yeah, read that earlier today. I'm not arguing against you btw, I posted that same url 9 hours ago	22:24
telmich	doener: sorry, did not want to make the impression that I think that you argue against me, just spread some information :-)	22:25
doener	:-)	22:26
	← Oejet left	22:26
	→ mithro joined	22:28
	→ janm joined	22:29
	← aroben_ left	22:30
	Ilari screwed up, causing ~15MB download to start...	22:33
Ilari	And even worse, that includes that "branch from hell" which can't be updated due to that git-rev-list bug...	22:35
	← ramza3 left	22:36
	→ alley_cat joined	22:42
gitte_	Ilari: which bug?	22:47
Ilari	gitte_: That git-rev-list pathlimiting/subrepo bug I told earlier.	22:49
gitte_	I wasn't there.	22:49
	→ jasam joined	22:51
Ilari	gitte_: The test repo I made for it goes like this: A is parent of B and C is parent of D. A has C as subrepo, B has D. Now, 'git rev-list B D ^A" crashes.	22:51
	Revision limiting, not path limiting...	22:52
	gitte_ goes and tries that.	22:52
Ilari	gitte_: Ooops. You need --objects there also.	22:54
	One can imagine what bug like that does to fetching...	22:55
	→ priidu_ joined	23:00
	← ferdy left	23:03
gitte_	Ilari: I cannot reproduce: http://git.pastebin.com/m6a2988ea	23:05
	Copy this into your git checkout's t/ directory, and start it with "sh t.sh".	23:06
	← aroben left	23:06
robinr	hmm, not missing a letter there? :/	23:06
gitte_	A letter?	23:07
	A cover letter?	23:07
robinr	sh t.sh	23:07
	never mind	23:07
gitte_	Well, this file is named "t.sh" here, obviously.	23:07
	Ilari: my script is buggy, it seems. It does not initialise the submodule correctly.	23:10
	Hah!	23:11
	Do it correctly, and you can reproduce it.	23:11
Ilari	gitte_: Got updated script?	23:11
gitte_	Ilari: yes: http://git.pastebin.com/m759e0435	23:12
	(Note that I already put in a gdb --args, so you _have_ to call it with -i -v)	23:13
	Error is in mark_tree_uninteresting(), revision.c, line 73.	23:14
Ilari	...Which assumes that if it isn't a tree, it is a blob?	23:15
gitte_	Yep.	23:15
Odd_Bloke	If I have a branch in a remote repository, what's the appropriate way to get a copy of it locally?	23:16
gitte_	I got a fix.	23:16
	Now comes the tedious part: writing a commit message.	23:16
	← alley_cat left	23:19
gitte_	Ilari: you want to be Cc'ed?	23:19
Ilari	gitte_: No need.	23:19
gitte_	Could you test?	23:20
Ilari	Where to get the patch?	23:20
gitte_	http://git.pastebin.com/m4d0c7121	23:21
	← bdiego left	23:21
	→ pigeon joined	23:22
	← HackyKid left	23:23
rhalff	how does delete work with git, is it possible to retrieve removed files from a repository revision where the file was still present ?	23:25
gitte_	Of course.	23:25
rhalff	probably not because that would make the repos huge after a while, right ?	23:25
gitte_	That's what SCMs are about.	23:25
	"git checkout <revision> <filename>".	23:25
rhalff	I'm talking about deleted files	23:25
gitte_	Yes, deleted files.	23:26
male	Or git show.	23:26
gitte_	It's called content tracker for a reason ;-)	23:26
context	rhalff: as long as they wered checked in for at least 1 revision, you can retreive it.	23:26
rhalff	so what if I I import a many big files one day, and then remove them, because it was a mistake	23:26
gitte_	Any proper SCM allows you to get back to the _exact_ state where you were when you committed.	23:26
context	rhalff: otherwise there would be no reason for using an scm in the first place ...	23:26
male	rhalff: man git-show	23:27
context	rhalff: if it was a mistake, in git you would reset --hard and drop the revision from the history, or branch-filter it out hopefully before pushing to a viewable repository	23:27
	→ myrizio_ joined	23:27
rhalff	context: ok :) svn has that also ?	23:28
	← myrizio_ left	23:28
context	rhalff: yes using svndumpfilter	23:28
gitte_	context: I think rhalff means "svn also stores all the files, even if they were deleted".	23:28
context	rhalff: you mean keeping deleted files ? of course ...	23:28
gitte_	Yes, it does.	23:28
rhalff	yeah I meant that, thanks gitte_	23:29
context	so does cvs , and hg, and every other possible scm i can think of ...	23:29
gitte_	Some are buggy, though ;-)	23:29
male	So do plain old tarballs.	23:29
gitte_	male: as long as you have them.	23:29
	male: think of Linux 0.02.	23:29
context	rhalff: it keeps deleted source code from files, why wouldn't it keep deleted files. ;)	23:30
rhalff	dunno, because I had trouble trying to get them back with subversion :)	23:31
Ilari	gitte_: The patch seems to fix the issue.	23:31
	→ RobotDeathSquad joined	23:31
robinr	rhalff: don't you ever imply that git has any of subversion's deficiencies :)	23:32
gitte_	Ilari: thanks.	23:32
Ilari	BTW: One I somewhat frequently run into, is that there is no way to clone some remote repsitory, using some existing repository as cache in just one command (it requires at least 6 commands).	23:36
	→ joevandyk joined	23:37
rhalff	robinr: I'm not, I expected an answer like, subversion deletes the files irrecoverably , but git doesn't.	23:37
joevandyk	I'm using git-svn. When I try to do a git-svn rebase, i get a bunch of ifles that "needs update".	23:37
	I'm using git-svn. When I try to do a git-svn rebase, i get a bunch of files that "needs update".	23:37
	oops, sorry	23:37
Ilari	joevandyk: You should commit first before rebasing.	23:37
joevandyk	Ilari, i did	23:37
gitte_	Ilari: git clone --reference does not work?	23:38
joevandyk	oh wait, all the files are ones that I deleted	23:38
Ilari	gitte_: Actually, I want to copy the cache too...	23:39
joevandyk	what happened is i have rails exported into vendor/rails. i deleted rails and did a new export, since i wanted to upgrade. then i did a 'git add . '	23:39
Ilari	joevandyk: 'git add' does not make a commit...	23:40
gitte_	Ilari: you mean the index? That does not make sense.	23:40
joevandyk	but rails moved some files around, so there's some files that used to exist that don't anymore. is there a way to automatically remove those deleted files?	23:40
	Ilari, I know	23:40
	Ilari, my question: how to tell git to remove a couple hundred files?	23:40
	Ilari, since they don't exist anymore	23:40
	→ RetroJ` joined	23:43
joevandyk	Ilari, in other words, say i removed a couple hundred files from various directories in my git project. how should i tell git to automatically remove those files?	23:45
	Ilari, i guess i could parse the output of 'git status', but was wondering if there was an easier way	23:45
	→ Fullmoon joined	23:46
Ilari	joevandyk: If there is no content in tracked files that couldn't be added, 'git add -u' might work.	23:46
RetroJ`	hi. I have created a git repository for a large project that contains subprojects. some of these subprojects are not my software at all. what would be a good way to organize things so that these subprojects can have independent version control in other locations, yet still be included in my repository in a clone, without duplicating the version control?	23:46
	← nud left	23:47
Ilari	joevandyk: There is also better way to obtain list of deleted files... I don't offhand know that...	23:47
joevandyk	Ilari, git rm `git status \| grep deleted \| awk '{print $3}'` did the trick	23:47
gitte_	joevandyk: if you need that more often, "git ls-files --deleted" is what you want.	23:49
joevandyk	gitte_, ah, sweet	23:49
	→ pigeon_ joined	23:53
	→ KirinDav joined	23:55

Logs Search ←Prev date Next date→

Channels Documentation