IRCloggy #git 2008-06-10

Provider of IRC logs since 2005.

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

	2008-06-10
dsal	OK. Start editing, then.	00:00
	← chris2 left	00:00
	→ loops joined	00:00
gass	dsal, humm ... this is wayyyyyy too complex	00:01
	← jkp_ left	00:02
loops	gass, i got bumped so i missed a bunch of the conversation.. but what you wanted sounds easy	00:02
	gass, you just want a new branch yes?	00:02
gass	loops, i know ... but i am a git newbie/ svn fan	00:02
dsal	it's this: git checkout -b mynewbranchname sometag	00:02
gass	dsal, i got that	00:03
dsal	That's all of it. What are you missing?	00:03
gass	but i also want that lenny branch is equal to before "Imported Upstream version 0.11~rc1"	00:03
	→ nat2610 joined	00:03
dsal	git checkout -b mybranchname 90d4675e	00:04
loops	gass, to create the lenny branch, git branch lenny <point you want it to start from>	00:04
	gass, or use dsal's checkout option to create it and check it out in one step	00:05
dsal	Must run to train. good bye and good luck. :)	00:05
	← dsal left	00:05
	→ xaiki joined	00:06
	← jamessharpe left	00:06
gass	loops, that is weird because there are 2 other "branches" called debian/version and the upstream.	00:06
	→ aeruder joined	00:06
loops	gass, why is that weird ?	00:06
gass	because i want a branch with master and debian, at the tag debian/0.10.4-2	00:07
	but mantaining the apart	00:07
	them*	00:08
	→ brosner joined	00:09
loops	gass, you want your master to start at debian/0.10.4-2 ?	00:09
	← aryx left	00:10
gass	yes, and upstream/0.10.4	00:10
	→ lorandi joined	00:10
loops	you can't have master be two things at once..	00:10
gass	but it seems to be ...	00:10
	one is only the debian dir	00:10
	← EiNZTEiN left	00:11
gass	the other is the upstream's source	00:11
loops	gass, so what you're struggling with is that the debian dir and the upstream source are separate?	00:12
gass	yes	00:12
	for example, tree etch has both also	00:12
	and i want to create tree lenny	00:13
	with both	00:13
	← lorandi left	00:13
gass	i branched upstream/0.10.4	00:13
loops	gass, are you still talking about the repo at git://git.debian.org/git/pkg-trac/trac.git ?	00:13
gass	yes	00:13
	← timocratic left	00:14
gass	loops, can you explain the heads, tags and the upstream/ debian/ origin ?	00:14
loops	okay.. i just cloned it.... when i do git branch -r i see origin/HEAD origin/etch origin/master and origin/upstream	00:14
gass	yes	00:15
loops	when i git checkout origin/master and then later git checkout origin/upstream they look similar, i don't see a debian dir	00:15
gass	true	00:15
loops	gass, so i'm struggling to understand what you're seeing and where you're trying to go :)	00:16
gass	lol ... i think i am not seeing the picture ... and mixing things ...	00:16
loops	gass, yeah, i think you've maybe not got a good picture of the ways Git is different from svn yet	00:17
gass	true	00:17
	well ...	00:17
	i want to branch upstream at tag debian/0.10.4	00:17
loops	gass, heads and tags simply point to different point in the history of the repo.. i individual commits in the DAG	00:18
gass	heads are the latest version	00:18
	tags are "in the middle" versions	00:18
	right?	00:18
loops	gass, okay.. so you want a new branch named "lenny" that diverges from origin/upstream at the tag debian/0.10.4 ?	00:18
gass	true	00:19
loops	gass... so simply create a new branch.. git branch lenny debian/0.10.4	00:19
	then you can check it out to see that it really worked :) git checkout lenny	00:19
	← jbunster left	00:20
gass	loops, one important question ... git clean, cleans all changes?	00:20
	→ spearce joined	00:20
loops	gass, git clean is primarily used to remove files that are _not_ version controlled. So it is quite dangerous, because there is no way to use Git to recover them once they are deleted.	00:21
	gass, git reset and/or git checkout is used to cleanup files that _are_ version controlled	00:21
gass	ok,, thanks	00:21
	← schacon left	00:22
loops	gass, i'm not sure it will help for the type of questions you have about Git, or if you've seen it, but you may want to take a peek at http://git.or.cz/course/svn.html anyway	00:23
gass	thanks loops	00:23
	it worked as you said	00:23
	checkout changes the branch?	00:23
	and rebase the tag?	00:23
loops	checkout is much the same as in svn.. it updates your working files to be equal to a given version/branch	00:24
	queso\|away → queso	00:25
loops	gass, so.. git branch lenny debian/0.10.4 creates a new branch named lenny that starts from the debian/0.10.4 tag	00:25
gass	i understood that now :P	00:25
loops	gass, git checkout lenny makes lenny your "current" branch and all your working files equal to that branch	00:25
gass	i was confusing a non confusing scene	00:26
loops	once you understand that then you can understand dsal's shortcut.. you can do both the above steps in a single instruction	00:26
	git checkout -b lenny debian/0.10.4 (which creates a new branch named lenny based on debian tag, and then checks it out)	00:26
	→ hein2 joined	00:26
	← Sho_ left	00:26
	hein2 → Sho_	00:27
loops	"git branch" by itself will show you all your local branches and put an asterisk * beside the branch you currently have checked out	00:27
	→ lorandi joined	00:27
	→ capisce joined	00:27
loops	gass, anyway, hope you've got your legs under yourself now.. Git really is rather easy once you get passed the initial shock of it.. good luck :o)	00:28
	← cmarcelo left	00:29
	→ panagos joined	00:29
gass	loops, thanks ... git really seems to be powerful ... and sometimes not that diferent from svn ... but a lot faster and feature full	00:29
	loops, one BIG feature, make several changes and commits offline	00:29
loops	gass, agreed. Once you get comfortable in Git, svn seems like a straight jacket	00:30
panagos	how can i make git-branch also display branches created from e.g. git-bisect? the -a flag dooesn't do that...	00:32
	← capisce_ left	00:32
Eridius\|work	try git show-ref	00:32
	→ capisce_ joined	00:32
	← texel left	00:33
panagos	Eridius\|work: that does the trick, although it is not as funky as git-branch :)	00:35
	→ lamont` joined	00:35
Eridius\|work	hehe	00:36
	git-bisect doesn't make branches, it makes refs	00:36
panagos	Eridius\|work: well, aren't they the same thing under git?	00:37
Eridius\|work	no, git-branch restricts itself to certain refs	00:37
	notably refs/heads and refs/remotes	00:37
	just like git tag uses refs/tags	00:37
	← capisce left	00:40
	→ thannoy__ joined	00:41
	→ at_random joined	00:41
at_random	hi	00:42
	→ peeja joined	00:42
	→ Mog joined	00:42
Mog	hi i have a small project with just two commiters and all of a sudden i cant commit i get the following error http://pastebin.ca/1043408 any ideas whats wrong / how to fix?	00:44
spearce	objects/ae must be owned by the other user.	00:45
	you'll need to make it writable by both of you, and set the core.shared property with git-config.	00:46
Mog	its in chgrped with us both in the group	00:46
at_random	is anybody using a git repo via samba that lives on a windows server?	00:46
spearce	Mog: but i don't think its writable by the group.	00:46
Mog	hrmm should be, i chgrped it when i made it	00:47
panagos	say i have branches a and b, and i merge b into a (so the tip of b doesn't change); does git keep track of which parent a was formerly pointing to? that could be useful e.g. when viewing a history of merges	00:47
spearce	at_random: it's an odd configuration, in theory it would work, but it wouldn't be recommended.	00:47
	← joeyb left	00:47
Mog	git-config core.shared brought up 0 results...	00:47
at_random	spearce: yea, i know...	00:47
spearce	Mog: http://www.kernel.org/pub/software/scm/git/docs/git-config.html, search for core.sharedRepository (yea, sorry i got the name wrong earlier)	00:48
Mog	no worries	00:48
	thanks for your help	00:48
spearce	panagos: yes and no. yes it stores the prior value of a locally in the reflog (git reflog show a; git log -g a) but it doesn't publish that data when you push or fetch.	00:48
	← at_random left	00:49
panagos	spearce: is there a history visualization tool that can show that? or, does this have anything to do with the colours of the joints between commits in git-gui?	00:51
spearce	there isn't really a good way to visualize the reflog because it can have cycles and so forth over time.	00:52
	but yes, merges show up as joints in gitk (which is git-gui's visualize function)	00:52
	← thannoy_ left	00:54
	← k0001 left	00:56
	← thannoy__ left	01:02
gass	loops, to commit to the server is git push, right?	01:02
	what command to i use to send the local commits to the server?	01:03
	→ sscj joined	01:04
Eridius\|work	gass: git push	01:04
Mog	spearce, can i just chmod -R g+rw the whole project dir	01:04
	to fix previous snafu?	01:04
gass	Eridius_, fatal: The remote end hung up unexpectedly ???	01:05
Eridius\|work	gass: it means exactly what it says	01:05
gass	no log?	01:05
Eridius\|work	what do you mean, log? The connection closed unexpectedly.	01:06
gass	humm ... maybe no permissions?	01:07
zlin	if I set user.email both globally and in a particular repo are both emails supposed to be shown by git config -l ?	01:07
	← agib left	01:08
Eridius\|work	zlin: git config shows the configuration space it sees. if you want to know what's in a specific file, you need to tell it the file to look at	01:08
	→ doener_ joined	01:08
Eridius\|work	otherwise it'll show you the "current" values	01:08
	which means global values overridden by local values	01:08
zlin	it shows both the global value and the local value on separate lines here	01:09
Eridius\|work	o_O	01:09
	oh you're right	01:09
	but if you use, say, git config --get user.email, it'll show you the latter value	01:10
zlin	ahh, yes	01:10
	← peeja left	01:12
spearce	Mog: yes	01:12
	← josb left	01:12
Mog	thanks	01:13
	its all better now	01:13
zlin	thanks btw. :p	01:16
	← pombreda left	01:17
	← wurble left	01:19
	← ketralnis left	01:21
	Mog → mog	01:21
	→ Leefmc joined	01:22
	← david_koontz left	01:23
	← sbeyer left	01:26
	→ Gorbulas joined	01:27
Gorbulas	hello	01:28
gass	Eridius\|work, lol ... no permissions on the git repo	01:28
Eridius\|work	heh	01:28
gass	Eridius\|work, i added a new branch locally, how can i move it to the repo	01:28
	?	01:28
Eridius\|work	by pushing it. read the manpages or tutorials	01:28
gass	git push says everything updated	01:29
	mog → Mog	01:30
	Mog → mog	01:30
gass	Everything up-to-date :(	01:30
Gorbulas	I'm trying to use git with a svn repository hosted by someone else that does not have /branches /tags /trunk etc. will git-svn still work correctly?	01:31
gass	Gorbulas, yes	01:32
Gorbulas	I just do the init without telling it where the tags and branches are?	01:33
	← lorandi left	01:33
	← mjc left	01:33
gass	Gorbulas, yes	01:34
	← digitalhobbit left	01:34
	→ adante_ joined	01:35
	→ lea_w joined	01:45
	← spearce left	01:45
	→ pombreda joined	01:45
	mog → Mog	01:45
	Mog → mog	01:45
	→ iiijjjiii joined	01:52
	adante_ → adante	01:52
	→ amishdude01 joined	01:53
iiijjjiii	are there separate git mailing lists for developers and users?	01:53
	← amishdude01 left	01:53
	← johnw left	01:55
kelvie_	iiijjjiii: nope	01:59
	this was discussed before on the mailing lists, and I think we're not going to go that route	01:59
	I'd think most users of git are developers anyways :p	02:00
	← nkallen left	02:00
	← bobesponja left	02:00
	→ FunkeeMonk joined	02:01
	← madewokherd left	02:01
	→ CodeOfficer joined	02:02
	← Eridius\|work left	02:03
	mog → Mog	02:04
	Mog → mog	02:04
	← DrNick left	02:04
	→ DrNick2 joined	02:05
	mog → Mog	02:08
	Mog → mog	02:08
	queso → queso\|away	02:08
	→ destruct_ joined	02:10
	queso\|away → queso	02:10
	← gitte left	02:12
	→ garrin joined	02:13
	→ gitte joined	02:14
	← kukks left	02:15
	→ ag joined	02:15
	← statim left	02:17
	→ VitaRara joined	02:17
	mog → Mog	02:17
	Mog → mog	02:17
	→ bdiego joined	02:20
	← destruct left	02:21
	→ pombreda1 joined	02:21
	→ jerbear joined	02:22
	← VitaRara_ left	02:22
jerbear	lea_w: you around?	02:22
	← vbgunz left	02:24
jerbear	i'm trying to debug a problem i'm having with gitweb... i've narrowed it down to a single command (git-for-each-ref) that doesn't work on a single repo... the command outputs nothing when run through the cgi script, but it does when running on the command line.. any ideas why this would be different?	02:24
	i have a number of other repos that work correctly, but this one doesn't for some reason	02:25
	← lea_w left	02:25
jerbear	anyone?	02:27
	mog → Mog	02:27
Gorbulas	check the environment?	02:28
	or permissions	02:28
jerbear	Gorbulas: what do you mean?	02:28
	Gorbulas: permissions look good, everything's readable by all	02:28
Gorbulas	maybe you have some odd environment variable set	02:29
	→ bobesponja joined	02:29
jerbear	Gorbulas: that would affect a single repo out of a number of others?	02:29
Gorbulas	or maybe the command is outputting to stderr instead of stdout	02:29
	no I'm just guessing	02:29
	does the repository name have weird characters in it?	02:30
	→ jbunster joined	02:30
jerbear	Gorbulas: interesting... redirecting stderr to stdout has revealed something... "error: refs/heads/master points nowhere!"	02:30
	how the heck could that have happened?	02:30
	← yofel left	02:31
	→ hidenowt joined	02:32
	← hidenowt left	02:32
jerbear	how would i fix this?	02:32
Gorbulas	are you sure the permissions are correct?	02:33
	on .git/refs/heads/master in particular	02:33
jerbear	i can push/pull from this repo... how would i be able to do this if the master branch was not connected	02:33
	← tenpaiyomi left	02:34
	Mog → mog	02:35
jerbear	ok, it was set to 660, setting to 644 fixed it... why would this be set? i certainly didn't do it, and i've been running into this over the past couple of days... permissions not looking correct suid bits set	02:36
	destruct_ → destruct	02:36
	← pombreda left	02:37
	→ huangjs joined	02:38
jerbear	any ideas?	02:38
	would git, itself, change the permissions, for some reason?	02:39
	→ jpieper joined	02:39
	← markkalderon left	02:40
	← jerbear left	02:41
	→ imyousuf joined	02:41
imyousuf	How can I delete a remote branch? :)	02:41
huangjs	Hi, I have a git repo shared through NFS, but when others pushed their changes, the newly created directories and files in .git/objects/ will have other usernames as owner. the problem is the default priviledge of these directories are 755 so others can't write to them. Is there a way to solve this problem?	02:41
imyousuf	I mean I have branch in my remote repo and I want to delete it, how do I do that?	02:42
	→ GeertB joined	02:54
Thumper_	imyousuf: git push remote :branch-to-delete	02:55
	← DrNick2 left	02:56
	→ DrNick2 joined	02:57
	→ [1]intripoon joined	02:59
imyousuf	Thumper_: should first delete the branch locally or just simply run the command?	02:59
Thumper_	no need to delete locally first	02:59
imyousuf	ok trying it	03:00
	→ spearce joined	03:00
imyousuf	Ah thanks Thumper_ it worked :)	03:00
Thumper_	imyousuf: wlcm :)	03:01
	mog → Mog	03:01
	Mog → mog	03:01
	mog → Mog	03:02
	Mog → mog	03:02
	→ kanru joined	03:02
Gorbulas	when I'm using git-svn, is there a way to change the e-mail addresses of people committing to svn?	03:02
	→ at_random joined	03:04
at_random	hi.. i had a strage problem. I made changed, commit, then did "git svn rebase" and got "Using index info to reconstruct a base tree...	03:05
	Falling back to patching base and 3-way merge...	03:05
	error: Entry 'app/views/users/_friend_profile_left_side.rhtml' not uptodate. Cannot merge.	03:05
	fatal: merging of trees a665d43e7a458dd24c4c09b5275c3a5687ad75a7 and e9dcae20b4b1a6a63b617b17d1534d39e763ddd6 failed	03:05
	Failed to merge in the changes.	03:05
	Patch failed at 0001."	03:05
	but when i did git stat, there were unmerged files	03:06
	no changes at all, actually	03:06
	sorry for typos... "made changes, commited, then..."	03:07
	← igorgue left	03:11
	mog → Mog	03:11
	Mog → mog	03:11
	→ igorgue joined	03:11
	→ nkallen joined	03:12
	mog → Mog	03:12
	Mog → mog	03:12
	← intripoon left	03:16
	[1]intripoon → intripoon	03:16
	← pombreda1 left	03:24
	→ pombreda joined	03:26
	→ joevandyk_ joined	03:28
	→ xenoterracide joined	03:30
xenoterracide	does git pull only merge with the current branch?	03:30
	← bdiego left	03:30
spearce	yes	03:31
	→ Deepspawn joined	03:34
	→ eternaleye joined	03:34
	→ fhobia joined	03:34
	← Deepspawn left	03:35
	→ bugfux joined	03:38
bugfux	how can I check out a file that was deleted in a previous commit?	03:38
	→ proxie joined	03:38
spearce	bugfux: `git checkout $thatcommit^ -- $path` where $thatcommit is the commit it was deleted in. the ^ suffix tells git to actually read from the parent of $thatcommit, where $path still exists.	03:39
	that will load the file into your index, making it appear staged for commit, and put it into your working directory.	03:39
bugfux	path is the path to directory right	03:39
	or no to the actual file, but the directory has to exist?	03:40
spearce	well, directory or file. if you give it a directory it gets all files in that directory recursively.	03:40
	← jbunster left	03:40
spearce	no, it doesn't need to exist first either. we'll make the paths we need as we go	03:40
	→ flaguy joined	03:41
bugfux	that worked perfectly	03:43
	← GeertB left	03:45
xenoterracide	vim uses the $PAGER value for git log right?	03:45
	by default	03:45
	→ pombreda1 joined	03:45
	→ stouset joined	03:45
xenoterracide	s/vim/git	03:46
	→ joevandyk__ joined	03:46
xenoterracide	sorry :P	03:46
spearce	yes, we use $PAGER, and default to less if $PAGER isn't set	03:46
	→ jmalonzo joined	03:47
	← janm left	03:47
	→ pombreda2 joined	03:49
	← bugfux left	03:53
	← bryanray left	03:54
	← stouset left	03:54
	← joevandyk_ left	03:55
imyousuf	spearce: regarding mavenizing jgit - I tried with pointing the test resources to the original location but without much result :(. Then I was thinking that I could create a MS lnk file and linux symlink to point to that test resources. what do you think about that?	03:56
spearce	hmm. i'd like to avoid symlinks, .lnk on windows is a crappy substitute	03:56
imyousuf	Yeah, but without copying the file this is the easiest	03:57
spearce	so can we move the test resources? is it just the .pack/.idx data?	03:57
imyousuf	there is another though	03:57
spearce	or are we talking about all of the test class sources?	03:57
imyousuf	yes, its justthe .pack/.idx	03:57
	s/another/another option/	03:57
spearce	oh. those never change. just move 'em.	03:58
imyousuf	ok :), thats what I did :)	03:58
gitte	.lnk do not work with maven, AFAICT.	03:58
	→ janm joined	03:58
gitte	You really cannot rely on symlinks on Windows.	03:58
	← flaguy left	03:58
gitte	Live by it, or die by it.	03:58
imyousuf	hmm	03:58
spearce	you really can't rely on anything on windows. except blue screens.	03:58
gitte	Right.	03:59
imyousuf	yeah the blue screen is quite reliable :)	03:59
gitte	You have symlinks in cygwin.	03:59
imyousuf	it will behace consistently :-D	03:59
gitte	But then, maven is a Java program which does not use cygwin at all.	03:59
imyousuf	s/behace/behave/	03:59
spearce	jgit has some weird code to better handle cygwin based git.	03:59
gitte	What does it do?	04:00
spearce	e.g. sometimes it forks cygpath.exe --windows	04:00
imyousuf	We can also avoid that by introducing a properties file for contains the path and the code will read the path from it by that way we can separate the test resource from the tst folder all together	04:00
gitte	Oh, that's bad. What happens if cygwin is not installed?	04:00
	← pombreda left	04:00
gitte	Does it try over and over again, making Windows _even_ slower?	04:00
spearce	then we assume you aren't using it and just do plain windows calls.	04:00
gitte	k	04:00
spearce	no, we're aren't that stupid. we only call cygpath.exe a few times as we startup opening a repository.	04:01
gitte	But you try once, right?	04:01
	Once per Git call, or once per repo?	04:01
	→ statim joined	04:01
spearce	yea, we scan %PATH% and stat for cygpath.exe on class init. so that sort of sucks.	04:01
gitte	Hrm.	04:01
spearce	if we find it, we call it like 5 or 6 times per repository we open to read .git/config, .git/objects, .git/logs, .git/refs, .git/packed-refs	04:01
gitte	I planned to use jgit in a project of mine.	04:01
spearce	think new-workdir in cygwin.	04:01
gitte	(Actually, the project that takes 104% of my time ATM.)	04:01
	Well, I wanted to have a better solution for new-workdir.	04:02
	But Junio seemed not too keen on the .git file idea.	04:02
	Which would have been just the 1st step, of course.	04:03
spearce	it wouldn't be hard to change how we decide cygwin is available to be more efficient. like make the user actually tell us they want cygwin by setting an environment variable before they start the jvm or something.	04:03
imyousuf	one question, does that mean jgit will not run on windows unless git over cygwin/msys is installed?	04:03
spearce	its more than just the new-workdir, objects/info/alternates when using cygwin paths can be an issue	04:03
	→ peeja joined	04:03
spearce	imyousuf: no, it works just fine without either installed.	04:03
gitte	imyousuf: no, spearce just said it will fall back.	04:03
imyousuf	ok :)	04:03
	← pombreda1 left	04:03
spearce	it just spends a few microseconds stating for a (non-existant) cygpath.exe.	04:03
	→ jmalonzo_ joined	04:04
	← Leefmc left	04:04
gitte	spearce: actually, the few microseconds are a bit longer, because of Windows' filesystems.	04:04
spearce	i'm aware of that.	04:04
	gitte is happy to have _no_ Windows box anymore.	04:04
spearce	but a few microseconds extra on JVM startup for a big IDE like Eclipse is a drop in the bucket.	04:04
gitte	I do not even have to do Windows for work, 'cause I haven't got one.	04:05
spearce	a few microseconds startup for a command line tool blows.	04:05
gitte	'xactly.	04:05
	And I plan to use it as a command line tools.	04:05
	s/s\.$/s/	04:05
	gitte can't even type anymore. Time for another beer.	04:05
imyousuf	spearce: Alternate solution to copying the file is we can introduce a properties file for contains the path and the code will read the path from it by that way we can separate the test resource from the tst folder all together (re-run - I was referring to the paths read in the tst code)	04:06
spearce	java for command line programs in general sort of blows due to the high cost to just initialize the jvm and load the core application classes.	04:06
gitte	spearce: not that bad anymore, what with Java6.	04:06
spearce	imyousuf: so i'm confused, just moving the resources in our source tree to another location isn't suitable?	04:07
gitte	And I thought that the pom.xml changes prevented us from needing symlinks...	04:08
	→ tjafk1 joined	04:10
gitte	BTW spearce: do you think it would be easy to teach jgit about shallow clones?	04:11
	← jmalonzo left	04:11
spearce	not too bad. i think the ugly part for shallow clones is the transpot part, not the revision walking machinary.	04:12
imyousuf	spearce: it works fine, but at some point if the test data is changed or added or new test data is required, then they will require copy-pasting as well. But we have a properties for pointing to the location, then we can put the test resources to a third directory (or keep it in the current one) and read the location prefix from a properties file or system properties.	04:12
	s/we have a/we may\/can have a/	04:12
spearce	imyousuf: i thought we were talking about moving it, not copying it. if we move it and add or change data in the future its only at one location.	04:13
gitte	spearce: okay, but that is basically off-loading the work to the server.	04:13
imyousuf	spearce: oh ok, moving it would not be suitable as that would require specifying different paths for different build system	04:13
	I think I will implement what I am trying to say and send out a patch :), that would make it easier	04:16
spearce	ok. i want to avoid having to store `pwd` into some file somewhere just so the test suite can find its input files.	04:16
	but other than that, i'm ok with moving those resources around if we need to, they aren't that big of a deal to rename. unlike the source code.	04:17
imyousuf	hmm	04:17
	← janm left	04:17
spearce	gitte: the shallow clone is just a set of grafts we need to cache in the repository and apply during revision walking, right?	04:18
	speaking of which, jgit doesn't honor the grafts file. :-\	04:18
gitte	Oh.	04:18
	→ mtodd_ joined	04:18
gitte	okay.	04:18
	DrNick2 → DrNick	04:18
gitte	It is not just for revision walking; those grafts have to be sent to the server when requesting an upload.	04:18
	(i.e. fetch)	04:18
	→ capisce joined	04:19
	→ nkallen_ joined	04:20
	← nkallen left	04:21
	← peeja left	04:23
spearce	i really don't think the graft data would be that hard to implement. may be easier than it is in c git. we have things like garbage collection and a class called ObjectId. no const unsigned char*. :)	04:23
	gitte will give it a try tomorrow.	04:23
	← mog left	04:25
	← FunkeeMonk left	04:27
	→ aroben joined	04:28
	← tjafk left	04:28
	← pombreda2 left	04:30
	← capisce_ left	04:30
	← queso left	04:33
huangjs	Hi, I need some help. I have my git repo on a NFS server shared with others. But when others pushed their commits, the newly created directories in .git/objects/ will have them as owners. Is there a way to ensure the owner is always me or add group writable priviledge to all files and dirs afer git push?	04:33
	→ awx_ joined	04:33
Gorbulas	huangjs: you should look at the mount options for nfs	04:33
gitte	set core.sharedRepository.	04:34
	But you have to chmod manually first.	04:34
huangjs	Gorbulas: I have no priviledge to change the NFS settings	04:34
gitte	Best would be with group sticky bit (g+s)	04:34
huangjs	gitte: I'm doing it manually, it's ...	04:34
	gitte: is there a git config to ensure that? or use post-update?	04:35
gitte	[05:31] <gitte> set core.sharedRepository.	04:35
huangjs	gitte: ok	04:35
	gitte can say it a 3rd time ;-)	04:35
huangjs	gitte: I know how to fix it. thanks!	04:37
gitte	yw!	04:37
huangjs	btw, I started using git since last week, it's really fast.	04:37
	→ mediogre joined	04:37
	lamont` → lamont	04:37
	← bobesponja left	04:38
	→ rafl joined	04:38
	→ spearce` joined	04:41
	← spearce left	04:41
	spearce` → spearce	04:41
gitte	Yep, it's fast alright.	04:41
	→ chendo[work] joined	04:42
gitte	So fast that even Mercurial seemed slow to me, recently.	04:42
chendo[work]	I'm trying to pull a set of changes that includes converting a folder to a submodule	04:42
	→ pergesu joined	04:42
	← nat2610 left	04:42
gitte	I did a "hg pull" on JDK, and I was _so_ annoyed when it came back only a few seconds later, and only to say that it was up-to-date!	04:42
chendo[work]	But when I try to pull, I get an incomplete pull with the error message: fatal: cannot read object e475cf78a49ede1ba6c83fdae7028d7d9da4cd85 'vendor/plugins/attachment_fu': It is a submodule!	04:42
	Merge with strategy recursive failed.	04:42
gitte	Yep.	04:43
chendo[work]	Any ideas?	04:43
gitte	submodule support is, uhm, not complete.	04:43
	I think, however, that in your case it is just the directory attachment_fu not being empty.	04:44
chendo[work]	Oh.	04:44
	Hmm	04:44
	← huangjs left	04:45
gitte	Am I right?	04:45
	← jim___ left	04:46
	← aroben left	04:46
chendo[work]	Uhh there's files in it now because it's a submodule, but the history shows that the entire folder was deleted	04:46
	But since submodule support is not complete...	04:46
	→ aroben joined	04:47
gitte	Well, I think that you are mistaken that it shows the history.	04:48
	I think that it is _still_ a directory, _not_ a submodule.	04:48
	You should be able to tell by looking for .git/ in that directory.	04:48
chendo[work]	Well we're just going to go back to what we were using before	04:48
gitte	If it exists, it is a submodule. If it does not exist, it is _not_ a submodule.	04:48
	Sure!	04:48
chendo[work]	By history I mean git log	04:48
	gitte wonders why you asked, if you go back anyway.	04:49
chendo[work]	Like the delete of the original attachment_fu worder was there	04:49
	Yeah I was wondering if it was a problem that can be solved	04:49
	But I'm just gonna be lazy and just ignore it until submodule support is 'complete'	04:49
	→ peeja joined	04:54
	→ pombreda joined	04:56
	← CodeOfficer left	04:58
	→ CodeOfficer joined	05:02
	→ swtaarrs joined	05:06
	← l0verb0y left	05:08
	← aroben left	05:08
gitte	chendo[work]: that will certainly get submodule support be complete in no time at all.	05:08
	gitte throws the hands in the air and leaves in despair.	05:09
	← gitte left	05:09
	→ aroben joined	05:09
mstormo	chendo[work]: Hmm, that's _really_ in the spirit of OSS..	05:10
chendo[work]	What?	05:11
mstormo	"I'm just gonna be lazy and just ignore it until submodule support is 'complete'"	05:11
chendo[work]	Well, I actually get paid because I work on the projects we're working on, not hacking OSS projects.	05:11
	→ mariusSO joined	05:12
mariusSO	chendo[work]: It _would_ have been nice if it rather said "Hey, that's something I'd like to work on! Give me a sec, and I whip up a patch!".. :-)	05:12
chendo[work]	Yeah it would also be nice if oil was cheap again, there was no world hunger, and there was a cure for AIDS and cancer as well ;p	05:13
	← nkallen_ left	05:19
	→ pombreda1 joined	05:21
	← peeja left	05:21
	→ spearce` joined	05:21
	← spearce left	05:21
	← at_random left	05:22
	spearce` → spearce	05:22
	Arme[0] → Arme[N]	05:23
	→ JonathanT joined	05:25
	→ mithro joined	05:27
	← eternaleye left	05:29
eMBee	good afternoon	05:31
mariusSO	morning	05:32
	(07:32 here ;-)	05:32
	→ tenpaiyomi joined	05:34
	→ robinr joined	05:35
	← pombreda left	05:37
	→ hobbes006 joined	05:38
	→ angasule_ joined	05:38
	→ Fullmoon joined	05:39
	← CodeOfficer left	05:39
	→ FunkeeMonk joined	05:40
	← Fullmoon left	05:40
	→ Fullmoon joined	05:40
	← Fullmoon left	05:41
	→ Fullmoon joined	05:41
	← Fullmoon left	05:41
	→ Fullmoon joined	05:41
	← tenpaiyomi left	05:42
	→ crab joined	05:42
eMBee	good afternoon	05:42
crab	hi.	05:42
	← JonathanT left	05:43
eMBee	what is the best workflow if i want to rebase a bunch of commits, but manually review each one of them,reorder, and possibly edit them?	05:44
	just cherry-pick one at a time?	05:44
tpope	you could rebase -i and set each to edit	05:45
eMBee	hmm, i think cherry-pick; show; amend; ... seems easier in this case	05:46
	→ tsc_ joined	05:48
eMBee	at least one commit should be squashed into another as well	05:48
	hmm, maybe i should try rebase -i to learn something new	05:50
	→ digitalhobbit joined	05:50
	→ RaceCondition joined	05:50
crab	embee: definitely rebase -i	05:51
eMBee	yeah, i just see in the manpage that it can split and squash commits too	05:52
	→ BabelO joined	05:53
eMBee	so it's more me being unfamilar with it that makes the other feel easier :-)	05:53
	better fix that ;-)	05:53
tsc_	Anybody know their way around the http-push code? I'm interested in extending it to push to amazon's s3	05:54
	→ piclez_ joined	05:56
	→ nat2610 joined	05:57
	← robinr left	06:04
	→ roham[rt] joined	06:05
	→ xaxxon joined	06:08
	→ ph^ joined	06:08
xaxxon	I have a git repo on my box. I want to back it up somewhere else. Is there a doc on how to do that? Basically a master/central repository to push to just to have it on another machine	06:09
	→ tenpaiyomi joined	06:11
	← tenpaiyomi left	06:12
	→ deavid joined	06:13
xaxxon	I guess just copying out the .git?	06:13
	→ Antiarc joined	06:15
Antiarc	Hey folks. Newbie question.	06:15
xaxxon	newbie anaswer	06:16
	answer	06:16
	scary part is I may be the only one here..	06:16
Antiarc	I just got back from vacation, and cloned a git repo onto my laptop to work on it. I just attempted a git push back into the repo it was pulled from, but the changes don't show up in my filesystem, though the commit is there in git log	06:16
xaxxon	:\	06:16
Antiarc	So, how do I propagate those changes back to the filesystem?	06:16
xaxxon	you don't	06:16
	that's a big nono in git	06:16
	→ jkp joined	06:17
xaxxon	but I'm not exactly sure why	06:17
	just that you don't push to a repo where you want to see the changes	06:17
doener_	http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	06:17
Antiarc	That seems horribly ignorant of common VCS usage.	06:17
	Thanks.	06:17
xaxxon	I think if you do a fetch from the other end it may work better?	06:17
	→ ajonat joined	06:17
xaxxon	common centralized vcs usagE?	06:18
Antiarc	Well, yeah ;)	06:18
	I'm coming from the SVN world	06:18
	Where I'd normally just do a checkin back to my "home" repo.	06:18
doener_	Antiarc: well, you can't push to a working copy in svn at all, so...	06:18
Antiarc	The local repos are fantastic for working offline, but I need to get it back into my primary development environment now ;)	06:19
	→ mcgrof joined	06:19
doener_	Antiarc: the repo _got_ updated by the push, just your working tree and index are out of date now	06:19
xaxxon	fetch	06:19
	← joevandyk__ left	06:19
Antiarc	Ah, okay, that makes sense.	06:19
xaxxon	whereever you want a working directory, think of that as a client, not a server	06:20
	wherever the stuff you want is is the server	06:20
Antiarc	Okay, that makes sense.	06:20
	xaxxon checks his shoulder to make sure linus isn't there waiting to smack him	06:20
xaxxon	if you pull from another repo to your central repo, it will get the change..	06:21
	← ajonat left	06:21
xaxxon	but it just won't be reflected in the FS	06:21
	← angasule_ left	06:21
	← johan-s left	06:21
	→ ajonat joined	06:21
xaxxon	on the central repo	06:21
Antiarc	I guess I was expecting the clone source to function as a server.	06:22
xaxxon	yep. and that's the mistake :) Every copy is equal	06:22
doener_	Antiarc: it can serve as a server, but only for fetch.	06:22
	(ok, you can push as well, but as the faq says, if you don't know exactly what you're doing, don't)	06:22
Antiarc	Oy, this is going to be a pain.	06:22
xaxxon	just fetch :)	06:23
	alwys fetch	06:23
Antiarc	The problem is getting the machines to see each other.	06:23
xaxxon	oh	06:23
	→ capisce_ joined	06:23
xaxxon	yeah, I have the same problem	06:23
Antiarc	I'm running a Linux VM inside of Windows on the laptop, and the networking is all funky.	06:23
doener_	Antiarc: you can put a bare repo $somewhere reachable from both	06:23
xaxxon	you can push to a repo and then clone another copy on that machine for FS use	06:23
	i.e. have 3 repos	06:24
doener_	then you can push to the bare repo and fetch from there	06:24
Antiarc	hm. seems a little roundabout but would work.	06:24
doener_	for two machines for the same dev it might look like a workaround, but once you got more people, it starts to make sense	06:25
xaxxon	and really, it's the same thing you'd have to do with svn	06:25
	you still need 2 client copies and a server copy	06:25
	right?	06:25
Antiarc	yeah. I was just confused about how git repos functioned.	06:25
doener_	no "oops, box is turned off", and you can then choose what to push to the bare repo, so none of your WIP stuff gets exposed	06:25
Arrowmaster	Antiarc: ohh switching to git finally?	06:27
Antiarc	I've been using git for about 2 months now, but just barely ;)	06:27
Arrowmaster	about the same	06:27
	← Fullmoon left	06:29
	→ thiago_home joined	06:29
	→ daaku joined	06:30
	← swtaarrs left	06:30
Antiarc	Hrm. git fetch from my primary environment to the laptop repo seems to do nothing. Because I did a git push before joining the channel, perhaps? ;)	06:30
doener_	Antiarc: exactly ;-)	06:31
Antiarc	chuckle and so the question becomes, what's the right way to revert that before I go screw this up further?	06:31
	← digitalhobbit left	06:31
Antiarc	pull a previous commit?	06:31
doener_	Antiarc: did you have any changes on the box you pushed to that weren't on the box you pushed from?	06:31
Antiarc	Nope.	06:32
doener_	git reset --hard	06:32
Antiarc	(which, in theory, makes this -really easy-)	06:32
	Aha!	06:32
	Okay, I see now.	06:32
	that resets to the state in the repo, at the expense of any local changes.	06:32
doener_	exactly	06:32
Antiarc	Thanks a ton.	06:32
	→ namenlos joined	06:32
mcgrof	crap, git-p4 submit --continue is not a real valid option	06:33
doener_	and the local changes you currently see are what your working tree/index are missing from the pushed stuff	06:33
mcgrof	anyone know a decent alternative? The only other options I see are -M and --origin	06:33
xaxxon	and in your "central" repo, all you need is the .git directory.. you can grab that, move it somewhere and delete all the FS stuff	06:33
Antiarc	Okay, cool.	06:33
doener_	well, or just "git --bare init" somewhere, and then push there	06:34
	← capisce left	06:34
	← spearce left	06:34
	→ njal joined	06:36
	→ robinr joined	06:36
njal	what is the recommended way of creating a sandbox branch for myself in a git repository? just create the branch using git branch, then deleting everything currently in it?	06:37
xaxxon	what are you deleting?	06:37
doener_	define "sandbox branch" please	06:37
xaxxon	we can wait :)_	06:38
njal	"sandbox branch" = (in my head, at least); small pieces of example code etc that I write for myself and fellow developers that is somehow related to the project - prototype code etc	06:38
Antiarc	that's just a branch	06:38
	branch, make your changes, commit to that branch, and switch back to your original branch/master branch	06:39
	← alecs left	06:39
xaxxon	you almost want a parallel comments section?	06:39
	← nat2610 left	06:40
xaxxon	more like railroad tracks than a branch	06:40
njal	xaxxon: thats true. I guess the concept of a sandbox branch doesnt really make sense, since I would probably require code that is in master branch while writing	06:41
xaxxon	there's not really anything like that, I don't think.. like.. here I am in my project and I want to see my docs for it, so I checkout the doc "branch" and now I have my docs	06:41
	now I check out my unit-test branch and I have the unit tests for that project	06:42
AlexB	Leftover Subversion-ism. :-)	06:42
	← pombreda1 left	06:43
	→ bentob0x joined	06:43
	→ jkp_ joined	06:46
	→ _zack joined	06:47
	← _zack left	06:48
	→ _zack joined	06:48
	← piclez_ left	06:51
	→ angasule joined	06:51
	→ priidu joined	06:54
	← priidu left	06:57
	→ jkp__ joined	06:57
	→ janm joined	06:58
	← jkp left	06:59
	→ priidu joined	07:00
	← robinr left	07:01
	→ Modius_ joined	07:01
	← chendo[work] left	07:03
	→ lea_w joined	07:11
	← jmalonzo_ left	07:11
	← jkp_ left	07:11
	← Antiarc left	07:14
	→ normalpe1son joined	07:18
	→ shawn__ joined	07:19
	→ Ted joined	07:20
	← thiago_home left	07:21
	← lea_w left	07:28
eMBee	what exactly happens (or is supposed to happen) on git fetch <new-remote-repo> ? i don't see any branches from the remote repo appear. do i need to specify them?	07:31
	← aroben left	07:32
eMBee	ok, seems pretty clear that i need to specify the branches or add the repo first, i just wonder then wjat fetch does without those. it does something.	07:35
	← jkp__ left	07:35
eMBee	seems like it does copy all commits	07:36
doener_	IIRC it stores the refs in FETCH_HEAD, I think...	07:36
	← igorgue left	07:37
doener_	* branch HEAD -> FETCH_HEAD	07:37
eMBee	yes, i was wondering which commits though, since no branches are specified, it could get none as well	07:37
doener_	so, yeah...	07:37
eMBee	like how does it know which commits to get?	07:38
doener_	well, seems that it fetched HEAD for me	07:38
	fetched from a local git.git clone	07:38
eMBee	yes, looks like the same here	07:38
	heh, that's exactly what i did as well :-)	07:38
	← nofxx left	07:39
	→ JonathanT joined	07:40
	→ csc` joined	07:45
	← RaceCondition left	07:48
	→ lea_w joined	07:48
	← krawek left	07:52
	← mtodd_ left	07:52
	→ mtodd_ joined	07:52
eMBee	hmm, sometimes i get confused by simple things: does A..B include A or not?	07:53
	i think not, because A should be reachable from A	07:54
	→ nud joined	07:58
	← fhobia left	08:00
	→ reel_ joined	08:01
	← tsc_ left	08:04
	← brosner left	08:05
	→ alenoosh joined	08:11
	→ nutrimatt joined	08:17
	→ RaceCondition joined	08:22
Arme[N]	I just merged a project(A) with subtree merge to a project(B) and then I changed A project through project(B) working copy, now I want to push those changes in to the A repository, How to do this? I can't find a real example of this situation in subtree merge HOWTO page	08:26
	← daaku left	08:26
	→ redondos joined	08:30
	→ ebel joined	08:31
	→ sbeyer joined	08:33
	thiago decides to repack: 192 packs and 105 MB worth of unpacked objects	08:34
	→ GregersP joined	08:34
thiago	eMBee: that's now how you read it	08:35
	eMBee: you should read "everything reachable from B, except what's reachable from A"	08:35
	since A is reachable from A, A.. will exclude it	08:35
	← mithro left	08:36
	→ mcella joined	08:43
	→ Fullmoon joined	08:48
	→ petteri joined	08:49
	dwmw2_gone → dwmw2	08:52
	pilif\|away → pilif	08:56
	→ charon joined	09:00
	→ koke joined	09:03
	← ajonat left	09:07
	→ ajonat joined	09:07
	→ ft joined	09:07
	← mediogre left	09:20
	← _zack left	09:25
	→ chris2 joined	09:25
	← Tv left	09:25
	→ n00b2000_ joined	09:27
n00b2000_	hi. I have a git repository on which I accidentally ran git-reset --hard <some old tag>	09:27
	(sorry for the message break) how do I revert back to the up-to-date head?	09:27
charon	n00b2000_: look for -g in 'man git-log', look at its output to figure out the '...' in reset --hard HEAD@{...}	09:30
thiago	if you have done absolutely nothing since, you can try this too:	09:32
	git reset --hard ORIG_HEAD	09:32
n00b2000_	charon, thiago - thanks, you saved my ass	09:34
	← n00b2000_ left	09:34
	jeckel → away	09:35
	lu_zero_ → lu_zero	09:36
beeble	what would be the best way if i want to "checkout" an old revision to do some testing, but be able to change back to the current HEAD?	09:37
mariusSO	beeble: If you want to do changes to that old revision when testing, create a branch for it. 'git checkout -b test_of_old_stuff <sha1 of old revision>'	09:38
	if not, just 'git checkout <sha1 of old revision>', and use a so-called "detached head"	09:39
	Then just 'git checkout <previous branch name>' to get back to what you had before	09:40
	→ CodeOfficer joined	09:40
beeble	mariusSO: thx, i should have read the message from git checkout <sha1> more carefully. seemed like an error but with the detached head in mind it looks right	09:43
mariusSO	The "error" with a detached head is that you cannot make a commit to it, unless you first create a branch for it.	09:44
	other than that, harmless	09:45
beeble	everything with a lot of output seems wrong at the first moment ;)	09:46
	but yes, it just want to warn me. so it's fine and good	09:46
	← thesaint4444 left	09:47
	→ Tv joined	09:53
	→ lea_w1 joined	09:53
	← lea_w left	09:53
	lea_w1 → lea_w	09:53
	→ _zack joined	09:54
	← lea_w left	09:56
	→ lea_w joined	09:56
Catfish	When I git stash && git stash apply, Xcode thinks that the project file has changed and wants to reload it	10:01
	Any guesses why?	10:01
bremner	Catfish: file access time?	10:02
Catfish	bremner: If that was the case, opening the project file in an editor & closing without saving would have the same effect, right?	10:03
	dwmw2 → dwmw2_gone	10:03
dato	uh, I'd expect git stash apply to alter the mtime, no?	10:04
Catfish	Even if the file hasn't changed?	10:04
bremner	Catfish: well, ls -l confirms that git stash apply updates mtime.	10:07
Catfish	hm	10:07
	On a totally separate note, doesn't Majordomo offer daily digest mails?	10:08
	awx_ → awx	10:10
	→ markkalderon joined	10:10
	← CodeOfficer left	10:11
	→ lea_w1 joined	10:12
	← lea_w left	10:12
	lea_w1 → lea_w	10:12
	← lu_zero left	10:14
	→ lu_zero joined	10:14
Debolaz	Hrmm.. Is it possible to get git to only pack x amounts of objects at a time? The problem is that even counting the objects in the repository takes too much memory.	10:17
	→ Ademan joined	10:20
	→ kukks joined	10:20
	→ capisce joined	10:22
	→ jkp joined	10:22
jkp	hi all...how do i clone a specific branch of a repository?	10:23
waawaamilkvampyr	jkp: not sure if this is what you're after, but:	10:25
	git clone -n [remote]	10:25
	git checkout [branch]	10:25
	?	10:25
	← mcella left	10:26
	→ mcella joined	10:26
	→ lea_w1 joined	10:27
	← lea_w left	10:27
	lea_w1 → lea_w	10:27
mcgrof	anyone know if Id tags are standard of some sort?	10:28
	like who the heck came up with it?	10:29
cypher23	what do you mean with "id tags"	10:30
	?	10:30
Arjen	jkp: You can't clone just a branch, you always clone the whole repository	10:30
waawaamilkvampyr	presumably $Id$ as is common in CVS ?	10:30
mcgrof	cypher23: #ident "$Id: //foo/bar/crap.c#5 $"	10:31
broonie	mcgrof: Not officially, though quite a few revision control systems use the same syntax.	10:32
mcgrof	broonie: know which one might have used it fist? Just curious	10:33
broonie	mcgrof: RCS or SCCS.	10:33
mcgrof	also, because of it it seems it can reate huge issues with fast-import scripts and wrapper for git	10:33
	← capisce_ left	10:34
Arjen	Why?	10:35
mcgrof	running into this same issue with git-p4 right now. If git would support it, which seems easy to do it remove sthis burden from wrappers and import scripts	10:35
	← imyousuf left	10:35
broonie	What is the issue which you are running into?	10:36
mcgrof	right now git-p4 is broken, because of this same issue, (patch pending, I'll send it to the list soon). Was trying to figure out the best way to deal with it but it would be all penalties on performance and they are just hacks	10:36
	git-p4 right now clones and removes the tags, however, the patch will feail on the depot as the tag was removed.	10:37
	The right solution seems to just have git deal with it IMHO	10:37
	as a flag at least	10:37
	otherwise your wrappers need to hack it up	10:38
	I just removed this check in git-p4 and its been working fine, except for cases where I have a series of commits and the tags will change and I have hunks with the tags within.	10:39
	another consideration is the directory move - the tags will change	10:39
	so I have to deal with this right no wmanually	10:39
	was looking for the right solution and I think if git could handle this internally, maybe as an option, it would help with wrappers	10:40
	← markkalderon left	10:40
mcgrof	what do you think?	10:40
	I'll just address this on the lists	10:41
cebewee	can perforce kill the tags on itself?	10:43
	for cvs/git-cvs-import you use -kk for cvs and -k for git-cvs-import and it works fine.	10:44
	→ koke_ joined	10:44
mcgrof	cebewee: I was looking into that, I couldn't find it :(	10:44
	cebewee: that would be the proper wrapper solution for git-p4 in the meantime	10:45
	git-p4 uses p4 print	10:45
	so it would have to be an option to it, and I don't see one	10:45
	there's -q flag which does: suppresses the initial line that displays the file name and revision, but that doesn't remove it from the source code print	10:46
	at first, I was considering using some sed, but even then, that wouldn' work as you still then have to retrieve the right Id when you want to apply the patch in perfoce, which it always leaves it	10:47
	so to submit into p4 your hunk needs to right tag	10:48
	if git had it, well this would resolve the issues. I guess the issue there is if this is a standard or not	10:49
	and if different version control systems use different implementations	10:49
	so hence my original question	10:49
	← lu_zero left	10:49
cebewee	i think such tags are broken, as they create conflicts when there are none.	10:50
	and you need to differentiate between binary and text files	10:50
	→ Xyrio joined	10:52
jettero	is there a git remote command to tell the remote to merge on a push from the source?	10:53
mcgrof	I don't care two shits about the tags, for all I care they can disappear from the face of te earth, but a solution for wrappers is still needed :(	10:53
	← koke left	11:00
	→ anathematic joined	11:00
	→ lu_zero joined	11:01
	← Xyri0 left	11:02
Debolaz	Gah, this object count is going in 10 objects per second...	11:02
	And there are 1.5m of them to count.	11:04
mugwump	repack --window=10000 ? :)	11:08
mcgrof	heh I just ran into a case where the p4 reduced the #version number from the tag instead of increasing it	11:08
	→ l0verb0y joined	11:08
jengelh	just that a superwindow does not help all that much than	11:09
	Lucine2 → Ariella	11:10
mariusSO	mugwump: Will make checking out files really slow, since you potentially have to apply 10k deltas to a file before you get the revision you'd want	11:10
mugwump	mariusSO: you're confusing it with --depth	11:11
	mariusSO smacks self	11:11
mariusSO	that's correct	11:11
mugwump	big windows are fine, I normally don't use more than about 200 for my brute-force packing attempts though	11:11
	don't really know where the typical point of diminishing return is	11:12
	← anathematic left	11:12
mariusSO	would be an interesting graph	11:13
mugwump	hmm, could perhaps calculate it on carbon belch day	11:14
mariusSO	same with unpack time in relation to --depth	11:14
	correlated with final repo size	11:15
mcgrof	no I take that back it didn't reduce the id number	11:18
Debolaz	mugwump: I use --window=100 now. Doesn't change the fact that git wants to swallow enormous amounts of memory just to look at the repository.	11:20
mugwump	which operation did that?	11:20
	→ jasam joined	11:22
thiago	suggestion: git tag also take a --contains argument, like git branch	11:22
Debolaz	mugwump: The counting stage of git-repack.	11:23
	So right now I'm tarring the repo up and moving it to a machine that has a few gigs of ram.	11:23
	→ markkalderon joined	11:24
	← RaceCondition left	11:24
Debolaz	Counting 10 objects per second just isn't acceptable.	11:24
mugwump	your system is thrashing then ? ie, lots of si/so visible on vmstat	11:25
	that does sound a bit odd, I would have thought it could just hit the index to count. have you looked at the source for hints?	11:26
	← markkalderon left	11:27
	← Sho_ left	11:27
Debolaz	I recall that when counting, it actually opens every individual file. I'm guessing it does so in an extremely inefficient way too, because tar is speeding through them without any issues.	11:27
	Recall from previous investigation on an unrelated issue.	11:28
mugwump	hmm	11:33
	I think the problem is it has to walk the entire object graph	11:33
Debolaz	tar will be done in an hour or so I think. git would had literally spent months on this.	11:33
mugwump	to see which objects are connected	11:33
Debolaz	mugwump: For counting?	11:33
mugwump	for repacking	11:33
	for counting which objects need repacking	11:34
	you're using repack -a -f ?	11:34
Debolaz	Yes.	11:34
mugwump	how many packfiles do you have?	11:35
	it does sound like you're running into an unfortunate worst case tbh	11:35
Debolaz	None. As I said, I've been unable to pack it.	11:35
	It will literally take months to make any packs at all on this repo.	11:35
mugwump	oh. what does git count-objects -v say?	11:35
	← roham[rt] left	11:36
mugwump	(that should use a faster count method, I think)	11:36
Debolaz	It's counting roughly 10 objects a second. That means the counting stage alone, and nothing else, will take 3 months.	11:36
mugwump	or, how many loose objects do you have, roughly?	11:36
Debolaz	There are 1.5m objects.	11:36
mugwump	ok	11:36
	which filesystem?	11:36
Debolaz	UFS2.	11:36
mugwump	heh	11:36
	→ Kbyte joined	11:37
mugwump	does that have directory hashing of any type?	11:37
	← sbeyer left	11:37
jengelh	even FAT would perform better! </s>	11:37
Debolaz	Yes. But you're missing a point I made before now, tar has no such performance problems.	11:37
	I'm tarring this repo up now at a very high speed, with compression.	11:37
mugwump	it's doing a different thing	11:37
	you're comparing a disk-level backup to a vacuum, in DB terms	11:38
	→ Sho_ joined	11:38
Debolaz	I realize there's a difference, I'm just saying I'm having a hard time seeing that the filesystem itself is the performance problem. Random access to any arbitrary object seems very fast.	11:38
	But directory hashing is used in the filesystem yes.	11:39
mugwump	ok. this is FreeBSD I assume?	11:39
Debolaz	Yes.	11:39
mugwump	I'm just wondering whether it might not go so slowly on Linux, I think it's inode/VFS stuff is faster for this particular case	11:39
Debolaz	I'll start trying other systems when the tar is ready.	11:40
mugwump	sure. this is why the auto-repacking was added, really	11:40
	you tend to blow the OS' dircache (or equivalent) with usage like that	11:40
	you might be able to set some kernel tunables and increase the size of the cache perhaps	11:41
	you're not importing ports are you? :)	11:41
Debolaz	But can't git simply not try to pack the whole object tree at once? Doing it chunks at a time seems to solve this issue.	11:41
dsop	can i something like an incremental blame? So like: this line first appeard in commit abc then modified in commit cd1 and commit db2	11:41
mugwump	sure, you can do that	11:41
Debolaz	I'm importing FreeBSD. :-)	11:41
mugwump	right, yes	11:41
	→ l0verb0y- joined	11:42
mugwump	git-fast-import also solves this in a much better way	11:42
	output all the revisions of an RCS backing file from a CVS repo in order, for instance, and you already get a whole lot of repacking on your first output run	11:43
	← Kbyte left	11:44
	→ Kbyte joined	11:44
Debolaz	But isn't there a program that can take a directory of objects and simply just compress them into a pack without having to start counting the whole repo, or any of the extra stuff, just compress a bunch of files?	11:45
	← Kbyte left	11:45
Debolaz	Heck, I don't even need compression, as long as the number of files are reduced.	11:45
mugwump	sure, git-pack-objects will wok like that if you like	11:45
	→ Kbyte joined	11:45
mugwump	but seriously, if you find yourself doing this multiple times, I recommend adopting the approach used by git-p4raw (on utsl.gen.nz/gitweb/?p=git-p4raw)	11:46
Ilari	git-repack without the -a option?	11:46
	→ mikeando joined	11:46
	→ lcapitulino joined	11:47
Debolaz	Ilari: I don't think that would help much.	11:47
	← lcapitulino left	11:47
mugwump	ie, break up the conversion into phases: 1) survey input metadata (like parsecvs) and place into Postgres tables, 2) export file blobs (to git-fastimport), 3) mine out history from metadata, place into further tables, 4) export branches to git-fast-import	11:48
	if you make all of the stuff that touches the database transaction-safe, you don't drive yourself nuts with repeated runs	11:48
	← mikeando left	11:50
	← Ted left	11:52
crab	hi.	11:53
	so i cherry-picked some changes from master into rel, but they have different commit ids, and now if i do "git log rel..master", that change is shown even though it's been merged.	11:54
	→ Ted joined	11:54
crab	what can i use to see only those changes that i really need to merge?	11:54
	← Fullmoon left	11:55
	← chris2 left	11:55
mugwump	why, with git log --left-right --cherry-pick rel...master, of course	11:55
crab	oh, of course.	11:55
	:)	11:55
	← Ted left	11:56
crab	except that it still shows me that commit.	11:57
	← l0verb0y left	11:58
mugwump	bah	11:58
	do the git-patch-id's for the patches show the same?	11:58
	also, I might have the order of the arguments backwards	11:58
crab	yes, patch-id shows the same thing. reversing the arguments didn't help.	11:59
mugwump	hmm. well, at the very least, git rev-list with those options should show a "+" or a "-" by the ones that do match	12:00
	obviously they're not just excluded like I was hoping	12:00
crab	git-rev-list shows only ">" lines.	12:01
mugwump	you are using three "."'s, right?	12:03
	you know, I think I've encountered cases where this didn't work	12:04
	because the "..." takes the latest merge base, not the first point of divergance	12:04
	that bit where it describes the merge base as:	12:05
	s/merge base/symmetric difference/	12:06
	git-rev-list A B --not $(git-merge-base --all A B)	12:06
	→ Fullmoon joined	12:06
mugwump	this is probably wrong	12:06
	a symmetric difference arguably should use an earlier commit than that	12:07
Debolaz	mugwump: Piping find objects/ into git-pack-objects yielded significantly better counting performance.	12:07
	← nud left	12:07
crab	mugwump: damn, i wasn't using three dots. sorry!	12:08
mugwump	Debolaz: of course, the chances of git doing a good job at compressing that lot in any kind of reasonable time are very slight	12:08
crab	thanks, that works.	12:08
Debolaz	mugwump: But we're merely talking about counting performance though, not compression?	12:09
mugwump	sure	12:09
	I found that repacking a repository like that yielded a 350MB pack for perl	12:09
	then, later, one produced where I'd fed the file revisions into git-fast-import in a per-file order (ie, all the revisions of a file sent in order), it got down to 110MB	12:10
	← intripoon left	12:10
mugwump	with the same repack options	12:10
	→ d0k joined	12:11
mugwump	the nature of the repack object window selection algorithm is that well corellated history packs better	12:11
	→ intripoon joined	12:11
	← intripoon left	12:12
	→ intripoon joined	12:13
	← koke_ left	12:14
mugwump	anyway, I have to go to bed - but do go and have a look at git-p4raw - and I'm more than happy to discuss this with you offline - FreeBSD is one thing I'd love to have a complete GitTorrent of :)	12:14
	best of luck, Debolaz !	12:15
	→ Beket joined	12:15
	← intripoon left	12:15
	→ lorandi joined	12:16
	← lorandi left	12:16
	→ intripoon joined	12:17
	← crab left	12:19
	→ lorandi joined	12:21
	→ EmilMedve joined	12:23
	→ juanjoc joined	12:25
	→ vanderkerkoff joined	12:27
	→ nud joined	12:27
vanderkerkoff	hello peeps. Some numptie forgot to not put our documents folder into git. I tried git rm -r 'public/\*ments/' but it's failing	12:28
	how do I remove the folder public/documents and all things in it?	12:29
Catfish	*ments?	12:31
vanderkerkoff	documents	12:31
	:-)	12:31
Catfish	Doesn't git rm -r public/documents work?	12:31
vanderkerkoff	two secs	12:31
	I'll tell you the error now	12:31
	fatal: pathspec 'public/documents' did not match any files	12:32
	git rm is for files I th ink	12:32
Catfish	And public/documents exists in your working copy?	12:33
vanderkerkoff	yep	12:33
	← alenoosh left	12:33
Catfish	Works fine for me on 1.5.5	12:33
vanderkerkoff	from this page it seems that git rm is for files	12:33
	http://www.kernel.org/pub/software/scm/git/docs/git-rm.html	12:33
	not for folders	12:33
Catfish	-r	12:34
	Allow recursive removal when a leading directory name is given.	12:34
	→ agib joined	12:34
Debolaz	vanderkerkoff: git doesn't really track folders.	12:34
vanderkerkoff	does that mean recursivley remove the directories?	12:34
Catfish	No	12:34
	It removes any files within that directory	12:35
vanderkerkoff	righto	12:35
	← lu_zero left	12:35
vanderkerkoff	I got you both, so if you want to remove a directory, how woulk you do that?	12:35
	but if git doesnt track it	12:35
Catfish	You just remove all files within it	12:35
vanderkerkoff	it's going to be hard	12:35
	:-)	12:35
Debolaz	vanderkerkoff: You don't, you just remove the files. There's no such thing as a directory in git. :)	12:36
Catfish	Which can be achieved by git rm -r directory	12:36
vanderkerkoff	here's my structure, public/documents/4/original/filename.ext	12:37
	the number changes with every file uploaded	12:37
	as does the filename	12:37
	so I'll have to manually go through all the number files and make sure there's no files in them wont I?	12:38
	using git rm	12:38
	hmm	12:40
	i think my local system must be bolloxed	12:40
	← garrin left	12:40
Arjen	What does 'git status' say, after the 'git rm -r public/documents' ?	12:40
vanderkerkoff	same as it does prior to it failing Arjen	12:41
	there's something wierd going on here	12:42
	I can drill right down in the folders to the docuement itself	12:42
	and git rm -r nameofdocument and it still gives errors	12:42
	→ vbabiy-laptop joined	12:42
vanderkerkoff	fatal: pathspec 'public/documents/1/original/AcademicOffences.doc' did not match any files	12:43
	I'm in the original folder	12:43
	here's the command	12:43
	git rm AcademicOffences.doc	12:43
	so I might think that git doesn't know about that file	12:44
	so git status	12:44
	looks ok	12:44
	:-)	12:44
Catfish	Are you using the right case?	12:44
	As in, does git think that the folder is 'Documents', and your local filesystem is giving 'documents'	12:45
vanderkerkoff	ahh	12:45
	there's only one folder, documents	12:45
Catfish	git ls-files public/documents ?	12:47
vanderkerkoff	Catfish, you beauty	12:48
	→ [1]intripoon joined	12:48
vanderkerkoff	you were right, git knows nothing of the entire documents folder	12:48
	:-)	12:48
	that's why you can't delete it	12:48
	:-)	12:48
	→ eikke joined	12:50
	← bentob0x left	12:50
eikke	is there any easy way to find out which commit removed/changed a given line?	12:50
	(ie in which commit a method was renamed)	12:51
vanderkerkoff	can you search in gitk for it?	12:51
eikke	no gui :)	12:51
Arjen	eikke: git log -S ?	12:52
	→ deavidsedice joined	12:52
	Arme[N] → Arme[0]	12:52
	Arme[0] → Arme[N]	12:52
eikke	Arjen: whats the syntax if 'string' then?	12:53
Arjen	git log -S'string to search for' -- file_to_search_in	12:53
	something like that	12:53
eikke	thanks	12:53
	→ nrl joined	12:58
	← Tv left	12:59
	→ koke joined	13:03
	→ stouset joined	13:05
	← intripoon left	13:05
	← ajonat left	13:07
	→ sgrimm joined	13:10
	→ intripoon joined	13:12
	→ redondos_ joined	13:13
	→ holizz joined	13:16
	← redondos left	13:17
	redondos_ → redondos	13:17
	← VitaRara left	13:18
	→ ajonat joined	13:21
SRabbelier	Arjen: and how do you do that if you want to search the logmsg instead of the diff?	13:25
Mikachu	log --grep=hello	13:26
SRabbelier	Mikachu: mhhh, and plumbing style that's git-rev-list --grep ?	13:26
Mikachu	i daren't say, there was some discussion recently about those two commands that confused me :)	13:27
SRabbelier	Mikachu: how about the '-S' switch to git log..?	13:27
Mikachu	didn't we just establish that searches the diffs?	13:27
	→ schacon joined	13:27
SRabbelier	Mikachu: yes	13:27
Arjen	SRabbelier: the --grep is in rev-list	13:27
SRabbelier	Mikachu: but, plumbing style, there's no -S in rev-list	13:27
	← schacon left	13:28
Arjen	It's a diff thing, and that's not in rev-list	13:28
Mikachu	i'm not going to say if any of them is plumbing or not	13:28
SRabbelier	Arjen: mhhh... is there a plumbing-style way to do -S?	13:29
Mikachu	but obviously you seem quite convinced already	13:29
SRabbelier	Mikachu: anything in git log is porcelain :P	13:29
	Mikachu: man git(7) tells me that git log is porcelain, so I can't use "git log -S"	13:29
	← [1]intripoon left	13:30
	Arme[N] → Arme[0]	13:30
Mikachu	SRabbelier: grep 'git log' *.sh	13:30
SRabbelier	Mikachu: heh, I know.. it's bad :(	13:30
Arjen	SRabbelier: Why can't you use it?	13:30
SRabbelier	Arjen: not supposed to use porcelain in my GitStats project :P	13:31
	← vanderkerkoff left	13:32
Arjen	SRabbelier: You can control the exact output of log with --pretty, so I don't see it as a real problem	13:32
	→ [sand-man] joined	13:32
	← proxie left	13:32
SRabbelier	Mikachu: also, it's only git-bisect.sh, git-merge.sh, git-request-pull.sh, git-stash.sh, git-submodule.sh; that are sinning, and only in a few places (about 10)	13:32
Arjen	Otherwise, use git rev-list and 'manually' do a diff-tree on the successive commits	13:32
SRabbelier	Arjen: yeah, I did, too slow :(	13:32
Mikachu	SRabbelier: there aren't any more users of rev-list	13:33
	11	13:33
Arjen	SRabbelier: Surprise, surprise :-)	13:33
	→ bryanray joined	13:33
SRabbelier	Arjen: not really, forking != cheap :(	13:33
Mikachu	SRabbelier: ask junio :)	13:33
SRabbelier	Mikachu: on how to do it?	13:33
Mikachu	if git-log is porcelain and/or plumbing	13:33
SRabbelier	Mikachu: I'll do it together with asking about "git branch --contains"	13:33
	there's no counterpart	13:33
	→ jerbear joined	13:34
Mikachu	git branch is not plumbing either :)	13:34
SRabbelier	Mikachu: it's under Main porcelain commands	13:34
	Mikachu: exactly	13:34
	Mikachu: and there's no plumbing counterpart of "--contains" either	13:34
	← panagos left	13:34
SRabbelier	and doing it manually is, you guessed it, slow..	13:34
	→ panagos joined	13:34
jerbear	can someone help me understand how alternates work?	13:34
	→ tenpaiyomi joined	13:34
	← ciaran29d left	13:35
	→ cmarcelo joined	13:35
	← intripoon left	13:35
Arjen	SRabbelier: Well, git log can even output NUL	13:35
	s between commits	13:35
Mikachu	jerbear: you write a path to another git repo in the alternates file, and when git tries to look up an object locally that it can't find, it looks in that object repo too	13:35
SRabbelier	Arjen: mhhh, yes, that's true.. mhh	13:35
Mikachu	jerbear: everything is written locally though	13:35
	jerbear: (if it can't be found in the alternate)	13:36
SRabbelier	Arjen: I guess it's porcelain but with plumbing tendencies :P	13:36
Arjen	SRabbelier: And with a --pretty=$format you can exactly control the output	13:36
	→ venger joined	13:36
Arjen	SRabbelier: So in this case, I don't see a real problem with using git log	13:36
	→ intripoon joined	13:37
jerbear	Mikachu: i'm trying to check out this gitweb fork.. it says that if i have an existing git repo, i can just use --reference when cloning it, but i'm not sure how to use it once it's been cloned	13:37
SRabbelier	Arjen: mhhh, sound argument, I'll poike the list though :)	13:37
venger	git gui has a nice little "Revert changes" selection that revert changes to a specific file. What is the cmdline equivalent to what it is doing ?	13:37
Mikachu	jerbear: use what? --reference?	13:37
SRabbelier	venger: git checkout -- file	13:37
Mikachu	jerbear: it's an option to clone, so i guess you can't then	13:37
	jerbear: it'll copy objects from the referenced repo instead of fetching it from the remote	13:38
venger	SRabbelier, interesting, thanks	13:38
SRabbelier	venger: yw	13:38
	venger: note that 'git revert' is totally different, it actually creates a -commit- that reverts another one	13:38
	→ jim__ joined	13:39
jerbear	Mikachu: is there a way to remove an alternate?	13:39
Mikachu	git gui should probably say something like 'undo work tree changes' to avoid confusion	13:39
	jerbear: i don't know if reference adds an alternate	13:39
SRabbelier	Can I specify a commit date on the commandline, or is GIT_[COMMITTER\|AUTHOR]_DATE the only way	13:39
	?	13:39
jerbear	Mikachu: is there a way to tell what alternates exist?	13:41
	→ eternaleye joined	13:41
Mikachu	jerbear: cat .git/info/alternates iirc	13:41
jerbear	Mikachu: ok thanks	13:41
Arjen	SRabbelier: I think only the env variables	13:41
SRabbelier	bummer	13:42
Arjen	(That's how commit-tree gets its information)	13:42
	SRabbelier nods	13:42
SRabbelier	ah well, on to manual labor then	13:42
	→ [1]intripoon joined	13:42
	← [1]intripoon left	13:43
	→ vbgunz joined	13:43
	→ broonie_ joined	13:43
	→ godog joined	13:44
Arjen	Manual labor?	13:44
SRabbelier	Arjen: setting the dates through export first, then commiting	13:46
rtyler	oh how manual :p	13:46
SRabbelier	very :P	13:46
	→ capisce_ joined	13:46
Arjen	SRabbelier: In a shell script?	13:46
SRabbelier	Arjen: nah, it's a one liner anyway	13:46
Arjen	GIT_COMMITTER_DATE=foo git commit-tree ... ?	13:47
	No export needed	13:47
SRabbelier	Arjen: I can still use 'git commit' :P	13:47
Arjen	I could be misunderstanding you completely, of course :-)	13:47
SRabbelier	Arjen: no you're right :)	13:47
Arjen	That's porcelain!	13:48
SRabbelier	but if I use export I only have to set the date once :P	13:48
Arjen	Heretic!	13:48
SRabbelier	Arjen: ah, yes, but this is manual labor as said	13:48
	not code ;)	13:48
Arjen	Heheh	13:48
	→ rleigh joined	13:48
	← jasam left	13:49
	→ [1]intripoon joined	13:50
	← hobbes006 left	13:51
	broonie_ → broonie	13:52
SRabbelier	Arjen: erm, is it right that GIT_AUTHOR_DATE is ignored when committing an empty tree? 0.o	13:53
	← stouset left	13:54
	→ lu_zero joined	13:56
SRabbelier	nvm, misformatted the date field :P	13:56
gass	how can i step into a new head?	13:57
	git checkout branch?	13:57
RandalSchwartz	step "into"?	13:58
	you mean, check out a branch?	13:58
	or create a new branch?	13:58
	new branch = git checkout -b newbranch oldbranch	13:58
gass	i discovered it ... git checkout origin/branch	13:59
RandalSchwartz	when you do that, you won't be able to commit	13:59
	you're only "looking" at that branch	14:00
	you can commit only to local braches	14:00
	so you just want to look at it?	14:00
	RandalSchwartz waits for gass to come back	14:01
	→ jmalonzo joined	14:02
	← capisce left	14:02
	→ capisce joined	14:02
	← l0verb0y- left	14:02
	→ markkalderon joined	14:03
	→ pombreda joined	14:04
	← eternaleye left	14:04
	→ jepler joined	14:06
	← intripoon left	14:07
	[1]intripoon → intripoon	14:07
	→ hobbes006 joined	14:07
	→ jmalonzo_ joined	14:08
jerbear	has anyone ever installed the gitweb fork that repo.or.cz uses?	14:08
	← reel_ left	14:08
pasky	any trouble there?	14:09
jerbear	pasky: are you talking to me?	14:09
RandalSchwartz	who else is talking? :)	14:10
	→ brosner joined	14:10
jerbear	just making sure :)	14:10
	yes, actually... it's mainly with getting it to work with git-browser	14:10
RandalSchwartz	[quiet room] [jerbear asks question] [pasky clarifies] [jerbear says "for me"?]	14:11
	RandalSchwartz boggles	14:11
	→ reel_ joined	14:11
jerbear	RandalSchwartz: i think we've clarified that i can be a bit dense... happy?	14:11
	rtyler gives RandalSchwartz a cookie	14:12
rtyler	now he'll be happy :D	14:12
pasky	jerbear: well, um, any trouble there? :)	14:12
	→ lawtech joined	14:13
	RandalSchwartz hands the cookie to pasky	14:13
jerbear	i've linked the git-browser dir to my web root, which it seemed to be looking for. when i go to graphiclog, it just shows the "loading" status at the top, and nothing happens	14:13
	← capisce_ left	14:14
	← deavidsedice left	14:14
	→ pormi joined	14:14
pormi	hello there	14:14
	← janm left	14:15
	→ Tv joined	14:16
jerbear	pasky: any tips for getting this working?	14:16
pormi	m..	14:17
	← pormi left	14:17
	← Catfish left	14:17
	← namenlos left	14:17
	← ph^ left	14:18
	→ Catfish joined	14:18
pasky	hmm	14:18
jast	jerbear, i did a small patch against git-browser iirc, after that it worked	14:18
pasky	jerbear: not immediately - well, check the settings, you might find it instructive to compare master and mob branches of git-browser	14:18
jast	don't recall the specifics though	14:18
gass	RandalSchwartz, here	14:19
	→ xyzzy_ joined	14:19
	→ Eduard_Munteanu joined	14:19
gass	RandalSchwartz, i want to change to that branch	14:19
	RandalSchwartz, and then edit it	14:19
jerbear	pasky: compare them? i don't understand	14:19
	compare what?	14:20
Eduard_Munteanu	Hi. I registered a repo on repo.or.cz. But I don't know how to tell git to use my SSH key to auth.	14:20
	→ bentob0x joined	14:20
Eduard_Munteanu	When pushing, it keeps asking for a password.	14:20
RandalSchwartz	gass - then you want to create a local branch	14:20
	as I said before	14:20
	and then you ignored	14:20
pasky	Eduard_Munteanu: check with regular ssh repo.or.cz first	14:20
RandalSchwartz	if you ask a question here, don't ignore the answers.	14:20
	→ jbunster joined	14:20
RandalSchwartz	git checkout -b newbranch oldbranch	14:20
gass	RandalSchwartz, sorry ... i thought i got it	14:20
jast	Eduard_Munteanu, if your key is in ~/.ssh/id_{d,r}sa it's used automatically. try ssh -vv [email@hidden.address] for lots of details.	14:20
	RandalSchwartz whacks gass with a LART stick	14:20
	← jmalonzo left	14:21
gass	RandalSchwartz, i have the branch created ... i just want to edit it	14:21
RandalSchwartz	no you don't	14:21
	you have a view on a remote branch	14:21
	that's not "created"	14:21
	you need to learn the difference between remote branches and local branches	14:21
gass	RandalSchwartz, i have to go out, be back in a minute	14:21
RandalSchwartz	see "git branch -a -v"	14:21
Eduard_Munteanu	jast, must my key have the comment in the form of [email@hidden.address] or it can be user@localhost?	14:21
gass	RandalSchwartz, so i have to check it out and re-create it?	14:21
RandalSchwartz	eh?	14:22
jast	Eduard_Munteanu, the comment can be anything as long as it's there	14:22
RandalSchwartz	you need to do what I said	14:22
jast	(perhaps not strictly anything but foo@bar definitely works)	14:22
RandalSchwartz	git checkout -b newbranch oldbranch	14:22
gass	RandalSchwartz, how can i edit the remote branch?	14:22
RandalSchwartz	how many times should I type that for you?	14:22
gass	ok, ok	14:22
	done	14:22
RandalSchwartz	you don't edit a remote branch	14:22
	you commit to a local branch	14:22
	have you been through the tutorial?	14:22
	probably a good time to do that now	14:22
	← markkalderon left	14:24
jerbear	pasky: what is the proper way to install this version of gitweb?	14:25
Ilari	gass: Remote branches form mirror(s) of other repository(ies). You don't want to edit those mirrors...	14:25
	→ gitte joined	14:25
	→ stouset joined	14:26
pasky	jerbear: just clone and build it :) but you seem to be confused about gitweb vs git-browser	14:27
	→ armedblade joined	14:27
pasky	i already gave my hints about git-browser and have nothing much to add	14:27
	→ ciaran29d joined	14:27
	← venger left	14:28
	armedblade → venger	14:28
jerbear	pasky: yes, git-browser has to be cloned separately from gitweb, right? there are no instructions as to how to integrate the two.... only how to set it up standalone	14:28
	→ jossipek joined	14:29
pasky	well, repo.or.cz's gitweb has it integrated :)	14:29
	→ ph^ joined	14:30
	← Modius_ left	14:30
jerbear	pasky: that what i cloned.. it didn't appear to be	14:30
	→ Modius_ joined	14:30
Eduard_Munteanu	jast, it says "debug2: we sent a publickey packet, wait for reply", but then goes on asking for a password.	14:30
	→ janm joined	14:31
jast	Eduard_Munteanu, directly after that debug2 message?	14:31
Eduard_Munteanu	jast, no, it goes on checking for DSA keys.	14:32
RandalSchwartz	are you using openssh, or that weird one that comes with solaris?	14:32
Eduard_Munteanu	RandalSchwartz, openssh	14:32
RandalSchwartz	ok - nevermind then	14:32
	I had trouble once because I thought I was using openssh	14:32
	and it was some freaked out old code	14:32
Eduard_Munteanu	jast, http://pastebin.com/d5f8e20c1	14:35
	← njal left	14:35
	→ [1]intripoon joined	14:37
	← jerbear left	14:37
	→ piclez_ joined	14:37
	← brosner left	14:38
Ilari	Eduard_Munteanu: What files are in '~/.ssh' besides config and authorized_keys?	14:39
pilif	Eduard_Munteanu: could it be that you are using one of these insecure private keys as generated by the defect openssl package?	14:39
Ilari	Eduard_Munteanu: Also, what distro?	14:39
pilif	they are often blacklisted	14:39
Eduard_Munteanu	Ilari, there are just these: id_rsa id_rsa.pub known_hosts	14:39
	Ilari, Gentoo. And I know about the Debian problem, I think it does not affect Gentoo.	14:40
pilif	it depends what the key was initially generated with	14:40
jast	that leaves another easy explanation: maybe you pasted the public key incorrectly on your repo account, perhaps with superfluous line breaks or something. you could try updating your public key on repo.	14:40
	← sgrimm left	14:40
Eduard_Munteanu	openssh-4.7_p1-r1 and openssl-0.9.8e-r3	14:41
Pieter	you can edit your keys now?	14:41
jast	yes	14:41
Eduard_Munteanu	Okay, I will insert the key again into the webapp.	14:41
jast	my patch got applied upstream \o/	14:41
Pieter	good :)	14:42
jast	yeah... i wrote the patch mainly for my own good ;)	14:42
	i was affected by that debian madness too	14:42
Pieter	madness? THIS IS SPARTA!	14:43
jast	sure :)	14:43
rtyler	git push into-pit	14:43
Eduard_Munteanu	Okay, will copy with cat ~/.ssh/id_rsa.pub \| xclip -i -selection clipboard and paste normally	14:44
	(this is what I use to paste patches into my email client)	14:44
Ilari	Eduard_Munteanu: '-selection primary' and middle-click the textbox (unless browser is buggy)? :-)	14:45
RandalSchwartz	useless use of cat!	14:45
	← jmalonzo_ left	14:45
Eduard_Munteanu	RandalSchwartz, yeah, you're right.	14:45
RandalSchwartz	xclip -i < ~/.ssh/id_rsa.pub	14:45
	I still wonder where people start learning to type that useless cat	14:45
Eduard_Munteanu	Ilari, I prefer using paste, but...	14:45
RandalSchwartz	maybe it's a windows thing	14:46
pilif	Eduard_Munteanu: what does the content of id_rsa.pub begin with? For authorized_keys it usually conists of one line beginning with ssh-rsa. I've seen public keys in the x509 text format one (--- BEGIN PUBLIC KEY---) which would not work.	14:46
	s/one/once	14:46
jengelh	use DSA keys	14:46
jast	rsa > dsa	14:46
Eduard_Munteanu	pilif, ssh-rsa stuff. Though the public key is x509-like.	14:46
	pilif, sorry, private key	14:46
jengelh	but RSA is encumbered :)	14:46
Mikachu	RandalSchwartz: i think you can even do < ~/.ssh/id_rsa.pub xclip -i	14:47
pilif	Eduard_Munteanu: that's ok. It's just the format of the public key	14:47
RandalSchwartz	in good shells, yes	14:47
Mikachu	RandalSchwartz: if you really want to have the filename to the left	14:47
RandalSchwartz	but I've seen broken shells that wouldn't do that	14:47
Eduard_Munteanu	jengelh, doesn't RSA work with repo.or.cz.	14:47
RandalSchwartz	so I no longer suggest that	14:47
Mikachu	ah	14:47
RandalSchwartz	some variants of bash, I think	14:47
	not being a bash user, I woudln't know	14:47
	I try to avoid most gnu abominations	14:47
jast	rsa works with repo, no worries	14:47
Mikachu	what shell do you use?	14:48
RandalSchwartz	zsh	14:48
Mikachu	aha, good :)	14:48
RandalSchwartz	only because it lets me run something on ssh	14:48
	pilif loves his zsh	14:48
RandalSchwartz	but I don't understand even 1/10th of zsh	14:48
	it's far too complex for a shell	14:48
	I need "ssh remotehost somecommand" to run stuff on remotehost	14:49
	csh and zsh are the only two I found to do that	14:49
	and I really really hate csh	14:49
	→ tro joined	14:49
Mikachu	the globbing syntax is so good i don't even know how to use find :)	14:49
Eduard_Munteanu	jast, nah, it does the same thing.	14:49
pilif	it's complex, but it's quite probable that the feature would want is in zsh	14:49
RandalSchwartz	so I was happy to do zsh	14:49
Eduard_Munteanu	jast, still asks for a password.	14:49
redondos	RandalSchwartz: example?	14:49
Eduard_Munteanu	Should I update my openssh & openssl?	14:49
	→ ceez joined	14:49
RandalSchwartz	example for hating csh?	14:49
	→ ironfroggy joined	14:50
redondos	redondos@thruster:~$ ssh nceurope.dyndns.org hostname	14:50
	nc-server	14:50
	just an example	14:50
ironfroggy	I'm tryin to determine what commit a particular change in a diff between two branches came from. is there a good way to determine this?	14:50
redondos	I am also a zsh user, but I don't get your point	14:50
RandalSchwartz	oh - run my per-host stuff before running the actual command	14:50
	such as updating path or setting envars	14:50
	since that's not considered a login shell, most classic shells won't run anything	14:51
	anyway... gotta drive.	14:51
redondos	bashrc is not read?	14:51
	RandalSchwartz runs off to work	14:51
redondos	yeah, anyway.	14:51
pilif	well... there's always .bashrc vs. .profile	14:52
RandalSchwartz	see my comment above about bash	14:52
	and gnu	14:52
pilif	like .zshrc vs. .profile	14:52
RandalSchwartz	I won't touch bash with a 100 foot pole	14:52
redondos	yes, ok, bash sucks.	14:52
RandalSchwartz	it sucks in the same way that csh sucks	14:52
pilif	actually, I'm inclined to agree on that one.	14:52
RandalSchwartz	designed by people who didn't understand that DWIM escape sequences are bad	14:52
	but that's another story	14:52
	RandalSchwartz wanders off for real	14:53
ironfroggy	can i search commits by the contents of the diffs?	14:53
	maybe that will help me narrow down this change	14:53
	→ njal joined	14:53
	← intripoon left	14:54
	→ CodeOfficer joined	14:54
bentob0x	what is the best, 'cleanest' way to use git on Windows? msysgit or cygwin?	14:54
pilif	ironfroggy: I don't know about that, but there's always git blame	14:54
jengelh	probably cygwin because it provides more nix semantics	14:54
	→ [2]intripoon joined	14:54
jast	the only downside with cygwin is that it pretends all files are executable	14:55
	→ bdiego joined	14:55
jast	so you should set core.filemode to false	14:55
	→ rdavila joined	14:56
	→ naeu joined	14:57
	→ intripoon joined	14:57
	stouset → stouset\|away	14:58
	Arme[0] → Arme[N]	14:59
	← pygi left	14:59
	stouset\|away → stouset	14:59
	← Kbyte left	15:01
	← CodeOfficer left	15:04
	← xyzzy_ left	15:04
	← Catfish left	15:04
	← Tv left	15:04
	← lu_zero left	15:04
	← Ademan left	15:04
	← BabelO left	15:04
	← adante left	15:04
	← cnk left	15:04
	← brendon_ left	15:04
	← mattf left	15:04
	← sscj left	15:04
	← Beket left	15:05
Mikachu	ironfroggy: git log -Sword	15:05
	heh, "word" is the word you want to search for there, that is	15:06
	← GregersP left	15:07
	→ CodeOfficer joined	15:07
	→ xyzzy_ joined	15:07
	→ Catfish joined	15:07
	→ Tv joined	15:07
	→ lu_zero joined	15:07
	→ Ademan joined	15:07
	→ BabelO joined	15:07
	→ adante joined	15:07
	→ sscj joined	15:07
	→ cnk joined	15:07
	→ brendon_ joined	15:07
	→ mattf joined	15:07
	→ cupsh joined	15:07
	→ afton joined	15:07
	→ jmspeex joined	15:07
	→ _bob_ joined	15:07
	→ bronson joined	15:07
	← [1]intripoon left	15:08
agib	does a tracking branch work for push operations too? I basically want to keep 2 branches on my local repository and github... (a master branch for deployment of my production rails application and a staging branch for deployment to my staging server). Does this make any sense?	15:08
	← pergesu left	15:08
mattf	afaik, tracking information isn't taken into account when pushing	15:09
vbabiy-laptop	Hey guys if want to reset a file to back to what it was a few commits ago how would i do that	15:09
Mikachu	vbabiy-laptop: git checkout <a few commits ago> -- thefile	15:09
mattf	if the local and remote branches have the same name push just works, if they don't you need to push to a specific remote ref	15:09
	agib, (above)	15:10
vbabiy-laptop	Mikachu: and what if i want to bring that forward to the current master	15:10
Mikachu	vbabiy-laptop: that's what i said how to do	15:10
gass	RandalSchwartz, i'm back :)	15:10
mattf	vbabiy-laptop, <a few commits ago> == HEAD^^^, a couple ago == HEAD^^	15:11
pilif	gass, but by now, he is gone	15:11
agib	mattf: how can I create a branch on a remote repository? (github)	15:11
mattf	agib, just push to it	15:11
gass	pilif, oh ..	15:11
	→ brosner joined	15:11
pilif	gass: he was pretty clear in the instrucations he gave though. Just do what he said and you'll be fine	15:11
	← ironfroggy left	15:12
pilif	s/instrucations/instructions/	15:12
gass	pilif, lol ... short time git user here. ...	15:12
	→ madewokherd joined	15:12
agib	mattf: git push origin staging_branch ?	15:12
	→ hector joined	15:12
agib	ahh ok	15:13
	thank you	15:13
mattf	agib, if you want to push your current branch, yeah	15:13
	→ [1]intripoon joined	15:13
	← [2]intripoon left	15:14
	→ tvachon joined	15:14
	→ [2]intripoon joined	15:14
gass	pilif, so i got it ... i think ... i do checkout -b local_branch origin/remote_branch	15:14
	pilif, and give the sabe name	15:14
	→ bobesponja joined	15:15
	← piclez_ left	15:15
	→ Cacheaway joined	15:15
gass	sabe = same	15:15
	→ piclez_ joined	15:15
SRabbelier	Well, it friggin took me 2 hours, but now I do have a super-sweet http://repo.or.cz/git-browser/by-commit.html?r=git-stats.git !	15:17
	← intripoon left	15:18
	← Sho_ left	15:18
	→ ebel_ joined	15:18
	← ebel left	15:20
	ebel_ → ebel	15:20
	→ intripoon joined	15:20
	→ pombreda1 joined	15:22
	→ markkalderon joined	15:22
	→ kumbayo joined	15:24
	→ Sho_ joined	15:25
	← venger left	15:25
Cacheaway	is there a way to undo a merge?	15:26
SRabbelier	Cacheaway: "git reset --hard HEAD^" will clobber your checkout and set it to before the merge	15:26
	→ jpeeler joined	15:26
SRabbelier	Cacheaway: assuming the merge is the last thing you did	15:26
Eduard_Munteanu	Cacheaway, a merge is marked by a commit, so you can use git-reset	15:26
Cacheaway	SRabbelier: thanks a lot	15:27
SRabbelier	Cacheaway: yw :)	15:27
Cacheaway	I was on the right track, just forgot about the ^ notation	15:27
SRabbelier	Cacheaway: ah, you couldof also used the reflog	15:27
	→ queso joined	15:29
Mikachu	Cacheaway: if the merge conflicted and you want to abort it, use just git reset --hard, no HEAD^	15:30
SRabbelier	Cacheaway: replying to what Mikachu said: that is because if there is a conflict no commit has been made yet	15:30
	← charon left	15:31
Eduard_Munteanu	How can I undo a git-push?	15:31
	→ venger joined	15:31
	← [1]intripoon left	15:32
	→ gza joined	15:32
gza	morning	15:32
Mikachu	Eduard_Munteanu: git push remote anysha1:refname	15:32
	← lea_w left	15:32
Eduard_Munteanu	Mikachu, well, the repo was empty before that, so can I use 00000 as the dest ref?	15:33
venger	if it want to pull all commits from "stage" (a repo based on master) into "master" would I run "git pull stage master"?	15:33
SRabbelier	Eduard_Munteanu: you can delete the branch	15:33
Mikachu	Eduard_Munteanu: to delete a branch, git push remote :branch	15:33
Eduard_Munteanu	Oh, thanks to both of you.	15:33
	← [2]intripoon left	15:33
SRabbelier	yw	15:33
	Mikachu: nice teamwork there :P	15:33
	← intripoon left	15:35
	← pombreda left	15:35
venger	or could I run "git merge stage" while master is the current HEAD?	15:38
vmiklos	venger: git checkout master; git merge strage	15:38
	exactly.	15:38
venger	ok thanks	15:38
	→ johan-s joined	15:38
	← ruphy left	15:38
gza	Anybody here using gitosis (or something else?) for an intranet?	15:38
SRabbelier	gza: probably a bunch of people, but it's best to just ask your question ;)	15:39
	→ bugfux joined	15:39
venger	vmiklos, so is it that git pull is preferred for pulling remote branch commits?	15:40
gza	Any links to gitosis sites?	15:40
	← JonathanT left	15:40
	← tvachon left	15:40
vmiklos	venger: pull is fetch + merge	15:40
	→ JonathanT joined	15:41
venger	aye	15:41
vmiklos	if you merge a local branch, then fetch is a no-op, so it's possible but isn't too logical	15:41
	iirc 'git pull . stage' would still work..	15:41
venger	gza: I run it. http://weblog.techno-weenie.net/2007/12/25/moving-to-git, http://scie.nti.st/2007/11/14/hosting-git-repositories-the-easy-and-secure-way,	15:41
	→ mando joined	15:42
	← Eduard_Munteanu left	15:42
biz	using gitosis, is there a way to split the 'writable =' line in the config into multiple lines? with lots of projects a single line is quite unclear on a short look	15:43
Mikachu	biz: did you try if it works already?	15:44
biz	Mikachu: No.	15:44
	→ chrisb joined	15:45
gza	venger: cool thanks...putting together a paper to argue moving to Git	15:47
	github, gitosis is one of the points	15:48
venger	well are you going to put it into practice first? one cool things is msys (MINGW32) on windows clients works quite nicely for me against my remote linux repo	15:49
gza	I hope we can try out git with a few projects and possibly use git-svn	15:50
	→ championdusk joined	15:53
	← hobbes006 left	15:53
championdusk	hello, how can I do a git pull and force all overwrites?	15:54
gza	most are on mac...but we have a few windows users...and yes that tool is point as well	15:54
vmiklos	git fetch origin; git reset --hard origin/master?	15:54
gza	venger: here's the document so far....thoughts? http://pastie.org/212286	15:54
vmiklos	or something like that	15:54
	← pombreda1 left	15:56
biz	Mikachu: Seems like simple multipe 'writable = foo' lines don't work	15:56
Mikachu	okay	15:56
venger	gza have you watched the Linus git youtube video? just add his points on how its different and why its better because its "distributed"	15:57
	← redondos left	15:57
gza	well I'm not really arguing SVN vs. Git....I have to argue Git vs. Mercurial	15:57
vmiklos	haha	15:58
	that's harder :)	15:58
gza	both are distributed...but I think there's more development and adoption of Git	15:58
	which should mean better docs, more productivity tools etc...	15:58
	→ krawek joined	15:59
	→ digitalhobbit joined	16:00
	← kumbayo left	16:00
	→ cannonball joined	16:02
gza	I think I have an error in that doc...	16:04
	← FunkeeMonk left	16:04
	→ pombreda joined	16:05
gza	I called gitosis an open source github like system...	16:06
	I think gitorious.org may be something more in line with that?	16:06
	→ [1]intripoon joined	16:06
	← mcella left	16:07
	← EmilMedve left	16:07
	→ EmilMedve joined	16:07
	→ tvachon joined	16:07
	→ ruphy joined	16:07
	→ fhobia joined	16:09
	→ hagna joined	16:13
hagna	why doesn't git use pack files by default? I have to repack and prune manually?	16:13
	→ alley_cat joined	16:14
	queso → queso\|away	16:14
	→ [2]intripoon joined	16:15
	→ k0001 joined	16:21
	→ RaceCondition joined	16:22
	← k0001 left	16:23
jbunster	hagna: no. It packs from time to time.	16:23
	→ k0001 joined	16:23
	→ warlock_mza joined	16:23
	→ d0k_ joined	16:24
dato	there is nothing like bzr shelve for git, is there? (is like an interactive stash, selecting what to stash away)	16:25
	← priidu left	16:25
hagna	jbunster: oh I didn't see that here http://www.kernel.org/pub/software/scm/git/docs/user-manual.html#git-concepts	16:27
	jbunster: good to know	16:27
	→ priidu joined	16:29
	← away left	16:30
	← Arme[N] left	16:31
	← piclez_ left	16:32
	→ piclez_ joined	16:32
	← d0k left	16:33
	← hagna left	16:36
	→ hobbes006 joined	16:37
	← [2]intripoon left	16:37
	→ timocratic joined	16:38
	← nud left	16:39
	← [1]intripoon left	16:40
	→ mithro joined	16:40
	← madewokherd left	16:40
	→ madewokherd joined	16:40
	→ ngng joined	16:41
	← pombreda left	16:42
	← naeu left	16:42
	→ bronson_ joined	16:43
	Fullmoon → Misha	16:43
	→ rubydiamond joined	16:43
	queso\|away → queso	16:44
	← reel_ left	16:46
	← koke left	16:48
	→ jeckel joined	16:48
	→ lea_w joined	16:49
	→ eternaleye joined	16:49
	→ igh joined	16:50
	→ EmilMedve1 joined	16:50
	→ gberdyshev joined	16:51
	← fhobia left	16:52
	← EmilMedve left	16:52
	EmilMedve1 → EmilMedve	16:52
holizz	How can I get hold of the summary that git push displays? i.e. the "931cfcb..bbd347d" part (from a post-update script). Or at least a way to convert a sha1 to an unambiguous partial sha1.	16:56
	← igh left	16:57
	← Misha left	16:59
	← _zack left	16:59
	← ebel left	17:00
	← cedricv left	17:02
	← warlock_mza left	17:03
	← bugfux left	17:05
	→ charon joined	17:05
	→ elliottcable joined	17:06
	← elliottcable left	17:06
	→ elliottcable joined	17:06
	→ cedricv joined	17:07
	← cedricv left	17:07
	→ l0verb0y joined	17:09
	← RaceCondition left	17:11
	← JonathanT left	17:12
	← mithro left	17:13
	→ docwhat joined	17:13
docwhat	Hello!	17:13
	queso → queso\|away	17:13
	→ eternaleye_ joined	17:13
docwhat	Is there a way to get the remote repository a branch (specifically the current branch) is tracking?	17:13
	→ hagna joined	17:15
	← eternaleye left	17:15
	← felixflores left	17:17
	← panagos left	17:18
	→ RaceCondition joined	17:20
	→ _graham_ joined	17:20
	← elliottcable left	17:23
	→ elliottcable joined	17:24
	→ felixflores joined	17:25
	← elliottcable left	17:25
	→ elliottcable joined	17:25
	← lea_w left	17:25
	→ koke joined	17:26
	← deavid left	17:26
	stouset → stouset\|away	17:26
	← elliottcable left	17:27
	→ elliottcable joined	17:27
	→ ropiku joined	17:28
ropiku	Hello, how can you manually resolve merge conflicts ?	17:28
Arjen	ropiku: Find the conflict markers ('>>>>>>>>>' and '<<<<<<<<<<'), fix it, git add the file, done	17:29
	docwhat: git remote show origin ?	17:29
	← vbgunz left	17:29
docwhat	no...	17:30
ropiku	Arjen, thanks, i'll try	17:30
	→ robinr joined	17:30
docwhat	Arjen: I'm looking for something that returns "origin/master" from the branch "master"	17:31
	oh wait, I typed your command wrong.	17:31
	Arjen: Yeah, that's close...The info could be gotten from there, though it's an ugly and less-robust-than-I'd-like parse.	17:32
	ruphy → [[	17:32
	[[ → ruphy	17:32
Arjen	docwhat: It was just off the top of my head, maybe there's some plumbing command	17:34
docwhat	I didn't mean to disparage; thank you for the help. :-)	17:35
	queso\|away → queso	17:35
	→ simmel joined	17:35
ropiku	And how can you fix a 'entry not uptodate' ? It doesn't add >> markers ?	17:38
	→ Fullmoon joined	17:38
	→ joevandyk_ joined	17:39
	→ thiago_home joined	17:39
	← elliottcable left	17:39
	→ elliottcable joined	17:40
	stouset\|away → stouset	17:40
	→ joeyb joined	17:40
Ilari	ropiku: What operation gave those 'entry not uptodate' things? Merge?	17:41
ropiku	Ilari, git pull (and git merge remotes/origin/master)	17:41
Ilari	ropiku: Pull/Merge with local modifications in tree is not recommended.	17:42
	→ Beket joined	17:42
Ilari	ropiku: ... Uncommitted modifications, that is.	17:42
ropiku	Ilari, thanks	17:43
	→ _zack joined	17:43
	→ vbgunz joined	17:44
Ilari	ropiku: AFAIK, with merge (including merging pull), it is not always an error.	17:44
joeyb	Has anybody seen issues where you add a submodule to a project that already has submodules and other folks who are doing fetches/pulls don't get the new submodule (despite git submodule init/update's being run) but people who do fresh clones do (get all the submodules)?	17:44
	(Sorry for barging in with a question straight away! :))	17:44
Ilari	ropiku: If you have local modificications, there is no easy way to then back off the merge without losing those modifications.	17:44
Tv	joeyb: you do know that it isn't automatic, i hope..	17:45
ropiku	Ilari, thanks, i'll commit then merge	17:45
gza	What are the names of the native Java and Ruby apis for Git?	17:45
Ilari	ropiku: Also see rebase (which IIRC is strict about local changes). It gives cleaner history (or causes complaints from others, depending on use.) :-)	17:46
	gza: AFAIK, Java is Jgit. And there are two Ruby ones...	17:47
	← joeyb left	17:47
	→ jaybee_ joined	17:47
jaybee_	Tv: thank you for your response! "automatic" in what sense? shouldn't a git submodule init + git submodule update trigger pulling down both updates to existing submodules' HEADs and drawing down any ones added (to the repo that you're pulling from)?	17:48
	jaybee_ → joeyb	17:48
Tv	joeyb: yeah i'm just making sure they understood to run those after merge	17:48
joeyb	Tv: yep! it's strange because clones have the new submodule without issue. but people who are fetching/pulling and doing the init/update routine have the .gitmodules file updated ... but the base dir for the new submodule is not created and of course nothing heads inside on a submodule update call...	17:50
	→ nud joined	17:50
	→ kewark joined	17:50
Tv	joeyb: sounds like they didn't run init+update	17:50
	joeyb: reproduce with a minimal example if you can	17:50
	→ bugfux joined	17:51
Ilari	Isn't that directory created by merge (since there's submodule entry now)?	17:51
	← krawek left	17:51
	→ entie joined	17:51
Tv	bah 2/3 of my usb ports flaked out :(	17:51
joeyb	Tv: I've confirmed that they did run them. Will do! And I'll scour their .git dirs to see if I can find the stopping point. Thank you for your help, just wanted to make sure I wasn't missing the point entirely! :)	17:51
	→ Eridius\|work joined	17:52
	→ pombreda joined	17:52
Ilari	Tv: What errors they produce?	17:52
gza	Ilari: JGit is a eclipse plugin?	17:52
joeyb	Ilari: I'd think so, but it's not there on a merge/pull and it's considered a new dir if you mkdir it manually. I'll dig in and report back if I find anything interesting...	17:52
Tv	[12658.309629] usb 7-1: new high speed USB device using ehci_hcd and address 6	17:52
	[12673.843061] usb 7-1: device not accepting address 6, error -110	17:52
	← joeyb left	17:53
Tv	Ilari: repeated 4 times with increasing addresses	17:53
	Ilari: ehci flaked, uhci works	17:53
	→ josb joined	17:53
Tv	Ilari: (right side 2 ports vs left side 1 port, thinkpad t61)	17:53
Ilari	Error -110... Hmm...	17:53
Mikachu	that happened a lot to me when i played with rockbox and kept (dis)connecting my ipod	17:53
Tv	$ errno 110	17:53
	ETIMEDOUT	17:53
Mikachu	eventually i'd run out of ports and have to reboot :)	17:53
Tv	i'm plugging in a usb hub..	17:54
Ilari	Hmm... And ECHI and USB support isn't a module one could reload? :-/	17:54
Tv	Ilari: it is but i don't want to risk oops right now	17:54
Mikachu	if you wanted to risk it, i would try reconnecting the hub	17:55
	but it might be a bad idea	17:55
Tv	oh i've been plugging things on/off the bad 2 ports	17:55
	just that timeout thing	17:55
	→ modoc joined	17:55
Tv	and i'm actively using the good port	17:55
jkp	how can i see the outstanding conflicts in an open rebase?	17:58
Tv	git status	17:58
	→ pygi joined	17:58
jkp	Tv: doesnt show me anything	17:58
	← elliottcable left	17:58
Tv	jkp: then you ain't got any	17:58
jkp	if i do a git rebase --continue...it tells me "You must edit all merge conflicts and then	17:58
Ilari	How badly oops screws the kernel varies. It can range from no apparent effect besides program crashing to whole machine locking up straight away...	17:58
jkp	"	17:58
Tv	hmm	17:58
	→ elliottcable joined	17:58
	← _graham_ left	17:59
jkp	Tv: shows lots of modified files in status, and i added the files i had to manually resolve	17:59
	but i cant see whats left to do	17:59
	← venger left	17:59
Tv	jkp: honestly, i usually use an emacs git mode that makes it look pretty ;)	17:59
	← Ademan left	17:59
Ilari	jkp: In what section the modified entries are?	18:00
jkp	Ilari: i dont follow	18:00
	they are all in "Changes to be committed"	18:00
	everything is green, ive added them all	18:01
gza	hmmm....can't find Java Git Api...google not finding anything	18:01
jkp	i shouldnt have to manually commit this lot should it?	18:01
	*should i	18:01
Ilari	jkp: 'git ls-files --unmerged' shows nothing?	18:01
	→ joeyb joined	18:01
	← xaxxon left	18:02
	→ ricky joined	18:02
jkp	Ilari: trying it again from the top	18:02
	← elliottcable left	18:03
	→ elliottcable joined	18:03
jkp	worked this time	18:05
	← elliottcable left	18:06
	→ elliottcable joined	18:06
	Debolaz tries various creative methods to pack this impossible to pack repo.	18:06
	→ eean joined	18:07
eean	How can I have a conflict when my local version doesn't have any changes?	18:07
	→ zhynn joined	18:07
	elliottcable → ec	18:07
Ilari	Debolaz: How it is 'impossible to pack'?	18:08
	Debolaz: Contains "exotic objects"?	18:08
Debolaz	Ilari: Huge repo, only 512MB ram on the machine.	18:08
	← rlb3_ left	18:08
Tv	Debolaz: i have this alias: repack-nice = !nice -19 ionice -n 7 git repack -d --window-memory=20m --max-pack-size=200	18:08
	← lu_zero left	18:09
Tv	Debolaz: very unscientific but seems to help	18:09
Debolaz	With git-repack alone, it only counts 10 objects per second.	18:09
	Way too slow to be useful.	18:10
Tv	Debolaz: huge amounts of loose objects on a non-hashed filesystem?	18:10
Debolaz	Huge amounts of loose objects, but the filesystem (UFS2) has hashing. And I was able to tar the entire repo in just three hours, it's only git that's slow.	18:11
	1.5m objects to be more specific.	18:11
modoc	I've pulled a repo from a coworker but am having trouble with a submodule.	18:11
Tv	you can also feed a list of objects to git pack-objects, to make it stop trying to find optimal order	18:12
	way less efficient packs but gets you going	18:12
modoc	the dir shows up but doesn't have a .git dir in it	18:12
zhynn	anyone have a few moments to help a git noob?	18:12
Debolaz	Tv: That's what I'm trying now.	18:12
Tv	zhynn: don't ask about asking, just ask	18:12
modoc	now git pull says 'cannot read object' ... 'it is a submodule'	18:12
	What can I try to fix this?	18:12
Debolaz	Tv: Doing it separately for 0, 1, 2*, etc, to ensure I don't get memory issues.	18:12
zhynn	Tv: I find I get better results when asking first. :)	18:12
Ilari	zhynn: Unless that request gets ETIMEDOUT... :-)	18:13
Tv	zhynn: generates unnecessary noise on this channel, and people might not even bother to encourage you to ask your real question	18:13
zhynn	I have done a git clone repo.git. then i do a git checkout v1.0 (a previous version). this seems to change the head reference, but how do I actually get the files of that version?	18:13
Tv	zhynn: also encourages mentality where your future questions are aimed at the one person who responded, instead of the whole channel	18:14
Debolaz	I really wish pack-objects were more verbose though. After it's done counting, there's no progress bar or explanation for what's happening for a long time afterwards.	18:14
	→ thannoy joined	18:14
	← josb left	18:14
Tv	zhynn: git checkout will change the working dir contents to match what you requested	18:14
zhynn	Tv: yes, that's true.	18:14
	→ deavid joined	18:14
Tv	Debolaz: i find it's usually statting the loose objects, at that point	18:14
zhynn	Tv: is there any output to reflect that other than the "HEAD is now at..." message?	18:14
Tv	Debolaz: strace is your friend (on linux)	18:14
Debolaz	Tv: Yes. I just want a progress bar.	18:15
Tv	zhynn: unix philosophy - tools are silent unless they fail	18:15
Debolaz	Tv: Some indication of how fast it's going.	18:15
	Tv: Also, ktrace > strace :-)	18:15
	strace needs that pesky procfs on most systems.	18:15
Tv	Debolaz: pack-objects has --progress and --all-progress	18:16
	Debolaz: if those don't do what you want, then you need to add that feature ;)	18:16
zhynn	Tv: yes, though it requires that you know all of the things that the tool is intended to do! I didn't know if checkout updated the reference, updated the working directory files, or both. The output told me one but not the other.	18:16
Debolaz	Tv: --progress and --all-progrerss doesn't mean what one would think they meant.	18:16
Tv	Debolaz: so you need to add it..	18:17
	→ drizzd_ joined	18:17
Tv	zhynn: "When <paths> are not given, this command switches branches by updating the index and working tree to reflect the specified branch"	18:17
	...	18:17
Debolaz	Tv: Hehe, maybe I should. I'm already running a custom svn, might as well throw a few more modifications on it.	18:18
	← RaceCondition left	18:18
Debolaz	Err..	18:19
	custom git	18:19
	Bah, hot in here. :-P	18:19
	Or even add a flag to disable the stat calls.	18:20
	→ lea_w joined	18:21
	← lea_w left	18:21
zhynn	Tv: Thanks for the help, i appreciate it. I'll get familiar with it, and it will make more sense to me. ;)	18:22
	→ lea_w joined	18:23
	← bentob0x left	18:24
	→ ehamberg joined	18:26
	→ daaku joined	18:27
	→ gberdyshev_ joined	18:30
	← ropiku left	18:30
	← drizzd left	18:32
	← _zack left	18:33
	← nud left	18:34
	← zhynn left	18:34
	→ sbeyer joined	18:34
	→ redondos joined	18:35
	→ _zack joined	18:35
championdusk	To [email@hidden.address] ! [rejected] master -> master (non-fast forward)	18:37
	error: failed to push to 'git@courtroomconnect.unfuddle.com:courtroomconnect/remotecounsel.git'	18:37
	wtf	18:37
	i used to be able to pull/push, now I can only pull	18:37
gitte	non-fast forward.	18:37
championdusk	ya, wtF does that mean	18:38
gitte	Typically means you have to pull first.	18:38
	Into your master.	18:38
championdusk	er, how can I force it	18:38
gitte	Into your master?	18:38
championdusk	ya	18:38
gitte	You need to know what branch you are on.	18:38
championdusk	I want the copy I have on my machine to be the new hotness	18:38
	master	18:38
gitte	And a pull succeeds?	18:38
championdusk	it did	18:38
	i don't want to pull again	18:38
gitte	And another pull says it is up-to-date?	18:39
	Probably not.	18:39
	← lea_w left	18:39
championdusk	oh	18:39
	there we go	18:39
	--force	18:39
gitte	Well, if you want to do bad things, go ahead.	18:39
championdusk	that's a bad thing?	18:40
gitte	But don't ask me to help you, then.	18:40
	Of course it is.	18:40
Tv	championdusk: you're overwriting work by others	18:40
gitte	Because you do not understand what you are doing there.	18:40
championdusk	i'm the only one working on this	18:40
Tv	doesn't look like it	18:40
gitte	You would lose what is on the remote and not local.	18:40
championdusk	ok god	18:40
Tv	championdusk: perhaps you rebase with reckless abandon?	18:40
championdusk	rebase?	18:41
Tv	i guess not, then	18:41
championdusk	quite	18:41
	hold on, I am going to need my monocle and top hat to continue this conversation	18:41
	← eikonos left	18:41
gitte	championdusk: why don't you just look what the branches look like?	18:43
	I mean, it is just a matter of "git fetch origin master && gitk FETCH_HEAD master"	18:43
	Then you should see exactly what is happening.	18:43
	And then you should have the perfect precondition to form a decision as to what to do next.	18:43
	→ p4tux joined	18:44
	← gberdyshev left	18:44
	→ RaceCondition joined	18:44
	← RobotDeathSquad left	18:44
championdusk	let me give her a whirl	18:45
	→ david_koontz joined	18:48
	→ entie_ joined	18:48
	← bugfux left	18:53
	← unreal left	19:01
	→ unreal joined	19:01
	→ eikonos joined	19:04
	← capisce left	19:04
	← entie left	19:05
	→ nud joined	19:06
	→ arun joined	19:07
	← felixflores left	19:09
	← bronson_ left	19:09
	→ felixflores joined	19:10
	→ peeja joined	19:10
	← daaku left	19:11
	← jbunster left	19:11
	← Fullmoon left	19:16
	→ lu_zero joined	19:17
	→ nadim joined	19:19
	← Xyrio left	19:20
	→ jkp_ joined	19:22
	← priidu left	19:22
	← arun left	19:27
	→ capisce joined	19:28
	→ lea_w joined	19:29
	← rubydiamond left	19:29
	→ JonathanT joined	19:29
lea_w	How can I create a tag object that points to another arbitrary object, not just a commit? (for testing purposes)	19:30
Ilari	lea_w: Doesn't git tag create such things (if you specify object to tag explicitly)?	19:31
Mikachu	git tag `git hash-object -w myfile` for example works fine	19:32
	er, with a tagname in there too	19:32
	but you can give it any sha1 you want	19:32
	→ blahed joined	19:33
lea_w	Right, thanks!	19:33
blahed	i am having an awful time trying to get a remote repo setup on leopard server	19:33
	anyone got a sec?	19:33
Ilari	blahed: Go ahead.	19:34
	→ arun joined	19:34
blahed	fairly sure i have everything setup the way it needs to be	19:34
Ilari	blahed: You used 'git --bare init', not 'git init', right?	19:34
blahed	i have a git user with git-shell	19:34
	Ilari: i used git init initially on a client computer and decided to move it to the server	19:35
	should i have just pushed it?	19:35
Ilari	blahed: You only copied .git part of it?	19:35
blahed	i just moved the entire folder	19:35
	scp	19:35
	← joeyb left	19:35
	← jkp left	19:36
blahed	the entire project folder to be more specfic	19:36
	← peeja left	19:36
Ilari	blahed: If it still has working tree attached, take the .git out of it 'mv foo/.git foo.git' and then delete the working tree on server (rm -r foo)...	19:36
	← vbgunz left	19:37
blahed	Ilari: now run git --bare init ?	19:38
Ilari	blahed: If you taken that .git of it, you already have a repository.	19:38
Mikachu	i think you have to edit the bare variable in config too	19:38
blahed	Ilari: i took the .git out of it	19:38
Ilari	No real need...	19:38
	→ darkmod joined	19:39
	← lea_w left	19:39
darkmod	when i do 'git submodule update' it seems to checkout HEAD^ not HEAD -- any suggestions	19:40
Ilari	blahed: Then there should be repository on server (likely set up for single user only)...	19:40
	← Sho_ left	19:40
Ilari	darkmod: It should use the revision recorded in supermodule...	19:40
darkmod	Ilari: where is it recorded -- I can't find it in the .gitmodule file -- is it in the index?	19:41
	→ yofel joined	19:41
Ilari	darkmod: Yes, it is in index. To update it, check out the revision you want in submodule and then 'git add path/submodule'. NO trailing '/'!	19:41
blahed	Ilari: maybe i have explained things wrong, i am trying to create a repository for my projects on a remote server that i have admin access to	19:42
darkmod	Ilari: oh, i see.. thanks, had a wrong idea of what git submodule update did. makes sense now. thanks!	19:42
blahed	Ilari: i have a /var/git setup on that server with git:git setup as owner:group	19:42
	Ilari: i want my projects stored in /var/git/	19:43
	← darkmod left	19:43
Ilari	blahed: What isn't working then?	19:43
blahed	Ilari: well let me just start fresh then.... all of that is setup user/group and the repo location /var/git/	19:44
	← Gorbulas left	19:44
blahed	Ilari: what should i do next? should i push from my client?	19:44
	Ilari: is that how i need to get it to the server?	19:44
Ilari	blahed: If you want multiple users (as in, distinct UIDs) to have push access, you should set setgid bit to all directories, give all directories g+rwx and set the shared flag.	19:45
	blahed: If you copied the repository files to server, it is on the server now...	19:45
	→ Sho_ joined	19:46
blahed	Ilari: okay well i can't clone.... so something isn't working... I want to start fresh	19:46
Ilari	blahed: What error does clone give?	19:46
blahed	fatal: The remote end hung up unexpectedly	19:46
Ilari	blahed: The URL syntax is something like 'ssh://user@host/var/git/repo.git'...	19:46
blahed	git clone [email@hidden.address]	19:47
Ilari	blahed: Try dropping that ':'...	19:47
blahed	same thing	19:48
gza	.	19:48
Ilari	gza: Found that Java API?	19:49
	blahed: And still don't work if you try 'ssh://git@10.20.30.25/var/git/intranet_content'?	19:50
bremner	blahed: can you try setting the user in ~/.ssh/config?	19:51
blahed	Ilari: still no	19:51
	bremner: setting what user?	19:52
bremner	blahed: git	19:52
blahed	bremner: i can do a direct ssh connection just fine to the server	19:52
	as the git user	19:52
	so i don't see how it could have anything to do with ssh	19:53
bremner	blahed: I am curious if the problem is with git's parsing of the url. We had this discussion with another user a few days ago	19:53
gza	Ilari: I'm in an argument right now about substandard windows support for Git...	19:53
	→ bryanl joined	19:53
blahed	bremner: well i am following a screencast from peepcode	19:53
	← RaceCondition left	19:53
Ilari	blahed: What is full path of repository (the directory containg stuff like 'objects' and 'refs') on server?	19:54
	← d0k_ left	19:54
blahed	/var/git/intranet_content	19:55
Ilari	blahed: Are permissions correct?	19:55
bremner	blahed: is that a bare repository? (i.e. does it have a .git subdirectory?)	19:55
blahed	Ilari: yeah..... chown -R git:git on the project	19:55
	bremner: it does have .git	19:55
	inside it	19:55
bremner	blahed: I expect that might be problematic	19:56
blahed	so, i should remove it?	19:56
bremner	blahed: actually, let me rephrase. I have only ever cloned remote bare repos.	19:56
	blahed: no, don't remove the .git	19:56
	← markkalderon left	19:56
bremner	blahed: try git clone --bare intranet_content intranet_content.git	19:57
Ilari	blahed: Wrong order.	19:57
blahed	well on the screencast he has a project on his client computer	19:57
	bremner curses	19:57
Ilari	blahed: 'git --bare clone intranet_content intranet_content.git'.	19:57
bremner	llari: thanks, I always mess that up/	19:57
blahed	he creates /var/git on the server, scps the project to the server, chowns it to git:git and then he clones it on the client	19:57
Ilari	That --bare after command might actually work in 1.5.6, but not all run that recent versions...	19:58
	← njal left	20:00
blahed	no that didn't work either	20:01
	is there not a way to debug this.... can i not do verbose or something to see what is going on	20:01
	→ cedricv joined	20:02
Ilari	blahed: If that clone on server doesn't work, then you got something seriously wrong...	20:02
blahed	Ilari: i should clone on the server?	20:02
Ilari	blahed: It was for debugging...	20:03
blahed	ok	20:03
Ilari	blahed: And maybe use that 'git --bare clone', so the clone would be usable for pushing...	20:03
	blahed: Also, maybe use git account on server so it gets perms right...	20:04
blahed	jesus, ........ subversion is calling me to come back	20:05
Ilari	blahed: Also, maybe check out 'gitosis'?	20:06
gram	and then when i get hooked on 16:9 they are going to take that away and it'll all be 2.39:1	20:06
	wrong channel	20:06
	→ ychat joined	20:07
	← felixflores left	20:07
blahed	Ilari: the bare clone did work	20:07
	i had to sudo it though because /var/git is owned by git	20:07
	→ blowery joined	20:08
	→ felixflores joined	20:08
	Modius_ → Modius	20:08
	← felixflores left	20:08
Ilari	blahed: Can you clone the copy you just made?	20:08
	blahed: From another computer?	20:08
blowery	is there a way to get a version number corresponding to the most recent sha1 for a subtree inside a git repo?	20:08
RandalSchwartz	subtree?	20:09
blowery	like if i have /foo/bar and /foo/baz	20:09
RandalSchwartz	sha1s apply only to the whole repo	20:09
	although you can ask what the sha1 is for a subtree	20:09
blowery	i want to get the equivalent of a revision number for /foo/baz/ and all the stuff inside it	20:09
	→ chris2 joined	20:10
	← ForsakenSoul left	20:10
blowery	to back up a bit, i have a directory full of javascript and i'm trying to build a unique url for that directory. i want it to be /js/$rev/blah	20:10
RandalSchwartz	hmm. not git-show	20:10
	where is it	20:10
blowery	with svn, we used svnversion -c	20:10
bremner	blahed: Iust tested it, and git clone user@hostname:/path/to/foo.git works for me	20:10
	s/lust/just	20:11
Ilari	bremner: Starting from filesystem root?	20:11
RandalSchwartz	ahh. git-rev-parse HEAD:	20:11
	that'll be the sha1 of the top of the tree	20:11
bremner	llari: yes	20:11
	→ peeja joined	20:11
blowery	RandalSchwartz: thanks, i'll poke through the man page for that	20:11
RandalSchwartz	git-rev-parse HEAD:perl will be the sha1 of the perl/* tree	20:11
pasky	or git ls-tree -d HEAD perl	20:12
bremner	llari: so user@host:/var/git/my_repo	20:12
pasky	timtowtdi	20:12
RandalSchwartz	why not just use the commit sha1?	20:12
	it's pretty hard to find a commit given a subtree SHA1	20:12
Ilari	bremner: You use sans-serif font, don't you? :-)	20:12
RandalSchwartz	(I think)	20:12
	although it's easy to go the other way	20:12
championdusk	ok, so when I try a 'git push' it says everything is up to date.. but it's not, there is a ton of stuff on my local machine that isn't in the repo	20:13
pasky	well, you can get a _set_ of commits given a subtree SHA1, but then it is especially not easy to decide which one to pick ;)	20:13
blowery	RandalSchwartz: i'd need to know the last commit to affect the subtree. if i could find that, it would work.	20:13
bremner	Ilari: oops, guilty	20:13
championdusk	git status, git add (enter new file name here) show that everything is already up to day	20:13
RandalSchwartz	that's an expensive operation	20:13
	and "last" in what context?	20:13
pasky	blowery: how do you define "last" wrt. branches?	20:13
	bremner kicks konversation	20:13
RandalSchwartz	right	20:13
	← janm left	20:13
RandalSchwartz	you do understand that git commits are a DAG, right?	20:13
blowery	pasky: most recent on a known branch	20:13
RandalSchwartz	branches are an illusion	20:13
Pieter	slow news day	20:14
RandalSchwartz	if you commit 1 2 3 on top of A B C	20:14
	then I commit 7 8 9 on top of A B C	20:14
Ilari	Most recent on given branch is well defined (its the commit pointed by branch tip pointer).	20:14
RandalSchwartz	and we merge	20:14
	is 3 or 9 later?	20:14
	and if you're only looking at your repo, 3 will be later	20:15
	and if I'm only looking at my repo, 9 will be later	20:15
	and what if you rebase	20:15
	← peeja left	20:15
pasky	blowery: anyway, I think git log --pretty=format:%h Documentation \| head -n 1 might be as good for you as anything?	20:15
RandalSchwartz	yeah "latest commit on a branch" means you aren't git-thinking yet	20:15
	← lorandi left	20:15
pasky	it will pick one of the sensible candidates at random ;)	20:15
RandalSchwartz	yeah - at random. good.	20:16
	kewark → krawek	20:16
Pieter	a branch is just a head	20:16
blowery	RandalSchwartz: ? none of that matters for me. i just need a number to shove in a url and i need to change that number when the subtree changes	20:16
	i don't really care about "latest"	20:16
RandalSchwartz	then just use the SHA1 of the subtree	20:16
blowery	sweet	20:16
RandalSchwartz	as already stated	20:16
	don't try to get a date from it	20:16
	it won't go out with you!	20:16
blowery	hah	20:16
	no worries, i dont need a date	20:16
jengelh	a time perhaps?	20:17
blowery	just a number to jam in a url	20:17
RandalSchwartz	although I'm wondering why you're publishing a subtree	20:17
	maybe your git repo is partitioned wrong	20:17
	perhaps you really wanted subprojects instead	20:17
blowery	it's part of a biggish web project	20:18
	currently hosted in svn	20:18
	pasky frowns at subprojects	20:18
	RandalSchwartz averts his eyes at the mention of SVN	20:18
pasky	subprojects are for extreme cases ;)	20:18
RandalSchwartz	eeek!	20:18
Pieter	quick, do the one / two problems quote!	20:18
blahed	bremner: no luck cloning remotely from the new clone on the server	20:18
	if that makes any sense	20:18
blowery	when we publish the app, we grab the latest svn commit for various staticish content subtree and use those to generate unique urls	20:18
	so like say i have some javascript in /js and the most recent svn commit in /js was 1234, we build a url like /js/r1234/blah	20:19
	RandalSchwartz continues to shield his eyes	20:19
jengelh	foreach (mention of svn) { roll }	20:19
blowery	hey man, i'm just explaining current process. i'm not defending it.	20:19
	:)	20:19
jengelh	but we ain't supporting it :p	20:19
Pieter	the problem is that git really doesn't work like svn. So if you want to really work with git, you'll have to drop svn and assume a new workflow	20:20
blowery	from what Randal's said so far, it seems the sha1 for the subtree is perfectly suited to what we really are trying to do, assuming it changes when anything in the subtree changes	20:20
thiago_home	blowery: use the SHA-1 of the commit	20:20
	blowery: it points to the root of the repository though	20:20
modoc	How can I see if a local branch references a remote branch?	20:21
	→ hagna_ joined	20:21
blowery	we also need this magical number to not change between commits that do not affect the subtree	20:21
bremner	blahed: hmm, sorry that's all I got. You already said ssh works, by which I assume you mean ssh git@host works.	20:21
Ilari	blowery: Yes, changing anything in subtree will change subtree SHA (if it does not, I would like to see those subtrees). :-)	20:21
blowery	awesome	20:21
blahed	bremner: yup	20:21
RandalSchwartz	it's technically feasible for it not to change a SHA1	20:21
	but it would be very difficult to find	20:21
	→ felixflores joined	20:21
RandalSchwartz	and pretty obvious what went wrong. :)	20:22
blahed	i am like...... almost ready to give up on this remote repository mess, but i need it for capistrano	20:22
	→ yofel_ joined	20:22
blowery	sweet. that's perfect. i never liked the way we did it with svn because the number increments .. anyway. off to try it.	20:23
	thanks for the pointers	20:23
thiago_home	modoc: cat .git/config	20:23
modoc	thiago_home: thanks	20:23
	→ markkalderon joined	20:24
Pieter	hmm	20:26
	who has heard of libgitcore?	20:26
	← yofel left	20:28
	yofel_ → yofel	20:28
blahed	attttttttthhhhhhhhhhhh	20:30
	i am ill	20:30
	← hagna left	20:31
	← tenpaiyomi left	20:32
Pieter	they've done quite some work at libifying git	20:33
	but I've never heard of it before	20:33
	← alley_cat left	20:33
	← Beket left	20:35
	← felixflores left	20:35
championdusk	something got f'd up in my git repo... I see all the files I just pushed, but when I pull, it says uptodate, it is not uptodate	20:36
	how can I check if im on the right version, etc	20:36
	i also have a 102 fever and my code review is in 30 minutes :)	20:37
	→ johnw joined	20:37
Ilari	championdusk: Are you pulling over HTTP?	20:37
championdusk	Ilari: not sure, all command line	20:37
Ilari	championdusk: Are you trying to pull to same repository you pushed from?	20:38
championdusk	yes	20:38
	i see my code in the repo, it is the most up to date	20:38
	however, pulling says upto date, and missed all of it	20:38
bartman	what branch are you on?	20:39
Mikachu	championdusk: are you pushing into a repo with a working tree?	20:39
championdusk	im a git newb (1 week) so I'm not sure	20:39
	← cedricv left	20:39
	→ bugfux joined	20:39
bartman	championdusk: run 'git branch', which branch has a star on it	20:39
championdusk	i dont remember doing anything other than the initial setup	20:39
	master	20:39
	only branch	20:39
	both machines	20:39
	mine + staging	20:39
bartman	and what does this say: git config remote.origin.push	20:40
championdusk	nothing	20:40
bartman	and how about: git config branch.master.merge	20:40
championdusk	the weird thing is, my new changes (like a quick edit to routes.rb) get caught and commited and pushed	20:40
	bartman: nothing	20:40
bartman	that's probably why it's not merging anything	20:41
	can you do a: git diff master origin/master	20:41
championdusk	bartman: which machine?	20:41
bartman	does it show the kinds of changes you'd expect to have after the merge?	20:41
	local	20:41
Mikachu	i think he pushed into a working tree and then committed, which makes git create a commit that reverts all the changes he pushed	20:41
championdusk	fatal: ambiguous argument 'origin/master': unknown revision or path not in the working tree.	20:41
	Use '--' to separate paths from revisions	20:41
bartman	championdusk: do you run commands on the remote?	20:41
championdusk	yes	20:41
	I am sshd in now	20:42
	→ joeyb joined	20:42
bartman	Mikachu is probably right then	20:42
championdusk	ok, how can I fix my stupidity	20:42
bartman	run gitk, confirm that you have a commit that removes your changes	20:42
championdusk	what is gitk	20:43
bartman	:)	20:43
championdusk	oh my god what the H is this	20:43
	→ mithro joined	20:43
	Mikachu gets the pop corn	20:43
bartman	Mikachu: can I have some?	20:43
championdusk	me too?	20:43
	i am on way too many percs	20:43
	to be dealing with this right now	20:43
	i have no idea what im looking for	20:44
	the thing is, it's IN the repo	20:44
	all my code is in the repo, I just need to get it onto the staging server	20:44
bartman	ssh to the server, do a git checkout -f	20:45
championdusk	now this might be a dumb question, but there have been guys working on the database for about a week straight	20:45
	mysql	20:45
bartman	I don't exactly know what's there, but I think you're pushing local maser to remote's master	20:45
championdusk	none of their stuff will get blown away, right?	20:45
	Your branch is behind the tracked remote branch 'origin/master' by 5 commits,	20:46
	and can be fast-forwarded.	20:46
bartman	you should probably tar up the dir first	20:46
championdusk	how can I backup the db too	20:46
	→ rikur joined	20:46
bartman	just tar up the tree	20:46
	the .git dir contains everything	20:46
championdusk	i'm using mysql	20:46
bartman	huh?	20:46
championdusk	right	20:46
bartman	what does mysql have to do with git?	20:47
	→ not-xjjk joined	20:47
championdusk	nothing	20:47
	i want to backup the mysql table	20:47
	bartman redirects championdusk to #mysql	20:47
championdusk	haha	20:47
	i see your point	20:48
	ok, i tared up my dir	20:48
bartman	you can always go back to that if you break something	20:48
championdusk	ya	20:48
	ok, so when I do git checkout -f, I get	20:48
	Your branch is behind the tracked remote branch 'origin/master' by 5 commits,	20:48
	and can be fast-forwarded.	20:48
bartman	ok, so let's fast forward it... git merge origin/master	20:49
championdusk	that did a bunch of crap	20:49
	→ peeja joined	20:49
bartman	it brought all the changes from origin/master into master and updated the working tree	20:49
championdusk	cool	20:50
	so now git pull?	20:50
bartman	do you need to?	20:50
championdusk	i don't know	20:50
bartman	I think that synced it up with the other repo	20:50
	you see 'origin/master' was already tracking the master branch on origin	20:50
	but git didn't know you wanted it merged	20:51
championdusk	that broke my entire app	20:51
	xyzzy_ → xyzzy42	20:51
bartman	when you said "that did a bunch of crap", did it by chance fail?	20:52
championdusk	no	20:53
	400 files changed, 20188 insertions(+), 14823 deletions(-)	20:53
	the db hook broke	20:53
bartman	wow	20:54
	did you expect that many changes?	20:54
	← holizz left	20:54
championdusk	yes	20:54
bartman	ok	20:54
championdusk	I installed rails 2.1 in the vendor dir	20:54
bartman	I see	20:54
	so what's not working?	20:54
	did you maybe not commit all the files on the first machine?	20:55
championdusk	it can't connect to the database	20:55
	i had the thing IGNORE my database yml file	20:55
	why does it do this crap	20:55
bartman	please narrow your questions to stay on topic :)	20:55
	(for this channel)	20:55
championdusk	i have a 102 fever, a deadline in 7 minutes, and a broken db	20:56
bartman	I know little about your database and less about your ruby app	20:56
	← blowery left	20:56
	pilif → pilif\|away	20:56
bartman	you're saying that it works on one end, but not the other?	20:57
championdusk	right	20:57
	it can't connect to the database	20:57
bartman	so on the computer on which you say it's complete, can you run 'git status'	20:57
	do you have a lot of files listed in 'Untracked files:'	20:57
championdusk	it overwrote my database.yml file in the repo, I thought .gitignore was supposed to... uhh.. ignore the file names you put in there	20:58
	sure, sec	20:58
	← eean left	20:58
	← peeja left	20:58
championdusk	nay, all clean	20:58
	→ ebel joined	20:58
	← docwhat left	20:58
bartman	so, ' git clean -d -x -n ' does not claim it will remove files?	20:58
championdusk	on local?	20:59
bartman	yes	20:59
	← not-xjjk left	20:59
championdusk	ya, some old vim .swp files	20:59
	but nothing important	20:59
	→ sbodo_w joined	21:00
bartman	ok, so everything is committed	21:00
	→ Eridius\|_ joined	21:00
championdusk	ya	21:00
bartman	let's do a manual push... git push origin master:refs/heads/tmp	21:00
	that should create a 'tmp' branch on the remote	21:00
	then ssh there and run: git checkout -f tmp	21:01
championdusk	well, this is the error i am getting	21:01
	Mysql::Error (Unknown database 'utf8'):	21:01
bartman	is git giving you this error?	21:01
championdusk	my biggest fear is wiping the database	21:01
	no, my rails app, I beleive you helped me :)	21:02
bartman	this is #git	21:02
	:)	21:02
championdusk	i know!	21:02
	thank man	21:02
bartman	good luck	21:02
championdusk	im not touch anything else	21:02
	until i know their database is safe	21:02
	→ not-xjjk joined	21:02
	← Eridius\|work left	21:03
	← xjjk left	21:03
	Eridius\|_ → Eridius\|work	21:03
	→ raalkml joined	21:06
	→ rukubites joined	21:06
	← raalkml left	21:06
	→ felixflores joined	21:06
rukubites	A question about git-svn: I want to set the svn property svn:eol-style=native on my new files (preferably automatically), using git. How?	21:08
Eridius\|work	AFAIK, you can't	21:08
	← felixflores left	21:08
rukubites	Well at least I don't ask easy questions. :)	21:09
	→ bobbit32 joined	21:12
	← bobbit32 left	21:12
	← not-xjjk left	21:13
	→ xjjk joined	21:13
	← hobbes006 left	21:15
	← thannoy left	21:16
	← l0verb0y left	21:16
rukubites	Anyone else chime in? I read a blog where someone said they were thinking of adding git-svn propget/set...	21:17
	→ blowery joined	21:17
	→ wagle joined	21:18
jengelh	blargh	21:18
rukubites	alas.	21:21
	← rukubites left	21:21
	← madewokherd left	21:21
	→ SRabbelier1 joined	21:22
	SRabbelier1 → SRabbelier	21:22
	→ holger_ joined	21:23
	→ schacon joined	21:24
	← _zack left	21:26
	→ madewokherd joined	21:28
holger_	newbie question re distributed development - i am a gsoc student who got his own repo which they told me to clone and work in that. it has a copy of master which doesn't update. (cont)	21:29
	→ docwhat joined	21:29
holger_	my understanding from http://www.kernel.org/pub/software/scm/git/docs/user-manual.html#sharing-development-examples was that i should clone the master repo and set up my private branches to be pushed to my private one.	21:30
	so what it the preferred way of doing this?	21:30
	← bugfux left	21:30
	← docwhat left	21:30
Ilari	holger_: Have two remotes registered, one for official upstream repo and one for private repo.	21:32
robinr	holger_: when you clone you get a bunch of tracking branches, origin/master etc plus a local branch initially set to the same commit as origin/master	21:33
	when you fetch only origin/master gets updated	21:33
	master in your clone is "your" branch	21:33
	← entie_ left	21:34
	← hector left	21:35
	→ \ask_ joined	21:35
holger_	yup, if i clone what they set up for me i only see a master branch. so to pull the real upstream master, i'd need to pull master from origin?	21:35
robinr	yes	21:36
mugwump	use git-branch -a	21:36
holger_	easy way to make that the default?	21:36
mugwump	you'll see all the branches it tracks	21:36
robinr	of course you can "see" it anyway	21:36
	gitk origin/master	21:36
	gitk --all etc	21:36
holger_	hmm. git pull origin master doesn't give me anything new either. should it track master on the remote side or am i supposed to set that up somehow?	21:39
mugwump	look at your .git/config	21:39
	take a look at the "pull =" line	21:40
	it's mind-numbingly simple once you see what it's doing :)	21:40
holger_	hmm. origin seems to be my own repo. so i need another remote with the real upstream repo and set branch master to that? (i tried googling that, but all the docs seem to assume that you want to publish what you already have or that you at least can set up your public repo yourself)	21:41
	i'm in the reverse situation. it's set up already and i need to figure out how to use it.	21:42
	→ mithro_ joined	21:43
	queso → queso\|away	21:43
Ilari	holger_: 'git remote add git git://repo.or.cz/alt-git.git', or something like that?	21:43
	holger_: Add '-f' there as well...	21:43
	→ whataboutmark joined	21:43
	← pombreda left	21:43
	← mithro left	21:43
	← stouset left	21:43
blahed	i need someone to help me debug my clone	21:44
holger_	basically, having it pull master from a different repo (the real master) while still publishing everything to my private would do i think.	21:44
	→ pombreda joined	21:44
blahed	tried to setup a remote repo on my internal server	21:45
	← whataboutmark left	21:45
Ilari	holger_: You have [branch "master"] in .git/config?	21:46
	holger_: Also, you have remote already registered for the official upstream repo?	21:47
holger_	Ilari: branch master yes. it came from the clone of my private repo and currently has remote = origin (which is the private)	21:48
Ilari	holger_: Alter that to say remote = <whatever nick the public repo has>	21:48
	← vbabiy-laptop left	21:49
	→ nedko joined	21:50
	← BabelO left	21:50
holger_	ah thanks. that seemed to work.	21:51
Ilari	holger_: Also, git pull can be used for 'one off' pulls (git pull <remote-nick-or-URL> <branch>). That fetches the specified branch and attempts to merge it into current branch.	21:52
	← deavid left	21:52
	← ph^ left	21:55
blahed	Ilari: I just need to copy the .git folder to my remote repo server?	21:55
	→ malc_ joined	21:55
Ilari	blahed: The recommended way is to create empty bare repo (git --bare init) on server (using the same user that will do the pushing) and then populate it using push...	21:56
blahed	Ilari: yeah, well the push failed... i think i am doomed..... i am going to try it with another user	21:56
	Ilari: i was still getting the fatal: The remote end hung up unexpectedly error	21:56
	→ bronson_ joined	21:57
Ilari	blahed: Try the following 'ssh user@server git --version'. What it prints?	21:57
	→ ph^ joined	21:57
	← joeyb left	21:57
	→ joeyb joined	21:57
blahed	Ilari: command not found	21:58
	← xjjk left	21:58
blahed	Ilari: and of course that doesn't work with the git-shell because that command isn't valid	21:59
Ilari	blahed: Does it work for some other user?	21:59
blahed	Ilari: not if you run the command like that, but if you get a shell and then run the command it is valid	21:59
Ilari	blahed: It must be valid if you run the command like that.	22:00
blahed	how do you mean?	22:00
	→ goibhniu joined	22:00
Mikachu	if you use git-shell, it would say fatal: unrecognized command 'git --version'	22:01
Ilari	blahed: The way git invokes the server program on remote side is essentially 'ssh user@host git-<something> <path-to-repo>'.	22:01
	← sbodo_w left	22:01
	→ peter_12 joined	22:02
peter_12	I'm seeing a few broken unit tests on OS X 10.5 with git 1.5.5. Should I worry?	22:02
blahed	Ilari: yeah, but if i am using the git user which has git-shell	22:02
	it shouldn't matter right?	22:02
Ilari	blahed: IIRC, there was issue with git-shell that it didn't properly initialize PATH...	22:03
blahed	Ilari: so what do you recommend	22:03
	not using git-shell?	22:03
Ilari	blahed: I don't even know how 'deployment systems' would react to finding that user shell is set to 'git-shell'...	22:04
Mikachu	peter_12: if it involves märchen, no :)	22:04
peter_12	Mikachu: "marchen"?	22:05
blahed	shouldn't matter	22:05
	if i can get GIT working, the rest will work	22:05
jengelh	märchen -- fairytales	22:06
Pieter	I removed the stupid Märchen file in my git clone	22:06
	It kept bugging me :)	22:06
mugwump	nothing to do with märz	22:07
jengelh	definitely not.	22:07
	→ strangy joined	22:07
Eridius\|work	Pieter: I just added it to .git/info/exclude ;)	22:07
jengelh	though the file existence provides testing that utf8 works correctly	22:07
Pieter	Eridius\|work: hmm, I sometimes get a "removed" status thing too	22:08
Eridius\|work	Pieter: if you removed the file, then you would get removed	22:08
	→ srid joined	22:08
Ilari	blahed: Bit dirty but should work: Symlink the git server programs 'git-upload-pack' and 'git-receive-pack' to some directory that is searched for executables by default...	22:08
peter_12	can I connect a remote git repository using ssh?	22:08
Eridius\|work	I just added it to .git/info/exclude so it doesn't think its's a new file	22:08
srid	I have some uncommited changes in my current directory. I don't want to commit them. How do I throw them away (and get back to last commited state of files)?	22:09
Eridius\|work	git reset --hard HEAD	22:09
jengelh	peter_12: git is transport agnostic as long as you have a fully bidirectional pipe	22:09
srid	but I want to restore only a particular set of files	22:09
Mikachu	pOSiX	22:09
blahed	Ilari: yeah, jeeze	22:09
doener_	srid: git checkout -f HEAD -- files	22:09
	s/-f//	22:09
	← destruct left	22:10
	→ cedricv joined	22:10
peter_12	so how would i clone a git repository over ssh?	22:11
Eridius\|work	peter_12: by reading the manpages or basic tutorials?	22:11
srid	doener_: thanks	22:11
peter_12	Eridius\|work: great answer	22:11
Ilari	peter_12: 'git clone ssh://user@path/path/to/repo.git'?	22:11
peter_12	Ilari: thanks	22:11
Eridius\|work	Ilari: oh come on, if you just hand out answers like candy, nobody will ever learn	22:11
	note that you don't need the ssh:// part, just user@domain:path/to/repo.git will work	22:12
	(i.e. scp-style)	22:12
	← ceez left	22:12
	→ xjjk joined	22:12
peter_12	Eridius\|work: then how does it know to use ssh?	22:12
	it is the default?	22:12
Eridius\|work	because the user@domain:path/to/repo.git format means ssh	22:12
peter_12	ok. I didn't know that was always the case	22:13
	← joeyb left	22:13
Mikachu	you would have if you looked in the man page	22:13
	← JonathanT left	22:14
	← csc` left	22:15
	← eikonos left	22:16
	← ph^ left	22:16
	← Eridius\|work left	22:20
	→ Eridius\|work joined	22:20
	queso\|away → queso	22:20
	→ peeja joined	22:21
	← capisce left	22:21
	← koke left	22:26
	→ eikonos joined	22:27
	← blahed left	22:29
	← peeja left	22:31
Debolaz	When I have created packs with git-pack-objects, can I remove the loose objects manually with a utility like find?	22:31
johnw	what do you mean by loose objects?	22:32
	the ones that are now in the pack?	22:32
Debolaz	The objects used to create the pack files, living in objects/\d\d/*	22:32
	Ok, I guess \d wasn't the right regexp sequence..	22:33
	But yeah.	22:33
johnw	i've never done that, but it sounds like you want to delete the same objects that you passed to git-pack-objects	22:33
Mikachu	Debolaz: why aren't you using git-repack?	22:33
Debolaz	Mikachu: Because it fails miserably on this repository.	22:33
	← championdusk left	22:33
Mikachu	Debolaz: how about git-prune-packed then?	22:33
	git-prune-packed - Remove extra objects that are already in pack files	22:33
Debolaz	Also, the pack files I've made, can I just move them manually into the pack/ directory?	22:34
	And then run git-prune-packed?	22:34
Mikachu	i think you need to put it in objects/info/packs as well	22:34
johnw	Debolaz: you can use this script to remove all your packs: http://rafb.net/p/zcJa6I91.html	22:35
mugwump	Debolaz: if I were you I'd just move the loose object directories out of the way	22:35
johnw	and then you can git-repack	22:35
mugwump	or delete them if you've got a tarball	22:35
Debolaz	johnw: But as I said above, git-repack doesn't work.	22:35
mugwump	johnw: this is a massive repo	22:35
Ilari	Debolaz: 'git-prune-packed'?	22:35
Mikachu	Debolaz: why does git-repack fail? is your repo corrupt?	22:35
johnw	even after you remove all packs?	22:35
Debolaz	Mikachu: No, the repo is too big.	22:35
johnw	ahh	22:35
	can't you tell repack to create multiple pack files?	22:35
Mikachu	why not use --max-pack-size=200 or so?	22:36
Debolaz	Ilari: ?	22:36
mugwump	there's a nice little conversation about it in the backlog if you're interested, johnw	22:36
Debolaz	Mikachu: Doesn't matter, it doesn't even get past counting the objects in reasonable time.	22:36
Ilari	Debolaz: IIRC, that command deleted loose objects that exist in packfile...	22:36
johnw	mugwump: thx	22:36
mugwump	from about 11 hours ago	22:36
Debolaz	Ilari: Ah, got a little confused there since it had been suggested, and I thought you wondered what I wanted to do with it... :)	22:36
	→ holger joined	22:37
Debolaz	But if I have a pack file, how do I install it into the repository?	22:38
Ilari	Debolaz: Pack files need to be named in certain way (git-pack-objects should name them right) and need to have corresponding .idx files moved there too...	22:38
Mikachu	Debolaz: i guess you want to read git-repack.sh then	22:38
Ilari	Debolaz: What is name of the packfile? pack-<40 hex digits>.pack?	22:38
Mikachu	Debolaz: easiest might be to replace the names=$(git pack-objects... line	22:38
Debolaz	Ilari: Right now, 0-.pack\|idx, 0 being named after objects/0 which was what I told it to pack.	22:39
Mikachu	ie, you have a list of objects in a file, so put $(cat file \| git pack-objects --non-empty $args "$PACKTMP")	22:39
Ilari	Debolaz: That won't work.	22:39
Debolaz	I realize that it won't work, I'm asking how I can make it work.	22:39
Mikachu	i've successfully done that sometime, don't remember why :)	22:40
	obviously copy git-repack to . first	22:40
Ilari	Debolaz: Can you give example filename?	22:40
Mikachu	i think any name is okay as long as you list it in the info file though	22:40
Debolaz	I replace the 0-.pack\|idx with pack-.pack\|idx, then move it into objects/pack ?	22:40
Mikachu	it's not like git is going to go and compute the sha1 of the pack file every time it accessses an object in it and go "aha, this is not the right sha1!111"	22:41
Ilari	Debolaz: If the names have that 40 hex digits, then that could work.	22:41
Debolaz	Ilari: The hex stuff is there yes, pack-objects generates it automagically after the prefix I specify.	22:41
	The 0- 1-, etc, has just been so I can keep track of what has been packed.	22:42
Ilari	Debolaz: Then just rename the files to pack-<...>.pack/.idx and move them to .git/objects/pack...	22:42
	→ wereHamster joined	22:42
Debolaz	Ilari: What about the info file stuff that has been mentioned, does this need to be fixed?	22:42
Mikachu	am i somehow less trustworthy than Ilari?	22:42
sbeyer	;)	22:43
wereHamster	is there an easy way to find out if a particular commit is part of a tree? in master there is a commit, and I'd like to check if the same patch was applied to a branch	22:43
	← eternaleye_ left	22:43
Ilari	Debolaz: What info file? If you mean the one in objects/info?	22:43
Debolaz	Hehe, I was about to comment how I didn't mean to exclude any help.. :)	22:43
jast	Mikachu, yep... your name is too close to a certain cartoon character.	22:43
Mikachu	jast: http://www.behindthename.com/name/ilari	22:44
Debolaz	git-update-server-info I guess does that job?	22:44
jast	interesting	22:44
Mikachu	that's a different info file ;)	22:44
Debolaz	That's the only info thing git-repack touches though?	22:45
Mikachu	oh hm, yeah maybe that packs file is only for http:// after all	22:45
	i tried moving it away and nothing broked	22:45
sbeyer	hmm, is it somehow possible to generate "local tags"?	22:45
	Mikachu shoots lightning bolts all around	22:45
Mikachu	sbeyer: git tag mytag sha1	22:46
	then just don't push it ;)	22:46
	mithro_ → mithro	22:46
mugwump	Debolaz: how big was your pack?	22:46
sbeyer	Mikachu: and how can I assure that git doesn't push the tag when I push a branch? ;)	22:47
mugwump	esp. vs source files	22:47
Mikachu	sbeyer: it never pushes tags when you push branches, it only pushes tags when you push tags	22:47
jengelh	git does not push tags by default	22:47
sbeyer	Mikachu: hmm I'm too unsure about "git push" to risk ;)	22:47
Debolaz	mugwump: Packing by objects/0*, the pack files gets roughly 180MB large.	22:47
Mikachu	Debolaz: git count-objects -v?	22:47
	→ eternaleye_ joined	22:48
mugwump	1.5million objects or so	22:48
	ok, so you're looking at about 2.8GB	22:48
sbeyer	Mikachu: usually I do "git push a_remote a_branch", but sometimes only "git push" and sometimes it works, too ;-) I've never cared about more detail in that ;)	22:48
mugwump	how big is the original cvsup?	22:48
Mikachu	sbeyer: you have to say git push --tags to push tags	22:48
	sbeyer: or the tagname	22:48
sbeyer	Mikachu: thanks	22:48
	← malc_ left	22:50
Debolaz	mugwump: It's a subversion repository, checking the size of it now.	22:50
	← peter_12 left	22:50
mugwump	oh, they downgraded to subversion?	22:50
Debolaz	3.4 GB SVN repo size.	22:51
mugwump	right	22:51
	have you converted all the branches within the svn repo to git branches?	22:51
Pieter	Debolaz: you packed those objects yourself?	22:52
Debolaz	Yes.	22:52
	Pieter: Yes, git repack can't handle it.	22:52
Mikachu	you mean you're too impatient	22:52
mugwump	Mikachu: shut up	22:52
Pieter	Debolaz: if you did it yourself, how did you figure out the ordering?	22:52
Mikachu	mugwump: ?	22:53
Ilari	Hmm... Might just taking body of packfile (everything after 12 byte header before 20 byte trailer) concatenating these bodies, writing header and trailer for new packfile and then indexing it work?	22:53
	← joevandyk_ left	22:53
mugwump	I'm surprised you didn't get more compression than that	22:53
Debolaz	Pieter: I didn't, I just do find objects/0* -type f \| perl -pe 'filtering' \| git pack-objects 0	22:53
Pieter	that explains the bad compression then	22:53
Debolaz	Right now I'm not concerned with compression at all.	22:53
Pieter	yeah, I'm just saying	22:54
Debolaz	I'm just concerned about reducing the amount of files in the repository so git will be able to deal with it.	22:54
	← holger_ left	22:54
	← ychat left	22:54
mugwump	sure	22:54
Pieter	(is this a public repo?)	22:54
mugwump	maybe the upstream repository is just compressed well	22:54
	ie, svn's deltas are arranged pretty well already	22:54
Debolaz	Pieter: Yes, I've put it online.	22:55
jengelh	git's are better	22:55
	← timocratic left	22:55
Pieter	mugwump: he got bad compression because he just used the filename ordering	22:55
mugwump	I guess that earlier note applies though - without being able to walk the full graph, the algorithm that finds likely good delta candidates is stymmied	22:55
	ffs there are a lot of people who think that people are doing things wrong in this channel	22:55
Debolaz	http://git.andersberle.com/?p=freebsd.git is the gitweb for the repo.	22:56
Ilari	Debolaz: Where those 1.3M objects all loose before repacking?	22:57
Debolaz	Ilari: Yes.	22:57
Ilari	Ouch...	22:57
Debolaz	Doing git pack-objects manually works well from a performance point of view however.	22:57
Pieter	yes, I'm not saying you shouldn't do that	22:58
Mikachu	mugwump: are you saying we shouldn't say anything when people are doing things wrong?	22:58
	fwiw, i didn't say anyone did anything wrong either, so i'm not sure who you're referring to	22:58
mugwump	no, I'm saying you should try and listen and understand what people are doing before saying they're doing it wrong	22:58
	stop taking me literally	22:58
Pieter	who said anyone was doing anything wrong?	22:58
jast	you're all doing it wrong! <- now i did	22:59
Mikachu	i only corrected the statement that git repack couldn't handle it, which it can, i can understand if you don't want to wait for it	22:59
	i don't like people telling me to shut up for no good reason	22:59
Ilari	Debolaz: Can you get 'git rev-list --all --objects' output dumped to some file?	23:00
mugwump	Debolaz: did you try the repack after removing all the loose objects?	23:00
Debolaz	mugwump: I'm not done creating packs yet.	23:00
mugwump	ah :)	23:00
Pieter	how long does the packing take?	23:00
	(I thought git-svn autopacked every 1000 commits or so btw)	23:01
	← wereHamster left	23:01
Mikachu	i wonder if passing --window=1 or so would help, since the 10 objects it tries will most likely be useless anyway	23:01
	← cannonball left	23:01
Debolaz	Pieter: A few hours per chunk, for a total of 16 chunks. Ie, a chunk is objects/0* 1* 2* etc.	23:01
Pieter	heh, painful :)	23:01
mugwump	Mikachu: well, there's always a chance :)	23:01
Debolaz	Well, it's not as painfull as waiting 3 months (That's the figure I calculated anyway) for git-repack to even just get done counting the object, let alone do anything usefull with them.	23:02
mugwump	how does that 180MB compare in disk space reduction? ignoring allocation overhead	23:02
Eridius\|work	Debolaz: the object counting isn't linear	23:02
Pieter	the object-count thing goes exponentially or o	23:02
	Hobbes → Guest25904	23:02
	→ Hobbes`` joined	23:02
mugwump	yes. thinking further about that time vs tar, it's down to the fact it's seeking randomly, and the OS can't hold the 1.5m directory entries in memory at once, so has to do lots of disk seeks for every object access	23:03
Pieter	mugwump: 180*16 = 2.9GB vs3.4GB svn	23:03
Debolaz	It had a more or less constant count of 10 objects per second.	23:03
	← jossipek left	23:03
mugwump	I meant, compared to the loose objects	23:03
Pieter	ah	23:03
Eridius\|work	Debolaz: it definitely speeds up at some point	23:03
Debolaz	mugwump: Working on that.	23:03
Pieter	wasn't there someone who had already created a freebsd repo in the what-vcs-system-should-we-use phase?	23:04
	→ timocratic joined	23:04
Pieter	ah, here	23:04
	500MB for /src	23:04
Mikachu	Debolaz: how long did the actual import take?	23:04
Debolaz	mugwump: 430 MB	23:06
	Meaning the git repo is close to 6.5 GB	23:07
mugwump	right	23:07
Debolaz	Mikachu: 4 days or so.	23:07
	← hagna_ left	23:07
	← charon left	23:07
Pieter	you can distribute the packing thing nicely though	23:08
	just send 15 other guys a 400MB tarball	23:08
Mikachu	hm, but repacking just a few hundred megs shouldn't take many hours...	23:11
Ilari	Machine trashing due to too large window/delta base cache?	23:12
gza	Pieter: did you go to Railsconf this year?	23:12
	→ Lytol joined	23:12
Pieter	gza: no	23:12
gza	ok different pieter then... ;)	23:13
doener_	Debolaz: just wondering, which OS/fs is that repo on?	23:13
Pieter	:)	23:13
	freebsd, I'd guess	23:13
	← yofel left	23:13
	← brosner left	23:13
Lytol	is there a way to query a remote server for commits that you haven't yet fetched (without actually fetching them)?	23:13
doener_	Pieter: I'd guess so, too. But you can't know for sure without asking ;-)	23:13
mugwump	doener_: sure you can, read the logs from 12 hours ago	23:13
Mikachu	Lytol: you can see the sha1 of the tip, but that doesn't tell you much	23:14
	← Guest25904 left	23:14
Pieter	Lytol: git-peek-remote will tell you if the branch heads are different, but not how many commits	23:14
doener_	mugwump: I wouldn't have known that it was mentioned there ;-)	23:14
Lytol	hmmm...	23:14
mugwump	however I think Linux' dircache would probably also suffer from a similar problem	23:14
	← Eridius\|work left	23:14
	← juanjoc left	23:14
	→ Eridius\|work joined	23:14
doener_	mugwump: ext3's dir_index can make a big difference though	23:14
mugwump	dircache. dircache.	23:15
Debolaz	FreeBSD yes, with UFS2.	23:15
jengelh	ext3 worst	23:15
	well, perhaps still above the ufses.	23:15
Mikachu	i started a git repack -adf on linux-2.6 3 minutes ago and it's already 15% done, and this computer is 5 years old	23:15
Pieter	I wonder why the freebsd cvs guy thinks he'd have to split op /src	23:16
jast	how many objects does linux-2.6 have?	23:16
Pieter	if it's only 400MB	23:16
Mikachu	i guess it might be much slower if the source objects are loose though?	23:16
	jast: 828684	23:16
jengelh	Mikachu: too slow :p	23:16
jast	that's very likely, yes	23:16
Pieter	you can also do something neat like writing the new pack to another disk	23:18
	or using fast-export \| fast-import as some sort of primitive packer :)	23:18
Mikachu	would be possibly useful if you could disable the page cache for that new file too :)	23:18
Ilari	And if fast-import tried to use previous version of each file as delta base?	23:19
	← chris2 left	23:19
Pieter	it doesn't really	23:19
	but, what if?	23:19
	it only has the stream as input, and no loose objects in it's git dir, so shouldn't matter	23:19
	→ unreal_ joined	23:20
Mikachu	wow, it suddenly jumped to 100% there	23:21
Ilari	Yea, it jumps up fast after it gets commit objects done...	23:22
Mikachu	hm, it first does compressing objects, then writing objects... does it keep the whole pack file in ram while compressing?	23:22
Pieter	yes	23:22
Mikachu	pack-objects is using 460M	23:22
	who came up with that clever design? :)	23:22
	← nud left	23:23
jast	isn't it mmapped?	23:24
Mikachu	no	23:26
	doesn't look like it anyway..	23:27
	← agib left	23:28
	← ebel left	23:34
	→ peeja joined	23:35
	← unreal left	23:36
	← \ask_ left	23:36
	unreal_ → unreal	23:36
	← bronson_ left	23:39
	← sbeyer left	23:40
	→ peter_12 joined	23:43
	← jeckel left	23:44
peter_12	I have been reading the git-checkout man page. The definition is recursive "git-checkout - Checkout a branch or paths to the working tree"	23:44
	is this the same as "svn switch", more or less?	23:44
Eridius\|work	how is that recursive?	23:45
	it's most commonly used to switch branches	23:45
	but you can also give it paths instead	23:45
peter_12	checkout is defined in terms of "checkout"	23:45
Eridius\|work	no, that's the summary	23:45
peter_12	ok	23:45
Eridius\|work	it's defining the command git-checkout in terms of the standard english word Checkout	23:45
mugwump	checkout is like switch, yes	23:45
peter_12	mugwump: thanks	23:45
Eridius\|work	if you give it paths it's like svn checkout	23:45
mugwump	I think there are differences to that usage	23:46
	if you 'git checkout REV path/', it will get that path/ from REV	23:46
	← robinr left	23:46
mugwump	but doing that with svn will leave the checked-out directory associated with the original SVN path	23:46
	← [sand-man] left	23:48
	← peeja left	23:48
	← unreal left	23:48
Debolaz	Yay, chunk 3 of 16 done.	23:49
	← cmarcelo left	23:49
Ilari	Debolaz: Repacking that repo?	23:50
Debolaz	Yeah, doing git pack-objects manually.	23:51
	→ unreal joined	23:51
	→ Xyrio joined	23:52
	→ drizzd joined	23:54
	→ warthog9 joined	23:55
	← gass left	23:56
	← strangy left	23:58
	→ strangy joined	23:59

Logs Search ←Prev date Next date→

Channels Documentation