IRCloggy #git 2008-01-08

Provider of IRC logs since 2005.

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

	2008-01-08
	→ agib joined	00:00
tpope	hey bigred-ian	00:00
bigred-ian	tpope: hey man	00:00
GSaldana	to pick edits we use the cherry pick command?	00:00
tpope	pull/merge merges the two branches together	00:00
	a new commit is created with two ancestors	00:01
Ilari	GSaldana: If one wants to pick whole commits then cherry-pick is fine.	00:01
	← YGingras_ left	00:01
GSaldana	else, i have to manually edit?	00:01
tpope	the way rebase basically works is, it removes each commit one by one, making them into patches, until it gets to a common ancestor	00:01
	then, it applies those patches in the correct order to the other branch	00:02
agreen	Hi folks --- I was asking here yesterday about why stgit trashed the commit history so I could no longer push to a remote repo I had cloned from, I learned why that was so. But I really like stgit, I am wanting to experiment with a style of the remote repo where it "mirrors" my local tree that I continually use stgit on	00:02
tpope	the end result of a rebase is one continuous history	00:02
Ilari	GSaldana: There are ways to pick whole files...	00:02
tpope	this is generally preferable	00:02
	BUT	00:02
bigred-ian	tpope: so if i didn't care so much about history. i could interchange the use of rebase/merge	00:02
tpope	it rewrites the history	00:02
	this is a bad thing if you've already shared it	00:02
agreen	What I think I need is a way to truncate the remote branch and then resend the branch I was working on as a fresh remote branch, accepting that I lost the remote history	00:02
	→ nashdj joined	00:03
Tv	agreen: look at what "+" means in refspecs, man git-push	00:03
tpope	so generally, you use rebase for simple things that you haven't shared, and merges for more complex things that you have	00:03
agreen	Tv: thanks, I am looking	00:03
tpope	if you don't care about history, you can substitute a merge for a rebase, yes	00:03
Ilari	GSaldana: Following might work: 'git show <commit to pick from> -- <file1> <file2> \| git apply --index'.	00:03
tpope	the reverse isn't necessarily true, due to caveats about rewriting	00:04
bigred-ian	tpope, so if i do a pull(or merge) instead of a rebase, then my branch would lose the history of what changes were made in the root branch	00:04
Tv	bigred-ian: pull = fetch + merge	00:04
tpope	the history isn't lost	00:04
	it's just shaped differently	00:04
bigred-ian	if history isn't lost, then why even have a rebase. why not just merge?	00:04
tpope	if you do a merge, it looks like an O. it splits and then rejoins	00:04
	if you do a rebase, it's a continuous line	00:05
	→ orospakr joined	00:05
Ilari	GSaldana: That does not commit through. But it does stage the changes ready for commit.	00:05
tpope	let me draw up the classic ascii art	00:05
	bigon → bigon`	00:05
	→ YGingras_ joined	00:06
	→ rhalff joined	00:06
Tv	time to pimp this, again: http://eagain.net/articles/git-for-computer-scientists/	00:07
	way better than ascii art ;)	00:07
	→ opsis25 joined	00:08
bigred-ian	Tv thanks	00:08
Ilari	Its bit of tradeoff: Rebase leads to nicer history but screws others if not used carefully. Merge leads to messier history but does not screw others even nearly as much...	00:08
agreen	Tv: you rock, I have the power :-)	00:09
	Tv: thanks a lot	00:09
Tv	maybe git should try to prevent rebasing commits that appear in any remote branch	00:09
	require --force to override or something	00:09
bigred-ian	thanks for clearing it up. the end result of rebase and merge are essentially the same. they do the same thing, just different ways of going about it	00:09
tpope	bah, well I did it anyways	00:10
	http://pastie.caboo.se/136460	00:11
	← YGingras_ left	00:11
Ilari	The new tip is going to be more or less the same. But other results of rebase are likely completely new things...	00:11
	→ YGingras_ joined	00:11
opsis25	when I move a file from one dir to another (using mv not git-mv) I see this in the log: "# deleted: [file name]" and then the it is also listed under the untracked files with the new location. am I suppose to commit both the old (non-existent) file and add the new one? will history be preserved? how do I specify a commit message for the move?	00:11
tpope	opsis25: yes, commit both	00:12
	git doesn't track renames in any special way	00:12
Ilari	It attempts to infer the renames if needed (and asked).	00:13
opsis25	so do a 'git commit [old file path] [new file path] -m "my message"'?	00:13
	no separate add?	00:13
gitster	"git add old new; git commit" is also fine.	00:13
	oops; "git add old" won't work --- you are not adding but removing ;-)	00:13
Ilari	opsis25: -m to commit shouldn't be used except in some quick-and-dirty testing or scripts.	00:13
	gitster asks "why not"?	00:14
opsis25	so if I try and get history on the renamed file (not even sure how to do that), will the old history be preserved across the move?	00:15
	← kristoffer left	00:15
	← YGingras_ left	00:16
	→ YGingras_ joined	00:16
opsis25	mmm... or doesn't git have single file history	00:16
	guess not, that's a paradigm shift	00:16
	← YGingras__ left	00:17
bigred-ian	tpope: thanks for helping me clear it up. they arent as different as i thought they were	00:17
	→ sverrej__ joined	00:17
tpope	they are different approaches to the same problem	00:17
	← HG` left	00:17
Ilari	Single file history and DSCMs don't mix that well... As was discovered several times during Linux kernel BK days...	00:17
bigred-ian	tpope: exactly. but they aren't totally different operations like i thought they were	00:17
opsis25	but if I do a "git log [my moved file]", the old history will still be there?	00:19
tpope	opsis25: it doesn't track renames because when you think about it, those are just one of many special cases, (splitting a file, joining two files, moving stuff from file 1 to file 2, etc etc etc)	00:19
	opsis25: but commands like git log have a --follow option that looks for files that were deleted/added in the same commit, and infers renames from that	00:20
	→ Ademan joined	00:20
	← timocratic left	00:20
opsis25	okay, thanks	00:21
	← agreen left	00:24
	← sverrej_ left	00:24
gitster	even better is "git blame -C" ;-) that looks for groups of lines (not the whole file) that were copied or moved from elsewhere.	00:33
	→ KirinDav joined	00:36
	← tvachon left	00:47
opsis25	last question about the move stuff: after doing the move will a "git-svn dcommit" include an SVN rename?	00:48
	← Ademan left	00:49
gitster	I doubt it, but I am not a git-svn person so take that with a grain of salt.	00:50
	← jmspeex left	00:54
	→ jmspeex joined	00:54
	→ priidu_ joined	00:55
	← vintrepid left	00:56
opsis25	just tried it, it marks it as a delete and a copy, not a rename. the history is preserved though because it is a copy.	00:56
	→ tvachon joined	00:57
aeruder	opsis25: svn renames are always a copy + delete	00:59
	→ \|kukks\| joined	01:01
opsis25	guess so, I was getting confused with the ViewVC page	01:02
	← devogon left	01:04
	← theCarpenter left	01:05
	← tvachon left	01:09
	→ tvachon joined	01:12
	← kukks left	01:18
	← GSaldana left	01:18
pigeon	how does git-svnimport handle svn's external item?	01:21
opsis25	it doesn't	01:21
	→ sempuki joined	01:21
pigeon	ah	01:22
opsis25	someone has used git submodules to support svn externals though	01:23
	http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2007-May/028406.html	01:23
	pretty complicated	01:23
pigeon	funny enough, that's exactly the same stuff what i wanted to import (ffmpeg)	01:24
	→ _kukks_ joined	01:25
	→ jesse joined	01:28
	← jesse left	01:28
	← kelvie_ left	01:30
	→ _kukks_3 joined	01:31
	→ kelvie_ joined	01:32
	← yeowman left	01:38
jfm3	I think I get git in theory, but I'm still playing with it to learn the details. I've a clone of the source and it's not compiling, breaking on de.msg. What am I doing wrong?	01:38
	→ jesse joined	01:38
	← jesse left	01:38
	→ Ademan joined	01:39
	← KirinDav left	01:40
	← \|kukks\| left	01:42
gitster	"apt-get install gettext"?	01:43
	← cedric_ left	01:43
	→ Adam12 joined	01:44
	← _kukks_ left	01:44
	← GyrosGeier left	02:01
	→ \|kukks\| joined	02:04
	→ soto joined	02:06
	→ zapnap joined	02:07
soto	Is the output of 'git diff -p' compatible with the patch command?	02:07
Ori_B	did you try it?	02:09
	← orospakr left	02:10
Ori_B	(quick answer: yes, but you need -p1 to ignore the "a/" that git-diff gives the file name)	02:10
soto	Ori_B: The git header is ignored by patch?	02:11
	Ah never mind, patch man page says it tries to skip any leading 'garbage'.	02:12
	Ori_B: Thanks	02:12
Ori_B	np	02:12
Ilari	Hmm... Does git diff need explicit switch to enable binary diffing? Those things definitely aren't compatible.	02:13
	→ yeowman joined	02:13
aeruder	Ilari: actually at one point there was a bit of a problem with the diff code in git	02:14
	mostly in that sometimes it didn't detect binary files correctly, and regardless, the diff code could fail to apply its own output	02:14
	← soto left	02:19
	← lamont left	02:20
	← _kukks_3 left	02:20
	→ FunkeeMonk joined	02:20
	→ lamont joined	02:21
	→ zed_ joined	02:22
sempuki	when doing a rebase, how can I skip a commit? I thought I could use git rebase --skip, but it keeps telling me "Dirty index: cannot apply patches (dirty: CONTRIBUTORS.txt)"	02:23
	→ kmap joined	02:24
aeruder	sempuki: that means just undirtyfy CONTRIBUTORS.txt and then use --skip	02:24
	i.e.	02:24
	← \ask_ left	02:24
aeruder	git checkout HEAD -- CONTRIBUTORS.txt	02:24
	git rebase --skip	02:24
	← tvachon left	02:24
	← rhalff left	02:26
sempuki	I ran:	02:26
	$ git checkout -f foo	02:26
	$git rebase bar	02:27
	how could it have been "dirty"?	02:27
aeruder	i know, and i'm saying that it tried to rebase, it failed	02:27
	it left it in a unmerged state	02:27
	you need to clean it up	02:27
	→ imyousuf joined	02:27
aeruder	imyousuf: hey, any luck?	02:27
sempuki	hrm, I guess I didnt/dont understand the meaning of "dirty"	02:28
imyousuf	hi aeruder	02:28
aeruder	sempuki: dirty = having local changes	02:28
imyousuf	yes, with help from gitster I could actually determine the reason of such behavior	02:28
aeruder	sempuki: when the rebase fails, it will put the file into an unmerged state, conflict markers and all	02:28
	i.e. diry	02:28
sempuki	ah, so that means everytime that it partially applies a diff its dirty? I thought that case was magically taken care of	02:28
aeruder	*dirty	02:28
sempuki	now I know, thanks!	02:28
imyousuf	aeruder: the git-clone script sets and exports the GIT_DIR environment variable which is used by git ls-files	02:29
aeruder	aha!	02:29
	that makes sense	02:29
imyousuf	since the GIT_DIR is set to the top level repo it does not work at level 1	02:29
	for test purpose i unset GIT_DIR before my recursion starts and at the end of the recursion I restored and exported the old value	02:30
	and it works fine now	02:30
aeruder	good deal	02:30
	glad you figured it out	02:30
	read: i'm lazy	02:30
imyousuf	thank you for the help as well aeruder	02:30
	btw do you and anyone in channel know if do unset GIT_DIR in git-clone what else should I set? I am not "setting" any environment variable	02:31
aeruder	mm... there's a GIT_CONFIG too, not sure if that is important for you though	02:33
	probably not	02:33
imyousuf	I see	02:34
	ok, if I search with [GIT].*=i with VI in the git-clone I get 3 (variables) sets and exports	02:36
	GIT_DIR, GIT_CONFIG and GIT_WORK_TREE	02:36
aeruder	o	02:36
	probably all important for you :)	02:36
imyousuf	:)	02:36
	For safety I will back them all up and again restore them after recursion :)	02:37
	← bobesponja left	02:38
imyousuf	what does this line mean? - export GIT_DIR && GIT_CONFIG="$GIT_DIR/config" git-init $quiet ${template+"$template"} \|\| usage	02:39
aeruder	which part	02:40
imyousuf	aeruder: the significance of && and \|\|	02:42
aeruder	&& is like ; but only when the previous command exits successfully	02:43
	\|\| is like ; but only when the previous command exits unsuccessfully	02:43
imyousuf	ok	02:43
	got it, thanks aeruder	02:43
Ori_B	ooh, neat. when did git get sumbodule support?	02:46
	and is there a tutorial on how to use it?	02:47
imyousuf	I think from 1.5.3	02:47
	the tutorial is in the manual Chapter 8 if I am not mistaken	02:47
aeruder	Ori_B: if you want to use submodule stuff	02:47
	i'd recommend using as new of a git as you can ;)	02:48
	← zapnap left	02:48
Ori_B	aeruder: nothing urgent, just something that I can see being useful in the near future	02:48
	1.5.3.7 seems to have the command, at least	02:48
aeruder	yeae, that'd be plenty new :)	02:49
imyousuf	aeruder: I just pushed the latest modification, the one that actually does what was intended originally. If you have time have a look and let me know what you think	02:49
	→ ironfroggy joined	02:50
	→ sunblush joined	02:54
	jwb → jwboyer	02:56
	→ tvachon joined	02:56
	← tvachon left	02:56
	→ stick joined	02:59
	→ cmacleod joined	03:02
	← cmacleod left	03:02
	← stick left	03:02
	→ cmacleod joined	03:02
	jwboyer → jw	03:06
	jw → jwb	03:06
	← tongueroo left	03:11
	cmacleod → stick	03:12
	→ KirinDav joined	03:13
gitster	When talking about features, comparing 1.5.3 and 1.5.3.7 is meaningless ;-).	03:14
	← sgrimm left	03:18
	← Ademan left	03:20
Debolaz	Are there any up to date technical comparisons of git and mercurial?	03:24
Tene	diff -r ~/src/git ~/src/hg	03:29
	;)	03:29
Debolaz	On the internet, everyone thinks they are a good comedian. :-P	03:30
Tene	On the internet, the only audience that matters is me. :-P	03:30
	← opsis25 left	03:47
	→ opsis25 joined	03:55
	→ nicolai joined	03:56
clee	Debolaz: git is better, of course.	04:02
	Debolaz: (what would you expect people in #git to say, dude?)	04:02
	← jfm3 left	04:08
Debolaz	I would expect them to answer the actual question or make jokes about it, not make up a question I didn't ask and answer that. :)	04:09
clee	Debolaz: dude, you're on IRC	04:09
	c'mon now	04:09
	← a-atwood left	04:10
	→ bartman joined	04:10
clee	Debolaz: seriously though, I'm not aware of any up-to-date comparisons per se, but most of the git fundamentals haven't changed too much	04:10
	I don't really follow hg development but I suspect that it hasn't gotten radically different either	04:10
	the big difference for me, and the reason I tend to prefer git for my own projects, is that (last time I checked) hg doesn't support multiple branches inside of a single repo/working copy	04:10
	so, if you want to branch for a new feature, you have to do a new clone in hg	04:11
	(again, my knowledge may be out of date!)	04:11
	whereas, in git, branching is so lightweight you may think that it didn't actually happen because it happens so fast	04:11
sempuki	the impression I had is that git and hg were rapidly approaching each other, and that if they arent the same now, they will be soon	04:14
	except one is in C the other python...	04:15
	but thats just an impression of a user	04:15
clee	sempuki: well, C/shell/perl/etc	04:15
sempuki	clee: and python is mostly implemented in terms of C is it not? :)	04:16
clee	yep	04:17
	well, except for IronPython and JPython.	04:17
sempuki	true	04:17
	Debolaz: go with whatever your friends/colleagues like/know best, or has the best documentation, because the hardest part is learning how to use it correctly	04:18
Debolaz	Still, what I want is a comparison of technical qualities, not personal preferences.	04:19
	Performance, features and maturity of features, etc.	04:20
clee	performance is basically a wash, though in a few cases I think git is still faster	04:20
	however, I heard that hg could compress version history slightly better than git at one point	04:20
	for most projects that doesn't really matter too much	04:20
	they're both far more disk-efficient than svn	04:20
	← agib left	04:26
	← \|kukks\| left	04:27
	→ k_dub joined	04:28
	← k_dub left	04:29
	→ tongueroo joined	04:30
imyousuf	vmiklos: I submitted another patch (recursion over submodules) today	04:30
	→ wayneeseguin joined	04:31
imyousuf	did you read that?	04:31
	← KirinDav left	04:31
imyousuf	vmiklos: btw thanks for the hints	04:33
	→ jesse_ joined	04:40
	← yeowman left	04:41
	← jesse_ left	04:46
	← bartman left	04:46
	← FunkeeMonk left	04:47
imyousuf	If I want to prepare a patch does the patches has to be with respect to a particular GIT version (e.g. 1.5.3.7) or git.git repo?	04:49
aeruder	imyousuf: see SubmittingPatches	04:51
	→ jesse_ joined	04:53
	jesse_ → anotherjesse	04:53
	→ agib joined	04:53
	← anotherjesse left	04:53
	→ bartman joined	04:55
	→ sgrimm joined	05:00
imyousuf	thank you aeruder	05:00
	→ fhobia joined	05:03
	← stick left	05:08
	→ cmacleod joined	05:09
	← cmacleod left	05:09
	→ bartman` joined	05:19
	← bartman left	05:19
	← zed_ left	05:20
	← RobotDeathSquad left	05:24
	← rkaway left	05:28
	← fhobia left	05:31
	→ rkaway joined	05:31
	→ kanru joined	05:32
	→ RobotDeathSquad joined	05:36
	← RobotDeathSquad left	05:36
	← YGingras_ left	05:39
	← bigred-ian left	05:43
	→ YGingras joined	05:47
	← YGingras left	05:50
	→ YGingras joined	06:01
	← tongueroo left	06:11
	→ fhobia joined	06:11
	→ FunkeeMonk joined	06:22
	← vbgunz left	06:29
	→ mediogre joined	06:30
wagle	how do you start a new root commit in a repository? alternatively, how do you get rid of a root commit, and make its child the root commit in a new (rebased?) branch of commits?	06:36
	→ TuxCM- joined	06:42
gitster	(1) you don't. you may have an independent history that originates in a totally separate repository and fetching from that repository may result in your repository having more than one root, though.	06:45
	(2) unfortunately you cannot. Such an initial mistake in an empty repository needs to be redone from scratch. This we may have to fix it, though. --- come to think of it, I suspect you may be able to "git commit --amend" the root commit, which would have the same efffect, though.	06:47
	→ vbgunz joined	06:50
imyousuf	after generating the patch using git format-patch -M <commit_sha1> how do I add it to the email? simply copy and paste?	06:58
	← TuxCM left	06:58
Tene	imyousuf: attach it	07:03
imyousuf	Tene: thanks, will attach the .patch file	07:03
Auris-	some how I've understood that you should not attach because then people cannot quote lines from the patch when they are commenting.	07:06
wagle	ok.. heres the problem.. back when i barely knew git, I wrote some scripts in the root directory of a git-repository. first thing i did was to branch from the remote commit. second thing i did was to delete everything, and commit that. after that, I started programming and committing. now we want to put some of those scripts into a completely different git-repository. i was able to fetch the branch as a branch with a new root	07:06
	, but need to get rid of the root and maybe its child before i can merge it. this is also an exercise is learning how to do hard/clever things with git.. 8)	07:06
floppyears	hi guys	07:07
imyousuf	Auris-: in that case how should I add the patch?	07:07
wagle	it could be done with cherry-pick if i could cherry-pick a new root, or cherry-pick from another repository	07:08
floppyears	if there are files with certain extensions that I want git to ignore can I use one global .gitignore file or do I have to have 1 .gitignore file in each directory ?	07:08
Auris-	imyousuf, I'm not really sure, sorry. I think there are guidelines at a website.	07:08
wagle	floppyears, i think a careful reading of "man gitignore" will answer your question	07:10
gitster	imyousuf: your earlier question about sending patches. Either read in your MUA (the procedure depends on the client, SubmittingPatches has some hints but I recently noticed that the kernel folks have more extensive insns in their Documentation/email-clients.txt file --- we may want to snarf it someday)	07:10
	or "git-send-email".	07:10
	→ devogon joined	07:12
floppyears	wagle: thanks. I already had the .gitignore setup to ignore the files I don't want. These files in the index I guess were just old	07:12
gitster	floppyears: ".txt" in .gitignore at toplevel will ignore all .txt files (e.g. foo.txt, bar/boz.txt). "/.txt" in .gitignore at toplevel will tell git to ignore all .txt files at the toplevel (e.g. foo.txt is ignored, bar/boz.txt is not). ".txt" in bar/.gitignore will tell git to ignore all .txt files in bar/ and down below.	07:12
	Quiz. What does "/*.txt" in bar/.gitignore ignore?	07:13
imyousuf	gitster: thank you, I will first try git-send-email and then read the kernel documentation. I will send the first email to my self for safety :)	07:21
	← agib left	07:23
	→ harinath joined	07:24
	Toad → NotToad	07:26
	NotToad → Toad	07:27
Auris-	gitster, ignores *.txt in bar/ but not down below?	07:30
	→ Dodji joined	07:31
gitster	Auris-: correct.	07:34
Auris-	yay.	07:35
	dwmw2_gone → dwmw2	07:36
wagle	how lame.. you can fetch into a brand-new repository, but you can't cherry-pick from it	07:40
floppyears	hi guys	07:45
	I use mantis to keep track of bugs	07:45
	Is there a way to integrate the 2 ? Meaning if I commit a message, and I specify the mantis bug #, it would then add a note or some info to mantis ?	07:46
	metze_away → metze	07:47
gitster	Will _all_ your commits be created locally (or at least on the machine that has access to your mantis installation)?	07:48
	If yes, probably post-commit hook would be the thing to use for such a hack.	07:49
	thiago → thiago_home	07:49
floppyears	gitster: awesome thanks gitster	07:50
gitster	I would not recommend that myself, though.	07:50
	I am assuming that you would have a centralized installation of mantis (perhaps it is not even a "distributed bug tracking system"). A better alternative is probably to have a repository on the same machine as you have that installation, and drive that "commit to mantis" from post-receive hook, after you push into that repository.	07:52
	That way, you can experiment all you want, even with a commit log message "This fixes bug #2326" and you will be safe until you actually push it out, even if you later find out that the fix was faulty and did not really fix it. The commit log while you are experimenting, rebasing and amending won't contaminate your mantis tracker.	07:53
halorgium	floppyears: you'd want to have a simulated centralised repo which drives the mantis integration	07:54
floppyears	halorgium: that's pretty much what I have. I do development on my ubuntu server which has the central repo of my code and also has mantis installed	07:55
	← mithro left	07:55
gitster	Then post-receive hook on that central repository is the way to go.	07:55
floppyears	thanks guys	07:56
imyousuf	gitster: is there a significant difference between `sed -n -e 's/^path = //p' .gitmodules` and `grep "path =" .gitmodules \| awk '{print $3}'`?	07:56
gitster	I think our scripts avoid awk.	07:56
imyousuf	ah I see	07:56
gitster	But not mergetool ;-)	07:56
	I should rewrite that someday.	07:57
imyousuf	and how can I use expr ' '.* instead of expr index ' '?	07:57
gitster	also yours have two processes piped together when one sed is enough.	07:57
imyousuf	sorry not how can but can I?	07:57
gitster	What are you trying to achieve?	07:57
	I do not like the way you quoted the string for shell safety to begin with.	07:58
imyousuf	finding whether there is space in the string	07:58
gitster	I feel much safer if a script does (1) replace all occurences of single quote with '\'' (sq bq sq sq) and (2) surround the result with single quote pair.	07:59
	See the comment at the beginning of quote.c.	07:59
imyousuf	ok	07:59
gitster	There may be 47 other ways to safely quote strings, but if the code sticks to one style that is known to work, it is less work for reviewers and people who later want to muck with the code.	08:00
imyousuf	I fully agree with it thus asked it	08:01
jeffpc	gitster: awk '/^path =/{print $3}' < .gitmodules	08:01
	:)	08:01
imyousuf	actually if -m "test commit" is passed as argument then when evaluating the git-commit I should send -m "test commit" instead of just commit	08:02
gitster	imyousuf: The first message sent to the list before you started to show your code to the list made it sound as if you are near-illiterate in shell scripting, but I found your scripts reasonably well disciplined.	08:02
imyousuf	that is why I did quoting	08:02
	I know reasonable amount of scripting but not as good as you guys	08:03
	I needed to learn from you all actually :)	08:03
gitster	Do you mean 'git submodule recursive commit -m "test commit"' results in 'git commit -m "test commit"'?	08:03
imyousuf	as I mentioned earlier I am primarily a java developer and you all have helped me a lot thanks for that	08:03
	not it does so, yes gitster	08:03
gitster	(all these in-sentence quoting makes this hard...)	08:04
imyousuf	yes, that is why I needed to wrap the test commit before eval	08:04
gitster	If you have the remainder of the command line in $@ (by shifting away "submodule", "recursive"), you can do eval '"$@"' ;-)	08:05
	without any of your own quoting.	08:05
imyousuf	I see	08:06
	I thought that might not work actually :)	08:06
	will give it a try	08:06
gitster	again, lack of in conversation quotes makes it harder to read it, but what I meant is you literally tell eval to evaluate dq dollar at dq.	08:06
	← madewokherd left	08:07
gitster	Or at that point, perhaps even without eval. Just "$@" ;-)	08:07
imyousuf	got it	08:07
	will try both and keep the shorter one :)	08:07
	← kanru left	08:08
gitster	shell knows its own syntax the best. The best strategy to safely quote strings is avoid having to quote at all.	08:09
wagle	wow.. you can merge from a branch from a different root, and it does what you want	08:10
gitster	pretty cool, isn't it?	08:12
	← anttih left	08:12
wagle	i was starting to expect that i'd set myself on a wild goose chase with these multi-root things, and that you probably couldnt merge from two different roots	08:13
	(why i used "want" instead of "expect" above)	08:13
gitster	what are you trying to achieve with the multi-root thing, may I ask? We have it primarily as a way to fix earlier mistakes of not starting two things that should have been done in the same history.	08:14
wagle	merging two different projectsw	08:14
	merging two different projects	08:14
	← fhobia left	08:14
	→ srid joined	08:14
	→ Zhou joined	08:15
wagle	the first complication was that i forked from another project first, and wanted to remove that historical baggage	08:16
	→ aarapov joined	08:23
	← sgrimm left	08:24
	← mediogre left	08:24
imyousuf	After editing a file if I do git-reset HEAD will revert my changes?	08:27
	→ marcel joined	08:28
imyousuf	*will it	08:28
Mikachu	only with --hard	08:29
	git reset on its own just resets the index	08:29
	so it will undo any git add you've done	08:29
	← harinath left	08:29
imyousuf	I see	08:30
	so git-reset --hard HEAD will revert changes, thanks Mikachu	08:30
	← Zhou left	08:33
	→ \|zeal\| joined	08:35
	→ Oejet joined	08:38
	← johan-s left	08:45
	→ rphillip1 joined	08:46
	→ thiago joined	08:48
	→ Rogan joined	08:56
Rogan	Hi folks, I am dealing with DOS-format files - i.e. that end with CR/LF. I have a bunch of files with trailing whitespace (apart from the CR) that I'd like to clean up automatically. However, git-apply --whitespace==strip also takes off the CR, which I don't want	08:57
	any idea how to strip trailing spaces, but not the ^M?	08:58
Mikachu	i'm not sure, but see the crlf stuff in man gitattributes	08:58
	← rphillips left	08:59
Rogan	Also, is it possible to get cherry-pick to do this automatically for me? At the moment, I am dinking around with doing the cherry-pick, resolving the conflicts, making a patch, reset --hard, apply the patch with whitespace=strip, then committing	08:59
	core.autorclf doesn't offer any solution for folks that actually WANT CRLF in the repo	09:01
Mikachu	ah, okay	09:02
Rogan	maybe the filter option would be what I'm looking for. Does anyone have any idea how it is supposed to work?	09:07
Mikachu	you could just run something like sed -i 's/ $//g' .c or whatever	09:07
	→ TuxCM joined	09:12
	→ twork__ joined	09:14
twork__	hi - git-describe will tell me the commit reachable to a tag, but is there a way to indicate whether a file is reachable to a tag (i.e., git-describe for files?)	09:15
	← TuxCM- left	09:17
	→ octopod joined	09:17
Rogan	twork__: I'd say that that would be a lot harder. You'd need to figure out which commits that file appeared in first, and then continue as usual	09:17
gitster	"git ls-tree -r $commit \| grep $file" ;-)	09:18
Rogan	well, for i in <commits>; do git ls-tree -r $commit \| grep $file ; done	09:19
	or thereabouts	09:19
twork__	i'm trying to emulate cvs tagging in git -- or maybe i'm going about it the wrong way.	09:20
gitster	what's being asked for? The first time a path appears in the chain of commits?	09:20
Rogan	yes, I think you are	09:20
	see gitattributes: $id$	09:20
	this discussion has come up repeatedly. General emulation of CVS tagging doesn't make sense in git	09:20
gitster	If you want to know the boundary commit whose parent does not have a path $path but the commit itself does, you would do "git log --diff-filter=A -r -- $path".	09:20
	→ johan-s joined	09:21
Rogan	gitster: you implemented smudge and clean filters in April last year, and I am trying to use them to allow me to maintain a CRLF ended file in my repo. How can I tell if the filter is actually being called?	09:22
gitster	How can you tell? Perhaps add a debug printf in your filter script? Run git under gdb?	09:22
Rogan	Ok, I'll try a debug filter that calls the real one	09:23
gitster	About the "fiters" there is one caveat. If you define a filter _after_ you check things out, "git checkout" won't change a thing. You would need to do something to cause the actual checkout to happen, as we try very hard not to check out things that did not change between the filesystem and the index.	09:24
imyousuf	gitster: making the submodule command change under master - ok?	09:24
Rogan	yes, I realise that. My checkout filter is just "cat". I have specified unix2dos as the checkin filter	09:25
gitster	imyousuf: ?	09:25
	you do not have to define them as pairs (cat can be omitted, that is).	09:26
Rogan	right, ok	09:26
imyousuf	gitster: introducing variable command to be used instead of case "$add,$init,$update" .... in git-submodule.sh	09:27
gitster	imyousuf: sounds like a useful clean-up.	09:27
	post 1.5.4, that is.	09:27
imyousuf	yes, you said that to me over email and I agreed	09:27
	not a issue when it will be released	09:28
Rogan	it seems that git-config doesn't throw an error if the config file is not writable	09:28
	so my attempts to set up the filter were silently ignored	09:28
imyousuf	so before resending my patches I will first send that, gitster. Will doing the changes in master branch	09:28
Rogan	which explains why the filter was not executed	09:28
gitster	Heh, that's a nice bug. I personally almost never use git-config for writing as I do not trust that code X-<. Patches would be very welcome as we are really trying to stabilize things before the release.	09:29
	→ theCarpenter joined	09:30
Rogan	to test: chmod -w .git/config; git config filter.mycrlf.clean=unix2dos; echo $?	09:30
	returns 1, but no error message is shown	09:30
	sorry, my C is non-existent	09:30
	shall I send a mail to the list?	09:30
gitster	That sounds like something else. I do not think we write things in place.	09:30
Rogan	hmmm, you are right, I think	09:31
gitster	That is, mode of .git/config itself should not matter, as we do (1) create .git/config-new and write into it, and then (2) rename .git/config-new to .git/config, or something like that.	09:31
	Unless .git/ itself is unwritable for you, but that sounds utterly implausible given you have bunch of stuff in your repository.	09:32
	Ah, returns 1 but no error message is shown.	09:33
	← aarapov left	09:34
	← sempuki left	09:34
Rogan	.git is definitely writeable, and actually, so is .git/config. Was my mistake there. But I still get the $1 return code, and no changes to the config file for some reason	09:35
gitster	Yup, can reproduce. I'd dig this if I have time sometime but not tonight (past midnight here)	09:35
	Thanks for reporting.	09:35
Rogan	off to bed :-)	09:35
	no problem. Thanks for all your work on git	09:36
	→ saimon joined	09:37
gitster	Heh, that's not a bug but PEBCAK. You fooled me.	09:38
	"git config filter.mycrlf.clean unix2dos"	09:39
	There is no = sign there.	09:39
Rogan	ah	09:39
	doh	09:39
gitster	doh, I did not notice myself (I said I do not use that, didn't I?)	09:39
Rogan	so the $?=12 is because a key was not found	09:39
	er, $?=1	09:39
gitster	There should be a better error message.	09:39
	So it is still a bug but is a bit different from what we originally thought.	09:40
	→ namenlos joined	09:41
Rogan	ok, I have the filters executing, but I'm obviously doing something wrong	09:42
	← sverrej__ left	09:43
Rogan	I have no smudge filter defined, but if I delete my working tree and force a checkout (git reset --hard), my entire working tree is shown as modified, and the CRLF has been removed from the files	09:43
gitster	t/t0021 may be of help of you have the source handy.	09:43
Rogan	ok, will check it	09:44
	Thanks	09:44
gitster	If you have clean filter that is unix2dos (does that mean you _WANT_ CRLF in the objects, not just checkout???), and cat as smudge (or no smudge), then...	09:44
	→ BHoltzman joined	09:45
	← BHoltzman left	09:45
gitster	reset --hard would read objects, and write it out as-is to your work tree.	09:45
twork__	Sorry for the delay -- what i'm after is for a some file, doesn't matter what it is -- it could be a collection of files to know whether or not it has been "tagged" (like you have in CVS). The functionality of git-describe using the SHA1s is fine for this, I'd just want to be able to say something like: git describe -- ./* and for all files in there to be looked against to see if they reached a tag or not.	09:45
	→ kanru joined	09:46
gitster	Then, when comparing the index with the work tree, the work tree is converted through smudge (so it will modify LF in work tree to CRLF) and then compare what is recorded in objects and the index. Maybe you have LF endings in the repository?	09:46
	That would explain "everything changed", because everything did. Your next check-in will make a tree full of CRLF terminated files.	09:47
imyousuf	If I have changed multiple files and want to revert the change of 1 of them, how should I do it?	09:48
gitster	"files" in the sense that the blobs you track from then on will be terminated with CRLF (which is an odd wish, though).	09:48
	"git checkout that-path this-path"	09:48
imyousuf	thanks gitster	09:48
gitster	that is before "git add that-path this-path". If you have done that,	09:48
	"git checkout HEAD that-path this-path".	09:48
imyousuf	got it	09:49
gitster	Read these as "check out that-path and this-path from the index", and "check them out from HEAD commit".	09:49
	This form is quite different from "switch to that branch: git checkout that".	09:49
imyousuf	ah thats nice	09:50
	→ mediogre joined	09:50
imyousuf	is git-add changed-file encouraged or git-commit -a?	09:50
gitster	Many new people actually find these two completely separate semantics confusing.	09:50
	← wagle left	09:51
gitster	depends on the way you work. both are equally supported.	09:51
imyousuf	I see, which would you consider as a better practice?	09:52
gitster	what's frowned upon by some people, although we do try supporting it equally well as other ways, is "git commit this-path that-path" (partial commit).	09:52
	all three have their uses.	09:52
	I do not think there is "better practice" vs "not so good one".	09:52
Rogan	gitster: Yes, I WANT CRLF endings in the repo, since I am interacting with an SVN repo operated by someone else	09:53
imyousuf	ah ok, thanks for the detailed explanation gits	09:53
gitster	Rogan: quite reasonable.	09:53
imyousuf	* gitster	09:53
Rogan	however, I didn't realise that when checking the status it would run my clean filter	09:53
	does that really make sense?	09:53
gitster	Of course. Otherwise, it wouldn't be "cleaning".	09:54
	What you are seeing is an inevitable glitch in the priming stage, if I am not mistaken. You have your objects without CR and you are trying to fix, aren't you?	09:55
	Debolaz ponders writing up a comparison of git and mercurial, not finding preexisting with any informative content of recent origin.	09:55
Rogan	no, the repo actually contains CRLF terminated files	09:55
	all I really want to do is make sure that all the changes that I commit maintain that CRLF ending	09:55
gitster	Maybe I am not getting what you are doing then. Your work tree files have CRLF?	09:56
	After reset --hard, that is.	09:56
Rogan	in the face of stripspace=fix which takes the CR away at the same time as it takes trailing spaces and tabs	09:56
	yes, they do	09:57
gitster	Maybe your unix2dos is not doing unix2dos? Does it take options (I do not know offhand how we are handling custom commands with options)?	09:57
Rogan	well, this is what I actually have	09:58
	vi -b file <- shows ^M	09:58
	cat file \| dos2unix \| unix2dos > /tmp/t	09:58
	diff file /tmp/t	09:58
	no differences	09:58
gitster	I am assuming you are on Unix?	09:59
Rogan	however, git diff file shows -...^M, +...$	09:59
	yes	09:59
	where $ is just the end of line	10:00
gitster	That's interesting.	10:00
Rogan	I run dos2unix first to get to a poin where the whole file is in unix format, rather than some lines with CRLF and some lines without (thanks to apply --stripspace)	10:01
imyousuf	can I added a body to an email which I am using to send out patch using git send-email?	10:03
	*add a	10:06
	→ a-atwood joined	10:08
arw_	imyousuf: --compose i guess	10:09
imyousuf	ah thanks arw_	10:10
arw_	or send an email before the patch series and pack the message-id into --in-reply-to	10:10
	→ janm joined	10:10
imyousuf	missed that in the option in the manual in midst of the rest :)	10:10
	→ chris2 joined	10:12
twork__	Is there anything in git which can take a path and spit out commits relative to that path?	10:14
DrNick	define "relative"	10:15
twork__	Well, commits that contains changes to files under that path.	10:16
gitster	"git log -- path"?	10:18
	if you want only commit object names substitute log with rev-list.	10:19
twork__	git cherry branch1 branch2 gives a list of SHA1s -- I'd like a list of files that are in that SHA1.	10:20
gitster	do you mean s/that SHA1/those commits/?	10:21
twork__	Yes.	10:21
	[thomas@ubuntu smoothwall]% git cherry guardian-update1 master	10:21
	+ c8fd4f01bb97df3500e4d03abbf42832b8593211	10:21
	+ 236c065c3a37f5d961aac5584594dd3937d62798	10:21
gitster	"for c in these-commit-object-names; do git-ls-tree -r $c; done"?	10:22
twork__	So given those commits, what files are in them?	10:22
	Ah ha.	10:22
	twork__ tries.	10:22
twork__	:) Thank you.	10:22
	→ reval joined	10:22
gitster	but is that what you really want?	10:22
	you will get tons of duplicates because in a usual history most of the files stay the same between commits.	10:23
twork__	Not quite what we want...	10:24
	← dysinger left	10:24
twork__	for instance, git ls-tree d5442 lists tonnes of stuff, yet the information from git show d5442 shows only one file -- it's that information I'd like to get.	10:25
gitster	so what are you trying to achieve? for example, "git log --name-only --pretty=format:%H -m $from_this_commit..$to_that_commit" would show only the names of paths that were affected by each commit.	10:26
	→ sverrej joined	10:26
	← YGingras left	10:26
	→ YGingras joined	10:27
	← Catfish left	10:27
	← Sho_ left	10:30
Rogan	ok, I've got myself into a VERY weird situation. I've previously resolved a merge, and saved it as a patch. It modifies one file (.java), and creates 2 others, one java and one html. I apply it with git-apply --whitespace=strip, and get no errors, other than whitespace warnings. "git status" shows the modified file, and the new .html file, but does not show the new .java file	10:34
	however, the file is there. Even if I explicitly add it, git refuses to acknowledge its existence	10:35
Mikachu	twork__: maybe you can do git show abc123\|lsdiff	10:35
	(depends on patchutils)	10:36
Rogan	I have checked my .gitignore, and it doesn't match anything there	10:36
	← YGingras left	10:37
	→ YGingras joined	10:37
maks_	hmm if i fetch somebody pull request where do i see those commits?	10:38
	← reval left	10:38
Mikachu	maks_: if you use 'git pull' you will merge their branch into your currently checked out branch	10:39
	'fetch somebody pull' doesn't quite parse here	10:39
maks_	yes i know the pull part Mikachu	10:40
	Mikachu: you get a pull request to merge someones branch, but do a s/pull/fetch/	10:40
Mikachu	if you run 'git fetch git://server.org/foo.git' you will fetch all commit objects and blobs, but you won't get a reference in any branch	10:40
maks_	where do i see those objects?	10:40
Mikachu	you have to specify where to store it	10:41
	git fetch git://blah master:masterfromotherserver	10:41
maks_	thanks Mikachu	10:43
gitster	With "git fetch $elsewhere; git log ..FETCH_HEAD", you do not necessarily have to ;-)	10:47
Mikachu	aha, so that's where it ends up :)	10:47
Rogan	any ideas why git might be ignoring my file? (I did check .gitignore)	10:49
Mikachu	did you check .git/info/excludes and ~/.gitignore ?	10:50
Rogan	exclude (no s) is the default with everything commented out, and there is no ~/.gitignore	10:51
Mikachu	oops	10:53
	i usually just press tab after the e :)	10:53
maks_	git log ..FETCH_HEAD is empty	10:54
	← imyousuf left	10:55
	→ imyousuf joined	10:55
	→ koki joined	10:57
maks_	hi koki :)	10:58
	← srid left	10:58
koki	maks_: hey!	10:58
	→ saimon_ joined	11:02
	→ srid joined	11:03
	← saimon left	11:05
	→ Kebianizao joined	11:08
Kebianizao	Hello	11:08
	I've done a git status and I have plenty of modifications after applying a patch	11:08
	← namenlos left	11:08
Kebianizao	I have all the files I want to git add listed in a file and I would like to feed git add with that file	11:09
	how could i do that?	11:09
Mikachu	if there are no spaces or funny business in the filenames you can probably get away with git add `cat file`	11:11
imyousuf	how can I specify a SMTP username and password in the git-send-email command?	11:12
twork__	after creating a branch and performing some commits and/or merges, how can I find the commit on master from which I originally branched?	11:14
	→ octopod_ joined	11:14
	← imyousuf left	11:14
Mikachu	git merge-base branch master	11:16
	why do you want to find it?	11:16
twork__	nearly what we wanted	11:17
Mikachu	you could just bring up gitk --all and click on the one you want :)	11:18
	→ jsimonss joined	11:18
twork__	basically, i have two branches created from master, but from different commits. when i merge from the second-created branch to the first-created branch, it has (logically) also merged commits on master which took place in between the two creations	11:19
Kebianizao	Mikachu: thanks it worked	11:19
twork__	what i was trying to do was merge just those commits on the second-created branch post creation of the branch	11:20
Mikachu	you would have to use rebase i think	11:20
Tv	twork__: the only way to do that is to first rewrite the commits on top of the new location -- aka rebase	11:20
Kebianizao	bye	11:20
	← Kebianizao left	11:20
twork__	before i can start trying to understand that (which looks like a brilliant suggestion), I need to revert the ballsed merge. also, i'll then want to push the resulting merged branch out to a shared repo	11:22
	which may or may not be pertinent	11:23
	the immediate problem is i can't find the SHA1 of the commit I originally branched from (which is what I want to revert to). merge-base isn't quite there, because of the commits on master which I (unintentionally) merged.	11:24
Mikachu	right, you'll have to reset --hard HEAD\^ or something first	11:24
	(don't do that on the wrong branch or twice)	11:25
	or you could just pass HEAD^ to git merge-base	11:26
	← octopod left	11:28
vbgunz	I am on part 2 of a tutorial. I was told to do this > '$ echo "hello world!" > file.txt' but got this returned > 'bash: !": event not found' why?	11:30
	in single quotes it seemed to work just fine...	11:31
broonie	! tells your shell to recall the last command with the prefix in front of the !.	11:31
	(so if I say 'ls ~' then 'ls!' it will do ls ~ twice)	11:32
	octopod_ → octopod	11:34
	→ captnmark_ joined	11:35
	→ ivazquez joined	11:37
ivazquez	Is there any way to add arbitrary metadata to a blob?	11:38
Mikachu	you can tag it, but you can't reach the tag from the blob, only the other way	11:40
vbgunz	Mikachu: I decided to try it, so I said "ls ~" and then tried all of the following:, bash: ls!: command not found, ls: !: No such file or directory, bash: syntax error near unexpected token `newline'	11:41
	heh, nothing repeated the command at all :/	11:41
Mikachu	wrong tabcomplete there	11:41
	→ ferdy joined	11:42
ivazquez	Unfortunately that doesn't help. I need it reachable via the blob.	11:43
Mikachu	i think that's not possible, since existing objects are read-only	11:43
twork__	how do i delete a branch from a remote (shared) repository? such that nobody can pull from it any more, etc. (yes, this is just a testing repository. :) )	11:45
Mikachu	push remote :branchname	11:45
twork__	it's not having any of that	11:48
	... or is it	11:49
	← captnmark__ left	11:52
	← jsimonss left	11:54
	→ jsimonss joined	11:54
	← \|zeal\| left	11:54
	→ lcapitulino joined	11:59
vbgunz	I have an offtopic question. hopefully, you guys understand. when I open a terminal session, is there any way I can create a personal .bash_history for it? e.g., .myGitHistory ?	12:05
Tv	vbgunz: help history	12:07
	← G0SUB_ left	12:07
	→ wagle joined	12:08
	→ IanL joined	12:14
	→ mithro joined	12:15
	→ YGingras_ joined	12:21
	← YGingras left	12:22
	→ ivazque1 joined	12:27
	→ namenlos joined	12:28
	← bartman` left	12:29
twork__	Thank you all.	12:30
	← twork__ left	12:30
	→ bartman joined	12:30
johan-s	hey folks, I wrote yet another git hosting thing: http://gitorious.org/	12:35
lu_zero	johan-s what's that?	12:35
johan-s	lu_zero: hosting git repositories and their clones basically	12:37
	← srid left	12:37
Tv	forgeries, eh?	12:37
	← theCarpenter left	12:37
johan-s	Tv: I'm not even sure thats a valid gramatical form of "forge"	12:38
lu_zero	sounds nice	12:38
johan-s	haha, forgeries means something else entirely of course, d'oh	12:39
Tv	bah yet another fake-linear history browser	12:39
johan-s	Tv: yes. the browser sucks, I want it more gitk like	12:39
lu_zero	Tv how's going with gitosis?	12:39
Tv	lu_zero: it seems a lot of people are happy with it, which is good	12:40
	→ cedric_ joined	12:40
Tv	gitweb integration is somewhat of a pain, still	12:40
	lu_zero would make ffmpeg using it once the other issues against migration are solved	12:40
lu_zero	noticed ^^;	12:40
Tv	because of the security barriers	12:40
	nothing a manual chgrp won't fix, but annoying	12:40
	→ jengelh joined	12:41
	← ivazquez left	12:42
	ivazque1 → ivazquez	12:42
lu_zero	I see	12:43
	← ironfroggy left	12:44
Tv	the thing is, gitosis by design can't do that by itself	12:45
	it isn't allowed to ;)	12:45
	i could automate the case of gitweb running as the same user	12:46
	but i'm not sure if that's common enough	12:46
lu_zero	Tv otherwise you'd need an helper	12:47
Tv	to chgrp? maybe	12:47
	but then, if you trust gitosis to chgrp things, that is to allow their web publication	12:48
	you might as well run gitweb as the same user	12:48
lu_zero	yup	12:48
Tv	i think the same-user-id model will get packaged in a nicer way	12:49
	and if you're not happy with that, you get to chgrp individual repos manually	12:49
	that's the best i can do, really	12:49
	← chris2 left	12:50
MadCoder	git commit -a -s -m'no we shouldn''t be removed'	12:52
	fatal: Out of memory? mmap failed: Bad file descriptor	12:52
	Created commit	12:52
	WTF ?!	12:53
Tv	MadCoder: reproducable?	12:53
jengelh	Is there a chance to do two stash applys?	12:53
Tv	jengelh: if nothing else, via git apply	12:53
MadCoder	Tv: not really, though it happens on a regular basis	12:54
	it's annoying	12:54
	though the commit is here, so it's merely annoying	12:54
Tv	oh wow	12:54
	→ toxx_ joined	12:55
MadCoder	yeah	12:55
jengelh	Tv: git-stash apply 'stash@{2}' => Cannot restore on top of a dirty state	12:57
	→ WorkingGeier joined	12:58
Tv	jengelh: try git stash show 'stash@{2}' \| git apply --reject	12:58
jengelh	git-stash show only prints a diffstat, not the actual patch	12:58
Tv	oh crud	12:58
	well you can even git show it	12:58
	i don't have stashes handy for playing, right now ;)	12:59
jengelh	heh	12:59
	git log -p 'stash@{2}' did the right thing.	12:59
vbgunz	Tv: thank you so very much for pointing that out, you sort of killed 2 birds with a single stone regarding my history :)	13:01
	← johan-s left	13:02
	→ johan-s_ joined	13:02
Tv	stashes are just commits. well, two commits: index and working tree.	13:02
	← johan-s_ left	13:08
	→ leethal joined	13:08
	← YGingras_ left	13:09
	→ YGingras_ joined	13:09
	→ johan-s joined	13:13
	→ Sho_ joined	13:18
	← z3ro left	13:22
	← mattly left	13:22
leethal	why is git saying lib/ is ignored? http://git.pastebin.com/m56d717c4	13:22
	I didn't set any global ignores or anything either	13:22
	→ z3ro joined	13:22
	→ YGingras__ joined	13:23
Tv	leethal: is it empty?	13:23
leethal	no, it has another dir in it	13:23
Tv	leethal: is that empty ;)	13:23
	find lib \! -type d	13:23
leethal	oh hey, no we're getting somewhere	13:23
	didn't even cross my mind that it recursively checked for empty dirs	13:24
	thanks =)	13:24
	→ theCarpenter joined	13:24
	← YGingras_ left	13:24
leethal	while I'm at it - is there any best practice (pardon me if it's not a practice at all) to version control empty directories?	13:24
	→ EmilMedve joined	13:25
Tv	git only tracks files, by design	13:25
	→ pngl joined	13:25
Tv	you put an empty .placeholder or .gitignore file in dirs you absolutely must get committed	13:25
	or just create them in the makefile	13:25
pngl	is there any reason why a git push would work, update the remote git log, but do no actual change to the files? (I'm getting started with git)	13:27
Tv	pngl: don't push to non-bare repositories	13:27
	← namenlos left	13:27
Tv	pngl: the working directory will not be updated	13:27
	shared repositories should be bare	13:28
pngl	so I have to clone --bare my local rep, scp it and then push will work?	13:28
leethal	I'm pretty noob at git, only know the basic stuff - is it possible to create the dirs with git itself when someone clones the repos? with some hooks or something?	13:28
Tv	pngl: you can just create a new bare repo from scratch and push to that	13:28
	pngl: scping repositories is sort of weird	13:28
leethal	the makefile suggestion sounds clean 'n stuff, but this system doesn't have makefiles =/	13:29
Tv	leethal: nope	13:29
leethal	I see	13:29
pngl	oh thanks for that... yeah I'm sure I'm using a lot of homebrewed methods that have nothing to do with the standart way of doing stuff	13:29
Tv	leethal: you don't want people running arbitrary code on your system when you just grab their sources, you know..	13:29
	→ mattly joined	13:29
leethal	mm, good point	13:29
pngl	thanks again I'll make a new bare on my remote server (is scp just weird or actually bad?)	13:31
hliusv561	pngl: It is possible to dedicate some prefix for pushes so they won't wind up goin to currently checked out branch. One can even specify patterns at client side how branch names are mapped...	13:31
jengelh	why scp, just push ssh://	13:32
Mikachu	i don't think push can create a remote repo?	13:32
hliusv561	pngl: It does not check the stuff copied and when copying repositories with their working trees transfers extraneous data plus trashes the working copy caches (fixable with one command).	13:33
	← octopod left	13:33
Tv	Mikachu: except with gitosis ;)	13:33
	but still, create it first, push then	13:33
	bigon` → bigon	13:34
	→ jtoy joined	13:34
jengelh	git addiction?	13:35
pngl	alright I'll see how this goes but it seems that what I need is to spend more time reading the docs	13:35
jtoy	how can I take the changes that I currently have and put them into a new branch? I forgot to create a new branch for the new code I have added	13:35
	→ cmarcelo joined	13:35
leethal	would suggesting to add a newb-friendly error message when trying to add empty directories to the repos be a silly idea? like, 'git only tracks files, can not add empty directory to repository' or whatever, instead of the 'cryptic' stuff about the dir being ignored.	13:35
Tv	jtoy: already committed or not?	13:35
jtoy	Tv: not committed yet	13:36
hliusv561	Actually, there are at least two ways to rebuild those caches with just one command ('git status' (the recommended way), and 'git update-index --refresh' (invoking just the low-level operation)).	13:36
Tv	jtoy: git checkout -b newbranch	13:36
jtoy	Tv: thanks	13:36
LotR	hmm. is there a webpage for the git.vim script other than mailing list archives?	13:36
	→ ironfroggy joined	13:36
hliusv561	git update-index is mainly used in scripts or for doing some really weird stuff.	13:37
	← mithro left	13:39
	← theCarpenter left	13:40
	→ Catfish joined	13:42
	← leethal left	13:45
soma-	what trac plugin is preferred for git?	13:47
	← YGingras__ left	13:48
	→ YGingras__ joined	13:49
	cbrake_away → cbrake	13:53
	bigon → bigon`	14:05
	→ dancor joined	14:06
dancor	i'm going to put a hook in a git-svn repo to automatically commit to svn anything push-ed to it	14:07
	however, the svn repo also has pre-hooks that can fail	14:07
	so is the best way to use .git/hooks/pre-receive and have that fail if the svn commit fails?	14:07
	bigon` → bigon	14:08
hliusv561	dancor: pre-receive is IIRC invoked before data is transferred.	14:09
dancor	hliusv561: hm, do i want 'update'?	14:10
pngl	I'm lost. Or rather I don't get the concept of bare repository. Where are the actual files?	14:11
	→ bigred-ian joined	14:11
dancor	i really just need to: detect failure, make sure the repo stays clean, and notify the person trying to push	14:11
hliusv561	pngl: Mainly inside those big *.pack files in objects/pack?	14:11
	← sunblush left	14:12
hliusv561	dancor: AFAIK, update can print messages to client and can abort the update.	14:12
pngl	hm. alright. so without a bare repository, my git push doesn't actually update the files. with a bare one, there are no files. what if I simply want to push changes from my local machine to my remote web server?	14:13
hliusv561	pngl: There are some hooks that can be enabled. One of them is invoked after pushing. It could copy a snapshot of newest pushed stuff to apporiate directory.	14:14
	→ cartman joined	14:14
hliusv561	pngl: At least 'post-update' hook can be used for that.	14:14
dancor	hliusv561: for me? yeah post-update seems like it might work better with git-svn	14:15
	i might have to do a little bit more clean up on failure there	14:15
hliusv561	dancor: You can't abort from post-update.	14:15
pngl	sounds good, I'll take a look at this. Strange though that there is a need to add hooks for what seems like a very simple operation to me.	14:16
dancor	hliusv561: but i can reset and tell the user they got pwnd right?	14:16
hliusv561	dancor: Actually propagating the failure to client would be nicer thing to do.	14:16
dancor	hliusv561: do you think it will be possible with git-svn?	14:16
hliusv561	dancor: Not just spew errors to client (and have client think it succeeded).	14:17
	pngl: Also, some data can live in those files which have 38 characters of "hex garbage" as their names under objects/xy/, where x and y are hex characters also.	14:18
dancor	update has problems in that it is once-per-file but it's not easy (that i know of, but i would be surprised if it is..) to do svn hook checks without actually committing files	14:20
hliusv561	dancor: I think there was some 'all-branches-at-once' hook also...	14:21
	pngl: It is not simple, as that working tree must not be trashed.	14:24
	pngl: It would be simple operation only if that working tree was clean on all paths modified. Anything else and merging could wind up being disaster.	14:25
	← jtoy left	14:25
	→ bobesponja joined	14:26
	→ doener joined	14:27
	← dkagedal left	14:27
	← cartman left	14:30
	→ careo joined	14:31
	← saimon_ left	14:41
	← ironfroggy left	14:42
	← doener_ left	14:42
	→ ironfroggy joined	14:43
	→ jtoy joined	14:46
	→ aroben joined	14:46
	← aroben left	14:46
dancor	can i get like a 'git show' of the change coming in in pre-receive	14:49
Tv	dancor: just let the fetch happen, review remotes/* before you merge it	14:50
	oh sorry you meant on push	14:50
	why?	14:50
	the hook gets the necessary info in stdin to do that, if that's what you really want	14:51
dancor	Tv: i'm trying to put a hook in a git-svn repo to automatically commit to svn anything push-ed to it, but fail if the git svn dcommit fails	14:52
	this seems difficult because git svn dcommit only works on a clean index (after all the hooks are called)	14:53
peper	how can I rever N last commits to the workdir?	14:53
Tv	peper: man git-reset	14:53
dancor	peper: as a new commit, or pretending they never occurred?	14:54
Tv	peper: oh yeah that too.. there's like three different things you might be after	14:54
peper	the latter	14:54
dancor	then git reset	14:54
Tv	1) starting a new branch 2) undoing commits you haven't shared 3) cancelling commits you have shared	14:54
peper	2)	14:55
Tv	man git-reset	14:55
peper	but I want the changes from the commits to be in the workdir	14:55
	alright	14:55
dancor	Tv: for this hook, one idea i had was to pull the change out at the pre-receive stage and patch it to a separate repo and dcommit from there	14:55
	i can't think of anything less hacky	14:56
peper	ta	14:56
Tv	dancor: well dcommit is going to rebase it anyway	14:56
	dancor: so the push is going to be an utter disaster	14:56
	dancor: what the client pushed will not be what you want to end up with	14:57
	← bigred-ian left	14:57
dancor	Tv: so you think separate repo is necessary?	14:57
peper	btw. how can I revert a git add? (either of a new filename or just changes)	14:57
Tv	dancor: i have no clue what you're trying to build	14:57
	dancor: but in general, git commits intended for git-svn dcommit should not be shared	14:57
	dancor: that includes pushing	14:57
dancor	Tv: svn <-> git-svn repo <-> pure git repo	14:57
Tv	peper: just like anything else	14:57
	(16:55:35) Tv: dancor: but in general, git commits intended for git-svn dcommit should not be shared	14:58
	dancor: think about that	14:58
dancor	why not	14:58
peper	Tv: man what? :]	14:58
Tv	dancor: because dcommit will mutate them	14:58
	peper: man git-reset	14:58
	peper: if it's the whole commit you want to undo	14:58
peper	now I am talking about reverting index changes only	14:59
dancor	Tv: i see. it adds a git-svn-id line to the message e.g.	14:59
peper	i.e. I did git-add foo	14:59
	midgar → bdiego	14:59
peper	and, oops I didn't want to do that ;]	14:59
	→ kristoffer joined	15:00
Tv	peper: see git status output	15:00
peper	Tv: "to be commited"	15:01
Tv	peper: ... aaand?	15:01
	→ kukks joined	15:01
peper	er, ok. simple git-reset did the trick ;]	15:01
Tv	so you didn't read it, then	15:02
	oh well	15:02
peper	I did	15:02
	then run git-reset	15:02
	and read again	15:02
	→ dkagedal joined	15:02
dancor	Tv: but there are certain benefits from a more-git-like setup that i want such as disk savings cloning from one repo on each machine that developers are on	15:02
	but i can see it's going to necessarily be more hacky than i had imagined	15:03
Tv	dancor: you can want all you want ;)	15:03
dancor	i intend to do more than want	15:03
Tv	dancor: switch over, then ;)	15:03
dancor	Tv: in time, and iteration	15:03
	← IanL left	15:05
careo	question for any git-svn users out there: what's the suggested way to start a new branch in svn with git-svn? or should I just start a new one with subversion directly and just use it in git?	15:05
Tv	careo: yup	15:06
careo	Tv: cool. thanks	15:07
	careo heads back into the dark pit of despair that is his svn checkout	15:07
Tv	careo: you can do that with a single command, just use full urls	15:08
	no checkout needed	15:08
careo	ack. right.	15:08
	thanks again	15:08
	Tv: one last dumb question and I'm out your hair: that's done, git-svn fetch found it with the "found possible branch point...", and is now pulling down revision 1-now. is that normal, or is something in my git-svn setup wrong that it can't just get the history from disk?	15:13
Tv	not sure	15:14
	git-svn has changed since i last used it for real	15:14
	for real = more than read-only trivial use	15:14
doener	careo: for the branching stuff, I have a script here to handle that. http://git.pastebin.com/d387cb2c	15:15
careo	ok. no worries. I trust that it'll sort itself out after a good gc, even if it does have to go through this temporary insanity.	15:16
doener	careo: that should look for the last commit in you current git branch that exists in svn, and create a new branch starting at that commit	15:16
	s/you/your/	15:17
careo	doener: ahh good! thanks	15:17
jengelh	How would I make gitweb show the shortlog of all recent commits rather than just the ones on the current branch?	15:18
doener	still needs the svn executable though. If I knew perl, I'd add that to git-svn... ;-)	15:18
careo	heh. git-svn needs svn anyway iirc	15:19
doener	careo: it can get away with just the perl bindings AFAIK	15:19
careo	wait. no, I don't. it uses uses the bindings directly	15:19
	careo goes for more coffee to guard against further errors	15:20
	→ orospakr joined	15:21
	← sverrej left	15:21
doener	jengelh: http://git.kernel.org/?p=git/git.git;a=shortlog;h=--all	15:21
	jengelh: seems that you can pass arbitrary arguments via "h"	15:22
jengelh	o_O	15:22
	metze → metze_away	15:22
	→ zapnap joined	15:24
	← zapnap left	15:24
jengelh	well it does what I expected	15:26
	but is this also possible for the shortlog on the summary page?	15:26
doener	oh well, not arbitrary, there is some checking it seems	15:28
	→ sverrej joined	15:28
	← sverrej left	15:29
ironfroggy	i have a file that was changed, staged for commit, and then the changes reverted, but it still shows up under changes to be commited	15:29
	how can i just clear anything that is staged?	15:30
	→ sverrej joined	15:30
	→ rmh3093 joined	15:30
kelvie_	it is safe to remove the tmp_packs right?	15:30
	→ YGingras_ joined	15:31
	← sverrej left	15:31
kelvie_	ironfroggy: git-reset HEAD <paths>	15:31
	or use git-gui	15:31
	→ p4tux joined	15:31
doener	jengelh: my perl sucks, but HEAD seems to be hardcoded in git_summary in gitweb	15:31
jengelh	i don't mind editing gitweb.cgi	15:32
	changed and works ;-)	15:32
	→ sverrej joined	15:33
	← kanru left	15:33
doener	jengelh: now using $hash instead of 'HEAD', I guess?	15:34
	← sverrej left	15:34
ironfroggy	kelvie_: thanks	15:34
jengelh	my %co = parse_commit("--all");	15:35
doener	ah, hardcoded again :)	15:35
jengelh	no chance for sneaky kiddies	15:35
	what's a fork?	15:35
eparis	i have 1 'public' branch and 10 'devel' branches each with 1 commit and a very nice commit message. git checkout public; git pull . devel-1; git log looks great; git pull devel-2; git log for the latest merge is crap: "Merge branch 'devel-2' into public" rather than the nice commit message I made when I created the branch. how do i make all these devel branches merge into my public branch and keep the nice commit messages?	15:35
rmh3093	is there a way to add custom dns info to a repo hosted by repo.or.cz	15:37
	i have a webredirect working in my dns	15:37
	which works ok to redirect cloning	15:37
doener	eparis: your nice commit message is still there, you're just looking at the merge commit that pulls in the other commit	15:37
rmh3093	but i cant push to that	15:37
doener	eparis: have a look at your repo in gitk to see that	15:38
	eparis: what happened is that the first merge happened to be a fast-forward, so the branch was just moved forward and no merge commit created	15:38
	eparis: but the second one is not a fast-forward, thus you see the merge commit in _addition_ to the original commit	15:39
careo	doener: in script there is a line "URL=$(echo -n "$URL" \| sed -e 's!//.*@!//!')", svn yelles about the "-n" argument. without it, it works pefectly.	15:39
eparis	doener: hmmm ok, is there a way to not have that merge commit and just make it look like i did a patch->commit->patch->commit work on the 'public' branch (aka when i push this public branch i don't want people to have to see my internal workings....)	15:39
careo	in your script, rather	15:39
doener	eparis: if you want to avoid merge commits, you'll have to rebase the devel branch before merging it	15:39
	bigon → bigon`	15:39
jengelh	Is there any description of why gitk colors branches halfway one color and in the middle changes to another?	15:40
doener	eparis: or, if you really just have a single commit in those branches, just cherry-pick them instead of rebasing and merging over and over again	15:40
eparis	doener: even better. but i'll go read about rebasing just to learn	15:40
doener	careo: svn yells about -n? that's passed to echo!?	15:40
eparis	doener: thanks!	15:40
careo	doener: to svn cp	15:40
doener	careo: svn cp gets -r ..... -m, I don't see an -n here	15:41
	careo: there's a single -n and that's passed to echo	15:41
careo	doener: strange. here what's sent is: svn cp -r ... -n ...	15:42
	← YGingras__ left	15:43
doener	careo: hm, so it's "svn cp -r ... -n ... -m ..."?	15:44
careo	yeah	15:44
doener	careo: hm, does your "echo" not support -n?	15:44
	careo: what does "echo -n foo" print?	15:44
careo	doener: just foo, -n's suported	15:44
	→ cannonball joined	15:45
doener	where does that -n come from then? confused	15:45
careo	echo "$DST" => -n reponame	15:46
	I'm somewhat confused too. but either way, without -n it's running happily. I'm in os x leopard though	15:47
Rogan	has anyone here used git-cherry?	15:47
doener	careo: maybe your /bin/sh has a broken echo built-in?	15:47
Rogan	I'm trying to understand what it is telling me	15:47
jengelh	Is not `git add .; git commit` like git commit -a ;-)?	15:47
doener	jengelh: IIRC "commit -a" is like "add -u", while "add ." won't notice deletions	15:48
vmiklos	git add . will add new files while git add -u will just add modified files to the index	15:48
jengelh	good to know	15:49
careo	doener: by god, it does. sh-3.2$ echo -n foo => -n foo	15:49
	← johan-s left	15:49
doener	ah yeah, new files are handled differently, too	15:49
	← mediogre left	15:49
doener	careo: well, then make that #!/bin/bash or /bin/echo	15:51
	:-)	15:51
	→ octopod joined	15:51
	bigon` → bigon	15:52
	← spuk- left	15:53
	→ johan-s joined	15:54
	→ harinath joined	15:57
	← johan-s left	15:59
	→ johan-s joined	16:00
	→ spuk- joined	16:02
	← johan-s left	16:02
	→ johan-s joined	16:03
	→ harinat1 joined	16:08
	→ aroben joined	16:11
	← harinat1 left	16:11
	→ twork__ joined	16:13
twork__	Hmm -- I was hoping git cherry-pick would copy the same SHA1 over, but it doesn't. Is there any way of copying an entire commit verbatim, including the same SHA1? We're trying to perform merge-tracking.	16:14
	→ z3om joined	16:15
doener	twork__: no, you're changing the parent which affects the hash	16:15
Tv	twork__: commit sha1s are cryptographical checksums of everything in history before that point	16:15
	twork__: sounds like you want to study the concept of topic branches	16:16
twork__	Thank you. :)	16:16
careo	is there a way to have git-svn mark that branches in svn have been deleted?	16:17
	← marcel left	16:19
	→ sunblush joined	16:22
	→ stillLotR joined	16:24
	← LotR left	16:25
	← harinath left	16:25
	→ HG` joined	16:26
	→ mjc joined	16:27
	→ puk joined	16:27
	→ stick joined	16:27
mjc	hmm. we have a huge "images" directory in our cvs repo, and I want to make it a separate repo... is there a good way to do that? the history from it is huge	16:28
	← a-atwood left	16:28
	← priidu_ left	16:28
mjc	when I do the git import	16:28
	er git-cvsimport	16:28
Tv	mjc: you can always copy the cvs repo and manipulate the ,v files	16:28
	but maybe git-cvsimport can do stuff	16:29
mjc	that'll eliminate it from the history entirely? I'm not familiar with cvs's inner voodoo	16:29
	(thank god)	16:29
Tv	and if you import the whole thing you can use filter-branch to split them later	16:29
mjc	how's that work	16:29
Tv	cvs versions individual files only	16:29
mjc	filter-branch	16:29
Tv	ask the man ;)	16:30
	→ devurandom joined	16:30
devurandom	Hello!	16:30
mjc	ahh, here we go	16:30
	thanks tv.	16:30
devurandom	I am getting "Odd number of elements in anonymous hash at /usr/bin/git-svn line 1760." (output, no warning/error) during git-svn-clone. What does it mean?	16:31
	Is that something bad?	16:31
	I.e. may my repository be corrupted?	16:31
Tv	devurandom: could be	16:32
	devurandom: but probably not	16:32
	devurandom: it's a bug; try the latest version, if that doesn't help write a good bug report to the mailing list	16:32
	devurandom: oh and do try to google for the error message, first	16:32
devurandom	I am using 1.5.4_rc2.	16:32
	← pngl left	16:32
	← zeom left	16:33
	z3om → zeom	16:33
	→ aroben_ joined	16:33
devurandom	Seems as if it really could be a bug: http://www.nntp.perl.org/group/perl.bugmongers/2004/09/msg204.html	16:33
Tv	well it is a bug	16:34
	that's clear	16:34
	what's on line 1760 of your copy of git-svn?	16:34
	← aroben left	16:34
mjc	"To remove commits authored by "Darl McBribe" from the history" funny but I think that should be removed :O	16:35
	aroben_ → aroben	16:35
	← lcapitulino left	16:36
devurandom	$svnsync = {	16:36
	With context:	16:36
	# see if we have it in our config, first:	16:36
	eval {	16:36
	my $section = "svn-remote.$self->{repo_id}";	16:36
	$svnsync = {	16:36
	url => tmp_config('--get', "$section.svnsync-url"),	16:36
	uuid => tmp_config('--get', "$section.svnsync-uuid"),	16:37
	}	16:37
mjc	[email@hidden.address]	16:37
devurandom	};	16:37
	← FunkeeMonk left	16:40
	← dsaxena left	16:40
	← p4tux left	16:41
	← sunblush left	16:42
	→ Phoop joined	16:43
	→ RobotDeathSquad joined	16:43
	→ robinr joined	16:45
	→ p4tux joined	16:45
	→ theCarpenter joined	16:47
vbgunz	forgive me if this is lame, in the man git-init page on the web, what does this mean: Use permissions reported by umask(2). ... specifically umask(2)	16:49
Tv	vbgunz: umask. look it up in any advanced unix book.	16:49
	vbgunz: man -a umask may or may not help	16:49
vbgunz	Tv: not really umask, when you see something with (2) ... whats that mean?	16:50
	e.g., git(7)	16:50
mjc	manual section	16:50
Tv	vbgunz: man section	16:50
mjc	man 7 git	16:50
vbgunz	man 2 umask: No manual entry for umask in section 2	16:51
	ok great (n) means n section and to get to it is man n ...	16:51
mjc	man 1 umask	16:51
	on osx at least	16:51
vbgunz	No manual entry for umask in section 1	16:51
	heh, am on Kubuntu 7.10	16:52
mjc	http://en.wikipedia.org/wiki/Umask	16:52
	Tv: git-filter-branch --index-filter 'git update-index --remove images' HEAD	16:52
Tv	there's no user command called umask, it's a shell builtin	16:52
mjc	that'll blow away anything named images from everywhere, no?	16:52
	or just the top level directory	16:53
	from all commits	16:53
Tv	mjc: i'd need to spend 5 minutes with the man page myself ;)	16:53
	→ saimon joined	16:53
mjc	ah	16:53
	it gave that with a filename as an example	16:53
	well I can't hurt anything I can just reimport :P	16:53
vbgunz	git-init(1) states to read umask(2)... but, there is no man for umask, correct?	16:54
mjc	you (or your distro) might not have installed the manual for it	16:54
	read the wikipedia link I posted	16:54
Tv	vbgunz: man you jump into weird conclusions	16:54
mjc	and here's an online copy of the man page http://ibm5.ma.utexas.edu/cgi-bin/man-cgi?umask+2	16:55
vbgunz	heh, "there's no user command called umask, it's a shell builtin", not sure, thought maybe this would mean no man for umask, sorry :/	16:55
Tv	vbgunz: which just means there's no man(1)	16:56
	err	16:56
	umask(10	16:56
	umask(1)	16:56
devurandom	Filed as http://marc.info/?l=git&m=119981060609053&w=2	16:56
mjc	there's a umask(1) on BSDs	16:56
Tv	mjc: do they have cd(1) too?	16:56
mjc	yes looks like 1 is for builtins	16:56
Tv	mjc: oh wow, what did i just say about jumping into weird conclusions	16:57
mjc	heh	16:57
	← theCarpenter left	16:59
mjc	git is so fast it makes my head spin	16:59
	→ \|kukks\| joined	17:00
	← p4tux left	17:03
	→ ivazque1 joined	17:04
mjc	well git-filter-branch without the update index bit is slow	17:05
vbgunz	the following in git-init(1) ...config variable "core.sharedRepository", '...configuration flag receive.denyNonFastforward' ... where do I find these? are these global or specific? do I edit something outside of .git fro this or edit something inside .git after init?	17:07
	→ maio joined	17:07
	→ harinath joined	17:07
maio	hi i deleted some files in working copy how do i get them back?	17:08
Tv	maio: git checkout HEAD path/to/file	17:08
mjc	vbgunz: .git/config	17:08
	or ~/.gitconfig	17:08
vbgunz	yeah, this is in regards to --share=...	17:08
mjc	then in the config file in your shared repo	17:08
	but you probably just want git init --shared	17:09
	or, gitosis	17:09
vbgunz	so, if I had new templates, I rerun git-init ... if I wanted to change persmissions, do I rerun git-init --shared... OR just change the config?	17:09
Tv	vbgunz: it won't overwrite anything existing	17:09
maio	Tv: thanks	17:10
Tv	vbgunz: if you want to change the hooks, change the hooks	17:10
	vbgunz: if you want to edit the config, edit the config	17:10
vbgunz	yeah I read that, I understand you would usually rerun it to pick up modified templates or something similair	17:10
Tv	it won't overwrite anything	17:10
vbgunz	Tv: in regards to --shared. git-init --shared=(s/false/group) won't work, I will need to change the config to make such a change correct?	17:11
	→ lcapitulino joined	17:11
	→ dsaxena joined	17:12
	← kmap left	17:13
mjc	dude it's just a text file	17:13
	→ sunblush joined	17:13
mjc	but it again sounds like what you really want is gitosis :P	17:13
vbgunz	hmm, rerunning git-init, changing --shared changed the config	17:13
	← WorkingGeier left	17:13
mjc	vbgunz: http://eagain.net/gitweb/?p=gitosis.git	17:14
	→ sgrimm joined	17:14
vbgunz	I am trying to learn core-git as best I can, trying to weed out any questions a.s.a.p., am currentyly on git-init, basically done and 19 commands to go... only to be told that isn't core :P	17:15
mjc	watch randal's presentation	17:15
	← kukks left	17:17
	→ _kukks_ joined	17:18
	→ kmap joined	17:18
	← ivazquez left	17:18
vbgunz	mjc: yeah, I am basically done with git-init. seems simple enough though it put umask on my todo, one more command and I'll go over your link again	17:18
	thanks	17:18
mjc	yep	17:18
	← wayneeseguin left	17:19
mjc	been using git for 6+ months; finally convinced the boss to begin migrating to it	17:19
vbgunz	mjc, quick question, where is the presentation, not in the last link you linked too?	17:20
Tv	vbgunz: http://tinyurl.com/2c9np	17:21
vbgunz	heh	17:22
	← kristoffer left	17:25
jengelh	Could someone look at git log d9a61d669c39d9f260a92554071aa960eb0c3e80?	17:25
	→ kumbayo joined	17:26
	→ wayneeseguin joined	17:27
	→ krsanky joined	17:33
	← \|kukks\| left	17:34
	→ theCarpenter joined	17:36
	← twork__ left	17:38
	→ desegnis joined	17:45
mjc	vbgunz: http://video.google.com/videoplay?docid=1251251453592758541	17:46
vbgunz	am 12 minutes into one presentation... may be the same one	17:47
mjc	linus gave another one that is more generic	17:47
	but funny	17:47
vbgunz	yeah, I think it is, I found this one http://www.techpresentations.org/Git	17:47
jengelh	linus also has a git talk on video.google	17:48
vbgunz	yeah, I saw the one with Linus a couple months ago, the presentation by him put Git on my todo.	17:48
jengelh	and there is another from James Bottomley on linuxworld.com/video	17:48
vbgunz	Linux went a little nuts in the one I saw about SVN, I think basically he called them retards or something equivalent	17:49
jengelh	yeah that one	17:49
mjc	for a reason	17:49
jengelh	obviously :p	17:49
	but the beat against svn was not as justified as the beat against cvs	17:50
mjc	yes it was	17:50
	:P	17:50
vbgunz	I never worked on any vcs, I chose Git after seeing the Linus video	17:50
mjc	even more so because of how most windows users use SVN	17:50
jengelh	mjc: well just imagine what were if Windows users used CVS	17:50
	it would be worse	17:50
mjc	I already know what that's like	17:50
jengelh	ya right they'd use clearcase :D	17:51
mjc	but most windows users that use SVN don't use trunk/branches/tags	17:51
	← harinath left	17:52
vbgunz	im lost, I don't speak svn, cvs and so far, in git, I know enough to ask for a cookie	17:52
jengelh	git-fetch cookie	17:52
Ilari	mjc: Yea, I have seen some SVN users. No branches, ocassionally no commit messages. Newer anything about 'Why', etc.	17:52
vbgunz	heh	17:53
	→ ivazquez joined	17:54
	→ madewokherd joined	17:54
	→ robfitz_ joined	17:55
	→ dysinger joined	17:56
	→ sdfahakan joined	17:56
careo	most longtime svn users I know don't even use branches	17:57
mjc	yeah	17:57
	← sgrimm left	17:57
sdfahakan	Hi, I want to merge 2 commits into 1: have two commits in sequence: one really adds new funcionality to code, second converts CR LF -> LF - the second one just makes mess - how can I merge the two?	17:57
careo	or if they do, they're about to find out why most other users dont ;)	17:58
	← Oejet left	17:59
	→ vintrepid joined	18:02
careo	I have a git-svn question I'm hoping someone can shed some light on. I have a nicely setup git-svn clone of our svn repo. git lists 72 remote branches, 71 of which are branches long-since deleted in svn. is there a way to tell git they're dead and to stop listing them?	18:02
Mikachu	you can rm them from .git/refs/remotes/svn	18:03
	they will only show up again if they're updated in svn i think	18:03
careo	cool	18:03
Mikachu	and when i say rm you should use mv, of course	18:03
careo	mv to?	18:03
	or just mv to some backup place in case that wasn't what I wanted to do?	18:04
Mikachu	yeah	18:04
	← puk left	18:06
	← robfitz left	18:06
	← ivazque1 left	18:08
	← johan-s left	18:09
careo	hmm. not quite so easy. I'll keep messing around and googling.	18:10
	→ harinath joined	18:11
devurandom	Hmm... My SVN branches and tags, how are they supposed to be displayed in gitweb? As heads?	18:12
	→ Lash\| joined	18:12
	← aroben left	18:14
vmiklos	nope	18:14
	gitweb hows local branches while svn branches are remote branches using git-svn	18:14
devurandom	So gitweb cannot show them?	18:14
vmiklos	shows*	18:14
	← harinath left	18:14
vmiklos	it can, but not as local heads	18:14
devurandom	Hm...	18:15
Randal	sdfahakan - use git rebase -i - there's a "squash" option	18:15
vmiklos	so they will not be listed	18:15
devurandom	So how do I make them show up?	18:15
Randal	beware - you will invalidate all future commits on the old branch	18:15
vmiklos	create local branches from remote branches	18:16
Randal	if those are the two most recent commits, it might be easier just to do the shuffling "by hand"	18:16
	→ TuxCM- joined	18:16
	Randal wonders if sdfahakan is still listening	18:16
	→ pennyk_ joined	18:19
pennyk_	hi, all	18:19
	BitKeeper is commerical software. It provide a simple free tool, bkf.	18:20
	Is it right that 'bkf clone' only gives a snapshot, not a whole repository clone?	18:20
devurandom	It's getting better... :( "Unable to determine upstream SVN information from working tree history"	18:21
Ilari	pennyk_: What the heck it does have to do with Git?	18:21
sdfahakan	Randal: thanks for your tip; I apologize, went out for a little while, will look at it now	18:21
pennyk_	Any way to convert a bk repo into git?	18:21
	→ alley_cat joined	18:22
pennyk_	I know we can convert CVS/SVN repo into git.	18:22
vmiklos	using bkcvs	18:24
devurandom	Hmm... And "git svn log" shows an empty log and gives "fatal: bad default revision 'refs/remotes/git-svn'" on the commandline...	18:24
	← TuxCM left	18:24
vmiklos	devurandom http://blogs.frugalware.org/vmiklos/2007/12/09/p320 <- maybe you want something like this. the second repo is hanled properly by gitweb	18:24
	+d	18:25
devurandom	Saw that yesterday already... But somehow I dislike having the same repository 3 times on the same server, which I all need to update when the first is updated... svn => git-svn => git	18:26
	git-svn-clone also checks the repository out?? Can I undo that?	18:27
thiago	git-runstatus stats each file twice. Is there a good reason for that?	18:28
sdfahakan	Randal: invalidate further commits - I realized this even before asking this question :-) so the rebase does not recalculate the commit hashes (IDs) that are descendants of the new merged commit? That would be bad	18:28
vmiklos	devurandom yes, it does - as git-svn fetch doesn't work without a working tree	18:29
	at least not here;)	18:29
sdfahakan	(... commit hashes _of commits_ that are descendants ...)	18:29
	← theCarpenter left	18:29
pennyk_	vmiklos: but I can't access that bk repo directly. All I can do is to access it throught Internet.	18:29
devurandom	vmiklos: I did "git --bare init && git --bare svn init && git --bare svn fetch" before and it seemed to work...	18:30
vmiklos	pennyk_ i don't know bkcvs deeply, i just know Linus convert the linux-2.6 repo to git using bkcvs:)	18:30
pennyk_	vmiklos: oh. Still, thx.	18:30
	→ tvachon joined	18:30
vmiklos	devurandom huh, git --bare svn fetch ;) i haven't thought of it	18:31
	$ git --bare svn fetch	18:31
	fatal: Not a git repository: '/home/vmiklos/scm/git/ooo-build.git'	18:31
	rev-parse --symbolic --all: command returned error: 128	18:31
	yes	18:32
	← dancor left	18:32
	vmiklos doesn't think you can avoid having two repos	18:33
	← mountie left	18:33
	→ mountie joined	18:35
devurandom	vmiklos: git --bare svn fetch works flawlessly here...	18:35
	Did you do git --bare init before?	18:35
	← saimon left	18:35
devurandom	I think you will get that error if you don't.	18:35
	→ tongueroo joined	18:36
devurandom	Heh... Now it gets funny...	18:37
	Odd number of elements in anonymous hash at /usr/bin/git-svn line 1760.	18:37
	r13 = 6d57857b7da6aeeda39329ccfd52a329fe47ce78 (tags/1.10a)	18:37
	Segmentation fault	18:37
	→ \|kukks\| joined	18:37
	← jtoy left	18:37
Ilari	devurandom: What got toasted by segfault?	18:38
devurandom	git --bare svn fetch	18:38
	Currently trying to reproduce... r13 takes a while, though...	18:39
	← tvachon left	18:39
Ilari	devurandom: Maybe allow coredumping 'ulimit -c unlimited' or something like that and then use file on the crashdump to find the actual crashed program?	18:40
devurandom	Why does git-svn-log take hours, but git-log shows me the log immediately?	18:40
	Ilari: Will try that.	18:40
	→ tvachon joined	18:41
	← janm left	18:41
	← nashdj left	18:41
devurandom	When I use git-daemon, I don't need git-update-server-info, right?	18:42
	And gitweb doesn't need it either?	18:42
Ilari	Maybe git svn log hits the remote repo in some really nasty manner, but git log only hits the local object database (and in really nice way).	18:42
loops	devurandom, correct.. you only need it to support dumb protocols such as http	18:43
	→ johan-s joined	18:45
	→ k_dub joined	18:45
	← johan-s left	18:45
devurandom	Ilari: How do I examine that core?	18:46
	← pennyk_ left	18:47
Tv	devurandom: what core?	18:48
devurandom	coredump. From git-svn	18:48
Tv	oh i see ilari talk about that	18:48
	gdb	18:48
	file can extract the program name too, yeah	18:49
devurandom	If I don't have gdb on the server? Can I do it from remote somehow?	18:49
Tv	devurandom: you'll need the core file, the binary that was running, and possibly all the libs	18:49
	→ anttih joined	18:50
	→ theCarpenter joined	18:50
devurandom	How do I get file to extract the program name?	18:50
	Here it just says "core.25748: ELF 32-bit LSB core file Intel 80386, version 1 (SYSV), SVR4-style"	18:50
	→ GyrosGeier joined	18:51
Tv	hmm, maybe it didn't know how to	18:51
	strings core\|less	18:51
	or just use gdb	18:51
	→ sgrimm joined	18:52
	← opsis25 left	18:53
	← k_dub left	18:53
devurandom	vmiklos: Can you explain to me what those configurations in the description on your blog mean?	18:54
	← _kukks_ left	18:54
	→ gitte joined	18:55
Ilari	devurandom: And if it was perl which segfaulted, good luck for debugging, you probably will need it...	18:57
devurandom	It probably was Perl... :(	18:57
	← theCarpenter left	18:57
Ilari	devurandom: If it wasn't, one important thing would be to recover the command line used to invoke it.	18:58
	devurandom: That was important step when I tried to debug that git-rev-list commit limiting crash.	18:59
	→ dancor joined	18:59
devurandom	What do the remote.origin.fetch config keys do?	19:02
	← gitte left	19:03
Ilari	devurandom: Control remote branch name mapping on fetching.	19:04
devurandom	So remote.origin.fetch +refs/remotes/tags/:refs/tags/ means refs/remotes/tags will be renamed to refs/tags on fetch?	19:05
	← careo left	19:06
Ilari	devurandom: I think so.	19:09
devurandom	Do I need to have refs/heads/master?	19:09
	Or would it also work if I just had refs/heads/trunk?	19:10
	→ careo joined	19:10
Tv	devurandom: as long as HEAD points to something sane	19:10
GyrosGeier	hmm	19:10
	can I instruct fetch to pull all remotes, or do I have to use one fetch command per remote?	19:10
	I'd like to build an ip-up rule for the laptop	19:11
Tv	GyrosGeier: you want git remote update	19:12
GyrosGeier	thanks	19:12
	well	19:12
	→ KirinDav joined	19:12
	GyrosGeier is at the beginning of a long hacking session now	19:12
	→ jesse_ joined	19:14
	← careo left	19:15
	→ timocratic joined	19:16
	→ careo joined	19:16
	→ archfear joined	19:18
vmiklos	devurandom aha. it seems if you do a git --bare init before git --bare svn init, then it does (how horrible)	19:20
	← kmap left	19:20
devurandom	"Then it does <work>" or "Then it does <crash>"?	19:21
vmiklos	the previous. at least it started	19:21
	without git --bare init it just bailed out with an error in 2 secs	19:21
	→ drizzd_ joined	19:23
devurandom	Yeah, that's true.	19:23
	It seems as if git-svn init doesn't do a full init or something like that.	19:24
	How do I tell it where the default ref is?	19:24
	s/it/git or gitweb/	19:25
vmiklos	it does if you don't use --bare ;)	19:25
	the default ref is refs/heads/master	19:25
devurandom	But then I will have 2 times the same head, no?	19:25
vmiklos	yes, one remote and one local	19:26
devurandom	Then trunk will be the same as master	19:26
vmiklos	remotes/trunk, yes	19:26
devurandom	vmiklos: Aren't they both local now, that I fetched them from my git-svn repos?	19:26
vmiklos	you have one local ref: master	19:27
	→ nud_ joined	19:27
vmiklos	and you have remote refs: remotes/trunk, remotes/branches/, remotes/tags/	19:27
	or so	19:27
devurandom	vmiklos: I don't have any refs/remotes/	19:27
vmiklos	huh:)	19:27
	where do you have your branches?	19:27
	and tags	19:28
	in the git-svn repo	19:28
devurandom	refs/tags/ and refs/heads/	19:28
	In git-svn I do have them like you said.	19:28
Tv	devurandom: are you perhaps using a very old git-svn?	19:28
devurandom	Tv: Probably outdated, yes. I says 1.5.4_rc2 ;)	19:28
Tv	hah	19:28
vmiklos	that isn't outdated at all:)	19:29
devurandom	Note the ";)" ;)	19:29
	vmiklos: But the only exported repos is the one where I fetch from git-svn	19:29
	I wonder how this will all work... sigh	19:29
vmiklos	exported repos?	19:29
devurandom	git-daemon-export-ok	19:29
vmiklos	well, it isn't that complicated	19:30
devurandom	Since now I am 100% not on the git-svn manpage way...	19:30
vmiklos	mkdir foo.git	19:30
	cd foo.git	19:30
	git --bare init	19:30
	git --bare svn init url	19:30
	git --bare svn fetch	19:30
devurandom	(That is my git-svn repos?)	19:30
	→ theCarpenter joined	19:30
vmiklos	git config remote.origin.url .	19:30
	→ aroben joined	19:30
devurandom	Oh... Now I fetch from myself??	19:31
vmiklos	git config remote.origin.fetch +refs/remotes/tags/:refs/tags/ && git config --add remote.origin.fetch +refs/remotes/trunk:refs/heads/master && git config --add remote.origin.fetch +refs/remotes/:refs/heads/ && git fetch	19:31
	yes	19:31
devurandom	Aha...	19:31
vmiklos	i haven't tested it yet as git --bare svn fetch is still running, but ideally this can work	19:31
devurandom	Sounds cool. :)	19:31
vmiklos	if you fully switch to git later, then you probably still want to remove the git-svn-related files, but that's a minor problem	19:32
devurandom	And after that I can follow the steps after "# Do the initial import on a server" from man git-svn ?	19:33
vmiklos	after that you already have an import (git --bare svn fetch did it)	19:34
	if you want to update it incrementally, then you need git --bare svn fetch && git fetch	19:34
	← drizzd left	19:34
devurandom	vmiklos: I mean on the clients.	19:35
	As that thing is thougth to be only a gateway/conversion-station, like described in the manpage.	19:35
vmiklos	on the clients you can just clone it as a normal git repo	19:36
	or do you want to use git-svn on the clients, too?	19:36
devurandom	vmiklos: I want to commit to svn, yes.	19:36
	→ Giel joined	19:36
devurandom	Hello Giel. :)	19:36
vmiklos	uh :/ no idea	19:36
Giel	hi devurandom :-)	19:36
vmiklos	i never used git-svn dcommit from a repo which is created by git clone	19:37
devurandom	vmiklos: I hope this doesn't mess it up then...	19:37
vmiklos	probably it's possible but you need someone who knows how to do this ;) (probably you should ask on the list)	19:37
Giel	devurandom: I think git-svn uses some special config options in .git/config which git-clone doesn't copy over	19:38
	→ grahal joined	19:39
	← desegnis left	19:39
	→ desegnis joined	19:40
mjc	devurandom: did those videos help you?	19:41
cehteh	is it possible to let submodules link to a head instead to a specific commit?	19:41
vmiklos	nope	19:41
devurandom	mjc: Videos?	19:41
mjc	or was that someone else	19:42
devurandom	Well, if you have something which might even remotely help me, I'd be glad to get it. :)	19:42
	← dsaxena left	19:42
cehteh	mhm	19:42
vmiklos	i already asked on the list earlier (as i think it would be a nice feature) but i got the answer "you don't want to do this" ;)	19:43
cehteh	well they dont know for what i want it :)	19:43
vmiklos	(which is true in some cases, but i think it would be nice if i could just link to a given branch)	19:43
grahal	is there a way that after a rebase, commits get lost?	19:43
cehteh	for software configuration thats certainly a bad idea	19:44
vmiklos	is it an interactive rebase?	19:44
devurandom	Yeah, cool...	19:44
	Now git-svn-rebase gives me: fatal: ambiguous argument 'HEAD': unknown revision or path not in the working tree.	19:44
cehteh	http://www.pipapo.org/webgit?repo=webgit&object=af66dbba9f836bd298b2dbe0cd476d0bd4859cbe	19:44
devurandom	Whatever that means...	19:44
cehteh	.. see submodule integration in the browser, would be nice to have the toplevel project just accumulate everything below as submodule which links to HEAD	19:45
	but well .. not really important	19:45
devurandom	And on git-fetch: fatal: 'origin': unable to chdir or not a git archive	19:47
	← archfear left	19:49
devurandom	I'll redo it again, using the method you described above. Maybe that helps...	19:49
	It still segfaults on me. Though I could swear it worked yesterday...	19:51
grahal	any idea how to look for lost dir/files in git if it seems like no commits point to them?	19:59
Tv	git lost-found	19:59
	← jengelh left	20:07
	← Phoop left	20:07
	→ jengelh joined	20:08
	← jesse_ left	20:08
jengelh	Running git-gc in a bare repository seems detrimental...	20:11
	← Tux32 left	20:11
jengelh	refs/heads/ is cleared	20:11
grahal	Tv: thanks... I see that some commits have portions of the files I wanted to restore... is there any way that I tie up everything?	20:12
vmiklos	so now you have packed refs:)	20:12
jengelh	vmiklos: that's not nice ^_^	20:12
	← sdfahakan left	20:12
vmiklos	why not?	20:12
Tv	grahal: usually git show COMMITISH:PATH	20:13
Mikachu	jengelh: see the packed-refs file and man git-pack-refs	20:13
jengelh	vmiklos: custom script scanned refs/heads/ for available branches :p	20:13
vmiklos	that's bad	20:13
	use git ls-remote (for example) for listing refs	20:14
Tv	git rev-parse	20:14
	for-each-ref	20:14
	there's plenty of tools	20:14
jengelh	nah, ls-remote is good. but 'remote'? should be local, not?	20:14
vmiklos	'git ls-remote .' or so	20:15
	but yes, Tv has better ideas:)	20:16
jengelh	but git-rev-parse is not one of them	20:16
	← Rogan left	20:16
Tv	well not for listing out the refs	20:16
vmiklos	git 'git for-each-ref refs/heads'	20:17
	s/^git //	20:17
	← theCarpenter left	20:18
	→ theCarpenter joined	20:21
	→ johan-s joined	20:23
hachi	can I use git submodules to represent the structure of a single subversion repo that has multiple projects under /trunk and each one split apart when tagged?	20:31
	http://nopaste.snit.ch/12073 # I'm trying to do it like this, but I'm kinda shooting in the dark with the idea. Maybe someone else has attempted it?	20:32
vmiklos	you can but probably you don't want so	20:32
	http://www.kernel.org/pub/software/scm/git/docs/user-manual.html#submodules	20:33
	have you read this?	20:33
	→ jesse_ joined	20:33
hachi	nope, that's my first failing, missed that part	20:33
	reading it now	20:33
vmiklos	basically you want submodules when it makes sense to stick to a given version of the submodule. if you create such a trunk.git, this will only cause problems for you	20:35
hachi	okay, so then this is the wrong answer	20:35
	I'm starting with what I consider to be a messed up svn repo, and trying to use git-svn with it	20:35
Giel	hachi: maybe you might want to clean it out in subversion first ?	20:36
hachi	http://code.sixapart.com/svn/mogilefs/ # under trunk there's 6 different projects, and then in tags each project has a prefix on the tagname	20:36
	I can't clean it out, that would destroy the history	20:36
	while I do own this repo, and in theory I could do whatever I want... the real problem is that other people use it	20:37
	→ _mountie joined	20:37
vmiklos	he means move the tags/branches to their own subdir for each project	20:37
	← mountie left	20:37
jengelh	something like git-svn --trunk=trunk/myproj --tags=tags/myproj perhaps?	20:37
vmiklos	maybe you can even do so:)	20:37
Giel	hachi: yes, as vmiklos says; I meant move all projects to a project/trunk:project/branches:project/tags structure	20:38
hachi	this works with git-svn? I assumed that it understood the history of everything in the filesystem, and would just get more confused	20:38
	tarbo → tarrybone	20:39
Giel	hachi: ah yes, actually git-svn might not like that at all...	20:39
	tarrybone → tarbo	20:39
vmiklos	i would try git svn init and then hacking .git/config manually. maybe that would work	20:40
hachi	I can make a copy of the repo and attempt it	20:40
	git-svn doesn't allow remotes to have * anywhere in the filepath, it must be in it's own component /*/ style	20:41
	but still, the first idea I'll also try and see if it works	20:41
vmiklos	i'm not exactly sure btw what projects do you have, ie there are tags/MogileFS-Client-*, but there is no trunk/MogileFS-Client	20:43
hachi	it's in api/perl	20:43
	trunk/api/perl/MogileFS-Client	20:43
vmiklos	ugh:)	20:43
hachi	yeah	20:43
vmiklos	so tunk is trunk/api/perl/MogileFS-Client, tags are tags/MogileFS-Client-* and no branches?	20:44
	+r	20:44
hachi	given the choice I might have put these each in seperate repos, but... we didn't	20:44
	it was a group decision, and we went with the 'easiest' out of what we thought was pragmatism	20:44
	yes	20:44
	well, there are branches, but for the sake of git-svn bridging I'll ignore them	20:45
vmiklos	k	20:45
hachi	the sucky part is I can't say tags/MogileFS-Client-* to git-svn	20:45
vmiklos	not sure	20:45
hachi	I'm trying to see if just saying tags/* will do the right thing, maybe git-svn will see tags/something and go "hey, it was a copy of something never in my trunk... I bet I'm supposed to ignore this"	20:46
vmiklos	$ git svn fetch	20:47
	Missing trailing '/' on left side of: 'tags/MogileFS-Client-*' (tags/MogileFS-Client-)	20:47
hachi	says right in the manpage that you can't do that	20:47
	or EOL on both sides of the *	20:47
vmiklos	maybe you can comment out this check in git-svn and see what happends :)	20:47
hachi	ep	20:47
vmiklos	-d	20:48
	→ cartman joined	20:49
	← Lash\| left	20:49
devurandom	I remember where the difference to yesterday is... I didn't use --use-svnsync-props but --rewrite-root yesterday...	20:50
doener	EOL? Now that I look at it, EOL makes no sense there. We're talking about the left side, the * must be enclosed in // or /: ... Or am I missing something?	20:50
devurandom	I will try to confirm that this is the issue and then send a mail to the list...	20:50
hachi	doener: I'm just recalling what the manpage says... I rememeber quite distinctly that it says EOL or /	20:51
doener	hachi: yeah it does, but it doesn't make sense to me	20:51
	← pasky left	20:51
doener	oh well, will just send a patch and see if someone corrects me ;-)	20:52
hachi	well, for the record, git-svn doesn't just dtrt in this case	20:52
Pieter	dtrt?	20:53
doener	do the right thing	20:53
Pieter	ah :)	20:53
jengelh	time dtwt.	20:53
	← kumbayo left	20:55
hachi	so, why does git-svn let you specify a name in the svn-remote section?	20:56
	the default is 'svn' when it makes the config for you	20:56
doener	hachi: you can have more than one	20:56
vmiklos	then say 'git svn fetch foo'	20:57
hachi	more than one what though? is this only for submodules, or is it for something else?	20:57
	I can fetch from more than one location sure, but what's the point	20:57
	← ebzzry left	20:57
vmiklos	there is no point in most cases:)	20:57
hachi	obviously I'm not within that set currrently	20:58
vmiklos	say you have foo-ng and foo project and you want to use foo but cherry-picking changes from foo-ng to foo	20:58
	then maybe you want this	20:58
	not something for you as far as i see ;)	20:58
doener	hachi: svk comes to mind	20:59
hachi	okay, so that's more for stitiching things together again once they are split apart	20:59
	and I'm still trying to split them apart, so... ignore that	20:59
	ignore that part of git-svn	21:00
	← desegnis left	21:05
	← theCarpenter left	21:07
	← Giel left	21:13
	→ \ask joined	21:13
	← jesse_ left	21:16
	→ reval joined	21:17
	_mountie → mountie	21:19
	← wagle left	21:22
	← reval left	21:23
	→ reval joined	21:26
	→ madewokherd` joined	21:28
	stillLotR → LotR	21:30
	← cartman left	21:31
	← madewokherd left	21:38
	← GyrosGeier left	21:40
	→ chris2 joined	21:41
	→ loswillios joined	21:41
loswillios	hi	21:42
	I've patched my linux-tree manually (with patch). Now I want a clean tree again, how would I do that?	21:43
	git-status and then git-revert on the single files isn't the right way I guess	21:44
doener	git reset --hard	21:44
loswillios	cool, thanks	21:45
	→ sverrej joined	21:45
	← sverrej left	21:46
loswillios	is that the recommended way (patching with 'patch') or is there a git-way too?	21:46
	← KirinDav left	21:47
madduck	copy src/events/handlers/{CTLSOCK_PING.py => NOOP.py} (83%)	21:47
	how the heck does git know this???	21:47
	i copied the file, changed it, added it.	21:47
tpope	loswillios: git-apply	21:47
	bdiego → midgar	21:47
tpope	it's atomic, among other things	21:47
doener	madduck: it just compares the contents	21:47
madduck	but the contents is not identical	21:47
doener	but above a certain threshold	21:48
tpope	it's 83% identical :)	21:48
madduck	holy cow	21:48
vmiklos	maybe a few lines of code with a big copyright header or so :)	21:48
madduck	vmiklos: yeah	21:48
vmiklos	then it's logical	21:48
loswillios	tpope: oki, will check it out	21:49
doener	madduck: the cool thing is that "git log -p -C" should even show you an exact patch for the stuff you changed, not just some "oh well, this file is new, of course _everything_ was added right now"	21:49
	→ dsaxena joined	21:49
madduck	this is pretty nifty	21:49
	→ KirinDav joined	21:51
	← ironfroggy left	21:52
	← spuk- left	21:53
hachi	how do I move master to point to something else remote?	21:59
	oh, you just switch branches	21:59
	← aroben left	21:59
	→ spuk- joined	22:00
hachi	so, git-svn didn't like my attempt to clean up the svn source tree at all	22:00
	the history is not looking very linear	22:00
doener	hm, this looks "unexpected": http://git.pastebin.com/d4d18a46f	22:01
	I had accidently fetched 2.6.16-rc2-mm1 earlier, removed the tag, gc --prune, and then tried to fetch 2.6.24-rc6-mm1	22:02
	but git decided to fetch both, the new one as well as the old one once again	22:02
	is that just me doing sth. stupid, or a genuine bug?	22:03
vmiklos	where is the problem?	22:04
	now you probably want git reset --hard v2.6.24-rc6-mm1	22:05
	← KirinDav left	22:05
	→ KirinDav joined	22:05
	→ theCarpenter joined	22:05
doener	vmiklos: git fetch fetched more than I requested	22:07
	vmiklos: I only requested v2.6.24-rc6-mm1 in that second run (which is in the paste), but git fetch v2.6.24-rc6-mm1 _and_ v2.6.16-rc2-mm1.	22:08
	← cmarcelo left	22:09
vmiklos	i don't think you can fetch only one tag	22:11
	well, you can but in fact you need the objects from a previous tag for a later tag, so it's no good not fetching the tag itself	22:11
	→ Tux32 joined	22:12
	← wayneeseguin left	22:12
doener	vmiklos: hm? I doubt that v2.6.16-rc2-mm1 is an ancestor of v2.6.24-rc6-mm1	22:13
	← lcapitulino left	22:13
doener	by your argumentation, it should've fetched all other tags as well	22:14
	→ kumbayo joined	22:14
	← soma- left	22:17
	← HG` left	22:17
vmiklos	yes, you probably did	22:17
	→ soma- joined	22:18
	→ wagle joined	22:19
vmiklos	well, i don't know that repo but only tags that do not point at objects reachable from the branch heads are not fetched automatically	22:19
	which means that in most cases tags are always fetched - without using fetch -t	22:19
doener	vmiklos: those -mm* tags are on topic branches. Upstream has e.g. tags v2.6.22-rc3, v2.6.22-rc4, and so on	22:20
	→ pasky joined	22:21
doener	at each tag, a branch like v2.6.22-rc3-mm1 is created, and a tag v2.6.22-rc3-mm1 is created on that branch	22:21
vmiklos	i would do: git remote add mm git://git.kernel.org/pub/scm/linux/kernel/git/smurf/linux-trees.git; git fetch mm; git reset --hard mm/v2.6.22-rc3-mm1 or so	22:22
doener	vmiklos: I don't want all that mm crap, thus I asked git to just fetch a single tag for me	22:22
	vmiklos: and the first run, requesting just v2.6.16-rc2-mm1 did just that	22:23
	but the second one, although instructed to fetch v2.6.24-rc6-mm1, did _also_ fetch v2.6.16-rc2-mm1 again	22:23
	→ Plnt joined	22:24
	← krsanky left	22:24
	← kumbayo left	22:26
vmiklos	hm	22:26
	← KirinDav left	22:27
gitster	if it was not by explicit "git fetch $url $this_tag" request but by lazy "git fetch" or "git fetch $nickname" request with configured remote.$nickname.fetch, I suspect that it is working as designed (you are triggering automatic tag following in that use case).	22:28
	→ rhalff joined	22:28
doener	preparing a script right now to reproduce	22:28
gitster	automatic tag following picks up tags that point at commits that exist in your repository automatically. That commonly is what people want. "git fetch --no-tags" should disable it.	22:30
doener	http://git.pastebin.com/d4f53698e	22:30
	← Plnt left	22:32
	← pasky left	22:32
doener	hm, that just fetches the tag though. For the kernel thing, it actually downloaded 5MB, so I doubt that that was just the tag	22:32
	→ Plnt joined	22:32
	← rhalff left	22:36
	→ mithro joined	22:36
hachi	http://hachi.kuiki.net/stuff/gitk.png	22:40
	that does not look right, dangit :(	22:40
	→ pasky joined	22:40
	← sunblush left	22:42
	← anttih left	22:42
	→ aroben joined	22:42
	← orospakr left	22:43
devurandom	Sadly another bugreport: http://marc.info/?l=git&m=119983162014061&w=2	22:43
	← Plnt left	22:44
	→ Plnt joined	22:44
vmiklos	devurandom git svn fetch finished here and it works properly - so thanks for for suggestions on how to do the conversion without a working directory - i've updated the blog entry	22:50
	→ metafollic joined	22:50
theCarpenter	crap	22:50
	← \ask left	22:50
devurandom	Seems as if I am the only one where it does not work. ;)	22:50
theCarpenter	one of my intro programming courses is mandating SVN use for our programming projects	22:51
devurandom	vmiklos: Which version of git are you using?	22:51
vmiklos	1.5.4.rc2.38.gd6da3 at the moment	22:51
	← nud_ left	22:51
	→ \ask joined	22:52
theCarpenter	we have to submit our full project repo and history with the actual final source code version	22:52
devurandom	Does that have any git-svn differences compared to 1.5.4.rc2?	22:52
	→ aroben_ joined	22:52
theCarpenter	can i publish git repos as SVN repo?	22:53
vmiklos	devurandom i don't think so	22:53
robinr	you can export to SVN	22:53
theCarpenter	wonderful	22:53
	ill talk to my professor and promote git a bit i think B)	22:53
robinr	your'e supposed to learn to use more than one SCM I guess	22:54
	SVN is just an example to practice on	22:54
theCarpenter	it is	22:54
	but none of the guys have heard about anytihng else	22:55
devurandom	vmiklos: Would you mind to try this on a svn dump of mine?	22:55
theCarpenter	except maybe perforce	22:55
	and by guys, i mean profs/TAs	22:55
	← aroben left	22:55
vmiklos	devurandom if you upload a tarballs somewhere, i can	22:55
	aroben_ → aroben	22:55
devurandom	Ok, sec.	22:55
	→ delroy joined	22:56
	← reval left	22:57
	← Fullmoon left	22:58
	→ ironfroggy joined	22:59
	← ironfroggy left	22:59
	→ KirinDav joined	23:00
	→ ironfroggy joined	23:00
doener	gitster: hm, can't reproduce git fetching the objects again... it now always just fetches the tag (as expected)	23:00
	cbrake → cbrake_away	23:01
	← ferdy left	23:02
	← delroy left	23:03
	→ federico__ joined	23:03
	← alley_cat left	23:03
federico__	I get a problem building gitk; anyone can help?	23:04
	make[2]: *** No rule to make target `giml.dtd', needed by `all-am'. Stop.	23:04
devurandom	vmiklos: "if you later switch fully to git, then you want to remove the git-svn metadata - doing so is the easiest if you just clone the repo" git-svn parts of the repository are not preserved when doing git-clone?	23:07
	→ duncanm joined	23:08
	← duncanm left	23:08
	← \ask left	23:08
Ilari	theCarpenter: If you can assign linear branches for commits, then it is possible to dump linearized version of history and commit that to SVN. It sure as heck doesn't look 'natural' though (it probably merges like crazy to SVN person).	23:08
robinr	linearizing history can be very pretty	23:09
	→ duncanm joined	23:09
robinr	it also gives you an excuse to drop those oopses	23:10
duncanm	i made a few commits on my new laptop, and i forgot to set GIT_AUTHOR_EMAIL, is there an easy way to fix those commits?	23:10
	i haven't pushed it anywhere, so it's all local	23:10
	→ kristoffer joined	23:10
Ilari	theCarpenter: And it is difficult to do in stable manner. But stability is presumambly not required here, as it is one-shot dump.	23:11
kelvie_	duncanm: git-filter-branch is probably what you're looking for	23:11
	→ cartman joined	23:11
duncanm	hmm, i'm on git 1.5.2.5, and i don't have that	23:12
kelvie_	hrm.. perhaps time to upgrade then? :P	23:13
	→ YGingras__ joined	23:14
	← YGingras_ left	23:14
vmiklos	devurandom correct	23:16
robinr	duncanm: that question was on the mailing list just one or two days ago	23:17
devurandom	vmiklos: So that is what I regenerate with the "git svn init && git update-ref" part described in the manpage?	23:17
robinr	and an answer	23:17
duncanm	robinr: oh nice - do you have a link?	23:17
devurandom	vmiklos: Got my message?	23:17
	duncanm is build git 1.5.3.x on ubuntu	23:18
vmiklos	devurandom private one?	23:18
duncanm	building	23:18
devurandom	Yes,	23:18
vmiklos	they're on ignore :/	23:18
	use /notice if you want something :)	23:18
robinr	duncanm: Look for "Retroactively change email signature?"	23:18
duncanm	cool	23:18
	→ flaguy joined	23:19
devurandom	Better?	23:19
	→ jsimonss_ joined	23:19
vmiklos	devurandom regarding "git svn init && git update-ref" i'm not sure what exactly they're doing - in fact no idea what is the recommended way to clone a git-svn repo properly	23:19
devurandom	Well, that is what man git-svn says...	23:20
gitster	If one spends time and energy to build something from source, 1.5.3.x is a wrong thing to build at this time. Why not do 1.5.4-rc2 (or even 'master')?	23:20
	← octopod left	23:20
gitster	Otherwise you are obsolete the time you finish building, and will be ancient by the end of the month.	23:20
duncanm	gitster: i guess you have a build	23:20
	← jsimonss left	23:20
duncanm	s/build/point/	23:21
kelvie_	just get git from git.git :P	23:21
duncanm	yeah	23:21
kelvie_	it's just 37 M	23:21
gitster	Thanks. We just got another tester ;-)	23:21
duncanm	and i guess i can remove the git package from ubuntu and just use a version in $HOME/bin	23:21
	i run git.git in my other box - i just got a eee pc for x'mas	23:21
kelvie_	well you need one bootstrapped to get it remember ;)	23:22
duncanm	sure ;-)	23:22
kelvie_	you don't want to be gitless	23:22
duncanm	heh, my domain name is a-chinaman.com, and i host my git stuff on git.a-chinaman.com	23:22
kelvie_	haha	23:22
duncanm	i didn't notice that odd that domain name is until recently	23:22
	s/that/how/	23:22
	← theCarpenter left	23:23
	→ sverrej joined	23:23
vmiklos	devurandom huh, how do i create an svn repo from that dump file?	23:24
devurandom	vmiklos: bzcat warzone2100.bz2 \| svnadmin load warzone2100	23:24
vmiklos	thx	23:25
	vmiklos does not know too much about svn administration :)	23:25
devurandom	I have not tested the bzcat part. If it fails, you can also bunzip2 it and then use cat.	23:25
	devurandom neither.	23:25
devurandom	I just learned it in the last 2 days.	23:25
jengelh	why cat	23:26
	bzip2 -cd	23:26
	← cartman left	23:26
Tv	because bzcat is easier to type	23:26
kelvie_	whoops git.git is 21M on my system	23:26
	forgot to repack	23:26
jengelh	21M is nothing :p	23:26
kelvie_	it is :P	23:26
	but it's almost half the size of my earlier quote	23:27
vmiklos	devurandom erm, it requires an argument and an empty dir is not enough	23:27
devurandom	vmiklos: "it"?	23:27
Tv	vmiklos: do you read documents, like ever?	23:27
vmiklos	svn load, sorry	23:27
devurandom	Ah, sorry.	23:27
	vmiklos: It has to be the following:	23:28
	svnadmin create warzone2100 && bzcat warzone2100.bz2 \| svnadmin load warzone2100	23:28
vmiklos	Tv well, i offered to help testing git-svn but i don't want to learn svnadmin :)	23:28
	devurandom thx	23:28
devurandom	I have to thank you. :)	23:29
duncanm	building git on an eeepc is not the fastest thing on earth ;-)	23:32
	hmm	23:33
jengelh	distro should provid eit	23:33
duncanm	MSGFMT po/de.msg make[1]: *** [po/de.msg] Error 127	23:33
	jengelh: ubuntu has only 1.5.2	23:33
jengelh	NOT ubuntu	23:33
	whatever ox ships	23:33
	← mjc left	23:34
	duncanm installs gettext	23:34
	→ mjc joined	23:35
vmiklos	devurandom duh, why didn't you just told me it's at svn://svn.gna.org/svn/warzone/trunk/ ;) (or is it different?)	23:35
	← mjc left	23:36
devurandom	That one is different.	23:36
vmiklos	ah, k	23:36
devurandom	I synced a copy, dumped the repos, edited it and will later ask it to be uploaded.	23:37
	When I sort out why all this problems arrise.	23:37
	(svn itself doesnt have any problems with it, btw.)	23:37
	vmiklos: And svn://svn.gna.org/... is usually not working. "Malformed network data" and similar responses.	23:38
vmiklos	ok, then it's better this way:)	23:39
	duncanm builds docs	23:39
hliusv561	vmiklos: Maybe try creating empty SVN repo there and try the load then?	23:40
vmiklos	already in progress:)	23:42
devurandom	hliusv561: Yeah, I forgot to tell him that he needs an empty repos.	23:42
Ilari	BTW: What's the "right" way to load "dumps" made using 'git bundle create foo.bundle --all'?	23:45
	← KirinDav left	23:46
	← kristoffer left	23:47
	← dancor left	23:51
vmiklos	is there any way to make --ext-diff the default when using git log -p?	23:51
duncanm	hmm	23:58
	how do i specify a <command> in git-filter-branch?	23:59
vmiklos	see the manpage, it has several examples	23:59
duncanm	git-filter-branch --env-filter 'git-commit --author "Duncan Mak [email@hidden.address] -- that doesn't work	23:59
	→ Eridius joined	23:59
Eridius	is there any sort of bug tracker? I've just found a bug in git-svn but I don't know if it's been reported before	23:59

Logs Search ←Prev date Next date→

Channels Documentation