IRCloggy #git 2007-07-19

Provider of IRC logs since 2005.

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

	2007-07-19
	→ orospakr joined	00:01
	← QuercusMax left	00:06
	← goloo left	00:09
	→ sgrimm_ joined	00:11
	→ QuercusMax joined	00:11
	← sgrimm left	00:12
	→ gitte joined	00:18
mnemoc	hi, may i convert a directory into an independent repo? (preserving history)	00:24
loops	mnemoc, it's possible, but not easy. It's much easier to join repos, than split one apart.	00:26
	← dduncan left	00:26
mnemoc	loops: ic, thanks :)	00:26
gitte	mnemoc: I'm not really understanding what you want to do.	00:26
	convert a directory into a repo?	00:26
loops	mnemoc, see git-filter-branch for a tool that should help you rewrite history to remove everything but the directory that you want to keep	00:26
	mnemoc, (obviously you'd do this in a clone of your repo, leaving the original in tact)	00:27
	gitte, he wants a new repo that only contains the contents of one directory of an existing repo.	00:27
	→ aroben joined	00:27
loops	(preserving history)	00:27
mnemoc	gitte: i'm going to import an svn repository in which was splited in subproject after a few years, and i'm triying to plan "safely"	00:27
	loops: thanks a lot for the hint	00:28
gitte	Ah. Yes, --subdir-filter of filter-branch seems to do what you want.	00:29
	It is only lightly tested, though.	00:29
thiago	that's something we'll have to decide for KDE. We move things around.	00:31
mnemoc	great, this way i can import the first 20k revisions in one pass and filter before continue with the rest :p	00:31
gitte	Yep.	00:31
	filter-branch is not exactly a speed daemon, but in a couple of weeks we will probably have a substantial speed up.	00:32
mnemoc	:)	00:32
gitte	jasam: I just read your comments on krh's patch. Very good!	00:39
	← codahale left	00:42
	→ felipe joined	00:44
	← Myrizio left	00:46
jasam	gitte, thanks!	00:53
gitte	Well, you deserve it!	00:53
	How are you? All going well?	00:54
jasam	yes	00:55
	← KirinDave- left	00:56
jasam	just I should move to another thing instead builtin-tag.c	00:56
gitte	I decided to continue working on filter-branch BTW...	00:56
jasam	fine!	00:56
	good!!	00:56
gitte	Yes. Are you finished with Junio's comments?	00:56
jasam	yes, but I'm thinking on the editor.c file	00:57
	I don't know if shipping it or just release builtin-tag.c	00:57
gitte	Release builtin-tag.c	00:57
	Definitely.	00:57
jasam	ok	00:57
	→ dduncan joined	00:58
jasam	something I did not well, was not numbering the changes to the last patch, now I don't know if depend on the new read_fd() or use the old read_pipe() instead	00:59
gitte	Heh. From time to time I confuse myself, too ;-)	00:59
jasam	what you would do in such case?	01:00
	→ dash_ joined	01:00
gitte	I'd make a temporary branch, and cherry-pick the patch.	01:00
	In your case, the builtin-tag.c patch.	01:00
	("git checkout -b temp junio/next" is one command that I use very, very often)	01:01
	If it works without cherry-picking the other patch, it is fine.	01:01
jasam	so, is the patch in next?	01:01
gitte	Which one?	01:01
	Just look at "git log junio/next --author=Carlos"	01:02
jasam	I'm pulling	01:03
	hey! I'm more used to do this now!	01:04
	before was a pain	01:04
	now just pulling and dig a little on the origin/branches	01:04
	→ rtmfd_icbm joined	01:05
jasam	and I'm using the same directory again and again to retrieve the last version and develop on it	01:05
	the only problem is that I always reset my changes before pulling	01:05
gitte	Wonderful!	01:05
	Why not use "git stash"?	01:05
jasam	is it recent?	01:06
mwc	jasam: it's coming in 1.5.3	01:06
	jasam: if you're using the stable vers, you don't have it yet	01:06
jasam	I'm not sure what that is for	01:07
	documentation talk about save	01:08
gitte	jasam: git stash will put your current changes into an own branch.	01:08
mwc	apparently mercurial has something similar. It lets you move some changes out of the way if you're not ready for a commit, do something, and then bring them back	01:08
gitte	IOW all that "git status" shows, will be gone from the working tree, but you will be able to reapply it with "git stash apply".	01:08
mwc	I wonder... does git-stash use a stack, or is it a one-shot deal	01:09
	IMAO, a per-branch stack is the way to go	01:09
jasam	but, could that be achieved also using branches or it has something different?	01:10
	→ mwc_ joined	01:10
	→ Romster joined	01:10
	← dash__ left	01:11
gitte	It could be achieved using branches.	01:11
	However, it also stashes away the index. So if you were in the middle of something important, but need a certain fix which is included in the upstream, do this:	01:12
	git stash && git pull <upstream> <branch> && git stash apply --index	01:12
	You will have merged your current work in progress with the upstream.	01:12
	"git diff" and "git diff --cached" will show the appropriate thing (if git stash apply succeeds, that is...)	01:13
jasam	ok, you can choose not to commit to save your temporary changes	01:13
gitte	Exactly.	01:13
	It is "sort of" committing onto a different branch.	01:13
	But better.	01:13
jasam	:-)	01:13
mwc_	right now, I just git-diff HEAD > ~/foo; git-reset --hard HEAD; blah blah blah; patch -p1 < ~/foo	01:13
gitte	No need to initialise a temporary branch, and "git show stash" will show you the combined diff between HEAD, index and working tree!	01:14
	mwc_: yes, that is what I used to do, too. Only i used "git apply".	01:14
	But it has one important shortcoming: it is no 3way merge.	01:14
mwc_	indeed	01:15
gitte	And sometimes, a threeway merge just rocks.	01:15
	← devogon left	01:16
jasam	well, I have to go, tomorrow I will send builtin-tag.c	01:17
	thank you for your help and explanation gitte	01:18
gitte	jasam: good! And you're welcome.	01:19
	I will be online tomorrow afternoon, until roughly 6pm your time...	01:19
jasam	I think I will connect before that	01:20
gitte	;-)	01:20
jasam	so I can ask you if I need!	01:20
gitte	Yes, that's the idea!	01:20
	Good night, jasam!	01:20
jasam	good night...	01:20
gitster	goodnight ;-)	01:20
jasam	thanks again	01:21
	← jasam left	01:21
gitte	Hey gitster!	01:22
	A couple of days ago, you wanted to talk to me?	01:23
gitster	Perhaps I did, perhaps I didn't.	01:24
	I think I did, but I do not remember what it was about.	01:25
	Which probably means either it was not so important, or it was already resolved.	01:25
gitte	Great!	01:25
	I thought about what you wrote with regards to filter-branch.	01:25
	And I agree, after quite some discussion with myself...	01:26
	So my plan for the next days is to teach filter-branch to rewrite multiple branches, to overwrite refs (putting the old values into refs/original/),	01:28
	to teach "rebase -i" to squash multiple commits nicely (not popup an editor everytime, but just the last time),	01:29
	and finally take a look at that worktree stuff.	01:29
	Also, I found a very subtle xdl_merge() bug, which I thought resolved.	01:30
gitster	Hmph --- which is...?	01:30
gitte	Remember the bug squashed in v1.5.0-rc1~145?	01:31
	It resurfaced again.	01:31
gitster	The same bug, or found a similar corner case?	01:33
	If the former we really should have added a test script or two...	01:33
gitte	Yes, that goes without saying.	01:34
	The only reason I did not include the test case (which was from Wine), was that it is too large.	01:35
	No longer a problem: this time, it was a merge in builtin-branch.c (IOW your latter case).	01:35
	I still have the "git branch --sort-by-date" in my local tree, and use it.	01:36
	And that stuff did not merge cleanly with the --track cleanup I did recently, even if it should have had.	01:36
	Again, a problem of the wrong attribution of an empty line.	01:36
	→ anculz joined	01:37
anculz	http://rectum.antiville.fr/	01:37
	http://rectum.antiville.fr/	01:37
gitte	spam.	01:37
	← lcapitulino left	01:38
gitte	Anyway, since it is easily reproducable from two certain versions of builtin-branch.c which are included in git.git, and a version that can be constructed by a very short patch, I will write a test case this time.	01:38
	← rtmfd_icbm left	01:39
	→ JoaoJoao joined	01:40
JoaoJoao	hello	01:40
gitte	Hello JoaoJoao	01:43
	← anculz left	01:44
	← JoaoJoao left	01:51
	← kc5tja left	02:10
	← QuercusMax left	02:33
	→ kanru joined	02:34
	→ saintdev_ joined	02:44
	← saintdev left	02:44
	saintdev_ → saintdev	02:44
gitte	Good night!	02:44
	← gitte left	02:44
	← MrPloddy29` left	02:45
	→ MrPloddy29 joined	02:48
	→ jcollie joined	02:50
	← MrPloddy29 left	02:52
	→ MrPloddy29 joined	02:53
	→ mosno joined	02:53
	← bronson left	03:19
	← madewokherd left	03:29
	→ jlf joined	03:35
	→ fhobia joined	03:37
	← aroben left	03:47
	→ spearce joined	03:50
	← jcollie left	03:53
	→ KirinDave- joined	03:54
	← tchan left	03:56
	→ tchan joined	03:56
	← fultilt left	04:11
	→ FunkeeMonk joined	04:13
	→ tcoppi_laptop joined	04:14
	← rkaway left	04:26
	← KirinDave- left	04:29
FunkeeMonk	how do I install the git man pages?	04:29
	→ rkaway joined	04:30
FunkeeMonk	oops, nevermind, found it	04:31
Randal	make install-doc	04:31
	→ aroben joined	04:32
	← aroben left	04:32
	→ aroben joined	04:32
	→ spearce_ joined	04:37
	← spearce left	04:38
	← aroben left	04:40
	← brother1 left	04:42
FunkeeMonk	argh, so many prerequisites for installing the docs	04:45
gitster	Merlyn, can I ask you a Perl question?	04:45
spearce_	FunkeeMonk: yea, i usually don't build them.	04:45
	spearce_ → spearce	04:45
	→ aroben joined	04:45
FunkeeMonk	argh, this is just too painful	04:49
	where can I download the html docs for git instead	04:50
spearce	clone git, there is an html branch which has the preformatted html documentation stored on it.	04:50
Randal	"html" ugh	04:56
	kids... not getting that "man FOO" is useful. :(	04:56
FunkeeMonk	I tried installing the man pages, its just too painful on Mac OS X	04:57
aeruder	FunkeeMonk: its a bit buggy (the xmlto stuff is in darwinports)	04:59
FunkeeMonk	aeruder: Yeah. Exactly my point. Painful. :-)	04:59
aeruder	http://dump.aeruder.net/docbook.catalog.xml -- throw that in your home dir as .docbook.catalog.xml	05:00
	when you compile git, try a line like	05:00
	XML_CATALOG_FILES=~/.docbook.catalog.xml make -j 3 PERL_PATH=/opt/local/bin/perl prefix=/usr/local all doc && sudo make PERL_PATH=/opt/local/bin/perl prefix=/usr/local install install-doc	05:01
Randal	it's working fine for me. :*	05:01
	oh well	05:01
aeruder	xmlto in darwinports has some issues	05:01
Randal	I'm usin git.	05:01
	→ kristoffer joined	05:01
Randal	using it	05:02
aeruder	yes, xmlto has some issues	05:02
Randal	Oh - I'm using fink xmotl	05:02
	fink xmlto	05:02
aeruder	ah, yea, that would make a difference ;)	05:02
Randal	the fink one works. :)	05:02
FunkeeMonk	ah okay	05:02
aeruder	anyway, with darwinports, the thing up there ^^^ works fine	05:02
FunkeeMonk	might just use fink	05:02
Randal	0.0.18 in fink	05:03
	not sure how that compares to ports	05:03
aeruder	its not an issue with xmlto per se, i think its just how they've configured/installed it	05:03
	it doesn't find any of the local docbook stuff without that file up there	05:03
	so it attempts to go out and download them all	05:04
	which asciidoc tells xmlto not to do	05:04
	so it just bombs out	05:04
Randal	hmm	05:04
	I have a shell script that runs on git.git without ever invoking xmlto	05:04
	and still provides manpages	05:05
	← saintdev left	05:05
aeruder	asciidoc runs xmlto internally i believe	05:05
spearce	Randal: from the man branch perhaps?	05:05
Randal	... cd /opt/git/src/git-man && (git-fetch -u; git-checkout -f; rsync -avxP man? /opt/git/man/)	05:05
aeruder	or the man branch ;)	05:05
Randal	yah	05:05
	that's a "man" branch checout	05:05
	I guess I could have simply done git-export man	05:06
	but this was before git-export, if I recall.	05:06
spearce	what's this git-export thing you speak of?	05:06
Randal	git-tar-export?	05:06
	something that makes tars	05:07
aeruder	git-archive ?	05:07
spearce	aaah. you mean git archive --format=tar. but ok.	05:07
Randal	yeah	05:07
	sorry	05:07
	git-archive --format=tar man	05:07
	that would have made my man-branch as well	05:07
	→ devogon joined	05:10
	← mwc left	05:13
gitster	In gitweb.perl, if I wanted to see if "sub gitweb_foo" exists at runtime, what is the kosher way to do this?	05:13
spearce	gitster: what was your encoding that git-gui choked on?	05:14
gitster	I think it was UTF-8, let me see...	05:14
Randal	if defined &gitweb_foo	05:15
gitster	Hmph, what if I have 'foo' part in a variable?	05:15
Randal	or if exists &gitweb_foo if you want to know if it has a prototype, although not ncesssarily an implementation	05:15
	oooh. please don't.	05:15
spearce	gitster: ok, i copied paul's table from gitk and UTF-8 works.	05:15
Randal	{ no strict 'refs'; if defined &("gitweb_$some_var"} ... }	05:16
	you're using symrefs at that point. please don't.	05:16
gitster	I cheated and did this:	05:16
	aeruder thinks there's a bracing error on that one	05:16
gitster	eval "gitweb_bc_feature_$name()";	05:16
aeruder	;)	05:16
Randal	No No NO	05:16
	do not use eval	05:17
gitster	I just wanted to say "if such a function exists, call it, but if there isn't I do not care" ;-).	05:17
Randal	if the answer is eval, you've asked the wrong question. :)	05:17
gitster	I wanted to see an answer that does not resort to eval because I knew it was wrong.	05:17
spearce	q: "How can i let a random web user destroy my computer?" a:" eval $input;".	05:17
	see, eval is the answer.	05:17
Randal	if (defined &{"gitweb_bc_feature_$name"}) { ... }	05:17
	within { no strict 'refs' }	05:18
	presuming you are sane and said "use strict" somewhere	05:18
gitster	Ouch, I was about to ask "do I need to be in "no strict refs"". Ok but that is much better than eval, I would say, even though $name is totally under my control at that point in the code.	05:19
Randal	it's still a sign of a bad design	05:19
	and by bad, I mean very, very bad	05:19
FunkeeMonk	hehe, "very bad" - reminds me of this - http://fishbowl.pastiche.org/2007/07/17/understanding_engineers_feasibility	05:20
spearce	right, like why not build a hash of $name => \&gitweb_bc_feature_name for each possible name?	05:21
Randal	exactly	05:21
gitster	what's so wrong to do introspection of the module?	05:21
Randal	it's pretty f'ing rare that you want to let data become code	05:21
gitster	meaning, having the language execution environment to build that hash for me?	05:21
Randal	and in those instances, you'l accept the loss of control	05:21
aeruder	gitster: fwiw, i say shrug	05:22
Randal	the trouble is for the most part, people cheat.	05:22
aeruder	that's half the fun of using something like perl/python/ruby/etc..	05:22
Randal	well, fun for you	05:22
	not fun for your maintainer	05:22
	glad that you have that feeling. your maintainer thinks otherwise :)	05:23
aeruder	sometimes its done to make maintainership easier	05:23
Randal	eh. no.	05:23
aeruder	like not having to maintain a huge mapping of names to functions when the functions are all named prefix_$name	05:23
Randal	I will bet the rent money on that.	05:23
	No.	05:23
	and let me restate that. definitely... no.	05:24
aeruder	oh, spare me the gospel according to Randal	05:24
	:-P	05:24
Randal	ok - fuck you then too.	05:24
	seriously... been there, done that.	05:24
	no t-shirt.	05:24
	I will not discount your appreciation of my 30 years of programming history.	05:25
	aeruder rolls his eyes	05:25
Randal	But I will clearly label it as such.	05:25
aeruder	wait, is this the part where you start ranting about perl book authorship, et al. ?	05:25
Randal	No. Just history.	05:26
	I got scars, man.	05:26
	you don't respect those, it's your life.	05:26
aeruder	shrug	05:26
Randal	I will call it for what it is though.	05:26
	how many years you been hacking code?	05:26
	better yet, for hire?	05:27
aeruder	is that how it is when you get old, you justify your worth on how long you've been doing something?	05:27
Randal	No, based on numbers of scars.	05:27
	And that is linear with years.	05:27
gitster	... and the fact that you survived ;-)	05:27
	→ lyakh joined	05:27
Randal	if you can't see that, you're still a kid. :)	05:28
	gitster - precisely	05:28
	I don't ask for people to believe me. Only to weight what I say based on my scars.	05:28
loops		05:29
Randal	smart peopel do that.	05:29
	idiots are doomed to rediscover that. :)	05:29
FunkeeMonk	All this talk about experience is reminding me of this guy: http://blogs.ittoolbox.com/pm/irm/archives/theory-p-the-philosophy-of-managing-programmers-4993	05:29
Randal	so aeruder - what's your history?	05:29
gitster	even if you are not smart, if you are old enough to have scars on yourself....	05:29
Randal	gitster - precisely	05:30
aeruder	what do you mean, what is my history, as i've already stated, length of time doing something does not any way shape or form relate to skill or knowledge	05:30
FunkeeMonk	"The elements of Theory P were not derived from casual observations. Rather, they are based on fifty years of practical experience in the field, in a variety of industries."	05:30
Randal	if you think that, you must be at a disadvantage here. :)	05:30
aeruder	so i'm not going to take your opinion on "I've been doing this longer than you buddy boy!! ho ho ho!"	05:30
FunkeeMonk	"Programmers exhibit an average intelligence level, no greater than any other professional with a college degree. They exhibit an average imagination."	05:30
	"Perhaps more than any other profession, programmers try to impress and intimidate others with their technical jargon. Such language usually masks inadequacies elsewhere."	05:31
Randal	aeruder - would it trouble you to answer the question for the onlookers?	05:31
FunkeeMonk	And those are words from a person with "fifty years of practical experience"	05:31
Randal	OK - I'm guessing so. In other words, you don't think enough of yourself to be able to answer that question.	05:32
	time for us to discount EVERYTHING you say.	05:33
aeruder	huh?	05:33
Randal	then answer the simple question, please.	05:33
	Ahh. no answer.	05:34
	love the youth. think they are brilliant. and yet...	05:34
aeruder	i've contributed 19 patches to git (albeit mostly documentation), 2-3 to gcc, i've contributed patches to lejos, xmms2, gnustep, bmp	05:34
Randal	Yeah, whatebver.	05:34
	answer the f'ing quesiton.	05:34
aeruder	i'm the maintainer of sidestep, a yet to be released (but will be in the next month or so) fork of gnustep-base	05:34
Randal	OK, how many years have you programmed?	05:35
	simple question.	05:35
	← kristoffer left	05:35
Randal	answer the f'ing question.	05:35
	or not.	05:35
aeruder	netclasses ( an asynchronous networking lib), talksoup (an irc client), zirc (an irc client written in zsh), maintainer of trustees kernel acls, pam-sessionrun, and two firefox extensions	05:35
Randal	and in your lack of answer, you say where you are.	05:35
gitster	It is like asking how old are you isn't it? Some people (including me) choose rather not...	05:35
Randal	someone to be discounted when push comes to shove.	05:35
aeruder	and what that has to do with age, i have no fscking clue	05:35
Randal	You have things to contribute, sure.	05:36
gitster	heh, ohloh agrees aeruder has 19 commits in git.	05:36
Randal	but if your observations contradict mine, the sane person will pick the elder.	05:36
	that's all I'm saying.	05:36
aeruder	uh huh	05:36
Randal	and since you're being a jerk about answeirng my question, that's even more clue not to listen to you.	05:37
	idiot.	05:37
aeruder	i'm 23	05:37
FunkeeMonk	But you're the one calling people names Randal	05:37
Randal	how many years of programming.	05:37
	I'm calling names because I'm tired of his games.	05:37
aeruder	hm, quite a few, maybe 10?	05:37
Randal	I apologize for that. but I hope my point was made.	05:37
aeruder	no, your point is usually masked in your arrogance	05:38
FunkeeMonk	As an observer, if I didn't already know who you were, I wouldn't have thought that you have 30 years of experience.	05:38
	Because of the namecalling.	05:38
loops	Randal, your experience in of itself is worthless to others. What you bring to the table in any given debate is what matters. So if your experience lets you give insight that wouldn't otherwise be offered, great. But saying.. trust me i have experience is a classic logical fallacy "appeal to authority"	05:38
aeruder	i can pull up some past conversations you've had with others where within two lines of a question you're calling them an idiot	05:38
	← yorgen15 left	05:38
Randal	if I did, it's beacuse they were.	05:38
	I make no concessions for peopel who haven't had experience.	05:39
FunkeeMonk	Seriously, everyone should read this post and see what 50 years of experience can bring you: http://blogs.ittoolbox.com/pm/irm/archives/theory-p-the-philosophy-of-managing-programmers-4993	05:39
Randal	it's not an appeal to authority. it's an appeal to experience.	05:39
aeruder	Randal: experience != some claim to being a jackass	05:40
loops	Randal, that is by definition an appeal to authority... the authority of experience	05:40
	→ yorgen15 joined	05:40
gitster	Being great yourself and being great in explaining things to others are sometimes different. Even when you are right, other people cannot understand why you are right (because they are inexperienced).	05:40
loops	Randal, it is just clearly wrong to think that your experience means you're always right on every issue. Therefore you must defend your positions with facts.. not your resume.	05:40
Randal	not at all	05:40
aeruder	ever heard that phrase you don't gain respect, you earn it? ;)	05:40
Randal	You know, at this point, I don't care.	05:41
	I have a history. I've made it to here.	05:41
gitster	Some people say "I have experience, trust me", because it is too much effort to explain the background from experience.	05:41
Randal	If you don't want to use that, fine.	05:41
aeruder	ok then	05:42
loops	gitster, sure. if you don't want to make the effort that's fine. But it's not a way to convince others of the merit of your point.	05:42
Randal	If you're so youthfully arrogant that you think that the elders have nothign for you, then in fact, I have nothing for you. Whatever.	05:42
spearce	gitster: i usually find myself saying "I have experience, trust me" and then immediately launch into the background anyway. makes for some long conversations. ;-)	05:42
Randal	All I can say is, it must suck to be you.	05:42
aeruder	Randal: i'm not saying that you have nothing for me, i'm saying that you come off so arrogant that you tend to piss people off faster than you can exude your wisdom	05:42
Randal	yeah, so says you.	05:43
gitster	I see many good people do that "trust me", but great ones explain themselves when people ask right questions in response to the "trust me".	05:43
Randal	good thing others are quick to recognize something else.	05:43
loops	Randal, that's not what i'm saying at least. I respect those who have earned their success.. and i can respect your achievements.. but that's very different from accepting everything you say as fact.	05:43
Randal	I don't wish that everythign I say be accepted as fact.	05:43
	I merely ask that you considre that maybe I have scars to back up what I say.	05:43
gitster	rather, "I'd want to accept but I'd like to 'understand'".	05:43
	and "trust me" does not lead to understanding, unfortunately.	05:44
loops	gitster, exactly.	05:44
spearce	no, it never does.	05:44
	→ janm joined	05:44
Randal	so I'm not just a guy making up random crap	05:44
	or a youthful guy speaking from theory	05:44
spearce	and what's the point of being an 'elder' if you cannot impart your tribal wisdom and battle scars on the younger generation, so that it lives on longer than you?	05:44
Randal	again, you can ignore that. Oh well.	05:44
	spearce - sometimes, it has to be a matter of folklore rather that complete jstification.	05:45
	gitster thinks Merlyn can always say "that is my work -- I am just having fun here without getting paid" ;-)	05:45
aeruder	accepting things because someone else says so isn't generally the fastest road to bettering yourself in anything in life	05:46
	loops changes channel title to The Tao Of Git	05:46
aeruder	heh	05:47
FunkeeMonk	how old is Bush?	05:47
	gitster changed the topic to: 1.5.2.4/1.5.3-rc2 \| Everyone asleep or clueless? Try [email@hidden.address] \| Channel log http://colabti.de/irclogger/irclogger_log/git \| Linus on git: http://youtube.com/watch?v=4XpnKHJAok8 \| Mailing list archives: http://news.gmane.org/group/gmane.comp.version-control.git/	05:47
Randal	aeruder - well - I'm not trying to prevent you from having to learn what I've learned on your own. but you can in fact save some time. )	05:47
	aeruder thinks FunkeeMonk is heading down a treachrous road	05:48
FunkeeMonk	I wonder if he uses this argument too when convincing others.	05:48
	"I have XX number of years in politics - how many years do YOU have?"	05:48
loops	aeruder, if you're honestly trying to help people save time, then explaining things without the use of "trust my experience" is much more effective.	05:48
spearce	no, his argument is he is the decider, and he decides things.	05:48
loops	oops.. aeruder -> Randal	05:48
aeruder	loops: i was gonna say, i have no experience :-P	05:49
	i won't be making those claims ever	05:49
	← doublec left	05:49
loops	Randal, your vast experience should give you the tools to prove your points without needing to cast yourself as an authority.. your experience will simply make your arguments compelling.	05:50
Randal	Yeah, if I wanted to spend time justifying my experience, sure.	05:51
	But really, again, I don't care. If you want to ignore my experience, that's your choice.	05:52
loops	Randal, it's not justifying your experience.. it's justifying your position on a given point.	05:52
	spearce tries to decide if octopus merges are useful to have in git-gui still...	05:52
Randal	You are free to relive my scars.	05:52
	I really don't f'ing care.	05:53
loops	spearce, why wouldn't they be?	05:53
spearce	loops: i'm rewriting the UI for the merge dialog, because it blows. ;-)	05:53
	i have a widget i usually use for selecting commits, but it picks only one commit.	05:53
Randal	I was just trying to help.	05:53
gitster	Nah, octopus was just historical curiosity without real life advantage.	05:53
spearce	you mean that famous 12 way merge that gittus is still not a fan of?	05:54
aeruder	12 way merge, heh	05:54
spearce	day-job has two 15-way's right after each other. just seconds apart. 'cause we needed a 30 way merge.	05:54
loops	lol	05:54
aeruder	very nice, very nice ;)	05:54
spearce	yea, its insane and could have been 30 individual merges.	05:54
aeruder	but i bet the gitk diagram looks cool ;)	05:55
spearce	yea. gotta tweak that thing out to allow like 60 branches at once and give it 75% of the view. :)	05:55
	(there's yet more branches that were much much later, but started further back in time....)	05:55
aeruder	ah, i've love to see a screenie of that if you ever think of it, even if its private and it was just a picture of the diagram, it'd be neat	05:55
	make a poster out of it and hang it on my wall :)	05:56
spearce	like the internet posters? or the linux kernel calls? we should do a famous git history poster. :)	05:57
aeruder	exactly	05:57
	get some nasty areas of several famous git projects (git, kernel, etc..)	05:57
spearce	ok, so no ocotopus in git-gui anymore.	05:57
aeruder	or heck, do something like linux kernel calls and have the entire gitk history diagram of git in microprint	05:58
	man, this empty dir thing i think is just going to turn into a mess	06:04
Randal	some stupid unix-like implementations. :(	06:05
	unlink on a dir shoudl always falk	06:05
	fail	06:05
spearce	Unless you paid good money for your operating system. :-)	06:05
aeruder	nah, i'm talking about the big "i want to add empty dirs" thread	06:06
spearce	yea, i know what thread you are talking about.	06:06
aeruder	just throws so many things out of whack	06:06
spearce	and Randal pointed out another that was equally er, strange.	06:06
aeruder	yea, some old unices, the dirs were just files	06:06
Randal	but never responded to unlink() that way	06:07
aeruder	yea, that thread is messed up too	06:09
spearce	i'm gonna unlink me a directory! ;-)	06:11
	aeruder unlinks spearce's directory first	06:11
spearce	heh, i still have a "." link to myself. you can't get rid of me!	06:11
	aeruder destroys ..	06:11
aeruder	your parent is dead	06:12
	gitster sees kids who never saw pre V7 unices ;-)	06:12
aeruder	wait, was i alive when that was around? :-P	06:12
spearce	i know all about that directory link mess. i'm just surprised a UNIX today still supports that mess. "just in case". i wonder what customer still uses UFS that way and won't replace their rmdir utility...	06:13
gitster	I bet Solaris still has ncheck/icheck/dcheck.	06:14
	aeruder would not be surprised at all if it weren't some multi thousand dollar engineering app	06:14
	← FunkeeMonk left	06:14
spearce	if i had a solaris box handy i'd look...	06:14
aeruder	s/weren't/were/	06:14
	those are generally the worst	06:14
gitster	Google is usually pretty good source -- people have solaris man pages available somehow.	06:14
spearce	i remember reading in their "halt" manpage one day about how its useful if your processor is on fire. i wonder how often that happened that they had to document it.	06:14
aeruder	use halt when a moth flies into your cpu and starts eating at the wirewrap terminals	06:15
	spearce: some of that old sun hardware is pretty robust, probably could start on fire, turn it off, let it smolder a bit, fire it back up and good as new (just slightly charred and funny smelling)	06:17
	→ lyakh_ joined	06:21
	← lyakh_ left	06:23
	← aroben left	06:26
spearce	no more octopus from git-gui. and now time for bed. because i am too tired to type the word 'bed'. ;-)	06:26
	← spearce left	06:27
	→ KirinDave- joined	06:27
	→ Pistakk joined	06:27
	→ etnt joined	06:28
	→ aroben joined	06:28
	← Pistahh left	06:29
	→ Pistahh joined	06:29
	← Pistahh left	06:31
	→ Pistahh joined	06:31
	← Pistakk left	06:32
	→ evilchelu joined	06:33
	← DrNick left	06:36
	← lyakh left	06:37
	← fhobia left	06:45
	→ evilchelu_ joined	06:54
	→ hackbert-hh joined	07:08
	← evilchelu left	07:08
	→ saintdev joined	07:14
	→ r0bertz joined	07:19
	saintdev → saintdev_	07:23
r0bertz	can git-format-patch set To: field?	07:27
Randal	why would i?	07:30
	it?	07:30
vmiklos	you are looking for git-send-email	07:31
Randal	Yeah, perhaps	07:33
r0bertz	ok, thanks	07:35
	← dduncan left	07:35
r0bertz	can git-send-email use msmtp?	07:36
	seems that msmtp doesn't qualify, cause it doesn't support -i	07:38
vmiklos	it uses the sendmail binary so it should be able to use any smtp	07:38
	umm	07:38
	that's possible:)	07:39
	← mwc_ left	07:41
	← KirinDave- left	07:45
	→ fhobia joined	07:47
	→ fultilt joined	07:54
r0bertz	if i have a series of patch, can i just 'git-send-email *.patch --to ...."?	07:59
	← johan-s left	07:59
osfameron	has anyone got a walkthrough or tips on how to bring an existing home directory into git version control?	07:59
	I'm getting confused by the fact that unlike svk you can't add a directory without adding all of its contents	08:00
	but... you can add its contents without adding the directory	08:00
vmiklos	r0bertz something like that, yes	08:00
evilchelu_	osfameron: you can add .gitignore files to ignore certain bits	08:00
osfameron	... which leads me to realise that I don't understand the repository model at all, and am probably doing things very wrong	08:00
r0bertz	thank you, vmiklos	08:00
osfameron	evilchelu_: oh right, that would be a workaround, yes	08:01
	is it wrong to add a subdirectory/file without having added its parent?	08:01
Tali	osfameron: git doesn't track directories, only files	08:03
	osfameron: so you can't "add a directory"	08:03
	osfameron: just add all the files you want to track	08:03
osfameron	Tali: ah	08:08
	Tali: but when I check them out, I want to be able to check out certain trees of files (for example, my homedir)	08:08
	→ cort joined	08:13
	← evilchelu_ left	08:21
	← hackbert-hh left	08:24
Tali	osfameron: sure, it just works	08:25
	osfameron: git always tracks the complete path to the file, but you don't have to tell it explicitly that it should care about parent directories. This is done automatically	08:26
osfameron	ok! that seems sensible (and I was pleased when it let me do that automatically)	08:27
	but I'm lacking some concepts to work out how to, for example, check out a copy of stuff I've checked in under /home/osfameron in a temporary directory	08:27
	man git-checkout says I need to give a repository location but I don't know how to refer to it (like // in svk)	08:28
Tali	osfameron: git checkout only works on your current working area and its associated repository	08:31
	osfameron: if you want to create a new working area in another directory/machine you can use "git clone"	08:31
osfameron	ah! ok, "checkout" is a false friend from sv[nk] then. Thanks	08:33
Tali	well, partly	08:33
	git checkout is the way to move data from the repository to your working area, in that sense it's the same	08:34
osfameron	like sv[nk] update ?	08:34
Tali	git clone is the way to create a mirror of the repository	08:34
	as each working area has its own repository in git you need clone to create a new working area	08:35
osfameron	does git clone imply having another .git/ directory ?	08:35
mugwump	that's right	08:35
	← sgrimm_ left	08:35
mugwump	osfameron, did you read my git-svn article? It's targeted at svk users	08:35
Tali	well, update is a different thing	08:35
mugwump	svn update is like git rebase :)	08:36
	osfameron panics	08:36
mugwump	actually it's worse, because you can't recover with svn if the update ruined your day	08:36
osfameron	mugwump: yes I did, but I think I need to reread that and a couple of the other tutorials a few times to work out what's going on, I'm getting very confused	08:37
mugwump	ok. I'd love feedback with what you thought was missing	08:37
	→ evilchelu joined	08:37
	← evilchelu left	08:37
	→ Dodji joined	08:38
osfameron	I don't like the idea of multiple .git/ repos, I get the feeling I'd mislay them :-)	08:38
	→ evilchelu joined	08:38
mugwump	they stick around with your checkouts though	08:38
	I used to find with svk that I felt I couldn't move stuff around in my src/ because svk would lose track of what it was	08:38
	and the original problem was really just irritation with grep -r et al	08:38
	you can separate it, write the glue if you really want to work like that. You just need to make sure that GIT_DIR gets set correctly	08:39
	osfameron sucks at sysadmin :-(	08:39
osfameron	which is part of the reason I want to impose discipline and version control on my homedir of course...	08:40
mugwump	sure. Just don't use 2gb mboxes like Joey Hess :)	08:40
osfameron	hehe, though having email records going back that long is quite cute I don't think it's realistic for me	08:41
	→ johan-s joined	08:42
mugwump	would work fine with maildir though	08:42
osfameron	I'm not even going to keep my svk version history because I can't figure out git-svnimport and I'm just going to start afresh with git instead	08:42
mugwump	and as you moved files around it would be the same sha1 so very little extra space :)(	08:42
	you can use git-svn on your mirror paths	08:42
	saves re-fetching them	08:43
	but yeah, clean slate is good	08:43
osfameron	ah, I'll look at git-svn	08:43
	svnimport seemed to want branches and tags and things	08:43
mugwump	well git-svn does too,	08:43
osfameron	which of course I didn't bother with in my homedir import originally	08:43
mugwump	oh right	08:43
	single branch works as well	08:43
	→ cedric__ joined	08:47
	→ sgrimm joined	08:48
Fritti	mugwump: where is that git-svn article?	08:56
	Fritti is interested as well	08:56
mugwump	utsl.gen.nz/talks/git-svn/intro.html	08:57
Fritti	thanks	08:58
	oh hey, I saw that once, and promptly forgot to bookmark it. Nice! :)	09:00
mugwump	I should really write more on dcommit	09:08
Fritti	just from reading (my git svn fetch is not complete yet :-) I'm a little confused on the section on how to track the upstream SVN repository	09:10
	http://utsl.gen.nz/talks/git-svn/intro.html#howto-track-svn	09:10
	is there a paragraph missing there?	09:11
mugwump	yes, there could be an intro paragraph there	09:11
	see the "Fetching more revisions" bit	09:11
Fritti	ah, ok, I didn't connect that section with the one I linked	09:12
	so just another git svn fetch	09:12
mugwump	yep	09:13
Fritti	mugwump: while cloning your article repository: error: File 3f727b6bdc6fe19c3e59a7fedb2a89963942004e (http://utsl.gen.nz/talks/git-svn/.git/objects/3f/727b6bdc6fe19c3e59a7fedb2a89963942004e) corrupt	09:17
mugwump	:(	09:17
	Some loose object were found to be corrupt, but they might be just a false '404 Not Found' error message sent with incorrect HTTP status code.	09:24
	is what my git reports	09:25
	and it's right	09:25
	lighttpd--	09:26
	→ nzkoz joined	09:26
Fritti	ah, ok	09:26
	→ sgrimm_ joined	09:30
	← sgrimm left	09:31
nzkoz	hey guys, I'm using git-svn to use git with one of my client's repositories, and it's ... well amazing really. But I have a question about the branch support, anyone around who knows much about it?	09:31
	specifically these guys have their release branch at /release not /branches/release, and I'm wondering if there's a way to get that imported into git alongside the other branches?	09:32
mugwump	nzkoz: set it up as a second git-svn remote	09:33
	I think that works	09:34
	however what you are asking for should be on the TO-DO list	09:34
	← aroben left	09:38
nzkoz	awesome, that appears to be working :)	09:40
	thanks mugwump	09:40
	← fhobia left	09:46
	← sgrimm_ left	09:48
	→ sgrimm joined	09:54
	→ nzkoz_ joined	09:57
	← nzkoz left	09:57
	→ chris2 joined	09:59
mugwump	hmm, git-diff-files doesn't show untracked files... anyone know the right plumbing command for that?	10:00
	git-runstatus is still porcelain	10:04
	git ls-files -o -m -d -t	10:05
	is what I was looking for	10:05
	→ sgrimm_ joined	10:07
	→ nud joined	10:14
	← rkaway left	10:16
	→ rkaway joined	10:19
	← sgrimm left	10:24
vmiklos	or just git status	10:31
	oh, plumbing, then nvm	10:32
	→ branstrom_ joined	10:34
	← nzkoz_ left	10:35
	→ ofri joined	10:43
	← branstrom left	10:55
	← fultilt left	11:01
Fritti	is it a problem to run git-update-server-info from 1.4.1 on the server, where I rsynced my 1.5.2.2 created repository?	11:04
	→ branstrom joined	11:06
Fritti	mugwump, did you get my privmsg?	11:18
mugwump	merged and updated, thx :)	11:20
	→ QuercusMax joined	11:20
	→ bentob0x joined	11:20
mugwump	lol, I just made a new commit on the webserver by mistake	11:23
Fritti	thanks! I wasn't sure it got across due to that freenode registration policy. And now I've done my first git development work! :)	11:23
mugwump	so I closed the shell	11:23
	git-pull utsl master	11:23
madduck	does anyone know why git's make install or install-doc creates /usr/man/man3/* ?	11:23
	and how to change that?	11:23
mugwump	make mandir=/usr/share/man	11:24
	see the package in feisty or lenny	11:24
	→ standel joined	11:24
	← standel left	11:25
	← branstrom_ left	11:25
	← ofri left	11:30
tsuna	$ ll .git/refs/remotes \| wc -l	11:33
	13	11:33
	$ git branch \| wc -l	11:33
	2	11:33
	Why don't I see remote branches in `git branch'?	11:34
	→ branstrom_ joined	11:34
aeruder	r0bertz: you can do something like this to set To/Cc	11:34
	git config format.headers "To: Junio C Hamano [email@hidden.address] [email@hidden.address]	11:35
loops	tsuna, git branch -r	11:44
	← QuercusMax left	11:46
aeruder	and git branch -a to show both	11:47
madduck	mugwump: well, it already installs all other manpages under $prefix/share/man	11:50
	just the man3 section is in share/../man	11:50
	mugwump shrugs	11:51
mugwump	don't see any special handling of that in the debian/rules	11:51
madduck	neither do i	11:51
mugwump	but Git(3pm) ends up in /usr/share/man/man3 on my system	11:52
madduck	i bet it's perl	11:52
	→ cmarcelo joined	11:53
madduck	piper:/usr/local/src/git> make install install-doc prefix=/usr/local/stow/git 2>&1 \| grep Git #[318]	11:53
	Installing /usr/local/stow/git/man/man3/Git.3pm	11:53
	← branstrom_ left	11:53
madduck	Writing /usr/local/stow/git/lib/perl/5.8.8/auto/Git/.packlist	11:53
	→ branstrom_ joined	11:53
	← branstrom left	11:54
mugwump	of course that might be ExtUtils::MakeMaker	11:54
madduck	INSTALLMAN3DIR = $(PERLPREFIX)/share/man/man3	11:56
	INSTALLSITEMAN3DIR = $(SITEPREFIX)/man/man3	11:56
	why oh wh?	11:57
	INSTALLVENDORMAN3DIR = $(VENDORPREFIX)/share/man/man3	11:57
	mugwump: debian/ubuntu use INSTALLDIRS=vendor	12:04
mugwump	because clearly only anal retendtive people doing dists care about using such newfangled paths as /usr/share/man	12:05
madduck	it's the MakeMaker default, so I don't think i need to bother submitting a patch of some sort.	12:05
	→ branstrom joined	12:05
mugwump	MakeMaker is a module with history	12:06
	going back to at least 1994	12:06
r0bertz	aeruder, thanks, how can i check all available git-config options?	12:15
	← sgrimm_ left	12:18
aeruder	r0bertz: there's a man git-config which lists most if not all of them	12:18
	unless you mean see what is currently set, in which case you can look at .git/config ~/.gitconfig	12:19
r0bertz	thank you, aeruder	12:19
aeruder	yep	12:19
	← MrPloddy29 left	12:19
	← tronical left	12:19
	← hpa left	12:19
	→ MrPloddy29 joined	12:21
	→ hpa joined	12:21
	→ tronical joined	12:21
	→ ruskie joined	12:21
	← branstrom_ left	12:25
	→ branstrom_ joined	12:25
	→ sgrimm_ joined	12:25
	← branstrom left	12:25
osfameron	mugwump: in your svn tutorial you mention that one of the ways of creating a branch is to: $ cp -a parrot parrot.my-branch	12:27
	that absolutely terrifies me :-)	12:27
mugwump	that's something that bzr, darcs, etc people are used to	12:32
	they just copy the whole damned thing, and it's a new branch	12:33
	by virtue of the fact it's a consistent snapshot and you start committing differently	12:33
	ie, the revision is independent of the location of the revision	12:33
osfameron	but how does git know which local repository it should talk to? And which path you are really in with respect to the root of the working copy?	12:35
mugwump	it's obvious	12:35
	if GIT_DIR is set, use that	12:36
	otherwise, look up until you see a .git	12:36
osfameron	oh, because ususally the .git repo is in the top level of the working copy?	12:36
	oh, or rather, above the working copy	12:37
	→ branstrom joined	12:37
mugwump	top level	12:37
osfameron	ok, and if GIT_DIR is defined elsewhere, then just carry on looking up until there is a wc directory named the same as the top level that the repo knows about?	12:41
mugwump	GIT_DIR is an env. var	12:41
	oh, right, no, it assumes you're at the top	12:42
	GIT_WORK_TREE is a new feature, which is how you specify that	12:42
	(might have the name wrong)	12:42
osfameron	oh... so if you are using GIT_DIR in env, you might be able to confuse a copy?	12:42
mugwump	right	12:42
osfameron	whereas if you cloned it then... oh, no, presumably that would depend on whether the repo was in the top level or elsewhere again	12:44
mugwump	clone doesn't need a working copy for the source	12:47
	→ krh joined	12:49
	→ gitte joined	12:52
	gitte is annoyed by that stupid empty directories thread.	12:53
	← branstrom_ left	12:54
	← nud left	12:58
	→ lcapitulino joined	12:59
	→ jcollie joined	13:01
aeruder	gitte: i just think its going to turn into a mess	13:06
	spearce has already been making some good points, like having to keep track of which empty dirs you specifically want to keep so that when you remove a file you donn't have to remove all the dirs too	13:06
Fritti	that seems strange	13:07
	if I do rm subdir/file I don't expect subdir to disappear	13:07
gitte	Fritti: that's not the same here.	13:07
aeruder	i do, if i added the file with add subdir/file	13:07
gitte	You need a marker that you want to keep the directory specifically.	13:07
Fritti	fwiw I ran into the 'git doesn't track empty directories' myself 2 days ago when I started to investigate/learn git	13:07
gitte	And such a marker exists _dammit_	13:08
	It is called ".gitignore".	13:08
aeruder	yep	13:08
Fritti	shrug	13:08
	to me, an empty directory is content	13:08
aeruder	add an empty .gitignore	13:08
gitte	And only because some bonehead of talker-not-coder does not want to get it does not mean that I should have to suffer from such a spam.	13:08
Fritti	it might not be for you	13:08
	and I'll work around it using .gitignore because I don't have the knowledge to improve the current patch	13:09
	but it feels like a workaround	13:09
	that's why people will keep complaining	13:09
gitte	Fritti: to mark which directory was _explicitely_ added, as opposed to _implicitely_ by a "git add dir/file", you need a _marker_.	13:09
	It's _not_ a workaround dammit.	13:09
Fritti	I know	13:09
gitte	It is the most elegant solution.	13:09
	You say it _feels_ like a workaround.	13:10
Fritti	it's not a workaround IF you accept the fact that an empty directory does not fit in the git model	13:10
	yes	13:10
gitte	But that's only because you do not look at it from the _technical_ angle.	13:10
Fritti	that doesn't exclude one another?	13:10
	exactly	13:10
aeruder	certainly a heck of a lot more elegant than changing all the porcelain to suddenly deal with directories	13:10
Fritti	I'm a new, dumb, unknowing user	13:10
	and I accept that	13:10
	and I accept the fact that I need .gitignore now that I use git	13:10
	but can you understand that people coming from svn might not understand that at first?	13:11
gitte	It is not like empty directories do not fit into the git model.	13:11
	They do not fit into the content model.	13:11
Fritti	and there will be lots more people coming from svn in the future	13:11
	ok	13:11
gitte	Yes, but they should be willing to listen _and_ understand, or go away.	13:11
	Sorry, I am venting here.	13:11
Fritti	:)	13:11
	I don't mind	13:11
aeruder	Fritti: while that is true, if svn users are coming expecting things to be like svn, why are they using git? :-P	13:11
Fritti	I find it funny that the subject is brought up on the list 1 day after I run into that myself	13:12
gitte	But do you agree that you need _some_ way to mark an empty directory as tracked?	13:13
Fritti	of course	13:13
gitte	Okay, we have such a way.	13:13
	It just _happens_ that you can use ".gitignore" for something else, too.	13:13
Fritti	sure	13:13
	I can also use 'THIS_FILE_NEEDED_BECAUSE_GIT_TRACKS_CONTENT'	13:13
	with or without leading dot	13:13
gitte	:-)	13:14
Fritti	that's why it feels a bit funny	13:14
aeruder	the way i see it, git can still be a lowlevel thing	13:14
gitte	Yes, but since we already have .gitignore why introduce (and complicate) new things?	13:14
aeruder	if people want to store perms, empty dirs, etc.. on git, just build a layer over it, have scripts that interpret a .extracrap in the top-level to handle that stuff	13:14
Fritti	of course I'd use .gitignore (especially because most of the time the dir will not be empty while building etc)	13:14
	aeruder: sounds like a perfectly fine solution to me	13:15
mnemoc	hi, does svnimport detect copies between branches and trunk?	13:15
aeruder	heck, i'd rather have a .emptydirs file at the toplevel	13:15
	when git does a checkout, it reads that in, and creates those directories	13:15
	doesn't require any porcelain changes	13:16
	just like .mailmap or .gitignore	13:16
Fritti	that would work in my case	13:16
	I "need" it for templating software installs (web software)	13:16
aeruder	man, i'm tempted to code that up and send it to the ml	13:16
gitte	aeruder: I don't get it why ".gitignore" would not solve the problem already.	13:16
aeruder	true, actually i like your patch	13:17
	having git-add add empty dirs by adding an empty .gitignore	13:17
	← rkaway left	13:18
aeruder	although david kastrup's reply makes absolutely no sense	13:18
	placeholder name should rather be "."	13:18
	→ nikodemus joined	13:18
aeruder	i think that placeholder is rather reserved	13:19
	anyway, work time	13:19
	→ ofri joined	13:19
	→ rkaway joined	13:20
Fritti	btw, I'm interested -- what's the deal with the 'commit notes' thing and why is it different from a commit log?	13:23
	(reading the list archives now)	13:23
loops	Fritti, commit notes can be added at anytime after the commit.	13:23
gitte	aeruder: completely agree.	13:24
Fritti	loops: aah, cool. that makes sense	13:24
loops	gitte, the only complaint i heard against .gitignore as a placeholder is that it's not automatic.. so untarring something and doing a git add . as in the tutorials would no longer work	13:28
chris2	what are commit notes used for in practise?	13:28
gitte	loops: with the patch I sent: "git add --add-empty-dirs ."	13:29
	Heck, it can even be made default. Not like it is hurting anybody.	13:29
	chris2: nothing as of now: they are not even in "pu".	13:29
chris2	i see	13:30
Fritti	if it's default, won't 'git commit -a' not add empty directories?	13:30
gitte	chris2: but the idea is to add "Acked-by"s that did not make it into the commit.	13:30
chris2	ah, nice	13:30
Fritti	double negative alert -- I meant, does 'git commit -a', with your patch, add empty .gitignores in empty directories?	13:30
gitte	Fritti: Dunno. But as I said, I am really annoyed by the subject.	13:30
Tali	Fritti: commit -a never adds anything	13:30
Fritti	k	13:31
loops	gitte, i have to admit i like Git the way it is, and all the special cases that have been documented so far in that thread seem excessive	13:32
	(i mean.. all the special cases that have to be handled when adding empty dirs)	13:32
Fritti	naive idea from a newbie: why not just special case the empty dir? and forget about special cases the moment something is added in that directory?	13:34
	i.e. 'git add dir; git commit; git add dir/file; git commit; git rm dir/file; git commit' -> no more dir or file	13:34
gitte	Fritti: and if someone does "git rm dir/file" like you suggested above, the directory vanishes, too?	13:36
Fritti	exactly	13:36
gitte	That is almost certainly not what a certain bonehead wants.	13:37
Fritti	as git does now	13:37
	no but it does give you the ability to say 'look! git supports empty directories! now go away!'	13:37
	:)	13:37
loops	gitte, your --add-empty-dirs should really be enough.	13:43
	→ engla joined	13:44
engla	hi. A question: What is the pu branch generally?	13:44
loops	engla, it's highly experimental features that may or may not make it into Git proper at some point	13:45
engla	ok. I thought i had seen that branch in many git projects	13:45
	but it's not special in any way?..	13:45
loops	no, not special.	13:46
engla	and: I want a clone of an svn repository in git. git-svn works fine. But can one set up this to be updated automatically at for example repo.or.cz	13:46
	← janm left	13:46
engla	ok, thanks	13:46
loops	engla, you could make a nightly job to run git-svn and push the results to repo.or.cz	13:47
	ie.. a crontab entry to: git svn update ; git push repo.or.cz	13:48
engla	true. I don't have a server-like computer, so I can't do that at this time. (going electricity-less to the cabin next month :)	13:51
	→ pedro_mg joined	13:51
loops	heh.. sounds nice.. i don't know then ;o/	13:52
	← devogon left	13:54
	← Pistahh left	14:10
	→ pedro-mg joined	14:11
	← pedro-mg left	14:13
	→ branstrom_ joined	14:15
	→ Pistahh joined	14:16
	→ Yuuhi joined	14:16
	→ jasam joined	14:24
	← pedro_mg left	14:24
	← jlf left	14:27
	→ WorkingGeier joined	14:28
	← mosno left	14:31
	← orospakr left	14:32
	← branstrom left	14:33
	→ devogon joined	14:40
	← pcapriotti left	14:42
	→ QuercusMax joined	14:47
	→ branstrom joined	14:47
jasam	gitte, is there a preference in C code, when comparing the value returned by open(), to write fd<0 vs fd==-1	14:51
	?	14:51
gitte	I'd always write "< 0", because it is shorter. And also resistant to buggy implementations, even if I never encountered one.	14:52
jasam	ok	14:53
	gitte, I did not work very much in the launch_editor() function	14:53
gitte	That can be done as an additional commit, no?	14:54
	For builtin-tag.c, there is not much use in refactoring that part.	14:54
	It only makes sense when unifying with builtin-commit.c.	14:54
jasam	ok, then	14:54
	I'm checking the changes I did to see if there is an error	14:55
	I applied my patch, added builtin-tag.c and then I replaced it with the new one to see them clearly	14:56
	gitte, an important question	14:57
gitte	Yes?	14:58
jasam	In the recent patch from Kristian, Nico said to move the shell script to contrib/examples	14:59
gitte	Yes.	14:59
jasam	instead of deleting it	14:59
gitte	I do not completely agree, but it is easy enough, no?	14:59
jasam	I have had to restore it, since I started from my first patch	15:00
gitte	No problem: "git checkout junio/next git-commit.sh && git mv git-commit.sh contrib/examples/"	15:00
jasam	I think that way is easy to compare the files to see the commands	15:00
	→ damjan\|work_ joined	15:01
jasam	I did git cat-file blob HEAD:git-tag.sh >git-tag.sh	15:01
	→ rtmfd_icbm joined	15:01
damjan\|work_	what can git use for publishing repositories? can it use webdav directly??	15:01
gitte	jasam: ah, of course: git-tag.sh, not git-commit.sh. With your method, you have to "git add" it somewhere.	15:02
jasam	gitte, I realized that my command changed the mode for git-tag.sh	15:02
gitte	damjan\|work_: Kind of. There is a howto in git's Documentation/howto/.	15:03
jasam	gitte, before apply my patch (not committing it), git-tag.sh existed, so restoring it leave it without changes	15:03
gitte	Yes.	15:04
	← branstrom_ left	15:05
	→ branstrom_ joined	15:06
jasam	gitte, now I have a lonely line "diff --git a/git-tag.sh b/git-tag.sh"	15:09
damjan\|work_	gitte: sweat	15:09
	← jcollie left	15:09
	← kanru left	15:11
gitte	jasam: yes, that is expected.	15:12
	You have to update the index. That is implicitely done by "git status".	15:13
jasam	haha, interesting side effect..	15:13
	→ KirinDave- joined	15:14
jasam	gitte, anyway, on the first question, Junio said: "The editor might have been launched but exited with non-zero status." about if (run_command(&child)) die("could not launch editor %s.", editor);	15:14
gitte	Yes. Just say "There was a problem with the editor %s" ?	15:15
jasam	but, what to do? he does not say that	15:15
	I wrote this:	15:15
	if (run_command(&child) <= -10000) ...	15:16
	and then the message is now true	15:16
	but perhaps that's not what Junio was saying	15:16
	← branstrom left	15:16
gitte	Didn't you say "... if (run_command(&child)) die( ... "?	15:17
jasam	the code first was if (run_command()) die(...)	15:17
	Junio talked about the "message"	15:17
gitte	Yes, I like that code better.	15:17
jasam	but my question is	15:18
gitte	But then, run_command() can return -1, too, and your message is printed, right?	15:18
	→ branstrom joined	15:18
jasam	in my version not	15:18
	→ orospakr joined	15:18
gitte	But why check for -10000 explicitely? Why not just have all errors handled in one go?	15:19
jasam	Junio complained also about other <= -10000 some lines below	15:19
	but he did not said it was about the -10000 constant (but perhaps he thought it...)	15:20
gitte	He probably read it as I did. I missed it the first time, too...	15:20
jasam	The other case was: if (ret <= -10000)die("unable to run %s\n", argv_verify_tag[0]); return -ret;	15:22
	he said: I wonder why you need to differentiate between ERR_RUN_COMMAND_* and non-zero exit status... Also do you need to "return -ret", instead of not negating?	15:22
gitte	For the sake of simplicity, I'd change that too.	15:22
jasam	ok	15:22
	I will remove the -10000 thing	15:23
gitte	if (ret) die ("unable to run %s", argv_verify_tag[0]); return ret;	15:23
	No magic constant, no newline, and no negating ret.	15:23
jasam	return 0;	15:23
	ok	15:23
mnemoc	hi, i was doing svnimport but after 882 i got disconnected from the svn server, now when i try to `git-svnimport -s 883 ...`. i was doing svnimport -i but there are some files beside .git/, how do i "Make sure your working directory corresponds to HEAD" ? (thanks)	15:24
	→ stouset joined	15:25
	← johan-s left	15:35
	← etnt left	15:36
	← branstrom_ left	15:36
	← evilchelu left	15:37
jasam	gitte, I have another problem, just the same	15:42
	→ drizzd_ joined	15:42
jasam	if (ret) die ("unable to run %s", argv_verify_tag[0]); return ret; <--- git verify-tag returns error status when the tag cannot verified (but the program worked)	15:43
gitte	Yes, but that is expected, no? If the tag cannot be verified, exit status must not be zero.	15:44
jasam	that is the reason because I needed to differenciate between ERR_RUN_COMMAND_* and zero status	15:44
	when git verify-tag fails, I supose it shows messages that the key is not valid	15:45
	so saying "unable to run git-verify-tag" is not correct	15:46
	"tag could not be verified"?	15:46
gitte	Ah, yes, you were talking about the message. Yes, that sounds saner.	15:47
	→ branstrom_ joined	15:48
jasam	hum, that's also better in some ways...	15:48
	:-)	15:49
	gitte, something I didn't tell you is that I unified the common code between list_tags and verify_tags	15:49
	using cb_data	15:49
gitte	Let's see it ;-)	15:50
jasam	hum, cb_data not, that's from list_tags	15:50
	I mean delete_tags and verify_tags	15:50
	if you see the last version	15:50
	there was a repeated code	15:51
	did you see it?	15:52
	there is not much else to see	15:53
gitte	Unfortunately, I did not really have time to read it...	15:53
jasam	just this: typedef int (func_tag)(const char name, const char ref, const unsigned char sha1);	15:53
gitte	But promise, your next patch will be studied in detail again.	15:53
	→ kanru joined	15:53
jasam	both functions, verify and delete, accept now many tag names as arguments	15:54
gitte	Very good!	15:54
	Did it delete more lines than it added?	15:54
jasam	I don't know	15:54
	because I also added things like this:	15:55
	if (snprintf(ref, sizeof(ref), "refs/tags/%s", *p)	15:55
	>= sizeof(ref))	15:55
	die("tag name too long: %.s...", 50, p);	15:55
	← drizzd left	15:55
jasam	to remove sizeof ref (line got too long)	15:55
	if (fn(*p, ref, sha1)) had_error = 1;	15:56
	common code had a lot of lines, the best thing is not the saved code, but the same version on both	15:57
	before, if you change something on one of them, then you could forget to change the same on the other	15:57
	the common code is the code that finds the tags	15:57
gitte	Yes, that is very good in general.	15:57
jasam	static int do_tag_names(const char **argv, func_tag fn)	15:58
	that is	15:58
	well, I will terminate this, so see you gitte	15:59
gitte	CU	15:59
jasam	I continue here working on it!	16:00
	I said for you	16:00
	← KirinDave- left	16:03
	← nikodemus left	16:03
	← branstrom left	16:06
	→ johan-s joined	16:09
	→ fhobia joined	16:16
	← rkaway left	16:16
	→ alley_cat joined	16:17
	→ rkaway joined	16:19
	→ jcollie joined	16:30
	→ ZeeGeek joined	16:30
puzzles	oh my god, it's ZeeGeek, how unexpected!	16:30
duncanm	hey puzzles	16:31
ZeeGeek	puzzles: omg, puzzles !	16:31
puzzles	yo	16:31
ZeeGeek	hey	16:31
	I cherry-picked two commits into master, then I can't push it to my remote repo, "src and dst refspec master matches more than one"	16:32
	← nashdj left	16:33
	← branstrom_ left	16:35
	→ praka joined	16:41
	← fhobia left	16:47
	→ DrNick joined	16:50
	→ KirinDave- joined	16:55
	→ aroben joined	16:58
	→ nikodemus joined	17:05
gitte	ZeeGeek: That problem is completely unrelated to your cherry-pick.	17:08
	What was your push command line?	17:08
ZeeGeek	I've been using "git-push myrepo master"	17:08
	here's what I did, I created a new branch from master called scm, then did something to that branch, then I cherry picked two commits from scm to master	17:10
	before all these, that command was working	17:10
gitte	Can you please look what "git ls-remote myrepo \| grep master" says?	17:11
ZeeGeek	yeah, sure, one sec	17:11
	refs/heads/master and refs/remotes/origin/master	17:12
	why are there two of them?	17:12
	KirinDave- → KirinDave	17:12
	← bentob0x left	17:13
moh	refs/heads/master is your local master, the remotes one is the remote one	17:13
ZeeGeek	so it's all right?	17:14
moh	they will differ if you commit changesets to your local tree, or if someone else pushes to the "origin/master"	17:14
gitte	ZeeGeek: I guess "git push myrepo refs/heads/master" will work.	17:14
ZeeGeek	yeah, they have different sha1 values	17:14
	gitte: trying	17:14
Randal	Hmm. Other than Linus, who would make the most sense for a skype interview about git?	17:15
	Junio?	17:15
ZeeGeek	gitte: aha, that worked!!	17:15
	gitte: thank you very much :)	17:15
	→ mtcx joined	17:20
ZeeGeek	gitte: btw, why did the old command stop working?	17:20
	→ bronson joined	17:21
gitte	ZeeGeek: probably because both "refs/heads/master" and "refs/remotes/origin/master" were uptodate.	17:21
	IOW once you changed _one_ of them, it stopped working.	17:22
	Randal: definitely.	17:22
	Randal: of course, Shawn Pearce has a lot to say about working with git, too, maybe even more than Junio.	17:22
ZeeGeek	hmm, I see. so the one you gave me should be the right way to do it	17:22
gitte	Shawn seems to be stuck in a shop with many "Aunt Tillie"s.	17:22
	ZeeGeek: yes.	17:23
ZeeGeek	thanks	17:23
gitte	ZeeGeek: you are welcome.	17:23
	→ theCore joined	17:24
Randal	I'm not sure how to ask this, but does Junio have something that mainstream americans would consider a thick accent?	17:28
gitte	gitster: that is a question directly to you...	17:28
Randal	heh yes.	17:29
	→ lyakh joined	17:31
WorkingGeier	an Australian once diagnosed that I had a "thick accent"	17:36
	only he pronounced it "thaeck"	17:37
Randal	heh	17:37
gitte	Randal: seems you made some friends last night ;-)	17:42
	gitte is reading the backlog	17:42
Randal	Yeah, I get on a soapbox from time to time.	17:44
	I should just get away from IRC entirely.	17:44
gitte	No, please don't!	17:44
	I for one enjoy talking to you (and sometimes fighting).	17:44
Randal	well, I'm getting too old and crotchety for the kids, apparently. :)	17:45
	"he can't possibly know what he's talking about... it's all NEW!"	17:46
gitte	I fully understand.	17:46
	There I am, trying to be nice to some Clueless in Seattle, to explain why the reasoning is wrong.	17:46
	And what do I get? I am _accused_ of condescending.	17:46
Randal	heh. yeah.	17:46
	part of the problem is that tone is not carried well in typed text.	17:47
	so people don't necessarily that I am respecting the person. in fact I care enough about person that I am taking my time to help them.	17:47
	but I will call ineffective or dangerous behavior when I see it.	17:47
	often because I have lived the consequences of that behavior in my ever-growing history.	17:48
	→ dduncan joined	17:50
	← ZeeGeek left	17:51
gitte	Yes, that is what I feel, too.	17:51
KirinDave	Excuse me, is there a way to see commits not associated with any current refs?	17:58
cehteh	git fsck will list dangleing commits	17:58
	the reflog might also be useful	17:58
KirinDave	Does git cherry-pick cascade?	17:58
cehteh	cascade?	17:59
	it only merges one single commit	17:59
	if you mean that	17:59
aeruder	you probably want git-fsck --no-reflogs to see dangling commits	17:59
	if you really want everything that's dangling	18:00
	KirinDave: well, you can use -n	18:00
KirinDave	Oh	18:00
	Oh shit	18:00
	He was using git 1.5.0.6 on a git 1.5.2 repo	18:00
aeruder	if you want to pick a few commits into a single commit	18:00
KirinDave	I've heard this is a bad thing to do.	18:01
aeruder	uh	18:01
	i doubt it matters	18:01
KirinDave	Well the commands you're listing don't exist, for one. :)	18:01
aeruder	the reflog might be new, git-fsck may be all you need	18:01
	i don't know when reflog was added	18:01
LotR	KirinDave: how is code-casting going?	18:02
KirinDave	LotR: Halted while I see if instead of a new project I should chip in with Lighthouse.	18:03
	LotR googles	18:03
	← robewald\|work left	18:04
LotR	lighthouseapp.com?	18:05
	→ robewald\|work joined	18:05
	← kanru left	18:07
mnemoc	there is any well integrated issue tracker for git?	18:08
cehteh	not that i know .. we use a tiddlywiki in git .. works well so far	18:17
	maybe someday extending that to so other system	18:17
	→ branstrom joined	18:23
	← KirinDave left	18:26
	← cedric__ left	18:30
	→ robfitz_ joined	18:30
Randal	hmm. Gitster hasn't answered yet. I'll drop him email.	18:31
	→ geppetto joined	18:31
Randal	email sent. :)	18:34
geppetto	So "git apply" says there is a --no-add option ... implying to me that by default "git apply" followed by "git commit" will do everything, but it doesn't seem to add anything by default ... is this expected?	18:35
Randal	the docs say something different for --no-add than what you just implied	18:36
	it's not "add" in the "git-add" sense	18:37
	you probably still need to add the changed files	18:37
gitte	... or use the --index option	18:37
Randal	Ahh, yes	18:37
	← robfitz left	18:39
geppetto	Randal: Ahh apply --index was what I was looking for, thanks	18:42
	gitte goes off for some dinner.	18:43
	← gitte left	18:43
	← moh left	18:45
	← chris2 left	18:45
	→ moh joined	18:49
corecode	what does less have to do with screen?	18:53
	→ Eludias joined	18:54
	→ KirinDave joined	18:55
LotR	KirinDave: not the lighthouse on lighthouseapp.com, right?	18:56
KirinDave	LotR: Yes that one	18:58
LotR	KirinDave: so is it free software after all?	18:59
	LotR got the impression it wasn't	18:59
KirinDave	LotR: Nope. Not free.	19:00
LotR	i see	19:00
KirinDave	LotR: Although anything I do would probably just talk to an API like lighthouses'	19:00
	→ robfitz joined	19:02
	← nikodemus left	19:11
	← robfitz_ left	19:12
	→ goloo joined	19:18
	→ nud_ joined	19:38
	→ Prax01D joined	19:40
	← Prax01D left	19:41
	→ Myrizio joined	19:45
	← cmarcelo left	19:46
	→ rlb3_work_ joined	19:46
	← robewald\|work left	19:49
	→ robewald\|work joined	19:49
	→ arrai joined	20:05
	← ofri left	20:11
	→ chris___ joined	20:12
	← mtcx left	20:13
chris___	hello, I'm looking for a little assistance: I type `git-rebase -i` and get... nothing. It did work yesterday, and I don't know what changed.	20:14
	how can I debug this?	20:14
	`git-rebase -i -v ` also gives nothing	20:15
	http://pastebin.ca/626672 is the output of `strace git-rebase -i`	20:19
vmiklos	sh -x git-rebease--interactive? (just an idea)	20:21
chris___	http://pastebin.ca/626679 is the output of sh -x git-rebase--interactive	20:24
	that "case $# in ... break" is suspicious, no?	20:25
	→ ofri joined	20:25
	← lyakh left	20:26
vmiklos	i had some hanging problem with git-rebase -i but it's fixed in git.git a long time ago	20:26
chris___	doh. EUSER_IS_IDIOT.	20:27
	I forgot that there is one mandatory argument.	20:27
vmiklos	:)	20:28
	→ markwa joined	20:28
	→ nikodemus joined	20:34
	→ madewokherd joined	20:34
	→ fultilt joined	20:39
	→ branstrom_ joined	20:42
	← mncharity left	20:46
	← QuercusMax left	20:48
	→ pedro_mg joined	20:49
	← branstrom left	20:58
	← pedro_mg left	21:00
	→ mwc_ joined	21:11
	← alley_cat left	21:13
	← cort left	21:16
	← lcapitulino left	21:19
	→ evilchelu joined	21:19
	→ hein joined	21:23
	← Dodji left	21:29
	← nikodemus left	21:31
	→ lyakh joined	21:33
	→ cannonball joined	21:36
	← Sho_ left	21:38
	hein → Sho_	21:38
Pieter	heh	21:44
	I had some git repo's once	21:44
	but, not anymore	21:44
	or in any case, not where I can find them	21:44
	:(	21:44
aeruder	find / -type d -name refs ? ;)	21:45
Pieter	I tread something like that	21:45
	pieter@bwide:~$ find . \| grep HEAD	21:45
	./git-repo/bonnenteller.git/ORIG_HEAD	21:45
	./git-repo/bonnenteller.git/logs/HEAD	21:45
	./git-repo/bonnenteller.git/HEAD	21:45
	that was all	21:45
	← jcollie left	21:45
Pieter	which is weird :)	21:45
	→ nud__ joined	21:46
	← nud_ left	21:46
	→ QuercusMax joined	21:47
	← davi left	21:50
	← stouset left	21:53
	← Eludias left	21:54
	→ nashdj joined	21:57
	→ alli joined	21:57
	→ davi joined	21:58
	→ aroben_ joined	22:01
	← WorkingGeier left	22:02
	← ofri left	22:02
	← aroben_ left	22:05
	→ aroben_ joined	22:06
	← arrai left	22:07
	→ nikodemus joined	22:07
	→ Prax01D joined	22:08
	← Prax01D left	22:09
	← alli_ left	22:09
	← nikodemus left	22:09
	← aroben left	22:18
	← orospakr left	22:20
	← davi left	22:24
	→ Lash\| joined	22:26
	← mwc_ left	22:27
	→ mwc_ joined	22:28
	aroben_ → arobebn	22:30
	arobebn → aroben	22:30
	→ lcapitulino joined	22:37
	→ GyrosGeier joined	22:53
	← goloo left	22:53
	← fultilt left	22:53
	← mwc_ left	22:54
	→ fultilt joined	22:58
	← cannonball left	23:01
	← rtmfd_icbm left	23:04
	→ mwc_ joined	23:08
	← fultilt left	23:08
	← lyakh left	23:08
	← mwc_ left	23:11
	→ rtmfd_icbm joined	23:12
	→ mwc_ joined	23:12
duncanm	tig is pretty cute	23:12
	it's a nice tool	23:12
	→ fultilt joined	23:13
	← Lash\| left	23:14
	← theCore left	23:17
	← fultilt left	23:21
	→ fultilt joined	23:26
	→ mwc__ joined	23:28
	← mwc_ left	23:28
	mwc__ → mwc_	23:29
	← Yuuhi left	23:33
	→ orospakr joined	23:34
	← QuercusMax left	23:36
	→ QuercusMax joined	23:37
	← QuercusMax left	23:37
	→ QuercusMax joined	23:38
	→ Lash\| joined	23:42
	→ groome joined	23:44
	→ cort joined	23:45
	← groome left	23:45
jasam	how can I change the encoding of the mails sent with Thunderbird? I'm afraid that I sent a windows-1252 patch with european special characters again	23:48
	I don't know if it can be correctly applied to be utf-8 as the original patch was	23:50
	← orospakr left	23:54
	→ orospakr joined	23:56
	← markwa left	23:57
	← evilchelu left	23:57

Logs Search ←Prev date Next date→

Channels Documentation