IRCloggy #git 2008-01-22

Provider of IRC logs since 2005.

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

	2008-01-22
wagle	chmod 664 /home/git/bsrflash.git/logs/refs/heads/master (?)	00:03
	← vbgunz left	00:03
	← meyering left	00:04
	← cannonball left	00:04
wagle	Spyro,	00:04
	← p4tux left	00:04
	← octopod left	00:05
	← Dodji left	00:06
	← chris2 left	00:07
	kristoffer → Ken\|JLIME	00:12
	← brosner left	00:18
	← GyrosGeier left	00:22
	→ meyering joined	00:24
	← toxx_ left	00:27
	← Cerebon left	00:28
	← robinr left	00:28
	← HG` left	00:30
	clee_ → clee	00:31
	→ robinr joined	00:33
	dwmw2_zzz → dwmw2_gone	00:33
	← robinr left	00:40
	→ jerbear joined	00:40
kelvie_	hrm does git-svn dcommit detect moves/renames?	00:42
	when committing to git-svn who doesn't know a delete+create is a move	00:42
	→ robinr joined	00:45
mugwump	kelvie_: should work on recent git, maybe try --find-copies-harder	00:46
	svn doesn't actually support renames	00:46
	just copies	00:46
pasky	thanks, fixed	00:50
	and that's it, I'm done with ns.inecnet.cz for goot	00:50
	*good	00:50
kelvie_	mugwump: svn mv?	00:53
	so then a copy + delete would be a rename?	00:54
mugwump	right	00:54
	So, there's copy, but no move, merge, ignore, etc :)	00:54
	← Ademan left	00:55
	→ Ademan joined	00:55
	← jerbear left	00:56
	→ yeowman joined	00:57
	← Ken\|JLIME left	00:58
	→ fhobia joined	01:11
	← timocratic left	01:13
	→ obvio171 joined	01:15
	→ kevwil joined	01:22
	← obvio_ left	01:22
	← devogon left	01:27
	dwmw2_gone → dwmw2_ULN	01:27
	← robinr left	01:28
	→ gitte joined	01:29
	→ pflanze joined	01:33
pflanze	Hello. I'm missing a tag for git 1.5.3.8 in git://git.kernel.org/pub/scm/git/git.git	01:34
	Also, it might make sense to publish the sha1 sums of the releases and/or provide signatures for the commmits.	01:35
	→ dAndy joined	01:35
pflanze	Then I could just update through git and still have all the safety I long for.	01:35
dAndy	Is there something that needs to happen to get post-checkout hooks to work?	01:35
	← tchan left	01:35
dAndy	i have an executable post-checkout file in .git/hooks, which does not get run	01:35
	→ tchan joined	01:36
dAndy	I see post-checkout here: http://www.kernel.org/pub/software/scm/git/docs/hooks.html	01:36
	nm, not in 1.5.3... http://colabti.de/irclogger/irclogger_log/git?date=2007-10-19,Fri#l1129	01:38
	hooray for logs	01:38
	→ bdiego joined	01:38
gitte	Is Kevin around here?	01:38
	← G_SabinoMullane left	01:38
Eridius	you mean me?	01:38
	Eridius is actually eating dinner, so if you want to say something, don't expect a response for about 25 minutes	01:39
gitte	If your surname starts with "B", I am likely to mean you.	01:40
	→ pabs3 joined	01:41
rumun	/dev/slmkep[kg=-fvnnb	01:41
	← rumun left	01:41
gitte	Oh, so it _is_ you!	01:42
	Could you do me a favour?	01:42
	→ vintrepid_ joined	01:43
gitte	So you don't answer me?	01:44
	Could you do me the favour of not answering to that _ugly_ and _stupid_ thread anymore?	01:44
	You are not helping _anyone_.	01:44
	Including yourself.	01:44
	pflanze realizes that tags are not automatically fetched	01:44
gitte	pflanze: "git fetch --tags"	01:45
pflanze	ok. That explains why I was missing the tags. thanks.	01:45
gitte	No problem.	01:45
	The thing is: usually tags _are_ fetched, if you fetched the commits they are referencing.	01:45
Mikachu	gitte:	01:45
	03:54 <gitster> Anyway, 1.5.4-rc4 is out. I'd appreciate whoever is the channel operator update the /title here.	01:45
	← gitte left	01:46
Mikachu	doh	01:46
	→ gitte joined	01:46
	ChanServ set mode: +o	01:47
gitte	Mikachu: you should have op status...	01:47
	Mikachu: why not keep it? I have a feeling you are a good operator.	01:47
	→ G_SabinoMullane joined	01:48
	← vintrepid left	01:48
gitte	pflanze: ther are a few possible reasons why you would not get tags, most of them boiling down to you fetching commits before they were tagged.	01:48
Mikachu	feel free to have me added to the chanserv access list, but i don't like being actively equipped with a @	01:49
	← yeowman left	01:49
gitte	pflanze: possible?	01:49
	Mikachu: I know the feeling.	01:49
	← Bira left	01:49
Mikachu	partly because i tend to do things by accident sometimes, and partly because people tend to think you are somehow authorative :)	01:49
gitte	Mikachu: but I do not know how to add you to the chanserv access list...	01:49
pflanze	gitte: yes man git-fetch explains it good enough for me. One should just RTFM..	01:49
Mikachu	so do we want both .8 and rc4 in the topic?	01:49
	gitte: i think only pasky can do that	01:49
gitte	Mikachu: rather only -rc4.	01:49
	Mikachu changed the topic to: 1.5.4-rc4 \| Homepage http://git.or.cz/ \| Everyone asleep or clueless? Try [email@hidden.address] \| Channel log http://colabti.de/irclogger/irclogger_log/git \| Mailing list archives: http://marc.info/?l=git \| Gits on git: http://tinyurl.com/2xq3ke \| You want $ID?: http://tinyurl.com/yqpgv9 \| Need to paste something?: http://git.pastebin.com	01:50
gitte	I think .8 was only a way to explain that 1.5.4 is not yet out.	01:50
	Mikachu: Thanks!	01:50
	← pabs3 left	01:50
	← kzak_ left	01:51
Mikachu	yeah, only pasky is level 30, and that's needed to modify the channel access list	01:51
semi_	that's a whole lot of levels :p	01:51
Mikachu	everyone else is level 29 heh	01:52
Eridius	gitte: I've made many offers to end the argument and work towards a solution, and nobody has responded to that. Don't blame me for the thread	01:53
	Eridius goes back to what he was doing	01:53
mugwump	but you will get blamed for colours in the channel! Mikachu, prepare the kb!	01:53
	Mikachu set mode: +c	01:54
tpope	technically inverse isn't a color	01:54
Mikachu	test	01:54
	did that come out inverted?	01:54
mugwump	not to me	01:54
loswillios	no	01:54
Mikachu	does anyone in here ever use colors legitimately?	01:55
tpope	hmm different networks handle it differently	01:55
gitte	Eridius: you have _not_ offered _anything.	01:55
	Eridius: just dumb arguments.	01:55
	Eridius: taking time from people who actually _do_ something, in addition to hanging out on mailing lists.	01:55
Eridius	gitte: apparently you haven't actually bothered to read my emails. now stop bothering me	01:55
tpope	it is my humble opinion that +c should be brutally enforced upon every channel of every server	01:55
gitte	Mikachu: I see.	01:55
	Eridius: could you stop bothering writing mails to the list, then?	01:56
Mikachu	tpope: on some networks they just silently drop the message instead of stripping the codes	01:56
gitte	Eridius:	01:56
	Eridius: It would be _much_ appreciated.	01:56
Eridius	gitte: can you stop being a prick? you haven't contributed a single thing of value to the thread, and you have the gall to tell ME to shut up. take your own advice	01:56
loswillios	Mikachu: heh, that's cool	01:56
gitte	Eridius: you have _not_ bothered to work _in the least_ towards a solution.	01:57
	Eridius: you just made a fool out of yourself.	01:57
	Eridius: and a PITA.	01:57
Eridius	gitte: once again, read my emails before you bother to respond. now shut up	01:57
vmiklos	omg	01:57
gitte	Eridius: and if _you_ read the thread, you'd know that I don't need to prove anything to you.	01:57
tpope	Mikachu: typically those drop it, but send a message back stating such	01:57
	just like if you spoke in a +m channel	01:57
vmiklos	now this continues here even on IRC :-)	01:57
Mikachu	tpope: hm that's true	01:58
gitte	vmiklos: I try to end it here on IRC.	01:58
Eridius	dear god you are annoying. I've read every single message in their entirety, which is more than can be said for you (or, apparently, others). Now I'm going to have to /ignore you since you can't seem to have the decency to just shut up as asked	01:58
gitte	Because this has been going on too long.	01:58
vmiklos	thank you :) (both)	01:58
tpope	I don't feel like running multiple irc clients just to test	01:58
gitte	Costing everybody too much time.	01:58
	Eridius: you read the mails? I had to _delete_ them, to get through to truly interesting stuff.	01:58
	Now, I do not have the power to kick you from the mailing list.	01:59
	There's got to be an end.	01:59
	→ garrin joined	01:59
gitte	And it will _not_ be me, giving in, and writing a patch that leaves an idiot happy.	01:59
	An idiot that succeeded in _pissing off_ thousands of people.	02:00
Mikachu	i think he did ignore you	02:01
	which is sort of amusing	02:01
	← tjafk2 left	02:01
	→ tjafk2 joined	02:01
	ChanServ set mode: +o	02:01
vmiklos	well, that's his problem not ours:)	02:02
gitte	I _did_ remember the name Eridius...	02:03
	He seemed not to rant here, but give occasional (if rare) good advice.	02:03
	Oh, so he knows about #git only since Jan 16th, this year ;-)	02:04
	Lol.	02:04
	So I am angry.	02:06
	So I wrote evil mails.	02:06
	So I lose karma.	02:07
	Hopefully others will benefit from my sacrifice ;-)	02:07
Mikachu	are you johannes?	02:07
gitte	Mikachu: no, I am gitte.... ;-)	02:08
	→ tjafk3 joined	02:08
	→ tjafk1 joined	02:09
gitster	bri-gitte?	02:15
	→ Bluefury joined	02:16
gitte	gitster: of yourse!	02:17
	gitster: the question is now, what does the "bri" stand for... ;-)	02:18
	→ imyousuf joined	02:18
gitster	that's half briton.	02:20
	← cmarcelo left	02:21
gitte	gitster: no need to insult me ;-)	02:22
	FWIW I'm probably moving back from this "island" to where I came from...	02:23
	gitster apologizes while laughing hysterically...	02:23
	gitte is lauging, too!	02:23
gitster	go back to better food?	02:23
	Mikachu doesn't get it	02:23
vmiklos	:-)	02:23
gitte	Hope so.	02:23
	Better wine, to begin with.	02:23
	gitte is stuck with red, since the white here either sucks, or is unaffordable.	02:24
gitte	gitster: good to have fun with you again!	02:24
	gitster: how are you holding up?	02:24
gitster	I am mired in day-job.	02:25
	don't remind me X-<.	02:25
gitte	(This is really an especially long rc-cycle)	02:25
	Uh, oh.	02:25
	gitte is doing his famous imitation of rainman.	02:25
	Mikachu hands gitte the car keys	02:25
gitte	Mikachu: what kind of car?	02:25
	← tjafk2 left	02:25
gitte	Now, don't watch the whole movie before answering!	02:26
Mikachu	i have no idea, i have some vague memory that maybe it is gray? :)	02:26
aeruder	wow, that was an almost sorta absolute statement maybe	02:26
	:)	02:26
	→ tjafk2 joined	02:27
gitte	I am an excellent driver.	02:27
Mikachu	just remember to drive on the left side	02:29
gitte	For those who did not know... "a 1949 Buick Roadmaster."	02:29
gitster	An overly long rc-cycle is unhealthy for everybody. I think after deciding to punt on http-push without multi for safety's sake, however, we can do the final any day now, I think.	02:29
gitte	Mikachu: does not matter that much... British drivers are insane.	02:29
	gitster: hey, I'm all for it.	02:29
	gitster: I will work on http-push nevertheless... if only for 1.6.0 ;-)	02:30
	gitte grins	02:30
	gitte grins like a Cheshire cat	02:30
Mikachu	gitte: but what colour?	02:31
	← lamont left	02:31
gitster	Good to have a committed volunteer. If we end up deciding on punting for 1.5.4, it would allow us to update the release notes from saying "will hopefully be fixed" to something a bit stronger.	02:31
gitte	Mikachu: orange... flaming orange!	02:31
	← tjafk3 left	02:31
gitte	gitster: you know me, so maybe you want to keep "hopefully" ;-)	02:32
mugwump	gitte: like this? http://www.internationalhero.co.uk/a/aku.jpg :D	02:32
gitte	lemmesee.	02:33
	→ Cerebon joined	02:33
mugwump	Well, it has the flaming orange part right	02:33
gitte	But it's not a cat!	02:33
mugwump	not really grinning in that shot ... although that character does have some devillish grins	02:33
gitte	I has can FLAMEZ!	02:34
mugwump	it's a shapeshifting master of darkness! it can take on any form	02:34
	even kitteh	02:34
	gitte really rolls on the floor now, laughing.	02:34
Mikachu	wireless keyboard?	02:34
gitte	Well, I stumbled back to it.	02:35
	Or crawled.	02:35
	Back to the topic:	02:35
	gitster: IMHO it would be not too bad leaving the http-push issue in peace for 1.5.4, given that not many people use this deprecated transport (and for a good reason).	02:36
	Much as I hate it, the crlf issue is not big enough to hold off a release.	02:36
	And the other issues, well, most cropped up pretty late in the rc cycle, so they should not be surprised to be held off for 1.5.5 or 1.6.0...	02:37
gitster	As I said I do not think we have anything lacking. Even Linus's in-core index changes will not in 1.5.4 even though it saw some review from real people.	02:38
	s/real/real core/	02:38
	I think it is not healthy for everybody to prolong the freeze too long.	02:39
gitte	Concur.	02:39
	AFAIAC holding off Linus' patches does not concern me much ;-)	02:39
Mikachu	i think the majority of people who like hacking on new stuff don't start fixing bugs during a freeze, they just go away	02:39
gitte	I'm running master + patches ;-)	02:39
	Mikachu: well, I listened, and I tried to find bugs, and tried to fix them.	02:40
	pull --rebase, for example.	02:40
	(Only the test case is missing).	02:40
	cvsexportcommit, for example...	02:40
	(But it was a bit too complicated, and I wanted to wait for 1.5.4 to start it)	02:40
	There was another issue I forgot, but it was not that important. Just a minor user interface issue.	02:41
gitster	You would exercise existing code while hacking on new stuff, and that's how some bugs are found. We should not discourage random private hacking too much, either.	02:42
gitte	Right.	02:42
	But on the other hand, you were correct to ask people to hold off their private issues in order to work on 1.5.4.	02:42
gitster	I've been wondering if we should shorten each cycle, concentrating on smaller number of tasks.	02:42
gitte	gitster: you have been particularly good at discerning what makes a new release, and what does not.	02:44
gitster	Oh, my not taking patches early is pretty much orthogonal. If I had an infinite bandwidth, leaving 'next' open during the whole rc period would be feasible, and it would make (1) people to play with existing code by building on top, and (2) keeping the 'master' stable because everybody understands their invention will not graduate during the period.	02:44
gitte	gitster: don't let others make you unsecure...	02:44
gitster	It's just it is not practical for me. And I suspect it is probably not practical for contributors.	02:45
	← dreivier1 left	02:45
gitte	For me, it was a healthy exercise.	02:45
	In spite of what I tried in the early 1.5.0 cycle, I really rebased to "master" this time.	02:45
	It's pretty stable.	02:46
	For some time, I chased that "commit <path>" bug, though.	02:46
	I was so relieved that Linus found the solution, because I did not seem to be able to.	02:46
aeruder	everytime i think i've seen the end of that stupid thread, there's another reply	02:47
gitte	ERIDIUS! STOP IT!	02:48
Mikachu	Eridius: stop it!	02:48
Eridius	GODDAMN IT YOU PEOPLE ARE FUCKING MORONS	02:48
gitster	Well, it was educational, I guess, if you choose to ignore cruft ;-).	02:48
Eridius	READ MY FUCKING EMAILS	02:48
	← tjafk1 left	02:48
aeruder	Eridius: WE ARE, YOU ARE A FUCKKING MORON	02:48
gitte	Eridius: stop your _FUCKING EMAILS_	02:49
aeruder	cough	02:49
Eridius	not once have you actually argued against a point I made, and not once have you actually responded to my COUNTLESS offers to end the thread and work towards a solution	02:49
	no, you'd rather bitch and moan and make straw man arguments	02:49
	so shut the fuck up	02:49
gitte	Eridius: you should _know_ by know.	02:49
aeruder	Eridius: yet you always have to have the last word	02:49
	just stop replying	02:49
Eridius	that you're an idiot? I do know	02:49
aeruder	and the thread will end	02:49
Eridius	aeruder: I tried, and I got 15 emails	02:49
	so no, it won't	02:49
aeruder	Eridius: ignore them, stop replying	02:49
gitte	Eridius: if one person tells you that you're an idiot, you are free to ignore it.	02:49
	But if more than _two_ say that.	02:49
	_Especially_ on a _technical_ list like that.	02:50
aeruder	or if you can't find a single person who says that you aren't an idiot	02:50
gitte	You.	02:50
	Better.	02:50
Eridius	if every person that calls me an idiot does so based on an argument that I DIDN'T FUCKING MAKE, then you all are retarded	02:50
gitte	Understand.	02:50
	gitte kicked Eridius (User terminated!)	02:50
	→ Eridius joined	02:50
mugwump	Eridius, calm down a bit. Yes, everyone in the world is irate and tends to flame	02:50
gitte	mugwump: I kicked him.	02:50
	It was enough.	02:50
Eridius	hi gitte	02:50
aeruder	he autojoined	02:51
	gitte kicked Eridius (User terminated!)	02:51
	→ Eridius joined	02:51
Mikachu	autorejoin is a bannable offense in my book	02:51
	gitte kicked Eridius (User terminated!)	02:51
	→ Eridius joined	02:51
Eridius	blame my client	02:51
	gitte kicked Eridius (User terminated!)	02:51
	→ Eridius joined	02:51
	gitte kicked Eridius (User terminated!)	02:51
	→ Eridius joined	02:51
	gitte kicked Eridius (User terminated!)	02:51
	→ Eridius joined	02:51
aeruder	while this looks like fun	02:51
	← Eridius left	02:51
	Mikachu set mode: +b	02:51
Mikachu	ah	02:51
mugwump	gitte, there is also /kb	02:51
	Mikachu set mode: -b	02:51
gitte	Thanks, Mikachu.	02:51
gitster	please.	02:51
	gitte, thanks for the link to google.nl talk video the other day ;-)	02:51
aeruder	i have a feeling that someone's going to have to block him from the mailing list as well	02:51
	since i'm guessing the next email will be about the irc channel...	02:52
gitster	I'd prefer you did not do that.	02:52
aeruder	yea, i was mostly joking :)	02:53
	→ Eridius joined	02:53
gitster	what I meant was not you but whoever did that /kick.	02:53
Eridius	I'm really confused here. Do you guys actually want people to contribute, or do you just get a kick out of alienating potential new contributors?	02:54
	← loswillios left	02:54
aeruder	Eridius: you can't possibly be even on the road to contributing with your constant email writing	02:54
	i don't know how you'd have the time	02:54
Eridius	seriously, every single fucking time I offered to end the thread and work towards a solution. NOBODY accepted my offer. obviously you don't care about solving the problem, you'd much rather yell at me	02:54
Mikachu	you can't "offer to end the thread", just send a patch	02:55
gitster	"http://video.google.nl/videoplay?docid=-4216011961522818645"	02:55
Eridius	aeruder: 95% of those emails were written from my laptop on the go. How many times must I explain that not all time is equal, that I can compose emails in situations where learning/writing code would not be feasible?	02:55
aeruder	Eridius: how about now? please?	02:55
	← kukks left	02:55
	→ offby1 joined	02:55
gitte	Eridius: if you want to have the last word, you don't fucking want to end the thread.	02:55
	Now go and think for a while before typing again.	02:56
gitster	Eridius: please understand that other people are monitoring e-mail not as time-killing-activity like you do.	02:56
Eridius	wtf? What are you calling a last word? me saying "lets stop arguing and work towards a solution?"	02:56
aeruder	Eridius: no, its the fact that your <let's stop arguing> is	02:56
Eridius	gitster: I assume that anybody monitoring the list is capable of deleting threads they don't want to read	02:56
gitster	They are looking for real issues, not interested in whose definition of "normalize" is correct.	02:56
aeruder	you're all wrong! I'm right! Conversation ends now!!!	02:56
Eridius	gitte: please grow up and stop insulting me in private messages	02:57
mugwump	Eridius: are you talking about using unicode normal forms like decomposing diacritics etc?	02:57
	mugwump wonders if people are simply talking past each other	02:57
gitster	being able to delete is not the issue. your forcing others (many others) to delete is.	02:57
Mikachu	mugwump: summary, osx renames files without telling apps, Eridius wants git to magically know when files have been renamed	02:57
gitster	People, just grow up and learn to ignore topic that you are not interested in.	02:58
gitte	Eridius: how _difficult_ is it to understand that instead of _everybody else_, _you_ are wrong?	02:58
mugwump	I'm just wondering if people are confusing case folding with unicode normalisation	02:58
	Gitzilla wonders if there's any popcorn.	02:58
Eridius	gitte: how difficult is it to understand that, well, you don't understand shit? You consistently misrepresent my arguments and ignore my pleas to work towards a solution instead of repeating the tired old arguments. Now shut the fuck up	02:58
mugwump	Linus seems to always be giving examples involving case varieties	02:59
gitte	Eridius: if you _think_ that you are achieving _anything_ by making people angry, you are a sad figure, really.	02:59
	gitster laughs hard.	02:59
Eridius	ok, once again I have to /ignore gitte	02:59
	gitte kicked Eridius (User terminated!)	02:59
	→ Eridius joined	02:59
	gitte kicked Eridius (User terminated!)	02:59
	← offby1 left	02:59
	→ Eridius joined	02:59
	gitte kicked Eridius (User terminated!)	02:59
	→ Eridius joined	02:59
	Mikachu set mode: +b	02:59
	gitte kicked Eridius (User terminated!)	02:59
	→ loswillios joined	02:59
	aeruder teaches gitte the /mode command :)	02:59
gitte	Yeah, teach me.	02:59
Mikachu	gitte: i know you're angry but try not to spam with repeated kicks :)	02:59
gitte	Teach me.	03:00
Mikachu	just write /kickban instead of /kick	03:00
gitte	Thanks.	03:00
aeruder	gitte /mode #git +b <mask>	03:00
	Mikachu: that's a client thing, not a server thing	03:00
gitte	aeruder: what's that mean?	03:00
aeruder	gitte: sets the +b on a mask in the channel, so the mask is of the form	03:00
	nick!user@host	03:00
gitte	Ah.	03:01
aeruder	basically when you see Eridius (n=Eridius@growl/Eridius) has joined #git	03:01
mugwump	Seriously though, it does look like Linus is arguing against case folding, and eridius is arguing for unicode normalisation	03:01
aeruder	you could do /mode #git +b Eridius!n=Eridius@growl/Eridius	03:01
gitte	It's the first time I really felt the need to kick somebody in order to continue the very sensible conversation we had before.	03:01
aeruder	or Eridius!@	03:01
mugwump	unicode normalisation is probably a sensible thing, but that's more of a filesystem interface hack	03:01
gitte	Thank you, aeruder.	03:02
Mikachu	mugwump: linus uses case folding as an argument i think, but kevin doesn't see the analogy	03:02
loops	kick/banning is bad form.. ignoring is better answer...	03:02
gitster	mugwump: but abc vs xyz example applies equally to both folding and normalization.	03:02
aeruder	mugwump: no, the issue is that hfs munges filenames when you write to it	03:02
gitte	mugwump: the problem is that it solves nothing.	03:02
Mikachu	loops: there are hundreds of us and only one of him	03:02
aeruder	and somewhere it turned into a 'that's ok'	03:02
loops	Mikachu, yeah.. but he won't talk forever.. esp if nobody wants to talk with him	03:02
Mikachu	evidence show the opposite :)	03:03
loops	heh	03:03
gitte	The git mailing list is a meritocracy by nature, and nobody should turn it into a temporacracy.	03:03
aeruder	see the mailing list	03:03
gitster	what evidence?	03:03
	what I saw was everybody took turns to feed a single troll.	03:03
	Please do not do that.	03:03
gitte	loops: we tried that. Did not work. Which is why I tried to ban him.	03:04
Mikachu	gitte was trying to ask him to stop replying on the mailing list, with limited success	03:04
	and i suppose it got worse from there	03:04
gitster	So don't even try to ask. Just ignore.	03:04
loops	gitte, i can understand your frustration.. i just don't think we want to go down the road of excluding people we disagree with.	03:04
mugwump	well, a project-global gitattribute could specify which unicode normal form is desired	03:04
	and then translate to and from that with the filesystem as required	03:04
gitte	gitster: he did not just answer. He actively fed. And I was fed up. I'm sorry, I'm only human.	03:04
	loops: I would _never_ exclude a person I am _just_ disagreeing with.	03:05
	I learnt _too_ much on the list.	03:05
aeruder	mugwump: i'm not sure how that handles repositories that may have two equivalent unicode strings but not equivalent binary blob filenames	03:05
gitte	I was _too_ often wrong, being corrected, learning from it.	03:05
aeruder	or maybe the answer is, it just wouldn't	03:05
mugwump	repositories like that won't have the gitattribute that specifies one or the other	03:05
	heh. timeline	03:06
gitte	This case was even worse than Kastrup: I did not _see_ the relevant emails between all those crap emails.	03:06
	That's _different_ from kicking people who _disagree.	03:06
aeruder	mugwump: perhaps that's true, but the problem is that it isn't a repository issue	03:06
gitte	It is kicking people who are _disruptive_.	03:06
aeruder	it is one certain filesystem issue	03:06
mugwump	sure	03:06
gitster	"http://video.google.nl/videoplay?docid=-4216011961522818645"	03:07
aeruder	so i don't think a .gitattribute would be appropriate	03:07
mugwump	but in the attributes, you specify whether you want your diacritics expanded or contracted by default	03:07
gitster	I am not so sure about that approach, though.	03:07
mugwump	then if git detects a corrupting filesystem, it normalises as appropriate	03:07
aeruder	mugwump: yes, but that means that everyone needs to change their repository to work with hfs+ correctly	03:07
	i agree with the readdir() hack	03:07
gitte	gitster: exactly. The dak problem went away... luckily.	03:08
aeruder	which is a hack,but hfs deserves nothing more	03:08
Mikachu	mugwump: as i understand there's some problems with knowing what "as appropriate" means even	03:08
gitster	I was initially opposed to readdir() hack, but I am inclined to agree.	03:08
gitte	mugwump: I really think the problem can only be solved by working around MacOSX' insistence to normalise filenames.	03:08
aeruder	gitster: i don't know what else you can do, you create one file name and mac os x makes up its own	03:08
gitster	The most attractive aspect of that approach is that it does not have to hurt people outside Mac.	03:08
mugwump	Mikachu: unicode normalisation is pretty well defined	03:09
	gitte: I agree, but I think you need to know what you are normalising to in the repo	03:09
gitster	And for the case of HFS+, it is even more cast-in-rock defined ;-). Which is a gift and a curse, I guess.	03:09
aeruder	mugwump: yes, but what do you do when the gitattribute is not defined	03:10
gitte	mugwump: that's the beauty of the readdir approach: it does not matter what you are normalising to.	03:10
mugwump	how can that work though?	03:10
	you don't know whether to do NFKC -> NFC, etc	03:10
Mikachu	mugwump: i only know what i read in that thread	03:10
mugwump	http://unicode.org/reports/tr15/	03:11
gitte	mugwump: git can compare by "similarity", or "equality" if you want.	03:11
mugwump	sure, but which do you write initially?	03:11
aeruder	mugwump: you write the binary blob	03:11
mugwump	given that nothing will be in the index to compare to already, you don't know which form you want	03:11
aeruder	like any other filesystem	03:11
mugwump	blob encoding is a different issue, I'm just talking about filenames	03:12
imyousuf	I have a small IRC related question, how can I become operator of a channel (definitely not this channel :))?	03:12
context	someone ops you ....	03:13
aeruder	imyousuf: you talk to the corrent op	03:13
gitster	"git add ." will be fine with readdir(3); "git add foobar" you will convert the user string "foobar", perhaps in get_pathspec(). There are 37 different little corners like "git update-index --index-info" you need to deal with...	03:13
mugwump	I guess you'd need to be able to override the conversion in local config	03:13
	37? You have counted? :)	03:14
imyousuf	ah ok, thanks	03:14
aeruder	i'm not sure how the readdir approach solves the problem of different way of encoding strings on another computer adds a file to the repository	03:15
	→ lamont joined	03:15
aeruder	i.e. I add a latin-1 filename to git on a sane filesystem	03:15
	how does git realize that the unicode mess that it reads from the hfs filesystem matches what is in the index (which i'm assuming would be the latin-1 representation)	03:16
mugwump	well git could detect (or be configured/.gitattribute'd) what encoding each (index/repo, filesystem) is in	03:17
	defaulting to C when ambiguous	03:17
aeruder	well, the filesystem can't really be setup in the .gitattribute	03:17
	← KirinDave left	03:17
mugwump	no, the .gitattribute only controls which encoding (and Unicode normal form) you want in the repo	03:17
context	cant the encoding alrady be defined in gitrc	03:18
gitte	mugwump: sorry, I was trying to bring somebody to reason... Time wasted.	03:18
mugwump	you would override that probably only if the repository doesn't have it	03:18
aeruder	mugwump: what happens if a latin-1 representation is added and a .gitattribute is later setup to make everything utf-8	03:18
mugwump	well, that's a bad commit	03:18
	:)	03:18
	they should also re-encode all the filenames in that commit where the .gitattribute is added	03:18
aeruder	i wonder if there is some hack to find out what hfs decided your filename should be	03:18
mugwump	but it's not just utf-8 vs locale	03:19
	you've also got 4 different normalised forms for Unicode	03:19
aeruder	yes	03:19
mugwump	any of which might be the canonical form for a project and/or filesystem	03:19
gitte	mugwump: but there's a function that can tell you if the forms are "equivalent"	03:19
aeruder	and that's why i halfway would prefer some sort of mapping of what is in the index to what os x munged it to	03:19
	← sgrimm left	03:19
mugwump	gitte: that's right, those functions are somewhere in the nameprep stack	03:20
gitte	mugwump: my suggestion with readdir() was about using those functions.	03:20
mugwump	sure... but you probably only know what the filesystem is using, and would have to guess what the project is using	03:21
Ilari	mugwump: the short answer to question on NFC vs. NFKC is that valid NFKC is valid NFC, but conversion to K-forms destroys information (more than just precise encodings of characters).	03:21
gitte	mugwump: so, if there is a file, whose name is equivalent to a name in the index, take the name in the index instead.	03:21
mugwump	that only works if there is already something in the index	03:21
gitte	mugwump: if the project includes people from HFS+, they know.	03:21
mugwump	clear your index, git-add ., and you've destroyed that	03:21
gitte	mugwump: the upside of this approach: case-insensitivity is handled by just another equivalence function.	03:22
Mikachu	what is NFKC? is it something else than NFD?	03:22
mugwump	Mikachu: see the TR15 link I posted above	03:22
Mikachu	sorry	03:22
	← vintrepid_ left	03:22
gitte	mugwump: if there is nothing in the index, you can take what is there... does not matter if it is normalised or not... beautiful, i'nt it?	03:23
mugwump	it does matter	03:23
Ilari	Mikachu: C means composed (single codepoint if possibled), D means decomposed (pull all decorations out to be combining), K means normalization according to look.	03:23
mugwump	because the tree IDs will vary depending on the history	03:23
aeruder	personally, i have an even better solution for git on os x	03:23
	and this is a good one	03:23
	mugwump drumrolls for aeruder	03:23
aeruder	we handle it in the same way that we handle case sensitivity clashes	03:23
Ilari	Mikachu: IIRC, K normalizations for example transforms ² -> 2	03:23
aeruder	we don't	03:23
Mikachu	oh okay	03:24
aeruder	if you use os x + hfs, then stay away from unicode filenames, or at least make sure they are added on the os x side	03:24
mugwump	Ilari: close - see Figure 6 in that document	03:24
aeruder	i just really think that when you are using a crappy filesystem that does things that are so very un-unix-like, you should be willing to put up with the limitations	03:24
	i use git on os x all the time, and you know what, I ignore that file that is always listed as missing on git's testsuite	03:25
	it is that easy	03:25
mugwump	Ilari: actually, not just "close", actually "accurate" :)	03:25
	s/actually//	03:25
	Anyway, I think it would be a good feature to allow a project to specify which encoding filenames are in	03:25
	Just so that tree IDs align etc	03:26
	← Bluefury left	03:27
mugwump	And with that information, the conversion required becomes deterministic	03:27
aeruder	but then wouldn't all git stat's be stuck with determing encodings on disk, converting them to your .gitattribute'd encoding, and then comparing with the index?	03:28
	← lamont left	03:29
Ilari	aeruder: You can probably cache what they look at disk into the index (it already caches stat info of working tree file).	03:29
aeruder	so why not just have the index cache the filename on disk that gets created and it solves the os x issue	03:29
	→ lamont joined	03:30
mugwump	aeruder: it's a deterministic, fast conversion, I'm not particularly worried about the performance impact.	03:30
	I don't think caching that info in the index would add anything	03:30
aeruder	is it deterministic when you don't know the encoding on disk?	03:31
gitte	The only big problem I see is that all of a sudden, filenames are no longer unique identifiers.	03:31
Ilari	aeruder: If Mac OS X normalized to NFC, it would solve 99% of issue (the remaining 1% is essentially unsolvable). But it normalizes to NFD...	03:31
mugwump	Not talking about being case/normalisation insensitive, though	03:31
	gitte: right, which is why I would want them normalised on the way in	03:31
	and preferably, on the way out too	03:31
gitte	mugwump: But I don't want HFS+ to dictate the way how git handles things!	03:31
mugwump	it doesn't have to	03:32
	I just want my project to be able to specify its filename encoding form	03:32
	in gitattributes	03:32
	and the rest is a platform issue for those users	03:32
pasky	04:29 -ChanServ(ChanServ@services.)- [Mikachu] has been added to the access list for #git with level [29]	03:32
Mikachu	i saw	03:32
pasky	oh did you? cool	03:32
Mikachu	well, it said "You", not [Mikachu]" :)	03:32
gitte	pasky: thanks.	03:32
pasky	;)	03:33
	Mikachu set mode: -o	03:33
	dwmw2_ULN → dwmw2_gone	03:33
gitte	mugwump: FWIW I really agree with others that it is not about a gitattribute.	03:33
	gitattributes are per-file.	03:33
mugwump	they can be per-directory as well, right?	03:33
gitte	mugwump: What you are talking about is really per-repository	03:33
	mugwump reads docs	03:33
mugwump	yes and no	03:33
	yes, for sanity	03:34
	→ macournoyer joined	03:34
gitte	Well, you _can_ extend attributes over a whole directory, yes.	03:34
	But inherently, attributes are _still_ per-file, even if you can say "* attribute=mugwump" ;-	03:34
	;-)	03:34
mugwump	It would be silly to not use a consistent encoding for filenames for a whole project, I agree	03:34
	but gitattributes seems to be the right mechanism	03:34
	for configuring what is intended to be the repository form, anyway	03:35
aeruder	man, i'm missing something or ... but how does this solve the os x issue?	03:36
mugwump	what you do locally - case/normalisation insensitivity via the readdir()+index thingy, etc, well that's a config issue	03:36
	(or, a platform default config issue)	03:36
	It would also mean Unix users can have case insensitivity if they want it! I can see them lining up to submit patches now	03:36
Ilari	Anybody knows how filenames with nontrivial NFD forms when added to project behave on other platforms?	03:36
aeruder	Ilari: there's only one filesystem (that I know of) that even looks at the encoding of your filename	03:37
mugwump	vfat has an encoding	03:37
aeruder	and that's the reason why I still don't see how this fixes, say, the git repo problem in the testsuite	03:37
mugwump	which file?	03:38
Ilari	aeruder: Yes, I know that native Linux filesystems don't even look at the encoding, but it still doesn't mean that working with them is pleasant....	03:38
gitte	mugwump: FWIW I think it would be wrong to handle it with .gitattributes.	03:39
	It would encourage enforcing it for _one_ file, but not for another.	03:39
mugwump	gitte, is there another in-repository mechanism for it, though?	03:39
gitte	I'd prefer a config variable which makes it independent of the filename.	03:39
mugwump	it would be pretty stupid to do it on anything other than a directory	03:39
Ilari	aeruder: Say you have files named char_å and char_ä in repo. If names are NFC-normalized, stuff like that isn't too bad. But it could get nasty if they are NFD-normalized.	03:40
Mikachu	the problem is that config variables aren't cloned, i guess	03:40
aeruder	# gitweb/test/Märchen	03:40
	mugwump: that file	03:40
mugwump	ok, so take ä - its NFD form is "a" U+0308	03:41
aeruder	yep, but the problem is	03:41
mugwump	its NFC form is U+E4	03:41
aeruder	does that mean that the git repo needs to set an encoding, because you can't say that you even have an ? until you assume that it is unicode	03:42
	especially if the default encoding is just none	03:42
	← Strogg left	03:42
	→ Strogg joined	03:42
mugwump	that's right - you'd have to config what the repository format was if the repository doesn't specify it itself	03:43
aeruder	mugwump: ok, so repositories would still more or less have to go out of their way to make themselves work on os x	03:43
mugwump	possibly a global default, eg "utf8+nfkc" or something	03:43
aeruder	even though the .gitattributes things could have benefitts besides...	03:44
Cerebon	I prefer to just use ufs disk images myself on osx	03:44
	← tongueroo left	03:44
aeruder	but the point is, at what point is it worth making git work with hfs, when you could just as easily recommit the file in a way that wouldn't get rewritten on os x	03:44
Ilari	My current opinion is that if filenames are in UTF-8, putting anything but NFC-normalized filenames in repository is insane.	03:44
aeruder	since either way you're changing the repositories to work with hfs	03:44
mugwump	this will enhance usability on any crappy filesystem	03:44
Cerebon	hfsx/hfs+j/hfs+ are the ones that do the mangling	03:44
Ilari	mugwump: Make that utf8+nfc...	03:44
mugwump	sure	03:44
aeruder	Ilari: you do you make assumption that the filenames are utf-8 at all though	03:45
mugwump	and perhaps people don't want any random encoded filenames in their project, and want git to reject them based on .gitattributes	03:45
	(or global repo encoding config t.b.d)	03:45
aeruder	i mean, maybe a system-wide default of utf8+nfs isn't bad, but you're still in the business of treating binary blobs (i'm referring to the filenames) as unicode on repositories that are already preexisting	03:46
Cerebon	as long as it is compile time, or config only on osx so it doesn't kick in for me I don't care what gets implemented	03:46
aeruder	repositories that may not have had their files added as utf8+nfc	03:46
Cerebon	could readdir optionally normalize to nfc on osx, or something weirder? I read the link about what limewire proposed. Seemed complicated.	03:48
aeruder	Cerebon: the problem is that then you have to make some assumption about what encoding the index is in	03:48
	i'm not saying i'm opposed to the .gitattributes thing that mugwump is talking about	03:49
	and in practice it could work towards solving some os x issues	03:49
gitte	Hey, people, I think that I can solve that problem	03:49
Cerebon	that is the thing though, it is hfs, not osx	03:49
Ilari	aeruder: Maybe I made a thinko somewhere and made statement that both makes sense for non-UTF8 encodings and doesn't explicitly assume UTF-8... :-/	03:50
aeruder	but i think at some point you also have to realize that hfs is just broken	03:50
Cerebon	ufs on osx has NO issues	03:50
aeruder	Cerebon: sorry, when i say osx, i mean hfs	03:50
gitte	But due to a person that annoyed me too much (you know that person), I decided to delay it.	03:50
aeruder	not because i don't know that other filesystems are fine, i'm just talking the typical osx issues ;)	03:50
gitte	It's actually a good thing to delay it, until 1.5.4 comes out.	03:50
	_Then_ we can talk shop.	03:50
Cerebon	aeruder: roger, just want to get that clear, no worries	03:51
aeruder	yea, it is clear	03:51
Cerebon	aeruder: I partitioned my drive specifically so I have no issues, this isn't a git only problem, svn/hg etc.. all have issues on hfs	03:52
aeruder	yes, i know it, i'm just of the opinion that you can't ever fully fix the crappiness that is hfs	03:52
	→ vbgunz joined	03:52
aeruder	so at some point the responsibility of worrying about it just falls on the maintainer's hands and whether or not they care to help out the os x users	03:52
Cerebon	that is partially why I personally think it is more of a user needs to do x issue, but I suppose I am in the minority there	03:53
aeruder	yes, like use git in a disk image	03:53
	os x makes that ridiculously easy to do	03:53
Ilari	mugwump: The reason I wanted utf8+nfc and not utf8+nfkc for UTF-8 filenames is that I think NFKC normalization destroys too much information for this application...	03:53
Cerebon	hdiutil create -size foo -foo -fs UFS -attach -volname foo foo.dmg does it	03:54
aeruder	exactly	03:55
Cerebon	I created a script to test out all of the os x filesystems and checkout the git source, plain hfs sucks worse than a black hole of suck, and ufs is the only sane choice until zfs comes along	03:55
Ilari	Cerebon: Is there case-sensitivity option on UFS?	03:56
aeruder	i doubt it	03:56
Cerebon	ilari: ufs behaves like it does on freebsd, e.g. it is case sensitive only	03:56
Ilari	Cerebon: Ah, good.	03:57
aeruder	hfs lets you do case sensitive too	03:57
	doesn't help with this issue	03:57
	Ilari doesn't like case-insensitive systems.	03:57
Mikachu	on linux you can make anything case insensitive with fuse :)	03:57
Cerebon	although on leopard the man page for newfs notes UFS is being deprecated	03:57
	aeruder: hfsx is the case sensitve, but name mangling still version of hfs	03:57
aeruder	its just, i don't know why some people (cough K.B. cough) thinks git should change for their filesystem	03:59
Cerebon	heh, he didn't like my disk image idea	03:59
	I will send the one that amounts to a ufs ramdisk! :)	03:59
aeruder	when really it comes down to an issue of a.) talking to the repo owners to make their data (filenames) work with os x b.) using a disk image	03:59
Cerebon	it'll be really fast!	03:59
	c: partition as ufs	04:00
	← csc` left	04:00
aeruder	i mean, we can solve this issue, but you'll still never be able to checkout the linux source on os x due to case insensitivity	04:00
mugwump	Ilari: yes, nfc is probably a better default for those users	04:00
aeruder	so do we implement some sort of work around for that?	04:00
Cerebon	If it bugs someone enough that they provide working code, so be it, I personally think it is wasting time in discussion myself	04:01
mugwump	aeruder: I think I'd have to see a real life use case for a clash before I'd worry about that	04:01
	← harinath left	04:02
mugwump	checking out linux source on Mac OS - do people actually do that?	04:02
Cerebon	it isn't like developers need to be pampered like most osx users might	04:02
mugwump	I mean, you probably couldn't build it on Mac OS	04:02
Ilari	What that filename mapping stuff is going to do for index though? I know that the format supports mandatory and optional extensions...	04:02
kylem	mugwump, you can.	04:02
Cerebon	mugwump: you might with a cross compiler	04:02
mugwump	sure, you can, but who would ? :)	04:03
	Anyway, I think it's something of a different problem	04:03
aeruder	mugwump: why not? i'm just saying at what point do you stop making workarounds for os x and just say too bad, talk to the repo person to fix it	04:03
	→ harinath joined	04:03
kylem	someone who wants to hack on linux but also wants working suspend and resume? :)	04:03
aeruder	mugwump: why, there could be clashes just the same with hfs rewriting the encoding	04:03
kylem	aeruder, you can just make a disk image with case sensitive hfs and use that.	04:03
mugwump	aeruder: yes - in which case on those filesystems, you can't make working copies for the project - tough.	04:04
kylem	(s'what i do...)	04:04
mugwump	Well, you can, if you can change the normalisation rules	04:04
	which is why you'd want it per-path, for the inevitable hacks	04:05
aeruder	mugwump: and so how is that any worse than 'you have a repo not really setup for your borked filesystem, too bad'	04:05
mugwump	.oO{ is there a local-equivalent of .gitattributes? }	04:05
Cerebon	I am evil, I am checking out the linux source on osx just to be a pain	04:05
mugwump	aeruder: well, it's quite different I think	04:05
	say that your filesystem is normalising to nfd	04:06
	and the project is nfc	04:06
	you can do it	04:06
aeruder	true	04:06
mugwump	you can apply that to latin-1 vs utf-8	04:06
aeruder	based on the assumption that you know the encoding of both	04:06
mugwump	exactly	04:06
aeruder	which you don't	04:06
	unless every git repo goes out of its way to specify	04:06
mugwump	users must be able to override it in a local .git/attributes or something	04:07
	(or specify it, in the regular case where it is not specified)	04:07
aeruder	mugwump: obviously the ability to do something isn't stopping K.B. from whining or he'd just use a disk image	04:07
	so overriding it on .git/attributes is a moot point	04:07
Cerebon	or newfs	04:08
mugwump	so, maybe your config is set to assume utf8+nfc on repositories where it is not specified	04:08
	and you've diligently patched git for your platform to know what the right encoding for the filesystem being written to is	04:08
aeruder	i think the only valid default you can have though is no encoding	04:09
mugwump	you mean C ?	04:09
aeruder	yes	04:09
	unless you're suggesting that on os x the default be different	04:09
	in which case i wouldn't horribly disagree	04:09
mugwump	only users with corrupting filesystems are forced to set it	04:09
	and then only if they are checking out repositories with high bits in filenames	04:10
	→ harinat1 joined	04:10
aeruder	i would be mildly happy with adding what you are talking about and just telling os x users to set the default attribute in .gitconfig	04:10
	where the default default attribute is C	04:10
	i think it could possibly fix some os x issues, they'd have to have an extra git config setup line though	04:11
mugwump	like I say, I think it is a generally useful feature to be able to enforce sanely encoded filenames	04:12
	it's not just catering to os x users	04:12
aeruder	yes, i'm just saying that the default encoding should be imo C, and changeable via a config option	04:12
Mikachu	can't you just make a pre-commit hook?	04:12
Cerebon	in that case might as well allow nfc/utf8 encoding	04:12
aeruder	(speaking of the case when the .gitattributes does not specify)	04:12
	i.e. if .gitattributes specifies, use that, otherwise fall back to the encoding in the config option which defaults to C	04:13
Cerebon	mugwump: ok, don't make me cross compile the 2.6 kernel, it checks out fine on ufs btw	04:13
mugwump	Cerebon: if a user knows how to set up a cross-compiler I bet they can format a ufs partition	04:14
Mikachu	Cerebon: git status reports no changes too?	04:15
Cerebon	yeppers	04:15
aeruder	but that's UFS, it is as expected	04:15
Cerebon	$ git status	04:15
	# On branch master	04:15
	nothing to commit (working directory clean)	04:15
aeruder	mugwump: what about that?	04:16
mugwump	aeruder: afaict you repeated back what I suggested, so I guess we agree :)	04:17
	or, at least, along the lines I was thinking	04:17
aeruder	sorry, just reiterating that i feel strongly that the default encoding should be C	04:17
	and if you're on an OS X machine, and that causes issues for you, you can git config --global filename.encoding utf-8	04:18
	or whatever	04:18
Cerebon	I would argue it should be the current LANG value	04:18
mugwump	that's a good thing to re-iterate	04:18
aeruder	Cerebon: that doesn't make sense	04:18
Ilari	aeruder: You mean binary encoding should be default?	04:18
aeruder	yes, of course	04:18
	if the repository has actual text strings in their repo	04:19
	they should setup a .gitattribute	04:19
	otherwise git needs to fall back on operating like it always has	04:19
	assuming they are binary blobs, not a text string	04:19
	you can override that of course using git-config if you feel otherwise	04:19
	← bdiego left	04:20
Cerebon	say I have a file named Tschüß, if I am running en_US.UTF8 it can store that or C, or am I talking out of my ass? I admit my knowledge of unicode falls down here	04:20
aeruder	Cerebon: no, the problem isn't necessarily how it stores it, it is the fact that let's say I have a latin-1 filename in my repo	04:20
	and no .gitattributes to say what encoding it is	04:21
	you can make no other assumption than C without possibly borking it up	04:21
	which is why the git-config should be explicit that you are telling git to stop treating filenames as just a binary blob of data as it always has and actually do the encoding stuff	04:22
Cerebon	got it, so would it be sane(r) to put the onus on the repository owner to specify "all files shall be encoded by name as latin-1, UTF8/6/etc..?"	04:22
	← harinath left	04:22
aeruder	Cerebon: yes, by specifying a .gitattributes (which is preferably in line with what format the current repository is in)	04:23
gitte	As I said, I'd appreciate it being a config option...	04:24
mugwump	http://git.pastebin.com/m3d522278	04:24
gitte	Not everybody has MacOSX.	04:24
	It's the same as with CR/LF: not everybody has Windows, so not everybody has to suffer.	04:24
mugwump	indeed	04:24
Cerebon	gitte: even those that do don't necessarily need this	04:24
mugwump	if you have a filesystem with C semantics, it is not required unless you want git to enforce encoding sanity for you	04:25
gitte	Exactly, and in that case, you should put it into /etc/gitconfig.	04:25
	← Industrial left	04:25
mugwump	sure, but you need to know which encoding the project is using so you know the correct behaviour!	04:25
aeruder	so what happens when I create a latin-1 filename	04:25
	and then git add it	04:26
mugwump	ok, so assuming the project says it wants utf-8 in the repo	04:27
	and you've configured via git-config (or even locale) that you are using latin-1 locally	04:27
	pathnames are transcoded on the way into the index	04:28
	and the way out	04:28
aeruder	ah, so what happens when you've configured C encoding for the repository and you check it out on a machine that specifies utf-8 for local files	04:28
	← flaguy left	04:28
aeruder	does it attempt to just flat out treat the data in the repo as utf-8 and write directly?	04:29
Cerebon	I would assume use iconv to reencode?	04:29
mugwump	aeruder: I cover that in the second para of the doc patch I nopasted	04:29
aeruder	ah, so you're saying (if I understand that correctly) that in that case I just mentioned, it refuses to check out the repository until I set C for my filesystem encoding?	04:31
gitte	The point is: if a couple of people decide to develop using Latin-1 filenames, or codepage 892 for that matter, git should not stop them.	04:31
Ilari	What about saying in commit object how to interpret the names in tree (since there is no way to extend trees and introducing new object type is too much work)? Binary (default), latin1 vs. UTF8-NFC... Or is implementing that too complicated due to transcoding (which OTOH seems necressary anyway)?	04:32
gitte	And if there are people developing on both Linux and MacOSX, I do not see why the Linux people should be affected by the MacOSX madness.	04:33
aeruder	gitte: i partially agree with you that os x should not be a concern	04:33
	If I'm on os x, making a repository on os x, then all the files will get added in the appropriate encoding	04:33
Cerebon	gitte: hell, I use osx and I agree, but I have used unix for a long time so it is a non issue if you ask me	04:34
Ilari	gitte: That was exactly my point earlier when strongly advocating NFC if UTF-8 is used...	04:34
aeruder	if I'm on os x, and I don't own the repo, i just deal with it if I can't convince the project owners that it is worthwhile to change	04:34
	Ilari: right now there is no encoding on filenames	04:34
Ilari	aeruder: I know that...	04:34
Cerebon	seems simpler if there isn't	04:35
aeruder	s/seems //	04:35
Ilari	aeruder: That's why binary is the default... To be compatible with existing practice...	04:35
gitte	Ilari: but the point is: NFC should _not_ be _unilaterally_ used.	04:35
aeruder	Ilari: I just see that the encoding thing if I'm understanding it correctly would be more of a mess than now	04:35
Mikachu	if i can ask a tangential question, can you convert between normalization forms with the iconv command?	04:36
aeruder	which requires you to set a repository encoding AND set a filesystem encoding	04:36
gitte	Ilari: only on systems that do not care about "code points" (as Linus points it, according to the specs) should there be possible exceptions.	04:36
aeruder	and complain when the two settings aren't compatible	04:36
Ilari	aeruder: You got to be able to tell what the encoding is somehow to be able to transcode it...	04:37
aeruder	Ilari: yes, i know, i'm just saying, let's say you want all files on your filesystem transcoded to utf-8	04:38
gitte	Mikachu: AFAIK yes (but I'm no expert on iconv)	04:39
aeruder	so you set the config option to utf-8, now you check out any of the bajillion repos with no .gitattributes, does it now refuse to check out since it doesn't know how to transcode?	04:39
	← fhobia left	04:40
gitte	Same problem as for crlf.	04:41
	tongue-in-cheek: join Steffen's discussion.	04:41
Ilari	I guess I should set up test repo that uses nontrivial NFD just to verify and possibly demonstrate or prove wrong my points about NFC and NFD...	04:43
gitte	Hey, that would be very good!	04:44
	I imagine we could even integrate it into a test case (since git is byte-wise precise ;-)	04:44
mugwump	ok, I've submitted the suggestion that (at least) aeruder and I casually agreed might work	04:49
Ilari	gitte: What would be the charset name? I don't see anything mentioning 'NFC' or 'NFD'... And all options regarding unicode seem just be different encodings like UTF-8, UTF-16 (big and little endian), UTF-32 (again the two variants)...	04:49
gitte	Ilari: does not matter, does it?	04:50
mugwump	currently all that is supported is iso-8859-1 and utf-8	04:50
gitte	Ilari: if you write a name in NFD on MacOSX, it will probably be written as NFD, right?	04:50
Mikachu	Ilari: they're the same encoding, UTF-8	04:50
gitte	And if you add it as NFC, it will be stored in the index as NFC...	04:51
Mikachu	Ilari: the charset is "unicode" for all of them too of course	04:51
gitte	So there's our test case.	04:51
Ilari	gitte: Yes, and I'm worried that using NFD will cause trouble on other systems...	04:51
Mikachu	gtk+2 (or pango) isn't very good at rendering NFD	04:51
	← bobesponja left	04:51
Mikachu	in my experience	04:51
Ilari	gitte: I was talking about using iconv to convert to NFC... I couldn't find any 'codec' supporting it.	04:52
gitte	Ilari: I _think_ that other systems are sane enough not to trifle with the way you specify the filename when creating the file.	04:52
	Oh, I know not enough about iconv to say if it has a codec, or how to convert between NFD<->NFC with it.	04:53
	Sorry. ;-)	04:53
mugwump	you might need another library for the normalization	04:53
Ilari	mugwump: Not to mention it requires fairly sizable tables...	04:54
mugwump	sure	04:54
Mikachu	Ilari: i only know you can use //TRANSLIT when doing utf8 to something less to transliterate	04:54
mugwump	don't have those installed, don't build against them	04:54
Mikachu	dunno if there's something ilke that for norm	04:54
mugwump	Hmm I think my patch got blocked	04:54
gitte	I'd imagine that there's something like "iconv_strcmp()"...	04:55
Cerebon	ok again, this is way more than we should NEED TO do, even if it was windows. I think the crlf thing had more merit than this	04:55
gitte	mugwump: I have the feeling that the spam filters had to be upgraded ;-)	04:55
mugwump	I probably naffed up the encoding	04:55
gitte	mugwump: you know because of whom ;-)	04:55
mugwump	oh has the flamewar switch been flipped	04:56
gitte	Probably all people with MacOSX are blocked ;-)	04:57
	Well, not really, vger is much nicer than me ;-)	04:58
Cerebon	gitte: can I at least tell him to just newfs a partition on his macs and shut up?	04:59
aeruder	please no	04:59
	:)	04:59
	if we all stop talking to him, he may stop e-mailing back	05:00
	aeruder crosses his fingers	05:00
	gitte too	05:00
	→ panagos joined	05:00
Cerebon	how are you two typing with crossed fingers?	05:00
	→ notahat joined	05:00
aeruder	very slowly	05:01
	← kevwil left	05:01
panagos	when i modify a file, do i also have to run git add <file> every time prior to commiting?	05:01
aeruder	panagos: yes, that is the best way, or you can also use git commit -a	05:02
gitte	Cerebon: I croos my left two fingers, and type with the right two fingers.	05:02
panagos	aha	05:02
gitte	panagos: you can also specify what to commit by "git commit <paths>".	05:02
aeruder	note that his typo involved a letter he would have typed with his left hand, but obviously is having some issues with the fully right-handed typing	05:03
	aeruder gives gitte bonus points for believability	05:03
panagos	aha	05:03
mugwump	heh, apology post from Eridius	05:03
aeruder	he can't just say	05:04
	i'm sorry for being a moron	05:04
	he has to sneak his point into it too	05:04
	thus reopening the whole freaking flamefest	05:04
imyousuf	aeruder: at least he said sorry :)	05:04
gitte	Well, at least he apologised.	05:04
	Maybe now people can ignore him.	05:05
panagos	i can add files to a coommit prior to commiting? sort of like gathering them together? && how?	05:05
imyousuf	actually he has more apologies pending :-D	05:05
aeruder	yes, he tacked i'm sorry into the same email he's sent 900 times now	05:05
gitte	(In peace, of course)	05:05
aeruder	panagos: yes, git add the files you want to be part of a commit	05:05
tpope	which one is the apology?	05:05
gitte	panagos: yes: the index is sort of a staging area.	05:05
aeruder	you can see what you have added is	05:05
gitte	tpope: [email@hidden.address]	05:05
aeruder	git diff --cached	05:05
	and then git commit	05:05
	will commit your cached changes	05:05
mugwump	ok, I've re-sent the patch with an extra header. I'll stop overnight now though :)	05:06
gitte	tpope: so there are really three areas: working directory -- index -- commit.	05:06
mugwump	There's a big Cc: list now	05:06
panagos	hmm, i see cogito had a pretty different approach when i used it	05:06
tpope	panagos: ^	05:06
mugwump	Oh whoops, didn't really mean to drop Linus from it	05:06
panagos	tpope: ?	05:07
	ChanServ set mode: +o	05:07
	Mikachu set mode: -b	05:07
aeruder	panagos: gitte was talking to you and meant panagos	05:07
	Mikachu set mode: -o	05:07
Mikachu	hope that is okay	05:07
aeruder	er, was talking to tpope and meant you	05:07
panagos	oh	05:07
mugwump	ah, got to the list that time	05:07
gitte	"I am a stubborn man, as I believe most of you are." -- reminds me of that scene in Charlie Wilson's War, when he's in Pakisan, asking for Whisky ;-)	05:08
aeruder	gitte: "I'm sure many people make that mistake" "No."	05:08
gitte	tpope: yes, cogito tried to hied the staging area from you. I think that was wrong. But others disagree.	05:08
tpope	bah well the web interface doesn't provide a way to retrieve by message idea	05:08
	panagos: ^	05:09
gitte	Mikachu: let's try.	05:09
tpope	again	05:09
	ChanServ set mode: +o	05:09
aeruder	heh	05:09
	gitte is having some issues tonight apparently ;)	05:09
panagos	...............	05:09
gitte	aeruder: right you are.	05:09
tpope	s/idea/id/	05:09
aeruder	or his tab complete is completing like this: pan<tab>^H^H^Htpope	05:09
gitte	aeruder: actually, on both issues ;-)	05:09
panagos	LOL	05:10
Mikachu	why did you op me again?	05:10
tpope	ha	05:10
gitte	Mikachu: so you can do things that I am too stupid to do myself ;-)	05:10
Mikachu	but i already unbanned him	05:10
gitte	I know, I read it.	05:10
Mikachu	pasky gave me my own op power	05:10
	maybe you missed it	05:10
	ChanServ set mode: -o	05:11
Mikachu	see	05:11
gitte	Yes, I missed it, I thought you were talking about "30" earlier, and he gave you "29" or somethin.	05:11
	Whatever... ;-)	05:11
	If nothing else, I publically showed you my trust in you ;-)	05:11
Mikachu	that was about you not being able to give it to me	05:11
	:)	05:12
	→ tongueroo joined	05:12
gitte	Yep!	05:12
	aeruder: be honest: how many movies do you watch, on average, per week?	05:14
aeruder	0.2 maybe :)	05:14
gitte	(IIRC you were pretty good on my pirates-of-the-carribean riddle, too)	05:15
	Wow.	05:15
aeruder	i just happened to watch quite a few over christmas break	05:15
	:)	05:15
gitte	Hehe.	05:15
	I just watched Charlie.	05:15
aeruder	yea, i watched it a few weeks ago, pretty good :)	05:15
gitte	Tomorrow's movie is "Kiterunner"...	05:15
	Well, Charlie was a bit biased, I thinkl	05:15
mugwump	I really enjoyed "The Devil Dared Me To" the other night	05:16
gitte	New film=	05:16
	s/=/?	05:16
aeruder	now i watch even less stuff considering i cancelled my cable :)	05:16
gitte	aeruder: good thing. I don't even have a TV. I only get DVDs from time to time. And the occasional cinema break.	05:17
	Ilari watches on average ~0 movies a week.	05:17
aeruder	but the movie theatre here is $4.00 matinee	05:17
gitte	With Charlie, I was wondering if people would have cheered also, if it was not Afghanistan, but Vietnam, and not Russians but Americans...	05:17
	And I really missed seeing OBL.	05:18
	Since all the monies by Charlie were really funneled through him... and he was particularly missing from the movie.	05:18
Ilari	aeruder: Yea, good thing, there are channels that are maybe 1% good stuff and 99% garbage. Then there probably are channels the other way...	05:23
aeruder	Ilari: plus my full utilities bill (electricity + cable services (inet)) came out to $65.00	05:23
Ilari	aeruder: Yea, dropping nonessential consumption is often a good thing.	05:24
gitte	Ilari: I like to believe that vger is the other way...	05:24
	← macournoyer left	05:24
gitte	aeruder: per-month or per-year?	05:25
Ilari	gitte: Was talking about TV channels. But I guess it can apply to mailing lists too..	05:25
aeruder	gitte: per-month	05:25
gitte	Ilari: lol... I got that wrong!	05:25
	aeruder: wow. You have a 80186 cluster going or what?	05:25
aeruder	gitte is still scheming to rid the world of Eridius, it is distracting to say the least	05:25
	gitte: hm?	05:25
Ilari	gitte: 80186 did exist, but those very pretty rare...	05:26
	Said 'probably' on the other way around, because one FTA channel here is more like 1% good, 99% garbage, but there aren't any really other way around (though at least one has more good than garbage).	05:29
gitte	aeruder: no, I'm just tired and slow by now... Besides, I want to think that I accomplished my goal.	05:29
	← nbrunson left	05:29
aeruder	heh	05:29
gitte	Ilari: yes, they were, and as far as I remember, they were quite some power consumers...	05:30
Ilari	gitte: Where they relatively worse than those Prescotts (Preshotts)? :-)	05:31
aeruder	heh, 285 kwh isn't bad on my apmt	05:31
Ilari	gitte: As heat generation = power consumption.	05:31
aeruder	i can live with it anyway	05:31
gitte	Hehe.	05:31
	You sure don't need heating during the winter time ;-9	05:31
aeruder	i apparently had 1100 kwh in september, so i'm doing pretty good, hehe	05:32
Ilari	gitte: Winter? What's that? You mean cold period?	05:33
	:-)	05:33
gitte	Ilari: where do you live that you don't have to suffer any winter?	05:33
Cerebon	it is -14f 6c here right now, if you are warmer, I hate you!	05:33
	aeruder suspects Cerebon hates a lot of people	05:34
	Cerebon suspects aeruder is right	05:34
	Cerebon suspects he should move to brazil and be happy	05:34
aeruder	minnesota ? :)	05:34
Ilari	gitte: ~60N. Notice the smiley...	05:34
Cerebon	ja, it sucks	05:34
aeruder	Cerebon: texas.... sucks to be you...	05:35
	but then again, i have to live in texas	05:35
	so that point goes to you	05:35
Cerebon	heh, evens out a bit with that one	05:35
	then again we have bridges collapsing at random so...	05:35
	→ saintdev_ joined	05:35
gitte	Cerebon: it is warmer here alright. But then, the sun is hardly up when it already goes down. Sucks.	05:35
	Ilari: 56°N here...	05:36
Ilari	Heating is less of a problem than cooling. One may live in place where very little heating is required, but cooling requirements in summer are massive.	05:37
	There	05:37
Cerebon	it can get over 100 here in summer	05:38
	← notahat left	05:38
Cerebon	which must sound warm in comparison to texas :)	05:38
Ilari	There's for example co-generation, which is very efficient compared to plain electricity generation if heating is required.	05:39
gitte	Cerebon: is that Celsius? Or do you still think in Fahrenheit ;-)	05:39
	← cworth left	05:40
Cerebon	sorry 100f 37ish C	05:40
	→ njero joined	05:40
Cerebon	Celsius tends to be a verboten thing here in america	05:41
	I say we just use kelvin and be done with it	05:41
gitte	I'm all for Calvin.	05:41
	But don't forget Hobbes.	05:41
	He's hilarious.	05:41
	← saintdev left	05:41
Ilari	Fun things to ask from those still using FFUs: How many grains there is in ounce and how many ounces in pound. :-)	05:42
Cerebon	heh, we don't use grains, but the ounces in a pound I can do	05:43
	but why is airbus using metric inches instead of pure metric?	05:44
	→ kanru joined	05:44
gitte	Dunno.	05:45
	In any case, time for me.	05:45
	Git night!	05:45
	← gitte left	05:46
	← RobotDeathSquad left	05:46
Cerebon	http://www.theonion.com/content/node/72502	05:50
	saintdev_ → saintdev	05:53
	← harinat1 left	05:55
	← njero left	05:56
	→ tongueroo_ joined	06:11
	→ notahat joined	06:17
	→ njero joined	06:22
	→ fhobia joined	06:25
	← tongueroo left	06:27
	← panagos left	06:31
	→ Eridius joined	06:32
	→ devogon joined	06:34
	→ robinr joined	06:39
	→ blake joined	06:40
	→ bronson joined	06:44
	dwmw2_gone → dwmw2_ULN	06:47
	→ kylem_ joined	06:50
	← robinr left	06:50
	← rhett left	06:56
	→ namenlos joined	06:56
	→ robinr joined	06:58
	← pergesu left	07:03
	← mithro left	07:03
	← kylem left	07:06
	→ garrin_ joined	07:07
	← garrin left	07:07
	→ pmezard joined	07:15
	← robinr left	07:20
	← tongueroo_ left	07:31
	→ panagos joined	07:31
	← pmezard left	07:37
	→ tongueroo joined	07:38
	→ FunkeeMonk joined	07:45
	← fhobia left	07:54
	← johan-s left	07:56
	← tongueroo left	08:08
	← notahat left	08:11
imyousuf	I have cloned a remote repo and started working on it. After several commits in my local repo can I do a diff with my index and remote repo? if I can, than how?	08:14
	s/than/then/	08:15
	← anttih left	08:16
	→ anttih joined	08:17
	dwmw2_ULN → dwmw2_gone	08:19
	← panagos left	08:19
	→ Eludias joined	08:23
	→ pmezard joined	08:25
	← madewokherd left	08:37
Mikachu	imyousuf: git diff origin/master probably	08:40
	→ octopod joined	08:40
	→ priidu_ joined	08:40
imyousuf	ah ok, Mikachu will check it and let you know.	08:40
	→ obvio_ joined	08:42
	← obvio171 left	08:42
	→ FauxFaux joined	08:45
	← sverrej_ left	08:49
	→ johan-s joined	08:53
hliusv561	:encoding utf-8 iso-8859-1	09:01
	oops.	09:01
	imyousuf: If you want to get individual patch for each commit, then maybe 'git format-patch' should be used...	09:02
imyousuf	hliusv561: I knew that as I have sent out some elementary patches	09:03
	I actually was asked this particular question by my boss :) in comparison with p4 and svn :)	09:03
	he was mentioning it as an advantage and as I did not know how to do it using git I could not reply back :), now after testing I can	09:04
	→ marcel joined	09:04
	← obvio_ left	09:06
imyousuf	Mikachu: wouldnt "git-diff origin/master" take the origin master branch from the local repo?	09:09
hliusv561	imyousuf: Yes, I uses local copy of origin master branch...	09:11
imyousuf	ah	09:11
	but I want to diff with the remote	09:12
	*remote's origin master branch	09:12
	so that if the original remote repo has changed and I have not pulled down I can still see the diff	09:12
	how could that be possible?	09:12
hliusv561	imyousuf: You need to fetch those changes first then.	09:13
imyousuf	I see	09:13
	→ kzak_ joined	09:15
	← jmspeex__ left	09:18
	→ TuxCM- joined	09:19
	← priidu_ left	09:23
	→ Dodji joined	09:24
	← Tux__ left	09:35
	→ jmspeex joined	09:35
	→ sverrej joined	09:44
	→ Fullmoon joined	09:59
Fullmoon	Can i make git ignore mode-hanges?	10:00
	*changes	10:00
mugwump	read Documentation/gitattributes.txt	10:01
hliusv561	Fullmoon: You have some filesystem which does not support +x/-x?	10:02
mugwump	I'm not sure if that system works for file modes though ... it's just a filter	10:02
gitster	core.filemode, if mode bits are unreliable.	10:05
	← kzak_ left	10:06
	← Spyro left	10:06
Fullmoon	Thanks	10:06
	The problem seems to be that rsync messes up my modes dispite passing in -a	10:06
gitster	if the issue is because a build modifies the source file, the build infrastructure should be fixed, though.	10:06
	→ kzak_ joined	10:07
gitster	ahh, rsync from windows or something?	10:07
Fullmoon	From a samba share	10:08
	ont-time thing fortunately, i am just restoring a backup	10:09
	→ felipec joined	10:11
felipec	is it safe to remove $GIT_DIR/logs ?	10:11
	giggle keeps showing the original branches, but I've removed them AFAIK	10:13
hliusv561	felipec: If you want to nuke reflogs... And you shouldn't just delete the branch files.	10:13
	felipec: If you use 'git branch -d' (or -D) to nuke a branch, its reflog is deleted at the same time.	10:14
	← pflanze left	10:15
felipec	hliusv561: hmm, I'll try again	10:15
	hliusv561: error: branch 'origin' not found.	10:17
hliusv561	felipec: Branch origin? What git version is that?	10:17
felipec	git version 1.5.3.7	10:19
hliusv561	felipec: Probably only cloned with obsolete version then...	10:19
	→ janm joined	10:19
hliusv561	felipec: And '.git/logs/refs/heads/origin' exists?	10:20
felipec	hliusv561: logs/refs/remotes/origin	10:21
	this worked: git branch -d -r origin/foo	10:21
	error: refs/remotes/origin/HEAD points nowhere!	10:22
hliusv561	felipec: Maybe take backup copy of that file and delete it.	10:23
	→ thiago joined	10:23
felipec	hliusv561: ok, but still those commits are shown in giggle	10:24
hliusv561	felipec: What's 'giggle'?	10:25
felipec	hliusv561: a GTK+ ui	10:25
	hliusv561: and git show is still showing them too... no branch though	10:26
hliusv561	felipec: For getting rid of extra remote branches, theres 'git remote prune <remote-nick>'.	10:27
felipec	hmm, so I have to try again	10:28
	hliusv561: same result	10:32
loswillios	is there a way to show the changes between .24-rc8-git4 and .24-rc8-git5? I'm currently using git-log --since="1 weeks ago" as a workaround	10:32
felipec	hliusv561: I'm doing a "git-filter-branch --env-filter" too	10:32
hliusv561	loswillios: 'git diff 2.6.24-rc8-git4..2.6.24-rc8-git5'?	10:33
	loswillios: Ah, but are there tags for them?	10:34
felipec	hliusv561: there are some refs/original, how do I remove those properly?	10:34
	← IRSeekBot left	10:34
loswillios	hliusv561: nope, only for release candidates. that's my problem	10:34
hliusv561	loswillios: If you could find their commit IDs, then one could get the diff...	10:35
loswillios	hliusv561: yeah. I just want to scroll through the commitlog and see what has changed. thought there was an easy way ;-)	10:35
	→ IRSeekBot joined	10:36
hliusv561	loswillios: Commit message log can be shown using 'git log'.	10:36
	loswillios: The problem seems to be that there's no easy way to know what those versions actually are.	10:37
loswillios	..since there is no tag you mean?	10:38
hliusv561	loswillios: No tag or way to get the commit ID from kernel.org pages...	10:39
	s/or/and/	10:39
loswillios	I see. so I'll stick to git-log --since="1 weeks ago"	10:39
felipec	how can commits exists without any reference?	10:42
hliusv561	loswillios: As example, changes between .24-rc7 and .24-rc8 could be obtained using 'git diff 3ce54450461bad18bbe1f9f5aa3ecd2f8e8d1235..cbd9c883696da72b2b1f03f909dbacc04bbf8b58' even if those tags didn't exist. The problem in that case that one can't even figure out those commit numbers.	10:42
felipec	and not being found by git-lost-found	10:43
hliusv561	felipec: Reflogs? Do those commit contain some sensitive data that you need to dispose as soon as possible?	10:43
imyousuf	I have a repo with history A-B-C-D-E and yesterday with help from gitster, Ilari and aeruder I was able to rewrite the history of master (the only branch in the project) to have B-C-D-E and the new commits have different (more accurately) new IDs (as expected). Now after this when I do cat .git/refs/original/refs/heads/master I get a SHA1 which is not equal to the Commit SHA1 of either B, C, D or E. I also executed git-log --pretty=raw \| grep $	10:45
felipec	hliusv561: no, but I don't want those to be shown in giggle	10:46
hliusv561	imyousuf: Of course it won't match, it's the original E before rewrite.	10:46
imyousuf	FYI: I actually faced problem when I tried to change the history to C-D-E (that is delete B all together)	10:47
thiago	imyousuf: you said that was expected.	10:47
	imyousuf: since it's expected, what's the issue?	10:47
imyousuf	thiago: I said the change in SHA1 was expected	10:47
thiago	imyousuf: so why are you surprised that it is changed?	10:47
imyousuf	but a git will still point to the head that was changed by git, is that ok?	10:47
hliusv561	felipec: Don't know about giggle, but commits not on any branch shouldn't be shown unless one asks for reflog view.	10:47
thiago	imyousuf: "a git" ?	10:47
imyousuf	s/a git/git/	10:47
	→ mithro joined	10:48
thiago	imyousuf: you still have the old ref, yes.	10:48
hliusv561	felipec: Do those commits appear on 'git log --all'?	10:48
imyousuf	yes I still have the old ref thiago	10:48
thiago	imyousuf: you have to delete it if you're satisfied with the rewrite.	10:48
felipec	hliusv561: yes	10:48
imyousuf	thiago: you mean I should delete the refs/original after the rewrite?	10:49
hliusv561	felipec: Maybe fire up 'gitk --all', find those commits and trace back the refs they connect to?	10:49
thiago	imyousuf: after you're satisfied you did not screw up.	10:49
imyousuf	btw after the rewrite there is no way I can revert	10:49
	so I have to be satisfied	10:49
felipec	hliusv561: how do I trace the refs?	10:50
thiago	you can revert by changing your ref back to what it was.	10:50
	that's the whole point of having refs/original in the first place	10:50
imyousuf	I see	10:50
	you mean I can change the SHA1 of the commit if not satisfied	10:51
hliusv561	felipec: Follow the lines starting from those commits upwards until you hit commit that has label.	10:51
thiago	no	10:51
	you can change the SHA-1 that a branch or a ref points to	10:51
	dwmw2_gone → dwmw2_ULN	10:51
hliusv561	felipec: And those lines ending in arrow, click the arrow and it will jump to where the line resumes from 'hyperspace'.	10:51
imyousuf	thiago: I think it will be more clear to me if you mention how I do it	10:52
thiago	imyousuf: git-update-ref, git-reset, git-branch, etc.	10:52
	imyousuf not familiar with git-update-ref checking its manual will get back asap	10:53
felipec	hliusv561: master	10:53
hliusv561	felipec: As check if you did it correctly, you should find the commit in 'git log master'.	10:54
felipec	hliusv561: but those are groupped in original/refs/heads/master, but when I click the arror it goes to master	10:54
	hliusv561: not in git log master	10:54
	← tjafk2 left	10:55
hliusv561	felipec: Aha. 'git filter-branch' saves the original branches under 'refs/original/*'...	10:55
thiago	imyousuf: you can use git-reset to reset to any commit, including those out of your current branch's history.	10:56
felipec	hliusv561: yes, that's why I was asking for a way to properly remove them	10:56
hliusv561	felipec: Maybe 'git branch' can delete those. If not, there's 'git update-ref'.	10:56
felipec	I just did rm -rf refs/original	10:56
hliusv561	felipec: You probably also want to nuke logs/refs/original (if it exists)...	10:57
imyousuf	thiago: let me see if I understand now correctly, I change the history to B-C-D-E and am not satisfied. So I do git-update-ref HEAD "$head" where $head=$(cat .git/refs/original/refs/heads/master) and I get back the old history (A-B-C-D-E)	10:57
thiago	imyousuf: yes.	10:57
	imyousuf: well, no	10:57
	git-update-ref refs/heads/master $head	10:57
imyousuf	great, now I know how to revert the change as well :)	10:57
thiago	or whatever other branch	10:57
	that doesn't change the checkout. You'll have to do a git reset later.	10:58
felipec	hliusv561: I did	10:58
thiago	or, in one step: git reset --hard $head	10:58
hliusv561	thiago: Isn't that equivalent to 'git reset --soft $head' if on master?	10:58
	→ tjafk1 joined	10:58
thiago	hliusv561: it is	10:58
	git-update-ref would be useful if you were touch a branch that is not your checkout	10:58
	if it's the one you have checked out, git-reset is better	10:59
imyousuf	thiago: what do you mean as checkout in this context?	10:59
felipec	hliusv561: git-update-ref -d original/refs/heads/master cf170818f79b5df87dbf33afde873f5cdffd8847	11:00
	error: unable to resolve reference original/refs/heads/master: No such file or directory	11:00
	→ harinath joined	11:01
thiago	imyousuf: type "git branch"	11:01
	imyousuf: the branch that appears with a * is your current check out	11:01
imyousuf	I see master	11:01
thiago	there you go	11:02
hliusv561	felipec: It's 'refs/original/refs/heads/master'. And didn't you delete it already with that directory deletion?	11:02
imyousuf	ok I understood what you meant	11:02
	how do I make the history rewrite permanent?	11:02
felipec	hliusv561: yeah, I did... trying again	11:03
hliusv561	felipec: Does it still show in 'git log --all' or 'gitk --all'?	11:04
	tjafk1 → tjafk2	11:04
felipec	hliusv561: yes	11:04
thiago	imyousuf: it's as permanent when git-filter-branch finishes as any other commit.	11:05
	imyousuf: that is, it's permanent if you don't change it yourself.	11:05
imyousuf	change it means using git-revert or git-update-ref?	11:05
hliusv561	felipec: 'gitk --all' and find the label it connects to maybe.	11:06
felipec	hliusv561: finally!	11:07
	git-update-ref -d did it :)	11:07
	hliusv561: thanks a lot	11:07
	is this channel archived?	11:07
imyousuf	thiago: The whole point why I wanted to rewrite the history is to save space as this rewrite will take place in a repo containing binary files only, so after some point of time the oldest history will be pointless (in my case) and thus I want to delete it not only just to delete the history but also to save the disk space consumed due to it	11:07
hliusv561	felipec: Yes, see the topic string.	11:07
	felipec: ... \| Channel log http://colabti.de/irclogger/irclogger_log/git \| ...	11:08
imyousuf	thiago: another question - is it safe if I do rm -rf .git/refs/original/refs? or what is the proper way for it?	11:08
thiago	imyousuf: remove the original refs. Or tell git-filter-branch not to create them	11:08
	imyousuf: the proper way is git-update-ref -d	11:08
felipec	imyousuf: that's what hliusv561 just told me :)	11:08
imyousuf	ok :)	11:09
	yeah I see the -d option in your case felipec :)	11:09
	thanks a lot thiago	11:10
felipec	any proper way to reset the reflog?	11:10
thiago	git gc will expire old reflog entries	11:10
	git update-ref -d deletes the reflog when it deletes the ref as well	11:11
felipec	thiago: yeah, but I just did a massive reorganization of the repo, so I guess the reflog sould be empty	11:12
mugwump	thiago: hey, I noticed a lot of gitweb traffic from sweden for kde.git	11:12
	trying to pull tarballs etc	11:12
	however gitweb isn't designed well enough for that to work for the size of tarballs people were trying to pull	11:12
	it will end up buffering the tarball	11:13
thiago	felipec: how did you reorganise?	11:13
	← G0SUB_ left	11:13
felipec	thiago: I removed a remote origin and filtered the author names	11:14
	→ wingo joined	11:22
wingo	good day	11:22
	let's say i committed something yesterday, but have no idea where it is now, in what branch or what	11:23
	← garrin_ left	11:23
wingo	is there a way to grep the object database for a string, or a committer?	11:23
mugwump	lots of ways, see git-rev-list man page	11:23
	and git-log	11:23
wingo	tx	11:24
mugwump	also gitk has a search box	11:24
thiago	wingo: git reflog will tell you what you have been doing	11:24
mugwump	eg, gitk --all --since="2 days ago"	11:24
thiago	wingo: but in general it's a good idea to remember what you're doing.	11:24
felipec	wingo: hi there! it's good that you are trying git :)	11:24
thiago	if you have so many branches that you don't remember what you committed and where, you're over your head and you should take a step back.	11:24
wingo	felipec: it's kicking ass :-)	11:24
felipec	wingo: git > everything out there ;)	11:25
	→ dave_ joined	11:25
wingo	thiago: yep, but i'm working with others that need to rebase their entire trees with filter-branch, it's a bit complicated :/	11:25
	tx all	11:25
mugwump	rebasing with filter-branch ... ooo, that's the wrong two words to use together	11:26
wingo	well	11:26
mugwump	so long as you know what you mean :)	11:26
wingo	i guess it's not precise	11:26
	i'm pretty sure that i do, although in this world there is always room for doubt :)	11:26
felipec	wingo: git-cvs?	11:26
mugwump	rebase means re-applying patches	11:26
	filter-branch is reorganising trees	11:27
wingo	felipec: no, a number of svn changes losing history, then a git import that threw away svn history	11:27
	complicated ;)	11:27
	the others only have history back to december, when they need to have it back a couple years more	11:28
mugwump	you found how to graft and then use filter-branch?	11:28
wingo	i actually didn't graft, i did it manually with filter-branch	11:29
	→ G0SUB_ joined	11:29
mugwump	yeah, I started out like that too	11:29
	grafting is a handy way to preview a filter-branch though	11:29
thiago	graft + filter-branch is much, much easier	11:30
mugwump	assuming that what you're doing is that simple	11:30
wingo	i couldn't actually find any info about grafting	11:30
	anything link handy?	11:30
mugwump	grep in Documentation/	11:31
wingo	aight	11:33
	← imyousuf left	11:38
felipec	what do you use to generate ChangeLog files from the git log?	11:38
	thiago uses nothing	11:39
felipec	thiago: do you touch any ChangeLog?	11:40
thiago	nope	11:40
	← jmspeex left	11:40
	→ cedricv joined	11:43
	dwmw2_ULN → dwmw2_gone	11:48
	→ captnmark_ joined	11:56
	← captnmark left	11:56
	→ lcapitulino joined	11:58
	→ alturiak joined	12:05
	← thiago left	12:09
	→ kukks joined	12:13
	→ bentob0x joined	12:31
pasky	Lazy Clone of http://git.or.cz/gitwiki/SoC2007Ideas never got implemented, right?	12:34
	metze → metze_food	12:35
	← dave_ left	12:40
vbgunz	anyone know why, when in a current project directory, the following does not completely destroy it? rm -fr ./* ... that command always leaves .git behind... explicityly deleting the parent works but I am just trying to empty the directory :/	12:45
Cerebon	dot directories are not globbed with *	12:46
vbgunz	Cerebon: I thought they were	12:48
Cerebon	I know ksh and zsh don't for sure, pretty sure bash doesn't	12:49
vbgunz	rm -frd ./* seemed to work, now I know	12:49
	Cerebon: thanks	12:49
Cerebon	I always do * .* if I want to empty	12:50
	there was a fun glob way of doing it in zsh that I can't find in my history	12:50
vbgunz	I've been meaning to learn gnu/linux a whole lot better. I cannot find a book that can teach a 6th grader and only teach the most vital command line programs found in at least 90% of all distros... probably why, it's the longest stale item on my todo :/	12:51
Cerebon	heh, go figure, setopt dotglob in zsh and it works like you want :)	12:52
	zsh: the emacs of shells	12:52
vbgunz	I think I have bash... last thing I want to do is mess with it, godforbid I lose it, I'll die :)	12:53
	→ kendy_ joined	12:53
kendy_	Hi there!	12:53
Cerebon	anyway off to work	12:54
vbgunz	have fun!	12:54
Cerebon	vbgunz: it is better than vi, I mean bash :)	12:54
kendy_	Looking at http://git.or.cz/gitwiki/SoC2007Ideas, was some work on the 'lazy clone' done in the meantime, please?	12:54
	Or is it still in the 'would be nice to have that, but noone has implemented it yet' state?	12:55
Cerebon	bash is not the shell you are looking for...	12:55
	← Cerebon left	12:55
vbgunz	I use vim for editing any text files and use yakuake for everything else...	12:55
Debolaz	tcsh > bash # Awesome history search makes a difference	13:05
	metze_food → metze	13:16
	→ cmarcelo joined	13:17
	→ janm_ joined	13:17
	→ Ryback_ joined	13:20
	→ Yuuhi joined	13:29
	← janm left	13:29
	← ironfroggy left	13:32
	← marcel left	13:35
lamont	sigh. so if I Delete a branch and haven't done a gc/pack/whatever yet, is there a way to go find the commit and recover the branch?	13:37
	s/commit/head/	13:37
	→ hein2 joined	13:38
	← Sho_ left	13:38
	→ jmspeex joined	13:39
	← mithro left	13:39
vmiklos	did you use git branch -d or git branch -D?	13:39
lamont	-D	13:39
vmiklos	that's the worse case.	13:39
lamont	yeah.	13:39
Debolaz	lamont: git-fsck should be able to recover it afaik.	13:39
lamont	-d would be trivial	13:39
	hein2 → Sho_	13:39
vmiklos	ah, reflog of HEAD?	13:40
lamont	Debolaz: so git-fsck dumps a bunch of dangling blobs and a pair of trees...	13:40
vmiklos	don't forget that not only branches have reflogs but HEAD itself, too :)	13:40
lamont	how do I see what a dangling blob is?	13:40
	rather, what's in the blob	13:40
vmiklos	git show <sha1>	13:40
lamont	thanks	13:41
	→ cworth joined	13:41
lamont	on the bright side, it wasn't that complex of a branch (just a rebase with a few conflicts)	13:43
felipec	git push without arguments automatically pushes to origin?	13:43
	lamont decides it'll be easier to just redo the branch-n-rebase	13:43
	← octopod left	13:44
	→ Cerebon joined	13:48
	→ Pengwn joined	13:53
	→ docelic joined	14:12
	→ mjc_ joined	14:13
	← bentob0x left	14:16
	← dysinger left	14:18
	→ hurikhan\|Work joined	14:18
	→ aroben joined	14:24
	→ dreivier_ joined	14:24
	← kanru left	14:27
	→ kylem joined	14:29
	← spuk- left	14:35
	→ dimsuz joined	14:36
dimsuz	is this the right channel to ask git-svn questions?	14:37
	→ ironfroggy joined	14:37
Debolaz	dimsuz: Sure. :)	14:38
	→ bdiego joined	14:38
dimsuz	cool. my git commit added some new files to git-svn generated repo. Now I'm doing "git-svn dcommit" and svn's pre-commit hook fails telling me that props eol-style and mime-type are not set. what to do? :)	14:39
	→ eetfunk joined	14:41
	→ ckm joined	14:41
	→ mjc__ joined	14:41
eetfunk	Hi all. Has the way of doing git --bare init changed? I get this error: fatal: Out of memory? mmap failed: Invalid argument	14:42
	← kylem_ left	14:45
hliusv561	eetfunk: That's a bug.	14:51
felipec	how do I do a git commit --amend for a specify comit (older)?	14:51
eetfunk	hliusv561: is there a way to bare it after init?	14:51
jengelh	eetfunk: just the .git directory is a bare.	14:52
	→ KipIngram joined	14:52
hliusv561	eetfunk: Sure. Grab the .git dir created (rename it) and set the bare setting in config file...	14:52
cehteh	felipec: git revert	14:52
eetfunk	hliusv561: cool! thanks.	14:52
hliusv561	eetfunk: setting 'bare' in section '[core]'.	14:54
	eetfunk: What git version?	14:54
	← dreivier_ left	14:55
eetfunk	1.5.2.5 (ubuntu) and 1.5.3.6 (osx)	14:55
hliusv561	eetfunk: Did you get that error on both?	14:55
	← mjc_ left	14:55
eetfunk	yes. same on both	14:55
	i doubt my mac is out of ram. i have 4 gigs :)	14:56
	eetfunk trying to clone, i get this: fatal: no matching remote head	14:56
eetfunk	what could it be?	14:56
vbgunz	what is this called? master..experimental, master...experimental?	14:59
	mainly the 2 dot 3 dot syntax	14:59
	I know how they differ but how can I get more info on them and more if they exists?	14:59
	I think I know, man git-rev-parse, found a hint in man git-log	15:01
	cbrake_away → cbrake	15:03
	→ kanru joined	15:05
	→ ivazquez joined	15:11
	→ spuk- joined	15:11
	→ EmilMedve joined	15:12
ivazquez	Okay, I messed up a commit and a push. Several files were erased and I tried a fresh clone and reset and revert but couldn't manage it since my git-fu isn't strong yet. What can I do to recover those files, short of reconstructing them from commit notifications?	15:12
	→ GyrosGeier joined	15:15
	→ kevwil joined	15:16
	→ kumbayo joined	15:18
cehteh	make ... 30 sec	15:21
	make doc ... 5 min	15:21
	:/	15:21
vbgunz	ivazquez: if you know what commit actually contained those files you need you can try checking them out e.g., git checkout tree-ish /path/to/file ... you may want to try it on a branch?	15:22
jengelh	make -j2147483648	15:22
	← ckm left	15:22
ivazquez	vbgunz: Excellent, perfect, MANY thanks.	15:24
cehteh	jengelh: there is no distcc for ascidoc and xmlfoo	15:24
vbgunz	ivazquez: glad that helped :)	15:24
ivazquez	Now I can git-add and do a commit and push for it?	15:24
vbgunz	yeah, should work	15:25
ivazquez	Looks like I don't need an add.	15:25
jengelh	cehteh: but you sure have a few extra cores in the local machine	15:25
cehteh	hey this is my laptop	15:26
	← EmilMedve left	15:28
vbgunz	I have an interesting issue. I had a file staged in the index and then checkedout an older version from HEAD. it happened to overwrite the index and left that latest file as a dangling blob... how do I checkout that dangling blob? in other words, how do I restore that version?	15:33
	well, I did this, git cat-file -p 513fe > file.txt ... nothing wrong with that? I did check -p first to make sure it looked good	15:36
	→ madewokherd joined	15:37
vbgunz	re-adding the file.txt, git fsck shows no dangling commits...	15:37
cehteh	when i 'edit' in an interactive merge and not commit/amend the edits i do they will be squashed with the next commit right?	15:43
	→ EmilMedve joined	15:49
	→ ckm joined	15:50
	← njero left	15:57
	→ jelmer joined	15:57
	→ careo joined	16:01
	← FunkeeMonk left	16:02
	← Pengwn left	16:05
jelmer	'evening	16:08
	I'm looking for the equivalant of "svn export" with git. Is there such a thing?	16:08
semi_	rm -rf .git	16:08
jelmer	I'd like to run this on a remote git repository I don't have a copy of locally	16:09
	→ priidu_ joined	16:09
semi_	then git-clone first	16:09
jelmer	That's going to be very inefficient bandwidth-wise, because I can't keep the local copy around	16:10
	there's no other way?	16:10
Mikachu	there's the --depth option to git-clone	16:11
jelmer	Mikachu, ah, thanks	16:11
	← mjc__ left	16:12
	← cmarcelo left	16:12
	← sverrej left	16:12
	← zrajm_ left	16:12
	← aazar left	16:12
	← philc left	16:12
	← kelvie_ left	16:12
	← kendy_ left	16:12
	← Catfish left	16:12
	← bsmith left	16:12
	← shawn left	16:12
	← clee left	16:12
	← xkr47 left	16:12
	← adante left	16:12
	← dsop left	16:12
	← chrizz- left	16:12
	← simmel left	16:12
	← moh left	16:12
	← floppyears left	16:12
	← pete__c left	16:12
	← CIA-6 left	16:12
	← Sho_ left	16:12
	← robfitz left	16:12
	← LotR left	16:12
	← Tux32- left	16:12
	← eetfunk left	16:14
	→ Sho_ joined	16:15
	→ robfitz joined	16:15
	→ LotR joined	16:15
	→ Tux32- joined	16:15
	→ mjc__ joined	16:15
	→ cmarcelo joined	16:15
	→ kendy_ joined	16:15
	→ sverrej joined	16:15
	→ zrajm_ joined	16:15
	→ aazar joined	16:15
	→ philc joined	16:15
	→ kelvie_ joined	16:15
	→ Catfish joined	16:15
	→ bsmith joined	16:15
	→ shawn joined	16:15
	→ clee joined	16:15
	→ xkr47 joined	16:15
	→ adante joined	16:15
	→ floppyears joined	16:15
	→ pete__c joined	16:15
	→ chrizz- joined	16:15
	→ CIA-6 joined	16:15
	→ moh joined	16:15
	→ simmel joined	16:15
	→ dsop joined	16:15
	← johan-s left	16:16
	→ langenberg joined	16:17
langenberg	Just a question, I haven't used git much, but am I right that here are two kinds of commits. Local commits and remote commits?	16:17
	So if you want to work really thight on a project with someone, you can better fetch from eachoter's repository then the remote's ?	16:18
Debolaz	langenberg: No, a commit is a commit. But you can push commits remotely, or pull them from a remote.	16:18
	langenberg: There's no neccesary central server in git. You can have one, but it's not required by design.	16:19
	→ octopod joined	16:19
langenberg	So you can do like, 5 commits (local) and then push those remotely?	16:19
adante	hi, i just tried to run git status, it said something like 'fatal: couldn't write to index'	16:20
langenberg	Will that cause for a burst of 5 commits. (like git-svn dcommit)	16:20
adante	now when i try to checkout it errors out	16:20
Debolaz	langenberg: Yes. You can push that directly to your peers repository (If he allows it), or he can pull them from you. Or you can both work against a common central repository.	16:20
adante	saying the index.lock exists	16:20
	there is nothing else accessing my repository (checked with lsof)	16:20
Debolaz	langenberg: Yes, when you push, you push all commits on your branch.	16:21
adante	is it safe to delete index.lock and continue? or what is the normal procedure in this case	16:21
siprbaum	adante: AFAIC it is safe IFF no other process is accessing that repo	16:21
adante	alright thanks	16:21
langenberg	Debolaz: I'm talking about a 'corporate' centralized situation here. At this point we just a central SVN repository. (and I'm using git-svn :) )	16:22
Debolaz	langenberg: Or to be more specific: It sets the remote branch to be at a certain point, and then transfers all objects neccesary to reach this point.	16:22
langenberg	Sometimes I want to have a really tight loop with my colleague, like for doing BDD together. How much commands would it need to see the work from eachother. With svn it's just: svn commit and svn update.	16:25
Debolaz	langenberg: Just publish your repository and he can do "git fetch" and he'll have your latest updates if he tracks any branches in your repository.	16:26
	ckm → stick	16:26
langenberg	Is that with a centralized or approach or doesn't that matter?	16:26
Debolaz	langenberg: That's a decentralized approach, you publish your own repo, not the centralized one. You can easily go through the centralized repository as well, but it's not neccesary since you can work directly against each other's repositories.	16:27
langenberg	That's really cool.	16:27
	→ nbrunson joined	16:28
langenberg	If I want to share my git repository, do I need to give someone a SSH login to my machine?	16:28
semi_	no	16:29
	if you just want to give read access	16:29
Debolaz	langenberg: The trick is to not try to memorize a bunch of commands, but to get a deeper understanding for how git works beneath the hood. Because ironically, git is based on a few very, very simple ideas, it just seems complicated because the commands on top can be somewhat weird.	16:29
langenberg	ah	16:30
	→ ilogger2_ joined	16:33
jengelh	git diff HEAD does what exactly?	16:33
	it seems to compare HEAD..workingcopy	16:34
	→ bobesponja joined	16:35
	→ RobotDeathSquad joined	16:39
	→ njero joined	16:41
	→ lamont joined	16:42
	→ tvachon joined	16:44
careo	Debolaz: so true. now that I get how the repository works (thank you gitk), git seems sooo much simpler than anything else	16:48
	So, I genuinely don't want to start any sort of flamefest, but I'm curious if anyone know of a good resource describing just how git and mercurial differ. specifically the things git has that hg doesn't. bonus points if it mentions the repository formats.	16:49
	s/know/knows	16:49
	→ tongueroo joined	16:53
Debolaz	careo: I've actually been thinking about writing up a comparison. All I've found are horribly outdated or don't give any useful information.	16:54
	→ tongueroo joined	16:54
careo	Debolaz: cool. I've been putting things up on our wiki internally (we're going to ditch svn here in a few months, and I'm hoping to steer them in the right direction ;) )	16:54
Debolaz	I still often recommend mercurial based on my perception that it's a lot easier to get started with, but I don't know how true that is anymore. It used to certainly be the case that mercurial was a lot more elegant and git was a lot better at branching.	16:55
careo	yeah. it's a great program, far better than svn. it's hard to explain to people that nave never used a distributed system why git is better than mercurial	16:58
vmiklos	and if they used?	16:58
careo	have, not nave. to early to type	16:59
vmiklos	i don't know people who used both for a long time - so they could make a not-so-subjective comparision	16:59
	→ dreivier_ joined	17:00
careo	vmiklos: yeah. that's what I'm finding.	17:00
vmiklos	probably because both is highly usable, so if you read enought docs, etc - so you really learn it, then probably you won't learn the other :)	17:02
	→ jmspeex joined	17:02
vmiklos	enough*	17:02
	→ tongueroo joined	17:02
careo	heh. true, true. I don't really feel very honest just skimming the mercurial docs. of course, some things are obvious. like the section on cherry-picking that says something to the effect of "need to ask someone how this actually works"	17:03
	→ chris2 joined	17:04
	→ njero joined	17:04
vmiklos	that's easy. if something not obvious from the git docs, then after you figured out, you have to send a patch to the docs :)	17:05
	→ vfe4t67yg joined	17:08
vfe4t67yg	ck BEket	17:08
	→ chris2 joined	17:18
	→ thiago joined	17:25
	→ timocratic joined	17:37
	→ raalkml joined	17:39
	→ kevwil joined	17:42
	→ Ingmar joined	17:43
	→ p4tux joined	17:52
	→ njero joined	17:53
	→ ilogger2 joined	17:53
	→ harinath joined	17:56
Catfish	Any suggestions on finding the most recent tag?	18:00
	Obviously this is complex when it comes to multiple branches, but just the basic case would do for me...	18:01
cehteh	Catfish: from C or shell?	18:02
Catfish	from shell	18:02
cehteh	git for-each-ref	18:02
	→ drizzd joined	18:03
Catfish	cehteh: Perfect, thanks	18:04
	→ Phoop joined	18:09
	→ robinr joined	18:12
RobotDeathSquad	how do you change the svn repo url that git svn is using?	18:18
Randal	"man git-svn"	18:18
	Oh. Hmm. I guess it's not clear	18:19
RobotDeathSquad	:)	18:19
Randal	Wait - See "CONFIGURATION"	18:19
	it's right there	18:19
	see "url"?	18:20
RobotDeathSquad	yeah, thanks	18:20
Randal	So I wasn't lying. :)	18:20
siprbaum	Randal: are you sure this works? git-svn stores the url in the commits, too.	18:21
	but i have to admit I would suggest this, too.	18:21
	at least I'd try it as a first attempt :-)	18:21
Randal	well - the real key was that I should have turned awayas soon as I saw it was a git-svn question	18:22
siprbaum	hehehe	18:22
	→ glguy joined	18:22
glguy	Why does git-mergetool show me the attempted merge result	18:22
	instead of letting me solve the three-way merge?	18:22
	→ tvachon joined	18:22
	→ alley_cat joined	18:22
siprbaum	?	18:22
glguy	so I started the git-mergetool to do the merge by hand because it wasn't obvious to git how to do it	18:23
siprbaum	it should show you tree files, the base, your version and their version	18:23
glguy	but it is showing me all the >>>>>>>>>>>>>>>>>>	18:23
	====== <<<<<<<<<<<	18:23
	cruft	18:23
siprbaum	what tool are you using (meld, emacs, vimdiff ....)	18:23
glguy	meld	18:23
	pristine local branch ----- ugly merge attempt ------ pristine right remote branch	18:24
siprbaum	not sure, but could it be that meld shows you the conflicts there and you have to select one ore the other from the files left and right?	18:24
glguy	no, the problem is that git-mergetool is not passing in a clean base version	18:24
	but a half-merged file	18:24
siprbaum	urgh. this sounds like a bug. AFAIR this never happend to me here (but I seldom use git mergetool and/or meld)	18:25
	you could try to redo the merge by doing git reset --hard (THIS WILL LOSE ALL YOUR LOCAL WORK; INCLUDING ALREADY MERGED FILES)	18:26
	and then git merge the_other_branch	18:26
	→ dreivier_ joined	18:29
	→ tongueroo_ joined	18:29
siprbaum	glguy: propably worth to try a different merge-tool if it still persists (just to check if this is an error in meld)	18:30
	→ stillLotR joined	18:31
Bartman007	Contexto, ¿por que estas tan enojado?	18:33
	doh, wrong window.	18:33
	← Phoop left	18:35
reuss	how do i get git to ignore every other file except secret.txt in a directory?	18:44
	→ stillLot1 joined	18:45
Randal	"every other file", like all the odd numbred files?	18:48
	check out "man gitignore"	18:49
	broonie suspects he means "all files other than..."	18:49
Randal	but I think you want "*" on one line, and !secret.txt on the next line	18:49
reuss	yeah, like broonie :)	18:49
Randal	in your .gitignore file	18:50
reuss	so it would be /my/path/!secret.txt ? ..	18:50
Randal	No	18:50
	in your .gitignore file in THAT directory	18:50
	and be sure to add .gitignore as well. :)	18:50
reuss	oh yeah - good point and the .gitignore should be added to the git repo :)	18:51
	→ tarbo joined	18:51
	→ LotR joined	18:51
Randal	well - once it's added, it doesn't matter if .gitignore matches it	18:51
	→ manithree joined	18:51
	→ sgrimm joined	18:53
	→ harinat1 joined	18:57
	← stillLot1 left	18:57
manithree	Besides contrib/continuous are there any CI systems that work smoothly with git?	19:00
siprbaum	what is a CI system?	19:01
	thiago was expecting someone else to answer and clarify that	19:02
tpope	continuous integration	19:03
	i.e., run your tests on each commit and/or push	19:04
siprbaum	Ah. Ok.	19:04
	← stillLotR left	19:04
	→ nud joined	19:05
	→ KirinDave joined	19:09
	→ harinat2 joined	19:14
	← harinath left	19:14
	→ ferdy joined	19:16
	→ lcapitulino joined	19:20
	→ pmezard joined	19:20
	→ devogon_ joined	19:28
	← sgrimm left	19:30
	← harinat1 left	19:30
njero	hi #git, I have a git repo and certain files are taking a really long time. I am guessing these are files that shouldnt be in the repo... only problem is I can't seem to find the large files... is there a command to rank files by size recursively	19:33
vmiklos	i don't think that's a task of an scm :)	19:34
Randal	you want the largest file in the current worktree?	19:34
vmiklos	you can write a short script to do so	19:34
njero	vmiklos: fair enough	19:35
	→ sgrimm joined	19:38
	← sgrimm left	19:39
	→ cappslocke joined	19:44
cappslocke	I've got a question guys	19:44
	I hear there are plugins to make SVN and Photoshop play nicely. Is there any such thing for git?	19:44
vmiklos	i don't think so	19:46
	how do you diff images?	19:46
	(not counting the binary diff)	19:47
cappslocke	shrug, it was off-handedly mentioned in a meeting. Im trying to determine whether we should move to SVN or git.	19:49
	I personally like the idea of git, and we're mostly a web-application development company with a bit of design work	19:49
	→ kreuter joined	19:50
vmiklos	i don't say git is worse than svn when you store images, but much of the extra functionality (merges, branches, etc) requires text content which won't work for images	19:51
	→ Fullmoon joined	19:53
zeffseries	jeffpc@freyr:~/hvf$ git-branch -r \| grep origin	19:55
	origin/master	19:55
	origin/todo	19:55
	jeffpc@freyr:~/hvf$ ls .git/refs/remotes/origin*	19:55
	ls: .git/refs/remotes/origin*: No such file or directory	19:55
	how is that possible?	19:55
	I don't have origin defined anymore to begin with	19:55
	(I don't want origin)	19:56
cappslocke	I agree, but you often never need to merge/branch graphics in that way. Usually only one or two designers are working on a photoshop file at a time, and will often just trade the files back and forth	19:56
	→ Phoop joined	19:57
	← Phoop left	19:58
thiago	zeffseries: packed refs	20:01
vmiklos	git remote rm origin is your friend:)	20:01
zeffseries	ah :)	20:02
vmiklos	or git update-ref -d for each remote ref	20:02
	zeffseries makes a mental note...don't edit the config file by hand	20:02
	Debolaz explicitly disabled refpacking.	20:02
zeffseries	well, at least the remote stuff	20:02
vmiklos	you can - if you know what you're doing:)	20:02
	→ Phoop joined	20:03
zeffseries	vmiklos: I never said I was going to follow my mental note :)	20:03
vmiklos	:]	20:03
blake	ssh zeverly	20:04
	oops, crap. :)	20:04
	that's embarassing	20:04
zeffseries	Password for blake@zeverly:	20:05
	:)	20:05
blake	blah	20:05
	:p	20:05
	This is a sign i'm either too tired or too sick to be working	20:05
	← Phoop left	20:06
	→ Phoop joined	20:06
	→ toxx_ joined	20:09
	→ raalkml joined	20:17
	→ timocratic joined	20:17
	← timocratic left	20:18
	→ khwu- joined	20:19
	← khwu- left	20:19
	→ khwu- joined	20:20
	→ DrNick joined	20:24
	← lcapitulino left	20:24
	→ sverrej joined	20:25
kreuter	are git newbie questions appropriate for this channel?	20:34
Debolaz	Well, you risk getting roasted but no pain no gain. :-)	20:35
gitster	nothing but newbie questions is appropriate ;-)	20:35
kreuter	very well	20:35
Debolaz	Consider burn marks a badge of honor.	20:35
kreuter	so, I'm on day number 4 or 5 in my transition to git, having used CVS, SVN.	20:35
	I'm starting to get the hang of it	20:35
	my question is this: my CVS/SVN workflow is usually to make several working copies, and to do parallel development in the separate working copies. in particular, this permits me to have several different builds on disk, in different places	20:36
	how can I do something like this with git?	20:36
semi_	branches	20:36
DrNick	no, just clone multiple times	20:37
semi_	DrNick: what's the advantage?	20:37
DrNick	he wants multiple working copies	20:37
Debolaz	semi_: Having several branches don't give you several working copies.	20:37
gitster	or contrib/workdir/	20:37
semi_	isn't that only a subgoal?	20:37
kreuter	right, I've figured out that branches don't do the trick, since there's just one file hierarchy	20:38
	which cloning do I want to do? a git clone -l -s ... ?	20:38
PerlJam	kreuter: why is it you like to have several different builds on disk at once?	20:39
DrNick	PerlJam: does that matter?	20:39
kreuter	PerlJam: the things I work on tend to have compile-time configurations, and when I encounter a bug in one configuration, I want to see if I can reproduce the bug under other configurations.	20:40
PerlJam	DrNick: not to his particular problem, but I'm curious	20:40
vmiklos	kreuter: avoid -s if possible. and -l is the default	20:40
	(if possible == unless you know what you're doing)	20:41
kreuter	vmiklos: I don't!	20:41
	can you explain the difference? in my head, the word "shared" seems like what I'm looking for.	20:41
semi_	kreuter: I do that too, but I just have several out of source build trees in one clone	20:42
raalkml	kreuter: it is just matter of implementation	20:42
	-s uses symlinks	20:42
vmiklos	i don't think so	20:42
semi_	but sometimes I clone the project under /tmp for testing something	20:42
raalkml	-l hardlings	20:42
	hardlinks	20:42
vmiklos	-s uses alternates	20:42
raalkml	:)	20:42
	oh, right	20:42
kreuter	semi_: what does "out of source build trees" mean?	20:43
raalkml	these. Has problems unless you're very careful	20:43
	← tvachon left	20:43
vmiklos	kreuter: if you clone foo.git to bar.git using -s, then you might want to keep objects in foo.git even if nobody refers to them (as you refer to them only in bar.git). then if you use git gc --prune in foo.git you can have problems	20:44
semi_	kreuter: in my repo, there's the src/ dir tracked by git, and then build/foo , build/bar etc that are built with different configure options	20:45
	and build is ignored by git	20:45
	in that repo	20:45
kreuter	semi_: ah, okay.	20:45
	vmiklos: I'm reading your last sentence over and over, and I don't really understand it.	20:47
vmiklos	git gc --prune searches for unreferenced objects and removes them.	20:48
kreuter	okay, I get that.	20:48
	what I don't get is how a clone of a repository relates to the repository.	20:48
	I guess "how a repository relates to its origin"?	20:49
raalkml	kreuter: the cloned repo does not have own objects, it references some from the source repo. You get into problems removing objects in source repo. git-prune removes objects	20:49
vmiklos	for example you have two branches in foo.git. then you clone, and you remove the branch in foo.git	20:49
	then you run git gc --prune. now you would think that the branch is still in bar.git but in fact it isn't	20:50
kreuter	oh. the branch will still be referenced in bar.git?	20:50
Mikachu	semi_: the problem with that is when you checkout another branch, the makesystem usually detects that and refuses to do make install without rebuilding	20:50
	→ nbrunson joined	20:50
	→ kristoffer_ joined	20:50
kreuter	vmiklos: well, you did predict what I would think :)	20:50
semi_	Mikachu: hmm, true	20:51
kreuter	vmiklos: now, which kind of cloning are we talking about? "shared" or normal?	20:51
kristoffer_	I've added a new git repository in same folder as my other repositories. Since I export everything through gitweb I expected it to appear once I added it to list file, but nada.	20:51
	Ive set exactly same rights, but still doesn't happen	20:51
Mikachu	given two commit sha1s, is there an easy way to tell if they have a .. or ... between them?	20:51
kristoffer_	anything inside .git that I need to set? I tried to restart apache but doesnt help	20:51
Mikachu	i guess i could see if a == mergebase(a, b)	20:52
raalkml	Mikachu: _how_ are you given them?	20:52
Mikachu	in the update hook	20:52
	i don't want to spam out stuff when i rebase a branch	20:52
vmiklos	kreuter: just use the default, don't use -s.	20:52
kreuter	vmiklos: okay!	20:53
raalkml	Mikachu: git rev-list sha1...sha2 \|wc -l ?	20:53
Mikachu	how is that supposed to help me?	20:54
raalkml	depends on what you want	20:55
Mikachu	i want to know if sha1 is reachable from sha2	20:55
raalkml	git rev-list sha1..sha2 \|head -n1\|grep ^	20:56
	git rev-list sha1..sha2 \|head -n1\|grep ^ >/dev/null	20:56
	hmm, that wont work if sha1 == sha2	20:57
Mikachu	or in any other case	20:57
	→ sgrimm joined	20:57
kreuter	so, when I clone a local repository, the two repositories will have the same set of objects, right?	20:57
kristoffer_	Anyone got an idea about my gitweb problem? What do I need to change if I add another repository?	20:58
Mikachu	initially, yes	20:58
kreuter	do commits to one repository appear in the other repository?	20:58
Mikachu	you have to push or fetch them over	20:58
kreuter	oh	20:58
Mikachu	once you do a commit, they don't have the same set anymore	20:59
kreuter	hm	20:59
	well, I'll try it.	20:59
	robinr s mailbox i glowing from the incoming flames	21:00
raalkml	well, it works if sha1 _is_ reachable... git rev-list sha2\|grep $(git rev-parse sha1)\|head -n1 but that is too brute-force	21:01
Mikachu	i'll just check if sha1 == $(git merge-base sha1 sha2)	21:02
kreuter	hm. only one branch appears in the newly cloned repository.	21:02
robinr	kreuter: try git branch -r	21:03
kreuter	okay. how can I change branches in the new repository?	21:04
robinr	try reading the tutorial	21:04
kreuter	I've tried a half dozen tutorials. :(	21:04
raalkml	git checkout branch	21:05
kreuter	there's so much to git, it's hard for me to grasp	21:05
raalkml	git checkout -b new-branch old-branch (to create a new branch and switch to it)	21:05
kreuter	doesn't that create a new branch?	21:06
robinr	kreuter: because there is so much I suggest a tutorial walkthrough instead of random testing	21:06
raalkml	yes. The second, with -b	21:06
kreuter	what I think I want to do is to switch to a branch that's in the origin repository.	21:06
raalkml	you can't. It is in the origin repo	21:07
kreuter	git checkout origin/branch ?	21:07
	oh	21:07
	okay	21:07
raalkml	but you can base your branch on it	21:07
	git checkout -b my origin/master	21:07
kreuter	oh, I suppose that makes sense.	21:08
raalkml	yeah, we think so too	21:08
kreuter	so why does git clone produce a new repository with only the active branch? couldn't it produce a new repository with all the origin's branches?	21:09
raalkml	it does	21:09
	git branch -r	21:09
	→ HG` joined	21:10
kreuter	but you also have to "git checkout -b" to create new branches...?	21:10
mugwump	new local branches	21:10
	read the user manual	21:10
raalkml	kreuter: well, you can work without them. But you wont like it	21:11
	→ robin_ joined	21:16
	→ gebi_ joined	21:16
	→ kylem_ joined	21:20
	→ mjc_ joined	21:20
	→ \ask_ joined	21:23
	→ pombreda joined	21:27
	→ riddochc joined	21:29
riddochc	Hi, folks.	21:29
	← robinr left	21:30
riddochc	I see via kernel planet that wincent colaiuta is reading some interesting old 2005 flamewars about git.	21:31
	And that history smiles on linus. ;)	21:32
kreuter	what's the recommended way to install git on Windows?	21:33
Mikachu	i don't think it is	21:33
raalkml	don't :)	21:33
kreuter	(there seem to be a few different porting strategies.)	21:34
	:(	21:34
raalkml	yeah. Cygwin and MSYS	21:34
kreuter	well, what's the recommended way of working around the absence of git on Windows then? :)	21:34
riddochc	Sounds like msys.git is pretty good.	21:34
raalkml	try MSYS	21:34
kreuter	ok	21:34
	→ madewokherd joined	21:35
raalkml	unless you have to work with cygwin	21:35
	they conflict	21:35
kreuter	do they? I thought they didn't. I have cygwin installed already. does git on cygwin work reliably (I don't care about performance)	21:35
	?	21:35
robin_	yes	21:35
kreuter	okay, thank you.	21:36
raalkml	yes	21:36
robin_	at least when compiled with NO_MMAP and NO_PREAD	21:36
	you must only work on LF-mounted directories btw	21:37
kreuter	alright	21:38
raalkml	aka binmode	21:38
robin_	yes	21:38
kreuter	isn't that the default for cygwin?	21:38
	→ Cerebon joined	21:38
robin_	yes, but that's not what everybody uses	21:38
kreuter	oh. I think I probably used the defaults	21:38
	(I'm not really a Windows user, I just play one on IRC.)	21:39
raalkml	can be dangerous, around here	21:39
robin_	:)	21:39
kreuter	I don't like it any more than anybody else. I like being the guy who broke things on Windows less, however.	21:39
raalkml	right answer	21:40
	→ PugMajere joined	21:40
	← glguy left	21:40
riddochc	I have a teeny Windows partition just so I can use Netflix's 'Watch Instantly' stuff. Because my system isn't fast enough for qemu to keep up with it.	21:41
robin_	yeah, everybody has an excuse...	21:42
	← robin_ left	21:42
raalkml	:)	21:42
	→ mithro joined	21:45
riddochc	Feh. It's not like I'd ever trust it with anything important.	21:46
	Gotta have some standards, after all.	21:47
	→ robinr joined	21:47
	← robinr left	21:50
	→ johan-s joined	21:50
	→ robinr joined	21:50
	→ vbgunz joined	21:51
	← robinr left	21:52
	→ tsc_ joined	21:53
	→ tvachon joined	21:53
	→ robfitz_ joined	21:53
	← pombreda left	21:57
	← riddochc left	21:58
jrockway	is there a way to import a single file's history from cvs to git?	21:58
semi_	import it to git first then git-filter-branch everything else out	21:58
	→ riddochc joined	21:59
	→ Sho_ joined	21:59
jrockway	it's kind of a giant repo (emacs)	21:59
tsc_	any msysgit folks around?	21:59
	← kristoffer_ left	21:59
jrockway	i think i'll just write a quick perl script to extract the file i need and add each revision to git	21:59
	that's good enough for my needs	21:59
	→ pombreda joined	22:00
	← HG` left	22:06
	→ bobesponja joined	22:06
	← cappslocke left	22:08
	→ csc` joined	22:09
mugwump	jrockway: you have access to the repository?	22:09
	→ cappslocke joined	22:09
	→ wagle joined	22:09
mugwump	Someone has already converted emacs AIUI - Randal was talking about using it	22:09
tsc_	http://git.sv.gnu.org/gitweb/?p=emacs.git	22:10
jrockway	oh nice	22:10
	i solved my problem with a bit of drudgery	22:11
	basically my patch was against an old cvs checkout	22:11
	so i updated that "knowing" that the file is dead, and it wouldn't change	22:11
	well it did, so my patch became fucked up	22:11
	anwyay, i found the right base and can now merge properly and get a real diff	22:11
	i was silly initially by not importing the original; i just made a new repo with my changes :)	22:12
	a 3 seconds savings of time a few weeks ago == a 15 minutes waste today	22:12
	→ hon joined	22:23
	→ robinr joined	22:24
	← Phoop left	22:26
manithree	Replying to my own question, it looks like the 0.7.7 version of buildbot will have git support	22:26
hon	Assume I have a few categories of files all in the same directory. Each category is really one thing. I prefer to have the sets under revision control, each category having its own history. Can I currently do this with git? If not, will it be possible AND reasonable in the future?	22:28
raalkml	yes. Create directories	22:28
	→ zapnap joined	22:34
	→ cmarcelo joined	22:34
hon	Will it be too ugly to keep all the files in the same directory?	22:36
raalkml	only if you paint them pink	22:36
kreuter	http://git.pastebin.com/d73d25c13 <-- using a git built from upstream source on NetBSD 3.0. any idea what the problem is?	22:37
raalkml	http://repo.or.cz/r/sbcl/kreuter.git	22:38
kelvie_	anyone nkow where I can find out how to add a svn branch to track with git-svn?	22:39
kreuter	bad url?	22:39
raalkml	yep	22:39
tsc_	try changing the '/w/' in http://repo.or.cz/w/sbcl/kreuter.git to an '/r/'	22:39
gebi_	kreuter: clone from git://repo.or.cz/sbcl/kreuter.git	22:39
kreuter	oh	22:39
	hm.	22:39
	← pombreda left	22:39
	→ sverrej_ joined	22:39
gebi_	it's strange that gitweb don't provide a way to get the repo too, but thats the way in git...	22:39
	→ ironfroggy joined	22:40
	→ pombred1 joined	22:40
raalkml	...the wain in gitweb	22:40
	way	22:40
kreuter	hm. the /w/ and /r/ are for writing and reading?	22:40
raalkml	no	22:40
	it just so	22:40
kreuter	ok	22:40
vmiklos	it's w = gitweb, r = repo	22:41
	where the repo is ready-only as well	22:41
	← mithro left	22:41
hon	How is modules support at the moment? (it wasn't ready last time I checked)	22:42
	← zapnap left	22:42
	→ jeffpc joined	22:42
raalkml	people like it (I do)	22:42
	→ mithro joined	22:43
	→ pombreda joined	22:43
	← mjc_ left	22:43
careo	kelvie: take a look in .git/config. there should be an [svn-remote "svn"] section. the branches= line looks promising	22:44
	kelvie: if you find out though, I'd love to know. I want t do do the opposite and stop watching some of them :)	22:44
	gebi_ → gebi	22:44
	← cappslocke left	22:47
mugwump	kelvie_: to add another branch, you can also set up another git-svn remote using the -i option	22:48
	the way that works isn't very flexible atm	22:48
gebi	hm... where is the horizontal scroll-bar in gitk?	22:48
	← sverrej left	22:48
gebi	in the source code window, i thought there where one in the past	22:48
	→ cappslocke joined	22:50
	→ CrimsonScythe joined	22:51
CrimsonScythe	hi guys. i've got a quick question. are there any web front-ends for git that have the same functionality as SVNManager has for svn? (i.e. create users, groups, repositories; change access permissions, etc)	22:52
DrNick	no	22:53
CrimsonScythe	bummer :-/	22:53
DrNick	mostly because git doesn't have users, groups or access permissions	22:54
tpope	everything in that list but repositories is generic to unix	22:54
mugwump	CrimsonScythe: see gitosis	22:54
yoh	Hi Git people... Quick question again: I've rebased (had to fix some commits which prevented proper merge) a branch and trying to push it onto remote... I am sayin git push -f origin +_tent/sg:_tent/sg but it still refuses to push saying that "denying non-fast forward refs/heads/_tent/sg (you should pull first)"	22:54
CrimsonScythe	so there aren't any ways in which to somewhat automate the creation and maintenance of repositories?	22:54
yoh	but I don't want to pull I guess, right?	22:55
	→ vatic42 joined	22:55
CrimsonScythe	mugwump: thanks. i'll check it out	22:55
	← wagle left	22:55
vatic42	hey, are you all willing to help out with gitosis issues?	22:55
CrimsonScythe	DrNick: so it's not possible to set access restrictions on parts of a repository?	22:56
tpope	I think I've helped you plenty already	22:56
vatic42	!	22:56
	you're stalking me tpope	22:56
tpope	excuse me	22:56
	why don't we see who joined first	22:56
kelvie_	mugwump: why isn't it flexible?	22:56
	the -i option	22:56
	rather I added it manually myself	22:56
vatic42	I'm just kidding	22:57
kelvie_	in the config file that is	22:57
vatic42	I'm trying to setup gitosis, but I'm having this problem: http://pastie.caboo.se/142155	22:57
tpope	well I was joking first too	22:57
vatic42	I know	22:57
	:)	22:57
mugwump	kelvie_: because what is there works :)	22:57
	patches to add better behaviour welcome	22:57
	← alley_cat left	22:58
tpope	vatic42: looks like it might perhaps be too old of git	22:58
	but I've never used gitosis	22:58
vatic42	hm	22:58
	how old is 1.1.3?	22:58
yoh	heh ok... I just removed that branch on remote side first bu pushing empty one...	22:58
tpope	ANCIENT	22:58
vatic42	damn ubuntu	22:58
tpope	if you said 1.4 I would have told you to upgrade	22:58
vatic42	why does ubuntu install 1.1.3? that's retarded	22:59
tpope	that must be the longterm support release or something	22:59
	because even debian stable has 1.4	22:59
vatic42	haha	22:59
	maybe the apt repos need to be updated	22:59
mugwump	see packages.ubuntu.com/git-core	22:59
kelvie_	mugwump: no no, I meant in what way is it not flexible? :P	22:59
vatic42	one sec	22:59
mugwump	that version was in dapper	22:59
kelvie_	not why can't it be more flexible :P	22:59
	← pombred1 left	22:59
mugwump	kelvie_: oh, well basically it just supports one trunk path, one tags path and one branches path - not arbitrary mappings like normal git push/pull	23:00
	so, if you want anything more complicated you need to set up multiple remotes	23:00
kelvie_	ah	23:00
	← CrimsonScythe left	23:01
vatic42	dang	23:02
	this is dapper	23:02
	→ z3ro joined	23:02
vatic42	I think	23:02
	it just hasn't been kept up to date	23:02
	not my machine, sorry	23:02
mugwump	you might be able to backport	23:02
vatic42	hm	23:02
mugwump	oh, look, there's a dapper-backports	23:02
	version, 1.5.3.6	23:03
	that's good enough	23:03
vatic42	k	23:03
	I will try to set that up	23:03
kelvie_	so if you import them as branches rather than a separate remote.. git will know their common ancestors and what not?	23:03
mugwump	kelvie_, hopefully	23:03
kelvie_	though theorhetically it'd know their ancestors even if they are a separate remote	23:03
mugwump	svn is somewhat fragile in this regard	23:03
kelvie_	theoretically *	23:03
mugwump	you usually end up having to make repairs	23:04
	← raalkml left	23:04
kelvie_	I'd just like it to not have to fetch redundant items	23:04
	fetching a branch (as a new remote) is sloooow	23:04
mugwump	use svnsync or svk to do the initial fetch	23:05
	svk 2.0 is very fast at fetching remote repositories. Can cut times down from days to hours	23:05
kelvie_	the thing is I already have the entire history checked out	23:05
	of trunk	23:05
	we branched for release a while ago	23:06
	a while being less than a month	23:06
	oh wow it finished	23:06
mugwump	It should deal with that fine. make sure --follow-parents is on	23:06
	→ kukks joined	23:06
kelvie_	mugwump: yeah, and it does :P	23:07
	it still has to do the initial svn checkout of the entire tree though	23:07
	which is 5 gigs	23:07
mugwump	you've repacked?	23:08
	or is a head checkout 5 gigs?	23:08
kelvie_	I repack often	23:08
	a head checkout	23:08
	my pack file is 9 gigs	23:08
	we check in binaries... so that can't be helped	23:08
mugwump	:o	23:08
kelvie_	and it's 12 years old	23:08
	.. 13	23:08
	wait 14 now	23:08
	→ pici_juvenile joined	23:11
pici_juvenile	bash: git: command not found	23:11
vmiklos	did you want to ask something?	23:12
mugwump	pici_juvenile: add ~/bin to your PATH	23:13
	mugwump guessing wildly ;)	23:13
pici_juvenile	mugwump: http://pastebin.ca/869167	23:15
gebi	kelvie_: i've another project which commits the _whole_ build-system (gcc+all libs in binary) with every official release g	23:15
	→ dsaxena joined	23:15
mugwump	pici_juvenile: ah.. you want git-core	23:16
vatic42	tpope: yeah that was the problem; thanks	23:16
pici_juvenile	oh...	23:16
mugwump	git is some crappy old GNU Interactive Tools program	23:16
	← pmezard left	23:16
Randal	git-er-done!	23:17
	pici_juvenile → FullMon-T	23:17
	FullMon-T → Death2Larry	23:18
vmiklos	every nick change enlarges your e-penis	23:19
	Death2Larry → pici_juvenile	23:20
pici_juvenile	any idea where I can find git-core help?	23:20
gebi	what help?	23:20
vmiklos	fyi, this channel is not about GNU Interactive Tools	23:20
ivazquez	There's some documentation linked to on the website.	23:20
mugwump	pici_juvenile: you need the git-doc package I think	23:22
pici_juvenile	mugwump: apt-get install git-core	23:23
mugwump	was that a w/w ?	23:24
pici_juvenile	unsure	23:24
	now, I get a different error message.	23:24
	Here comes about 6 lines:	23:25
	root@HappyTrees:/home/enigmata/wifi# git clone http://intellinuxwireless.org/repos/ipwraw.git	23:25
	Initialized empty Git repository in /home/enigmata/wifi/ipwraw/.git/	23:25
	/usr/bin/git-clone: 310: curl: not found	23:25
	Cannot get remote repository information.	23:25
	Perhaps git-update-server-info needs to be run there?	23:25
	root@HappyTrees:/home/enigmata/wifi#	23:25
	← toxx_ left	23:26
	← nbrunson left	23:26
	→ nud_ joined	23:27
mugwump	install curl	23:27
	← pombreda left	23:27
	→ pombreda joined	23:28
ivazquez	Wow. No wonder people say the .deb distros don't suffer from "dependency hell".	23:30
	They don't have any :P	23:30
pici_juvenile	???	23:30
	← vatic42 left	23:30
pici_juvenile	mugwump: that did it. thanks	23:30
	→ flaguy joined	23:30
	→ Ademan joined	23:31
mugwump	debian has the dependency correct, it's just those "who needs QA on packages anyway" Ubuntu ones that are broken :)	23:32
	→ vatic42 joined	23:32
	← cappslocke left	23:32
	← khwu- left	23:35
kelvie_	gebi: wow.. and you checked that out in git?	23:37
semi_	ivazquez: what do you mean?	23:38
	oh, curl is just recommended, not depended on	23:39
	ivazquez: Depends is just for packages that are absolutely necessary, Recommends are almost always wanted and they are installed by default on at least my system, Suggests are things that you often want but some people don't	23:40
	→ timocratic joined	23:42
	← nud left	23:43
	→ HG` joined	23:49
gebi	kelvie_: yea :)	23:50
	but it's only 630MB .git	23:51
	← HG` left	23:54
	← kreuter left	23:55

Logs Search ←Prev date Next date→

Channels Documentation