IRCloggy #git 2009-05-10

Provider of IRC logs since 2005.

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

	2009-05-10
	← gotgenes left	00:00
	← drewolson left	00:01
	← joshsdc left	00:03
	← catalis left	00:03
	→ dsyzling joined	00:03
	→ drizzd joined	00:04
	← dsyzling left	00:04
	← WALoeIII left	00:06
	← toxx_ left	00:07
	→ TokyoDan joined	00:07
	← towski left	00:08
	→ towski joined	00:08
	← wezyde left	00:08
	→ quoin joined	00:09
	→ WALoeIII joined	00:10
	→ towski_ joined	00:10
	→ ssvb joined	00:11
	← towski left	00:11
	→ wsc_ joined	00:12
	← wwwald left	00:13
	← wshimmy1 left	00:15
	→ gotgenes joined	00:15
	dereine → dereine[OFF]	00:15
	← justatheory left	00:16
	→ wshimmy joined	00:17
	← LiamH left	00:22
	→ shana joined	00:24
	← mjf left	00:25
	← vipul left	00:25
	← wsc left	00:28
	→ etienne9 joined	00:29
	← etienne9 left	00:30
	→ xyz joined	00:31
	← TokyoDan left	00:33
	← eletuchy left	00:34
	→ JasonWoof joined	00:35
	← dreiss left	00:35
	← johan-s left	00:35
	→ alanhaggai joined	00:36
	← alanhaggai left	00:36
	→ alanhaggai joined	00:38
	← uau left	00:39
	← qhoxie left	00:42
	← ericindc left	00:44
	→ wsc joined	00:45
	→ uau joined	00:46
	sebleier → sebleier\|away	00:46
	← zirpu left	00:46
	← MisterN_ left	00:49
alanhaggai	Hi. Is there any way to add a branch to track a Subversion repository?	00:53
	← bryanray left	00:53
	← brennen_ left	00:57
	→ brennen joined	00:57
	← kuadrosx left	00:59
	← wsc_ left	00:59
	→ Minimiscience joined	01:01
	← rettub_ left	01:01
	→ wsc_ joined	01:02
	← lolage left	01:02
	← dtangren left	01:05
	← schacon left	01:06
	→ wsc__ joined	01:06
	→ gaveen_ joined	01:06
	→ rettub_ joined	01:07
	← tatsuyao left	01:07
Fissure	alanhaggai: you mean like git-svn?	01:07
alanhaggai	No.	01:07
	Fissure: I imported a Subversion repository to GitHub. Then, I cloned it.	01:08
	So, the cloned one has no idea of the Subversion repository.	01:08
	How can I pus commits that have been made by someone else to the Subversion repository?	01:09
	s/pus/push/	01:09
	So I was thinking of adding a remote tracking branch for the Subversion repository. I am not sure how to create such a branch.	01:09
	Fissure: I do not know if it is possible either.	01:10
	← aziz left	01:10
	← gaveen left	01:12
	← brennen left	01:15
	← wsc left	01:16
	← SRabbelier left	01:18
	← wsc_ left	01:25
	→ Beket_ joined	01:26
Beket_	Hey people. Is there a way to git cherry pick the last N commits of a branch ?	01:26
	or perhaps not use cherry-pick at all	01:27
	← JasonWoof left	01:27
	← ceej left	01:27
Beket_	i just want to grab the last N commits of a branch	01:27
bdrewery	I think git-rebase can do that	01:32
Beket_	really ?	01:32
	xm	01:32
	← paulboone left	01:34
	← parasti left	01:37
	→ ilogger2_ joined	01:46
	→ ceej joined	01:48
	→ joevandyk joined	01:49
	← ceej left	01:51
	→ ericindc joined	01:52
	→ JasonWoof joined	01:58
	→ mw joined	02:00
	→ joshsdc joined	02:00
	→ [1]intripoon joined	02:06
	→ wsc joined	02:07
	→ bryanray joined	02:13
	→ brennen joined	02:14
	→ paltman94 joined	02:16
	← mw left	02:20
	→ cytrinox joined	02:21
	→ mw joined	02:21
	→ MachX joined	02:22
	→ lasso joined	02:22
	[1]intripoon → intripoon	02:24
	← lasso left	02:24
	→ bcardarella joined	02:31
	← brennen left	02:31
	→ rekoerb joined	02:32
	→ niki joined	02:35
	→ mlins joined	02:35
	→ brennen joined	02:35
	→ bdiego joined	02:39
	→ jelly-bean joined	02:42
jelly-bean	if i change a file but haven't committed and i want to undo those changes with git (same as git reset --hard HEAD except for one specific file or a couple files only) how can i do that?	02:43
	git checkout HEAD <file> ?	02:44
Beket_	git checkout HEAD -- file	02:44
jelly-bean	Beket_: ok, thx.	02:44
	← jelly-bean left	02:46
Beket_	if u hsve git added it as welll u need git rm --cached <filename>	02:46
	← MachX left	02:46
	← rekoerb left	02:46
	→ tjafk1 joined	02:52
	← mw left	02:53
	← brennen left	02:54
	→ brennen joined	02:54
alanhaggai	Hi. I am trying to add a remote Subversion repository tracking branch.	02:55
	I did a `git svn init <repo>` and then `git fetch git-svn`. I am thinking of doing the fetch occasionally and merging it with master.	02:55
	Is it a correct workflow, or can you suggest another?	02:55
	→ brennen_ joined	02:59
	→ JPohlmann1 joined	03:01
	→ AkumaStreak joined	03:06
Run	If I do 'git push' in branch master, it tries to push branch build ... is that a topgit thing? This is insane :/	03:08
	http://codepad.org/4VJu3xBU	03:10
	→ clumzee joined	03:11
clumzee	clueless ;)	03:11
	I need to clone a repository in a bash script, does anyone know how to automatically respond to the verify rsa fingerprint?	03:11
	or disable it during a clone	03:12
RandalSchwartz	log in once manually	03:12
	and say "yes"	03:12
	that's a safety thing	03:12
	don't try to work around it	03:12
	← ericindc left	03:13
	← brennen left	03:13
	→ [1]intripoon joined	03:13
clumzee	its going to be running on hundreds of machines ;)	03:15
SamB	why ?	03:15
clumzee	why not	03:16
SamB	I mean, what for?	03:16
RandalSchwartz	first, ask yourself why that check is there	03:16
clumzee	in a cloud of application servers	03:16
RandalSchwartz	and then decide if you really wanna work around it	03:16
clumzee	yes.	03:16
RandalSchwartz	if you're just pulling, use git: protocol	03:16
	not ssh	03:16
	← brennen_ left	03:17
RandalSchwartz	heck... you can even use http protocol	03:17
clumzee	when i have servers that come up and down automatically and I know the finger print ahead of time, it would be sweat to just 'ignore it'	03:17
SamB	wouldn't it be better for it to be in known_hosts ?	03:17
clumzee	can you add a finger print to a known_hosts i thought it had to be all the hashed stuff	03:18
	RandalSchwartz its a private repo on github, I think its always over ssh	03:18
SamB	that's where they go	03:18
	when you say yes	03:18
	so ... just do it once and say yes, then you can copy that entry to every known_hosts file or something ...	03:19
RandalSchwartz	why not just set up an rsync server	03:19
	is this always pull-only?	03:19
clumzee	yeah but its not a nn:nn:nn:nn: etc, its the all hashed up funness	03:19
RandalSchwartz	sounds like you're abusing git	03:19
SamB	clumzee: and ?	03:19
	I guess it's just a longer fingerprint ...	03:20
	in a base other than hex	03:20
clumzee	hrm maybe.	03:20
	RandalSchwartz, I could do that, but id like to have as little moving parts as possible, and i like being able to text my servers a rev number and them all grip it	03:21
RandalSchwartz	yeah, sounds like you want an rsync server	03:21
	or any of a dozen other solutions	03:21
	not git	03:21
	→ shze joined	03:21
RandalSchwartz	git would be abuse there	03:21
SamB	RandalSchwartz: overkill, sure ...	03:21
	but why not ?	03:21
RandalSchwartz	there are mass-configurator solutions	03:21
	why?	03:21
	why use a hammer to pound in a screw	03:21
	doesn't make sense	03:21
SamB	if he's already using git to control the files ...	03:22
RandalSchwartz	but he doesn't need the git repo on each machine	03:22
SamB	why not use it to update them as well ?	03:22
RandalSchwartz	he just needs the rsync	03:22
SamB	well, true, he doesn't need the entire history ...	03:22
RandalSchwartz	and an anon rsync server is trivial	03:22
	so have one master, edited with git	03:22
SamB	I'm not sure how anon is better	03:22
RandalSchwartz	and 1000 slaves that view it with rsync	03:23
	← clumzee left	03:23
	→ alama joined	03:24
alama	can one convert a bzr repo to a git repo?	03:25
	i'd like to put up some of my old stuff on github.com	03:26
	but some of my old repos are bzr	03:26
	→ xeno__ joined	03:27
	→ cilly joined	03:27
	→ eno__ joined	03:28
	← intripoon left	03:29
	[1]intripoon → intripoon	03:29
	← paltman94 left	03:32
	← JPohlmann1 left	03:32
shze	I want to cherry-pick a commit from another repository, but despite reading man pages and searching I can't find how to do this. can anyone help me?	03:33
bdrewery	git remote add someother URL	03:33
	git cherry-pick hash	03:33
	git fetch someother too	03:33
shze	oh, ok, so I missed the very first command. thanks, I'll try	03:33
	eno__ → eno	03:36
	→ cdmwebs joined	03:37
shze	cherry-pick returns fatal: Could not find <hash>; do I have to create a local branch from the remote one first?	03:37
bdrewery	did you fetch?	03:38
shze	before cherry-pick? no.	03:38
bdrewery	yeah fetch first	03:38
	→ orafu joined	03:40
shze	worked great. thanks a lot!	03:40
bdrewery	:)	03:41
	→ ashleyw joined	03:45
	→ jelly-bean joined	03:46
jelly-bean	what's the cmd to quickly git branch to a temp swap/switch branch	03:46
	or something like that	03:47
	← shze left	03:47
	→ brennen joined	03:49
bdrewery	git stash probably	03:51
jelly-bean	bdrewery: yep	03:55
	→ justatheory joined	03:57
	← justatheory left	03:59
	→ seangrove joined	04:02
	← bdiego left	04:02
	← jelly-bean left	04:04
	← brennen left	04:05
	← cdmwebs left	04:06
	→ wsc_ joined	04:19
	→ brennen joined	04:28
	← wsc left	04:29
	→ ivanoats joined	04:30
	→ dreiss joined	04:39
	→ fujin joined	04:45
	← brennen left	04:46
	→ schacon joined	04:48
	→ arohner joined	04:50
	→ hyperair joined	04:55
	← schacon left	04:55
	← hyperair left	05:01
	→ hyperair joined	05:01
	→ curvature joined	05:01
	→ killerchicken__ joined	05:03
	← alama left	05:04
	→ bobmcw joined	05:04
	→ mlins_ joined	05:05
	→ eno__ joined	05:07
	→ rubydiamond joined	05:12
	→ g3d joined	05:14
	← mlins left	05:16
	← eno left	05:18
	→ towski joined	05:19
	← ivanoats left	05:20
	eno__ → eno	05:26
	← arohner left	05:29
	→ harinath joined	05:30
	→ unixluser joined	05:31
unixluser	hello. theres a git repo that i want to clone.. but i only want to clone a specific branch of it? how do i do this?	05:31
cehteh	mkdir project; cd project	05:34
	git init	05:34
	git pull git://url branchname	05:35
cxreg	iirc, that will merge the branch into your master, though?	05:36
unixluser	.git/config only mentions core	05:37
	not the branch name	05:37
	nor remote details	05:37
cxreg	yes, "git pull" with args doesn't always do what you expect it to do	05:38
	maybe something more like "git init; git remote add remotename git://url; git fetch remotename branch; git checkout -b remote/branch branch"	05:39
	→ brennen joined	05:39
unixluser	'remotename' would typically be origin, right?	05:40
cxreg	could be, doesn't have to be	05:41
	that's the default name it gives clones, though	05:41
	← ashleyw left	05:41
	→ ashleyw joined	05:41
	→ aspotashev joined	05:43
unixluser	cxreg: the last command (git checkout..) results in fatal: git checkout: updating paths is incompatible with switching branches.	05:43
cxreg	what version of git?	05:44
	sounds like 1.5 maybe	05:45
cehteh	cxreg: yes but master is empty after a git init .. alternatively you can just fetch that branch and reset your master to it	05:45
cxreg	sure, if you're ok with master being something that's not remote/master	05:46
unixluser	yeah, i'm fine with that	05:48
	as long as pushing it online will go properly to the branch	05:48
cxreg	push expects that the branch names are symmetric, unless you tell it otherwise	05:48
	imho, naming fetched branches a different name leads to trouble	05:49
unixluser	ot	05:49
	it's a very basic repo	05:49
	basically remotely created to store dotfiles	05:50
	for different distributions, with each having a different branchname	05:50
	i'm just looking for an alternative to standard git clone; git branch -D master	05:50
cxreg	to push, you'd need to "git push remote master:remotebranchname"	05:51
	→ brennen_ joined	05:51
unixluser	yeah, i just avoid 'master:' altogether	05:51
	→ ashleyw_ joined	05:54
	← brennen left	05:56
	← wsc_ left	05:57
	→ sitaramcell joined	05:58
	→ brennen__ joined	05:59
	→ rubydiam_ joined	06:01
cxreg	unixluser: my last command had a typo, btw which is probably what failed	06:01
	git checkout -b branch remote/branch	06:01
	← rubydiamond left	06:02
	← ashleyw left	06:02
unixluser	nope, didn't work either	06:03
	git 1.6.3	06:03
	→ rubydiamond joined	06:04
	→ gaveen__ joined	06:04
	← rubydiam_ left	06:05
	→ WALoeIII joined	06:07
	← brennen_ left	06:08
	→ dduncan joined	06:09
	← unixluser left	06:09
	← brennen__ left	06:16
	← dduncan left	06:17
	← AkumaStreak left	06:18
	← sitaramcell left	06:25
	→ WALoeIII_ joined	06:31
	← WALoeIII left	06:32
	→ wsc joined	06:37
	→ kate21de joined	06:43
	→ fargiola` joined	06:43
	fargiola` → fargiolas	06:44
	← joshsdc left	06:45
	← ashleyw_ left	06:50
	→ elmob joined	06:55
	← dreiss left	06:59
	→ r0bby joined	07:01
	→ foutrelis_ joined	07:01
	→ GreenAsJade joined	07:01
GreenAsJade	Is it "basically a bad idea" to clone another repo into a subdirectory of an existing repo?	07:02
	Or should that fundamentally "just work"?	07:02
	→ resmo joined	07:02
	← cilly left	07:02
RandalSchwartz	you want submodules	07:03
GreenAsJade	I have a project into which I did this (not thinking too hard, obviously). Then I pushed my project back to where it comes from. The 'sub repo' didn't get pushed/	07:03
RandalSchwartz	then it works	07:03
AAA_awright	GreenAsJade: Sounds like submodules	07:03
RandalSchwartz	without that, no it's a bad idea. :)	07:03
GreenAsJade	Ah - OK, "doh"	07:03
	GreenAsJade tips his hat respectfully to RandalSchwartz, runs off to learn The Right Way	07:04
GreenAsJade	hmm - so I can	07:07
	t even take a shortcut and copy the other "submodule" in eh, 'cause that's already a 'repo'	07:07
	(I meant "darn I'll just plonk the files in")	07:07
Arrowmaster	part of the submodule implentation is that git completely ignores other git repos in subdirs	07:09
	→ angerman joined	07:10
GreenAsJade	So, before I get carried away (now that I have read something about them) ... if I "submodule add" rather than just splat the module repo in, then git push the project, the submodule will get pushed too, in it's right place in the project?	07:12
Arrowmaster	not really	07:13
GreenAsJade	rats :)	07:14
Arrowmaster	but information about where others can clone that repo from and what exact commit you were on in that repo when you commited to your project will be stored	07:14
GreenAsJade	I better ask how to do what I want to do then :)	07:14
	I have a project, which is deployed by git push	07:14
	→ fargiola` joined	07:15
GreenAsJade	I have created a component which needs to be in this project, and it is in a separate repo (and is pushed to github seprately)	07:15
	the first project needs to contain the files of the second when it is deployed	07:15
Arrowmaster	is this done with a custom script or are you using a larger project designed to do it? (theres a few)	07:15
	← fargiolas left	07:15
	fargiola` → fargiolas	07:16
GreenAsJade	so far no scripts, just git push ...	07:16
	→ mtkd joined	07:16
GreenAsJade	"git push" pushes the component to github. "git push" deploys the project to it's destination	07:16
	Unfortunately, when I copied the new component into the right place in the project, it doesn't get pushed	07:16
	Does this make sense, or is it gibberish?	07:18
Arrowmaster	sorta yes sorta no	07:18
GreenAsJade	Strange, this was all feeling good and natural up till today!	07:18
Arrowmaster	when you say 'deploy' you mean a working copy on a server somewhere is getting updated to use the new version right?	07:18
GreenAsJade	Right - the server runs off it's copy of the repo, which is the "master". git push sends the new commits up to it	07:19
Arrowmaster	right	07:19
	thats not possible without some kinda script on that server to update the working copy	07:19
GreenAsJade	yow.	07:19
	→ dizpater joined	07:20
GreenAsJade	OK, I must be missing something about how "people normally do what I'm trying to do" then, I'll go take a look at that avenue	07:20
	← dizpater left	07:20
Arrowmaster	GreenAsJade: faq non-bare	07:20
Gitbot	GreenAsJade: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare	07:20
GreenAsJade	(the component is a rails plugin, so there must be some other sensible way to get a rails plugin into a rails project other than copying it's repo :)	07:21
Arrowmaster	GreenAsJade: like this http://github.com/guides/deploying-with-capistrano	07:21
GreenAsJade	thanks for taking the time...	07:22
Arrowmaster	ive never used capistrano and i have no clue if it supports git submodules but it seems to have some sort of dependancy system	07:25
	→ NoirSoldats joined	07:25
Arrowmaster	but as a general rule its probably never a good idea to commit another project directly into yours just so its available when other solutions such as submodules exist, it keeps things a lot cleaner that way	07:26
GreenAsJade	I think "work out how to install your own rails plugin, instead of ploking it in, dude" is the right advice for me :)	07:26
	→ Fragsworth joined	07:27
	gaveen__ → gaveen	07:30
	← gaveen left	07:30
Arrowmaster	GreenAsJade: it whould help if i actually fully read a page before linking it, the github guide for capistrano mentions enabling submodule support so yes it supports git submodules	07:32
GreenAsJade	:)	07:32
	→ wsc_ joined	07:37
	← wsc_ left	07:37
	→ patrikf joined	07:40
GreenAsJade	so ... if I push my component to github, then install it in my project from there, using the rails install, it arrives properly by some magic :)	07:44
	So my development flow now includes a push to github of the component and a pull back if I change it, but at least it works!	07:45
Radar	GreenAsJade: when you clone your project you will need to do git submodule update --init in order to get the submodules into your app.	07:45
GreenAsJade	I don't have control at the deployment-target end. I can only git push to there	07:45
	But sounds like I should read aobut --init	07:45
	→ spuk- joined	07:50
spuk-	hello, I'm trying something like: 'export GIT_DIR=/tmp/test.git ; git init --bare; git add ~/foo', I'm getting "fatal: This operation must be run in a work tree" everytime, isn't that supposed to work ?	07:52
patrikf	spuk-: no	07:53
	→ ia joined	07:53
RandalSchwartz	why aren't you inside test.git ?	07:53
	ahh. ~/foo is not inside GIT_DIR	07:53
	that's why you can't do that	07:53
	unless your ~ is /tmp :)	07:53
patrikf	spuk-: you cannot normally commit to bare repositories	07:53
RandalSchwartz	ahh, that too	07:54
	← wsc left	07:54
RandalSchwartz	bare = no workdir	07:54
patrikf	spuk-: if you must for some reason, you can use "git hash-object -w" in conjunction with "git update-index --cacheinfo" I think	07:54
	→ binjured joined	07:54
spuk-	hmm, that might make sense	07:55
	thank you both	07:55
	→ dreiss joined	07:56
	→ fargiola` joined	07:56
meder	RandalSchwartz: oh, you're a git one too eh	07:59
RandalSchwartz	eh?	08:00
	→ ccherret1 joined	08:00
RandalSchwartz	meder?	08:01
cehteh	randal is everywhere :P	08:03
	← mtkd left	08:04
	→ aspotashev\|eeepc joined	08:04
	→ wsc joined	08:05
	→ hummingkid joined	08:05
meder	oh nothing, just saying i see you in two channels	08:06
	← bcardarella left	08:06
RandalSchwartz	I'm in a lot more than that	08:09
	smalltalk - perl - git - postgresql - openbsd - opensolaris (etc etc)	08:09
	and jquery	08:09
	and whatever else we've recently interviewed on FLOSS Weekly	08:09
cehteh	you wherent on #cinelerra or #lumiera, shame on you :P	08:10
meder	well yeah	08:10
RandalSchwartz	I was a bit overwhelmed on that show	08:10
cehteh	haha excuses	08:10
	cehteh hides	08:10
RandalSchwartz	my line of questions wasn't as good as I wanted. a bit frustrated	08:10
	did much better on OLE nepal and Xen, I think	08:11
	part of it was the delay to the UK	08:11
	and 4 people on skype. that's never good	08:11
cehteh	yeah i think the whole one was little unprepared .. aaron asked longer ago, but then just reappeared 2 days before	08:11
	i am happy that we send joel eg	08:11
RandalSchwartz	well - it'll certainly raise the visibility	08:12
cehteh	and dont call him a kid next time, that was a bit rude :)	08:12
RandalSchwartz	heh	08:12
	he's a kid relative to me	08:12
cehteh	he is our youngster	08:12
	→ dodo_the_last joined	08:13
RandalSchwartz	I'm glad I nailed Leo on the 29.97 though	08:13
	given that Leo is a TV broadcaster and technician :)	08:13
cehteh	30000/1001 to be exact	08:13
RandalSchwartz	oooh	08:13
	RandalSchwartz bows in cehteh's general direction	08:13
cehteh	whenever i see a video lib which handles this incorrectly i barf on it	08:14
	meder can't handle the context	08:14
RandalSchwartz	... To reduce interference between the chrominance signal and FM sound carrier required a slight reduction of the frame rate from 30 frames per second to 30/1.001 (very close to 29.97) frames per second, and changing the line frequency from 15,750 Hz to 15,734.26 Hz.	08:14
	you're right	08:14
	wow	08:14
cehteh	lumiera aims to be a tool which doesnt make compromises in exactness and quality	08:15
	you dont want your video sync drift away	08:15
	← fargiolas left	08:15
cehteh	not even microseconds	08:15
RandalSchwartz	yeah... I've had stuff get off after a while	08:16
	probably poor computations	08:16
cehteh	well if its more noticeable then blame your cheap cam	08:16
	RandalSchwartz tries to figure out what channel he has selected	08:17
RandalSchwartz	is this #git?	08:17
cehteh	for some the 44.1 sound carrier drifts with temperature	08:17
	but the computer expects it to be constant	08:17
	RandalSchwartz has a bad irc client that doesn't show current channel	08:17
cehteh	hehe ok :)	08:17
	cehteh goes one coding	08:17
	→ ciskje joined	08:17
RandalSchwartz	ahh yeah, #git	08:18
cehteh	there is #openvideo for general video talk	08:18
	'nuff offtopic here	08:18
RandalSchwartz	like I need one more channel to irritate peopel in :)	08:18
	← WALoeIII_ left	08:19
	→ WALoeIII_ joined	08:20
	→ brennen joined	08:20
	← aspotashev left	08:21
	→ aspotashev__ joined	08:21
Fissure	RandalSchwartz: wow, must be a really bad client	08:24
	and i felt lame for using finch...	08:24
	→ radarek joined	08:28
	← dreiss left	08:28
	→ mithro joined	08:32
	→ brennen_ joined	08:32
	← mithro left	08:33
	→ bdowning joined	08:34
	→ ethercrow joined	08:35
JEEBcz	is there an easy way to get the changelog f.ex. between the versions 7554705 and 400740b in git?	08:35
RandalSchwartz	git log 755..400	08:36
	← rubydiamond left	08:36
RandalSchwartz	maybe git log 755...400	08:36
	→ rubydiamond joined	08:37
	→ mithro joined	08:37
	→ therrg joined	08:37
JEEBcz	ok, I'll try :)	08:37
	← aspotashev\|eeepc left	08:37
	→ naeu joined	08:40
	← brennen left	08:41
	← foutrelis_ left	08:49
	← JasonWoof left	08:51
	→ Bogh joined	08:52
Bogh	morning!	08:52
	→ brennen__ joined	08:55
Ilari	...	08:56
	→ knittl joined	08:56
	← therrg left	08:57
	→ smeevil joined	08:58
	← ciskje left	08:58
	← brennen_ left	08:58
	→ sitaram joined	08:58
	→ solydzajs joined	08:58
	→ kahmalo joined	08:59
	← sitaram left	09:01
	→ brennen_ joined	09:07
	→ lolage joined	09:08
	← angerman left	09:08
	→ gilimanjaro joined	09:11
	← brennen__ left	09:13
	← seangrove left	09:14
	→ d0k joined	09:22
	→ brennen__ joined	09:26
	→ DavidKlein joined	09:27
	→ tatsuyao joined	09:27
	← patrikf left	09:28
	→ mbroeker joined	09:29
	→ seradin joined	09:32
	→ khelll joined	09:32
	→ seangrove joined	09:32
seradin	hey guys, suppose i have a git-repository at xyz.com/rep.git now i want to check this out on a client machine, edit something and commit the changes: But when i do a git --init followed by a git remote add origin ssh://user@xyz.com/rep.git/ and a git fetch ... then nothing is fetched to my local directory? Why ?	09:34
khelll	i'm having this error when trying to clone from bare git repo for first time:	09:35
	fatal: The remote end hung up unexpectedly	09:35
	Bad port ''	09:35
	fatal: The remote end hung up unexpectedly	09:35
	→ JPohlmann1 joined	09:35
	← brennen_ left	09:36
cehteh	khelll: "Bad port"	09:36
	seradin: fetch doesnt update your working tree	09:36
seradin	cehteh: and how can i do what i want to do?	09:36
cehteh	first you need to tell what you want to do :)	09:37
	branches in git are local	09:37
seradin	i want to check out the stuff from my repository in order to edit and commit it	09:37
cehteh	there is your local master branch and the remote master branch .. and these are distinct	09:37
seradin	okay, and how do i pull from one to the other?	09:38
cehteh	git pull	09:38
	which is essentially a git fetch + git merge	09:38
seradin	git pull tells me that it doesn't know what to do	09:38
cehteh	git pull remote branchname maybe then	09:38
	your new local repo is brand new (you told you did git init)?	09:39
seradin	yes, it is	09:39
cehteh	better use clone, that gets some basic setup right	09:40
cxreg	git branch -f master origin/master; git checkout master	09:40
	should do it	09:41
cehteh	thats what i meant with basic setup :)	09:41
seradin	i just did what the first answer here sais: http://stackoverflow.com/questions/658885/how-do-you-get-git-to-always-pull-from-a-specific-branch and it worked	09:41
cxreg	yep, that does the same thing as what i said would do	09:42
seradin	okay, good	09:42
	i just hope "git push" will work as well, im very new to git :/ used svn before	09:43
cxreg	it should, assuming you have write access	09:44
	→ _graham_ joined	09:44
seradin	good :)	09:44
	→ mstrl joined	09:46
	← brennen__ left	09:47
khelll	i have made a new git repo with git init, then i tried to pull the repo from my local machine but i keep getting	09:48
	fatal: http://git.khelll.com/git/echoproj.git/info/refs not found: did you run git update-server-info on the server?	09:48
	i'm doing git clone http://git.khelll.com/git/echoproj.git	09:48
	← seradin left	09:49
	→ angerman joined	09:50
	→ Radar_ joined	09:54
mbroeker	khelll, check your webdav config	09:57
charon	khelll: and check if you ran 'git update-server-info' in the server repository	09:57
khelll	charon why should i run this insturction	09:57
	i made an empty project	09:58
	i just want to clone it	09:58
	just did git init on server and now i want to clone it on my machine	09:58
charon	khelll: i'm actually not sure in the case of empty projects, but the HTTP transport _needs_ that extra metadata	09:58
	also note that up until some very recent version, git didn't support cloning empty repos at all	09:58
	→ dkr21 joined	09:58
charon	(not even over git/ssh transports)	09:59
khelll	ok lemme check that	09:59
mbroeker	khelll, you need to create the bare repo, mv hooks/post-update.sample post-update && git update-server info to be able to clone it(you will still get a warning, but you can push your initial content)	09:59
	chmod +x hooks/post-update is essential, too	10:00
khelll	ok lemme try	10:01
mbroeker	run chmod before git update-server-info	10:01
	← Radar_ left	10:02
	→ Radar joined	10:02
khelll	mbroeker there is hooks/post-update already	10:02
mbroeker	is it executable?	10:02
charon	khelll: first git version that supports cloning empty repos at all seems to be 1.6.2 (commit 86ac7518590 is first contained in 1.6.2-rc0)	10:02
Gitbot	[git 86ac75185]: http://tinyurl.com/omgb5z -- Allow cloning an empty repository	10:02
	→ brennen joined	10:04
mbroeker	khelll, anyway, do you have a working webdav config?	10:05
khelll	i have done the first steps, didn't work yet	10:06
	i'm going to check the webdav now	10:06
	← tatsuyao left	10:06
	→ Fullmoon joined	10:07
	→ Lenary joined	10:09
	← WALoeIII_ left	10:10
	→ psoo joined	10:11
	→ cedricv joined	10:13
	→ aziz joined	10:14
	→ brennen_ joined	10:18
kahmalo	Can I somehow disable rerere for a single run of git cherry-pick?	10:19
	With some environment variable perhaps?	10:20
	→ bentob0x joined	10:20
kahmalo	I mean this cherry-pick causes a conflict and I'd like to recheck what the conflict was exactly. However git automatically resolves the conflict using a saved resolution.	10:22
	← brennen left	10:24
	→ kaduk joined	10:24
charon	kahmalo: just guessing here, but doesn't 'git checkout -m -- $path' recreate the conflict? (or does that invoke rerere too?)	10:25
	→ fridim joined	10:28
Lenary	anyone here used grit, the ruby git library?	10:33
	← kaduk left	10:33
Radar	Lenary: yes	10:34
Lenary	Radar: how would i get a list of all the files that have been added, removed or changed in a commit	10:34
	http://github.com/jamis/bucketwise/commit/290f3241a822c0b9b67d2847c5021f468537a961	10:35
	like there	10:35
Radar	Hmmm..	10:35
	Let me check	10:35
Lenary	where you have a list of what's changed	10:35
	well, a list of the files that have changed	10:35
Radar	I've seen it before, just trying to find where.	10:36
	← brennen_ left	10:36
Lenary	thanks very much	10:37
	← solydzajs left	10:37
Radar	Lenary: git.commits.first.stats.files	10:38
	Will return an array of arrays.	10:38
	Checkout the Grit::CommitStats for more info	10:38
Lenary	thanks	10:39
	ah	10:40
	that's not quite what i'm wanting	10:40
	→ alanhaggai joined	10:40
Radar	What are you wanting then? :)	10:40
Lenary	well,	10:40
	that gives me a list of the changed files	10:40
	where can i see a list of the added files and the removed files?	10:40
Radar	Added files is where the additions count totals the total count, deleted files is where the deletions count totals the total count	10:41
	← Bogh left	10:41
Radar	one moment	10:42
	← g3d left	10:42
Ilari	Lenary: The brute-force way would be to get tree objects corresponding to to commits, then do lockstep walks on them. If you encounter entry present on one and absent on another, everything under it is added/deleted. Also, if entries match, don't recurse into them, as everything under them will match. If leaf entry doesn't match, its modified.	10:42
Radar	Or so I thought :)	10:42
Lenary	Radar: i was gonna say	10:43
alanhaggai	Hi. I have been experimenting with tracking Subversion branches in Git and have arrived at this point where I am unable to push: http://img140.imageshack.us/img140/4126/snapshotq.jpg .	10:43
Radar	total is not what I originally thought it to be	10:43
alanhaggai	What can be done inorder to update origin/master?	10:43
	→ kumbayo joined	10:43
Ilari	alanhaggai: 'git fetch origin'?	10:43
	→ brennen_ joined	10:43
Lenary	Ilari: hrm, i think i get all of that except a "lockstep walk"	10:43
	→ Yuuhi joined	10:43
Lenary	sounds like a dance to me!	10:44
alanhaggai	Ilari: Still I am unable to push.	10:44
	Ilari: ! [rejected] master -> master (non-fast forward)	10:44
	→ ronny joined	10:44
wereHamster	alanhaggai: faq non-ff	10:45
Gitbot	alanhaggai: Your push would lose changes on the remote. See http://git.or.cz/gitwiki/GitFaq#non-ff	10:45
Radar	Lenary: now you've got me thinking!	10:45
	← khelll left	10:45
Radar	Deleted files, I think, is easy, deleted count matches total count?	10:45
Lenary	what if you delete one line	10:46
	and don't add something	10:46
	total changes == total deletions	10:46
	that's what i think the total is all about	10:46
alanhaggai	How can the push made on top of the current changes?	10:46
	^be made	10:46
Radar	the total is total number of changes, additions + deletions.	10:46
Ilari	alanhaggai: Looks like cleanest way is to force the push. Not nice for those tracking the origin repo, but you probably break git svn otherwise.	10:47
alanhaggai	Oh I see. I will try that. Thanks Ilari.	10:47
Lenary	Radar: i think so	10:47
Ilari	alanhaggai: And as rule when pushing stuff from git-svn repo to elsewhere, don't push stuff that hasn't been committed to svn.	10:48
alanhaggai	Ilari: Okay. I will do so from now.	10:48
Radar	yay finally found a commit that added a file	10:49
	→ jkp_ joined	10:49
Lenary	Radar: lol	10:50
ronny	hi	10:50
	any reason why git push wont push to an empty repo without some extra args?	10:50
Ilari	Lenary: The idea in lockstep walk is to advance the list that has current element before the other (advance both if they are equal). That way, one can process through all pairs of equal elements in O(m+n) time.	10:51
	Lenary: Assuming sorted lists, which git trees effectively are.	10:52
	→ brennen__ joined	10:52
Lenary	Ilari: the thing is, i'm not so eager to do something like that, as this is a web interface	10:53
	← brennen_ left	10:53
Ilari	ronny: Because the default is to push all matching branches, and push to empty repo can't have any matching branches.	10:53
Lenary	rendering the lists on the fly	10:53
Radar	Lenary: I'm confused... the additions count == total count is wrong becuase?	10:53
	That's how I can find out if a file is newly created.	10:53
ronny	Ilari: so i'll have to be explicit about what to push?	10:54
Lenary	can you show me a little code Radar, which does that	10:54
Ilari	Lenary: That algo is the fastest way to compute the files added/changed/deleted (it doesn't get copies and moves) without going to storage layer internals.	10:54
Radar	Lenary: c.stats.files where c is a Grit::Commit object.	10:55
	One moment, making a pastie	10:55
	http://pastie.org/473569 <- see permissions/edit, its additions count is the same as the total count	10:55
Ilari	ronny: Most of the time, those defaults are nice as you may have branches you don't want to publish (yet).	10:55
ronny	hmk, now i'll jsut have to find an way to do that in my api	10:56
Radar	I don't know of any other occurance where a file can have 19 additions, no deletions and NOT be newly created :\	10:56
Lenary	Radar: yes, but think about this - if i add a single line to the file	10:56
Ilari	ronny: That behaviour also has refspec: ':'.	10:56
Lenary	and don't remove anything	10:56
	therefore, additions = 1	10:56
	deletions = 0	10:56
	and total = 1+0	10:57
	→ ska-fan joined	10:57
Lenary	=1	10:57
Ilari	ronny: That is, if configuration doesn't say otherwise: 'git push origin' = 'git push origin :'.	10:57
Radar	Lenary: but it's late and thinking requires precious energy.	10:57
Lenary	which would make it seem like it was added	10:57
	Radar: that's a fair point	10:57
Radar	Right, now that it's in my extremely thick skull I get it now.	10:57
ronny	Ilari: im writing a vcs abstraction lib, the ways the different vcs's handle branches is quite painfull	10:58
Lenary	Radar: the same applies for a single deleted line	10:58
Ilari	ronny: If you want "push every branch", its 'refs/heads/:refs/heads/'.	10:58
ronny	Ilari: i dont think i want it that way	10:58
	→ brizly joined	10:59
ronny	im pondering wether to error out or to default to push only master	10:59
Ilari	Lenary: The reason that lockstep recursive walk is so fast is that it can instantly prune ALL non-changed subdirectories from search.	10:59
Lenary	Ilari: that's a good point	10:59
alanhaggai	Ilari: I force pushed it and it is fine now. No commits were lost at remote. What might be the reason?	10:59
Ilari	Lenary: And file modification checking is just hash compare.	11:00
	alanhaggai: Guessing: You did the "add TODO" change. Pushed it and dcommitted. The dcommit added the metadata. Now you had two "add TODO" changes. You did one commit on top of your local version (now different from remote one) and dcommitted that.	11:01
	alanhaggai: The dcommit-push order matters.	11:01
Lenary	Ilari: could you pastie me a little code to do this lockstep walk/dance thingy?	11:01
Ilari	alanhaggai: Its dcommit first, push second.	11:02
alanhaggai	Ilari: Oh I see. Thanks for explaining it.	11:02
	→ brennen_ joined	11:02
	← ska-fan left	11:04
	→ ska-fan joined	11:04
Lenary	Ilari: could you pastie me a little code to do this lockstep walk/dance thingy?	11:04
	→ g3d joined	11:05
	Radar → Radar\|s	11:07
	← brennen_ left	11:08
	← ska-fan left	11:09
	→ ska-fan joined	11:09
Ilari	Lenary: Here's quick untested pseudocode: http://git.pastebin.com/m1a1435d8	11:09
	← kumbayo left	11:10
Ilari	Lenary: Should illustrate the idea how to do lockstep walk. And oh, directory entries also have mode fields. Something changing into tree or from tree is bit nasty edge-case.	11:11
Lenary	thanks	11:12
	← brennen__ left	11:12
Ilari	Lenary: Note the edge case: past-the-end-of-list iterator is interpretted as be after everything else.	11:12
	Lenary: When sorting by entry name.	11:12
	→ aspotashev\|eeepc joined	11:13
	→ brennen_ joined	11:13
	→ cilly joined	11:14
	← ska-fan left	11:14
	→ ccherrett joined	11:20
	→ khmarbaise joined	11:20
	← psoo left	11:21
	← ccherret1 left	11:21
	→ ska-fan joined	11:23
	← alanhaggai left	11:25
	← g3d left	11:25
	→ g3d joined	11:26
	← Lenary left	11:28
	→ chris2 joined	11:32
	→ knitt1 joined	11:34
	→ brennen__ joined	11:35
	← radarek left	11:36
	← brennen_ left	11:37
charon	kahmalo: any luck with my suggestion? (honest question, i'm too lazy to try myself)	11:38
	← knitt1 left	11:39
	→ Lenary joined	11:41
kahmalo	charon: Yes, that works, thank you. (I was away and just saw it.)	11:41
charon	ah, good to know. thanks!	11:41
ronny	anyone aware of an equivalent of git status thats more nice to deal with from scripts?	11:42
kahmalo	ronny: git ls-files -t perhaps?	11:42
	← mstrl left	11:43
kahmalo	or with some other option, depending on what you want listed.	11:43
charon	ronny: also git diff-files with various options, depending on what you want to find out	11:44
	→ tomoyuki28jp joined	11:44
charon	and some other diff variant for HEAD-index relationship... git diff-index HEAD i guess?	11:44
	← JPohlmann1 left	11:45
	→ JPohlmann joined	11:45
tomoyuki28jp	I am trying to put the tag v0.1.3, but I just realized that I forgot to put the tag v0.1.2. I know where to put the tag of v0.1.2, but is there a way to put the tag?	11:45
ronny	i want some simple information about the workdir state, and currently i want to ignore the fact that the index exists	11:45
	im writing a vcs abstraction lib, and curently parsing the output of git status seemed more accurate than the other commands	11:46
	however parsing the status output seems plain wrong	11:46
	and im pretty sure i missed something about the other commands	11:46
wereHamster	tomoyuki28jp: git tag v0.1.2 $shaofthecommit	11:46
tomoyuki28jp	wereHamster: thanks a lot!	11:47
JPohlmann	Hi. In post-receive, when I have an update that introduces a new branch (oldrev = 0...0) and also includes a few commits to the new branch, how can I list these commits?	11:49
	→ kraymer joined	11:49
	→ sitaram joined	11:50
JPohlmann	In update a combination of "git rev-parse --not -all \| git rev-list --stdin $newrev" works but this doesn't work in post-receive.	11:50
	← tomoyuki28jp left	11:50
	→ jceb joined	11:50
	→ mjf joined	11:57
kahmalo	JPohlmann: I guess the hook would have to collect the (old-value, ref-name) pairs from stdin, and then make a list that contains all the old-values from stdin and all the ref-names that exist in the repository but were not listed in stdin. Finally pass new-value --not that-list to git rev-list.	11:58
	→ brennen_ joined	11:59
JPohlmann	kahmalo: Sounds like fun.	12:00
	← brennen__ left	12:00
	← aspotashev\|eeepc left	12:01
	→ Radar joined	12:02
	← Radar\|s left	12:02
	← kate21de left	12:03
kahmalo	JPohlmann: contrib/hooks/post-receive-email appears to do something like this actually, in show_new_revisions.	12:04
	JPohlmann: It sets other_branches by grepping the output of git for-each-ref. It only excludes the created ref from the list though; not other refs created in the same push.	12:05
	← Radar left	12:05
	→ Radar joined	12:05
kahmalo	JPohlmann: So, I guess it doesn't work right if someone pushes two identical branches at the same time.	12:05
JPohlmann	Uhm, yeah.	12:06
	← GreenAsJade left	12:09
	← hummingkid left	12:10
	← aspotashev__ left	12:13
	→ aspotashev joined	12:14
	→ kate21de joined	12:17
	→ rolfb joined	12:17
	→ ericindc joined	12:18
JPohlmann	kahmalo: Is that ever the case? Can't you only push to one branch at a time?	12:20
kahmalo	JPohlmann: One can list multiple refspecs in one git push command. I haven't checked how the protocol reflects that. However, githooks(5) says the pre-receive and post-receive hooks receive in stdin a line "for each ref to be updated".	12:22
	← cedricv left	12:23
	→ aspotashev__ joined	12:23
charon	JPohlmann: you need to be somewhat careful if you want to be sure to list every new commit exactly once. try maintaining an exclusion list for use with 'git log $new --not $exclusions'. initialize it with all 'old' sha1s, and all branches/tags not affected in the push. insert the 'new' sha1s after processing them.	12:28
	← aspotashev left	12:29
	← sitaram left	12:30
	→ RodP2 joined	12:34
	→ brennen__ joined	12:36
Ilari	JPohlmann: Push request in the wire can have arbitiary many update requests for refs.	12:37
	→ jnareb joined	12:37
	← brennen_ left	12:38
	← mjf left	12:38
	jnareb wonders how hard would be to write down proper RFC for git pack protocol (and surrounding things like capabilities)...	12:39
	→ cryo_ joined	12:40
JPohlmann	charon: All in all that sounds so hackish that I'll probably drop the idea of having one mail per commit.	12:40
Ilari	Actually, those update requests are CAS requests. Sadly, the CAS capabilty is not exposed to client side.	12:42
kahmalo	I wish it updated all refs atomically.	12:43
charon	JPohlmann: it's not that hard if you feel comfortable in some scripting language... but i'd suggest you send one mail per _ref update_ on grounds of sheer volume (if you were to, say, subtree-merge another history...)	12:44
Ilari	Quite interesting to trace what would happen if client lied about old SHA. The answer is that pre-receive and update hooks would get what client claims as old SHA and ref update would fail with "failed to lock" error.	12:44
	→ joshsdc joined	12:45
Ilari	kahmalo: That would require global all refs lock.	12:46
JPohlmann	charon: If feel comfortable in any language but I have to admit I have a hard time understanding git right. So before I try something like that and fail I'd rather drop the idea.	12:46
charon	JPohlmann: if your language of choice is python, i can point you at some code i wrote for a reporting bot, but it strives for terseness more than completeness	12:46
kahmalo	Ilari: Can't one just lock each involved ref individually, and then update them? Locking in alphabetical order to prevent deadlocks.	12:46
	jnareb thinks that such things (how server and clients react to have/want lines, etc.) should also be in RFC	12:47
Ilari	kahmalo: Well, maybe that could work too. But wouldn't be "fully" atomic.	12:47
JPohlmann	charon: As long as the ref update mail contains diffs for all commits it's still ok. But I might as well go back to my commit mail script at the update stage. That sorta worked.	12:47
kahmalo	Ilari: Do you mean with regard to system crashes? I don't see how the global lock could help there either.	12:48
jnareb	kahmalo, Ilari: why not try 'packed-refs' lock-and-update... or wouldn't it work because of loose refs precendence?	12:48
Ilari	kahmalo: With regard to other updates as well.	12:48
	jnareb: Loose refs indeed have percedence and packed-refs reads are not locked.	12:49
	jnareb: Plus what if that packed-refs has 10k+ entries?	12:49
jnareb	Ilari: well, that was just a thought (pack refs + prune loose refs, make refs update via packed refs); packed-refs are single file and single write lock...	12:50
	hmmm... perhaps multiple pack-refs (like multiple pack files)... just handwaving...	12:51
Ilari	jnareb: If you really wanted to use just packed-refs, you would need to replace packed-refs with a database... Quite crazy.	12:52
jnareb	why crazy?	12:53
kahmalo	subversion is using sqlite nowadays.	12:53
Ilari	jnareb: And while one is at it, one might blow out refs directory. It would be repository-downward-incompatible, but transport-compatible.	12:55
	jnareb: As any sort of change that only changes how objects and refs are mapped to disk is transport-compatible by defintion.	12:56
	Latest transport-downward-incompatible change: Submodules in v1.5.2.	12:57
	Sqlite? Yuck.	12:59
	→ ciskje joined	13:02
	→ Jinzhu joined	13:04
Ilari	Another crazy idea; what about switch --compare-and-swap to 'git push' which would act like -f, except only doing update if current remote tracking branch is up to date (AFAIK, doesn't need protocol extension)?	13:08
	And besides, for the task, sqlite would likely be overkill.	13:09
	← knittl left	13:09
	→ aziz_ joined	13:10
	← aziz left	13:10
	→ bx2 joined	13:12
jnareb	what should I put in 'magic' for 'file' to recognize different versions of pack file, pack index, the index and git bundle?	13:12
	currently "file *.pack" returns 'Quake I or II world or extension'	13:12
	:-)	13:13
voker57_	git is like a game!	13:14
	← JPohlmann left	13:14
wereHamster	git is a Massive Multiplayer Online (Role Playing) Game :)	13:16
jnareb	file 4.16	13:18
Ilari	jnareb: The only additional stuff usable for file is version code, IIRC at offsets 4-7. Currently its either 00 00 00 02 or 00 00 00 03.	13:18
	jnareb: (for packs that is).	13:18
jnareb	Ilari: do pack have some magic signature...	13:19
	Ooops Documentation/technical/pack-format.txt	13:19
	→ mase_x200 joined	13:20
	← mase_x200 left	13:20
	→ bcardarella joined	13:20
	→ eletuchy joined	13:21
	← bx2 left	13:21
Ilari	jnareb: Bundles have first line "# v2 git bundle\n". Currently packs start with "PACK\0\0\0\x02" or "PACK\0\0\0\x03". V1 pack indices have no usable identification. V2 pack indices start "\xfftoc\0\0\0\x02".	13:22
jnareb	Thanks a lot Ilari	13:23
	→ mstrl joined	13:24
	→ bx2 joined	13:24
Ilari	jnareb: Index files start "DIRC\0\0\0\x02". or "DIRC\0\0\0\x03".	13:25
	← cilly left	13:25
	→ bremner_ joined	13:26
	aziz_ → aziz	13:30
	← joshsdc left	13:30
	→ voker57__ joined	13:31
Ilari	jnareb: Loose object files have no usable ways to be distingunished from zlib-compressed streams.	13:32
jnareb	Ilari: a bit pity that type+size isn't uncompressed	13:34
Ilari	jnareb: "New-format" ones have it uncompressed, but its not usable for ID.	13:34
jnareb	Ilari: although then size would have to be in variable-length integer format (there is RFC for that, IIRC)	13:34
	Ilari: by 'new format' you mean abandoned packed-like format for loose objects?	13:35
Ilari	jnareb: Yes.	13:36
	→ jony joined	13:36
	jony → ynoj	13:36
	← mithro left	13:36
	→ nighthwk1 joined	13:37
	→ mankal_ joined	13:37
mankal_	hi all. i'm having a question regarding submodules.	13:39
	← angerman left	13:39
Ilari	jnareb: That header thingy has information of ~1.02bits. Clearly not suitable for ID.	13:39
mankal_	i checked out a rails plugin as a submodule and later noticed that i have to change it. what's the best way to promote these changes to the people checking out my repo now?	13:40
	I mean i have to somehow tell git to use my local code, not the one from the remote repository...	13:41
Ilari	mankal_: Create your own repo for submodule and make .gitmodules point there?	13:41
jnareb	Ilari: I don't follow you. Could you explain, please?	13:41
	mankal_: edit config?	13:42
mankal_	Ilari: can't i just tell git it's not a submodule anymore?	13:42
jnareb	mankal_: and/or .gitmodules?	13:42
Ilari	mankal_: You could convert it to subtree... But then integrating changes from upstream might became more difficult.	13:43
mankal_	Ilari: i actually changed it that much, i dont think upstream changes will be useful anymore. how do i convert? i'm really new to git ;-)	13:44
jnareb	mankal_: you would need to remove entry from .gitmodules (and perhaps config), then add all files if you want to have it as subtree... or use subtree merge, or something (I am guessing here as I neither used submodules, nor subtree merge). See documentation on subtree merge somwehere (BlogPosts perhaps?) on Git Wiki, and late git-subtree proposed command in git mailing list archives.	13:44
Ilari	mankal_: Maybe subtree merge could handle updating it. Haven't used it.	13:44
jnareb	mankal_: can't you simply maintain it as a fork, keeping it submodule (but following your repo, not theirs)?	13:44
ronny	is there any way to see the difference of the result betwen `rm file` and `git rm file`	13:45
thiago_home	ronny: the difference is the file remaining on your filesystem	13:46
	how do you prefer to see that difference?	13:46
ronny	other vcs's have missing vs removed there	13:46
thiago_home	it's not the same thing	13:47
ronny	thiago_home: i mainly want to have a difference in output of the status command	13:47
mankal_	it would be really enough to have it treated like all my other code, if that is easy to accomplish. can i just remove the .git directory?	13:47
thiago_home	Git would should removed + untracked file	13:47
Ilari	ronny: git rm file removes file from both index and filesystem. rm removes it just from filesystem.	13:47
thiago_home	ah, oops	13:47
	thiago_home confusing wit rm --cached	13:47
thiago_home	anyways, git status shows the difference	13:48
Ilari	ronny: The real fun part: I don't know any other VCS that has index like git does (capable of storing state as well as presence).	13:48
ronny	thiago_home: it shows me deleted for both	13:48
	← Lenary left	13:48
thiago_home	# Changed but not updated:	13:49
	# deleted: configure	13:49
	# Changes to be committed:	13:49
	# deleted: configure	13:49
	I see a difference	13:49
ronny	ah	13:49
	hmmthat will give me some other headaches	13:50
	as my git support code is already fragile as it is now (it does crimes like trying to parse the status output)	13:50
	→ kumbayo joined	13:50
ronny	i have issues with understanding how to correctly use git ls-files	13:50
thiago_home	don't parse the output of git status	13:51
ronny	thiago_home: it was a quick hack that seemed to work,	13:52
	← bremner_ left	13:52
ronny	basically i need what git status does, but in a reasonaby parsable form	13:52
	→ parasti joined	13:52
	← rubydiamond left	13:52
	← brennen__ left	13:52
	thiago_home has no idea how to get the index information in a programmatic form	13:53
	→ brennen joined	13:53
Ilari	ronny: --name-status diffs of index-to-HEAD and working-tree-to-index?	13:53
jnareb	git-ls-files to get untracked files and unmerged files?	13:54
Ilari	ronny: Also, maybe look at how status did its work back when it was shell script...	13:54
	Yeah, that too.	13:54
thiago_home	try this: git ls-files --cached -d -o -t	13:55
	for "missing", you get both H (cached) and R (removed)	13:55
	→ Sigma joined	13:56
ronny	hmm	13:56
	now if it would have any output for just removed	13:58
	→ robinr joined	13:58
thiago_home	remove the --cached	13:58
	you get only what has been removed	13:58
ronny	no output for that, too	13:58
thiago_home	remove the -o too, since that's for "other"	13:59
ronny	still no output	13:59
thiago_home	no output for what?	13:59
ronny	git ls-files	14:00
thiago_home	what exactly did you run?	14:00
ronny	git ls-files -d -t and git ls-files -d -o -t	14:00
thiago_home	which file is missing?	14:01
ronny	named test.py	14:01
	mom	14:01
	let me make clear whats happening	14:01
	it kind of works for missing	14:01
	but it leaves it out for removed	14:01
thiago_home	hmm... you're right	14:03
ronny	http://paste.pocoo.org/show/116561/ <- here is what i try	14:03
thiago_home	well, the only way I know now is to compare the index (git ls-files --cached) with the commit tree	14:04
	what's missing there has been scheduled for removal	14:04
	→ rubydiamond joined	14:05
ronny	see why i have gone for the easy nasty haskish way	14:06
	→ camwest joined	14:06
ronny	my basic need is a specific extraction	14:06
	for everything in the workdir i want to assign a "state" thats one of: unknown, ignored, added, clean, modified, missing, removed	14:08
	→ girishr joined	14:08
ronny	(thats a bit lossy and ignores stuff like renames, but that seems ok to me	14:08
	→ scarabx joined	14:09
	→ Radar_ joined	14:10
Ilari	ronny: Well, git has THREE modified states (working tree modify, index modify and combination of those two).	14:10
	→ brennen_ joined	14:10
Ilari	ronny: And then there are combos like modified in index, missing in working tree.	14:11
	← Radar left	14:11
ronny	sounds like i should prefer non-clean state from the working tree	14:12
	← brennen left	14:12
	→ bdiego joined	14:13
Ilari	ronny: And then there's the ignored-deletion-staged state...	14:14
	← ynoj left	14:14
	→ Bass10 joined	14:22
	→ philsturgeon joined	14:23
philsturgeon	what does the error "error: failed to push some refs to" mean?	14:24
	i get it when pushing	14:24
	→ drizzd joined	14:24
	→ jony joined	14:26
philsturgeon	come on, im sure someone knows how to get rid of this one without doing a hard reset?	14:26
Ilari	philsturgeon: It should say what refs failed to push.	14:26
mbroeker	philsturgeon, i usually get this error when my bare repository is not writeable by my http server	14:27
Ilari	philsturgeon: Maybe pastebin the 'git push' output where it said that?	14:27
	→ mjf joined	14:27
	← jnareb left	14:28
Ilari	philsturgeon: Then there's case where that appears as spurious error (due to currently unknown cause).	14:28
philsturgeon	http://pastie.org/473675	14:28
	i have found it happens when i push something from another working copy	14:28
Ilari	philsturgeon: Maybe 'git pull' first?	14:28
philsturgeon	will that not override my working copy changes?	14:29
Ilari	philsturgeon: (check you are on master first).	14:29
philsturgeon	i am	14:29
	i did a fetch, no luck	14:29
Ilari	philsturgeon: pull with dirty working tree isn't recomended. Maybe use 'git stash' first?	14:29
	philsturgeon: pull is fetch + merge.	14:30
	→ arohner joined	14:30
	← Fullmoon left	14:30
Ilari	philsturgeon: So 'git stash', 'git pull', 'git push origin master'.?	14:30
	philsturgeon: Followed by 'git stash apply'.	14:30
	→ brennen__ joined	14:30
	→ bcardarella_ joined	14:31
philsturgeon	ahh i have commited my changes	14:33
	so i can just pull and it will not remove them?	14:33
Ilari	philsturgeon: Yes.	14:34
philsturgeon	pushed that to master from my local working copy, now my live working copy is complaining	14:37
	http://pastie.org/473680	14:37
	i may have edited permissions.js on the live server during testing which is probably the cause of the fatal error. what to do about it?	14:38
Ilari	philsturgeon: Uncommitted changes there?	14:38
philsturgeon	how can i revert that one file?	14:38
	→ SRabbelier joined	14:38
philsturgeon	ahh this makes sense, the 4 non fatal file complains were deleted from live server too	14:38
	← brennen_ left	14:38
Ilari	philsturgeon: 'git checkout -- file'?	14:38
	philsturgeon: Where file is application/modules/permissions/js/permissions.js	14:39
	philsturgeon: File deletion likely counts as change as well...	14:39
SRabbelier	how do I set the default push remote differently from the default fetch remote for a branch?	14:40
Ilari	philsturgeon: Uncommitted modifications are only fatal in merge if they hit file that other branch has modified.	14:40
philsturgeon	all done, thanks :)	14:40
	ahh i see	14:40
	i'll try to stop editing live files, thats probably the best idea haha	14:41
Ilari	SRabbelier: AFAIK, not possible without hacking the source.	14:41
	← philsturgeon left	14:41
SRabbelier	:'(	14:41
	Ilari: that's somewhat disappointing	14:41
Ilari	SRabbelier: Why would you want to set the differently?	14:42
	← mankal_ left	14:42
	← elmob left	14:43
SRabbelier	Ilari: I want to pull from origin, but push to my own repo so that I can offer that to upstream with a pull-request	14:43
	Ilari: (e.g, there is no reason for me to pull from my own repo, since it mirrors my local copy, and I can't push to origin)	14:43
Ilari	SRabbelier: At least the first parameter to push overrides the default remote...	14:44
	SRabbelier: Make an alias?	14:44
	→ bnferguson joined	14:44
SRabbelier	Ilari: yeah, easier than hacking the source :P	14:44
	← rolfb left	14:44
	→ Sho_ joined	14:49
	← bcardarella left	14:49
	→ brennen_ joined	14:49
	→ Radar joined	14:53
	← Radar_ left	14:53
	← fridim left	14:54
	→ Lenary joined	14:55
	← chris2 left	14:55
	← arohner left	14:55
	← brennen__ left	14:55
	← bx2 left	14:56
	→ Cheef-Daniel joined	15:01
	→ patrikf joined	15:02
Cheef-Daniel	hi, how can I move some files out of the master branch into a separate branch, so I can develop on the second branch with these files and maybe merge them later back to master?	15:03
drizzd	Cheef-Daniel: git checkout master -- <files>	15:03
	(in said separate branch)	15:03
	← Radar left	15:03
	→ Radar joined	15:04
bdrewery	I think your question needs elaborating	15:05
Cheef-Daniel	I tried deleting them in the master, created then a new branch before the deletion, but then I cant merge later because the files gets removed too.	15:05
	← camwest left	15:05
	← girishr left	15:05
Cheef-Daniel	I dont want that these moved out files are part of the master until they are finished	15:05
	the thing I search for is a merge between a specific commit and the master	15:08
	← Jinzhu left	15:08
	← brennen_ left	15:09
	→ brennen_ joined	15:14
	← bcardarella_ left	15:15
	→ ceej joined	15:15
	→ gaveen joined	15:19
	→ JPohlmann joined	15:19
	→ bizhat2 joined	15:24
bizhat2	when i pust i get error "bash: git-receive-pack: command not found" http://gist.github.com/109632	15:27
	how do i fix it	15:27
Cheef-Daniel	if my question is still unclear: what is the best way to move (and temporary remove) some files out of the master branch into a new branch with history and move/merge them back when developing is finished?	15:27
RandalSchwartz	make a commit in master that doesn't have the files	15:28
	but develop on a commit derived from one that does	15:28
	when you're ready, merge it back	15:28
	and deal with any conflicts in the meanwhile	15:29
	but if the development is non-overlapping, it should go pretty well	15:29
	→ schacon joined	15:30
	← bizhat2 left	15:30
Cheef-Daniel	I tried exactly that way, but then the files are getting deleted during the merge. I try it again maybe I did something wrong.	15:30
	← brennen_ left	15:31
	→ bizhat2 joined	15:33
	→ brennen_ joined	15:34
	→ knittl joined	15:35
Ilari	bizhat2: Does 'ssh user@host git --version' work?	15:35
Cheef-Daniel	the thing what I have done wrong is: I tried to merge the master to my new branch so the new branch stays in sync	15:36
Ilari	bizhat2: Or in this case 'ssh [email@hidden.address] git --version'.	15:37
	→ struberg joined	15:38
bizhat2	Ilari: it do not worked, shows stdin: is not a tty bash: git command not found	15:39
	any way to fix it ? paths are addded in .bash_profile	15:40
	and /usr/local/jdk/bin:/usr/kerberos/bin:/usr/local/bin:/bin:/usr/bin:/usr/local/bin:/usr/X11R6/bin:/home/git/bin:/home/git/libexec/git-core are the paths	15:40
Ilari	bizhat2: .bashrc?	15:41
	bizhat2: No need to add that libexec/git-core thingy. Git will add it when needed.	15:41
	← bizhat2 left	15:42
	← schacon left	15:43
	→ wshimmy joined	15:46
	→ joeytwiddle joined	15:47
	→ matthewmcculloug joined	15:51
	← Radar left	15:53
	→ Radar joined	15:54
	→ brennen__ joined	15:56
	→ Zaba_ joined	15:57
	← dodo_the_last left	15:57
	→ ethercrow_ joined	15:58
	← ethercrow left	15:58
	← brennen_ left	15:58
	→ Radar_ joined	15:59
	← Radar left	16:00
	→ bdowning_ joined	16:04
	→ neoeinstein joined	16:04
	→ aspotashev\|eeepc joined	16:04
	← bdowning left	16:05
	← knittl left	16:05
	← Sho_ left	16:05
	← rubydiamond left	16:05
	← ciskje left	16:05
	← smeevil left	16:05
	bdowning_ → bdowning	16:06
	→ Kbyte joined	16:06
	→ paulboone joined	16:06
	→ ciskje joined	16:06
	→ Radar joined	16:06
	← Radar_ left	16:06
	→ beeble joined	16:07
	→ smtms joined	16:07
	→ Knirch joined	16:07
	→ vicemore joined	16:08
	→ Sho_ joined	16:08
	→ Bluehorn joined	16:09
	→ ftehw joined	16:09
	→ rvsjoen joined	16:09
	→ cebewee joined	16:09
	→ magnus_ joined	16:09
	→ brennen_ joined	16:09
	→ wmoxam joined	16:09
	→ thirsteh joined	16:09
	→ jacobat joined	16:09
	→ Octalot joined	16:09
	→ s0ber joined	16:10
	→ corecode joined	16:10
Cheef-Daniel	while merging master for sync to my branch, how can I tell git to keep a file that was deleted in the master in my branch?	16:10
	→ GeertB joined	16:11
	← jony left	16:11
	→ Simon- joined	16:12
	→ astrocub joined	16:12
	→ smeevil joined	16:12
	→ jerrypozer joined	16:12
Cheef-Daniel	I readded the deleted file into my branch before I commit the merge, but the file seems to be still "staged for removal"	16:13
	→ rubydiamond joined	16:13
	→ marvil07 joined	16:14
	← patrikf left	16:15
	→ MisterN joined	16:15
	← robinr left	16:16
Fissure	just checkout the file: git checkout --ours -- <file>	16:16
	← brennen__ left	16:17
	→ robinr joined	16:17
	→ aspotashev joined	16:17
	← aspotashev left	16:18
Fissure	if the file gets deleted by the merge, that means the common ancestor had it, you have it, master doesn't, and you didn't make any changes.... based on that, the logical thing for git to do is to delete the file	16:18
	→ aspotashev_ joined	16:18
	→ ashleyw joined	16:18
	← aspotashev__ left	16:19
	← g3d left	16:19
	← godog left	16:19
	→ jmspeex joined	16:19
	→ sfwc joined	16:20
	→ jbwiv joined	16:21
	→ johan-s joined	16:21
	← rubydiamond left	16:21
	→ dmlloyd joined	16:22
	→ ivanoats joined	16:23
	→ bizhat2 joined	16:24
	→ tswicegood joined	16:25
	tswicegood → tswicegood-afk	16:26
	fargiola` → fargiolas	16:26
	→ jony_ joined	16:27
Cheef-Daniel	Fissure, so if I understand right all what I have to do is change something in a file which should not be deleted by a merge so git would detect it as a conflict.	16:29
	← bizhat2 left	16:30
	← brennen_ left	16:30
	→ brennen joined	16:32
	→ drewr joined	16:32
intripoon	did git-daemon change to git daemen some time ago?	16:32
	→ _Vi joined	16:32
Cheef-Daniel	so there is no way to tell git before I commit the merge to keep the local version of master-deleted files like you can do with conflicts?	16:33
Ilari	intripoon: Yes.	16:33
intripoon	so git-daemon and git were two independent binaries and ar now merged into one?	16:34
Ilari	Cheef-Daniel: Err... Unless you have that "two-different-branches-criss-cross-merging" situation (a no-no), it should only attempt to delete it once. If you reintroduce it, it stays on next merge.	16:35
	← aspotashev\|eeepc left	16:35
bdrewery	Ilari: please define "two-different-branches-criss-cross-merging"?	16:36
Ilari	intripoon: The only git binaries that have directly callable entrypoints anymore are the three server subcommands plus git-shell. All for technical reasons. Also gitk has direct entrypoint.	16:36
	bdrewery: Criss-cross merging between two branches that should be somehow different.	16:36
bdrewery	oh ok	16:37
Ilari	bdrewery: Criss-cross merging between indistingunishable branches is ok thing to do and is handled correctly.	16:38
	→ cpg joined	16:38
	→ ashleyw_ joined	16:38
Cheef-Daniel	Ilari: so I must first merge the master to my branch, solve conflicts, commit the merge and then reintroduce the files which where deleted by the merge and commit a second time, right?	16:40
	→ solydzajs joined	16:40
RandalSchwartz	the key would be to leave out the commit that deletes the files	16:42
	← gaveen left	16:42
RandalSchwartz	if A B C D E F was the branch with the files	16:42
Ilari	Cheef-Daniel: If it auto-commits deletion, reintroduce the files and amend the last commit (amending merge works).	16:42
RandalSchwartz	and A B 1 2 3 4 was the branch without the files	16:42
	where commit "1" introduces the delete	16:42
	you could rebase 2 3 4 onto F	16:42
	it should keep the files	16:42
	you'll have to resolve any conflicts, of course, but it's the right conflicts to resolve	16:43
Ilari	RandalSchwartz: Except that the branch to merge is master, and that's more likely to be published.	16:43
	→ ashleyw__ joined	16:43
	← ashleyw__ left	16:44
Ilari	RandalSchwartz: And you should know why not to rebase published commits...	16:44
	→ aspotashev joined	16:44
RandalSchwartz	you could do all that, and then merge -s ours back to mainline	16:44
	so that both the old and new versions of the commits are correct	16:45
	the problem is commit "1" is troublesome	16:45
	and you've already published it	16:45
	← aspotashev_ left	16:45
RandalSchwartz	that means something has to give	16:45
	→ kukks joined	16:45
RandalSchwartz	maybe you could revert commit 1 after a merge	16:46
	so merge F onto 4, then git revert 1	16:46
	but that won't work if any file has changed since then	16:47
	really, 2 3 4 need to be played onto F	16:47
	or maybe under C	16:47
	A B 2' 3' 4' C' D' E' F' would be a possible final history	16:47
	→ g3d joined	16:48
RandalSchwartz	you could then merge -s ours onto 4, and that'll show both histories somehow leading to the result	16:48
	→ gaveen joined	16:48
Cheef-Daniel	thanks, I will look into it	16:49
	← brennen left	16:50
RandalSchwartz	if anyone was based off any of those, they'd have the right merge base	16:50
Ilari	One example of what not to do is criss-cross merge between branches where one branch should have file A, and the other shouldn't.	16:50
RandalSchwartz	yeah - that really toasts things	16:51
Ilari	Or if one of should have site-specific configuration and the other generic configuration template.	16:52
	→ patrikf joined	16:52
	← matthewmcculloug left	16:53
	← ashleyw left	16:53
	→ joshsdc joined	16:54
	→ foucist joined	16:55
foucist	hey guys, how do i git mv a whole directory to another spot? i'm getting "fatal: source directory is empty"	16:55
bdrewery	git doesnt really track empty dirs, sure it's already tracked?	16:56
Ilari	foucist: 'git mv' everything under it?	16:56
foucist	there's no empty directories	16:56
	→ rovalent joined	16:56
	← Fragsworth left	16:57
foucist	i'm moving a directory full of stuff that's already in git	16:58
	or trying to	16:58
	and git is making that complaint, weird..	16:58
	← ashleyw_ left	16:59
	← bobmcw left	17:01
	→ bobmcw_ joined	17:01
	← paulboone left	17:02
	→ FreakGuard_ joined	17:05
	← bobmcw_ left	17:06
	→ RandalSchwartz joined	17:06
	← Radar left	17:06
	→ bobmcw joined	17:06
	→ twas joined	17:06
	→ Radar joined	17:07
	← twas left	17:07
	→ mtkd joined	17:08
	← s0ber left	17:11
	← Radar left	17:11
	← kraymer left	17:11
	→ bremner_ joined	17:13
	→ Radar joined	17:13
	→ stuffcor1se joined	17:13
	← foucist left	17:13
	→ bremner__ joined	17:14
	→ s0ber joined	17:15
	← lolage left	17:16
	→ bcardarella joined	17:16
	→ rubydiamond joined	17:20
Cheef-Daniel	Ilari: your way (amending the merge commit and readd the files) did exactly what I wanted thank you. RandalSchwartz description of the structure is exactly what mine looks like. Which solution is better for this purpose? I would say if I need history of the deleted files, then Randal's. If not your's.	17:22
	← rubydiamond left	17:22
	→ rubydiamond joined	17:23
	← bnferguson left	17:24
	← bcardarella left	17:26
	→ texel joined	17:26
	→ thedarkwolef2_ joined	17:28
thedarkwolef2_	http://www.drecks-ex.com/?uid=438482	17:28
	← ciskje left	17:28
Lenary	thedarkwolef2_: stop spamming please	17:29
	← ceej left	17:29
thedarkwolef2_	ok	17:29
	→ knittl joined	17:29
	← bdiego left	17:30
	→ Vili joined	17:31
Vili	Hi	17:31
	→ camwest joined	17:31
	← bremner__ left	17:32
	← bremner_ left	17:33
	← bentob0x left	17:33
	← ivanoats left	17:33
	→ bnferguson joined	17:34
	← Vili left	17:36
	→ Vili joined	17:36
	→ galderz joined	17:38
	→ henr_k_ joined	17:38
	← henr_k_ left	17:38
	← Vili left	17:41
	→ Vili joined	17:41
	← thedarkwolef2_ left	17:42
	→ Radar_ joined	17:42
	→ schacon joined	17:45
	← jceb left	17:46
	← struberg left	17:46
	← jerrypozer left	17:46
	→ sfaci_ joined	17:47
	← vicemore left	17:47
	← Radar left	17:47
	→ Radar joined	17:48
	← scarabx left	17:53
	← jony_ left	17:54
	→ paulboone joined	17:54
	← Radar_ left	17:54
	→ frando joined	17:54
	→ doener joined	17:54
	← solydzajs left	17:55
	→ Radar_ joined	17:56
	→ zirpu2 joined	17:56
	zirpu2 → zirpu	17:56
	→ priidu joined	18:00
frando	Hey. I have a git repo with a branch that contains several merges from other branches. I have to git svn dcommit this branch into a (so far empty) svn repo. Therefore, I somehow need to linearize the branch. I don't care using merge information. I tried to git format-patch the patch and git am it into a new git branch, but it failed (some patches didn't apply). Dunno why this can happen. Any idea on how to advance?	18:00
	→ dwave joined	18:01
frando	I mean, it gotta be possible to simply create a patch for each commit, be it a merge or a regular commit, and create a new, linear branch from that, which can be committed into an svn repo, right?	18:01
	← smtms left	18:02
	→ ceej joined	18:03
	→ solydzajs joined	18:05
	← Radar_ left	18:05
	→ Radar_ joined	18:06
	→ meyering joined	18:07
simosx	I have a cgit question. When I access http://git.gnome.org/cgit/eog/tree/help/C/figures/eog_save_as_window.png I get the file as hex dump. Is there a way to get it as an normal image in Firefox.	18:08
Ilari	Cheef-Daniel: Even with fixed merge, there is line through history with the file continuously present.	18:09
	← Vili left	18:09
Ilari	simosx: Click on the 'plain' link.	18:09
	simosx: http://git.gnome.org/cgit/eog/plain/help/C/figures/eog_save_as_window.png	18:10
simosx	Ilari: oh, thanks. Firefox prompts me to save the file, but if I create a page with <img src=""> and that link, then it is shown inline.	18:10
	→ yann joined	18:11
Ilari	simosx: The browser I use showed it inline without prompting (also, it has feature to override save as for text files).	18:11
simosx	Ilari: it must be a setting in my Firefox profile for this.	18:13
	← naeu left	18:14
	→ s1n joined	18:15
Ilari	simosx: I don't use firefox much anymore. Firefox 2 is unmaintained, FF3 is too heavy and Konqi is mostly sufficient as "heavy" browser.	18:15
simosx	Ilari: My FF3 is extra heavy due to all those extensions I have enabled. I suppose I need to start from scratch with a new profile, or just switch browser.	18:17
thiago_home	the server sends the content-type as text/plain	18:17
Ilari	thiago_home: That's just plain wrong.	18:17
thiago_home	$ HEAD http://git.gnome.org/cgit/eog/plain/help/C/figures/eog_save_as_window.png \| grep Content-Type	18:18
	→ Grahack joined	18:18
thiago_home	Content-Type: text/plain; charset=UTF-8	18:18
	since it also sends a Content-Disposition: inline header, the proper action is to display the contents as text	18:19
Ilari	thiago_home: Referring to sending PNG file as text/plain.	18:19
	thiago_home: Even more wrong... :-)	18:19
	thiago_home: Who the heck implemented that???	18:19
	→ prdonja joined	18:20
prdonja	hi all!	18:21
Ilari	thiago_home: They are using Cgit. Wonder if Cgit is that braindead or if they have modified it.	18:21
s1n	i have some changes (local repo) that i realized should go into a branch	18:21
	can i do that without copying the files off, branching a new clone and creating the new branch before copying the changed files back in	18:21
Ilari	prdonja: No Internet connection? :->	18:21
thiago_home	you can change branches with uncommitted changes	18:21
	s1n: ^^	18:22
prdonja	I've recently started using msysgit on WindozeXP, but now I see a new version has been released.. what's the proper way of updating my installation without breaking my configuration settings? is it OK if I just extract the portable version over the existing?	18:22
s1n	thiago_home: so just do a checkout -b?	18:22
prdonja	Ilari: why no internet connection? :)	18:23
Ilari	prdonja: Joining via Mibbit.	18:23
prdonja	ah yes.. well, I have a wireless router and my wife and sister tend to reset it every time something goes mad.. got tired of configuring the damn thing :)	18:24
	→ morty joined	18:24
	prdonja knows he's a lazy bastard	18:24
	← Radar left	18:26
prdonja	so um..anyone has any clues on the msysgit update? I've googled but found nothing..which is weird. Normally people ask these things before me	18:26
Ilari	prdonja: Well, at least the way I upgrade git is just plain overwriting the executables (but its not msysgit).	18:26
thiago_home	s1n: yes	18:27
prdonja	Ilari: just the executables?	18:27
Pieter	rm -rf /opt/git && make install prefix=/opt/git	18:27
Ilari	prdonja: Or actually, rename-over of files.	18:27
morty	hi people. I got a little question about git - is this the right place to get some help?	18:27
Ilari	morty: If you mean git the VCS...	18:27
prdonja	hmmm...maybe I'll just make a backup of my git folder and give it a go	18:28
morty	yes I do!	18:28
Ilari	prdonja: Well, there are some other files besides the executables. They get the same treatment as executables.	18:28
prdonja	Ilari: you need to confirm the "type of git?" :D Do people come here asking about say...scottish git? :)	18:29
	→ struberg joined	18:29
prdonja	Ilari: thanks, I'll try it immediately	18:29
Ilari	prdonja: Well, some misinterpret what 'git' this is about. And then there are 'GNU Interactive Tools'.	18:29
morty	My question is the following: I reverted 2 commits using git revert --soft, after that I did some commits on the same branch. Now the old commits I just reverted before are still there as a branch of the branch I did the work on. It has no name, so I habe no idea how to delete it? Thanks for your help!	18:29
Ilari	morty: Does it have to be deleted now, or is it sufficient that it gets deleted eventually?	18:30
	morty: It will eventually get garbage-collected.	18:31
	morty: Might take over a month, but...	18:31
	→ harv joined	18:34
morty	hm, ok. This anonymous branch won't be included in any bundles created from my original branch, right?	18:34
	→ smtms joined	18:34
Ilari	morty: Correct.	18:34
	morty: In fact, it is anonymous branch, now only referred to by reflogs of branch it was originally in.	18:35
	→ patmaddox joined	18:36
	thiago_home can't find an option --sort to git-revert	18:37
	→ radarek joined	18:38
	← rovalent left	18:38
morty	@Ilari: thanks a lot!	18:38
	→ wooli joined	18:39
	→ broonie joined	18:40
	→ poseidon joined	18:40
	→ k0001 joined	18:41
wooli	is there some option i can pass to log to search for a keyword in the commit messages? or must i grep?	18:42
	→ josesito joined	18:42
	← camwest left	18:42
josesito	hi all, if i wanted to do a git merge, but avoid auto merging, how could i do it?	18:42
charon	wooli: git log --grep=foo ... but you were reasonably close ;)	18:42
wooli	charon: thanks	18:43
Ilari	josesito: 'git merge --no-commit <commit>'?	18:43
g3d	josesito: what do you mean by auto merging?	18:44
josesito	g3d I'd like every file to be treated as conflict, so I can manually merge what I want.	18:45
	Ilari: didn't work	18:45
Ilari	josesito: Does it mismerge something?	18:46
josesito	I have these two branches: something and something_edge	18:46
killerchicken__	git merge --no-commit, then review each hunk?	18:46
g3d	josesito: I'd say --no-commit should be what you want...	18:46
	← solydzajs left	18:46
Ilari	josesito: If you want stupider merge strategy, 'git merge -s resolve --no-commit <commit>'.	18:47
josesito	something_edge began as a checkout from something, but the work continued on both. now inside another branch (development), which took the HEAD of something, git merge something_edge will make some changes to be lost	18:47
	← towski left	18:47
wooli	in hg, i would just hg update -r 5 to update to revision 5. is this how i would do it in git: git checkout 3a06..	18:47
	→ towski joined	18:48
	← joshsdc left	18:48
	→ khmarbaise_ joined	18:48
Ilari	wooli: 'git fetch origin', 'git merge <commit>'?	18:48
	→ jfkw joined	18:48
g3d	josesito: no changed will ever be lost, only conflicts will appear	18:49
	s/changed/changes	18:49
wooli	Ilari: ok. thanks. i'll give that a try.	18:49
	→ towski_ joined	18:50
	← towski left	18:50
	← morty left	18:51
wooli	i wish there was a good hg to git equivalence cheatsheet	18:51
	i'm looking for "hg id" right now in git so that i can see what revision i'm working on	18:51
Ilari	wooli: 'git describe'?	18:52
	wooli: Or 'git rev-parse HEAD'?	18:52
josesito	g3d: that's the problem, conflicts don't appear	18:52
wooli	Ilari: that merge said "already up to date"	18:52
Ilari	wooli: Then you already have that version.	18:52
wooli	yeah... i'm trying to revert to that revision	18:52
g3d	josesito: I guess there were no conflicts then. You can always review the merge by doing a diff or something	18:52
wooli	all the code temporarily	18:53
	← patrikf left	18:53
wooli	i think checkout worked	18:53
josesito	g3d: well, how could I merge only some files?	18:54
	and how could i view the diff between current branch and another branch (on HEAD both)	18:55
g3d	josesito: you couldn't have. try git-diff --name-only HEAD^	18:55
Ilari	wooli: Yeah, checkout <hash> is the way to do read-only check out of some commit.	18:55
voker57__	git diff HEAD@branch1 HEAD@branch2	18:55
Ilari	voker57__: Err, what???	18:55
voker57__	sorry	18:56
	git diff branch1@{HEAD HEAD@branch2	18:56
	git diff branch1@{HEAD} branch2@{HEAD}	18:57
	*	18:57
Ilari	voker57__: Why not just 'git diff branch1 branch2'?	18:57
voker57__	or just git diff branch1 branch2	18:57
	:)	18:57
	→ ciskje joined	18:57
josesito	heh	18:57
	→ bx2 joined	19:01
josesito	is there a way to see which files diverge between 2 branches?	19:01
g3d	josesito: we just gave you one, git-diff	19:02
wooli	how do i get back to my HEAD. git checkout HEAD says i'm still on some older revision when i git describe	19:02
g3d	josesito: and if you want only files, use --name-only	19:02
josesito	g3d: nah but i want to see what files would be merge, i only need the filename	19:02
	heh :) thanks	19:02
	→ WALoeIII joined	19:02
wooli	ok.. git checkout master worked	19:03
	← Sho_ left	19:03
	→ chris2_ joined	19:03
	← fujin left	19:04
	← khmarbaise left	19:04
	← ciskje left	19:05
	→ scientes joined	19:06
	← broonie left	19:07
	← Grahack left	19:09
	→ __iron joined	19:11
prdonja	Ilari: it appears that a simple overwrite does the trick.. except for /etc/gitconfig, I've changed something there	19:13
	Ilari: thanks for your help	19:13
	← prdonja left	19:13
	→ ereslibre joined	19:14
	→ dodo_the_last joined	19:15
	← MisterN left	19:16
	→ JasonWoof joined	19:21
	→ schmrz joined	19:21
	→ flaguy48 joined	19:25
	← rubydiamond left	19:27
	→ Khelben joined	19:29
	← Radar_ left	19:30
	← texel left	19:31
	← dwave left	19:32
	← paulboone left	19:35
	← priidu left	19:37
	← struberg left	19:37
	→ tatsuyao joined	19:39
	← galderz left	19:42
	← seangrove left	19:45
	→ Cheef-Daniel_ joined	19:47
	→ struberg joined	19:47
	← DavidKlein left	19:48
	← Cheef-Daniel left	19:48
	Cheef-Daniel_ → Cheef-Daniel	19:48
	→ arohner joined	19:49
	← Sigma left	19:51
	→ jnareb joined	19:52
	← eno left	19:53
	→ eno joined	19:55
	→ serfurj joined	19:57
jnareb	I was re-reading http://code.google.com/p/support/wiki/DVCSAnalysis and I have noticed a bit worrying thing	19:57
	In the footnote about benchmarking clone and pull (fetch) performance, comparing Git and Mercurial, we can see that "dumb" HTTP protocol used by Git is much slower than Mercurial wire protocol (hg-serve) over HTTP	19:57
	But what its worrying that in this benchmark Git is slower also when using git:// protocol (8.7 git vs 81 hg seconds for cloning, 2.8 git vs 1.5 hg seconds for the pull) over this Chicago <-> Mountain View link (51 ms ping time)	19:57
	→ MisterN joined	19:58
	← josesito left	19:59
thiago_home	how is 0.5 seconds worrying?	20:00
	→ solydzajs joined	20:01
JasonWoof	thiago_home: where'd you get 0.5 from?	20:01
ndim	how is 81-8.7 (aka several times faster) worrying?	20:01
	← d0k left	20:02
killerchicken__	I think that should be an 8.1	20:02
	not 81	20:02
	← schmrz left	20:03
	→ bobesponja joined	20:03
	→ nighthwk\| joined	20:04
thiago_home	0.6, sorry	20:04
	but it could be 0.5 due to rounding	20:04
jnareb	oops, 8.1	20:04
thiago_home	8.66 - 8.14 = 0.52	20:04
jnareb	well, the benchmark is rather lacking in the details...	20:04
thiago_home	so, I ask again: how is 0.5 to 1.3 seconds worrying?	20:05
	← kahmalo left	20:06
	→ stuffcorpse joined	20:06
jnareb	that is (2.8 / 1.5 = 1.9 times slower is what is worrying	20:07
	but I agree that is probably within the margin of error	20:07
	← stuffcor1se left	20:07
thiago_home	the relative value there is meaningless	20:08
	← flaguy48 left	20:08
thiago_home	I ask for the third time: how is 1.3 seconds going to make your life more difficult?	20:08
Pieter	I hate 1.3 seconds	20:08
	← Khelben left	20:09
jnareb	well... we will lose the 'fastest SCM' title... :-P	20:12
	→ webcoder joined	20:13
	→ Sho_ joined	20:13
jnareb	Bit off-topic: can ns3 (network similator) be used to benchmark network perfomance of DVCS?	20:13
Ilari	jnareb: BTW: The latencies probably don't matter that much there, especially for cloning.	20:15
	← nighthwk\| left	20:15
	→ nighthwk_ joined	20:17
	→ broonie joined	20:19
uau	the time could be spent either in the network part or by the remote server to create the pack to send	20:19
Ilari	The main components of the time are remote end assembling pack to send, sending it across (mostly network-bound) and then resolving deltas.	20:22
	→ bmalee joined	20:23
	→ osfameron joined	20:27
	→ godog joined	20:27
	→ Radar joined	20:31
	jnareb wonders whether the caching for git-daemon project would help there...	20:31
Ilari	jnareb: Probably.	20:33
	→ naeu joined	20:34
	→ paulboone joined	20:35
	← khmarbaise_ left	20:40
	← webcoder left	20:40
jnareb	I hope all will go well with this year Git's GSoC projects...	20:42
Ilari	With modern offset-delta packing, one could resolve deltas in pack while transport stream is being received. That might speed things up.	20:45
jnareb	Ilari: you mean changing received pack from possibly thin to thick one? Or something else?	20:46
Ilari	jnareb: Nope, meaning computing all object IDs on-line.	20:46
	jnareb: Nowadays, delta object IDs are computed off-line.	20:47
	→ spearce joined	20:47
Ilari	Hi spearce.	20:48
spearce	morning	20:48
jnareb	Ilari: I don't follow you... where (when) client computes objects ID when receiving pack, and why?	20:49
Ilari	Well, with faster links, it could require too much CPU power to keep up with the transport stream but as there is CPU power left (unless transport stream arrives very fast) one could possibly speed things up that way.	20:49
	← g3d left	20:50
jnareb	Ilari: is it when client calculates and creates pack index?	20:51
Ilari	jnareb: Right now object ID calculation for deltas occurs after whole pack is available.	20:51
	→ qrush joined	20:52
	→ ljungk joined	20:53
	← arohner left	20:53
	← ericindc left	20:53
	→ arohner joined	20:54
jnareb	Ilari: would this require multicore (one core is saving network stream to disk, second is resolving objects ID for deltas)?	20:57
Ilari	jnareb: No.	20:57
spearce	what are we talking about?	20:57
thiago_home	improving git pull/clone performance by parallelising operations	20:58
Ilari	spearce: Resolving deltas during object transfer.	20:58
thiago_home	to gain 1.3 seconds of precious pull time :-)	20:58
jnareb	spearce: the discussion started with me noticing that git performs slightly slower than Mercurial wven when both are using "smart" in http://code.google.com/p/support/wiki/DVCSAnalysis (but within measure error, probably)	20:59
	spearce is reading the log now trying to get caught up	20:59
	← mjf left	20:59
spearce	so basically Ilari's right about the problems.	20:59
	there's two areas where git clone sucks:	20:59
	1) remote side computing the set of objects to send in the pack.	21:00
thiago_home	sucks?	21:00
spearce	the gsoc cache daemon project is supposed to improve that.	21:00
	2) the delta resolving phase once the pack data is local.	21:00
	→ fridim joined	21:00
thiago_home	I mean, the difference in time is less than saying "oh, it's finished?"	21:00
	← drizzd left	21:00
spearce	thiago_home: iirc hg just has to stream the files out, it doesn't have the server computation demand that git has.	21:00
	in the test they were running, yea, network bandwidth is the bottleneck, not CPU on either side.	21:01
	→ Leemp joined	21:01
thiago_home	the only bad thing about the git transfer I see is the computing of the pack on the server side	21:01
spearce	no consider git.kernel.org, if they are a busy mirror, CPU for the pack enumeration may be more of a limit than their pipe at OSU OSL.	21:01
jnareb	does mercurial send optimal packs, then?	21:01
spearce	jnareb: it doesn't have a concept of packs.	21:01
thiago_home	if you don't have a properly packed repository, that creates load	21:01
	← RodP2 left	21:01
spearce	their on disk format permits them to just concat files from disk to the network.	21:02
	it has all sorts of downsides though; i still much prefer the git on disk format over hg format.	21:02
	← gilimanjaro left	21:02
Ilari	Regarding multicore, With it one could keep up with even faster streams when doing delta resolution online. Of course, the programming is more challenging. But it should speed things up even without mutlithreading.	21:02
	→ djszapi joined	21:03
djszapi	Hello!	21:03
Ilari	...	21:03
thiago_home	you don't need multicore	21:03
djszapi	Is there any description to set up git server ?	21:03
thiago_home	you don't even need multithread	21:03
spearce	djszapi: man git-daemon	21:03
thiago_home	the network transfer is likely to be much slower than the delta resolution	21:03
djszapi	http://blog.commonthread.com/2008/4/14/setting-up-a-git-server	21:03
Gitbot	djszapi: the git-daemon manpage can be found at http://git.or.cz/man/git-daemon	21:03
Ilari	thiago_home: Right.	21:03
jnareb	whet I meant, how Mercurial sends only update (only difference) for pull / fetch...	21:03
thiago_home	so you can use the idle time in-between select(2) to do that work	21:03
djszapi	i saw this	21:03
spearce	jnareb: git also only sends update during pull/fetch.	21:04
Ilari	djszapi: For ssh, you need to have git and sshd installed on remote side.	21:04
	djszapi: Good test is 'git user@host git --version'.	21:04
djszapi	Thank for you	21:04
jnareb	errr... what I meant is that I wonder how mercurial does send only difference, when it is streaming files from disk	21:04
thiago_home	same way as git	21:05
spearce	thiago_home: yea, you could use select to do it, but it would probably just be easier to fork a second process, so you don't slow down the network link when computing a large-ish object.	21:05
thiago_home	the client says where it ended	21:05
	so the server can calculate what the update is	21:05
	spearce: yeah, it would be easier	21:05
spearce	yea, but in hg the "update" is just a contiguous byte range in some subset of files	21:05
	so hg can pick and transfer by just grabbing a byte range in the affected path names	21:05
	git still has to do object enumeration to build the pack	21:06
	of course, hg "renames" a file by copying the entire #@!(#!@( history. :-)	21:06
jast	as also demonstrated by svn, to great effect	21:06
thiago_home	svn is completely different	21:07
	it never transfers history	21:07
jnareb	hmmm... I wonder how well it works if history is nonlinear	21:07
	→ ashleyw joined	21:07
spearce	jnareb: as i understand it, you may actually get a revision sent to you that you already have, because for the remote peer, it fell into that bucket you asked for, but on your side, its already in your file.	21:08
thiago_home	unless you meant that it transfers the whole file on rename	21:08
jnareb	Ilari: it would be challenging to code 'object ID resolving for deltas during streaming from network to disk' so it degrades well: from ulticore to single core, from slow network to fast network. But what a challenge!	21:08
spearce	the "friendly revision numbers" that hg offers are based on the position of the revision in the local file, iirc. which is why numbers for the same revision differ between repositories; the revision falls at a different location in the same file, based on merge ordering.	21:08
jnareb	spearce: so Mercurial can send more than strictly necessary when repositories differ on "friendly revision numbers" aka local commit numbers, e.g. due to different ordering of doing merges?	21:09
djszapi	and where is the git server config file ?	21:10
spearce	jnareb: that is my understanding of it, but i'm not as much of an expert on hg as i am on git. :-)	21:10
	← hyperair left	21:10
Ilari	jnareb: The scalabilty w.r.t. cores is more difficult. OTOH, it should paralelize (BTW: I have written code that performs encryption with multiple cores using multiple processes)...	21:10
djszapi	and i haven't got git-daemon to start after boot automatically	21:10
Ilari	djszapi: Create/install suitable boot script?	21:11
jnareb	djszapi: the config for git-daemon is config for inetd or xinted (plus a bit of git config file)	21:11
	djszapi: unless you run service not from (x)inetd	21:11
	← ereslibre left	21:12
jnareb	djszapi: then in service start/stop script	21:12
djszapi	once more	21:12
	i havn't got gitd init script, nor configuration file for it	21:12
jnareb	in git-daemon manpage there is example configuration for inetd	21:13
Ilari	djszapi: There is no configuration file for git daemon. All configuration options available for git daemon itself are from startup command line.	21:13
	← solydzajs left	21:13
jnareb	Ilari: what about daemon.* config variables?	21:14
	but git-daemon is configured via command line parameters (used either in (x)inetd config file, or in git-daemon service script you have to write)	21:14
spearce	jnareb: they are only overrides for per-repository options, and sometimes still require the daemon to be configured to permit override. e.g. receive-pack requires you at least allow override receive-pack on the command line first	21:15
djszapi	To set up git-daemon as an inetd service that handles any repository under the whitelisted set of directories, /pub/foo and /pub/bar, place an entry like the following into /etc/inetd all on one line:	21:15
	i don't have any /etc/inetd	21:15
Ilari	Oh, those. But mostly those are better left alone.	21:15
djszapi	maybe xinetd ?	21:16
	→ gotgenes joined	21:16
	← radarek left	21:17
Ilari	djszapi: The command line part for xinetd is similar. The service config is totally different.	21:17
	← FreakGuard_ left	21:17
Ilari	djszapi: But probably that config file has some configs that can be modified.	21:18
	→ FreakGuard joined	21:18
djszapi	http://vim.pastey.net/113964	21:18
jnareb	Vs lbh nfxrq Oehpr Fpuarvre gb qrpelcg guvf, ur'q pehfu lbhe fxhyy jvgu uvf ynhtu.	21:19
thiago_home	ROT-13?	21:19
	← Radar left	21:19
jnareb	Ooops	21:20
spearce	damn, is it legal for a git repository to be missing $GIT_DIR/objects/pack ?	21:20
jnareb	thiago_home: ROT13, from {somebody} Facts	21:20
djszapi	Ilari: i can't understand you	21:21
	sorry	21:21
thiago_home	spearce: if it's never been packed before, why not?	21:22
	→ hyperair joined	21:22
	→ thumper joined	21:22
Ilari	djszapi: If you have xinetd, presumably you have some services configured (even if disabled).	21:22
doener	spearce: at least "git repack -ad" creates it, if it's missing	21:22
spearce	thiago_home: yea, i think that's the case, its legal. sigh	21:22
	spearce goes off to patch jgit	21:22
djszapi	Ilari: what do you want to say with it ? what can i do for git ? :)	21:23
thiago_home	jnareb: increase your security, use double ROT-13!	21:23
	:-)	21:23
jnareb	djszapi: services can be started from service script (like e.g. named or httpd) or started from network services daemon, e.g. inetd or xinetd	21:23
djszapi	yeah	21:23
	→ davi joined	21:23
djszapi	but what must i write to xinetd.conf ?	21:23
jnareb	that... I do not know, sorry	21:24
djszapi	:(	21:24
Ilari	Actually, chaining block cipher twice in row doesn't imporve security. That's the reason why its triple-DES and not double-DES.	21:25
	djszapi: Perhaps your distro has 'git-daemon-run' or similar package?	21:25
djszapi	yeah	21:26
	git-daemon	21:26
jnareb	djszapi: check out "man giteveryday" (or "git help everyday", or Documentation/everyday.txt), but it might be bit dated	21:27
	← aspotashev left	21:27
djszapi	No manual entry for giteveryday	21:27
	sorry	21:27
jnareb	/usr/share/doc/git-*/everyday.txt	21:28
djszapi	it doesn't exist	21:29
jnareb	HEAD:Documentation/everyday.txt	21:29
	→ ereslibre joined	21:29
jnareb	djszapi: Documentation/everyday.txt in your git repository (or in git.git repository on repo.or.cz or git.kernel.org)	21:30
	→ earcar joined	21:31
	← kumbayo left	21:31
	→ aspotashev\|eeepc joined	21:32
thiago_home	who's running the Gitbot here?	21:32
	thiago_home would like a gitbot in his channel too	21:32
djszapi	http://vim.pastey.net/113965	21:32
robinr	spearce: any idea abouyt: http://code.google.com/p/egit/issues/detail?id=87 Just a case of needs large enough pipe?	21:32
djszapi	hit installed these files	21:32
	← smeevil left	21:32
djszapi	*git	21:32
context	is post-commit called when you push to a repo	21:32
jnareb	http://gist.github.com/109753	21:33
context	trying to get this hook working that sends commit info to lighthouse	21:33
spearce	robinr: oh, yea. maybe...	21:33
Ilari	context: No.	21:33
spearce	robinr: except TransportLocal l.173 has code to increase the allocation of the out_r buffer.	21:34
jnareb	thiago_home: try searching for Gitbot on git wiki...	21:34
context	ilari: ahh that'd explain a lot	21:34
	thnx	21:34
	← gaveen left	21:34
thiago_home	jnareb: thanks, I will	21:34
Ilari	context: Its pretty logical. Push doesn't commit, so it won't call post-commit either.	21:34
spearce	robinr: MAX_HAVES * 46 + 4 should have been sufficient; we +1024 for slop.	21:34
djszapi	jnareb: i can't see that file in the given repo	21:34
jnareb	djszapi: what file?	21:35
djszapi	Documentation/everyday.txt	21:35
spearce	robinr: the other thing is, i don't think we play these buffer games when running over TCP. IOW, we are deadlocking because our code is wrong, not because our buffer is insufficiently sized.	21:35
	← yann left	21:36
djszapi	http://www.kernel.org/pub/software/scm/git/docs/everyday.html	21:36
	maybe this one	21:36
jnareb	yup	21:37
	→ gaveen joined	21:37
djszapi	i have got man git-daemon, however i don't have git-daemon binary	21:37
Gitbot	djszapi: the git-daemon manpage can be found at http://git.or.cz/man/git-daemon	21:37
djszapi	it's really interesting	21:37
	→ elad` joined	21:37
djszapi	i've red it already	21:37
elad`	How do I check the diff between the thing I just pulled and whatever I had beforehand?	21:38
	I did git pull before doing git diff.	21:38
jnareb	djszapi: what distribution are you running? perhaps it does have git-daemon packed in other binary package	21:38
djszapi	archlinux	21:38
jnareb	elad`: pull not fetch?	21:38
djszapi	i can't find that	21:38
jnareb	elad`: git diff ORIG_HEAD	21:38
	djszapi: try searching in "git --exec-path"	21:39
djszapi	/usr/lib/git-core	21:39
	/usr/lib/git-core/git-daemon	21:40
	this one	21:40
jnareb	djszapi: git-daemon should be there... unless your packaging system puts git-daemon in separate package	21:40
elad`	Thanks.	21:40
	→ ericindc joined	21:41
	← Cheef-Daniel left	21:41
jnareb	elad`: or, in more general, try reflog	21:42
	← kate21de left	21:43
spearce	huh. i have a server running jgit whose window cache has 128 open files, but claims to have only 764.32 kb of data loaded, which should be at moast 95 files.	21:43
	oh, wait, a ton of those pack files are really tiny, probably smaller than the 4096 packedgitwindowsize.	21:44
	← SRabbelier left	21:44
jnareb	[censored] [censored] thiose [censored] wiki spammers, [censored]	21:45
robinr	jnareb: i heard there'd be free space at Guantanom soon...	21:45
	← earcar left	21:45
robinr	spearce: you mean the code isn't parallell enough?	21:46
spearce	robinr: what are talking about, the fetch deadlock or my stupid remark about windowcache?	21:47
robinr	the fetch dealock	21:47
	→ pigeon joined	21:49
	→ gilimanjaro joined	21:51
	→ yoann joined	21:51
	→ angerman joined	21:55
	→ bremner_ joined	21:55
JasonWoof	everyday.html has this example: git format-patch -k -m --stdout v2.6.14..private2.6.14 \| git am -3 -k	21:57
	seems wacky to me. shouldn't that just be a simple rebase command?	21:58
spearce	robinr: i don't see why we deadlock. but, we could try taking that "while(resultsPending > 0)" block inside of BasePackFetchConnection.negotiate (l.362) and call it if pckIn.available returns > 0, trying to suck up server responses as soon as they are available, rather than waiting until the end of a block.	21:58
djszapi	git stream tcp nowait nobody /usr/bin/git /usr/lib/git-core/git-daemon --inetd --verbose --export-all /pub/foo /pub/bar	21:59
	git: 'stream' is not a git-command. See 'git --help'.	21:59
	← ceej left	21:59
Ilari	djszapi: "git stream tcp nowait nobody" are instructions for inetd. The next argument is actual binary, followed by its args.	22:00
	→ hummingkid joined	22:00
	← Leemp left	22:00
jnareb	djszapi: here is excerpt from "Everyday ..." dealing with xinetd configuration for git-daemon: http://gist.github.com/109753	22:01
	← mtkd left	22:01
djszapi	bash-3.2# git stream tcp nowait nobody	22:01
	git: 'stream' is not a git-command. See 'git --help'.	22:01
	← ereslibre left	22:03
thiago_home	that looks like an /etc/inetd.conf line	22:03
spearce	robinr: what vm was that?	22:03
djszapi	thiago_home: but i haven't got /etc/inetd.conf	22:04
jnareb	xinetd != inetd	22:04
	djszapi: and there is example of using git-daemon as standalone server in git-daemon(1)	22:05
djszapi	jnareb: Don't tell me, thiago_home've said wrong	22:05
	→ josesito joined	22:06
josesito	I want to delete a remote branch, how do i do that? and if i wanted it back, how could i restore it?	22:06
djszapi	jnareb: yeah but i don't want to operate this server according to the ip number	22:06
thiago_home	josesito: git push remotename :branchname	22:07
	josesito: to restore it, simply push the same SHA-1 to that branch again	22:07
	← angerman left	22:07
thiago_home	you must have that SHA-1 in your repository, of course	22:07
josesito	thiago_home: uh?	22:07
	can i look the reflog to get the sha-1?	22:08
thiago_home	yes	22:08
	← nighthwk_ left	22:08
jnareb	errr... I think that you cannot push arbitrary sha-1...	22:08
	← nighthwk1 left	22:08
thiago_home	well, it has to be a commit SHA-1	22:08
	→ nighthwk1 joined	22:08
josesito	where can I look it for?	22:09
robinr	spearce: sun 1.6.0 (12) I think	22:09
	→ Radar joined	22:09
djszapi	jnareb: which line do you mean ?	22:09
	→ smashwilson joined	22:09
thiago_home	josesito: what do you mean "look for it" ?	22:09
	josesito: you should know what SHA-1 you want to push	22:10
robinr	It was quite a small fetch btw	22:10
	← bryanray left	22:10
spearce	robinr: i think i may see how this can overflow	22:10
josesito	thiago_home: (sorry for my grammar) well, if i wanted to restore the deleted branch, i'd like to restore it to it's last state, where would i look for that?	22:10
	jnareb has trouble demuxing the discussions	22:11
thiago_home	josesito: nowhere	22:11
josesito	fail!	22:11
thiago_home	josesito: you simply know what you want to restore it to	22:11
spearce	if multi_ack is enabled (and it would be, both sides support it) the UploadPack end can send ACK X continue lines for each have line. but the BasePackFetchConnection only reads each end() it had sent previously	22:11
josesito	i'm lost	22:11
	← naeu left	22:11
	→ psoo joined	22:12
thiago_home	josesito: don't delete stuff if you're not sure you want to delete it	22:12
spearce	thus the output stream from UploadPack->BasePackFetchConnection can become full, and cause the writer to freeze	22:12
	← gotgenes left	22:12
spearce	when UploadPack freezes, eventually BasePackFetchConnection also has to freeze, because its output stream becomes full	22:12
josesito	thiago_home: I'm sure I want to delete it, I just don't know yet if the merge went 100% successful	22:12
djszapi	?	22:12
josesito	It was supposed to be a tmp branch.	22:12
thiago_home	josesito: what merge?	22:13
spearce	ok, i don't know how the #@*!(!@ this works in C Git without them also being able to reach a deadlock state	22:13
thiago_home	this is the first you're talking about merges	22:13
	← Lenary left	22:13
thiago_home	spearce: sockets in Java?	22:14
	← gilimanjaro left	22:14
spearce	thiago_home: yea, well in this case, its a pipe within the process.	22:14
josesito	thiago_home: let me explain. I made this tmp branchto try something out.	22:14
thiago_home	spearce: no non-blocking mode, right?	22:14
jnareb	djszapi: inetds config and git-daemon as standalone service is described in git-daemon(1) manpage; xinetd config is described in "everyday" (which you can read via gitweb, you can read on kernel.org in provided link, or exceprt in my pastebin/gist).	22:14
josesito	then merged it into my master branch.	22:14
djszapi	http://vim.pastey.net/113968	22:14
spearce	thiago_home: yea, no non-blocking mode.	22:15
djszapi	and this isn't working	22:15
josesito	oh nevermind, I'll just let someone else does it...	22:15
	← _graham_ left	22:15
spearce	thiago_home: but, iirc, c git does this with standard tcp/ip sockets or a local socketpair() and doesn't require nonblocking sockets to implement the protocol	22:15
djszapi	B[B[B[B[B[Bbash-3.2# git add .	22:16
	bash-3.2#	22:16
thiago_home	and nothing sets it in C to non-blocking mode?	22:16
	thiago_home greps the source	22:16
	→ gotgenes joined	22:16
thiago_home	no non-blocking in C either...	22:16
	→ twas joined	22:16
thiago_home	josesito: if the merge was successful, you're going to see it in your "git log master" history	22:17
	→ dan joined	22:17
robinr	hmmm, I'd call it blocking...	22:17
	because that's what's happening	22:17
JasonWoof	What's the best way to apply the changes from a single commit (tagged foo) to my working tree (without committing it like cherry-pick would)	22:18
thiago_home	blocking sockets can deadlock if both ends are stuck writing	22:18
spearce	thiago_home: yes, clearly, and that's what happens here	22:18
JasonWoof	I could do: git diff foo^..foo \| patch -p1	22:18
	→ cryolab joined	22:18
JasonWoof	but I'm guessing there's a more git-ish way to do this...	22:18
spearce	i just can't explain how C git never seems to deadlock, when it seems the protocol is just asking for deadlock	22:18
thiago_home	they could as well if they were both stuck reading, but that would indicate a network failure (packets disappeared) or bad protocol	22:18
	the solution for that is to either use non-blocking or to use two threads	22:19
	← jnareb left	22:19
robinr	could that grouping into 31 be something	22:19
thiago_home	TCP socket buffers are something between 40 and 128k in buffer; I don't know about local sockets.	22:20
robinr	s/32	22:20
thiago_home	pipes have very small buffers though	22:20
spearce	oh, s**t. i think we don't read enough data in the client side	22:20
	→ shd joined	22:20
	← chris2_ left	22:20
	← thumper left	22:20
thiago_home	Qt implemented non-blocking sockets; Java requires the use of threads, which is much more complex	22:22
	spearce starts writing a patch	22:22
thiago_home	and then I have to deal with a lot of programmers coming into #qt and having problems in their applications because of threads	22:22
	← neoeinstein left	22:23
robinr	thiago_home: NIO has non-blocking sockets	22:23
	← josesito left	22:23
	→ neoeinstein joined	22:24
spearce	robinr: i think i have it fixed...	22:25
robinr	:)	22:25
	← WALoeIII left	22:26
	← nighthwk1 left	22:26
	← cryolab left	22:26
context	whew. feels good to get work done	22:27
thiago_home	robinr: interesting	22:29
	very complete API	22:29
	→ harinath_ joined	22:29
robinr	and a lot harder to use	22:29
	but then, no pain, no gain	22:30
	→ offby1` joined	22:31
robinr	I know the TAO (Corba) orb used async I/O. I just spawned off data, checking wheter it was fine later on, at the time it actually depended on the result	22:32
	→ Paf joined	22:32
	← psoo left	22:32
thiago_home	robinr: the API I had been envisioning to fix some of the performance bottlenecks for QIODevice would be similar to this	22:32
robinr	quite interesting to trace	22:32
	thiago_home: using the "reactor" pattern?	22:34
thiago_home	no, the channel pattern	22:34
	and a scatter/gather buffer	22:34
	basically I want to be able to do zero-copy memory transfers between the many levels of I/O operations	22:35
	← cryo_ left	22:35
	← ljungk left	22:36
robinr	ACE was the name of it	22:36
thiago_home	ASCII-Compatible Encoding?	22:36
robinr	Adaptive Communication Environment	22:38
spearce	wait, thiago_home hasn't heard of ACE ?	22:38
robinr	it's C++ too	22:38
	← bremner_ left	22:38
spearce	i thought any C++ guy would have heard of it...	22:38
robinr	don't be too hard now	22:39
thiago_home	never heard of it	22:39
	for me, ACE is related to IDNA	22:39
robinr	corba is dead, hence everything it used	22:40
thiago_home	RFC 3490, 3491 and 3492	22:40
	← bnferguson left	22:40
	← bx2 left	22:41
robinr	not even poker?	22:41
thiago_home	the reactor pattern looks a lot like the signal-slot mechanism in Qt	22:42
	← harinath left	22:43
	→ pencilo joined	22:45
przemoc	Hi. I'm using cvsimport and now after incremental running I'm confused, because origin/master isn't updated, only master and origin. Is it normal?	22:45
	← ericindc left	22:45
robinr	sort-of, I'd guess, though signals and slots aren't "usually" asynchronous, with some exceptions, right?	22:46
spearce	robinr: patches sent	22:48
	→ spenser joined	22:48
spenser	hello	22:49
robinr	spearce: nice	22:49
thiago_home	if you use the reactor pattern in a multithreaded environment, it would be like having a Qt event loop with socket notifiers dispatching signals to other threads	22:49
robinr	junit tests too?	22:49
thiago_home	I see no difference there	22:49
	← paulboone left	22:49
robinr	the reactor pattern doesn't need the overhead of threads, that's one of it's points	22:50
	it's about scalability	22:50
	→ drizzd joined	22:51
thiago_home	so dispatch to any available thread?	22:51
	← fridim left	22:51
	→ dwave joined	22:52
spenser	I'm trying to use git to deploy my website with a post-receive hook. I followed the tutorial at http://toroid.org/ams/git-website-howto . However for some reason git is not honoring my GIT_WORK_TREE environmental variable or git config core.worktree variable. I'm using git version 1.5.4.3 from Ubuntu Hardy.	22:52
Ilari	Hmm... How is the protocol just asking for deadlock? What reads and writes run in parallel?	22:54
spenser	Has anyone seen this issue before? Should i file a bug somewhere? Also, how do I work around htis problem. The error I recieve during my commit is fatal: /usr/bin/git-checkout cannot be used without a working tree. error: hooks/post-receive exited with error code 1	22:54
	→ alien joined	22:54
robinr	I think it can, but there is some similarity to notifiers. ACE has a event loop too	22:55
alien	hello, can svn sumbodules be added to a git repo?	22:55
spearce	Ilari: client sends "have" lines in parallel to the server sending "ACK continue" lines. if there isn't sufficient buffering on at least one side for the 2*32 have lines, then you can deadlock.	22:55
robinr	alien: there is no such concept	22:55
przemoc	spenser: try newer git, https://launchpad.net/~git-core/+archive/ppa	22:55
spenser	przemoc: will do.	22:55
parasti	huh, nice straight-forward URL	22:56
alien	robinr: then how could I track a couple of external svn or git project from a central git project?	22:56
	→ flaguy48 joined	22:56
przemoc	spenser: in short add deb http://ppa.launchpad.net/git-core/ppa/ubuntu hardy main	22:57
	→ Ademan joined	22:57
przemoc	spenser: to /etc/apt/sources.list	22:57
	spenser: run: sudo apt-key adv --recv-keys --keyserver keyserver.ubuntu.com E1DF1F24; sudo apt-get update; sudo apt-get install git-core	22:58
Ilari	spearce: And client only reads the continue lines after blasting slew of haves?	22:58
spearce	Ilari: correct. client can blast up to 64 have lines before reading anything from the remote peer	22:58
	there was a bug in jgit where we had sufficient client buffering, but we stopped reading from the peer too early	22:59
	← hyperair left	22:59
spearce	the client only needs 2948 bytes of buffer.	22:59
	→ rjc_ joined	23:00
spearce	in the case of TCP sockets, i think most TCP implementations allocate at least that for the send queue.	23:00
	for a local pipe, I don't know.	23:00
Ilari	spearce: Local pipe is 4096.	23:00
	spearce: For pipes. Don't know about sockets (IIRC, it uses those).	23:01
spenser	przemoc: I'm installing it now	23:01
spearce	Ilari: basically the bad thing about the protocol is you must have at least 2948 bytes on the client send queue to avoid deadlock. C git is betting that most OS kernels will give them that with a TCP socket or a local pipe.	23:02
spenser	I think that fixed it	23:02
	przemoc: I believe that fixed it	23:02
	→ tjafk2 joined	23:02
	← mstrl left	23:03
Ilari	spearce: Doesn't it actually distribute over all four buffers (TX and RX buffers of both half-connections)?	23:04
spearce	Ilari: yes, that too	23:04
spenser	przemoc: yes its fixed in the ppa. Should I file a bug on the package or do anything else?	23:04
spearce	given that Nagle's algorithm should still be enabled on the send side, and the size of a TCP/IP packet, with the TX buffer on client and RX buffer on the server, that 2948 bytes probably is already easily handled in any TCP implementation.	23:05
	→ Leemp joined	23:05
thiago_home	nagle is asynchronous	23:06
przemoc	spenser: I'm not sure whether they'll backport git changes to older release. You may try and in the worst case you'll waste some of your time.	23:06
spearce	thiago_home: but my point was just that although we're doing these tiny 46 byte writes, they should coalesce into about a full IP packet.	23:06
	at which point, we're only talking about what, 2, maybe 3 IP packets needed worth of buffering?	23:07
robinr	isn't the amount of data you can write over TCP related to the window size ( a lot larger tha 2948 bytes)	23:07
	?	23:07
Ilari	spearce: Full TCP/IP payload is in practice 1460 bytes.	23:07
spearce	oh, yea, good point robinr, that too	23:07
thiago_home	Nagle is about trying to get 1460 bytes	23:07
spearce	the window size for the stream at this point should be a lot larger than 2948 bytes	23:08
thiago_home	then it fires off a TCP/IP packet	23:08
spearce	and 2948/1460 = 2.019 packets... so we should be able to buffer one packet in the TX buffer, fire it off, and buffer one more, fire it off, before the client needs to start reading.	23:09
spenser	przemoc: So file a bug on launchpad against the specific hardy package of git-core?	23:10
przemoc	spenser: There are other worthy ppas. Consider OOo 3 (https://launchpad.net/~openoffice-pkgs/+archive/ppa), svn 1.5 (https://launchpad.net/~mdiers/+archive/ppa) and so on...	23:12
	spenser: yes	23:12
	← voker57__ left	23:12
	← dwave left	23:12
	→ voker57__ joined	23:13
	→ nighthwk1 joined	23:13
	→ joshsdc joined	23:14
	→ eno__ joined	23:16
	→ nabilelisa joined	23:17
struberg	spearce: when I use TreeWalk, i.getDirCacheEntry().getLastModified() always gives me "Tue Oct 13 17:55:45 CEST 2037" (2139062145139) is this ok?	23:17
spenser	przemoc: I am filing the bug right now.	23:17
struberg	try to implement SimpleRepository#status with TreeWalk ...	23:17
spearce	struberg: aaah. that's a racy git condition.	23:17
	← MisterN left	23:18
nabilelisa	hi all, does anybody know how i can get my ant build script to compile to a different directory based on my current git branch?	23:18
spearce	struberg: basically we can't be 100% sure that the last modified time can be relied upon, so we give you a time that is very unlikely to match the working directory file.	23:18
struberg	spearce ah makes sense	23:19
spearce	struberg: when the index entry doesn't match the working directory file, you need to check the content to see if it hashes to the same objectid.	23:19
	← tjafk1 left	23:19
spearce	struberg: if it does, a "good" status program would then update the last modified time in the index, and write the index back to disk, to save from doing that in the future.	23:19
	struberg: where racy git triggers is if the index and the working directory file were written in the same filesystem timestamp.	23:19
struberg	spearce ah kk, hoped to only need the DirCacheIterator, but better I use the DirCacheBuildIterator?	23:20
spearce	e.g. if the filesystem only tracks second accuracy, and they are written in the same second, there is a good chance that an overwrite of the file would also happen in the same second, and not be noticed	23:20
struberg	yup I'm aware of that	23:20
spearce	struberg: maybe you want to use the build iterator here, so you can do the update	23:20
robinr	$ cat /proc/sys/net/core/wmem_max	23:21
	131071	23:21
killerchicken__	nabilelisa: well, you could parse the output of git status, for example?	23:21
struberg	spearce, yeah, once I get the basic parts running, then I'll working on writing them back to the DirCache ;)	23:21
robinr	means I can write 128k before write blocks?	23:21
	← Radar left	23:21
	← frando left	23:21
nabilelisa	killerchicken__: cool, interesting idea. i'll look into it, thanks. do you have any sample code handy i can start with?	23:21
	← schacon left	23:23
	→ ericindc joined	23:23
spenser	przemoc: Bug #374658 has been filled on the issue. Would you mind looking at it and making sure I don't need to add anything else?	23:26
	← eno left	23:27
robinr	#¤"#¤% mercurial plugin, always hangs eclipse for a while on startup	23:28
spearce	robinr: isn't that a feature? :-)	23:28
killerchicken__	nabilelisa: sorry, missed that question. Something like git status \| grep "On branch" \| awk '{ print $4 }' maybe? Note that that is rather brittle, of course (doesn't work if you're not on a branch currently, for example)	23:29
	(the output will be empty for that case)	23:30
nabilelisa	killerchicken__: thanks, i was thinking of doing it like that actually, but then i got a bit more ambitious and am looking into http://ant.apache.org/faq.html#batch-shell-execute to see if i can automate it as part of my ant target	23:30
	i dont know if it'll work, but it would seem cleaner i think	23:31
	← struberg left	23:31
nabilelisa	thanks for your input though - i appreciate it.	23:31
	← scientes left	23:31
killerchicken__	yeah, as I said, I don't really know ant	23:31
Ilari	nabilelisa: Better way to get current branch is probably to first try to read HEAD using symbolic-ref. Then you have the detached case still.	23:32
nabilelisa	Ilari: hrm, sorry im not sure i understand that (im newish to both git and ant)	23:33
	killerchicken__ has never used symbolic-ref before. I guess I should take a peek	23:34
	→ struberg joined	23:34
Ilari	nabilelisa: 'git symbolic-ref -q HEAD'. If it returns 0, then it prints FQRN of current branch to stdout. If it returns 1, run 'git rev-parse HEAD' to get the detached commit ID.	23:34
robinr	spearce: makes me use the uninstall feature	23:35
killerchicken__	hm, symbolic-ref is missing from the bash autocomplete thing	23:35
	heh, maybe it only has porcellain commands	23:35
Ilari	nabilelisa: Better yet: `git symbolic-ref -q HEAD \|\| git rev-parse HEAD`	23:36
	nabilelisa: That outputs either FQRN or commit ID for detached case.	23:36
nabilelisa	Ilari: what's FQRN?	23:37
robinr	spearce: btw, I'll fix that bug I mentioned in the Eclipse newsgroup tomorrow	23:37
Ilari	nabilelisa: Fully Qualified Ref Name. For example 'refs/heads/foo' for local branch 'foo'.	23:38
spearce	robinr: oh, the origin/HEAD symref not reporting? ok	23:38
robinr	yep	23:38
	→ jelmer_ joined	23:38
Ilari	Wonder how Jgit would react to HEAD being symbolic link?	23:38
przemoc	spenser: It's almost fine, but your wish for sure won't be granted, because only neccessary change (cherrypicked commit) can be applied if any. Ubuntu maintainers doesn't change major release version of their packages.	23:38
nabilelisa	Ilari: oh, i think i see	23:38
	robinr >> 0xBED;	23:39
	→ handstandmilitia joined	23:39
handstandmilitia	longtime listener first time caller.	23:39
	← jelmer_ left	23:39
	→ bnferguson joined	23:39
nabilelisa	Ilari: yes, i've just done a small test, it seems that it would help me. is can you think of any reason why this would be 'better' than using awk? i feel slightly more comfortable with that, but i'm not really experienced enough to know what i'd be missing out on by taking that approach	23:40
handstandmilitia	I was curious if its possible to 'checkout' or get a specific revision for an arbitrary repository if I know the revision I want.	23:40
Ilari	nabilelisa: Git status output is not intended for machine parsing.	23:40
spenser	przemoc: How can I cherry pick the commit that fixed this problem?	23:40
handstandmilitia	Im essentially just looking to get a snapshot and im not interested in cloning and getting the whole revision history	23:40
spenser	przemoc: That way I can just tell them which change to incorporate?	23:40
nabilelisa	Ilari: ah, so the formatting could change in future updates?	23:40
	← ashleyw left	23:40
nabilelisa	i see	23:40
przemoc	spenser: seeing git 1.6.x in hardy's official repo is impossible, however fixed git 1.5 is possible	23:40
nabilelisa	Ilari: well, that's a good reason i guess	23:41
Ilari	handstandmilitia: There is remote snapshot feature, but most of the time it doesn't work.	23:41
przemoc	spenser: yes, it will be helpful for them	23:41
handstandmilitia	ilari haha awesome	23:41
spenser	przemoc: Is there a revision history somewhere?	23:41
	→ ceej joined	23:41
handstandmilitia	ilari know the syntax?	23:41
Ilari	handstandmilitia: It pretty much only works with ssh and file protcols. And it doesn't work if remote end is using gitosis (git@foo or gitosis@foo are mostly those).	23:42
przemoc	spenser: sure, http://git.kernel.org/?p=git/git.git;a=summary	23:42
handstandmilitia	im working with a github repo in this instance	23:42
Ilari	handstandmilitia: Also, if the repo has gitweb, then that might have snapshot enabled.	23:43
	handstandmilitia: IIRC, github has snapshot download from web interface.	23:43
spenser	przemoc: I'm going to search this and see if I can find a commit then attach the commit sha-1 in the comments.	23:44
handstandmilitia	ilari dont know the syntax though? I suppose I can poke around for it... the gitweb feature is cool, but i actually have a script run it	23:45
przemoc	spenser: if there will be any interest, git-core maintainers will solve the problem even without specyfing commit(s), because they're familiar with git (or at least shoud be)	23:45
Ilari	handstandmilitia: 'git archive --remote=URL <commit>'.	23:45
handstandmilitia	an sweet, i was actually looking at github just now and was about to write a script to wget it lol	23:45
spenser	przemoc: I understand but they give me an OS for free the least i can do is try to help out.	23:45
	→ wshimmy1 joined	23:45
	← dkr21 left	23:46
Ilari	handstandmilitia: But most of the time, that just plain doesn't work because remote rejects it.	23:46
handstandmilitia	well if it rejects it i'll either write a script to auth to github and get it from the web or i'll just download it manually and have my script download the tarball from wherever i store it...	23:46
	that might even just be the better option huh...	23:46
przemoc	spenser: praiseworthy attitude	23:46
	← patmaddox left	23:47
handstandmilitia	ilari thanks for the info, 1000 internets to you.	23:47
	http://audibble.files.wordpress.com/2008/11/freeinternet.jpg	23:47
killerchicken__	please refrain from posting the link 1000 times - I'm sure Ilari knows where the refresh button is :-)	23:49
handstandmilitia	ilari you are totally on the ball: http://groups.google.com/group/github/browse_thread/thread/cfcbcb1dc5f41f16?pli=1	23:50
Ilari	CTRL+R? :-)	23:50
handstandmilitia	github doesn't support it out of caching reasons, easier to cache the tarball via the web	23:50
	thanks for the tips!	23:50
	→ Pewpewarrows joined	23:50
Ilari	Getting the latest tarball from archive is actually pretty CPU efficient.	23:51
	Except that the compression can't be cached... And it will blow bandwith like crazy without it...	23:52
	→ Emachman joined	23:52
spenser	przemoc: Can't seem to find a commit with that topic maybe a ubuntu patch is causing the error is there anyway for me to check the ubuntu patches if any?	23:52
Emachman	Hi, is there a way to remove a commit from the history? I accidentally committed some private info	23:53
	← handstandmilitia left	23:53
	← tatsuyao left	23:54
	← Bass10 left	23:54
killerchicken__	Emachman: have you published it already? meaning, did you push it anywhere?	23:55
przemoc	spenser: https://launchpad.net/ubuntu/+source/git-core	23:55
Emachman	I pushed it already to github	23:55
	← robinr left	23:55
przemoc	spenser: there under hardy release you can find ubuntu diffs	23:56
	offby1` → offby1	23:56
killerchicken__	Emachman: ok. you can do an interactive rebase locally, take out the commit, and force pushing	23:56
przemoc	spenser: but I doubt that maintainers broke it	23:57
	→ JohnnyL joined	23:57
spenser	przemoc: I bet someone fixed it without telling anyone in git.git.	23:57
Emachman	killerchicken__: How would I do that?	23:58
spenser	przemoc: It's probably going to be too dificult for me to find then.	23:58
	przemoc: Well I tried.	23:58
killerchicken__	Emachman: how long ago was that commit? Have you merged since?	23:58
Emachman	April 28	23:58
killerchicken__	well, I meant "have you done a lot of stuff since then"	23:59
Emachman	1 commit	23:59
	When I was removing the private info	23:59
killerchicken__	ah, so no big deal	23:59
	← wshimmy left	23:59

Logs Search ←Prev date Next date→

Channels Documentation