IRCloggy #git 2010-01-06

Provider of IRC logs since 2005.

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

	2010-01-06
cbreak	and all have their own issues	00:00
_Vi	Arafangion, Basic things can be done with plain "git gui" and "gitk".	00:00
cbreak	if you want a gui, it's highly likely that it'll be for one platform only	00:00
	either that, or it won't fit to any platform	00:00
Arafangion	_Vi: Last time I tried, git-gui wouldn't work properly. (But then, I was using binary files and crap that git wasn't designed for)	00:01
_Vi	Arafangion, (And "gitk" "git gui" looks almost the same on Linux, OS X and Win32 (both cygwin and msysgit)).	00:01
Arafangion	gitk is a nice history viewer, true, but that's really all it is.	00:01
cbreak	it looks like crap on OS X. git gui :)	00:01
_Vi	Arafangion, In "git gui" you may stage changes linewise easily.	00:02
Fazer2	I checked http://gitorious.org/about/faq and added [branch "master"] etc. group to my .git/config file	00:02
	but it didn't worked	00:02
_Vi	cbreak, Why "like a crap". It even renders UTF-8 (after configuration).	00:02
Fazer2	*didn't work	00:02
	← ideogon left	00:02
cbreak	yeah. but... it's german	00:02
_Vi	cbreak, And I can see all controls and all text./	00:02
Arafangion	_Vi: Some of my files were microsoft word documetns.	00:03
cbreak	gitx looks better.	00:03
Arafangion	gitx looks very nice.	00:03
_Vi	cbreak, Does it have some useful additional features for ones that are happy with CLI and "git gui" or only look&feel?	00:04
Arafangion	Although it doesn't look like it's something that's intrinsically non-portable.	00:04
cbreak	_Vi: I don't know. I rarely use it. What it does is show the history or allow commiting	00:04
	maybe more	00:04
	Arafangion: it's not portable because Cocoa is Mac OS X only	00:05
_Vi	cbreak, Even "egit" for Eclipse shoes the history and allows committing.	00:05
	s/shoes/shows/	00:05
cbreak	yes. even the git command line app does that :)	00:06
	but the presentation and UI is different	00:06
	I don't think there's anything you can do with git that you can't do with git on the CLI	00:06
Arafangion	cbreak: Could still be ported, although true, to a different gui library.	00:07
_Vi	cbreak, (Obvious fix: "I don't think there's anything you can do that you can't do on the CLI")	00:07
	← dgarr left	00:07
	← mattions left	00:08
gmarquezWRS	NewbAsks: does anyone know what updates my projects.list file (gitweb or gitolite)?	00:08
cbreak	Arafangion: cocoa is not only GUI. You'd have to rewrite the whole app in something other than Objective-C. Might as well do separate projects	00:08
_Vi	Arafangion, In Wikipedia or somewhere there's table of different UIs for Git.	00:08
Arafangion	cbreak: There are objective-c compilers for many systems.	00:09
	(Or rather, one major compiler, which has been ported)	00:09
cbreak	I don't think that'll work	00:09
	Objective-C is not much without it's framework	00:09
	And I am not sure how well supported some runtime extensions from apple are	00:10
	← warthog9 left	00:10
Arafangion	You might indeed have some difficulty with Objective C 2.0 things, with garbage collection and all that.	00:11
	← numberGrey left	00:11
	→ anjen1 joined	00:11
Arafangion	Anyway, it's non-trivial, that's the main point that I should accept.	00:12
killerchicken	sitaram: hey, next gitolite question, this time less stupied, I hope: We currently have repositories for different sub-projects and different developers grouped together in a directory structure like this: repositories -> project -> devname\|projectname	00:12
	← xlistener left	00:12
killerchicken	sitaram: can I use gitolite to allow access to such subdirectories?	00:12
	→ numberGrey joined	00:13
	← r11t left	00:13
	→ patmaddox joined	00:13
	← Weasel[DK] left	00:13
killerchicken	sitaram: for example, can I say "repo project1/dev1" to give permissions to the repository dev1 in the subdirectory project1?	00:13
	→ \|Inthewings\| joined	00:14
	← catalis left	00:14
Fazer2	so can you suggest how can I solve my problem with push?	00:14
_Vi	Fazer2, What URL?	00:15
	← sgh left	00:16
Fazer2	_Vi: [email@hidden.address]	00:16
floppyears	hi guys	00:16
	→ dgarr joined	00:16
floppyears	is there a good command line tool that I can use to help me during merges and conflict resolution?	00:16
_Vi	Fazer2, "$ ssh -v [email@hidden.address] -> "......", "Permission denied (publickey).". What for you?	00:16
cbreak	vim is good	00:17
_Vi	floppyears, "git status", "git add"?	00:17
	→ chalkbag joined	00:17
_Vi	floppyears, You also may use your primary editor (for text files).	00:17
	→ jmcantrell joined	00:17
	→ jerbear joined	00:17
floppyears	_Vi: no, I mean something kinda of merge or diff tool	00:18
Fazer2	_Vi: it doesn't say anything :-/	00:18
cbreak	git diff does show diffs	00:18
	git merge does merges	00:18
	← rektide left	00:18
	→ Tricon joined	00:18
	→ rektide joined	00:18
_Vi	Fazer2, Does it hang or exit?	00:18
cbreak	when there are conflicts, you have to resolve them yourself	00:18
_Vi	floppyears, If conflict, it changes the file, instering "<<<<<<<", "=====", ">>>>>>>>" so you can find them and fix.	00:19
Fazer2	_Vi: it doesn't say anything for a couple of seconds and then shows me command prompt	00:19
	← jerbear left	00:19
_Vi	floppyears, After fixing just "git add".	00:19
	← jmcantrell left	00:19
_Vi	Fazer2, Even with "-v" options?	00:19
floppyears	I know, but I know you can use gui tools to assist you and show you the diffs. You can then walk through the file and make the modifications or chose whichever version you want	00:19
Fazer2	_Vi: yeah, even with -v	00:20
	→ catalis joined	00:20
floppyears	for example, I think you can use kdiff as a diff program to assis you during three way merge	00:20
_Vi	Fazer2, 1. Try "ssh -v -v -v [email@hidden.address] 2. Try "ssh -v -v -v [email@hidden.address]	00:20
	elliottcable → ASHEEP	00:20
	ASHEEP → NOTREALLY	00:20
_Vi	Fazer2, E.g. test if your SSH client OK at all.	00:21
	NOTREALLY → elliottcable	00:21
Fazer2	_Vi: the first connection says nothing; how can I test my SSH client?	00:21
	← ph^_ left	00:21
_Vi	Fazer2, (Also you may try "ssh -v -v [email@hidden.address] -p 2222". It should succeed after asking the password)	00:22
Fazer2	_Vi: that last one doesn't work either	00:23
_Vi	Fazer2, Are there any firewalls or AVs or other things may intercept? Has you tried reinstalling msysgit? (The Faith of Windows users).	00:23
Fazer2	_Vi: hehe	00:24
_Vi	Fazer2, Has you tried just reboot Windows?	00:24
Fazer2	_Vi: I don't have AV	00:24
_Vi	s/Faith/Fate/	00:24
Fazer2	_Vi: but I have a firewall	00:24
_Vi	Fazer2, Looks if it block something.	00:24
	→ mnyon joined	00:25
	→ xlistener joined	00:26
_Vi	Fazer2, What network utilities do you have? Do you have Putty? Do you have Netcat?	00:26
Arafangion	cbreak: Interestingly, gitx is in objective-c++.	00:26
cbreak	even worse	00:26
floppyears	what do you guys use with git-mergetool	00:27
Arafangion	Indeed, unless it is mostly C++ with a thin objective-c binding to the carbon, but that doesn't seem to be the case.	00:27
Fazer2	_Vi: I disabled windows firewall, but I guess there's still a router firewall	00:27
_Vi	Fazer2, Is your laptop also on Windows with the same msysgit?	00:27
Fazer2	_Vi: I don't have Putty nor Netcat	00:27
	← JakeSays left	00:27
_Vi	Fazer2, Does you laptop also uses the same firewall and the same router?	00:28
	Fazer2, Try all that "ssh -v ..." on laptop.	00:28
	→ pklingem2 joined	00:28
Fazer2	Fazer2: I think it has the same Msysgit, but I worked on my laptop in the different town and I don't have it with me now	00:28
	← d2dchat left	00:28
_Vi	Fazer2, Download&install Putty. It is the Standard SSH Client for Windows.	00:29
	Fazer2, (at least very popular)	00:29
Fazer2	_Vi: ok	00:29
	→ r11t joined	00:29
	← InTheWings left	00:29
_Vi	Fazer2, If Putty will work connecting to SSH, the network&firewall is not the cause.	00:30
	→ jumbers joined	00:30
jumbers	How can I roll back a single, uncommitted file?	00:31
	I accidentally edited a file and now I can't pull down the upstream	00:31
_Vi	jumbers, What do you mean by "rollback" and "uncommited"?	00:31
	jumbers, "git checkout -- path/to/file"?	00:31
	→ JakeSays joined	00:32
	→ timcharper joined	00:32
	→ Fantuz83 joined	00:35
Fazer2	_Vi: Putty seems to work	00:36
	_Vi: for instance, I can connect to [email@hidden.address] and it asks me for password	00:36
_Vi	Fazer2, Tried updating/reinstalling msysgit?	00:37
	Fazer2, (You also can migrate to Cygwin/Git).	00:37
	← Twisted_Mentat left	00:37
	← pantsman left	00:37
	← rettub left	00:37
Fazer2	_Vi: I'll try it now	00:37
	→ Twisted_Mentat joined	00:38
	← smuf left	00:39
	→ schacon joined	00:40
	← patmaddox left	00:40
	→ smuf joined	00:40
	→ rettub joined	00:40
	← hipe left	00:40
	→ loinclot_ joined	00:41
	→ robinr joined	00:43
Fazer2	_Vi: it worked :-D	00:43
_Vi	Fazer2, Updating or reinstalling?	00:43
Fazer2	_Vi: reinstalling	00:43
_Vi	Fazer2, Strange error. May be the ssh binary was damaged?	00:44
Fazer2	_Vi: it still doesn't give me info what exactly was wrong, but what can I expect from windows ;-)	00:44
_Vi	Fazer2, (no malware installed? >"I don't have AV')	00:44
Fazer2	_Vi: I'm a consious user	00:45
	_Vi: I don't download pirated apps and so on	00:45
	_Vi: but maybe I'll make a scan tomorrow, just to be sure	00:46
_Vi	Fazer2, (Closed RPC DCOM, NetBios and all unknown ports, not running as admin priveleges, not running IE?)	00:46
Fazer2	it's 1:45 AM here =_=	00:46
	→ JoeConyers joined	00:46
	→ Kelly_zhou joined	00:47
Fazer2	_Vi: heh, I don't know what RPC DCOM and NetBios are, as for the rest, I'm using mostly open source apps	00:47
	← project2501a left	00:47
Fazer2	_Vi: thanks for the help and good night :)	00:48
	← rektide left	00:48
	→ rektide joined	00:48
	← Fazer2 left	00:49
	← altrux left	00:50
	→ jelmer joined	00:54
	← jrmuizel left	00:54
	← nitrate_ left	00:55
	→ Geralt joined	00:59
	← nadavoid left	00:59
	← catalis left	01:01
	← r11t left	01:01
	→ sourcemaker_ joined	01:01
	← Tricon left	01:01
	→ keex joined	01:02
keex	anyone here using git-cola ?	01:02
	← Swimming_Bird left	01:02
_Vi	keex, I only have it's name in Wikipedia.	01:02
keex	hmm	01:02
	too bad... :/	01:03
_Vi	keex, What's the problem?	01:03
	→ Swimming_Bird joined	01:03
	→ rchavik joined	01:03
	← mnyon left	01:03
keex	the context menu of the diff view does not show "stage/unstage selected lines"	01:03
	I have pulled to the newest version..	01:04
_Vi	keex, "git gui" does.	01:04
keex	I had that functionality once..	01:04
_Vi	keex, Previous version showed?	01:04
keex	oh yeah!	01:04
	← Swimming_Bird left	01:05
keex	_Vi, not the previous, but 239423 versions before... unfortunately i've never paid attention to the actual version	01:05
_Vi	keex, It they has broken it, downgrade to the older version and file a bug.	01:05
	→ maglos joined	01:05
_Vi	keex, "git bisect".	01:05
keex	hehe	01:05
jujule2	hi	01:05
keex	yes, a perfect application for bisect	01:05
	jujule2, hi	01:06
jujule2	how to 're-checkout' a submodule ? it appears as modified in git status but id like to cancel my changes to be able to pull again :)	01:06
_Vi	jujule2, "git submodule update" fails to do this?	01:07
jujule2	error: pathspec '[...]' did not match any file(s) known to git.	01:07
	cant get it	01:08
_Vi	jujule2, Strange error. What pathspec?	01:08
jujule2	what do you mean ?	01:08
_Vi	jujule2, You may try to checkout the submodule manually.	01:08
	jujule2, " pathspec '[...]' did" "[...]"=?	01:09
jujule2	wow :)	01:09
_Vi	jujule2, The problem has been resolved?	01:10
jujule2	no. cant understand your cmd	01:10
_Vi	jujule2, 1. Run "git diff" to see the old commit id. 2. chdir to submodule directory, 3. "git checkout <that_commit_id>"	01:11
_pointlessjon	i think that was a question	01:11
	← RaceCondition left	01:11
	→ loincloth joined	01:11
_Vi	_pointlessjon, Yes, that ("[...]"=?) was a question.	01:12
keex	got to go	01:12
	c ya	01:12
	← keex left	01:12
jujule2	same error :/	01:12
_Vi	_pointlessjon, >'cant understand your cmd' Is it about "git submodule update"?	01:12
	jujule2, What error on what command?	01:13
	→ RaceCondition joined	01:13
jujule2	git checkout in submodule path with last commit#	01:13
_Vi	jujule2, The checkout gives "error: pathspec ..."?	01:14
jujule2	yes. i use github. something must be broken in my urls	01:14
_Vi	jujule2, What does "git status" say in the submodule directory?	01:14
	jujule2, What URLs?	01:15
jujule2	# Not currently on any branch.	01:15
_Vi	jujule2, What git version? (git --version)	01:15
jujule2	my urls i mean the submodule git repo url maybe, i check	01:15
	git version 1.5.6.5	01:15
	← numberGrey left	01:15
_Vi	jujule2, Outdated. Current is 1.6.x.	01:16
	← dreiss left	01:16
_Vi	jujule2, May be first try update Git (to ensure we are talking about the same thing)	01:17
jujule2	well. using the github commit id it works. :8 sorry for the noise	01:17
	← schierbeck left	01:17
jujule2	but still appears as modified in git status in the main repo	01:18
_Vi	jujule2, What does "git diff" say about it?	01:18
	← eletuchy left	01:19
jujule2	-Subproject commit be8d929e17337a51639253f2c2944d4307441992	01:19
	+Subproject commit 8af073a8006ef5e30c17ef2a59abd2f2d9571cdc	01:19
_Vi	jujule2, 1. Enter subproject directory, 2. "git checkout be8d929e17337a51639253f2c2944d4307441992". Fails?	01:19
jujule2	yes	01:19
	← kpreid_ left	01:20
_Vi	jujule2, 1. Enter subproject directory, 2. "git fetch", 3. "git checkout be8d929e17337a51639253f2c2944d4307441992". Also fails?	01:20
	→ kpreid_ joined	01:20
_Vi	jujule2, What does "git fetch" say?	01:21
jujule2	note: some time ago the submodule url changed but i edited all config files. works ok on others machines. maybe something gone bad on this one	01:21
	→ jinzhu joined	01:21
jujule2	git fetch just says nothing	01:21
_Vi	jujule2, May be just remove offending submodule and recheckout it?	01:22
	jujule2, "git submodule update --init".	01:22
jujule2	i just rm the dir ?	01:22
_Vi	jujule2, Just remote the dir and then "git submodule update --init".	01:23
	jujule2, (But I'm not sure about submodules in Git 1.5)	01:23
	→ r11t joined	01:23
jujule2	i still get the same error and diff	01:24
	possible to reset the index for a submodule ?	01:24
	← name left	01:25
	← kennethlove left	01:25
_Vi	jujule2, May be better try to update Git before coping with particular problem?	01:25
	→ EricInBNE joined	01:26
	← spvensko left	01:27
	← loinclot_ left	01:28
	→ spvensko joined	01:29
	← ereslibre left	01:31
	← defectiv left	01:31
	→ numberGrey joined	01:34
	→ rgr joined	01:34
	→ jrmuizel joined	01:36
rgr	I have a project which is a clone of a bare reop via ssh. I this project I add a submodule "lib". The question I have is whether I should "add" "lib" to the parent project or not? Using magit any changes in lib are not shown in a git status for the parent project anyway.	01:36
	s/reop/repo	01:36
	→ onigiri_ joined	01:36
	← _pointlessjon left	01:36
	← voxpelli-laptop left	01:36
	← numberGrey left	01:37
	← cenuij left	01:37
_Vi	rgr, "add" when doing what? When changing the parent project?	01:37
	← Ryback_ left	01:38
rgr	when doing a status of the "container" peoject (the one were lib has been added a submodule) the very presence of the lib subdir means its listed as a new untracked file.	01:38
	→ squarism joined	01:38
_Vi	rgr, New untracked file? Are you adding a submodule or checking it out?	01:39
	rgr, Are you introducing the submodule or only updating it?	01:39
	← Geralt left	01:41
	→ martiin joined	01:41
	→ rolfb joined	01:42
	← amerine left	01:42
	→ jmcantrell joined	01:45
	→ Swimming_Bird joined	01:45
	← Fantuz83 left	01:47
ShadeHawk	How to download and install JGit and EGit?	01:48
_Vi	ShadeHawk, EGit - build from sources in Eclipse.	01:48
	→ dogmatic69 joined	01:48
_Vi	ShadeHawk, (EGit) And copy files into Eclipse's dierctory.	01:48
dogmatic69	hi all...	01:48
ShadeHawk	_Vi: thanks	01:48
dogmatic69	im running vista and have set my ide to linux line endings and now git is giving warning all the time	01:49
ShadeHawk	_Vi: I have thought that it would come as a nice package...	01:49
_Vi	ShadeHawk, May be it comes, may be not. Haven't used it for long.	01:49
	s/used it/experimented with it/	01:49
dogmatic69	the error is warning: LF will be replaced by CRLF in <file>	01:50
	← rolfb left	01:50
dogmatic69	and does this for every file i add	01:50
_Vi	dogmatic69, What line endings your checked in files are actually has?	01:51
sitaram	killerchicken: if you're asking "do all repos have to be in a flat structure at the top level or can they also be within directories, etc?", yes they can	01:51
dogmatic69	_Vi: it should be linux style... LF afaik... windows is normaly crlf?	01:51
_Vi	dogmatic69, Use "cat -v filename" or hex editor to verify.	01:52
	← timcharper left	01:52
_Vi	dogmatic69, Everything is normally LF, but many things in Windows prefer (or can handle only) CRLF.	01:52
	→ ideogon joined	01:52
rgr	_Vi: when I add the submodule it creates a dir "lib".	01:52
dogmatic69	_Vi: that cat -v just showed me the file?	01:53
_Vi	dogmatic69, May be after you have switched your IDE to LF it starts to insert LFs preserving old CRLFs?	01:53
	← t3h_awesome left	01:53
_Vi	dogmatic69, Without "^M"s at the end of each lines?	01:53
	dogmatic69, Has you set some git attributes for CRLF handling?	01:53
killerchicken	sitaram: ah nice	01:54
	sitaram: Is that somewhere in the docs? I didn't find it	01:54
dogmatic69	_Vi: ye it looks normal just like in my ide	01:54
_Vi	rgr, Yes. And what's wrong?	01:54
killerchicken	and how do I specify it?	01:54
dogmatic69	_Vi: if i do a old file that was set to dos line endings it has the ^M	01:55
	→ johnw joined	01:56
	→ numberGrey joined	01:56
	→ cedricv joined	01:56
sitaram	killerchicken: I guess I thought it was too obvious :) Just say foo/bar	01:56
dogmatic69	_Vi: so wat you think? seems it is the correct line ending but git is converting it/	01:57
killerchicken	sitaram: ok, I wasn't sure	01:57
	becuase you asked that only regular chars be used as repo name	01:57
_Vi	dogmatic69, May be try setting core.autocrlf to false?	01:57
	killerchicken is officially asking most stupid questions ;)	01:57
	← Alagar left	01:57
_Vi	dogmatic69, Has you touched any crlf-reladed setting or attribute?	01:58
	dogmatic69, (when you was using old mode with CRLFs)	01:58
	← kmartiin left	01:58
dogmatic69	_Vi: in git, maybe... a while ago	01:58
rgr	whats wrong is that the submodule appears as an item. Thas caused issues with pulling. The other issue is that a git status does not show the status of the submodule and I was ondering if it can. Yes, IF I pushed the entire submodule then the "Lib" appears changed in the parent. Is it possoible to tell git to check all subnmodules too( e.g f1 changes in proj2/lib,	01:58
_Vi	dogmatic69, So may be now it's time to turn it off?	01:58
	→ reprore joined	01:59
	← reprore left	01:59
dogmatic69	i guess so... now just to figure out how	01:59
_Vi	dogmatic69, May be you should also convert the whole thing to LF?	01:59
dogmatic69	how you mean?	01:59
_Vi	dogmatic69, (To avoid variance of line endings)	01:59
	dogmatic69, Filter all files to substitute CRLFs with LFs globally.	02:00
dogmatic69	that is what i wanna do now, go through each file and make them lf	02:00
	can git do that?	02:00
_Vi	dogmatic69, May be even filter all commits to solve the problem radically.	02:00
dogmatic69	cool... i would not know where to start though	02:01
_Vi	dogmatic69, something like "find -type f \| xargs perl -i -pe 's/\r\n/\n/g'" can do it.	02:01
	← \|Inthewings\| left	02:01
_Vi	dogmatic69, All CRLF files will be changed. You may either commit it. Cleaner solution will be convert ALL commits to LF (from the first one to the last one).	02:02
	→ Theravadan joined	02:02
	← rektide left	02:03
_Vi	dogmatic69, But this (filter-branch for LF in all commits) may be really long.	02:03
dogmatic69	_Vi: ok, i did: git config --global core.autocrlf false and commited something and there was no errors	02:04
	thanks :)	02:04
	not shure about doing it globaly aprat from opening each file now	02:04
	← bentob0x left	02:05
	→ djork joined	02:06
	→ rektide joined	02:06
dogmatic69	_Vi: what is "find -type f \| xargs perl -i -pe 's/\r\n/\n/g"	02:06
	how would i use that?	02:06
	→ meanburrito920_ joined	02:07
_Vi	dogmatic69, "find -type f" - find all files; "perl -pe 's/\r\n/\n/g'" - change all CRLF to LF;	02:07
	dogmatic69, "perl -i <commands> <files>" - change files in place.	02:07
	dogmatic69, All together: find all files in current directory and filter then through changer of CRLF to LF.	02:08
	dogmatic69, (xargs <command> is call <command> for each incoming line)	02:08
dogmatic69	_Vi: sorry.. im really no good at this cmd stuff.. would i do this as one command? "find -type f \| xargs perl -i -pe 's/\r\n/\n/g"	02:10
	→ project2501a joined	02:10
	→ pinnacle joined	02:10
_Vi	dogmatic69, >"how would i use that?" Go to repository root and paste in the console. You probably should change "file -type f" to something like "find . -type f -and -name '*.cpp'".	02:10
dogmatic69	cool thanks for the help	02:11
	much appreciated :)	02:11
_Vi	dogmatic69, Yes, but be preparet as it will convert binary files as well.	02:11
	← Bryson left	02:11
_Vi	s/preparet/prepared/	02:11
dogmatic69	:S so all my images will die?	02:12
_Vi	dogmatic69, 1. You may use for everything but restore images later. 2. You may change "find" option to make it find only your code, not binaries.	02:12
	← project2501a left	02:12
dogmatic69	cool	02:13
	→ project2501a joined	02:13
	← djork left	02:13
_Vi	dogmatic69, Simple thing will be 1. Convert everything (pictures broken). 2. Add and commit only code, 3. Restore pictures.	02:13
	← sumangurung left	02:15
killerchicken	sitaram: hrm. Something seems to not work with the subdirectories	02:15
sitaram	hmm?	02:15
	in what sense?	02:15
	aah	02:15
	→ drogue joined	02:15
dogmatic69	_Vi: ye... ill prob just have to do it manualy cos i go reformat my code also...	02:15
killerchicken	I added "repo test/test" with rw perms	02:15
	I cannot clone this repository	02:15
	→ djork joined	02:16
	→ AAA_awright_ joined	02:16
_Vi	dogmatic69, BTW reformatting the code can also be (partially) automatized.	02:16
	→ sumangurung joined	02:16
sitaram	what does "ssh git@server" say?	02:16
dogmatic69	my ide does that... i just have to open the file... but i can open a bunch and do all so its not to bad.	02:16
killerchicken	wargh	02:17
	→ thmghtd joined	02:18
dogmatic69	thanks for the time _Vi	02:18
killerchicken	I tried the wrong server. I'm so stupid today (I hope it is just today? ;/) I won't ask a new question.	02:18
	have been sitting in front of this for half an hour now	02:18
	← rgr left	02:18
killerchicken	sitaram: sorry :(	02:19
sitaram	:)	02:19
killerchicken	but thanks for the tool. It's working great	02:19
sitaram	ywyw!	02:20
	you're welcome!	02:20
	→ doxtor joined	02:20
doxtor	Hello, I have two branches and have overwritten newer files in one branch by git checkout from second branch. Is there a way to recover overwritten files?	02:21
killerchicken	"Why do you not publish your code?" "I hate users asking dumb questions" :-)	02:21
	doxtor: if you haven't commited them, git checkout -- <path>	02:21
	will check them out from current HEAD	02:22
_Vi	doxtor, "git checkout <first_branch> -- path/to/file"	02:23
	← lll left	02:23
	← dogmatic69 left	02:23
	← AAA_awright left	02:24
doxtor	That doesn't recover the newest version	02:26
	files won't even appear in "Changes to be committed:" list	02:26
_Vi	doxtor, Where has you committed the newest version?	02:26
doxtor	_Vi: I haven't committed	02:27
_Vi	doxtor, The file is actually unchanged?	02:27
	doxtor, Not commited? "checkout" should complain.	02:27
	→ mithro joined	02:27
_Vi	doxtor, Has you used general checkout or with file argument?	02:27
doxtor	with file argument	02:27
	I wasn't switching branches	02:27
_Vi	doxtor, So the previous version seems to be just lost.	02:28
doxtor	yeah	02:28
_Vi	doxtor, It was only in index (or only in work tree) and you have explicitly checked it out.	02:28
doxtor	looks like it	02:29
	← m104 left	02:29
doxtor	funny thing is I have compiled new version still around and no source	02:29
_Vi	doxtor, How precious is it? How much time the manual restoration can take?	02:29
doxtor	I guess 2 hours	02:29
_Vi	doxtor, Try "git fsck" for missing blobs.	02:30
	s/missing/dangling/	02:30
	doxtor, May be one of them will be yours.	02:30
	doxtor, Has you at least staged your version of file to index?	02:30
doxtor	no	02:31
	→ lll joined	02:31
_Vi	doxtor, (So don't try "git fsck")	02:31
doxtor	ok, thanks	02:31
_Vi	doxtor, /* A shoot in a foot */	02:31
	doxtor, Probably the easiest way will be just change it again.	02:32
	→ dreiss joined	02:32
	← WinterWeaver left	02:32
	→ upul joined	02:33
	← martiin left	02:34
	← johnw left	02:34
	← smuf left	02:38
	→ jlilly joined	02:38
	← remi left	02:39
	→ Bryson joined	02:40
killerchicken	sitaram: hah, I'm so bad at keeping promises. I see that I can add custom hooks, that are copied to all repositories. Is there a feature (or is one planned) to specify which hooks should be executed? The issue is that for some of our repositories, we want commits sent to a mailing list, for others, we want them to go to a different list, etc.	02:46
	→ fullTummy joined	02:46
killerchicken	We could of course put all that logic in the hook itself, but that seems bad.	02:46
fullTummy	hi all	02:46
killerchicken	Making hooks available via easy-install and then enabling/disabling them via config would be cool.	02:47
	← ShadeHawk left	02:47
ilumi	what does this error mean?	02:47
	/usr/bin/ld: cannot find -lexpat	02:47
	collect2: ld returned 1 exit status	02:47
	make: *** [git-remote-curl] Error 1	02:47
fullTummy	expat shared library not found	02:47
_Vi	ilumi, That you can't find "libexpad" library.	02:47
ilumi	what do i need to install?	02:48
_Vi	ilumi, Try installing it. Something like "libexpad-dev".	02:48
	ilumi, What system?	02:48
ilumi	ubuntu	02:48
fullTummy	get windows 7	02:48
	it comes with it :D	02:48
ilumi	lol, no	02:48
_Vi	ilumi, apt-get install libexpat1-dev	02:48
	ilumi, Works?	02:48
killerchicken	sitaram: I could make the hook depend on git config options, I guess.	02:49
_Vi	ilumi, (Actually, "sudo apt-get install libexpat1-dev")	02:49
	→ Twisted_Mentat_M joined	02:49
	← Twisted_Mentat_M left	02:49
ilumi	_Vi: works, thank you	02:50
	← Bryson left	02:50
	→ mrpeterpeter joined	02:50
	← WALoeIII left	02:51
	← mrpeterpeter left	02:51
	→ VVSiz_ joined	02:52
	← em-dash left	02:55
	← sunaku left	02:57
	← coolcat left	03:00
	← fullTummy left	03:01
	← xlistener left	03:01
	← loincloth left	03:01
	→ ken___ joined	03:01
	AAA_awright_ → AAA_awright	03:02
	← hummingkid left	03:03
	← djork left	03:04
	→ xlistener_ joined	03:05
	xlistener_ → xlistener	03:05
	→ montylounge joined	03:06
	← upul left	03:08
	← VVSiz left	03:09
	→ nadavoid joined	03:10
	→ ali1234 joined	03:12
sitaram	killerchicken: that is what teukka does; he added the feature to add repo specific config to gitolite.conf and he then uses that. You could also just check $GL_REPO inside the hook code and do different things accordingly	03:12
	→ d2dchat joined	03:13
killerchicken	sitaram: right, but that would mean updating the hook for everyone whenever I add a new repo ;)	03:13
	← jrmuizel left	03:15
sitaram	killerchicken: yes; the config option is better for most long term uses. I'm trying to design a clean way in which you can specify hooks for different repos but it's not high on my list right now due to work and other pressures	03:15
killerchicken	sure	03:15
	ah	03:15
sitaram	killerchicken: clean means it should not cause even more divergence between master and wildrepos :-)	03:15
killerchicken	when I place a new hook in src/hooks, I have to always do easy-install again, right? I cannot do that via an admin account? I think the docs are a little ambigious here	03:16
sitaram	killerchicken: you have to easy install again. That is a wart that will almost certainly get fixed; this should be doable by pushing admin repo, ideally. When I originally designed it, the idea was "how frequently will people be changing hooks? Not much"	03:17
killerchicken	no	03:17
	I think it should not.	03:17
	That would be a security issue, wouldn't it?	03:17
	Here is a hook that erases logs	03:17
	here is a hook that spawns a shell	03:17
	etc	03:17
sitaram	killerchicken: there are two ways to look at that	03:18
killerchicken	I don't want my admins to be able to mess with the log	03:18
sitaram	killerchicken: good point	03:18
	killerchicken: you want to separate people who have shell access to the server from people who are able to mess with the admin repo	03:19
killerchicken	that's what I'm doing	03:19
	← hyperair left	03:19
	→ hyperair joined	03:19
killerchicken	none of the admins except myself have shell access.	03:19
	and I don't use @SHELL	03:19
	at all	03:19
	But it seems that if they could get arbitrary code to be executed, that would be bad.	03:20
	like, really bad.	03:20
	← aziz left	03:20
	→ cytrinox_ joined	03:20
sitaram	yes; now that you mention it, I have a feeling that was my original motivation	03:20
killerchicken	In this case, let me update the docs.	03:21
sitaram	or not; I only see a security note saying "if you mess with the update hook, you lose all your per-branch permissions" so you make a very valid point	03:21
	yes please	03:22
	send me a patch via email if you have time	03:22
killerchicken	I was going to send a pull request?	03:22
sitaram	sure	03:22
killerchicken	:q	03:22
	ups	03:22
	(Too bad, I prevented myself from getting those $$$ easily by making you aware of the issue before you put it in. Man, I'm so dumb)	03:23
	← tedoc2000 left	03:23
killerchicken	:-)	03:23
	← Travis-42 left	03:23
sitaram	killerchicken: wouldn't have worked: that security warning says if you can do something the config file says youshouldn't...	03:23
killerchicken	ah	03:24
	← project2501a left	03:24
killerchicken	hm	03:24
	damn. nitpicker	03:24
sitaram	killerchicken: also, currently it doesnt even apply... better nit?	03:24
	(I mean I havent coded it yet; you stopped me!)	03:24
killerchicken	that's what I was saying :P	03:24
sitaram	ah yea	03:24
	← whaley left	03:25
	→ sunaku joined	03:25
	← chalkbag left	03:27
	→ jhelwig_ joined	03:28
	← jhelwig left	03:29
zahna	did the puppet.conf file even change from 0.24 to 0.25?	03:30
	jhelwig_ → jhelwig	03:31
	→ jrmuizel joined	03:32
	← maglos left	03:32
	→ Tricon joined	03:34
	→ jessep joined	03:36
	← tfarina left	03:36
	← cytrinox left	03:37
	cytrinox_ → cytrinox	03:37
	← GeertB left	03:38
	← schacon left	03:39
	← spearce left	03:43
	← sunaku left	03:45
	→ catalis joined	03:45
	→ sunaku joined	03:46
	→ cedric_ joined	03:46
	← sunaku left	03:51
	→ sunaku joined	03:52
	← cedricv left	03:52
	← thmghtd left	03:52
	→ tjafk joined	03:52
	← schambers left	03:52
	→ upul joined	03:53
	← tjaway left	03:53
	→ schambers joined	03:54
	← hobodave left	03:56
	→ j416 joined	04:01
	← spvensko left	04:02
jayne	ugh. Has anyone gotten gvimdiff to work as a difftool on Win32?	04:02
	if so, how?	04:02
	→ spvensko joined	04:04
	→ gno` joined	04:05
jayne	specifying a .cmd file for difftool.<tool>.cmd causes inexplicable syntax errors; specifying cmd /c gvim.bat starts a new cmd without passing the /c option; specifying gvim.exe doesn't work because it doesn't use -f; using start causes some _really_ weird path quoting errors.	04:05
	← gno` left	04:07
offby1	welcome to Windows	04:08
	← kukks left	04:11
	← spvensko left	04:11
jayne	offby1: this is very clearly a git problem, not a Windows problem	04:11
	← ken___ left	04:11
jayne	gvimdiff itself works fine, and if I bypass difftool and do it manually, it works (again) fine	04:11
	I can even use GIT_EXTERNAL_DIFF with just git diff and it works fine.	04:12
	but git difftool blows up in the ways described above	04:12
offby1	I suspect it's that git doesn't do anything special to pass the arguments properly	04:13
jayne	the gvim.bat case appears to be caused by sh trying to blindly read the argument as some sort of sh script	04:13
	← scarabx left	04:18
	← anjen1 left	04:21
	→ hobodave joined	04:23
	→ cenuij joined	04:24
	← hakunin_ left	04:25
jayne	that's why... echo "$(git config difftool.$merge_tool.cmd \|\| git config mergetool.$merge_tool.cmd)"	04:26
	→ warthog9 joined	04:26
	→ hakunin joined	04:27
	← hakunin left	04:27
	→ hakunin joined	04:28
	→ mm_202 joined	04:28
	← erikvold left	04:29
mm_202	Is there an easy way to get the most recent commit hash from a [different] branch?	04:29
	I want to do a git cherry-pick, but am getting tired of copy/pasting the hash from the previous commit.	04:29
	→ hakunin_ joined	04:30
_Vi	mm_202, git log <branch> \| head -n 1	04:30
	← hakunin_ left	04:30
_Vi	mm_202, You may supply any commit-ish to "cherry-pick"	04:30
	→ hakunin_ joined	04:30
	→ \|Rolenun\| joined	04:30
mm_202	_Vi: ooo, yes, that worked!	04:31
_Vi	mm_202, Examples: master, master^, your_branch, your_branch~2, HEAD@{1}.	04:31
	← Rolenun left	04:31
_Vi	mm_202, You may use "git cherry-pick <branch>"	04:31
mm_202	Yep, that's what I did.	04:31
	→ hakunin__ joined	04:32
	← agib left	04:32
	← aresnick left	04:32
jayne	git-mergtool--lib rather stupidly assumes that if the given non-builtin diff tool isn't in the current directory that it doesn't exist at all	04:33
mm_202	Is there any 'shortcut' to: git add [file], git commit [msg], git stash, git co [branch1], git cherry-pick [branch2], git co [branch2], git stash apply	04:33
jayne	if ! type $merge_tool_path then die (to paraphrase)	04:33
mm_202	sorry, git co = git checkout	04:33
	← doxtor left	04:33
	← hakunin__ left	04:34
_Vi	mm_202, I think 100% of aliases of "git co" point to "git checkout".	04:34
	← montylounge left	04:35
	→ hakunin__ joined	04:35
mm_202	_Vi: Im sure as well, but I got yelled at for it once in here, so I just try to avoid aliases altogether	04:35
	← schambers left	04:35
	→ hakuni___ joined	04:36
	→ schambers joined	04:36
mm_202	_Vi: or am I the only one that uses that 'work flow'?	04:36
	← hakuni___ left	04:37
	← rettub left	04:40
	→ Rolenun joined	04:40
	→ johnw joined	04:40
	→ mmoya joined	04:41
	← nothingHappens left	04:43
jayne	offby1: yeah, the problem is sh.exe eats all arguments starting with /	04:43
	→ loincloth joined	04:43
jayne	and (I don't get this part) tries to convert them into a filename	04:44
	→ rettub joined	04:44
jayne	so "attrib /s foo" gets turned into "attrib s:/ foo"	04:44
cbreak	makes sense	04:44
	windows doesn't have a root file system	04:45
	but drive letters	04:45
offby1	jayne: that sounds crazy ...	04:45
jayne	offby1: I agree	04:45
cbreak	so absolute paths starting with / have to be converted	04:45
offby1	but who would be so foolish as to convert _all_ arguments, as if each was a path?	04:45
jayne	cbreak: not all that starts with / is a path	04:45
	→ tedoc2000 joined	04:46
	← jhelwig left	04:46
cbreak	yeah. could be a random string that happens to start with it	04:46
	but not everything that starts with \ is an escape sequence either	04:46
	← hakunin left	04:46
cbreak	shells seem to assume some things	04:47
	like * being a glob pattern, or \ an escape sequence start, or / denoting the root volume	04:47
	in this case	04:48
	I guess	04:48
	← hakunin_ left	04:48
	→ jhelwig joined	04:49
	← \|Rolenun\| left	04:49
	← d2dchat left	04:50
jayne	I got it \o/	04:50
	cmd \\\\/c gvim.bat -your -args -here	04:51
	so having done that... now why does git status think the files are different when git diff reports them char-for-char identical? (No, it's not CRLF) :-(	04:52
	← jhelwig left	04:52
	→ jhelwig joined	04:52
	→ harryxu joined	04:52
cbreak	openssl sha1?	04:52
	tab vs spaces?	04:52
	(-> whitespace)	04:53
	← jmcantrell left	04:53
offby1	sounds like core.filemodes or core.crlf	04:53
cbreak	tried git diff?	04:53
	← harryxu left	04:54
	→ harryxu joined	04:54
	← harryxu left	04:54
	→ __Joker joined	04:54
	→ harryxu joined	04:55
	← hakunin__ left	04:55
	← loincloth left	04:55
	→ loincloth joined	04:56
jayne	the version on each side of git diff has the same sh1	04:57
	← JoeConyers left	04:57
Ilari	jayne: In what section(s) is that file listed as 'modified'?	04:58
jayne	changed but not updated	04:58
Ilari	jayne: But not 'to be committed'?	04:58
jayne	correct	04:58
Ilari	jayne: And 'git diff -- <file>' lists same hash twice in index line?	04:59
jayne	no, it's different there	04:59
	→ thomaz joined	05:00
	→ psankar joined	05:00
Ilari	jayne: Is there any diff hunks at all in diff output?	05:01
	← cloudhead left	05:01
	→ cloudhead joined	05:01
	← cloudhead left	05:01
	→ tedoc2000_ joined	05:02
jayne	seemingly the entire file.	05:02
	I just reset --hard and the problem seems to be gone though	05:02
	← dgarr left	05:03
	← jessep left	05:03
Ilari	jayne: Yes, because all uncommitted changes were overwritten.	05:03
jayne	I tried changing autocrlf just to see if I could reproduce it, but no	05:03
	Ilari: there were no visible uncommitted changes	05:03
Ilari	jayne: Maybe the file indeed got CRLFs from somewhere and autocrlf was off (changing it without checking out again seems problematic).	05:04
	← Yuuhi`` left	05:04
	→ Yuuhi`` joined	05:04
Rolenun	lol, you guys are killing me with the crlf discussion :D	05:04
	→ john268 joined	05:04
jayne	whatever it was, it's gone now	05:05
	→ sdboyer_ joined	05:06
Rolenun	I'd be wary about your GUI altering the files line endings automatically too. it's a common windows complaint.	05:06
	→ agenteo joined	05:07
	→ munichlinux joined	05:07
jayne	the file I have now is the same sha1 as I got on both sides of the diff earlier, and changing the line endings still doesn't get me the mysterious "wrong" sha1	05:07
Rolenun	Ilari: it was funny because I just spent a couple hours evaluating jgit's lack of autocrlf support. it has to be the biggest issue with jgit imho. hehe	05:08
jayne	I wish there was a "this repository's line endings are (LF\|CRLF)" config option and a "my checkout should have (LF\|CRLF) line endings" option. It'd be less confusing, I think.	05:09
	← tedoc2000 left	05:09
jayne	having the two set differently would imply that git should attempt conversion	05:09
Rolenun	hmmm	05:09
	← tedoc2000_ left	05:10
jayne	ooo. branch switching is _way_ faster in 1.6.5. nice.	05:11
	← jelmer left	05:11
Rolenun	... it's git lol, it's always fast :D	05:11
jayne	Rolenun: previously on Win32, it was painfully slow	05:11
	it's always been fast on *nix	05:11
Rolenun	ohhh, you're using windows :D	05:12
	Rolenun comforts you.	05:12
Rolenun	:D	05:12
Ilari	jayne: The repository line endings are _always_ LF. CR/LF/CRLF is checkout property.	05:13
	→ spvensko joined	05:13
	← sdboyer left	05:14
	→ \|Rolenun\| joined	05:14
Ilari	jayne: Or that's the way it should be. And there should be option to convert on add but not on checkout.	05:14
jayne	Ilari: the objects are supposed to always be LF, but reality often disagrees (ime)	05:14
Ilari	jayne: Plus of course tool to fix line endings.	05:14
	← zahna left	05:14
Ilari	jayne: Yeah, CRLFs get into repo a lot.	05:15
	← Rolenun left	05:15
	\|Rolenun\| → Rolenun	05:15
	← jfkw left	05:15
	→ obscurite joined	05:16
obscurite	quick q - how do you see what would change upon doing "git pull" on remote branches without actually doing the pull?	05:16
Rolenun	sorry about that, irc client yelled yippee! and jumped offline :/	05:16
	→ spielberg joined	05:16
Ilari	obscurite: Fetch first.	05:16
obscurite	ilari - is that like a dry run?	05:17
Ilari	Rolenun: Or connection glitched or server glitched? :-)	05:17
jayne	pull == fetch + merge	05:17
Rolenun	pull = fetch + merge	05:17
	lol	05:17
obscurite	Ilari: jayne Rolenun okay, that sort of helps, i think. :)	05:17
spielberg	quick question, does git handle large binary files well? hg just screwed me on a repo that had some 200+ MiB files :[	05:17
Rolenun	Ilari: yep, it's one of those issues ;)	05:18
jayne	I'm worried about upstream stuff, so I almost never use pull. Instead I fetch, inspect, then merge as separate steps.	05:18
bob2	it won't corrupt them	05:18
	but then you'll have to push around 200mb files	05:18
obscurite	Ilari: let's say I fetch and it's a disaster, how do I go back?	05:18
jayne	spielberg: does anything handle large binaries well?	05:19
spielberg	well hg just shits out when im trying to push over network.. just hangs then nothing	05:19
Ilari	obscurite: No real need to go back.	05:19
obscurite	Ilari: what if I need to get rid of the changes to my local copy?	05:19
Ilari	obscurite: What is updated is either scratch branches (blow them away) or remote tracking branches (leave them as is).	05:20
	← pkw left	05:20
obscurite	Ilari: okay, guess I still don't get the flow. reading...	05:20
Ilari	obscurite: Or FETCH_HEAD (leave it alone, it will be overwritten eventially)...	05:20
	obscurite: Do you have remote for repository you want to fetch from?	05:20
	← parolang` left	05:21
	→ tedoc2000 joined	05:21
obscurite	Ilari: I made changes on my local repos (dev box), and now i want to send them to the master working copy on the server	05:21
	← spielberg left	05:21
	→ parolang` joined	05:21
obscurite	Ilari: so far I added, committed, and pushed on my dev box	05:21
bob2	don't push to non-bare repositories	05:21
obscurite	Ilari: both machines are pointing at the same repos. my dev box is remote.	05:22
	bob2: what should I be doing insted of pushing?	05:22
	bob2: just commit?	05:23
Ilari	obscurite: Push to bare (no working copy) repo.	05:23
	obscurite: Or pull both ways.	05:23
obscurite	Ilari: well too late, my newb ass pushed to non bare with a working copy. now what?	05:23
Ilari	obscurite: Does that repo have any uncommitted changes you want to keep?	05:23
bob2	create a bare repository and push to that	05:23
	then have the non-bare one from the bare one	05:24
	er "one pull from"	05:24
	← johnw left	05:24
	→ \|Rolenun\| joined	05:24
	← parolang` left	05:24
obscurite	Ilari: when i do git diff at the base dir of the working copy I get nothing	05:24
	Ilari: so I think that means no	05:24
	→ parolang` joined	05:24
Ilari	obscurite: On repo you pushed _to_?	05:24
obscurite	Ilari: yes	05:24
	→ jessep joined	05:24
Ilari	obscurite: Some hook that autoupdates it?	05:25
obscurite	Ilari: i think it's just up to date	05:25
	← parolang` left	05:25
obscurite	Ilari: except for my changes	05:25
	← sumangurung left	05:25
Ilari	obscurite: Does 'git log' list the commits (on that repo)?	05:26
obscurite	Ilari: yes	05:26
	Ilari: my last commit is not there	05:26
	bob2: my idea was that by "pushing" I was making my stuff available to the other repos	05:27
Ilari	obscurite: Well, the hallmark of push-to-HEAD screwups is that commits pushed appear in 'git log', but changes don't appear in working copy and 'git status' shows pushed files as 'modified'.	05:27
bob2	obscurite: pushing does not update the working copy	05:28
obscurite	Ilari: when I do git remote show origin I see my local repos is set to push to master on the other box, (master pushes to master)	05:29
	Ilari: but git log on the other box does not show the commit	05:30
	← Rolenun left	05:30
Ilari	obscurite: Is the commit to be pushed on master on local side?	05:31
	obscurite: (git log master)	05:31
	← j416 left	05:31
obscurite	Ilari: I made the commit and push locally. git log master locally shows my commit. git log master on the server does not.	05:32
	← Bass10 left	05:32
	→ pkw joined	05:32
obscurite	Ilari: does this mean i'm not looking at the master I actually pushed to somehow?	05:33
	Ilari: both have the same remote URL	05:33
jnl	is there something like a combination of rev-list and for-each-ref? the shell-code-feature of for-each-ref, but for any commit?	05:33
Ilari	obscurite: And push showed that it was successful? IIRC, current git master versions deny pushes to checked out branches by default.	05:33
	jnl: At least rev-list (and log) have custom formats.	05:34
	→ nadavoid_ joined	05:35
obscurite	Ilari: ahh, I pushed to a diff repos. the server is just another remote branch.	05:35
	Ilari: I see in the push output that it's an unfuddle repos actually	05:35
jnl	Ilari: ah, yes, it has --pretty, but quoting would break for example	05:36
obscurite	Ilari: now my main concern is to undo any damage i've done	05:36
	Ilari: with my stupid push	05:36
	← segher left	05:36
obscurite	Ilari: unless in this case what I did is OK	05:37
Ilari	obscurite: Well, firstly get shell on that repo to examine the damage.	05:37
obscurite	Ilari: it's a hosted git repos. no shell.	05:37
Ilari	obscurite: 'git diff', 'git status', 'git log'.	05:37
	obscurite: That repo has no working copy then.	05:37
	obscurite: So only way you could have screwed it is by pushing stuff that shouldn't be there.	05:38
obscurite	Ilari: right, the main repos is on unfuddle.com. the server one I was looking at is a working copy on another remote.	05:38
	Ilari: so is it OK to push to the main repos that has no working copy?	05:38
	Ilari: and the pull to the server remote repos	05:38
	← tedoc2000 left	05:39
Ilari	obscurite: Repo with no working copy is meant to be pushed to and fetched (pulled) from.	05:39
obscurite	Ilari: so maybe I didn't mess up. heh	05:39
Ilari	obscurite: Repos with working copy are meant for making new commits, not for pushing to.	05:39
obscurite	Ilari: so back to my original question. in a working repos, if I fetch, I'm not actually changing my working copy?	05:40
	→ erikvold joined	05:41
Ilari	obscurite: Correct.	05:41
	obscurite: In fact, fetch has extra safety check, so fetch-to-HEAD is blocked unless specifically allowed via command-line option.	05:42
	obscurite: (effects of fetch-to-HEAD would be just like push-to-HEAD).	05:42
obscurite	Ilari: I was looking for the equivalent of a "dry run"/revert in svn	05:42
Ilari	obscurite: If you have remote for the repo, just 'git fetch <remote>' and then examine '<remote>/<branch>'.	05:43
	obscurite: Then one can for example do 'git log -p ..<remote>/<branch>' or 'git diff ...<remote>/<branch>'.	05:43
	obscurite: First is per-commit breakdown of what 'git merge <remote>/<branch>' would merge. The latter is diff of all changes to be merged.	05:44
	obscurite: Just fetching changes but not merging/rebasing is the closest to dry run.	05:45
obscurite	Ilari: which URL is <remote> pointing at?	05:45
Ilari	obscurite: 'git remote -v' should list that.	05:45
	← catalis left	05:47
obscurite	Ilari: so on the server, I do "git fetch", which is set by default to pull from the main repos	05:47
	Ilari: then git log diff, etc? it's all master branch, so not sure what to use for args	05:47
	Ilari: seems like I shouldn't specific <remote> or <branch> in any case since there are defaults	05:48
	specify*	05:48
	→ catalis joined	05:48
Ilari	obscurite: There probably are no defaults unles remote in question is origin.	05:48
obscurite	Ilari: it is origin. git remote -v shows origin of the main repos on both dev box and server	05:49
	Ilari: and all branches are master	05:49
Ilari	obscurite: Then 'git fetch' for fetch should be enough (check that the URL it gives is the right one).	05:50
	obscurite: And 'git log -p ..<remote>' and 'git diff ...<remote>' operate with remote-specific default branch.	05:51
	obscurite: (which is master probably in this case).	05:51
	← nadavoid left	05:51
	nadavoid_ → nadavoid	05:51
Ilari	obscurite: Actaully, those commands operate with '<remote>/HEAD' which is symbolic ref to default branch.	05:51
	obscurite: Fetch is one of the very few commands that can even designate branch in another repository.	05:52
obscurite	Ilari: okay, git fetch seemed to get all the objects	05:52
	Ilari: plain git log -p doesn't show anything	05:52
Ilari	obscurite: 'git log ..origin'?	05:53
	obscurite: Displays commits that are in origin/HEAD (default for 'origin') but not in HEAD (current branch).	05:53
obscurite	Ilari: I see my commit there now	05:53
	Ilari: in git log ..origin	05:53
Ilari	obscurite: Add -p option to get diffs.	05:53
	obscurite: Or 'git diff ...origin' to get one diff for it.	05:54
obscurite	so fetch just pulls in the origin head info?	05:54
	(with my remote default of origin i mean)	05:54
Ilari	obscurite: Fetch copies branches from another repository. There are defaults to create the effect of mirrors of branches from other repositories.	05:55
	obscurite: 'git fetch foo' would mirror all local branches of remote 'foo'.	05:55
obscurite	ahh, so it just makes the remote branch available locally for viewing	05:56
Ilari	obscurite: Yes. And that mirror can be accessed off-line. And there is 'git remote update' that updates mirrors for all repositories.	05:56
	obscurite: *all remotes	05:57
obscurite	Ilari: I'm guessing you wouldn't do this for origin/HEAD but I guess you could use this to work on different remote branches?	05:57
Ilari	obscurite: Usually remotes are fetched whole, not individual branches.	05:57
obscurite	Ilari: huh okay	05:58
Ilari	obscurite: 'git fetch <remote> <branch>' is usually a mistake.	05:58
obscurite	Ilari: and so the reason this is so great for the linux kernel is I can work on a whole bunch of people's remotes and merge them easily?	05:59
Ilari	obscurite: Yes. If one knows how to back off bad merges, one could just use pull also.	05:59
	→ project2501a joined	05:59
obscurite	Ilari: thanks for all the git lessons, you rock!	06:00
	Ilari: so now that my git log ..origin looks great, I do a git merge to update locally?	06:00
	→ htns joined	06:00
	→ em-dash joined	06:01
htns	Hi. I used to know this but can't remember and find it either. After I've done a patch < horriblepatch , and now want to revert back to the previous state, I can do git checkout -f but how do I also remove all the newly added files? [reset --hard doesn't do it]	06:01
Ilari	obscurite: 'git merge origin'.	06:02
	← justin-george left	06:02
obscurite	htns: good question. i'm interested as well.	06:02
	→ justin-george joined	06:02
Ilari	htns: man git-clean ... But thats one of the most dangerous commands.	06:03
	← em-dash left	06:03
Gitbot	htns: the git-clean manpage can be found at http://git.or.cz/man/git-clean	06:03
htns	thanks Ilari , i will read now	06:03
Ilari	htns: IIRC, the only command that has confirmation switch.	06:03
obscurite	Ilari++	06:03
	← ilumi left	06:04
Ilari	htns: And hint: 'git apply --index <patchfille>'.	06:05
	htns: That even stages the changes for you.	06:05
htns	Ilari, nice, I will try that	06:05
Ilari	htns: (for future patch applications, of course).	06:06
	→ em-dash joined	06:09
	← madewokherd left	06:11
	← em-dash left	06:12
	→ em-dash joined	06:13
	← mmoya left	06:14
	← drogue left	06:14
	← spvensko left	06:15
	→ priidu joined	06:17
	→ tedoc2000 joined	06:19
	← nighthwk1 left	06:20
	← wolog left	06:21
	← obscurite left	06:22
	← jrmuizel left	06:24
	← Arafangion left	06:28
yhager	I would like to cherry-pick a few commits - can I see the diff they will create combined?	06:28
bob2	cherry-pick takes -n	06:30
	← RaceCondition left	06:30
	← pklingem2 left	06:32
	mdel → mdel-away	06:34
	← niki left	06:35
	← harryxu left	06:36
	→ lresende joined	06:36
	← justinweiss left	06:38
	← agenteo left	06:39
	← johan-s left	06:41
	← tedoc2000 left	06:42
	← sudoer left	06:42
	← ankit9 left	06:43
	→ bambule joined	06:49
	← thomaz left	06:49
	← krawek left	06:50
	cedric_ → cedricv	06:51
	← cedricv left	06:51
	→ cedricv joined	06:51
	← munichlinux left	06:52
	← AkumaStreak left	06:55
	→ Haarg joined	06:56
	→ loinclot_ joined	06:58
	← jessep left	07:00
	→ marc joined	07:03
	← numberGrey left	07:03
	→ schmukkk joined	07:04
	→ sudoer joined	07:05
	→ numberGrey joined	07:07
	→ jessep joined	07:08
	← project2501a left	07:08
	→ em-dash_ joined	07:09
	← em-dash left	07:09
	em-dash_ → em-dash	07:09
	← bambule left	07:10
	→ oganso-- joined	07:12
	← jlilly left	07:12
oganso--	anyone familiar with easygit?	07:13
	→ ntoll joined	07:13
sitaram	(speaking for myself)) not lately... whats the question?	07:13
	← loincloth left	07:14
	← Adlai left	07:14
	→ sis joined	07:15
oganso--	sitaram: heck i cant do eg push but i can git push	07:19
	eg just says that "Error: No recorded commits to show yet"	07:19
	although git push bla bla works fine	07:19
	→ bwwww joined	07:22
	→ ph^_ joined	07:25
sitaram	hmm no too deep for me, but eg has an "explain" or "verbose" or some such mode where it shows what actual git commands are running; try it	07:25
	(from vague memory)	07:25
oganso--	sitaram: oki :D	07:26
	→ bwwww_ joined	07:27
	← ideogon left	07:27
	← nadavoid left	07:29
	← sis left	07:29
	← __Joker left	07:29
	→ jumoit1 joined	07:30
	← mithro left	07:30
	→ jumoit2 joined	07:30
	← jumoit left	07:30
	← catofzen left	07:32
	→ munichlinux joined	07:33
	→ project2501a joined	07:34
	← dreiss left	07:35
	← laj left	07:36
	← ph^_ left	07:36
	→ Tordmor joined	07:38
	← jumoit2 left	07:39
	→ jumoit joined	07:39
	← onigiri_ left	07:41
	← priidu left	07:42
	← sudoer left	07:43
	→ thiago_home joined	07:43
	← bwwww left	07:45
	bwwww_ → bwwww	07:45
Mage_Dude	I did some stupid stuff and pushed a horrible commit to my git server. Now I get fatal errors which seemed to result in dangling blobs. I'm not quite sure how to get myself out if it.	07:46
	→ j416 joined	07:46
	← jumoit left	07:47
	← jumoit1 left	07:48
Ilari	Mage_Dude: Dangling blobs aren't errors... And what fatal errors?	07:48
Mage_Dude	Bad tree object	07:49
	→ oganso joined	07:49
Mage_Dude	git upload-pack aborts due to repo corruption on remote side.	07:49
oganso	backzz	07:49
Ilari	Mage_Dude: What git versions on local and remote side?	07:50
	Mage_Dude: (at least local side version)?	07:50
Mage_Dude	Weird. Looks like I have 1.6.3.3 locally and 1.6.0.4 on the remote	07:51
	← xlistener left	07:52
Ilari	Mage_Dude: Can you run 'git fsck --full \| egrep -v "^dangling"' on remote side?	07:52
Mage_Dude	Does it matter if I upgrade first? Or at all?	07:53
Ilari	Mage_Dude: Shouldn't matter.	07:53
	→ martins joined	07:54
	← pinnacle left	07:54
	→ rseifert joined	07:55
Mage_Dude	Hmmm I'm using gitosis to manage the remote side. I can't directly access the repository it seems	07:55
	← Chillance left	07:55
	→ laj joined	07:55
	→ bentob0x joined	07:56
	→ d2dchat joined	07:57
Mage_Dude	Ilari: Is there any way to do it without direct access?	07:57
	→ chalkbag joined	07:57
	→ dreiss joined	07:58
	← oganso-- left	07:58
	→ ixti joined	08:01
	← warthog9 left	08:06
	← em-dash left	08:11
	→ ph^_ joined	08:11
Ilari	Mage_Dude: You have any grafts (.git/info/grafts) on local side?	08:11
Mage_Dude	I don't have an info directory?	08:13
	← sitaram left	08:13
	→ sitaram joined	08:13
	→ pantsman joined	08:15
Ilari	Mage_Dude: If you don't have info there, then there can't be grafts file... Maybe ask those who admin the repo hosting system for help?	08:16
Mage_Dude	Well crap. Since that's me, I guess I should just dump the repo and be more careful about what I commit	08:17
	→ wolog joined	08:17
	→ em-dash joined	08:17
	← bwwww left	08:18
Ilari	Mage_Dude: Those who admin it should be able to get shell on it.	08:18
Mage_Dude	Well, I only have an exclude inside the info directory	08:20
Ilari	Mage_Dude: Git can push broken commits in presence of grafts. Also Jgit pushing is somehow borked and sometimes pushes something totally broken.	08:22
	← mm_202 left	08:22
	← JasonWoof left	08:22
Ilari	Mage_Dude: Bad tree object? Anything else about that object?	08:22
Mage_Dude	Let's see: could not read, bad tree, waitpid failed, git upload-pack aborting due to possible remote corruption, error in sideband demultiplexer	08:24
	← bry left	08:25
Ilari	Mage_Dude: What's the exact could not read message?	08:25
Mage_Dude	error: Could not read 41cdf7e45e62facdd616c838a969c04b2d91d8ce	08:26
	→ drizzd joined	08:26
	← d2dchat left	08:26
	→ d0ugal joined	08:26
Ilari	Mage_Dude: Try 'git cat-file -t 41cdf7e45e62facdd616c838a969c04b2d91d8ce' on local side.	08:27
	→ d2dchat joined	08:27
Mage_Dude	tree?	08:27
Ilari	Mage_Dude: It should be tree. Just to check if it exists.	08:28
	→ Adlai joined	08:32
	← xenoterracide left	08:35
	→ xenoterracide joined	08:35
	→ yhager_ joined	08:36
	marc → marc_kdab	08:40
	← JEEB left	08:44
	→ chrisr joined	08:44
	→ JEEB joined	08:44
	→ daaku joined	08:44
	← Twisted_Mentat left	08:45
	→ cilly joined	08:46
	→ warthog9 joined	08:46
	← jessep left	08:47
	← __bert__ left	08:47
	→ __bert__ joined	08:47
	← __bert__ left	08:47
	← dreiss left	08:48
	→ Axius joined	08:48
	← yhager left	08:48
	← Axius left	08:49
	→ JasonWoof joined	08:50
	← chalkbag left	08:50
	← tekky left	08:51
	← thiago_home left	08:55
	→ mmoya joined	08:55
	→ eletuchy joined	08:56
	→ ankit9 joined	08:56
	→ rhett joined	08:57
	← psankar left	08:58
	→ oganso- joined	08:58
	→ antonfh joined	08:58
	→ becomingGuru joined	08:58
	← squarism left	08:59
	← JEEB left	09:00
antonfh	Hello	09:00
	VVSiz_ → VVSiz	09:00
Ilari	...	09:03
	← r11t left	09:04
	← numberGrey left	09:06
	m-takagi → m-takagi_	09:08
	← oganso left	09:11
	→ shruggar joined	09:12
	→ Twisted_Mentat joined	09:12
	→ nadim joined	09:13
	→ anno^da joined	09:13
antonfh	any discussion on Git today	09:14
	→ psankar joined	09:14
	→ yhager joined	09:15
	← laj left	09:17
	→ OtherJakeSays joined	09:18
wereHamster	hm.. this is #git, what else would we be talking about here?	09:18
	→ gitte joined	09:18
	← lemonchicken left	09:19
	→ lemonchicken joined	09:19
	← pantsman left	09:20
	→ Scorchin joined	09:21
	→ tekky joined	09:22
	→ loincloth joined	09:23
	→ lucsky joined	09:23
	← JakeSays left	09:25
Ilari	Crazy idea: Configuration option which would cause upload-pack/receive-pack to redirect requests...	09:26
	→ tvw joined	09:26
	→ hummingkid joined	09:28
	← mmoya left	09:28
Ilari	And that would work with just valid configuration file without even having valid repository there.	09:29
	← yhager_ left	09:30
	→ aleksei joined	09:31
aleksei	If i created a branch off master, which tracks an svn trunk.. and now i need to merge that branch into master to later dcommit to svn. What's the safest way to do it?	09:32
	→ loinclo__ joined	09:32
	← loincloth left	09:32
Ilari	aleksei: Merges and SVN don't mix.	09:33
wereHamster	aleksei: git checkout master; git merge whatever-branch; (optionally, but recommended: git svn rebase); git svn dcommit	09:33
	← becomingGuru left	09:33
	→ becomingGuru joined	09:33
	← simNIX left	09:34
aleksei	actually master hasn't diverged at all from the branch.. so i suppose that won't cause many problems in svn	09:34
	→ Bryson joined	09:34
	← Bryson left	09:34
	→ priidu joined	09:36
aleksei	thanks wereHamster	09:37
	← loinclot_ left	09:39
antonfh	have simple Question ... is Git merely to log changes and change History .... when I have a remote master and my local and I commit changes and push it to the remote ... does it update the files on the remote ? in other word keep syncronised versions of my project locally and remote	09:39
Ilari	antonfh: Push does not update files in working copy.	09:40
	antonfh: In fact, push (and fetch) completely ignores the working copy on both sides.	09:40
thiago	it doesn't update files	09:40
antonfh	ok so its not like I can work on my laptop ... track changes and do push and think my remote server will now have new files	09:40
tetha	how should that work, if your server's working copy might have changed?	09:41
	← ToxicFrog left	09:41
	← justin-george left	09:42
antonfh	I thought it does something like rsync and updates copies .... but I just set up my local as a clone ... maybe that is wrong ?	09:42
	→ squarism joined	09:42
antonfh	I thought the push and fetch would update the remote (with push) and the fetch would update the local with what ever files changed	09:43
	← aleksei left	09:43
Ilari	antonfh: $faq non-bare	09:43
Gitbot	antonfh: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare	09:43
tetha	I for myself usually put up 3 repos in such a situation	09:43
	← tvw left	09:43
j416	antonfh: read about the difference between workdir and repository	09:43
antonfh	ok thanks	09:44
j416	your workdir is in a sense just a temporary place to store files you are currently working on	09:44
	→ oganso joined	09:44
j416	when you want to "save" what you have done, you commit it to the repo.	09:45
	then, the files in your workdir are useless, you can always check them out from the repo at any time	09:45
	(except for any files you have ignored/not added)	09:45
	s/useless/redundant/	09:46
	:)	09:46
antonfh	will have to get into that ... I've only started using Git on Monday ... so its not all transparent yet	09:46
Ilari	antonfh: AFAIK, the versions in current git master would refuse push to checked out branch by default.	09:47
	→ bauruine joined	09:47
	← em-dash left	09:47
	← countd left	09:48
	← hobodave left	09:48
j416	antonfh: the reason the workdir is not automatically updated on push, is that there may be changes - and there is no way to handle possible conflicts without human interaction. There are no humans on the remote side, just git. So it can't be sanely done.	09:49
	(say the workdir files on the remote for some reason have changed - you couldn't just assume that those changes are irrelevant and delete them	09:50
antonfh	going to read up on the non-bare Wiki that was mentioned also ... ok so its not like you just simply have a update and push going to copy your local files to remote	09:50
	→ countd joined	09:50
j416	(although you can achieve this, using hooks, but there are other, better ways)	09:50
	→ globalkeith joined	09:50
j416	antonfh: your repo is pushed, not your workdir	09:51
	s/repo/repo branch(es)/ even	09:51
antonfh	... and out of curiosity what would those better ways be ? .... yes I think my mindset was it pushes files	09:51
j416	rsync for example	09:51
antonfh	lol ... ok so my hack was not that bad ... cause that is exactly what I did yesterday	09:52
j416	have a bare repo someplace	09:52
	then get a fresh copy, and rsync that to where you are going to use it	09:52
	(for example web server, which is what I'm imagining in this use-case)	09:52
antonfh	setup a rsync ... but then thought maybe Im doing this all wrong and completely miss the Git function	09:52
	→ intripoon joined	09:52
j416	if you are the only dev, it's easy to just rsync your repo to the server (for backup, or web serving)	09:53
	(if web serving, you might want to ignore the .git dir in one way or another)	09:53
antonfh	yes my case now its mostly backup for now ... and getting confused with what I have updated and when and where	09:53
j416	using git push/fetch is more useful in collaboration	09:54
	not so useful for simple backup	09:54
	because you will lose things like your reflog	09:54
	which can be good to have if you need to restore	09:54
antonfh	so Git is really helping me keep record of changes ... so the push as I have it will merely update my repo and have logs local and remote for me (but not files)	09:54
	← psankar left	09:54
j416	not logs	09:55
	history of changes	09:55
	from which you can generate logs, of course	09:55
antonfh	yes ... just wrong use of term ... meant history	09:55
j416	the data (files etc.) will of course also be pushed	09:55
	that's the whole point.	09:56
	← oganso- left	09:56
antonfh	can 2 people create a repo in a folder ... like I created one in say /www/dev ...git init ... so any changes staged and commited is marked as my username/name ... what if someone else does work in same folder and they update changes ... how will it log it as "they" did it and not me when I go update and commit it (or "they" do)	09:57
	← Tricon left	09:59
antonfh	or will they basically create their own custom repo in that folder ... where the default is simply .git ... you need to create one for yourself like .antonfh and .otherdev etc	10:00
	← JasonWoof left	10:01
	→ gzy joined	10:02
	← ankit9 left	10:03
	← tetha left	10:04
	→ v0n joined	10:04
	→ tetha joined	10:05
	→ nitrate joined	10:12
	→ Sonderblade joined	10:13
	→ tbf joined	10:16
	→ Sky[x] joined	10:16
	← Twisted_Mentat left	10:19
	← cota left	10:19
	→ laj joined	10:19
	← globalkeith left	10:20
	→ Paraselene_ joined	10:20
	← Guest19622 left	10:21
	→ ToxicFrog joined	10:21
	→ NET\|\|abuse joined	10:23
	← eletuchy left	10:23
divVerent	how can I find the fork-point between two branches like git-cherry does?	10:24
teuf	divVerent: git merge-base maybe	10:25
	← Scorchin left	10:25
	→ Scorchin joined	10:26
divVerent	looks good, in my case gives the exact output I want...	10:26
	but when i merged from one branch into another multiple times	10:27
	will it find the last merge?	10:27
charon	note that the merge-base may not be unique (git merge-base --all ...)	10:27
divVerent	I need the newest one :P	10:27
	but well	10:27
	CAN be it non-unique?	10:27
	oh... and can I undo a merge by simply git-resetting to a commit before it?	10:28
	→ vbgunz joined	10:29
charon	divVerent: http://pastebin.com/m4e6c7618 what is the merge-base of M1 and M2? (X is a crossing, not a commit)	10:29
	→ name joined	10:29
divVerent	ah, okay	10:30
charon	as for the "newest" one: yes. read the definition in the manpage	10:30
	→ curtana joined	10:30
divVerent	that absolutely can't happen in my use case	10:30
	"git-merge-base" finds as good a common ancestor as possible between the two commits. That is, given two commits A and B git-merge-base A B will output a commit which is reachable from both A and	10:31
	B through the parent relationship.	10:31
	there's no better definition here...	10:31
charon	man git-merge-base	10:31
divVerent	but well, "as good as possible" probably describes what I mean	10:31
Gitbot	charon: the git-merge-base manpage can be found at http://git.or.cz/man/git-merge-base	10:31
charon	"One common ancestor is better than another common ancestor if the latter is an ancestor of the former."	10:31
	"A common ancestor that does not have any better common ancestor is a best common ancestor, i.e. a merge base."	10:31
	there's your definition	10:31
divVerent	actually... to be more specific	10:32
	I want the merge-base that is in a given branch	10:32
	if there was a crossing	10:32
charon	umm	10:33
divVerent	just, is it recorded at all that, in the lower picture, 2 was once HEAD of the branch B?	10:33
charon	in my ascii art, both root commits (the "o") are "in" (i.e., ancestors of) M1 and M2	10:33
	no, a commit in git does not have an intrinsic branch membership	10:33
divVerent	okay	10:33
	→ killerchicken_ joined	10:33
divVerent	then I will just use git-merge-base as is	10:33
	and assume that this case does not happen :P	10:33
charon	notably, it may be part of any number of branches at any given time, including none at all, and this number may (and usually will) change over the life of the commit	10:34
divVerent	I doubt anything will go wrong in my script anyway if using a "wrong" merge base	10:34
	as it's still cosistent	10:34
	there isn't a simple argument to git-log to specify the merge base?	10:35
	like, a name like HEAD?	10:35
	← killerchicken left	10:35
thiago	git name-rev	10:35
	→ mbroeker joined	10:36
charon	divVerent: only in the A...B syntax	10:36
	which means A B --not $(git merge-base --all A B)	10:37
	← curtana left	10:37
	← hummingkid left	10:37
	→ bizhat joined	10:38
	← mbroeker left	10:39
divVerent	if two commits have no merge base, I assume I get an error	10:40
	(probably can only happen if something got messed up)	10:41
	→ esc joined	10:41
charon	actually, my copy of git.git has a dozen root commits	10:41
	← bizhat left	10:41
charon	you'll get a nonzero exit status and no output	10:42
wereHamster	antonfh: it's better if the two users log in using different accounts, and set up their username in ~/.gitconfig	10:42
esc	hi, i would like to store the SHA of the HEAD in an environment variable	10:42
	sofar i came up with:	10:42
	SHA=$( git --no-pager log --pretty=format:%H HEAD^..HEAD )	10:42
	any better way?	10:42
wereHamster	esc: git rev-parse HEAD	10:42
	→ bizhat joined	10:42
	← Kelly_zhou left	10:42
esc	wereHamster: just what i was looking for :-)	10:42
	→ hummingkid joined	10:44
	← bizhat left	10:44
	→ oganso-- joined	10:45
divVerent	yeah, this works now... except that I just noticed I want to be able to revert even commits from before the merge base, so my script has to parse more anyway	10:46
	→ nick_ joined	10:46
	nick_ → Guest24316	10:46
charon	divVerent: just for the record, i question your motives. why would you have to mass revert commits all the time?	10:47
Guest24316	hello	10:47
	i jsut tried to run "git revert HEAD "	10:47
	and got "fatal: Dirty index: cannot revert "	10:47
divVerent	no, not mass	10:47
	managing a stable tree	10:47
Guest24316	Can anyone help me how to fix this?	10:47
divVerent	and I may want to undo a decisionm to accept a previous commit	10:48
	EVEN if that commit was from before I started making the stable tree	10:48
	→ balint joined	10:48
divVerent	but well	10:48
	I don't think I will ever NEED that	10:48
	just want to be able to track it correctly in case I need to	10:48
wereHamster	Guest24316: you probably have uncommited changes in your index	10:49
	Guest24316: also, are you sure that 'git revert' is what you want?	10:49
Guest24316	yes	10:49
	← squarism left	10:49
	→ squarism joined	10:49
Guest24316	as i have to many conflicts and no real work worth keeping	10:49
wereHamster	Guest24316: so you have already pushed HEAD to a public repo, correct?	10:50
Guest24316	i just want to revert my previous commit	10:50
	→ eletuchy joined	10:50
Guest24316	and found i need to run git revert HEAD	10:50
	is there another way?	10:50
wereHamster	if you haven't pushed HEAD yet there is no reason to use git revert	10:50
	you can just reset HEAD~1 and be done	10:50
Guest24316	what does that do?	10:50
wereHamster	(or reset --hard, depending on whether you want to destroy changes in your working tree as well or not)	10:51
Guest24316	so run	10:51
	"reset --hard"	10:51
	and it does what?	10:51
wereHamster	Guest24316: man git-reset	10:51
Gitbot	Guest24316: the git-reset manpage can be found at http://git.or.cz/man/git-reset	10:51
wereHamster	Guest24316: also read the synopsis of man git-revert !	10:53
Gitbot	Guest24316: the git-revert manpage can be found at http://git.or.cz/man/git-revert	10:53
	→ d0k joined	10:55
	→ loincloth joined	10:58
	← oganso left	10:58
	← ph^_ left	10:59
	→ nonken joined	11:00
	→ ph^_ joined	11:01
esc	is there an easy way to see if a file has changed between revisions?	11:06
	i am using:	11:06
	TRUTH=$( git --no-pager log --pretty=format:%H --diff-filter=M HEAD^..HEAD -- file)	11:06
	→ jelmer joined	11:06
	→ gma joined	11:08
	← gma left	11:08
	→ gma joined	11:09
	← Theravadan left	11:11
	← loinclo__ left	11:14
	← jinzhu left	11:15
	← Guest24316 left	11:15
	→ desmond joined	11:20
desmond	Can someone help me. I have a file that wants to be checked in but i have /logs/* listed in my gitignore how can i remove it from list of files that wants to be checked in	11:21
	modified: log/frontend_dev.log	11:21
	any help?	11:21
	← stepnem left	11:22
pasky	hmm	11:23
	UPC just terminated the agreement to host repo.or.cz	11:23
	→ stepnem joined	11:24
	← desmond left	11:25
	← chrisr left	11:29
	emmanueloga → emmanuel_oga	11:31
	emmanuel_oga → emmanuel__oga	11:32
wereHamster	deavid: tracked files are not aftected by gitignore	11:32
deavid	?	11:33
	→ struberg joined	11:33
wereHamster	er, sorry	11:33
	← eletuchy left	11:33
wereHamster	desmond already left -.-	11:33
	pasky: why? Did they state a reason?	11:33
	emmanuel__oga → emmanuel___oga	11:34
	emmanuel___oga → emmanueloga	11:34
pasky	wereHamster: they probably want to clean up the server room and the server there originally had a different purpose	11:38
	→ Pupeno joined	11:38
	→ CVertex joined	11:42
	← CVertex left	11:42
	← Sky[x] left	11:44
	→ psoo joined	11:44
	→ Sky[x] joined	11:46
	← loincloth left	11:48
	dwmw2_gone → dwmw2	11:48
	→ loincloth joined	11:48
	→ bizhat joined	11:49
	← bizhat left	11:50
	nonken → nonken_afk	11:51
	nonken_afk → nonken	11:51
	→ jord joined	11:51
	→ smuf joined	11:52
	← jord left	11:52
	→ chrisr joined	11:54
	← ndim left	11:55
	← Sky[x] left	11:57
divVerent	just doing mad stuff... checking out a 2GB repository through git-cvsserver :P	11:57
	seems insanely fast, both compared to cvs and svn	11:57
	← ph^_ left	11:58
divVerent	4.5MB/sec	11:58
	but let me guess - it also has the same DISadvantages as cvs, like, if someone pushes during the checkout, the user gets an inconsistent copy	11:58
	→ capndiesel joined	11:59
_Vi	divVerent, The manual says it don't support all "commands and options".	11:59
	← David-T left	11:59
divVerent	sure	11:59
	I am just trying it "because I can"	11:59
	→ ph^_ joined	11:59
divVerent	and if I am going to use it at all, it'll be for read-only access	11:59
	actually... I do see one good use case for it	11:59
	for a read-only repository e.g. on a game server	12:00
_Vi	divVerent, Try to push something while you "checking out with 4.5 mbps".	12:00
divVerent	as cvs does not store history or (like svn) copies locally	12:00
	it would take much less HDD space	12:00
	→ Ryback_ joined	12:00
divVerent	makes qutie a difference if the game data takes up 2GB or 4GB on your dedicated server :P	12:00
_Vi	divVerent, Binary files in the repository?	12:00
divVerent	yes	12:00
	all files of the game are in there	12:01
_Vi	divVerent, Anti-pattern.	12:01
divVerent	why?	12:01
	they need versioning too	12:01
	like, "when did this file change" and stuff	12:01
	what's the alternative? :P	12:01
	(talking about model files, textures, NOT program binaries)	12:01
_Vi	divVerent, Better source text files that generates your binary (unless it is something like "photo taken by camera / downloaded from internet").	12:02
divVerent	_Vi: sure, if we can, we do that	12:02
	but that's not possible for e.g. music	12:02
	or textures made e.g. in photoshop or gimp	12:02
	← squarism left	12:02
_Vi	divVerent, At least may be the code and such big blobs should not be in the same repository.	12:02
divVerent	but the code depends on the big blobs	12:02
	like, often code changes and media changes go along with each other	12:03
	and hell would freeze over if changed independently	12:03
	like, if you change the name of a tag on a weapon model	12:03
	or if a new weapon is added, the data files for it need to be immediately there too	12:03
	← mdel-away left	12:03
_Vi	divVerent, I think games should be fault-tolerant against deleting it's data files. E.g. if no texture then render the level without this texture. No movie => skip this movie.	12:04
divVerent	of course this would happen	12:04
	with hell freezing over I mean behaviour that would be reported as bug by users/artists	12:04
	like, muzzle flashed attached at the wrong position of the gun (due to tag name change)	12:05
	it wouldn't lead to crashes	12:05
	just to wrong behaviour	12:05
_Vi	divVerent, So any can make minor code changes (supposing it's FOSS project) without downloading big repository and using data files he already has.	12:05
divVerent	it is FOSS	12:05
	but well	12:05
	the code is really not compatible with the data of the previous release	12:06
	or rather, if you do that, you get all sorts of missing texture replacement images and such stuff	12:06
	and colorful diamonds for missing models	12:06
_Vi	divVerent, Data packs should be versioned. Of course you can't use very outdated data pack.	12:06
divVerent	well	12:06
	we release about twice a year	12:06
	but there is LOTS of development and changes between releases	12:07
_Vi	divVerent, Early stage of development?	12:07
divVerent	no	12:07
	call it active development	12:07
	many people with many ideas	12:07
cbreak	git does not die when you feed it binary	12:07
divVerent	suree	12:07
	it works fine for me	12:07
	and even now, the git clone'd repository is smaller than the svn checkout'd one	12:08
	→ Zephyrus joined	12:08
_Vi	divVerent, You also may use "git submodules".	12:08
divVerent	now that could be more useful	12:08
_Vi	divVerent, 1. Code checkout will know what data pack to use (it will have pointer to commit-id).	12:08
	divVerent, 2. You may automatically checkout and update the code and data packs (with all histories for both).	12:08
	→ mithro joined	12:08
divVerent	that may be useful, but the directory structure of the game is really not fit for that	12:09
	e.g. art files are in multiple subdirs	12:09
_Vi	divVerent, 3. With some little quirks, you may keep code with all history, but data packs without history and updated no so frequent.	12:09
divVerent	like textures, models, sounds	12:09
	← gma left	12:09
divVerent	and code is in one subdir called "qcsrc", but depends on .cfg files in the main directory	12:09
cbreak	symlinks?	12:10
divVerent	I like being able to start the game directlry from a checkout, without having to e.g. create a pack	12:10
	windows users	12:10
	← upul left	12:10
_Vi	divVerent, Main directory is git repository, "sounds", "models" are subrepositories.	12:10
divVerent	sure	12:10
	I just mean	12:10
	we'd need one subrepository for each type of content	12:10
	that would work then	12:10
	but probably get out of hand	12:10
_Vi	divVerent, >"windows users" I wanted to offer using of symlinks...	12:10
divVerent	_Vi: sure, but symlinks do not work on windows	12:11
cbreak	windows does have links	12:11
divVerent	of directories, not of files	12:11
cbreak	I never got them to work though	12:11
_Vi	divVerent, You also may reorganize and create something like "data" and "src" directories. Big blobs in "data", textual and little things in "src".	12:11
divVerent	_Vi: that would be really huge changes to the project	12:11
_Vi	cbreak, NTFS symlinks are only for directories. And you can't copy it well.	12:11
divVerent	if we want it to still be able to run directly from a checkout	12:11
	→ jollyroger_ joined	12:12
divVerent	sure, one could restructure the repository like that, and provide a simple .sh file that creates symlinks to merge the trees correctly	12:12
	just, that won't help on windows	12:12
_Vi	divVerent, It will be "git clone .....; cd ...; git submodule update --init; make&&run".	12:12
cbreak	you would have to change your loader method to consider data as root and not .	12:12
divVerent	_Vi: I just don't want "make" to copy the cfg files	12:12
	wqe have users who'd change the copies, and not the originals, then :P	12:12
cbreak	git can manage symlinks I think	12:13
divVerent	even on windows?	12:13
cbreak	as long as you're on an OS that has them	12:13
_Vi	divVerent, Cygwin/Git or msysgit?	12:13
divVerent	well, the quakec source can even easily be moved out... the game reads no file of that	12:13
cbreak	(-> any OS besides windows I guess :)	12:13
divVerent	making "data" the root would be fine	12:13
	_Vi: and how to make the GAME recognize the symlinks? :P	12:13
_Vi	divVerent, Cygwin/git creates cygwin symlinks that work with cygwin.	12:13
divVerent	but the game does not work with cygwinb	12:13
	→ stamina_ joined	12:14
divVerent	it's compiled with mingw	12:14
	and it's absolutely not desired to make it depend on cygwin1.dll	12:14
_Vi	divVerent, (in the game) 1. Try to use a file. 2. If fails, read the singe line in file, open the file with name we just read.	12:14
	← tekky left	12:15
divVerent	there comes the security issue :P	12:15
	→ tek joined	12:15
divVerent	the game FS code refuses to access anything outside the data root	12:15
_Vi	divVerent, Also, cygwin's symlinks are actually Windows shortcuts.	12:15
divVerent	now THAT is good to know	12:15
	the game FS code allows symlinks in zip files actually	12:15
	but these then only reqwrite the path relative to the root	12:15
	→ ehabkost joined	12:15
divVerent	and don'T allow stuff like .. in them	12:15
	so the "data" directory can't be left by that	12:15
	well, one workaround would be putting the configs and other small text files to the QC source	12:16
_Vi	divVerent, May be configuration and media files should kept separate?	12:16
divVerent	and make the "make" command create a pk3 in the data directory with them	12:16
	nobody would open that pk3 and edit the configs there	12:16
_Vi	divVerent, E.g. configuration it user directory, media in system directory.	12:16
cbreak	nobody? :)	12:16
divVerent	there's two types of configuration	12:16
	user configuration already is in the user dir	12:16
	but there are also cfg files to CREATE the configurable variables for the user config	12:17
	and that give them defaults and descriptions	12:17
_Vi	divVerent, In Quake3 there's packs with both media and configs. User can repack them if it wants.	12:17
divVerent	sure, user can	12:17
	still, some of the config values are media dependent	12:17
	most are code dependent though	12:17
	→ mmw joined	12:17
_Vi	divVerent, By using packed files you can not bother with filesystem issues.	12:17
divVerent	sure	12:18
_Vi	divVerent, Default configs should be kept with code, not with media.	12:18
divVerent	but I don't want to repack everything just to try out one little change	12:18
	creating that pk3 takes quite a while	12:18
	_Vi: not possible in the quake engine	12:18
	configuration variables can only be created with config commands like "set" and "seta"	12:18
_Vi	divVerent, Create two packs: first for media, second for configs. Copy configs pack into proper place.	12:18
divVerent	but not with code	12:18
	or rather, code CAN create them, but only once executed	12:19
	as long as the menu is shown, the server code is not run yet	12:19
	and on a dedicated server, the menu code never runs	12:19
	_Vi: sure, that would work	12:19
	just see no advantage to keeping it all in one repository	12:19
	separately checked out code is useless	12:19
_Vi	divVerent, BTW "recreate configs if deleted" may be good approach. It will allow users to fix broken configs easily.	12:20
divVerent	as it needs to match media exactly	12:20
	or funny bugs happen	12:20
	actually	12:20
	exactly is saying too much	12:20
jollyroger_	hey guys, i have a problem with git and my ssh-keys: i have two git-hub accounts: one for me and one at the company i am working for.	12:20
	the problem: on a "git push origin master" i get: ERROR: Permission to my_company/repo denied to my_private_username	12:20
divVerent	but we often enough ALREADY get bug reports from people doing "svn update" only in the code directory	12:20
	← nitrate left	12:21
jollyroger_	apparently git is mixing it all up because it is trying to connect to the "my_company"-github account with "my_private_username"	12:21
_Vi	divVerent, >"or funny bugs happen" Bugs of discrepancy between code and media should be easy and noticable. Like message "I can't find textures/qqq.texture".	12:21
divVerent	_Vi: even then, first startup would miss some variables until the first game is started	12:21
	_Vi: we have these messages	12:21
jollyroger_	but I just dont see why, here is my (short) ssh-config: http://pastie.org/768683	12:21
divVerent	they are red and yellow	12:21
	nobody notices them	12:21
	mostly because crappy mappers of custom maps, or people who just take a quake3 map pk3 that references quake3 media and publish that as map for our game, cause lots of these messages to ALREADY appear on most servers	12:22
_Vi	divVerent, If the game don't start, user looks at the message. Quake3, for example, repeats the main error message in blinking style at the bottom.	12:23
divVerent	if it does nto start at all, sure noticed	12:23
	← niki_work left	12:23
divVerent	but if it causes more subtle errors, like walls appearing in white instead of the right texture, or models missing and appearing as colorful diamond	12:23
_Vi	divVerent, And users that use repository checkout should expect that the game may just not run and be prepared to problems.	12:23
divVerent	_Vi: sure	12:23
	we just ALREADY have a high "user error" rate in bug reports	12:23
_Vi	divVerent, Change the default texture from "plain white" to "tiled text with colourful error message"	12:24
divVerent	so I'd rather like to ensure 100% that the whole game stuff is consistent	12:24
	even if it often doesn't need to be	12:24
	but still, submodules may work for that	12:24
	→ segher joined	12:25
	← SockPants left	12:25
_Vi	divVerent, Just require users to run additional command (like "git submodule update").	12:25
divVerent	and when committing/pushing	12:25
	what if I e.g. add a texture, and make code use it at the same time	12:26
_Vi	divVerent, While compilation you may also add check if the submodule is checked out correctly and up to date and show big fat warning if it isn't.	12:26
divVerent	and want to prevent that anyone gets the new code without the texture	12:26
	"checkout the commit specified in the index of the containing repository"	12:26
	basically, my question is	12:26
	how to set that commit	12:26
_Vi	divVerent, s/prevent/provide good warning/	12:26
	divVerent, Do you know how to work with Git submodules?	12:27
divVerent	not yet	12:27
	just have the manpage	12:27
_Vi	divVerent, Main repository stores: 1. Commit SHA1 of submodule, 2. URL to know where to download it.	12:27
	nonken → nonken_afk	12:28
divVerent	right	12:28
	git submodule add creates 2	12:28
	how to edit 1?	12:28
_Vi	divVerent, When you add some new model to the data pack, you 1. Add it to submodule, 2. Commit it, 3. Update commit-id in the main repo.	12:28
divVerent	right	12:28
_Vi	divVerent, Something like "git add path/to/submodule"	12:28
	divVerent, It will substitute one SHA-1 for another.	12:28
divVerent	ah, okay	12:29
_Vi	divVerent, So you may keep the code and data in sync. (not not forced to to that).	12:29
divVerent	so git add submodulepath updates the commit hash to the one currently checked out in the submodule directory	12:29
	→ niki_work joined	12:29
	→ nitrate joined	12:30
_Vi	divVerent, Also (as users may easily skip submodule update) you should change makefile to check if everything's OK and show bug warning with instructions how to update the submodule.	12:30
divVerent	sure	12:30
_Vi	divVerent, (Yes, you may use "git" commands in makefiles)	12:30
divVerent	of course	12:30
	can e.g. make a "make update" command for our svn users	12:30
cbreak	submodules seem to create lots of detached heads for me	12:30
divVerent	that does git pull, git submodule update	12:30
	→ tstclair joined	12:30
cbreak	so I am not sure how well that works with people who want to commit things	12:30
	if you commit something to a detached head, you'll lose it sooner or later	12:31
	← niki_work left	12:31
	→ bjornbjorn joined	12:32
	→ __bert__ joined	12:32
_Vi	divVerent, Yes, committers need to be careful. They need more commands to execute.	12:32
	← martins left	12:32
_Vi	(going offline)	12:32
divVerent	not sure if I want that though	12:32
	but well	12:32
	the only files that are in the svn as "regeneratable binaries", the usual antipattern, are the bsp files of the maps	12:32
	I hate doing that	12:32
	but, recompiling them takes a day per map	12:32
	→ Geralt joined	12:33
bjornbjorn	I'm trying to fetch a sub folder from a git repo .. tried doing a clone but it tells me it's not a git archive	12:33
	← _Vi left	12:33
	← balint left	12:33
broonie	bjornbjorn: You can only fetch the full repository.	12:33
	nonken_afk → nonken	12:35
divVerent	hehe, exactly the problem why submodules would be better if possible	12:35
bjornbjorn	broonie: ok, so here's the issue .. I have a CMS which has a folder called /third_party/ which contains addons. I want to fetch from the git repo into that directory, but for it to work I need to retrieve a subfolder .. also, developing directly there is handy since I don't need to run a script everytime I want to test my addon .. I'm used to SVN where I'd just checkout the subfolder and bob would be my uncle you see	12:35
broonie	You probably want to use submodules here.	12:35
	→ kmartiin joined	12:36
bjornbjorn	checking	12:36
killerchicken_	bjornbjorn: what you want is narrow clone support, but that's not there yet	12:36
divVerent	LOL, just had to think of a funny way to abuse svn externals... as TAGS :P	12:36
	of course, git doesn't need that, as it has real tags	12:37
	→ niki joined	12:37
bjornbjorn	k, thanks for the input guys	12:37
	→ RaceCondition joined	12:39
	→ schierbeck joined	12:40
	← RobertLaptop left	12:40
	→ Axius joined	12:40
	→ bibinou joined	12:40
	← Axius left	12:41
	→ ndim joined	12:41
	→ djszapi_ joined	12:42
	← djszapi_ left	12:42
	→ kamen joined	12:42
kamen	How can I rewrite the whole git tree in that regard to delete the trailing whitespaces from the lines of files ?	12:43
	→ wallunit joined	12:43
kamen	I assume I should use git filter-branch for it, but how ?	12:43
wallunit	I am writing a merge driver for git and have a question. Is it possible to call the builtin merge driver "text" if my custom merge driver is not able to handle the merge?	12:44
	← mmw left	12:44
	← tek left	12:44
divVerent	possibly by calling git mergetool program again?	12:44
	→ tekky joined	12:44
divVerent	oh wait, sorry	12:45
kamen	It causes for me problem because if I clone a git tree and I checkout an earlier version I see some files that contain modifications.	12:45
wallunit	not merge tool. I am talking about merge drivers	12:45
	→ [Po]lentino joined	12:45
	→ artagnon joined	12:45
divVerent	yes, just noticed these are different	12:45
kamen	and When I look at for the modifications I see trailing whitespaces those make the differences.	12:45
	so I don't know exactly how to solve this situation.	12:46
	← kmartiin left	12:46
vmiklos	kamen: use --tree-filter	12:46
cbreak	write a script that removes them from the whole working directory	12:46
	and put that script into a tree-filter	12:46
vmiklos	yep :)	12:46
artagnon	Does Git ever plan to become platform-independent? I just glanced at the source: there are a bunch of shell scripts.	12:46
	← NET\|\|abuse left	12:47
kamen	hm, weird.	12:47
vmiklos	shell scripts can be portable, so that's not an issue.	12:47
divVerent	which platform doesn't support them?	12:47
kamen	Okay, I will follow your suggestions, but it's interesting definitely.	12:47
	so How can it be if I do a checkout for earlier version some modifications occur in the working tree.	12:47
artagnon	vmiklos: Oh, I see. What about references to things like /dev/null?	12:47
kamen	I wait that operation I can change among any versions.	12:48
	→ NET\|\|abuse joined	12:48
artagnon	divVerent: Any non-POSIX platforms?	12:48
divVerent	artagnon: sure, which?	12:48
	← Paraselene_ left	12:48
divVerent	there are sh.exe for windows that support /dev/null internally, I think	12:48
kamen	but If some modifications occur for it, I can't checkout other version because of the dirty git tree.	12:48
artagnon	divVerent: I see.	12:49
divVerent	but even if not - then one should simply make that part in git platform independent	12:49
	← marc_kdab left	12:49
divVerent	by using a OS distinction to use the right null device	12:49
artagnon	divVerent: What's blocking a native Windows implementation then? Unless I'm mistaken, it needs a POSIX-like emulator on Windows.	12:50
kamen	is it an expected behaviour of it or do I dod something wrong ?	12:50
divVerent	basically, git expects a posix environment, but this should work with msys just fine	12:50
	without all the cygwin layer stuff	12:50
	→ BigMack83 joined	12:50
divVerent	artagnon: there is none?	12:50
	maybe the networking :P	12:50
	artagnon looks	12:50
divVerent	that differs between windows and unix	12:50
	of course	12:50
	→ parasti joined	12:50
divVerent	the other reason could be, "cygwin1.dll works, and you can simply take all the other required programs from cygwin"	12:51
	← tehbaut left	12:51
artagnon	Hm. Yes.	12:51
divVerent	http://en.kioskea.net/faq/sujet-499-git-for-windows	12:51
	→ tehbaut joined	12:51
	→ malumalu joined	12:51
divVerent	actually, this is a windows native build :P	12:51
charon	artagnon: if you really want to help speed up "more native" implementations, google for libgit2 and contribute code there	12:52
	or just help with msysgit, they also chronically lack developers	12:52
artagnon	charon: Ah, thanks for the pointer!	12:52
divVerent	msysgit should be "almost like native" in the operations that already use an executable	12:52
	and probably still slow in the shell script stuff	12:53
artagnon	I'm not particularly interested in porting it quite frankly. I'll need a Windows VM to port it... which itself is a big turn off	12:53
	nonken → nonken_afk	12:53
	← gdaa left	12:53
	→ gdaa joined	12:53
artagnon	Again, thanks for the information. I'll get back to writing some "janitor code" to get started.	12:54
	← pielgrzym left	12:55
kamen	is it an expected behaviour of it or do I dod something wrong ?	12:55
	→ pielgrzym joined	12:55
kamen	I expect that if I checkout another version from the tree then I shouldn't get modifications in some file.	12:55
	or do I think wrong ?	12:56
esc	does anyone know of a git-tutorial in german language?	12:56
artagnon	Also, is there some issue tracker for Git where I can pick up small bugs to fix? Some warm-up stuff?	12:56
	I saw the TODO branch, but it's not very useful.	12:56
divVerent	just checking other alternatives for this... there is git clone --depth 1 to get a "shallow clone"	13:00
charon	lurk on the list for a week or on IRC for a few hours ;)	13:00
divVerent	this is all nice, but when I "git pull" inside it	13:01
	← Geralt left	13:01
divVerent	can I later truncate it to be a shallow clone again?	13:01
	← j416 left	13:02
	← anno^da left	13:02
	→ hendry joined	13:02
hendry	i like using `git diff --color-words --no-index orig.txt /tmp/edited.txt`, though I need to show this output to someone without a terminal.	13:03
	does anyone know how to convert colour escaped terminal output to coloured HTML?	13:03
	→ johan-s joined	13:04
	→ maleknet joined	13:05
artagnon	divVerent: See http://git.or.cz/gitwiki/SoC2009Ideas#Lazyclone.2BAC8remotealternates	13:05
	divVerent: ... and finally this ofcourse: http://git.or.cz/gitwiki/SoC2009Ideas#narrowclone	13:06
	← wallunit left	13:06
	→ Geralt joined	13:07
divVerent	do yes	13:07
	already followed the link to that	13:07
	don't worry... I don't TOTALLY like the idea of using git-cvsserver for read-only repository copies yet :P	13:08
	← dennda left	13:08
divVerent	because, I use a quilt patch set inside that, and sometimes quilt messes up and I have to "svn revert"	13:08
	i.e. git reset --hard	13:08
	→ ankit9 joined	13:09
	→ wallunit joined	13:10
	← tstclair left	13:11
divVerent	yes, narrowclone would be exactly what I need	13:13
	→ jmpc joined	13:13
	← malumalu left	13:13
artagnon	Hm. It'll probably come up as an SoC '10 project again	13:13
	← jmpc left	13:16
divVerent	how well does tortoisegit work?	13:17
	I don't have windows so I can't try	13:17
	← Geralt left	13:17
divVerent	is anyone here using it?	13:18
killerchicken_	hendry: HTML::FromANSI (perl module) might help?	13:18
	← hummingkid left	13:20
	→ hummingkid joined	13:20
	→ tstclair joined	13:21
hendry	killerchicken_: thanks, will look into it	13:21
	→ remi joined	13:22
	← catalis left	13:22
	→ kmartiin joined	13:23
	→ girishr joined	13:23
	← ankit9 left	13:23
	→ LiamH joined	13:26
	→ Guest12364 joined	13:26
	→ auslander joined	13:26
	→ akahn joined	13:27
	Guest12364 → Laraaa	13:27
	→ gambler joined	13:28
	→ tek joined	13:28
	← tekky left	13:28
	Laraaa hi?	13:30
	← Laraaa left	13:32
	→ kraymer joined	13:32
divVerent	ah, finally found an apparent bug in git-cvsserver	13:32
	cvs update -dPC (cvs equivalent to git reset --hard) downloads every single file	13:33
	→ rgr joined	13:33
	← hummingkid left	13:33
divVerent	it SHOULD know better, as cvs shows the only modified files as "M" in cvs update	13:33
	and deleting these and cvs update yields the result I want	13:33
	→ jlilly joined	13:33
divVerent	well, not surprised	13:33
	given what an ugly hack git-cvsserver is :P	13:34
rgr	I've got myself in a pickle and have no idead what to do. I tried to push some changes but had forgotten to pull. But when I try to pull, I get "error: Entry '.gitignore' would be overwritten by merge. Cannot merge.". Whats the next beginners step?	13:34
	→ marc_kdab joined	13:40
	← countd left	13:41
	→ DavidKlein joined	13:44
	→ maattd_ joined	13:45
	← maleknet left	13:45
	→ krawek joined	13:45
	← artagnon left	13:48
	← Lisimba left	13:48
	→ RobertLaptop joined	13:48
	→ _waseem joined	13:48
	→ kesor joined	13:49
kesor	hi folks, is there a way to invert git grep in "git log --grep" ?	13:49
	→ GeertB joined	13:50
_waseem	Hi, I have a file say 'foo' in my repository. From now onwards I want all the changes made to it ignored. How do I add it to the .gitignore?	13:52
	nonken_afk → nonken	13:52
	→ hummingkid joined	13:52
	← kmartiin left	13:52
	→ kssddy joined	13:54
divVerent	how to list unversioned files?	13:56
	→ kmartiin joined	13:58
divVerent	basically, like git-status, but machine readable	13:58
	← ixti left	13:58
Ilari	divVerent: 'git ls-files -o --exclude-standard'?	13:59
	→ snearch joined	14:00
	→ whaley joined	14:01
	→ mosno joined	14:01
	→ kumbayo joined	14:02
	← project2501a left	14:02
	→ Bass10 joined	14:04
	← jlilly left	14:07
	← maattd left	14:07
	maattd_ → maattd	14:07
	→ pinnacle joined	14:08
	→ jlilly joined	14:08
	→ simNIX joined	14:08
	→ anno^da joined	14:10
divVerent	thanks	14:10
	→ r11t joined	14:11
	← hummingkid left	14:12
	← rgr left	14:12
	→ aresnick joined	14:12
	→ justinweiss joined	14:16
	→ kristian42_ joined	14:17
	→ brodyberg joined	14:17
	← kristian42 left	14:18
	nonken → nonken_afk	14:21
	nonken_afk → nonken	14:22
	← jlilly left	14:23
	← schierbeck left	14:23
	→ ereslibre joined	14:24
	← _waseem left	14:25
	→ Geralt joined	14:25
	→ reprore joined	14:27
	→ TeckniX joined	14:27
	→ nothingHappens joined	14:27
	→ ronny joined	14:28
ronny	hi	14:28
	→ Simoyd joined	14:28
ronny	is there any default way to add custom metadata to a commit?	14:28
wereHamster	ronny: what kind of metadata?	14:28
	→ ereslibre_laptop joined	14:28
	→ Lisimba joined	14:28
	nonken → nonken_afk	14:29
ronny	wereHamster: anything necessary to deal with 2-way interaction with hg/bzr	14:29
wereHamster	ronny: I'd add a line similar to git-svn-id to the bottom of the commit message	14:29
	→ jfkw joined	14:30
	← kamen left	14:30
ronny	wereHamster: well, i dont want to crap stuff into git log	14:30
	← Aldwuin left	14:30
Simoyd	hey, I was wondering if there was some way to change a merge commit so that it's linear like a rebase. Is it possible to remove the reference to one of the parents? I understand this may cause problems if the commits along the no longer referenced branch commits are merged in again later, but this is for the purpose of cleanly promoting something and I'm sure I wont have those commits anymore.	14:30
wereHamster	I don't know if git notes could be useful in your situation	14:30
	← kmartiin left	14:32
ronny	git notes looks interesting	14:32
wereHamster	Simoyd: why don't you use rebase?	14:32
ronny	i wish there was more documentation	14:32
Simoyd	I would like to be able to resolve conflicts in 1 step (the way a merge works) not in possibly multiple steps (like a rebase)	14:33
	nonken_afk → nonken	14:33
Simoyd	can I "rebase" somehow in 1 step?	14:33
wereHamster	Simoyd: merge --squash?	14:33
Simoyd	cool	14:34
	I think that's exactly what I want =D	14:34
	thanks =D	14:34
ronny	hmm, whats the first git version to ship notes	14:35
	← kristian42_ left	14:36
	→ ctusar joined	14:36
wereHamster	looks like v1.6.2	14:36
ronny	i have 1.6.4 and no notes command	14:37
	← rchavik left	14:37
wereHamster	then it is v1.6.6	14:37
	→ malumalu joined	14:37
	→ spearce joined	14:38
ronny	hmm, and there is really absolutely no way for tools to add custom metadata?	14:38
wereHamster	yeah, git notes ;)	14:38
	← Shred001 left	14:38
ronny	notes are more like text messages tho	14:38
wereHamster	so what kind of metadata do you want to attach?	14:39
	binary blobs?	14:39
	as long as it is text, you can put it into the notes	14:39
ronny	hmk	14:39
	and how do i get it away from the user?	14:39
	the data is only usefull for the tools	14:39
	→ disappearedng_ joined	14:40
wereHamster	you could create your own 'notes' and store the data somewhere inside .git/..	14:40
	→ tom joined	14:40
	→ aziz joined	14:40
disappearedng_	Hey so when I do a push, I always have to do a git reset --hard to get to the HEAD, is there a better way of doing this?	14:41
wereHamster	disappearedng_: faq non-bare	14:41
Gitbot	disappearedng_: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare	14:41
tom	how can i switch to a branch? i checked out from github and "git checkout failure_backend_support" says "error: pathspec 'failure_backend_support' did not match any file(s) known to git."	14:41
	→ Shred00 joined	14:41
wereHamster	tom: 'git branch -a' to show which branches you have	14:41
	tom: if the branch begins with 'origin/' then use 'git checkout -t origin/foo'	14:42
	← BigMack83 left	14:42
tom	wereHamster: thanks, that was it	14:43
	→ fcuk112 joined	14:43
	← girishr left	14:43
tom	wereHamster: just wondering, i use "git checkout" to check out files as well (when i mess them up), why is the same command used for both things?	14:43
	← Swimming_Bird left	14:44
tom	so -t says "fatal: git checkout: --track and --no-track require -b"	14:44
wereHamster	tom: ah, so you have an older version of git.. 'git checkout -t -b foo origin/foo'	14:44
Simoyd	ok so lets say I have branch A (master) and branch B (some feature). They both have common ancestors. If I checkout B and merge A then checkout A and merge B --squash, is it possible for the second merge to ever have a conflict (assuming I am the only one modifying anything). I think it's not possible, but I want to make sure =D.	14:45
disappearedng_	um what is the post-update file for? how do I use it?	14:45
wereHamster	and about checkout doing two things, it's just like that historically...	14:45
	Simoyd: no	14:46
	disappearedng_: man githooks	14:46
Gitbot	disappearedng_: the githooks manpage can be found at http://git.or.cz/man/githooks	14:46
cbreak	disappearedng_: a hook?	14:46
wereHamster	Simoyd: well, on second thought, I don't know if that really is true when using --squash..	14:47
	← ereslibre left	14:47
disappearedng_	I can't find git hook	14:47
wereHamster	it's not a command	14:47
disappearedng_	oh ok	14:47
wereHamster	hooks are automatically run by certain git commands	14:47
	→ jrmuizel joined	14:47
tom	wereHamster: i messed somthing up and now git branch -a show me the same branch twice. can i just git branch -D? (nothing critical, i could also recheckout again if needed...)	14:48
wereHamster	tom: did you run 'git checkout -t -b origin/foo' ?	14:48
tom	wereHamster: "fatal: A branch named 'origin/failure_backend_support' already exists."	14:49
wereHamster	I did not ask you to run it, I asked you whether you ran it..	14:49
tom	but as i said i messed up and the branch now exists twice :)	14:49
	→ kifkif joined	14:49
wereHamster	ah, what the hell: git branch -D origin/failure_backend_support; git checkout -t -b failure_backend_support origin/failure_backend_support'	14:50
kifkif	Hi, I made some local changes and tried to push them to the remote repository. The commit is shown in the remote log, but the changes are not taken in account in the remote files. How can I update the remote files so that the local changes are applied?	14:50
wereHamster	kifkif: faq non-bare	14:51
Gitbot	kifkif: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare	14:51
tom	wereHamster: yup, working, thanks :)	14:51
kifkif	wereHamster, I didn't quite get how to manage bare repositories and still access my files....	14:51
wereHamster	kifkif: create a bare repo and push/pull commits from/to it..	14:52
	← Adlai left	14:52
	← ereslibre_laptop left	14:52
kifkif	wereHamster, ok, thanks a lot (and sorry for the basic question...)	14:52
	→ Adlai joined	14:52
	← Tordmor left	14:54
	→ mnyon joined	14:55
	→ Mezner joined	14:55
	→ Scorchin_ joined	14:57
Simoyd	is there any way to save my state in the middle of a merge operation, switch to a different branch, do some work, then change back and continue from the middle of the merge operation?	14:57
	← Scorchin left	14:57
Simoyd	(let's say I've merges 3 of 10 files or something)	14:57
PerlJam	Simoyd: git stash? (not sure how the merge affects that)	14:58
Simoyd	hmm	14:58
	→ bantic joined	14:58
	← mithro left	14:58
	← parasti left	14:59
	→ rkirti joined	14:59
Simoyd	well I can't find anything in the documentation saying it wont work.... so that obviously means it will work	15:00
disappearedng_	so just a quick question,	15:00
	← chrisr left	15:01
	← Scorchin_ left	15:02
	→ Axius joined	15:02
	← antonfh left	15:02
	→ ehabkost_ joined	15:03
	← LiamH left	15:04
	→ rgr joined	15:04
rkirti	I have a rather naive query - I commit make a commit X followed by Y and then push to github. But I now want to revert Y. So I do a git-reset --hard X on my local copy, but I cant push this to github.	15:05
PerlJam	rkirti: you can if you know how :)	15:06
	rkirti: but changing the history of somethign you're already shared is not very nice to those who pull from your repo	15:06
rgr	I had a submodule in a git project and it was added as a submodule "item" to the parents git repo. This caused a horrible to and from of updates since when client A changed the submodule and checked it it then client B pulled it and that pullingin cause the parents "lib" entry to update causing client A to have to pull that and to and fro and to and from if that rings any bells with anyone here. So the basic question is : whats the best way to remove a submodile from	15:06
Simoyd	rkirti: the "git push" command has a "-f" parameter for that purpose... I'm not sure what is used exatly to push to github though...	15:07
	← jrmuizel left	15:08
	→ jlilly joined	15:08
	← __bert__ left	15:08
	← Axius left	15:08
cbreak	rkirti: don't use reset if anyone else uses your repository	15:08
	← jlilly left	15:09
cbreak	if they already pulled from it, and you force push the reset, they will just push it again next time they push	15:09
	→ jlilly joined	15:09
	→ zoniq joined	15:11
rkirti	PerlJam, cbreak, Simoyd: thanks. I ll try the -f option. As to why I want to reset - we are a team using github to code together, and one person made a really horrible last commit. Noone else is currently using the code. I tried pulling, resetting and pushing back - doesnt work. he isnt around here atm, hence the fuss	15:11
	→ madewokherd joined	15:11
PerlJam	rkirti: is it so horrible that you can't just leave it there but correct it in a subsequent commit?	15:11
cbreak	the propper way to revert is git revert	15:11
	if you want to reset	15:12
	EVERY other developper that already pulled that commit has to reset it away as well	15:12
	→ Axius joined	15:13
	← ph^_ left	15:14
rkirti	PerlJam: yes. The code has been moved around and sphagetti-sed	15:14
PerlJam	rkirti: sounds like you do just want to "git revert"	15:15
	(rather than going throught the complications of a reset)	15:16
	← Modius left	15:17
PerlJam	(git revert would leave the errant commit in history, but record a new commit that undoes the effect of the errant commit)	15:17
	← akahn left	15:17
	← ehabkost left	15:17
	→ kukks joined	15:18
	← johan-s left	15:18
rkirti	PerlJam: ok, thanks	15:19
	→ akar joined	15:19
	← robinr left	15:20
	← snearch left	15:20
	← jollyroger_ left	15:22
	→ Turias joined	15:22
	→ LiamH joined	15:22
	→ hobodave joined	15:23
	← Axius left	15:23
	← akar left	15:24
	← Thumper_ left	15:24
	← Adlai left	15:25
	→ Adlai joined	15:25
	→ Swimming_Bird joined	15:25
	← crankharder left	15:26
	→ cwillu_at_work joined	15:27
	→ Thumper_ joined	15:27
cwillu_at_work	What's the difference between "GIT_PAGER=/usr/bin/less git log" and "git log \| /usr/bin/less"?	15:28
	The mouse wheel works in the latter and not the former	15:28
	cwillu_at_work is coddled by his distro's defaults	15:29
rgr	what does submodule init actually do? When I add a submodule form a bare repo the files are already extracted to my new sumodule directory.	15:30
	← rkirti left	15:31
	→ rkirti joined	15:31
cbreak	cwillu_at_work: maybe it's because git log \| less passes the data to less via standard input	15:32
	→ vals_ joined	15:32
cwillu_at_work	cbreak, as opposed to?	15:32
cbreak	so it won't get the mouse wheel commands that also get passed there	15:32
mosno	cwillu_at_work, your distro sets git environ vars?	15:32
cwillu_at_work	mosno, nope	15:32
cbreak	as opposed to any other FD	15:32
	cbreak is guessing	15:32
	← tango_ left	15:32
cwillu_at_work	cbreak, forgive my ignorance :)	15:33
	cbreak, how do keystrokes get to less in the former case then?	15:33
	i.e., you can still scroll around using the keyboard	15:33
cbreak	standard input?	15:33
	I have no idea how it works in detail	15:33
	→ JoeConyers joined	15:33
	← madewokherd left	15:34
	← hobodave left	15:34
	→ robinr joined	15:34
cbreak	but for example, mplayer can read things from standard input. If you pipe movies to it, the key commands stop working, because they are interpreted as movie	15:34
	→ aroben joined	15:34
cbreak	or, other theory	15:35
	← aroben left	15:35
cwillu_at_work	hmm, gnome-terminal magic	15:35
cbreak	git log \| less passes the scrolling to git	15:35
	→ aroben joined	15:35
cbreak	git log passes them to less, exec'ed by git log	15:35
	maybe.	15:35
cwillu_at_work	turning off the "use keystrokes to scroll on alternate screen" option disables scrolling on \| less	15:35
	→ madewokherd joined	15:35
cwillu_at_work	so gnome needs to think that git is a screen'ish thing	15:36
	← loincloth left	15:36
cwillu_at_work	okay, I'll leave you guys alone now :p	15:37
cbreak	it's all different on OS X's terminal :)	15:37
	← rseifert left	15:37
	cwillu_at_work hacks in a bash alias	15:38
Simoyd	I KNOW I've asked this question before, but I forgot and can't figure it out again.... how do I get the list that shows what files are unmerged (the one that shows each file in 3 rows with the hash on the left).	15:38
	i think it was ls-tree something...	15:39
	→ notjohn joined	15:40
divVerent	in a bare git-svn managed repository	15:40
	am I allowed to change the HEAD?	15:40
	hagger:/srv/gitosis# less repositories/nexuiz.git/HEAD	15:41
	ref: refs/heads/master	15:41
Simoyd	git ls-files -u!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1 I found it	15:41
divVerent	I basically want gitweb to show another branch by default	15:41
	← tchan left	15:41
	← jelmer left	15:42
divVerent	just am not sure if hell breaks loose once I touch that	15:42
	→ Travis-42 joined	15:42
	← ronny left	15:43
	← becomingGuru left	15:44
esc	in the context of 'git log' what exactly does '--diff-filter=M' mean	15:44
mosno	aww, jc.png is cute	15:44
	→ jrmuizel joined	15:45
	→ jelmer joined	15:45
	→ Modius joined	15:46
rgr	what does submodule init actually do? When I add a submodule from a bare repo the files are already extracted to my new submodule directory.	15:46
	→ mbernstein joined	15:47
charon	divVerent: HEAD matters when figuring out the "current" branch for the 'git svn rebase', info, etc commands, but you can't really do much except 'git svn fetch' in a bare so it doesn't matter	15:48
	← pinnacle left	15:48
charon	esc: the diff shown with 'git log' is always w.r.t. the commit's parent(s)	15:48
divVerent	charon: okay	15:48
charon	er, shown with the -p option to git-log, i should say	15:48
divVerent	git checkout -t -b div0-stable origin/div0-stable - and is there a shorthand for this?	15:49
	→ jrmuizel_ joined	15:49
wereHamster	divVerent: in 1.6.6 you cna just do checkout origin/foo	15:49
divVerent	ah, good, thanks	15:49
	← stamina_ left	15:50
divVerent	hm... can I make another branch than master default for checking out in a bare repository?	15:50
	→ tchan joined	15:50
divVerent	or did I just do that by changing HEAD?	15:50
	→ altrux joined	15:50
charon	divVerent: there are some protocol-related caveats but yes, HEAD determines the branch checked out after a clone of the repo	15:51
	← jrmuizel left	15:51
	jrmuizel_ → jrmuizel	15:51
divVerent	good	15:51
esc	charon: i see here that --diff-filer=M excludes for example merge commits.	15:52
	← reprore left	15:53
	← d0ugal left	15:53
	→ reprore joined	15:53
	→ tedoc2000 joined	15:54
	← bauruine left	15:54
	← Tureba left	15:55
	→ spvensko joined	15:55
	→ ruediger joined	15:55
	→ irc2samus joined	15:56
	→ ereslibre joined	15:59
	→ stamina joined	15:59
charon	hmm, i find that counterintuitive	16:00
	i'm also midly surprised that 'git log --no-walk --name-status' and 'git show --name-status' do not give the same results on a random merge	16:00
	← munichlinux left	16:01
	→ munichlinux joined	16:01
	sdboyer_ → sdboyer	16:01
	→ ereslibre_laptop joined	16:02
	← LiamH left	16:03
	→ LiamH1 joined	16:03
	LiamH1 → LiamH	16:03
	→ tvw joined	16:03
	nonken → nonken_afk	16:04
	nonken_afk → nonken	16:04
	← niki left	16:07
	← kesor left	16:08
	← psoo left	16:08
	→ keystr0k joined	16:09
	→ dreiss joined	16:09
	→ mwlang joined	16:09
	→ loincloth joined	16:10
mwlang	Greetings, all. I have a simple question for reverting a dumb mistake I just made...	16:10
	I somehow moved an entire folder of files into a subdirectory of a project I have under git's control.	16:10
	← Adlai left	16:11
mwlang	I did a "git add ." and saw that I picked up about 100 new files. How do I undo this to remove that accidental sub-folder from git's control?	16:11
	→ parasti joined	16:12
PerlJam	mwlang: git rm --cached <folder> # IIRC	16:12
	mwlang: you haven't done a commit yet?	16:12
	→ jschoolcraft joined	16:13
mwlang	I have not.	16:13
	last command I ran was git add .	16:13
	so I'll try that.	16:13
jschoolcraft	I'm using git svn, and I need to push my local changes to a specific svn branch (that shows up in remotes) what do I do? git push origin remote_branch?	16:14
	well, nm, no origin with svn	16:14
mwlang	PerlJam: thanks! That worked perfectly.	16:15
divVerent	Is there a command to list all branches that haven't been merged into branch X?	16:16
	← bibinou left	16:18
	→ Yuuhi``` joined	16:18
	← aziz left	16:19
PerlJam	divVerent: why do you want this?	16:19
divVerent	find all unmerged feature branches	16:19
	so they don't get "forgotten"	16:19
	→ Sky[x] joined	16:20
divVerent	or is there a better way to track that?	16:20
	← tchan left	16:20
	→ ladyinwhite joined	16:21
	→ smckinney joined	16:21
	← jrmuizel left	16:21
	→ jrmuizel joined	16:22
patrikf	i usually delete merged feature branches	16:22
	← ladyinwhite left	16:22
divVerent	patrikf: okay, that works too :P	16:22
	← Yuuhi`` left	16:24
	→ bcardarella joined	16:25
	← munichlinux left	16:25
wereHamster	divVerent: git branch --merge	16:25
divVerent	ah, thanks, found that	16:26
	← mosno left	16:26
	← tbf left	16:26
	→ mattions joined	16:27
divVerent	assume I have a bare git-svn checkout, and make a receive hook that git-svn dcommits all changes	16:30
	what trouble will I run into (like commit/update loops?)	16:30
wereHamster	divVerent: don't do that...	16:31
divVerent	why?	16:31
wereHamster	divVerent: it's recommended to run svn rebase before dcommit	16:31
divVerent	it looks like we can't convert everyone in the project to git	16:31
	because apparently there is no good way to use git on Windows	16:31
	as they say	16:31
	so basically, we at least need a read-only svn mirror of the git repository	16:32
wereHamster	have you tried msysgit? tortoisegit?	16:32
divVerent	msysgit? no	16:32
	they refuse to use commandline	16:32
	tortoisegit? the guy says it can't be installed together with tortoisesvn on Windows 7	16:32
	wereHamster: but I can't git-svn rebase if I have a bare repository, right?	16:33
	→ tchan joined	16:33
wereHamster	no, you can't	16:33
	← mwlang left	16:33
	→ MacGyverNL joined	16:33
wereHamster	I'm not even sure if you can dcommit without a working tree	16:33
divVerent	well... once we HAVE a read-only svn mirror	16:34
	we always can make a simple "emergency path" to commit	16:34
wereHamster	where will the windows folks commit to if the svn is read-only?	16:34
divVerent	like, svn diff \| curl ....	16:34
	and a CGI script that tries to apply the patch and commit+push	16:34
wereHamster	ugh..	16:35
divVerent	of course, if tortoisegit WOULD work for these guys, that wouldn't be an issue	16:36
	still, IS there a way to make a read-only svn mirror?	16:37
	→ guest_007 joined	16:37
divVerent	git-cvsserver makes a "sort of working" read-only cvs mirror, which still would be better than nothing for these users	16:38
	but well, I tested it for a short while and already found stuff that does not work :P	16:38
	← jschoolcraft left	16:38
guest_007	how to undo the git-add ? i decided not to include some files, but can't find how to remove them (i did not commit yet)	16:38
wereHamster	maybe you can use git svn set-tree somehow. But sending svn diffs through curl to a cgi script seems like a hack	16:38
divVerent	(cvs update -C considers ALL files modifed and redownloads them; cvs checkout doesn't allow checking out a subdirectory, although one CAN get that result by manually creating the CVS subdirs... both seem to be problems with git-svn-server)	16:39
wereHamster	guest_007: git reset <file>	16:39
divVerent	wereHamster: well, another option would be sending svn diffs via email, and using a key combination in mutt to apply them once read :P	16:39
wereHamster	and are you sure your windows users would be willing to run 'svn diff \| curl' but not git on the commandline	16:39
	← daaku left	16:39
divVerent	wereHamster: that's the smallest problem, for this one can make a simple batch file	16:39
guest_007	wereHamster: the problem - i did not do initial commit :) so no HEAD and can't reset	16:40
wereHamster	guest_007: git rm --cached <file>	16:40
	divVerent: how many windows/unix developrs are there?	16:40
	→ cedric_ joined	16:40
divVerent	about 60% unix, 40% windows	16:40
	← cedric_ left	16:41
guest_007	+1 unix	16:41
wereHamster	so maybe it would make sense to keep all code in a svn repo and use git-svn on unix	16:41
divVerent	but then we still need to be able to commit to svn from git	16:41
	which is basically the same problem	16:41
	also, I do want to be able to create topic branches	16:41
wereHamster	and share them with your windows developers?	16:42
divVerent	that is not entirely needed	16:42
	but if not, there must be a simple way to commit them to the svn trunk	16:42
	like one can merge them into master in git	16:42
wereHamster	you can keep your topic branches in a git repo, but when the topic is ready you dcommit it into svn	16:42
divVerent	and that works reliably?	16:43
wereHamster	why woudn't it..	16:43
divVerent	then why can't I do that in a hook script? :P	16:43
	basically, in one that's called whenever pushing to master	16:43
wereHamster	I just think it's a bad idea to run svn dcommit from a script, that's all	16:43
	← wolog left	16:44
divVerent	and personally I find it a bad idea to have working copies that are a git clone, but have a git svn init setup	16:44
	→ ph^_ joined	16:44
divVerent	as one wrong command causes weird stuff	16:44
wereHamster	heh?	16:44
divVerent	like, git svn fetch in such a working copy, and then pushing, may annoy the server	16:44
	as it would run its own git svn fetch	16:44
	basically, I don't want clients to deal BOTH with git and with svn	16:45
wereHamster	yeah, deal only with svn	16:45
divVerent	and having a very complicated setup that works neither like git nor like svn	16:45
	but I want topic branches :P	16:45
	I basically want to have git workflow, but there's svn users too	16:46
	→ jumoit joined	16:46
	→ hobodave joined	16:46
	← hobodave left	16:46
guest_007	wereHamster: Thanks :)	16:46
wereHamster	if you really want to commit from a hook, I'd use set-tree and not dcommit...	16:46
divVerent	would be just as fine for me	16:46
	→ hobodave joined	16:46
wereHamster	you probably don't care about the history in svn all that much anyway	16:46
	← cedricv left	16:46
divVerent	exactly	16:47
	just don't want stuff to bounce back and forth	16:47
	i.e. if I commit to svn	16:47
	and git-svn then fetches	16:47
	it shouldn't consider the just added commit as new	16:47
	as that'd cause the changes to be applied twice, which probably causes conflicts	16:47
	I do see the caveats for set-tree though :P	16:48
	looks like a race condition between fetch and set-tree	16:48
	well...	16:48
	even better would be getting that tortoisegit problem solved	16:48
	then the windows users can use git	16:48
	→ kmartiin joined	16:48
wereHamster	well, git svn runs rebase after dcommit to get rid of those duplicated commits	16:48
	← ereslibre_laptop left	16:49
	← godsflaw left	16:49
divVerent	actually, commit-diff looks fine for me too	16:49
	but well	16:49
	top solution would still be getting a git GUI working on windows	16:49
wereHamster	you did report the issure with tortoisegit, right?	16:50
	← kukks left	16:50
divVerent	I don't have the issue	16:50
	so I can't report it	16:50
	some guy in the team has it	16:50
	and all he can say is that "it does not work"	16:50
wereHamster	did he?	16:50
divVerent	no	16:51
wereHamster	if not, it likely won't get fixed	16:51
divVerent	sure	16:51
	maybe there is no issue even	16:51
	and he did something wrong when installing	16:51
	→ cedricv joined	16:51
wereHamster	get him to submit a ticked to the tortoisegit team	16:51
	← cedricv left	16:51
	→ bryanl joined	16:52
wereHamster	lol, you're going through all that trouble just because he can't install that little piece of software..	16:52
divVerent	well	16:52
	← MacGyverNL_ left	16:52
	← krawek left	16:52
	→ kukks joined	16:52
divVerent	supposedly it cannot be installed together with tortoisesvn	16:52
	I cannot check that	16:52
bryanl	if i have a branch that is ahead of it's remote branch by one command and i run "git push foo bar", why does it say everything is up-to-date when it clearly isn't	16:52
wereHamster	bryanl: maybe you are not pushing into the branch that your local branch is tracking.	16:53
	did you run 'git remote update' after pushing?	16:53
bryanl	wereHamster: i did	16:54
	here is what happens: http://gist.github.com/270406	16:55
wereHamster	and that 'git remote update' didn't help?	16:55
bryanl	nope	16:55
	← tazle_ left	16:56
wereHamster	are you fetching through http?	16:56
	← tchan left	16:56
bryanl	no over ssh	16:57
	← maattd left	16:57
	→ tazle_ joined	16:57
	→ cedricv joined	16:58
	→ alley_cat joined	16:58
	→ munichlinux joined	16:58
	aroben → aroben\|lunch	16:59
	← tek left	17:00
	→ tekky joined	17:00
	→ amerine joined	17:01
disappearedng_	what is the easiest way to solve non-bare repository ? Like when someone push while I am editing.	17:01
	→ cedric_ joined	17:02
	→ sudoer joined	17:02
wereHamster	disappearedng_: don't push into non-bare repos.. ?	17:02
	→ livid joined	17:02
wereHamster	bryanl: sorry, no idea	17:02
	← rkirti left	17:02
	→ rkirti joined	17:03
killerchicken_	bryan1: did you try git fetch sneezy	17:03
	→ tchan joined	17:03
	→ cedric__ joined	17:04
livid	anyone have any experience using git with mainly binary files or for engineering drawings (i.e. non-software)? any help/recommendations for setting this up would be great	17:05
	← vbgunz left	17:05
wereHamster	livid: git is generally not so good at managing binary files	17:05
	← cedricv left	17:05
wereHamster	compression does suffer a lot	17:05
livid	wereHamster: I am sort of trying to decide what version control software to use - any recommendations on things to try? I plan on trying SVN as well	17:06
wereHamster	if you really want to use git, disable delta compression for the binary files (can be done through git attributes)	17:06
	→ cedricv joined	17:06
wereHamster	svn sounds good for that type of files	17:06
	← tchan left	17:06
	← wallunit left	17:06
livid	wereHamster, yea, that is kind of what it seems like, but it is good to hear it from another source :)	17:06
	→ efphe joined	17:07
efphe	hello, i have a very stupid question (i'm new on git): which is the git command to have the same effect (dropping changes) to this svn sequence: `rm file && svn up fiile`	17:08
teuf	git checkout -- file	17:08
killerchicken_	efphe: git checkout -- <file>	17:08
efphe	oh	17:09
	good	17:09
	→ kristian42 joined	17:09
efphe	thank You killerchicken_, sorry for the really basic question	17:09
killerchicken_	efphe: no worries	17:09
	let me explain a little what it does	17:09
efphe	sure	17:10
killerchicken_	it means: "update the working tree with the file specified, and use the file that is currently in HEAD (the commit that is currently checked out, without any local modifications)"	17:10
	HEAD is implied, because you didn't give a commit	17:10
	→ ereslibre_laptop joined	17:11
wereHamster	killerchicken_: s/HEAD/index/	17:11
	← cedric_ left	17:11
	← cedric__ left	17:11
efphe	thank you again	17:11
killerchicken_	wereHamster: hah.	17:11
efphe	i was surprised when `rm file` caused a deletion on the repository	17:12
	→ jessep joined	17:13
	← kifkif left	17:13
killerchicken_	efphe: hm?	17:13
	← EricInBNE left	17:13
efphe	if I remove a file in a git repo	17:13
	it wants to delete it also on the repository	17:14
wereHamster	no, that's not true	17:14
	(unless you use git add -A or commit -a)	17:14
killerchicken_	only if you git rm it	17:14
	→ nadavoid joined	17:14
killerchicken_	or git add... right.	17:14
efphe	yes	17:14
	but in fact this is different from svn and initially i was a littel bit confused	17:15
	ok	17:15
killerchicken_	git isn't trying to match svn's behaviour, in fact, many people working on git don't know how svn works.	17:15
	← rgr left	17:15
wereHamster	'rm *; git commit' fails with 'nothing to commit'	17:15
efphe	now i know how to restore files with checout	17:15
ereslibre_laptop	efphe: it kind of assumes you wanted to use "git rm", so you just need to do "git ci -a" and you are done, the file also is removed on the repo	17:15
	← cedricv left	17:15
ereslibre_laptop	wereHamster: rm *; git ci -a, should remove all files though ;)	17:16
wereHamster	yes	17:16
	← kmartiin left	17:17
sourcemaker_	I am playing around with git... and now I think I have unterstood the functionality :-)	17:19
	Now I am trying to build a small script to backup my .kde data...	17:19
efphe	sourcemaker_, if you are new as me and you need server side repos, take a look at gitolite	17:20
sourcemaker_	efphe: now... I will use git as local repository	17:20
	is this enought? git add . && git commit -a -m`date '+%d.%m.%Y-%H:%M:%S'` (daily script as cron)	17:22
wereHamster	sourcemaker_: use git add -A; git commit -m '...'	17:22
sourcemaker_	wereHamster: ok	17:23
	when I execute this script every day... can I switch back let's say 3 days before?	17:24
	or do i need tags or branches...?	17:24
ereslibre_laptop	sourcemaker_: you don't need tags nor branches. git update -rHEAD^^^	17:24
	→ squarism joined	17:25
ereslibre_laptop	sorry	17:25
	git co HEAD^^^	17:25
wereHamster	update -r ? git is not svn...	17:25
ereslibre_laptop	wereHamster: yep, too many scm in parallel atm ;)	17:25
	← bryanl left	17:25
	→ d0ugal joined	17:26
	→ patmaddox joined	17:26
ereslibre_laptop	sourcemaker_: "git co HEAD^^^" is equivalent to "git co HEAD~3". which is: please, move 3 revisions back in history	17:27
	← dreiss left	17:27
wereHamster	and detach HEAD	17:28
	to go bach where you were before you can either use 'git checkout -' or 'git checkout master' (assuming you were on the 'master' branch before)	17:28
	→ coolcat joined	17:29
	→ schierbeck joined	17:29
ereslibre_laptop	sourcemaker_: as an aside note, if you are configuring cron for doing one run per day, you are getting too many inforation in the commit message. you don't need hours, minutes or seconds because you are doing it once a day and always at the same time, you only need day, month, year	17:30
	← oganso-- left	17:31
	← munichlinux left	17:31
	→ rseifert joined	17:32
	→ rgr joined	17:33
	→ tchan joined	17:33
	← rgr left	17:33
	→ rgr joined	17:34
	aroben\|lunch → aroben	17:35
	← sdboyer left	17:35
	→ wolog joined	17:36
sourcemaker_	ereslibre_laptop: thanks... you right :-)	17:39
	→ Axius joined	17:39
efphe	cu and ta again	17:40
	← efphe left	17:40
	→ bryanl joined	17:40
	→ maattd joined	17:42
	→ tbuehlmann joined	17:42
	← rgr left	17:43
	→ ixti joined	17:43
	← irc2samus left	17:45
	→ bibinou joined	17:45
sourcemaker_	ereslibre_laptop: so I create my git repository in .kde directory and execute the script every day?	17:47
ereslibre	sourcemaker_: note that you need to first cd .kde	17:52
	sourcemaker_: but yes, that is the idea	17:52
sourcemaker_	ereslibre_laptop: and how can I revert in some days...	17:53
	→ rgr joined	17:53
	→ tek joined	17:53
	← tekky left	17:53
ereslibre	sourcemaker_: you can use git revert	17:53
sourcemaker_	ereslibre_laptop: and all meta files are stored in .git... so if I would like to remove git... i only delete .git?	17:55
	→ __iron joined	17:56
	← Mezner left	17:56
	→ Alagar joined	17:57
	→ Mezner joined	17:57
	→ Adlai joined	17:58
	→ chalkbag joined	17:58
	← Mezner left	17:59
	→ Mezner joined	17:59
	→ BigMack83 joined	17:59
ereslibre	sourcemaker_: yes, if you are going to do a "rm -rf .git" FIRST make sure you _are_ on the revision you want to keep yourself. Note you won't be able to later go back again	18:00
	→ justin-george joined	18:00
sourcemaker_	ereslibre_laptop: ok thanks for helping me... :-)	18:00
	ereslibre_laptop: I will activate the script now :-)	18:00
	← squentin left	18:01
	→ drizzd_ joined	18:02
sourcemaker_	ereslibre_laptop: sorry... last question... I create only a simple git repository with "git init" ... without -bare -shared and so on?	18:03
	nonken → nonken_afk	18:05
	OtherJakeSays → JakeSays	18:05
	→ parolang joined	18:05
	← ToxicFrog left	18:09
	nonken_afk → nonken	18:09
	← brodyberg left	18:09
	← Aku left	18:09
	→ silas joined	18:10
silas	Hello.	18:10
	I think it is not possible to have two urls for the same submodule, in the case that checkout from a url doesn't work. Have you ever have this problem?	18:10
	→ squentin joined	18:11
	← anno^da left	18:11
	← bryanl left	18:11
	→ m104 joined	18:12
	← drizzd left	18:12
	← shruggar left	18:14
ereslibre_laptop	sourcemaker_: no need of those. --bare is usually for servers setting up a repo, and never used --shared. A raw "git init" should work out of the box ;)	18:14
	→ Akufen joined	18:14
sourcemaker_	ereslibre_laptop: thanks	18:14
ereslibre_laptop	sourcemaker_: np	18:15
	→ Kipras__ joined	18:15
	← jrmuizel left	18:16
	nonken → nonken_afk	18:16
	nonken_afk → nonken	18:16
	→ kipras_ joined	18:19
	← Kipras__ left	18:19
	→ ShadeHawk joined	18:19
	nonken → nonken_afk	18:19
	← amerine left	18:20
	← madewokherd left	18:20
ShadeHawk	warthog9: Do you have any comments to my take on your gitweb patches, early part (http://thread.gmane.org/gmane.comp.version-control.git/136077)?	18:20
	→ pantsman joined	18:20
ShadeHawk	warthog9: if so, now is the time before those patches land in 'next' (http://permalink.gmane.org/gmane.comp.version-control.git/136165)	18:21
	nonken_afk → nonken	18:22
	→ madewokherd joined	18:24
	← mattions left	18:24
	→ jumoit1 joined	18:25
	← jumoit1 left	18:26
	→ dreiss joined	18:27
	← jumoit left	18:28
	← gitte left	18:29
	→ _Vi joined	18:30
	← d0ugal left	18:31
	→ Axius_ joined	18:32
	→ bry joined	18:32
	→ dacqueries joined	18:33
Simoyd	ok sorry for spam, but it's easier than explaining and noone is really talking	18:33
	c:\dp\Projects\fakerepo>git status	18:33
	# On branch blah7	18:33
	# Your branch and 'master' have diverged,	18:33
	# and have 2 and 1 different commit(s) each, respectively.	18:33
	#	18:33
	# Changes to be committed:	18:33
	# (use "git reset HEAD <file>..." to unstage)	18:33
	#	18:33
	# modified: one.txt	18:33
	# modified: three.txt	18:33
	# modified: two.txt	18:33
	#	18:33
	c:\dp\Projects\fakerepo>git diff --name-only	18:33
_Vi	Simoyd, pastebin.	18:33
Simoyd	c:\dp\Projects\fakerepo>	18:33
	why does that happen?	18:33
	I just did a merge and added everything	18:33
	oh	18:33
	that would have been a good idea sorry	18:33
	next time I'll use that sorry	18:33
	why does git diff report nothing after I've git added all the merge conflicts?	18:34
_Vi	Simoyd, What "that" happened?	18:34
Simoyd	"that" => the diff reporting nothing	18:34
patrikf	Simoyd: git diff = diff between index and working tree	18:34
	→ thiago_home joined	18:35
_Vi	Simoyd, May be "one.txt", "tree.txt" and "two.txt" has just updated timestamp?	18:35
patrikf	Simoyd: maybe you want git diff HEAD = diff between HEAD and working tree	18:35
	Simoyd: or maybe you want git diff --cached = diff between HEAD and index	18:35
_Vi	Simoyd, Yes. Default "git diff" is diff of index and working tree.	18:35
Simoyd	ok I understand now.	18:36
_Vi	Simoyd, You need either "git diff HEAD" or "git diff HEAD --cached"	18:36
Simoyd	thanks alot =D	18:36
	was really confused there hehe	18:36
	→ krawek joined	18:36
	← Axius left	18:37
	← rgr left	18:37
	→ aspotashev joined	18:37
	← ixti left	18:39
	← Adlai left	18:39
	→ ixti joined	18:39
	→ pklingem2 joined	18:39
Simoyd	I love this channel, you guys always solve all my problems	18:40
	→ numberGrey joined	18:40
_Vi	Simoyd, When I was starting to use git, the channel was solving all my problems as well.	18:41
	→ Adlai joined	18:41
_Vi	s/to use/using/	18:41
	← NET\|\|abuse left	18:44
	→ snearch_ joined	18:44
_Vi	Can I enable push over HTTP for everyone, but with denyDeletes denyNonFastForwards for anonymous users? Like in Wiki.	18:45
	→ mmoya joined	18:45
warthog9	ShadeHawk: yeah, I've been on vacation and trying to get the stuff I missed dealt with	18:46
spearce	_Vi: only if you are using git 1.6.6 and its smart-http push support. you can't do that for webdav based push for older than 1.6.6.	18:46
wereHamster	_Vi: pre 1.6.6 git doesn't even support push over http..	18:46
spearce	wereHamster: it does, it uses WebDAV, but that means you can't do branch controls like denyDeletes	18:46
_Vi	spearce, Yes, I mean with http-backend. I've just tried it.	18:46
	→ jrmuizel joined	18:47
	→ dfr\|work joined	18:47
	← jrmuizel left	18:47
spearce	yea, http-backend should honor hooks just like ssh does, underneath they both are running git-receive-pack.	18:47
	→ froschi joined	18:48
_Vi	spearce, Yes, I validated that hook works. Can I verify HTTP username in a hook?	18:48
spearce	yea, the CGI environment is inherited by the hook	18:48
_Vi	spearce, (Oh, forgot that it's just an environment variable in CGI)	18:48
	→ rgr joined	18:49
_Vi	spearce, (BTW someone told me that I should CC patches to spearce in order to be applied upstream. I've sent resent one with CC but still no answer. Just wait more?)	18:49
	livid → alex_m	18:50
spearce	patches to me sometimes have a lot of lag because i tend to work on things in spurts. what patch are you talking about?	18:51
	→ seanh_ joined	18:51
_Vi	spearce, I want hotkeys in "git gui" for "Revert changes" and "Unstage from commit". So I've created a patch.	18:51
	spearce, It's very simple. Ctrl+J, Ctrl+U respectively.	18:51
Simoyd	question: Lets say I add all merge conflicts to the cache, but dont commit. how do I tell that I'm finishing a merge commit and dont just have some files on the cache for a normal commit?	18:52
seanh_	Argh! I work sometimes from home sometimes the office, and I just came into the office and realised I forgot to push the work I did at home, I'm gonna have to go back! Worse, it's the second time in as many days!	18:52
spearce	oh, yea. that. git-gui, so its on the backlog queue. :-)	18:52
Simoyd	seanh_: set up a vpn so you can pull from that home repo	18:53
seanh_	So I'm thinking of using Dropbox, a program that syncs your files automatically, but keeping a git repo in Dropbox? I don't trust it. I'm thinking something like a detached working tree, with the working tree in the Dropbox folder but the git repo itself outside of Dropbox	18:53
wereHamster	Simoyd: git knows that you attempted a merge commid, you don't have to tell that	18:53
Simoyd	yea, but I want to know :S	18:54
	:(	18:54
	is there any way to tell?	18:54
wereHamster	if .git/MERGE_HEAD exists AFAIK	18:54
seanh_	Simoyd: the home computer isn't even turned on! But you're right, I have a low-power laptop at home, I should use it as a file server	18:54
wereHamster	don't know if there is an easier way	18:54
	btw, why do you want to know that?	18:54
_Vi	spearce, 1. How long should I wait before considering patches lost? 2. Should I send patches when earlier remains unanswered?	18:55
Simoyd	wereHamster: I'm creating a GUI for idiots in the company here to use	18:55
	→ corni joined	18:55
Simoyd	gitGUI is still way too complicated for them	18:55
	→ David-T joined	18:55
ShadeHawk	Simoyd: what OS	18:55
wereHamster	for unix or windows?	18:55
Simoyd	windows	18:55
	← rkirti left	18:55
Simoyd	c# wpf	18:55
ciaranm	for git add -i, would anyone else find it useful if every fifth line in long lists of filenames were shown in bold?	18:55
ShadeHawk	smartGit and git-cola?	18:56
spearce	Simoyd: what's funny is, git-gui was created to help the idiots at my company where the command line was too complex ;-)	18:56
	_Vi: with git-gui, i can easily be behind by a couple of weeks. :-(	18:56
	_Vi: but gentle reminds work.	18:56
_Vi	spearce, >"git-gui was created to help the idiots ..." I think it's useful for power users too.	18:56
Simoyd	haha that's hilarious =D	18:56
_Vi	spearce, For example, it's handy to do "stage seleted line"/"unstage seleted line" in git-gui.	18:57
spearce	_Vi: oh, i agree, that's why i still use git-gui, just to help me review and stage code for commit	18:57
_Vi	spearce, (And it's not handy that I can't do "stage seleted lineS" (multiple lines))	18:57
	← zoniq left	18:58
Simoyd	tbh I dont understand why people find it so hard to understand git	18:58
	it's easy and intuitive	18:58
_Vi	spearce, Sometimes "stage hunk" stages too lot, but there too many "stage line" for that.	18:58
	← cwillu_at_work left	18:59
	→ [1]intripoon joined	18:59
_Vi	spearce, Also "git gui" is handy as you may enter commit message (beforehand) and it will not be lost.	18:59
	Simoyd, Git was the first VCS I learned to use well. May be it matters?	19:00
spearce	its also the first i've learned to use well, and i've used CVS, Perforce, SVN, BitKeeper, ClearCase, PVCS ... :-)	19:01
Simoyd	I've used subversion... not extensivly, and it's intuitive too, but git is millions times better	19:01
	← intripoon left	19:01
Simoyd	but I'm really good at git now =D (thanks to this channel)	19:01
	→ [2]intripoon joined	19:01
	[2]intripoon → intripoon	19:02
wereHamster	oh, clearcase :( I hope you dont have any permanent damage from using it :P	19:02
	← __iron left	19:02
_Vi	spearce, I've only used CVS some time before (just to commit to sourceforge and forget). And after Git I've learned Subversion.	19:02
	(Our company is migrating from ClearCase to Subversion)	19:02
	→ __iron joined	19:02
	← froschi1 left	19:02
	← Pupeno left	19:03
drizzd_	don't get me started on clearcase	19:03
	→ samgranieri joined	19:03
_Vi	drizzd_, "git-cc clone $URL"?	19:04
	→ mbroeker joined	19:04
	← snearch_ left	19:04
	← [1]intripoon left	19:05
	← seanh_ left	19:05
	→ jrmuizel joined	19:06
	← samgranieri left	19:06
	→ samgranieri joined	19:07
	← jrmuizel left	19:08
	← jessep left	19:11
rgr	when you use a submodule in a project It seems it would generally be better to ignore its directory name in the parent project or all sorts of issues occur when you pull a new submodule but later try to pull the parent - the problem being that the submodule "directory" is marked as "modified" causing merge issues. Any best practices or comments to help me here?	19:11
DrNick	commit the modified submodule	19:12
_Vi	rgr, "Modified" submodule means that version of submodule you are using is not the same as stored in main project.	19:12
madewokherd	git bisect is smarter than I remember	19:13
_Vi	rgr, You should "git add path/to/subm" to stage the new version of submodule.	19:13
	→ mikem23 joined	19:13
divVerent	I read something interesting... and wonder if git has support for that	19:14
	this is regarding clearcase and other commercial systems :P	19:14
	many support "locking" files	19:14
	mostly to avoid having two people work on the same, e.g., image file	19:14
rgr	I realise I can commit in the parent to commit the "submodule" but I dont really see what it brings me. I wonder if better to just ignore it in the parent.	19:14
divVerent	as these are unmergable	19:14
	does git provide some sort of "advisory locking", or is there a frontend doing that?	19:15
	I mean, sure you can just add e.g. filename.lck files, or pull and check if they are there	19:15
	but are there already any frontends doing that?	19:15
_Vi	divVerent, Locking with DVCS?	19:15
divVerent	well	19:15
	it'd be useful in commercial environments	19:16
	otherwise probably not	19:16
	but well, not talking about enforced lockas	19:16
	as that would simply make no sense	19:16
	certainly I wouldn't recommend e.g. locking C source files before editing	19:16
_Vi	divVerent, May be just use something not tied to Git?	19:16
divVerent	that would inhibit work	19:16
	← madewokherd left	19:16
divVerent	_Vi: well, in the Nexuiz development, we currently use IRC topics for that purpose	19:17
	like, "<nickname> is currently working on <thing>, do not touch" :P	19:17
_Vi	divVerent, E.g. "try_lock.pl path/to/my/file.png" -> "Connecting to corporate server..." -> "Error: Alerady locked by Dohn Joe".	19:17
divVerent	but yes	19:17
	that's the idea	19:17
	could e.g. be tied to many text editors as file-open hook or macro	19:17
drizzd_	_Vi: clearcase does not have atomic commits, a meaningful export to git is impossible	19:17
divVerent	and to image editors too probably	19:17
_Vi	drizzd_, "git-cc" was on the rights of joke.	19:18
divVerent	just asking if such a system already exists	19:18
_Vi	divVerent, How Git should help such lockings?	19:18
divVerent	drizzd_: well... cvs doesn't have that either	19:18
	→ cedricv joined	19:18
divVerent	_Vi: git metadata could be used to find the lock server	19:18
_Vi	divVerent, Git only will tell about locking failures (merge conflict in binary file).	19:18
divVerent	otherwise, probably not much	19:18
	→ jrmuizel joined	19:18
_Vi	divVerent, Empty commits like "git commit --allow-empty -m 'I am locking data/images/qqq.png'"?	19:19
divVerent	just asking, as git is a system which may actulaly need an external tool for this	19:19
	while most VCSes don't need that	19:19
wereHamster	divVerent: just because someone is working on that file doesn't mean his work will interfere with your. So what would that information be useful for?	19:19
divVerent	wereHamster: well	19:19
	if we both are working on e.g. an image file	19:19
	one of the person will lose his work in the end, or have to redo it	19:19
broonie	Any distributed VCS is going to need some external system if you want locks.	19:19
divVerent	but well	19:20
	I am clearly against locking mergable files	19:20
	it'd be just better handling for unmergable stuff	19:20
	to avoid duplicate work	19:20
	→ cedric_ joined	19:20
divVerent	svn apparently provides a locking facility, but I never tried it	19:20
	cvs has one, but I hate it :P	19:20
drizzd_	divVerent: yes, the same applies to cvs, but in cvs you just do 'cvs commit', IIRC, and that will commit everything. there is no analogous command in clearcase, so checkin's are usually done sequentially.	19:20
_Vi	divVerent, (Thinking how to add lockings to Git with minimal efforts)	19:20
divVerent	_Vi: well	19:20
	personally, I am thinking of git integration like this:	19:21
	- git commit and git add would refuse to work on files of certain configurable name patterns (or other selection) if not locked	19:21
_Vi	divVerent, E.g. how user should notify the server that it lock some file. Over some Git thing? Over separate channel?	19:21
divVerent	- the lock command would be in the "git namespace"	19:21
	of course... there's one obvious problem	19:21
	you could commit it first, then someone else locks the file, before you push	19:21
	→ kain joined	19:21
divVerent	so you basically would, as user, have to make sure that you only remove the lock after pushing	19:22
	certainly wouldn't be trivial	19:22
broonie	You'd need to define some sort of workflow around it, probably with porcelain that forced push/pull.	19:22
divVerent	git-add could be wrapped to require locking on certain fails	19:22
	*files	19:22
kain	hi there. how can I join two commits to figure as one before pushing?	19:22
	← [Po]lentino left	19:22
	→ cedric__ joined	19:22
divVerent	and git-push could remove the lock on any file that was already committed	19:22
	broonie: exactly what I am thinking of	19:22
	just asking whether it already exists	19:23
	currently I don't need it	19:23
_Vi	divVerent, 1. Acquiring a lock is some push. 2. Releasing a lock is also some push. Something like this?	19:23
divVerent	_Vi: the locks don't need to be on the git repository	19:23
	it's more about integration with the git commands	19:23
_Vi	divVerent, E.g. push empty commit with special message. Commit hook on server side will reject the push in any case, but preset different messages.	19:23
divVerent	and then, a commit hook on the server could check if the file was locked by the right user	19:24
_Vi	divVerent, Imagine there's a hook on Git server.	19:24
divVerent	and only accept the commit if the files were locked (i.e. those that need it)	19:24
	actually, that'd be mostly transparent	19:24
	← cedricv left	19:24
divVerent	client would basically only need a little script to send the lock commit	19:24
_Vi	divVerent, First user make any change to the file it wants to lock (not useful change).	19:24
divVerent	still, this would have to integrate with editing tools too	19:25
_Vi	divVerent, Then he tries to push the change to server. Server rejects and sets a lock.	19:25
divVerent	can git be instructed to make all working copy files of certain patterns read only?	19:25
	that'd basically enforce workflow then	19:25
	as a "lock" tool would make the file writable	19:25
_Vi	divVerent, Then user modifies file and pushes it again. Then server accepts the push and removes the lock.	19:25
divVerent	of course, for this to be useful and not annoying	19:26
	locks should be advisory, and only be asked for on certain file types	19:26
	locking e.g. C source is nonsense	19:26
_Vi	divVerent, Something like this? /* last 6 messages I've written without reading answers */	19:26
divVerent	however, once a lock is set, anyone attempting to edit certain files should be notiifed	19:26
	_Vi: mostly	19:26
	but the challenge would be making it non-annoying :P	19:26
_Vi	divVerent, We need special script to make this not annoying.	19:27
	← tek left	19:27
divVerent	is it possible to set file permissions to 444 on certain file types in the working copy?	19:27
	→ tekky joined	19:27
_Vi	divVerent, Using a hook, yes.	19:28
	divVerent, (hooks in all user's repositories)	19:28
divVerent	oh	19:28
	← sourcemaker_ left	19:28
divVerent	that sounds like an additional required action after initial clone	19:28
	tekky → tek	19:28
_Vi	divVerent, Yes.	19:28
divVerent	well, could be e.g. an initialization script inside the repository	19:28
_Vi	divVerent, (For example, submodules also require additional actions after clone)	19:29
	← Lisimba left	19:29
divVerent	well, when I have time, I'll code that	19:29
	just wonder if it will be any more useful than what I already have	19:29
	which is using IRC topics to "lock" :P	19:29
_Vi	divVerent, First it needs to be well-designed.	19:29
divVerent	well, I see that the design works now	19:29
_Vi	divVerent, (Thinking about GOOD design).	19:29
divVerent	of course	19:30
	→ Twisted_Mentat joined	19:30
divVerent	before I'd start implementing it, I'd actually write the design down, and look for problems	19:30
_Vi	divVerent, Using empty commits always rejected commits to notify server. Is it good?	19:30
divVerent	it's just... I now see it's POSSIBLE to do it even cleanly	19:30
	_Vi: not entirely :P	19:30
	but funny hack to avoid a server process	19:30
	→ mw44118-1ork joined	19:30
_Vi	divVerent, E.g. using push to lock something. Already bad?	19:30
	← schierbeck left	19:30
divVerent	no	19:31
	← dacqueries left	19:31
divVerent	or well	19:31
	yes, as it requires a commit first	19:31
	unless you do something really weird with push	19:31
_Vi	divVerent, Always-rejected push. Already bad?	19:31
divVerent	like attempting to create a special branch name	19:31
	← kumbayo left	19:31
divVerent	always-rejected push sounds bad to me	19:31
_Vi	divVerent, Special branches...	19:31
divVerent	but better than extra server process	19:31
	→ teloscientist joined	19:31
teloscientist	newbie question for git	19:31
_Vi	divVerent, BTW you may have branches with competely different layout.	19:32
divVerent	sure	19:32
teloscientist	I have a depository. I want to invite a developer to access it.	19:32
_Vi	divVerent, (thinking about using special locking branch)	19:32
	→ cedricv joined	19:32
_Vi	teloscientist, What's URL?	19:32
divVerent	well, branches with different layout is bad design too :P	19:32
teloscientist	Do I ask them to generate a keypair, then input the keypair into my process	19:32
divVerent	are you using gitosis?	19:32
teloscientist	unfuddle.com	19:33
	← cedric__ left	19:33
divVerent	ah, okay... keypair?	19:33
	is it ssh based?	19:33
	then they'd generate a key with "ssh-keygen" or on windows puttygen	19:33
	and you'd add it there	19:33
_Vi	divVerent, Using informative branch names? (push origin HEAD:file_data_images_foo_png_is_locked).	19:33
teloscientist	and here is where my ignorance shines through... A	19:33
divVerent	_Vi: that'd be VERY irritating in gitweb :P	19:33
teloscientist	so they send me their keypair, and I input it	19:33
	?	19:33
divVerent	teloscientist: basically yes, this is how it works for ssh	19:34
	you never create ssh keys for someone else	19:34
teloscientist	great	19:34
divVerent	but they create it, and you add their public key	19:34
_Vi	divVerent, Hiding branches from gitweb. Using refs not in "refs/head", but in "refs/locks" or something.	19:34
divVerent	tell them to NOT send you their private key	19:34
	in case of openssh, you want the id_rsa.pub, not the id_rsa	19:34
teloscientist	it seemed weird that they would give me their key, but if that is how it is, that is how it is,	19:34
	thanks!	19:34
divVerent	explicitly tell them to not email the private key	19:34
	the correct file is one line long and starts with ssh-rsa or ssh-dsa	19:34
	if it has -----STUFF---- in it it probably is wrong	19:35
teloscientist	I should as for their .pub not .rsa	19:35
	← Geralt left	19:35
	← Travis-42 left	19:35
divVerent	_Vi: refs/locks, looks like we're getting somewhere	19:35
_Vi	divVerent, 1. After every update some hook starts at user side and assigns "-r" to all binary files.	19:36
	← tek left	19:36
_Vi	divVerent, 2. User runs "./acquire_lock file.png".	19:36
divVerent	right, that chmods +w the file, and pushes the current head as branch with proper name	19:37
	→ radii joined	19:37
_Vi	divVerent, 3. The script does "git push origin <some_my_commit>:refs/locks/hash_of_file_pathname"	19:37
	← mmoya left	19:37
divVerent	now comes an idea... a push fails if that head already exists but cannot be fastforwarded to the rev you specify	19:37
teloscientist	@div - thanks a bunch!	19:37
divVerent	can that be used to make the locks atomic?	19:37
_Vi	divVerent, 4. If the push is rejected as non-fast-forward (the lock exists), the locking fails.	19:37
divVerent	exactly	19:38
	but how to make it always non-fast-forward?	19:38
	→ mstrl joined	19:38
divVerent	if the lock already exists that is	19:38
_Vi	divVerent, 4b. The push succeeds. Scripts makes "+w".	19:38
	divVerent, 5. Users edits the file	19:38
divVerent	stupid workaround: create an initial parent commit, and lots of child commits, one for each user ID	19:38
	and always push the "user ID commit" :P	19:38
	that ensures fastforward is not possible	19:38
_Vi	divVerent, 6. User runs "release_lock my_file". It uses "git add my file", "git push origin master:master" and "git push origin :refs/locks/hash_of_file_pathname".	19:39
	divVerent, (Sorry, forgot "git commit")	19:39
divVerent	to remove the lock, execute two steps in sequence: first push a commit that has ALL the user ID commits as parent	19:39
	then push-clear the branch	19:40
	that'd be race condition free too	19:40
_Vi	divVerent, BTW this scheme requires almost no changes on the server side.	19:40
divVerent	right	19:40
	for the ensuring that no fast forward can happen, it'd require some dummy commits at initialization	19:40
_Vi	divVerent, (You'll think it to the end and implement yourself, right?)	19:40
divVerent	one common parent, one for each user, and then lots of merges of all the users into one common head	19:40
	right	19:40
	← samgranieri left	19:40
Oxf13	question, I've got a repo with a number of in-repo branches. Is there a way to easily look at a tag object and tell which branch the tag was created on?	19:41
divVerent	looks like good design now with that	19:41
	← mw44118-work left	19:41
divVerent	adding users to that system is simple too	19:41
	← cedric_ left	19:41
_Vi	divVerent, The main idea is requiring fast-forward pushes (the default), lock==branch_existence and using hashes of pathnames.	19:41
divVerent	right	19:42
	although I may rather url-escape the pathname	19:42
_Vi	divVerent, Each user is some commit that we may push to refs/locks.	19:42
divVerent	so the branch names are human readable	19:42
	← cedricv left	19:42
divVerent	right	19:42
	← teloscientist left	19:42
divVerent	and to get lock-removing atomic... I just noticed we don't NEED that :P	19:42
_Vi	divVerent, Inside that commit there is "This is lock of Mary C. Jane"	19:42
divVerent	because, if you remove a lock, there's a precondition that you have created the lock	19:43
	of course, all this assumes centralized workflow	19:43
	loincloth → off_	19:43
	off_ → loincloth	19:43
_Vi	divVerent, BTW there may be hook on server which checks if users has locked the file it is pushing.	19:43
divVerent	sure, that is the simplest part	19:44
	does git track file access rights?	19:44
	like executable, writable?	19:44
_Vi	divVerent, No, only "+x" or not "+x" and only for files.	19:44
divVerent	okay, so like svn	19:44
_Vi	divVerent, Internally it has storage for file mode, but it is not used.	19:45
	← r11t left	19:45
	→ mattions joined	19:46
	← alley_cat left	19:46
divVerent	are there other ways to associate attributes with files?	19:46
	and I mean in the files, not in a .gitattributes file?	19:46
	think "svn properties" or "alternate data streams"	19:47
_Vi	divVerent, "chmod +x w", "git diff" -> "diff --git a/w b/w" "old mode 100644" "new mode 100755". But it's only for executable bit.	19:47
divVerent	sure	19:47
	would be annoying in many use cases otherwise	19:47
	like everyone once has had with cvs :P	19:48
	→ madewokherd joined	19:48
_Vi	divVerent, Be like "etckeeper". Store all modes separately and reapply them on checkout.	19:48
divVerent	I still wonder how git-cvsserver can be so damn fast	19:48
	← tg left	19:48
	→ tg joined	19:49
divVerent	I now REALLY use it for my dedicated game server	19:49
_Vi	divVerent, <joke>Don't say the word "CVS" on #git!</joke>	19:49
divVerent	to save the HD space for the history	19:49
	_Vi: heeh	19:49
	← charon left	19:49
divVerent	I still wonder HOW that git-cvsserver thing can work at all	19:49
	and there's many things I won't even think about, like pushing using it	19:49
_Vi	divVerent, You may use normal history for code and "git cvsserver" checkout for data. But it's when the data will stabilize.	19:49
	→ ignorabimus joined	19:50
divVerent	also, I wonder why it was made at all	19:50
	why didn't the same people make a git-svnserver instead?	19:50
	cvs is practically dead today	19:50
	_Vi: well	19:50
	in this specific case	19:50
	I never touch any of the checked out files	19:50
	but just pull	19:50
	→ alley_cat joined	19:50
divVerent	I know I could just as well use rsync and excluding the .git directory to keep it current	19:51
	→ charon joined	19:51
divVerent	but I still am amazed how fast this cvsserver thing is	19:51
	takes 7 minutes for what git checks out in 16 minutes, and svn does in 30	19:52
_Vi	divVerent, You can still use submodules and checkout specific commit with cvsserver. But it will be useful only when you need to update submodule much rarer than main code.	19:52
divVerent	checkout specific commit with cvsserver?	19:52
	→ JasonWoof joined	19:52
_Vi	divVerent, (You may write your script instead of "git submodule update")	19:52
divVerent	cvs can't do it	19:52
	so how can cvsserver? :P	19:52
_Vi	divVerent, How do you checkout something with git-cvsserver? (I don't know CVS not git-cvsserver).	19:53
divVerent	like, does cvsserver add a specific syntax that allows specifying git revision hashes?	19:53
	cvs -d:........ checkout reposiotryname	19:53
	and	19:53
	cvs update -r1.2.3.4 filename	19:53
	but cvs has per-file revisions	19:53
	and nonatomic commits	19:53
_Vi	divVerent, Give some real example of checking out something with git-cvsserver.	19:53
divVerent	one of the main disadvantages of it is the lack of any way to ensure consistency	19:53
_Vi	divVerent, "cvs -d:........ checkout reposiotryname" - what's after "-d" and what's "repositoryname"?	19:54
divVerent	http://141.2.16.23/cgi-bin/gitweb.cgi?p=darkplaces.git;a=summary	19:54
	this contains a command	19:54
ignorabimus	nonatomic commits? oh, I'm spoiled obvousely.	19:54
divVerent	except that I typo'd it	19:54
	put the checkout between the two -d options :P	19:54
_Vi	What is "div0-stable"?	19:54
	← jas4711 left	19:54
divVerent	a stable tree of the darkplaces engine	19:54
	→ sockmonk joined	19:54
divVerent	which itself is hosted in a svn	19:54
	it's the branch name, basically	19:54
_Vi	divVerent, Can't it be like "e1d673c476de7c52da87275c3e9403abb5f42c92"?	19:55
divVerent	oh, never tried that	19:55
	but if that works, one has to always do a full new checkout	19:55
	hehe... I see it actually works... "sort of"	19:55
	git-cvsserver creates a new database for that "branch name" :P	19:55
_Vi	divVerent, You have have special "CVS-helper-branch" in the repo.	19:55
divVerent	_Vi: yes	19:56
	the repo was originally cvs	19:56
	then converted to svn	19:56
	and this is a git mirror of it	19:56
	with an added branch	19:56
	← robinr left	19:56
_Vi	divVerent, <joke>Then it'll go reach nirvana</joke>	19:56
	s/go //	19:56
divVerent	hehe	19:56
	well	19:56
	up to now it was always possible to preserve history :P	19:56
	so when switching to git, we'll probably preserve history again	19:57
_Vi	divVerent, "git-svn".	19:57
divVerent	but, git allows to delete these stupid branches without bad side effects :P	19:57
	→ Aldwuin joined	19:57
divVerent	sure, this is a git-svn	19:57
	← bcardarella left	19:57
divVerent	a bare repo that a cron job "git-svn fetch"es into	19:57
	but that cannot commit to svn	19:57
	← Axius_ left	19:57
divVerent	all pushes go into the non-svn branches	19:57
	like the div0-stable	19:57
_Vi	divVerent, ? "git svn dcommit"?	19:57
	← jrmuizel left	19:58
divVerent	_Vi: what will happen if I put that in the cronjob? :P	19:58
	well, currently it's an unidirectional mirror	19:58
_Vi	divVerent, It will commit on the schedule.	19:58
divVerent	although I would like to make it bidirectional	19:58
	→ xlistener joined	19:58
_Vi	divVerent, Let it commit to the special "from-git" branch.	19:58
	→ jrmuizel joined	19:58
_Vi	divVerent, ... then make this branch main thing ...	19:58
	divVerent, ... then completely remove legacy SVN.	19:59
divVerent	actually... I think there is a path to remove legacy svn even without that	19:59
	first, I think we'll work on the git, but only in newly created branches	19:59
	→ Paraselene_ joined	19:59
divVerent	and use dcommit on the clients to commit to svn	19:59
	then, we eliminate the svn once everyone can do that	19:59
	disadvantage of that is the stupid dcommit on the client	20:00
	which I would indeed prefer on the server	20:00
	what is this "from-git" branch?	20:00
	→ anno^da joined	20:00
_Vi	divVerent, When you eliminate the "rebase-genic" SVN it should become shinier.	20:00
	divVerent, A branch in SVN repository.	20:00
divVerent	oh, a normal one	20:00
	→ robinr joined	20:01
divVerent	what use is it to commit to a new branch in svn?	20:01
	← mattions left	20:01
divVerent	then people checking out svn trunk won't get it	20:01
_Vi	divVerent, "git svn branch from-git".	20:01
	divVerent, (Assuming there's "branches" and "tags" and git-svn knows about them)	20:02
divVerent	yes	20:02
	that will then create a new branch on svn side too	20:02
_Vi	divVerent, It will create new branch at svn side. May be also some on git side.	20:03
	→ chalkbag_ joined	20:03
_Vi	divVerent, It will do like "svn copy http://qqq/trunk/ http://qqq/branches/from-git/".	20:04
	← ignorabimus left	20:04
divVerent	sure	20:04
	I just don't see the use of that	20:04
	then svn users commit into trunk	20:04
	and I git-svn fetch, then git pull, and will eventually get it	20:04
_Vi	divVerent, The data from trunk get fetched to Git repository and then dcommited to that branch.	20:04
divVerent	but when I push from git, it'll get dcommit'ed into the git-svn branch	20:04
	then the svn users working on trunk won't get my changes	20:04
_Vi	divVerent, Who merges? Who resolves conflicts between Git users and SVN users?	20:05
divVerent	ah, that is what i was missing	20:05
	okay	20:05
	so basically, we'd have to svn merge on the other side	20:05
	we can't, the svn server does not support mergeinfo	20:05
_Vi	divVerent, SVN users need to get a diff from from-git branch and apply it to main thing.	20:05
divVerent	or can we do this on git side?	20:05
	hm... but when that is done	20:06
	then git will see this as a new changeset	20:06
	and apply it again	20:06
	wouldn't that cause conflicts?	20:06
	← chalkbag left	20:06
_Vi	divVerent, (But then we may get problems on git side when the changes return back from SVN...)	20:06
divVerent	which hell do we end up in if the git repo dcommits to svn trunk?	20:06
_Vi	divVerent, (It MAY consider that "This patch is already applied, skipping", but may be not)	20:06
divVerent	especially not if it needed manual resolution	20:07
_Vi	divVerent, To the trunk? What if conflict?	20:07
divVerent	if conflict, show some sort of error	20:07
	is what I would prefer	20:07
	so the user has to wait for the next git-svn fetch, fix the conflicts, and push again	20:07
	just... not sure how this can be accomplished	20:08
	as the change is already committed, just not pushed yet	20:08
_Vi	divVerent, In any case SVN will generate waves of disruption to Git users.	20:08
divVerent	still, the client can probably still rebase it	20:08
	yes	20:08
	for me it'd suffice if svn can be maitnained as a read-only copy of git for a while	20:08
	many people want to check out	20:08
_Vi	divVerent, Because of "git svn" is "rebase-rebase-rebase-rebase".	20:08
divVerent	but not that many need to commit	20:09
	_Vi: funnily not in my git-svn mirror :P	20:09
	but well	20:09
	it only fetches, works on a bare repository	20:09
	← kain left	20:09
_Vi	divVerent, Bare == No merges.	20:09
divVerent	sure, it's the published repo from that website	20:09
_Vi	divVerent, (except of fast forward)	20:09
divVerent	still works fine for me	20:10
	→ rudi_s joined	20:10
wereHamster	still solving that svn issue?	20:10
divVerent	well, I consider it solved "as good as possible"; but we're talking about ideas to make it better	20:10
	is how I see it	20:10
_Vi	(Increasing the IRC log buffer)	20:10
divVerent	anyway, also just working on the "other trail"	20:11
	and booted up my XP VM	20:11
	and installed tortoisesvn	20:11
	and now tortoisegit	20:11
	let's see how it breaks, if it breaks	20:11
	too many tortoises make the system slow?	20:11
	→ Tricon joined	20:11
_Vi	divVerent, Because of tortoise sounds similar to "turtle".	20:12
	→ martiin joined	20:12
divVerent	doesn't it even mean the same?	20:13
	okay, one conflict between tortoisesvn and tortoisegit I see: both associate with .diff	20:13
	→ brodyberg joined	20:13
tetha	from what I know, on likes to be around and on land, while the other prefers the deep ocean	20:14
divVerent	does not sound serious to me	20:14
tetha	but I can't rembember which ones which	20:14
ifvoid	zeker ls je ook nog wat van oss moet weten	20:14
divVerent	well, I am German, and I am pretty sure both are "Schildkröte" :P	20:14
ifvoid	oeps	20:14
tetha	jep, but on is 'landschildkroete' and one is 'seeschildkroete' :)	20:14
divVerent	ah :)	20:14
tetha	I think however that tortoise is the ocean type	20:14
_Vi	/Considered "turtle" as animal and "tortoise" as computer program before this/	20:15
divVerent	so the tortoise that achilles could not catch up with will never meat the ninja turtles	20:15
	*meet	20:15
	← Adlai left	20:15
divVerent	turtle is on sea, tortoise is on land, says dict.leo.org	20:15
	→ Adlai` joined	20:16
tetha	hehe	20:16
	buyt they might have met, as leonardo travelled to south america	20:16
	;)	20:16
divVerent	actually...	20:17
_Vi	</offtopic>	20:17
divVerent	given how much they are on land and on water	20:17
	shouldn't the turtles be the tortoises then?	20:17
	Never seek for realism in ANYTHING shown on a rectangular screen, or on paper.	20:18
	← bibinou left	20:18
	→ loinclot_ joined	20:18
divVerent	hm...hm...	20:18
	sh-3.2$ git push	20:18
	To ssh://gitosis@141.2.16.23/darkplaces.git	20:18
	← smuf left	20:18
divVerent	! [rejected] master -> master (non-fast forward)	20:18
	I never touched anything in master	20:18
	but master got advanced on remote	20:18
_Vi	divVerent, Just update.	20:18
divVerent	is that normal?	20:18
	sure, did that and it worked	20:18
	just, I never touched master in my checkout	20:18
_Vi	divVerent, Somebody else has touched?	20:19
divVerent	and never will	20:19
	nope	20:19
	it got touched on remote	20:19
	after a git pull, the push did exactly nothing (as it should)	20:19
	basically, this tells me that I should probably have used a more specific push command	20:19
_Vi	divVerent, Somebody else has pushed, hasn't?	20:19
divVerent	as the default seems to try to push everything	20:19
	yes	20:19
	→ johnw joined	20:19
divVerent	someone else has	20:19
	but that is fine by me	20:19
	but I also pulled	20:20
	just, div0-stable was the active branch when I pulled	20:20
	and not master	20:20
	so I did git checkout master, git pull, git checkout div0-stable, git push	20:20
	then it was fine	20:20
	just... do I _have_ to do that?	20:20
	I don't work in master	20:20
	"for every head that exists on the local side" blabla	20:21
	is the default of git push	20:21
	can I make master not exist locally, and not be regenerated on next pull?	20:21
_Vi	divVerent, Personally, I always use fully qualified push and don't use "git pull".	20:21
divVerent	otherwise, "git push origin div0-stable" does what I want, but it's so much to type	20:21
Fissure	you can set push.default to change the behavior	20:21
wereHamster	divVerent: maybe you should set push.default to tracking	20:21
	or something else than the default	20:21
divVerent	ah, that sounds good	20:22
	← Fissure left	20:22
_Vi	divVerent, Just don't use "git pull".	20:22
	divVerent, fetch&merge.	20:22
divVerent	then an unqualified push won't attempt to push master?	20:23
	→ khmarbaise joined	20:23
	loinclot_ → loincloth_	20:24
wereHamster	divVerent: man git-config to see what values you can set push.default to and what they mean	20:24
	→ kkoehne_ joined	20:24
divVerent	ah, good	20:24
	← hyperair left	20:24
	→ hyperair joined	20:25
Gitbot	divVerent: the git-config manpage can be found at http://git.or.cz/man/git-config	20:25
	→ robin joined	20:25
_Vi	divVerent, It is especially useful when you are tracking multiple upstreams (like official linux kernel, zenkernel and "wireless testing" variant)	20:25
	← robinr left	20:25
divVerent	why do I want to git-fetch and merge separately	20:25
	instead of doing pull?	20:25
	robin → Guest98310	20:25
	← rudi_s_ left	20:26
divVerent	that is, what is the difference	20:26
_Vi	divVerent, Do be sure what will it actually do.	20:26
thiago_home	divVerent: git pull = fetch + merge	20:26
	so there isn't a difference	20:26
	→ thm joined	20:26
divVerent	thiago_home: well, the fetch part probably is the default	20:27
	→ ruediger_ joined	20:27
divVerent	and the merge part is probably origin/branchname	20:27
thm	can I teach git-mergetool somehow to call xxdiff with -m ?	20:27
_Vi	thiago_home, The difference is what branches will it process and where will it merge. Is using "git pull" a good idea when tracking multiple upstreams?	20:27
	← jast left	20:28
_Vi	thm, echo -e '#!/bin/sh\nxxdiff -m "%@"' > ~/bin/xxdiffm; chmod +x ~/bin/xxdiffm; # Something like this?	20:29
thiago_home	_Vi: I don't see why it would be a bad idea.	20:29
	→ loinclot_ joined	20:29
	← ruediger left	20:29
divVerent	to me it looks like git-pull does the same as git-fetch with the same arguments, then merges using git-merge HEAD FETCH_HEAD basically...	20:29
_Vi	s/\%/$/	20:30
divVerent	but the fetch is the same as if you'd use git fetch, right?	20:30
	← tvw left	20:30
divVerent	but well, in my specific use case I often don't need to merge anyway	20:30
	as I am the only committer in div0-stable, and I only read from mastere	20:31
_Vi	divVerent, (Still don't know what FETCH_HEAD does mean).	20:31
thm	_Vi: no. that doesn't work.	20:31
	loinclot_ → loincloth__	20:31
_Vi	thm, Has you substituted the wrong "%@" with "$@"?	20:31
thm	yep	20:32
wereHamster	_Vi: when you run 'git fetch origni foo' git doesn't store the branch in origin/foo but in FETCH_HEAD	20:32
_Vi	thm, This script should just call xdiff with "-m" and the rest arguments.	20:32
thm	_Vi: I understand. but that doesn't seem the way git-mergetool accepts a new/modified command.	20:32
	→ jast joined	20:33
_Vi	wereHamster, OK. I usually do just "git fetch origin" without refspec and then use "origin/" branches.	20:33
	thm, The problem is necessity to specify "-m" or something else?	20:33
wereHamster	_Vi: 'git fethc origin' uses the configured refspec, which is usually refs/heads/:refs/remotes/origin/	20:34
thm	_Vi: it says: Unknown merge tool xxdiffm. (but xxdiffm is in the path)	20:34
_Vi	wereHamster, "+refs/heads/:refs/remotes/origin/" to allow forced. I consider this refscec as correct and other ones as strange.	20:35
wereHamster	other ones?	20:35
	→ mmoya joined	20:36
_Vi	_Vi, Like if it not in "remotes" or the directory name not the same as remote name or when it tries to fetch right in "refs/heads".	20:36
	← loincloth left	20:36
_Vi	s/^_Vi/wereHamster/	20:36
	← sudoer left	20:36
	← schmukkk left	20:37
	ruediger_ → ruediger	20:38
_Vi	wereHamster, E.g. after "git fetch xxx" I expect to have all remote "refs/heads" in my "refs/remotes/xxx" (even if not fast-forward). So the typical scheme is "git fetch xxx && git rebase xxx/<branch_name>"	20:39
	→ ctpaterson joined	20:41
	← kkoehne left	20:42
ctpaterson	Apologies if a stupid question; but I'm VPNed into a newly setup git repository, have put my public key into the keygen dir, and my user into the gitosis.conf; when I locally add the remote to my machine and try to do a push; I'm getting prompted for git@servername's password. My key has [email@hidden.address] and the server I'm VPNed into can not resolve mymachine.local -- does this explain why I'm prompted for	20:43
	→ Scorchin joined	20:44
wereHamster	ctpaterson: any comments in the key are irrelevant. What counts is that you can authenticate with the key	20:44
_Vi	ctpaterson, May be put your private key to $HOME/.ssh/id_dsa?	20:44
ctpaterson	I don't think so...this is the key I took from our previous repository server, which was local to our network...copied the conf, copied the keys. The NEW repo is in a land far away, and we VPN to it...and now I'm plagued with being prompted for git's password.	20:45
_Vi	ctpaterson, Do you have the private key?	20:46
	← kristian42 left	20:46
wereHamster	ctpaterson: can could use 'ssh -vvv user@server' to see which keys your client tries to send to the server, or why the server is rejecting the keys	20:46
ctpaterson	Looks like - I've got id_rsa and id_rsa pub in the .ssh dir of my local machine.	20:47
_Vi	ctpaterson, Is that "id_rsa.pub" the same as server knows?	20:47
	← d0k left	20:47
ctpaterson	_Vi: checksums match	20:48
_Vi	ctpaterson, Then use "ssh -v -v -v" to see messages.	20:48
	← loincloth_ left	20:48
ctpaterson	Used -vvv ...many, many messages...going through now, looking for something I understand...	20:48
_Vi	ctpaterson, Look for ones with "id_dsa.pub" and "private key" or "public key".	20:49
ctpaterson	ssh -vvv seems to use my private key... refers to id_rsa, anyway.	20:50
_Vi	ctpaterson, Ans still asks a password?	20:50
cbreak	ssh-add helps?	20:50
ctpaterson	I may have misunderstood here...when I launched ssh - it wasn't in the context of git... Is ssh another way to add a remote?	20:51
_Vi	ctpaterson, Is your private key protected with a passphrase?	20:51
ctpaterson	_Vi: no	20:51
	← alley_cat left	20:51
wereHamster	ctpaterson: git uses ssh, and running ssh directly is a way to diaglose authenetication problems	20:52
ctpaterson	Also; the user I use on the remote machine is not the same username as I use locally. On my machine, I'm me...on the server, I'm an admin.	20:52
cbreak	ssh -u admin otherserver.local	20:52
	no	20:52
	ssh -l admin	20:52
keystr0k	how do I generate a patch file without any git information associated... I use git, but the patch is being submitted to an open source project... I don't know much about this stuff.	20:53
_Vi	keystr0k, "git diff -p"?	20:53
keystr0k	-p... I'll look into that	20:53
	_Vi: thanks	20:53
ctpaterson	cbreak: I'm sorry..."ssh -vvv -l admin@server" ?	20:53
charon	keystr0k: gnu patch will understand the patch if it was generated with default settings, despite the 'diff --git' header	20:54
	→ keex joined	20:54
	← nothingHappens left	20:54
keystr0k	charon: okay... and the commit info should be ignored too?	20:54
charon	indeed, but that also happens transparently	20:54
keystr0k	perfect! thanks charon	20:55
charon	keystr0k: extended git features like renames, binary patches and such cause problems	20:55
	but these aren't on by default	20:55
keystr0k	git diff --relative signup/signup.module	20:55
	→ sgh joined	20:55
	→ alley_cat joined	20:55
keystr0k	looks pretty straightforward... cool beans	20:55
	← ifvoid left	20:56
	→ kristian42 joined	20:56
	Adlai` → Adlai	20:57
ctpaterson	Have I run up against a wall on this?	20:59
	→ ToxicFrog joined	21:00
	→ jas4711 joined	21:00
	→ jumoit joined	21:00
mugwump	ctpaterson: run sshd -d on the server	21:01
	then connect to it	21:01
	that should tell you why the key is being rejected	21:01
_Vi	ctpaterson, Are permissions of "authorized_keys" and parents OK?	21:01
ctpaterson	_Vi: 644 for the admin account for the keys	21:02
	← jas4711 left	21:03
ctpaterson	wugwump: sshd is running under root...is there a log I can look at under its current exec?	21:03
	→ jas4711 joined	21:03
_Vi	ctpaterson, "og-w" for authorized_keys, ".ssh" directory, your home and "/home" directory and "/".	21:03
ctpaterson	_Vi: "authorized_keys"? I don't see that anywhere. In keydir, known_hosts o and g already are denied w permission.	21:05
_Vi	ctpaterson, /var/log/auth.log	21:05
	ctpaterson, Where has you stored your pubkey on server?	21:05
	← squarism left	21:05
	← GeertB left	21:06
	← kraymer left	21:06
_Vi	ctpaterson, Put your pubkey into ~/.ssh/authorized_keys	21:06
ctpaterson	_Vi: gitosis-admin/keydir/	21:06
	_Vi: I don't require a "me" user on the server, do I?	21:07
_Vi	(Still don't know if "gitosis" just some hosting for git or some piece of software)	21:07
	ctpaterson, on server: "cat /etc/ssh/sshd_config \| grep AuthorizedKeysFile"	21:08
ctpaterson	_Vi: gitosis is (according to my book) "a tool for managing a remote git server and its repositories".	21:08
	→ GeertB joined	21:08
	← jrmuizel left	21:09
	→ jrmuizel joined	21:09
ctpaterson	_Vi: sshd_config has %h/.ssh/authorized_keys	21:10
_Vi	ctpaterson, Mine too.	21:10
	ctpaterson, So put your pubkey into $HOME/.ssh/authorized_keys	21:10
	ctpaterson, (Pubkey of the _connecting_ part, not server's one).	21:11
	nonken → nonken_afk	21:11
	→ drizzd joined	21:12
phadviger	hi... when i do a git pull i get the following error: Your configuration specifies to merge the ref '4444-test-branch' from the remote, but no such ref was fetched. i already did a git remote prune origin, what else do i need to do to clean up this error?	21:15
	← reprore left	21:16
	nonken_afk → nonken	21:16
ctpaterson	_Vi: Well, putting my key into the authorized_keys file instead of the one that was there did seem to get me past the authorization...but as a hack - not expandable. Still, I think this led me to the solution. On our former server, it's got a whole mess of keys in that file (one per developer), and in the new server, ours are missing. I think it's about wrestling gitosis to update that file properly.	21:17
_Vi	ctpaterson, What is not expandable?	21:18
ctpaterson	_Vi: Thanks very much for your help	21:18
	→ shruggar joined	21:18
	→ mjf joined	21:18
	→ belak joined	21:18
_Vi	ctpaterson, You can put multiple keys into this file.	21:18
belak	Is there a way to output the current command for a repo? Like so I can get it with bash?	21:18
ctpaterson	_Vi: Oh, I'm sure it is - there are just warnings everywhere about how gitosis generated this file, and I'm not to touch it. One way or another, though, I can get this problem solved now. It's not too hard to figure out how gitosis is creating the file.	21:19
_Vi	belak, What is "current command for the repository"? How you doing it in bash?	21:19
ctpaterson	_Vi: My main problem was I couldn't figure out why my key wasn't accepted...now I know, thanks to you.	21:19
_Vi	ctpaterson, I don't know what relation has Gitosis to SSH and it's keys.	21:19
ianmcorvidae	gitosis manages the .ssh/authorized_keys file of its user for you	21:20
ctpaterson	_Vi: gitosis is supposed to manage all that for me so my life is easier.	21:20
belak	It's just a git pull	21:20
	sorry	21:20
	git fetch	21:20
_Vi	ctpaterson, So, to remove "the hack" you should now get why Gitosis hasn't placed your key to authorized_keys file.	21:21
belak	what about cat .git/refs/heads/master	21:21
	?	21:21
ctpaterson	_Vi: Yup...or just take the one from former server. I prefer to figure out what step I've clearly missed...but I've gone over the procedure a few times now.	21:22
_Vi	belak, Don't know what do you want to see. What is a "command"?	21:22
	← drizzd_ left	21:22
ianmcorvidae	belak: what do you want as a return value from this command?	21:23
	← loincloth__ left	21:23
belak	It's for an update script for the kernel	21:23
ianmcorvidae	that does not answer either question	21:23
	answer the questions we ask or we cannot and will not help you	21:23
belak	I want it to do a git pull and then run the make only if the commit id has changed	21:23
	I was typing.	21:24
	Sorry	21:24
ianmcorvidae	'git rev-parse <revision>' will give you the sha1 of that revision	21:24
_Vi	belak, So you want something like "git pull \| grep Updating && make"?	21:25
ianmcorvidae	if you don't mind doing a pull instead of a fetch you could do a git pull <origin> then see if git rev-parse HEAD has changed	21:25
	what Vi suggests would work as well	21:25
belak	I like seeing the output of the pull/fetch...	21:25
_Vi	belak "git pull \| tee /dev/stderr \| grep Updating && make"?	21:26
belak	Duh. Thanks. Sorry, I'm kinda new to git..	21:26
ianmcorvidae	otherwise you could do something like git fetch origin then check git rev-parse HEAD vs. git rev-parse origin/HEAD	21:26
	if you, like me, think grepping the pull is a bit of a hack :)	21:27
belak	Yeah	21:27
	← mmoya left	21:27
belak	I'm gonna go with ianmcorvidae's idea... comparing the 2 variables	21:27
	Thanks	21:27
_Vi	Why there's no "git pull --porcelain"?	21:27
	→ schierbeck joined	21:28
	← ctpaterson left	21:28
_Vi	(going offline)	21:28
	← _Vi left	21:29
	← Oxf13 left	21:30
	← pklingem2 left	21:30
	→ psoo joined	21:30
Jaxoreth	Why not run make unconditionally?	21:31
belak	Because I don't want to compile the kernel if I don't have to and it's already compiled...	21:31
	← CarlFK2 left	21:31
	→ blaenk joined	21:32
blaenk	hey guys is there a quick way to revert to a previous commit (the immediate previous one)	21:32
	← GeertB left	21:32
blaenk	I just commited something I'd like to 'take back'	21:32
Jaxoreth	Isn't the point of make not to regenerate products that are already up to date?	21:32
	blaenk: man git-reset	21:33
Gitbot	blaenk: the git-reset manpage can be found at http://git.or.cz/man/git-reset	21:33
blaenk	I was specifically looking for a shortcut, I forgot the shortcut for the previous commit	21:33
Jaxoreth	To undo a commit, `git reset --soft HEAD^`	21:33
blaenk	it's something like HEAD^ or something	21:33
	thank you Jaxoreth	21:33
	just a quickie, ^ is one back I believe so I can do ^^ for two back, I forgot what ~ is?	21:34
	← johnw left	21:34
	→ loincloth joined	21:34
Jaxoreth	commit~3 is commit^^^	21:34
	← ntoll left	21:34
blaenk	oh I see so it's like a shortcut, thanks Jaxoreth	21:34
	← ShadeHawk left	21:35
	→ Oxf13 joined	21:36
	← orafu left	21:36
	← belak left	21:36
	← mstrl left	21:37
	← kipras_ left	21:37
	→ agenteo joined	21:37
	← alley_cat left	21:37
	→ orafu joined	21:37
	→ corp joined	21:39
	← krynnotaur left	21:40
corp	hello. i'm trying to get a git post-receive hook to work	21:40
	i have a shell script that just contains: echo $3	21:40
	$3 in theory should contain the ref	21:40
	← maattd left	21:40
corp	but it's just blank	21:40
	→ maattd joined	21:41
	→ cedric_ joined	21:41
	← malumalu left	21:41
	← kpreid_ left	21:42
	→ GeertB joined	21:42
	→ kpreid_ joined	21:42
corp	anyone know why this might be?	21:43
blaenk	Jaxoreth: uh oh, you mind helping me, when I made that commit I wanted to revert, I pushed it too, so now this 'git reset --soft HEAD^' commit wont push	21:43
	gives the rewind error	21:43
	non-fastforward	21:43
	→ niki joined	21:43
	← auslander left	21:44
	→ cedricv joined	21:45
	aroben → aroben\|away	21:46
Altreus	k so the ancientness of the version notwithstanding (1.5.2.4), when I run git rebase --skip it tells me the conflicted file is dirty and it can't continue. What does this mean, in practical terms?	21:47
	← smckinney left	21:47
Altreus	Does it mean that the next patch relies on this patch, so if I skip it it won't work?	21:47
	I thought this might be the case but it doesn't seem to apply	21:48
	← khmarbaise left	21:48
	→ timcharper joined	21:48
	← psoo left	21:48
	← cedric_ left	21:49
Jaxoreth	blaenk: You have two options: either rewrite history, or revert on top of the bad commit	21:49
	→ bambule joined	21:50
Jaxoreth	If nobody else is using your upstream repo, you can safely rewrite history.	21:50
blaenk	how would I go about reverting on top of the bad commit	21:51
	← mw44118-1ork left	21:52
blaenk	you understand what happened right, I did the reset --soft on my end to revert a commit that I had already pushed, so when i tried to push the reset commit, it wouldn't let me	21:53
Ilari	blaenk: First look up the lost commit from 'git reflog', and then 'git reset --soft <commit>'.	21:53
	blaenk: Reset does not make new commits!	21:53
blaenk	weird, cause I did git status and it said a file was modified	21:53
	and I didn't edit anything in the file	21:53
	okay so I get the commit sha of what commit?	21:54
	the commit i tried to push after i did git reset?	21:54
	or HEAD^: Updating HEAD	21:54
Ilari	blaenk: There should be line for reset in reflog, right?	21:54
blaenk	well there's	21:55
	HEAD^: Updating HEAD	21:55
Ilari	blaenk: One before that HEAD^: Updating HEAD line.	21:55
blaenk	nope	21:55
	nothing shows up regarding any reset	21:55
	the one before that is my own commit	21:55
	← keystr0k left	21:55
	→ infoG joined	21:55
	→ botanicus_ joined	21:55
Ilari	blaenk: That 'HEAD^: Updating HEAD' is the line the reset caused.	21:56
infoG	hi	21:56
blaenk	right that's what I figured	21:56
	but you say there's another line I should look for?	21:56
	← rgr left	21:57
Ilari	blaenk: The one immediately after that 'HEAD^: Updating HEAD' line.	21:57
	blaenk: (its like for your commit).	21:57
blaenk	ohhh right	21:57
	right	21:57
	the message for it is 'removing previous changes' so yeah	21:57
	so I should get the sha for that one	21:57
	← cilly left	21:57
blaenk	and again do git reset --soft <the sha>	21:57
Ilari	blaenk: Or the 'HEAD@{nnn}' it shows.	21:58
infoG	i made some changes to a local copy of a project that has since been updated. how do i cancel my changes and mirror the mainline exactly?	21:58
blaenk	and that should put me back at the position my remote origin is at?	21:58
Ilari	infoG: man git-reset (some modes are quite dangerous).	21:58
Gitbot	infoG: the git-reset manpage can be found at http://git.or.cz/man/git-reset	21:58
Ilari	infoG: (and one needs the commits locally, for that presumably 'git remote update').	21:59
infoG	ok thanks	21:59
blaenk	Ilari: ?	22:00
Ilari	infoG: And usually you don't want to cancel your changes, you want to merge them.	22:00
	blaenk: Right.	22:00
infoG	well, my changes have been worked around	22:00
	so they're not needed any more	22:01
blaenk	Ilari: after I do that reset, I should just commit and then push should work?	22:01
Ilari	infoG: There's also rebase (one can just skip commits that have appiled).	22:01
	← fcuk112 left	22:02
Ilari	blaenk: After that reset, you presumably want to do 'git revert HEAD' to undo what the last commit did.	22:02
blaenk	okay	22:02
	← Tuomas left	22:02
blaenk	after the reset, reflog didn't change, this is normal right	22:02
Ilari	blaenk: That reset should have added new entry to reflog.	22:02
	→ Tuomas joined	22:03
blaenk	:/	22:03
Ilari	blaenk: Did the bad commit apper into 'git log'?	22:03
blaenk	I don't understand waht you mean by bad commit, the commit that's not letting me push?	22:03
	yeah it's in git log right now	22:03
	← cedricv left	22:03
blaenk	look watch	22:03
	the weird thing is that	22:03
	now the original commit is not there, the commit that I wanted to revert	22:04
	you know what I mean?	22:04
	eh this is confusing, watch, I originally made a commit, decided I didnt wnat it so I did the reset, but the reset then didnt let me push, the ORIGINAL commit at the beginning of this sentence is the one that is no longer there in my git log	22:04
Ilari	blaenk: Reflog entries don't disappear automatically except by expiring (and the expiry time is quite long).	22:04
	→ parolang2 joined	22:04
	→ onigiri_ joined	22:05
blaenk	no i mean git log	22:05
	it's no longer in git log	22:05
	so what im saying is, my end now looks fine it's just that i cant push	22:05
	→ gitte joined	22:05
Ilari	blaenk: Push will be broken until one gets it back there.	22:05
blaenk	I kind of feel like just re-cloning	22:05
	oh okay so I should just reclone	22:06
Ilari	blaenk: Can you pastebin the reflog output so I can look?	22:06
blaenk	and then do the changes manually	22:06
	sure thing	22:06
	hyperair groans. the kernel sure has a lot of objects	22:06
timcharper	blaenk: did you push the commit you decided you didn't want ?	22:06
Ilari	timcharper: Yes, they did.	22:07
blaenk	yeah that's what I said from the beginning, my mistake I know I know	22:07
timcharper	(I'm assuming that you did)	22:07
	okay, has anybody else pulled that commit?	22:07
hyperair	1318016 390000 (vsz followed by rsz for git repack)	22:07
blaenk	but I mean, I could've just pushed a new commit where I manually reverted the changes by editing the file, but I figured there was a 'git shortcut' for that	22:07
	→ krynnotaur joined	22:07
	← BigMack83 left	22:07
blaenk	which is when someone told me to do git reset --soft HEAD^	22:07
timcharper	blaenk: has someone pulled it ?	22:07
Ilari	blaenk: That's ok if it wasn't pushed, but it was...	22:07
	← univate left	22:08
blaenk	after that, git status showed a file was modified (not me) and then I commited it and when I tried to push it gave the non fast worward one	22:08
	no one has pulled it I'm the only one on this repos	22:08
	→ agile joined	22:08
timcharper	Right, because git reset without a --hard preserves the working directory in the state.	22:08
Ilari	blaenk: Well 'git push -f <remote> <branc>	22:08
	← shruggar left	22:08
Ilari	?	22:08
	← capndiesel left	22:08
blaenk	I can just reclone the repository from the origin (github) and manually revert the changes by editing the file but I figured there would be a fix	22:08
	I guess I should just do that, reclone, I don't want to get things messy	22:08
timcharper	blaenk: you can do this: git reset --hard origin/master	22:08
blaenk	http://pastie.org/769404 reflog	22:08
Ilari	blaenk: No, you shouldn't.	22:09
timcharper	that will be kind of the same thing as a reclone	22:09
blaenk	reset --hard origin/master sounds right, should I?	22:09
	→ univate joined	22:09
timcharper	yes,	22:09
Ilari	blaenk: 'git reset --soft HEAD@{2}'.	22:09
	blaenk: Watch out: reset --hard is _destructive_.	22:10
timcharper	yes, probably best to run a git stash first	22:10
blaenk	okay so I'll do what you say llari then what	22:10
	← lucsky left	22:10
timcharper	(if you use --hard)	22:10
Ilari	blaenk: Even worse idea.	22:10
blaenk	what?	22:10
Ilari	blaenk: Oh.	22:11
	blaenk: After soft reset, presumably 'git rever HEAD'.	22:11
	*revert	22:11
blaenk	okay hope this works	22:11
	← RobertLaptop left	22:11
Ilari	blaenk: Well, except that local changes can come to mess things up...	22:11
	← rseifert left	22:11
	→ anjen1 joined	22:11
	← name left	22:12
blaenk	okay the revert correctly reverted the tag styling, llari	22:12
	(annoying how I can't tell what characters are in your name to tab complete haha)	22:12
Ilari	blaenk: Use serif font.	22:12
blaenk	so should I try to push now llari?	22:12
Altreus	suggest changing fonts	22:12
blaenk	I like this font	22:12
Altreus	It's an i or an l isn't it really	22:13
blaenk	should I try to push now?	22:13
Altreus	Try both	22:13
Ilari	blaenk: Yes. Push should work now.	22:13
	← jrmuizel left	22:13
	→ crazy3k joined	22:13
blaenk	YES IT WORKED!	22:13
	you're the best Ilari :)	22:13
	did the previous reset you told me help to do this or was it irrelevant	22:13
	the one where I did git reset --soft 'commit after HEAD^ in reflog'	22:13
	→ sourcemaker joined	22:14
Ilari	blaenk: The proper way would have been either to reset + force-push or revert + push.	22:14
	← mbroeker left	22:14
Ilari	blaenk: The first way tends to mess others up.	22:14
blaenk	force push sounds messy	22:14
anjen1	newbie question: How often do you do commits? Or better, how often *should one do commits? Looking for "best practice" conept as a newbie	22:14
blaenk	yeah I know the proper way but I mean from what you said	22:14
	→ Dr1Horrible joined	22:14
blaenk	thanks Ilari	22:14
Dr1Horrible	I want to ignore everything except a couple of files or directorys, so I have * in gitignore with below that a list of files and directories preceded by a !, still it doesnt unignore my directories	22:15
timcharper	does Emacs have any hooks that you can tie into that get called when the window is focused?	22:15
	← froschi left	22:17
Ilari	anjen1: Quite often (try to get logical chunks). The staging area is quite useful for that. One can later split/merge/edit commits.	22:17
timcharper	Dr1Horrible: according to the man page for gitignore, ! Overrides any previous pattern	22:18
	have you tried #git? They are very helpful folks over there.	22:18
Dr1Horrible	timcharper: It does for files, but not for directories	22:19
	Isnt this #git?	22:19
Ilari	timcharper: Err... This is #git. :-)	22:19
anjen1	Ilari, thanks. That's what I had in mind, just wasn't sure if it was considered bad form to have many commits. I figure since I'm kinda new, when I do something right I want to know I can get back there if a fudg womething up later...	22:19
	*something	22:19
timcharper	oh, it is? well, em, why yes it is. as you were.	22:19
Ilari	anjen1: I usually don't even compile-test what I make, compiling the final result and backporting fixes.	22:20
anjen1	ooh...did I ask in the wrong forum?	22:20
	ilari, understood.	22:20
timcharper	I was wondering why there is so many git questions going on in #Emacs.	22:20
	^is^were	22:20
anjen1	ah...once the conv starts in a forum, it kinda goes until someone resets it, if you kjnow what I mean...off topic banter	22:21
	← gitte left	22:21
Ilari	anjen1: And besides, commits are fast.	22:21
	← Guest98310 left	22:21
anjen1	ilari, gotcha. so do you also push them with the same frquency?	22:22
	*push them to git that is	22:22
Ilari	anjen1: No. Test first.	22:22
anjen1	ilari...understood...will do. good advise	22:22
Ilari	anjen1: Fixing mistakes before pushing is much easeier than after pushing.	22:22
anjen1	ilari, btw, everryhting gets teted before I commit	22:22
	*everything gets tested	22:23
Ilari	anjen1: Yeah. Good to test it before commit (it needs testing before pushing anyway).	22:23
	→ jrmuizel joined	22:23
	← LiamH left	22:23
anjen1	ilari, I'm wondering this: 1. I write/modify code. 2. I make sure it works and doesn't mess things up. 3 commit completed subtask...4 push	22:24
	ilari, so is that acceptable method?^	22:25
Ilari	anjen1: Commits can group into larger groups that are pushed at once.	22:25
	← crazy3k_ left	22:25
Ilari	anjen1: Such as 'remote helper extensions' for git itself, which was group of 8 commits.	22:25
	← loincloth left	22:25
anjen1	Ilari, beautiful. the exact answer I was hoping for. I *was wondering if multiple commits would push without showing history, but I suppose that's sillly...wouldn't make sense	22:26
	← DavidKlein left	22:26
	→ Lisimba joined	22:26
Dr1Horrible	Does no one know how to unignore directorys?	22:26
wereHamster	!directory/	22:26
Dr1Horrible	wereHamster: That doesnt work:S	22:26
warthog9	ShadeHawk: ok I'm starting ti import your changes into my tree for review, sorry this has taken so long	22:27
wereHamster	Dr1Horrible: maybe it's easier to add just '*' to gitignore and explicitely track the few files you want	22:27
	→ schmukkk joined	22:27
Ilari	anjen1: Individual branches are updated atomically in push. So fetcher never sees intermediate states for individual refs.	22:28
Dr1Horrible	But it's my ~ amd tjem tjere will be files in there that i dont want tracked	22:28
parolang2	Sorry guys, I'm new to get. I have two repositories that mirror the same data on two machines. I git fetch it from the other machine but it doesn't update the working tree. The git fetch seems to work, but what do I need to do to update the working tree?	22:28
	→ makeworkeasy joined	22:28
parolang2	Err, I'm new to *git :)	22:29
anjen1	ilari. Understood. but intermediate commit history is sitill recorded, yes?	22:29
	← orafu left	22:29
wereHamster	parolang2: git merge (or git pull, as pull = fetch+merge)	22:29
Ilari	anjen1: Yes.	22:29
anjen1	ilari ttyy	22:29
	→ orafu joined	22:29
anjen1	mt	22:29
parolang2	wereHamster: Okay, thanks	22:29
anjen1	ilari, thanks for the lesson. I finally get it...or at least this part ;)	22:30
parolang2	wereHamster: Hmm...didn't update. I'll try to figure it out.	22:31
wereHamster	parolang2: what exactly did you run?	22:31
Ilari	anjen1: Sometimes groups get grouped as well. I have seen >100 commit "groups" (at least one gets that impression looking at commit timestamps, which span about 7 seconds).	22:31
	← Ratler left	22:33
	→ Ratler joined	22:33
parolang2	wereHamster: I did git pull from my other machine.	22:33
wereHamster	parolang2: no error message?	22:34
parolang2	wereHamster: no	22:35
wereHamster	probably your branch tracking is set up wrong	22:35
	git doesn't fetch or merge the correct branch\	22:35
anjen1	ilari, not sure what a commit group is yet	22:35
parolang2	wereHamster: Okay, that sounds like it could be it. I'm reading the user manual :)	22:35
Dr1Horrible	I now have every directory unignored except the ones with only subdirectorys	22:36
	← intripoon left	22:36
	← madewokherd left	22:36
	→ _Vi joined	22:39
	← infoG left	22:41
Ilari	anjen1: Commits that are pushed/sent/committed at once. Not really recorded.	22:42
	→ amerine joined	22:43
anjen1	ilari. ok. gotcha. I'll do a few minor commits and push to see for myself. Prolly a good thing to do for experience anyway.	22:43
	← tedoc2000 left	22:46
	aroben\|away → aroben	22:46
bentob0x	is there a way that when you pull a set of patches into master, that instead of having one big merge patch, it goes through all the patches one by one and stops at every conflicts (a bit like a git-rebase for instance)	22:46
	→ mmoya joined	22:46
	→ tedoc2000 joined	22:47
	← Dr1Horrible left	22:48
	→ drewolson joined	22:48
	→ project2501a joined	22:48
	← notjohn left	22:48
	→ Nazcafan joined	22:50
Nazcafan	hello	22:50
_Vi	bentob0x, "git pull --rebase"?	22:50
Nazcafan	I made a copy of my $HOME on an external drive and back	22:51
Ilari	bentob0x: No easy way (but I have done that once).	22:52
Nazcafan	but now when I git status in ~/my_git_local_repo it says: fou@leila:~/publis$ git status	22:52
	fatal: Not a git repository (or any of the parent directories): .git	22:52
	I don't know what went wrong	22:52
_Vi	Nazcafan, Have you copied the ".git" directory?	22:53
Nazcafan	_Vi, it seems so, I can see it when I ls -a	22:53
_Vi	Nazcafan, Is there "HEAD" inside it? "objects", "refs"?	22:54
	Nazcafan, What's in "HEAD"?	22:54
	\|Rolenun\| → Rolenun	22:55
Nazcafan	_Vi, there is no HEAD	22:56
	_Vi, here are the contents: http://paste.debian.net/55842/	22:56
	← pantsman left	22:56
	← __iron left	22:56
Nazcafan	there is a "head" though (small case)	22:56
_Vi	Nazcafan, Rename it to uppercase one.	22:57
	← priidu left	22:57
Nazcafan	oh?	22:57
_Vi	Nazcafan, Now works?	22:57
	→ eletuchy joined	22:57
_Vi	Nazcafan, "mv .git/head .git/HEAD"	22:57
	Nazcafan, Is it Windoze?	22:57
Nazcafan	_Vi, worked	22:58
	no, it is a debian distro	22:58
	← hendry left	22:58
Nazcafan	but I used a fat32 harddrive as a mean of backup	22:58
teuf	:win éç	22:58
Nazcafan	_Vi, how come the case was wrong?	22:58
_Vi	Nazcafan, Case-insensitive FAT32.	22:59
	← maattd left	23:00
Nazcafan	_Vi, however, I have some folders for which the capital case was saved	23:00
_Vi	Nazcafan, I also have all files lowercase on my FAT32.	23:00
Nazcafan	s/saved/preserved	23:00
	→ RandalSchwartz joined	23:00
_Vi	Nazcafan, They are >8 characters in filename.	23:01
	Nazcafan, So LFN used.	23:01
	← martiin left	23:01
	→ Chillance joined	23:01
_Vi	Nazcafan, There's special extension of FAT to recognize something vs Something vs SOMETHING for short filenames. It may be unimplemented for your vfat driver.	23:02
	← sockmonk left	23:02
Nazcafan	_Vi, ok	23:02
	_Vi, thanks for your help	23:02
	there seems to be a problem, though	23:03
	aroben → aroben\|meeting	23:03
	← jrmuizel left	23:04
	→ RobertLaptop joined	23:05
	← TeckniX left	23:06
	→ EricInBNE joined	23:06
	← drizzd left	23:06
Nazcafan	_Vi, git says almost all my files have been modified	23:07
	_Vi, which is wrong	23:07
_Vi	Nazcafan, "git reset --hard"?	23:07
	← lresende left	23:07
_Vi	Nazcafan, Do you have some useful changes?	23:07
	Nazcafan, "git diff"?	23:07
Nazcafan	_Vi, wait, some files have Really been modified	23:07
_Vi	Nazcafan, What about "git diff"?	23:08
	← eletuchy left	23:09
	→ WALoeIII joined	23:09
Nazcafan	ahaa	23:10
	all the diffs show a mode difference	23:10
_Vi	Nazcafan, +x everywhere?	23:11
Nazcafan	_Vi, you are right	23:11
	bloody vfat?	23:11
_Vi	Nazcafan, Just "find . -type f \| xargs chmod -x"?	23:11
	← ixti left	23:11
_Vi	Nazcafan, Now I use vfats only as boot partitions. There are GRUBs and kernel images there.	23:12
	Nazcafan, (And also on removable drives)	23:12
	→ giallu joined	23:13
Nazcafan	_Vi, it is a removable drive, actually	23:13
_Vi	Nazcafan, Why to you need complete working copy of repository on removable medium?	23:13
	→ drizzd joined	23:14
Nazcafan	_Vi, I didn't ask myself that, I just backed up everything in $HOME before installing another operating system	23:14
	nonken → nonken_afk	23:15
	← corp left	23:15
	← aresnick left	23:16
	← mnyon left	23:16
	→ tvw joined	23:17
GSaldana	i need some help setting up gitweb. I get "403 forbidden - No projects found" on gitweb's main page	23:18
	← drewolson left	23:18
	← mmoya left	23:20
	← russell_h left	23:21
	← philips left	23:21
	← jlilly left	23:22
	← tedoc2000 left	23:23
	← corni left	23:24
	→ tedoc2000 joined	23:24
	← magcius left	23:25
	→ magcius joined	23:25
	← stamina left	23:26
Nazcafan	_Vi, thanks for your help, everything seems to be on rails for now	23:27
	→ fualo joined	23:29
	→ eletuchy joined	23:35
	→ spaceonline joined	23:35
	→ unreal_ joined	23:36
	← unreal left	23:36
	← anno^da left	23:37
Nazcafan	_Vi, thanks again, good night	23:37
	← Nazcafan left	23:37
	← ehabkost_ left	23:39
	→ rkirti joined	23:39
	→ cnk_ joined	23:39
	→ cloudhead joined	23:40
rtyler	any tig users around?	23:40
	← onigiri_ left	23:41
	→ corp joined	23:42
cnk_	Is there someplace other the .gitmodules and .git/config that stores references to submodules? I have some differences in submodules on different branches of my code. I don't have any trouble checking out my submodules if I create a tracking branch and switch to it before I do my submodule init/update. But capistrano does a 'git reset --hard $TAG' and then the submodule init. And that method is generating complaints "No submodule mapping found in .gitmodules for	23:43
	path..." for one of the submodules I removed	23:43
	← bantic left	23:43
	→ onigiri_ joined	23:43
_Vi	cnk_, There are also commit_ids in trees.	23:43
	unreal_ → unreal	23:44
cnk_	_Vi and how are those related to the submodule init process?	23:44
_Vi	cnk_, The commit id to be checked out is taken from a tree.	23:45
cnk_	Ok.	23:45
	← drizzd left	23:45
cnk_	Actually what I think I don't understand is the difference between 'git reset --hard' and creating a branch	23:46
	← orafu left	23:46
mugwump	submodule checkouts don't behave in general like other files when changing branches and resetting	23:46
	→ orafu joined	23:46
mugwump	this is a longstanding deficiency in the	23:46
	implementation	23:46
	patches welcome :)	23:46
_Vi	cnk_, Try to emulate submodule checkout by hand.	23:46
cnk_	mugwump: has there been any recent work on submodules? Trying to figure out if just upgrading my version of git would make it magically disappear	23:47
mugwump	no idea, check the changelog	23:47
cnk_	I can reproduce the Capistrano problem by hand	23:47
	← altrux left	23:48
cnk_	I guess the most productive thing right now is to dig around in the Capistrano code and see if it would be easy to make it do its branch switching the way I have always done it by hand	23:48
	m-takagi_ → m-takagi	23:48
mugwump	you can fake it with hooks	23:49
	there should be both pre-checkout and post-checkout hooks in newer gits	23:49
	→ adaran joined	23:49
	→ tfarina joined	23:49
mugwump	that's probably better, and then submit your hacks to the list for discussion, maybe it will prompt someone to fix it	23:49
adaran	if there are two proceses trying to access a git repository, one will fail trying to lock it using the index.lock file. is there a way to tell git to wait until the lock becomes available instead of exiting with an error?	23:50
	← aroben\|meeting left	23:50
cnk_	mugwump: would love to but I don't know how to do enough digging to find the source of the confusion. But thanks for the info. I think the correct solution for my current problem is get stuff deployed so that my new code can move to "master" and not trigger the issue.	23:51
	→ cedricv joined	23:51
_Vi	adaran, (Hack: make FUSE filesystem which hangs accessing task instead of reporting already existing file)	23:51
	← cedricv left	23:52
adaran	_Vi, possible, but a lot of work. i'm lazy!	23:52
_Vi	cnk_, Tell you how to do submodule update by hand?	23:52
adaran	_Vi, though i've got no real solution other then the "spin for a bit, retry and hope for the best" approach	23:53
cnk_	git clone; git branch --track bname origin/bname; git checkout bname; git submodule init	23:53
	works great	23:53
_Vi	cnk_, It worked? Than OK. (I mean emulating "submodule update" with low-level commands).	23:53
cnk_	I think what is triggering the error is capistrano's use of git reset --hard	23:54
	→ Pupeno joined	23:54
_Vi	cnk_, Should "git reset --hard" ever touch submodule's directory?	23:55
cnk_	I don't know	23:55
_Vi	cnk_, I thought it doesn't touch submodules except of with "git submodule" command.	23:55
cnk_	I guess what I should start with is looking into where git submodule init gets its information	23:55
	← parolang2 left	23:56
	→ cedricv joined	23:56
_Vi	cnk_, And AFAIK you need "git submodule update --init" to recheckout a submodule.	23:56
cnk_	I had assumed it just read the .gitmodules file - which in both cases is fine and does not mention the removed submodule	23:56
	this is a new clone, so I am not rechecking out anything	23:56
_Vi	cnk_, (and "git submodule update --init" is universal command both for first and subsequent checkouts_	23:56
cnk_	it is the submodule init that is erroring	23:57
	← sgh left	23:57
	← Mezner left	23:57
_Vi	cnk_, Do you still has a problem to experiment with or everything's fine now?	23:57
cnk_	I still have the problem and can reproduce at will	23:58

Logs Search ←Prev date Next date→

Channels Documentation