IRCloggy #git 2007-08-13

Provider of IRC logs since 2005.

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

	2007-08-13
	→ Romster joined	00:08
	→ Fhoxh joined	00:17
	← doublec left	00:21
	← ferdy left	00:36
	← jeffpc left	01:20
	→ stouset joined	01:29
	→ nzkoz joined	01:31
	← amf left	01:34
	← _spanner_ left	01:45
	→ nashdj joined	01:53
	← aroben left	01:54
	← orospakr left	02:00
	→ vial_afk joined	02:09
	→ doublec joined	02:12
	← vial_afk left	02:13
	dwmw2_gone → dwmw2_PVG	02:21
	← jasam left	02:34
	→ com4 joined	02:34
	→ bdiego joined	02:37
	→ Beket joined	03:00
com4	How can I pull updates from my origin in a bare repository?	03:05
	cd /	03:06
Beket	git branch --track mylocalbranch origin/branchX ?	03:06
com4	Won't that track a specific mirror?	03:08
	I'm new to git	03:08
Beket	so am I, perhaps the more advanced users may shed some light	03:08
com4	but what I'm trying to do is have my repositories that I'll use over ssh, but I would like to publish them over http for others to pull	03:08
	so i created my personal copy, then i git clone --bare /home/..../project	03:09
	→ KirinDave_ joined	03:12
	KirinDave_ → KirinDave	03:13
Tv	com4: you don't normally pull to bare repos	03:46
	→ aroben joined	03:46
Tv	com4: pull somewhere else and push to the bare repo	03:46
	← aroben left	03:46
	→ aroben joined	03:47
Tv	com4: or just push directly, if you control the source repository	03:47
com4	Tv: Ok... that's what I ended up doing :P	03:47
	Tv: out of curiosity, can I push a branch to the public repository?	03:48
Tv	if you have write access, why not	03:48
com4	hah, thanks	03:48
	← kylem left	03:58
	← nzkoz left	03:59
	← bdiego left	04:07
	→ janm joined	04:13
	→ robinr joined	04:19
	← kanru left	04:19
	← fhobia_ left	04:22
	← rkaway1 left	04:27
	→ rkaway1 joined	04:29
	← Fhoxh left	04:30
Beket	Suppose I create a local branch and I want to push it to a remote repository. Any hints? Thanks	04:40
	← Rednaxela left	04:42
Beket	Ah I found it, thanks anyway :)	04:55
	→ syscrash_ joined	04:56
	← syscrash left	05:02
	← madewokherd left	05:03
	← nink\|gone left	05:08
	← PugMajere left	05:08
	→ PugMajere joined	05:09
	→ nink\|gone joined	05:09
	→ lyakh joined	05:20
	← spuk left	05:23
	→ robewald\|home joined	05:28
	← robin left	05:32
	dwmw2_PVG → dwmw2_gone	05:34
	← aroben left	05:40
	← doublec left	05:45
	→ doublec joined	05:45
	→ devogon joined	05:47
	← lyakh left	05:56
	→ meandtheshell joined	06:02
	← meandtheshell left	06:02
	→ kanru joined	06:03
	← dduncan left	06:12
	→ meandtheshell joined	06:17
	← KirinDave left	06:18
	→ dkagedal joined	06:22
	→ bentob0x joined	06:27
	← robinr left	06:30
	→ dkagedal_ joined	06:48
	← dkagedal left	06:50
	→ Dodji joined	06:53
	← hliusv561 left	06:56
	← doublec left	06:57
	→ evilchelu joined	07:06
	→ KirinDave joined	07:06
	→ reuss_ joined	07:13
reuss_	how do i pull from a remote tracking branch? .. "git-pull origin refs/remotes/CVS/HEAD" works, but is there a way where you do not have to specify it with refs/.... ? .. like git-pull origin CVS/HEAD?...	07:15
	← dkagedal_ left	07:15
	→ elmex joined	07:17
	← reuss_ left	07:24
	← matled left	07:25
	→ matled joined	07:25
	← LotR left	07:25
	→ bdowning joined	07:35
	← bdowning_ left	07:35
	← KirinDave left	07:39
	← Romster left	07:43
	→ Romster joined	07:44
	→ mgrimes joined	07:49
	→ branstrom_ joined	07:51
	← branstrom left	07:51
	← johan-s_ left	07:55
	→ LotR joined	07:56
	← kanru left	07:57
	→ kanru joined	07:58
	← janm left	08:00
	→ hliusv561 joined	08:03
	→ dkagedal_ joined	08:27
	→ johan-s joined	08:41
	→ reuss_ joined	08:49
	→ aruiz joined	08:58
	← mgrimes left	08:58
	← elmex left	09:03
	→ elmex joined	09:05
	← johan-s left	09:06
	← rkaway1 left	09:13
	→ rkaway1 joined	09:14
	→ nikodemus joined	09:19
	→ johan-s joined	09:23
	← johan-s left	09:31
	→ johan-s joined	09:31
	→ haskill joined	09:35
	→ Roomster joined	09:43
	→ Oeje1 joined	09:45
	← palmcron left	09:45
	← bentob0x left	09:48
	← Romster left	09:48
	→ bentob0x joined	09:50
	→ ferdy joined	09:54
	← Oeje1 left	10:06
	← johan-s left	10:06
	→ johan-s joined	10:10
Beket	I pushed my local branch back to mainline, but nothing appeared in log. Is it because of a fast forward merge? How do I fix it ?	10:27
	→ whee joined	10:31
hliusv561	Beket: Do those commits appear in log of branch you pushed them to?	10:33
Beket	hliusv561, yes they do appear in branche's log	10:34
hliusv561	Beket: Pushes (unless forced) are always fast forward. And even when forced, they never create merges.	10:35
Beket	Thanks hliusv561, I didn't know that. I guess next I'll merge locally and the push	10:35
	time*	10:36
hliusv561	Beket: The rule is: You have to have working tree (and assocated index) to perform merges.	10:37
	Beket: Because otherwise there is absolutely no way to handle conflicts.	10:38
Beket	One is for sure: I need more reading	10:38
	← Sho_ left	10:42
	→ tanuki joined	10:42
	→ Sho_ joined	10:48
	← Roomster left	10:50
Beket	hm, it is not possible to delete a remote origin/branch	10:51
	→ Roomster joined	10:51
Beket	just exclude it from your listing with git-branch -d -r origin/branch, right ?	10:51
evilchelu	hi, i have a question about stuff between two svn repositories	10:56
	i wrote a decent description of it here: http://pastie.caboo.se/87168	10:56
	erm, s/decent/bit more decent/ :)	10:57
	→ chris2 joined	11:11
fultilt	Beket: You can delete it from the remote. See "man git-push". Basically, "git push origin :mybranch".	11:11
	← fultilt left	11:11
	→ jasam joined	11:14
	Roomster → Romster	11:16
jasam	hi, "git diff" shows differences between two files (as "diff" command does") when called from the tests for Git, how can it achieve this? I cannot find how that behaviour is changed, or a reference in documentation, it is a mistery for me :-/	11:32
	s/mist/myst/	11:36
	← cortana left	11:39
	→ cortana joined	11:40
	← stouset left	11:51
	→ integral joined	11:52
	← jcollie left	11:52
	← integral left	11:53
	→ stouset joined	11:53
	← stouset left	11:55
	→ stouset joined	11:56
	→ henr_k joined	12:08
	→ dorileo joined	12:14
	→ lcapitulino joined	12:17
Beket	Suppose I changed all my file permissions in my source tree, how do I push these changes remotely ? Do I have ti git add * and then git commit ?	12:17
	← pdmef left	12:19
	→ pdmef joined	12:20
	→ mountie joined	12:25
hliusv561	Beket: If you don't have done other changes, you can do "git add -u" and then "git commit". But beware that git only stores the executable bit.	12:25
Beket	double thanks hliusv561 :)	12:26
	→ jcollie joined	12:31
	← jasam left	12:33
	← Beket left	12:33
	→ Yuuhi joined	12:38
	← stouset left	12:42
reuss_	... git-pull origin refs/remotes/CVS/HEAD ... works for me .. but is it the "right" way to pull from at remote branch? .. isnt there something like git-pull origin CVS/HEAD ? ...	12:47
	← binary42 left	12:57
	← johan-s left	13:10
	→ johan-s joined	13:13
	→ bdiego joined	13:26
	→ schofield joined	13:34
	← jcollie left	13:36
	← schofield left	13:37
	→ stouset joined	13:41
	← Sho_ left	13:41
	→ kylem joined	13:45
	→ spuk- joined	13:49
	← tanuki left	13:51
	→ d0k joined	13:58
	← rtmfd left	14:04
	→ rtmfd joined	14:04
	→ sirPalook joined	14:08
	← lcapitulino left	14:16
	→ lcapitulino joined	14:17
	→ KirinDave joined	14:24
	→ orospakr joined	14:30
	← sirPalook left	14:38
	← z3ro left	14:38
	→ fredrik joined	15:06
	← dkagedal_ left	15:14
	→ goloo joined	15:16
	← fredrik left	15:18
	← dorileo left	15:22
	← branstrom_ left	15:24
	→ branstrom joined	15:31
	→ Oeje1 joined	15:31
	← reuss_ left	15:33
mnemoc	hi, how can i escape a ; on the value of a config variable?	15:36
	→ maurizio joined	15:39
	← nashdj left	15:39
	→ binary42 joined	15:49
	← ft left	15:54
	syscrash_ → syscrash	15:56
	← orospakr left	15:57
	→ alley_cat joined	15:58
	→ orospakr joined	15:59
	← KirinDave left	16:00
	← johan-s left	16:02
	← binary42 left	16:07
	→ kumbayo joined	16:12
	→ binary42 joined	16:17
	← kanru left	16:24
	→ fhobia joined	16:28
	→ KirinDave joined	16:32
	→ bla joined	16:41
bla	Hi.	16:41
	→ strangy joined	16:44
	← fhobia left	16:45
	← aruiz left	16:45
bla	Is it possible to share git repos via plain http? Or there REALLY has to be special daemon for it?	16:46
	→ kanru joined	16:47
bdiego	bla, you only need apache (with DAV enable) to use http	16:52
loops	bla, you can publish a git repo via http, you only need the daemon if you want the benefits of publishing a repo using the native Git protocol	16:52
bla	Hm.	16:56
	bdiego, there's already a DAV enabled for SVN support. (I don't want to place there another module...)	16:57
	I'm just unable to git clone or pull from http:// address.	16:57
	loops, you mean I can e.g. tar git repos on http, then download it and use, yes?	16:57
loops	bla, you only need DAV support to allow updating of the repo. if you just want to publish a repo, it's not needed	16:57
bla	So copying repos on http and calling git clone http:// should work.	16:58
loops	bla, no. git supports cloning over http. git clone http://yoursite/repo	16:58
bla	Can you try calling: git clone http://thera.be/my_public/my_projects/HPPB/	16:58
	Please?	16:58
	Maybe i've just f* something up. ;)	16:59
	It should be bare .git directory?	16:59
	Yep.	16:59
	Right.	16:59
	Works.	16:59
	Thanks. ;)	16:59
loops	bla, make suer you run git update-serer-info after each update to that repo	16:59
	bla, chmod +x .git/hooks/post-update should do that automatically for you each time	17:00
bla	If that apache already serves SVN repos with DAV shouldn't it be possible to push/pull from GIT repos there?	17:00
	→ jackbravo joined	17:00
bla	Some paths should be specified I guess.	17:00
loops	i suspect you could make it work, yeah	17:01
bla	I see there example, but it was commented out for SVN. I though that GIT would use like SVN it's own module.	17:01
	← rlb3_work left	17:01
bla	Cool that it doesnt. ;d	17:01
loops	you might want to take a peek inside of Documentation/howto/setup-git-server-over-http.txt, might have something to help	17:02
	← strangy left	17:02
	→ robin joined	17:03
loops	bla, for whatever it's worth cloning over the native git protocol (which means you need to run the git daemon) has some nice benefits. If it's possible, it's usually worth the effort to run it.	17:04
	→ dorileo joined	17:11
	→ aroben joined	17:12
	→ praka joined	17:13
	← KiBi left	17:14
bla	;-)	17:19
	It would be possible.	17:19
	For 'group coding' I usually create SSH account with git repos to which everybody can push.	17:20
	→ Beket joined	17:20
bla	I just wanted to find a method of sharing GIT repos without giving someone more access.	17:20
Beket	mob:)	17:21
	→ _bdiego joined	17:21
	← maurizio left	17:21
	← bdiego left	17:26
	→ GyrosGeier joined	17:28
	→ johan-s joined	17:29
	← Beket left	17:32
LotR	what are the benefits of git-daemon over serving over http?	17:32
vmiklos	being able to push, shallow clones, less network traffic, etc	17:34
GyrosGeier	LotR, the git daemon creates packf containing just the objects people need on the fly	17:34
	vmiklos, git-daemon can push?	17:34
	I thought that was ssh only still	17:34
vmiklos	afaik yes	17:34
	also http+dav can	17:34
	→ Aeternus_ joined	17:35
	→ DmitryK joined	17:37
DmitryK	Is repo.or.cz down?	17:37
kergoth	is there a way to dump all refs from which a given commit is reachable from? git-name-rev seems to do it, but it only shows the most recent tag.. what i'm more often looking for is the first tag it was reachable from, so i know when it went in, in the versioning history	17:37
	s/ from\?//	17:37
lcapitulino	DmitryK: I can't reach it either :(	17:39
	lcapitulino fears	17:39
DmitryK	pasky: Hi, is repo.or.cz down?	17:40
	→ lyakh joined	17:43
	← praka left	17:45
	← DmitryK left	17:50
	→ rlb3_work joined	17:54
	→ jengelh joined	17:54
jengelh	What makes a pack (during `git pull`) a thin pack, and when does it become thick/fat?	17:54
	http://rafb.net/p/Vzyob570.html	18:00
robin	strange git1.5.3-rc4/cygwin problem...	18:05
	I've been snapshotting version using git; i.e. git add path;git add -u;git commit -m "<blah>"	18:05
	and for some reasons a number of files show up as new files, but commit never writes them so they stay "new" forever yielding empty commits	18:06
	instead of no commit when there is no change.	18:06
	← kanru left	18:07
robin	i've not had time to produce a small example yet, but i suspect the optimization junio did earlier, since 1.5.2 did not have the problem	18:08
	→ kanru joined	18:12
	→ ft joined	18:17
	_bdiego → bdiego	18:17
	→ DmitryK joined	18:22
	← DmitryK left	18:24
	→ dduncan joined	18:24
	→ KiBi joined	18:26
hliusv561	robin: You don't happen to have GIT_INDEX_FILE set or something?	18:30
	→ eikke joined	18:30
eikke	is there any git-cvs like git-svn ?	18:31
	← capisce left	18:32
siprbaum	eikke: you could use git-cvsimport and git-cvsexport (i've used it once for a project at the uni)	18:32
	→ DmitryK joined	18:35
	← DmitryK left	18:35
	← KiBi left	18:35
	→ KiBi joined	18:36
	→ Arjen joined	18:36
robin	hliusv561: no	18:38
	but rm'ing .git/index first resolves the problem	18:39
	(I think....)	18:40
	← kanru left	18:40
eikke	siprbaum: any idea: I did git-init-db, then try git-cvsimport, but says the origin branch doesnt exist	18:42
	git-branch origin doesnt work as there is no master branch yet *-)	18:43
	and it'd be kinda stupid to make a first commit using some dummy file I guess	18:43
robin	eikke: start without a git repo	18:43
siprbaum	yep. thats what i would suggest, toot	18:43
	→ kanru joined	18:44
eikke	right, must have done something wrong :)	18:44
	→ fultilt joined	18:44
eikke	hmm, now the dir is empty	18:50
	master branch	18:50
	git-pull . origin says its uptodate	18:50
	pasky changed the topic to: 1.5.2.4/1.5.3-rc4 \| Everyone asleep or clueless? Try [email@hidden.address] \| Channel log http://colabti.de/irclogger/irclogger_log/git \| Linus on git: http://youtube.com/watch?v=4XpnKHJAok8 \| Mailing list archives: http://news.gmane.org/group/gmane.comp.version-control.git/ \| repo.or.cz down for this night (DNS trouble)	18:52
kergoth	hmm, from a script, how to i check if a given commit is reachable from a given ref?	18:52
	pasky changed the topic to: 1.5.2.4/1.5.3-rc4 \| Everyone asleep or clueless? Try [email@hidden.address] \| Channel log http://colabti.de/irclogger/irclogger_log/git \| Linus on git: http://youtube.com/watch?v=4XpnKHJAok8 \| Mailing list archives: http://news.gmane.org/group/gmane.comp.version-control.git/ \| repo.or.cz down for this night (DNS trouble, ip = 62.24.64.27)	18:52
	← bla left	18:52
	← aroben left	18:53
	pasky changed the topic to: 1.5.2.4/1.5.3-rc4 \| Everyone asleep or clueless? Try [email@hidden.address] \| Channel log http://colabti.de/irclogger/irclogger_log/git \| Linus on git: http://youtube.com/watch?v=4XpnKHJAok8 \| Mailing list archives: http://news.gmane.org/group/gmane.comp.version-control.git/ \| repo.or.cz's DNS is in trouble, ip = 62.24.64.27	18:53
pasky	:)	18:53
	→ aroben joined	18:54
robin	kergoth: try git-rev-list	18:54
	← aroben left	18:56
	→ nud joined	18:56
kergoth	k, thanks	19:03
robin	kergoth: not sure it helps, seems it is too smart .. :(	19:04
	→ reuss_ joined	19:04
kergoth	hmmm	19:04
	← robin left	19:09
whee	is it easy to create a new branch without a parent?	19:11
gitster	it is not. why would you want to do something like that?	19:14
whee	I want to have a repository with not-really-related code for some tools	19:15
gitster	a repository's history can have more than one roots, but the normal reason such roots exist in a repository is because they originated as a different history from an independent repository and then later merged.	19:15
whee	but seperated somehow, maybe submodules would be better	19:15
clee	submodules ftw	19:15
whee	they're related in that they are for working with the same project, but they are standalone utilities	19:16
gitster	why would you even want to have a single repository to host the independent bits, it sounds like you are not even merging them.	19:16
	← d0k left	19:16
whee	there may be a point where they would share a lot of code and I'm lazy and want one repo to worry about :)	19:16
gitster	if the answer is "to give a single distribution points for users", then the standard procedure is to create them in separate repositories and publish in a single repository as different (and independent) branches.	19:17
	html/man/todo branches in git.git are managed that way.	19:17
whee	would submodules also accomplish that?	19:17
gitster	because the only reason they are in git.git is because of that "single distribution point" convenience.	19:17
reuss_	does git-cvsexportcommit handle multiple file additions/deletions? ..	19:18
gitster	subprojects are orthogonal. you do not have to have submodule commits in the same repository.	19:18
	← robewald\|home left	19:19
whee	I guess here's what I'm going for: the utilities don't depend on each other, might depend on a common library, and I'd like to be able to clone one utility or all of them easily	19:20
	ideally, it'd be possible to tag them all at once with some common tag	19:20
	→ robewald\|home joined	19:20
	→ doublec joined	19:21
gitster	sounds like you are talking about a superproject that is tagged with the common tag, that points at individual utility projects as subproject, and common library either as part of the superproject or as another subproject.	19:22
whee	indeed it does. I'll play with that setup	19:25
	thanks	19:25
bemo	gitster: did you get my earlier note (yesterday, maybe?) about "docbook2x-texi" not being available in FC4? (just another datapoint -- no action required)	19:29
gitster	saw it in the scrollback. did "make && make install" barfed for you? I believe the info stuff is purely optional but otherwise we have a serious problem.	19:29
bemo	nope, "make && make install" work fine.	19:31
gitster	Ok, then I wouldn't worry too much about it, thanks ;-)	19:31
bemo	no problem.	19:31
	→ maurizio joined	19:32
	→ lyakh_ joined	19:32
	← robewald\|home left	19:33
	← lyakh left	19:34
	← dduncan left	19:38
	→ adante joined	19:40
	← reuss_ left	19:42
	→ aroben joined	19:46
	→ tongueroo joined	19:51
	→ rjbs joined	19:53
rjbs	I want to use git-svn to interface with work's svn.	19:54
	The man page suggests that I can use svn+ssh://, but I get an error when I try to: fatal: I don't handle protocol 'svn+ssh'	19:54
	→ robin joined	19:57
context	people use svn+ssh ? :x	19:57
	← rtmfd left	19:57
	→ rtmfd joined	19:58
	← Oeje1 left	19:59
	→ crime_minister joined	20:01
robin	whee: I don't agree with gitster. it is easy to create a completely standalone branch. However you must do a simple thing without git commands	20:01
EvanCarroll	Anyone know how to solve: Cannot get the repository state from [email@hidden.address]	20:03
	lyakh_ → lyakh	20:03
whee	robin: I don't know how much I really want standalone branches if the subproject feature will work	20:04
	I thought it would be nice to be able to see the log of everything combined, though; I don't know if the subproject stuff does that	20:05
robin	so we'll keep it a secret until then :)	20:05
vmiklos	hehe:)	20:06
EvanCarroll	do I have to do something funky to share a clone over ssh?	20:06
robin	EvanCarroll: no	20:07
	is "scraper" your ".git" directory?	20:07
EvanCarroll	no but I tried ./scraper/.git too ;)	20:07
	still no joy	20:07
	I learned that lesson though	20:07
robin	show us the full clone command	20:08
EvanCarroll	I don't have the command I used to clone anymore	20:08
	I'm trying to git pull	20:08
	I used this command to git-remote-add	20:09
	git remote add -m -f chason ssh://dealermade@chason:/home/dealermade/Desktop/scraper/.git	20:09
robin	I do not have ':' between host and path	20:09
EvanCarroll	ah!	20:09
	thats an scp thing.	20:09
	I figured it would be the same way.	20:09
	still no joy though	20:09
robin	scp doesn't use ssh://-format	20:10
EvanCarroll	right.	20:10
robin	and without "/.git"?	20:11
EvanCarroll	-- anyway, I have remove the command, deleted the old in ./.git/config and added a new git remote add -m -f chason ssh://dealermade@chason/home/dealermade/Desktop/scraper/.git and still no joy	20:11
	doesn't work without the '/.git' either	20:11
	← rjbs left	20:11
EvanCarroll	hrm.	20:12
	i think this i smy mistake	20:12
	do you `git pull` from something that was `git remote add` to begin with ?	20:13
vmiklos	host:/dir, not host/dir if i'm not wrong	20:14
EvanCarroll	thats what I did first.	20:14
	it didn't work either	20:14
	robin suggested I try the latter	20:14
robin	works here: git remote add -f -m xxxxx ssh://robin@10.3.4.2/home/robin/repos/EGIT.git	20:14
	oops, wrong one	20:14
	this one works here: git remote add -f -m xxxxx ssh://robin@10.3.4.2/home/robin/repos/EGIT.git	20:15
	← rtmfd left	20:15
robin	how hard can it be....?	20:15
	git remote add xxx ssh://robin@10.3.4.2/home/robin/repos/EGIT.git	20:15
EvanCarroll	it adds ok, It just does not pull.	20:15
	now I'm getting bash: git-upload-pack: command not found	20:15
	even though git-upload-pack is in my path ...	20:16
robin	can't be	20:16
gitster	some ssh daemon installations spawn commands without reading user's .bashrc or .bash-profile.	20:16
robin	not only some	20:17
vmiklos	openssh reads .bashrc	20:17
robin	that's standard unix behaviour for what I know	20:17
	.bashrc	20:17
	yes	20:17
EvanCarroll	ah thats most likely it.	20:17
	that is exactly where I added ~/bin	20:18
	to $PATH	20:18
	so how do I add ~/bin to path automagically?	20:18
vmiklos	:)	20:18
	export PATH=$PATH:~/bin?	20:18
EvanCarroll	maybe I'll prefix= make or something, ubuntu is planning on hacking auth_pam to add ~/bin to $PATH	20:19
	vmiklos: granted, and where do i put it..	20:19
vmiklos	~/.bashrc	20:19
EvanCarroll	vmiklos: the very nature of this convo, is if i put it in ~/.bashrc ssh won't pick it up.	20:20
gitster	You don't have to. Look for remote.<name>.uploadpack in the configuration manual we ship. This is probably a FAQ.	20:20
	← nikodemus left	20:20
	→ palmcron joined	20:21
vmiklos	ah yes, i just thought about he'll want to pull, too	20:21
	push*	20:21
gitster	There is corresponding remote.<name>.receivepack for this exact reason. Don't we have a FAQ entry there at the wiki?	20:21
	... Ah, or.cz is down or something due to DNS glitch?	20:22
EvanCarroll	the joys of being the only modern package to default install to ~/bin ;)	20:22
	down =[	20:22
	→ Sho_ joined	20:22
gitster	Unfortunately, it doesn't change a thing if the default installed to /usr/local/bin if sshd does not look at there.	20:23
EvanCarroll	good question. (does it by default look in there)	20:24
gitster	New people would really appreciate it if people add FAQ entries after they learn new things here on #git to the wiki (when it comes back up).	20:24
	← aeruder left	20:24
vmiklos	gitster: yes, it's in the FAQ	20:25
EvanCarroll	got a bot in here?	20:28
	I like bots, that was I can get the solutions I need, without being to much a hinderance	20:28
	→ YGingras joined	20:29
EvanCarroll	9/10 git errors I get are either too broad for google, or non-existant, ie this doesn't work and just fails	20:29
chris__	how can I merge in a branch where a binary file changed?	20:29
gitster	after trying to do such a merge, what did you get?	20:30
	→ rkaway2 joined	20:30
	← rkaway2 left	20:30
	← rkaway1 left	20:30
	→ rkaway1 joined	20:30
gitster	"conflicted, please fix up by hand"?	20:30
chris__	error: Cannot merge binary files: ...fatal: Failed to execute internal merge...Merge with strategy recursive failed.	20:30
gitster	Ok.	20:31
YGingras	hi, I have a branched that I initially cloned with scp, not git clone, iirc, now I try to merge it but I get "/tmp/yould/ - not something we can merge", is there anything that I could do?	20:31
gitster	Now, it wants you to create a merged file in your work tree and "git add" the result. In order for you to do so, you have access to (upto) three versions of that file, in the index.	20:31
chris__	my command was: $ git-merge --no-commit --squash remotes/trunk	20:32
gitster	What does "git ls-files -u" say?	20:32
YGingras	git log shows common commits in the history	20:32
chris__	gitster: nothing	20:33
	gitster warns YGingras that the last comment is not meant for him -- he is talking to chris__.	20:33
YGingras	gitster: I figured that out : )	20:34
	← doublec left	20:34
YGingras	should I pull from my new repo and try to merge from the old head?	20:35
gitster	YGingras: I am getting an impression that it is not such a fundamental error but an error in your command line syntax.	20:36
chris__	gitster: (I'm using 1.5.3.rc2.4.g726f9)	20:36
gitster	"now I try to merge it but I get ..." does not say what you did to "try to merge".	20:36
	You might have said "git merge /that/directory", but that is a plain syntax error. git merge does not take repository name as parameter.	20:37
YGingras	gitster: maybe, I did: rsync -av remote:~/yould/ /tmp/yould/; cd ~/yould; git merge /tmp/yould	20:37
	→ jasam joined	20:37
gitster	Most likely you meant "git pull /tmp/yould $some_branch_name"; which branch you tried is not clear to me but probably it is clear to you ;-)	20:38
	chris__: I see merge-recursive is not being nice there. It should say "I cannot merge this path", but that is not a good reason to abort the entire merge. "git merge -s resolve <other parameters you gave it>" would probably be able to unblock you in the meantime.	20:39
YGingras	gitster: excellent!	20:40
gitster	did it work (perhaps $some_branch_name == 'master')?	20:41
chris__	gitster: It did give me the path, but it didn't tell me about the other paths it left in a conflicted state.	20:41
	← spuk- left	20:41
YGingras	gitster: but it fastforward the whole thing, would it warn if there was anyconflicts? Aren't pull always fast forwarding?	20:41
	gitster: I did "head"	20:42
	→ spuk- joined	20:42
EvanCarroll	gitster: hrm, you what might be kind of sexy. -- the man for git-fitch-pack says it will first look the file in your path, on the remote machine if your path is ~/bin though it will look for the expanded ~/bin, rather than ~/bin -- it would be nice if it let ssh expand ~/bin on the remote machine rather than doing it for ssh ( because I'm not sshing as the $USER )	20:43
	git-fetch-pack	20:43
gitster	YGingras: if "git pull /tmp/yould head" fast-forwarded, then (1) the rsynced repository had a branch called "head" and (2) that was truly ahead of your current branch.	20:43
EvanCarroll	I guess that makes the assumption all ssh clients can do home-expansion but it seems to be a safe assumption	20:44
gitster	chris__: "It" did give me the path? Meaning...? I was looking sideways so I am not following you, sorry. You tried "-s resolve" instead and now you see the binary file's path in the "ls-files -u" output?	20:44
YGingras	gitster: that's what I presumed but I still had a little doubt. I'm glad that it was the case. Thanks for your help	20:44
	→ madewokherd joined	20:44
chris__	gitster: no, sorry, I meant that the original merge did report the failed path, I just elided it from the error report.	20:45
gitster	chris__: yes, I know, but merge-recursive I think punts after that and does not give you the rest of the paths that it _could_ have handled, which was what I meant by "not being nice".	20:46
chris__	gitster: yes	20:46
gitster	I'd "reset --hard HEAD" and retry the merge with "-s resolve" if I were you.	20:46
chris__	I did a git-reset --hard ORIG_HEAD and then tried again	20:46
	fatal: Untracked working tree file '.gitignore' would be overwritten by merge	20:47
	I think the reset --hard didn't really undo what had been done.	20:48
gitster	Yeah, that is what I meant by merge-recursive not being nice. It probably left it as a part of that failed merge, but because it rudely died in the middle before writing the index out, there is no cue for "reset --hard" that the file should be removed. Remove it by hand after making sure that you should not have had that file in your tree before you started that merge.	20:49
whee	is there a nice doc somewhere describing typical submodule usage? git.or.cz appears to be down :\|	20:49
	← crime_minister left	20:49
	→ _spanner_ joined	20:50
gitster	At this point typical is probably something the interested users are working to establish with their projects right now ;-) Write a nice doc somewhere when you are done ;-)	20:50
	← bentob0x left	20:50
	← alley_cat left	20:52
whee	:)	20:52
gitster	I know cgit.git binds git.git as a submodule, so the author may have some notes in which he brags about the way he manages that repository.	20:52
whee	I am trying it now for maintaining a repository of my kernel config + submodule of the kernel tree	20:52
gitster	Also msysgit folks use git.git as a subproject.	20:52
hliusv561	There does not seem to be recursive commit. Or is it just plain insane idea?	20:56
	← meandtheshell left	20:58
robin	what would that be?	20:59
	for subprojects?	20:59
EvanCarroll	how do you set the upload path in your .git/config?	21:00
	you know the damn killer about this is the the tutorial git in 5min uses ssh, and no where in there do they have to set an "upload-path" or a fetch-path or any other goffyness.	21:02
hliusv561	Recursive commit: Descend to checked out submodules, commit them (if needed), update path with new commit and work upwards. Commit messages are sight problem though.	21:04
robin	should be the same meessge then	21:04
gitster	EvanCarroll: I think that is because most sshd admins are kind enough to their users to make sure their .bashrc or whatever is read in non-interactive sessions.	21:07
mnemoc	hi, how should i escape a ';' as part of a config value?	21:08
EvanCarroll	ok i think i got this working	21:08
	I think it works with the git-peek-remote	21:08
hliusv561	EvanCarroll: Also, are you sure that the shell on remote side is bash? Ocassionally one runs into environments where bash is not default shell...	21:09
EvanCarroll	hliusv561: yes, this is just off-the-shell ubuntu.	21:09
	gitster:when I manually ssh in, git-* is in the path.	21:09
	gitster: I'm not sure what is happening as far as ssh, if it has a non-interactive login that won't process ~/.bashrc too, or something.	21:10
	that has to be it	21:11
gitster	"when you manually ssh in", you are probably talking about an interactive session.	21:11
	try "ssh that-site /bin/env" or something like that to see what the sshd does for your non interactive connection.	21:11
	Some sshd installations honor ~/.ssh/environment; you may want to check that as well.	21:11
	→ amf joined	21:11
cortana	remember also that bash does some annoying working around old broken versions of rsh that screws up which dot files are processed when you ssh or rsh in	21:12
EvanCarroll	I'd rather know how do i add upload-pack into my .git config for a [remote foo]	21:12
amf	I cloned a SVN repo, and it created the master branch for me, is this a private branch? I don't think I understand the difference of the origin and master branches	21:12
	← lcapitulino left	21:13
EvanCarroll	amf: iirc origin is an exact replica of the head of the repo you clone, while master is the default head of your own repo.	21:13
gitster	Something like:	21:13
	[remote "origin"] upload-pack = "/mnt4/home/evan/bin/git-upload-pack"	21:13
	← otavio left	21:14
	gitster reminds evan that this is to be done on the local side that invokes "git push", not on the remote side that runs sshd.	21:15
EvanCarroll	ok this is making me want to scream.	21:17
	Cannot get the repository state from ssh://dealermade@192.168.1.112:/home/dealermade/Desktop/scraper/.git	21:17
hliusv561	EvanCarroll: Is your default shell "/bin/sh" (regardless weither it is symbolic link)?	21:17
gitster	That URL has extra colon after the IP address which tells you to connect to port 0.	21:18
	→ Pistahh joined	21:18
EvanCarroll	hliusv561: no default shell is /bin/bash	21:19
gitster	Using "user@host:path" form like "scp" feels much cleaner than URL-ish form ssh://host/path, by the way.	21:19
EvanCarroll	gitster: I was doing user@hast:path	21:20
	guess Ishould try wo/ ssh:// then?	21:20
gitster	"Cannot get the repository state from ssh://dealermade@192.168.1.112:/ho..."???	21:20
EvanCarroll	gitster: doesn't work either way	21:20
	the path is /home/dealermade.... the user is dealermade, the host is the ip or (chason) taking out the dns to avoid complexity	21:21
chris__	gitster: this is me first time using -s resolve; it's really much less smart than recursive, no?	21:21
gitster	chris__: not that much for everyday use if you are organized well (i.e. no cross-cross merges nor renames)	21:22
EvanCarroll	gitster: any ideas? same shitty gitish error either way	21:22
	Here is a good joke for you: Ever hear of a descriptive git error? No? -- Because there isn't one, jokes on the users... get it?	21:23
gitster	it feels like some ssh connection error to me, but if you are getting it from git-fetch perhaps running it under sh -x might give us more clues where it breaks.	21:23
EvanCarroll	gah	21:23
	gitster: no more clues	21:24
	just did it	21:24
gitster	Also I've never used user@host form; I do not offhand recall if we added the support to do that.	21:24
robin	it work at least	21:25
	how else would be tell which use to use?	21:25
	← chris2 left	21:25
EvanCarroll	gitster: i thought you suggested it, I'll go back to ssh:// form then	21:25
gitster	When I push as somebody else, I always have ~/.ssh/config entry that maps the "Host", "User", and "IdentityFile" entries.	21:25
robin	IC	21:26
gitster	What I meant was that I do not offhand know either ssh:// or host:path form supports it.	21:26
	But robin says it works for me, so...	21:26
whee	I use ssh://user@host/path and that works	21:27
chris__	gitster: I don't understand. It seems like it left conflicts for lots of trivial merges. 274 files worth. most of which are simple insertions or deletions.	21:28
gitster	Ok, then probably comparing your .git/config with Evan's would probably be a better way to figure this out.	21:28
hliusv561	chris__: Yes, resolve is quite stupid.	21:28
chris__	for a normal merge, I'd expect 6 conflicts, all involving the same code modified.	21:29
	It'd be nice if I could use merge recursive, but fall back to resolve for the paths that fail.	21:30
EvanCarroll	gitster: I'm 100% sure the user@host:path works.	21:31
gitster	Good to hear it is working, then.	21:31
chris__	In fact, I thought that's how merge worked.	21:31
EvanCarroll	git-peek-remote --upload-pack=/home/dealermader/bin/git-upload-pack ssh://dealermade@chason/home/dealermade/Desktop/scraper/.git works.	21:31
	however, no variant of git-pull works.	21:31
	err thats the ssh:// form rather	21:32
gitster	chris__: Now I do not understand. Where do these 270+ paths come from?	21:32
EvanCarroll	err damit	21:32
	one second.	21:32
	that command does not work	21:32
	git-peek-remote --upload-pack=/home/dealermade/bin/git-upload-pack dealermade@chason:/home/dealermade/Desktop/scraper/.git	21:32
	that command does work	21:32
	that proves that ssh://user:host/path does not work	21:32
	gitster: ssh://user:host/path is nautilus (gnomefs) addressing iirc.	21:33
chris__	gitster: I did: `git-merge --no-commit --squash -s resolve remotes/trunk` and it left 274 unmerged files.	21:33
EvanCarroll	user:host@path is scp addressing	21:33
	nautilus is made to look like uri	21:33
robin	: is for port in generic URLs	21:33
gitster	chris__: that much I already knew. What I do not understand is where that 274 vs 6 difference comes from. Do you have many criss-cross merges?	21:34
EvanCarroll	so my conclusion is the newest git does not in fact work over ssh, without depending on modifications to the ssh-server on the remote side.	21:34
	which is mega shitty.	21:34
robin	works here	21:34
EvanCarroll	you must have modified your ssh-server to include the path to git somehow.	21:35
robin	no	21:35
EvanCarroll	yes.	21:35
robin	just installed it in a location that is included in the PATH	21:35
gitster	we haven't changed much in that area so if it does not work for you then that is not newest vs older. If it does not work it does not work, and we need to know why it doesn't for you and why it does for the rest of them.	21:35
chris__	gitster: 6 was an estimate, it turns out I get 9 conflicts by doing the same merge with svn. And I don't know what you mean by criss-cross merges, but I think the answer is no.	21:36
gitster	Then, we can have a change to make it work also for you.	21:36
EvanCarroll	gitster: because the rest of you can access git commands by going `ssh user@host cmd` and I can not.	21:36
	and there is no way I can provide the path. in git pull, only in git-peek*	21:36
gitster	git-fetch takes --upload-pack command line and git-pull as well, so that might be worth a try, but this really should work with the config settings as well. At least that is how it is designed to be used.	21:38
	→ aeruder joined	21:39
EvanCarroll	I've tried it, no matter what path I give it, it fails	21:39
gitster	you earlier said that peek-remote with --upload-pack command line option works?	21:40
robin	EvanCarroll: try setting the path in ~/.ssh/rc	21:40
EvanCarroll	gitster: yes, thus the problem.	21:40
	← maurizio left	21:40
gitster	Does "ls-remote" with --upload-pack work? That is what is executed from git-fetch.	21:40
robin	~/.ssh/environment is another place. See man sshd(8)	21:41
gitster	ls-remote is a thin wrapper around peek-remote.	21:41
EvanCarroll	awkward.	21:41
	here is one bug, well probably not a bug, but bad error reporting by design.	21:42
	git-upload-pack will not allow you to prefix ssh address with ssh://	21:42
_spanner_	I have a packed archive where there are lots of commits that are unreachable except via the reflog. I can see them with git-fsck --unreachable --full --no-reflogs ... but how to remove them?	21:43
EvanCarroll	maybe git just has no uniform ssh support, and every one of the 150E150 apps use a slightly modified UI	21:43
	← tongueroo left	21:43
gitster	Actually ssh:// was bolted on (nobody needed it but somebody wanted it) without much thought.	21:44
	If there is something that is "uniform ssh support", that is host:path.	21:44
EvanCarroll	yea, as an outsider I get the impression that happened a lot.	21:44
	I like the error message, for adding ssh:// fatal: The remote end hung up unexpectedly	21:44
	because the remote end was successfully contacted.	21:44
	← kumbayo left	21:45
	→ tongueroo joined	21:46
EvanCarroll	git is officially the highest level app I've ever had to trouble shoot with ltrace.	21:47
robin	isn't that what one usuall does with servers? :)	21:48
	← amf left	21:49
	jwb → jwb_gone	21:49
EvanCarroll	no, they usually defer to logs or send a -v or something sane to the UI.	21:49
	http://rafb.net/p/WyfPnx94.html check that out	21:51
	last two cmds	21:51
robin	oh, strace/ltrace and wireshard are so simple	21:51
	wireshark	21:51
EvanCarroll	shark?	21:52
robin	looks wierd	21:52
	EvanCarroll: new name for ethereal	21:53
EvanCarroll	I know what it is, I use it often.	21:53
	the difference is I use strace/ltrace/low level network apps, when I'm developing, not when I'm wanting to get other stable software to work.	21:54
	then I resort to logs, and verbosity of the program.	21:54
robin	I use whatever I need	21:54
	logs first ofcourse	21:55
	→ nashdj joined	21:55
robin	i.e. the simplest things first	21:55
	the git programs don't log much	21:55
EvanCarroll	gitster: you check out last paste?	21:57
evilchelu	hi, i did git-submodule add remoterepo localdir. how do i pull the submodule from remoterepo?	21:57
hliusv561	EvanCarroll: maybe turn tracing on? "GIT_TRACE=1" or something...	21:58
	← GyrosGeier left	21:59
evilchelu	oh, figured it out... git pull in the submodule dir	22:00
	← whee left	22:02
	→ whee joined	22:02
	← _spanner_ left	22:04
EvanCarroll	gitster: no , thats exactly the problem	22:09
	gitster: git-ls-remote is borked.	22:09
	← Dodji left	22:12
EvanCarroll	gitster: you got it.	22:12
	the git-* bash scripts blow, I think I'm going to start a project to rewrite them in perl	22:12
Tv	EvanCarroll: won't that make a windows port just as annoying as the shell thing, if not more?	22:13
EvanCarroll	like in 07' ORing args is a pretty bad idea.	22:13
	Tv: no active state makes it easy to wrap perl scripts.	22:13
	but who cares about windows anyway, the immediate problem is this doesn't work for me, and it is a bug in git-ls-remote.	22:14
	→ nikodemus joined	22:15
EvanCarroll	I guess I hit the nail head on in my estimation of git UI	22:17
	← stouset left	22:17
EvanCarroll	this thing does all of the upload-pack by itself., git-peek-remote, the one with the working --upload-pack is C	22:18
	so no copy and pasting code =[	22:18
evilchelu	anyone having problems getting to git.or.cz? it seems down here	22:20
	→ raoul_ joined	22:22
hliusv561	evilchelu: I think it (or actually its DNS) is down...	22:22
EvanCarroll	yes, it is down.	22:22
	← whee left	22:22
EvanCarroll	anyone happen to have IP?	22:22
	I wanted to git the source and go to starbucks, I'm serious about rewriting the scripts into perl ;) use Moose and give them nice interface.	22:23
evilchelu	EvanCarroll: i have the tgz from yesterday if you want	22:23
raoul_	hi all - how is $id$ supposed to work via gitattributes? The manpage alludes to it. I've just been adding: # $Id$ for example in my file, checking it in, assuming that would get expanded in place each time, but it doesn't even get expanded. The documentation is a little light. Any ideas?	22:24
EvanCarroll	nah, I'll wait for git	22:24
evilchelu	actually, i think that's not on the same site	22:25
hliusv561	raoul_: I tried trying the same some time ago, but I couldn't get it to work...	22:25
raoul_	Hmm. Interesting.	22:25
	I'm assuming the idea behind it is akin to what CVS has?	22:26
	← nud left	22:26
	robfitz_ → robfitz	22:27
rehges	evancarroll: pasky.or.cz is down and the backup DNS claims it doesn't know or.cz at all	22:27
EvanCarroll	raoul_: go watch the linus presentation @google edu	22:29
raoul_	Eh?	22:31
	I was referring to this mythical $Id$ attribute, not GIT itself.	22:32
	Or was the context too hard to grasp?	22:32
	→ zen-master joined	22:34
	← orospakr left	22:38
Tv	raoul_: and he was referring to why you shouldn't do that..	22:40
	← lyakh left	22:40
raoul_	Oh, why? Is there something wrong with $Id$'s use?	22:41
Tv	most often very much so	22:41
	i wonder if anyone has written a concise "$Id$ considered harmful"	22:42
raoul_	So why is it supported in GIT? My understanding from the gitattributes manpage was that, unlike CVS, the GIT $Id$ mechanism just expands the SHA1. That's fine.	22:42
Tv	because somebody really really wanted it, i guess	22:43
	that doesn't make it a good idea ;)	22:43
hliusv561	raoul_: And actually the numbers it puts there are for the blob, not for the commit...	22:44
	→ jbellis joined	22:44
	← binary42 left	22:44
jbellis	is there a git analog to "hg grep" ?	22:44
hliusv561	jbellis: 'git grep'?	22:44
Aeternus_	isn't the problem that that doesn't really make sense as you are then effectively changing the files that made up the SHA1 in the first place :\? or am I completely misunderstanding the issue?	22:44
evilchelu	hey, any git-svn and git-submodule users around? i'm trying to dcommit a repository with submodules and i get a "bad file" error from git-svn	22:44
jbellis	cool :)	22:44
	← ferdy left	22:45
raoul_	Yes, my mistake, The blob.	22:51
	→ crime_minister joined	22:52
	← goloo left	22:56
	→ whee joined	23:00
	← Yuuhi left	23:11
EvanCarroll	I think the easy solution to all of my woes is just to override the (very, (very)) stupid default of ~/bin	23:11
	doesn't seem like ~/.ssh/rc will change it	23:11
	← jbellis left	23:16
	→ doublec joined	23:20
	→ robfitz_ joined	23:22
whee	which doc should I be reading if I want to take a branch and reorder the commits on it?	23:27
EvanCarroll	error: refs/remotes/chason/HEAD points nowhere!	23:29
	ok, any idea?	23:29
	shouldn't HEAD be set by git to whatever the remote has in working dir	23:29
	← doublec left	23:31
	← robfitz left	23:33
	→ doublec joined	23:36
	← jackbravo left	23:37
	← henr_k left	23:38
	→ jackbravo joined	23:38
	← nikodemus left	23:41
	→ matley joined	23:43
matley	how can i get git now?	23:44
EvanCarroll	wait for the dns issue to get fixed...	23:45
bdowning	Use the IP address, which is in the IRC topic? (62.24.64.27)	23:49
	→ robfitz joined	23:52
matley	bdowning, i dont see anything on http	23:56
bdowning	You may have to add it to your hosts file as repo.or.cz for http to work beacuse of named-based virtual hosting...	23:57
	Ssh should work fine though.	23:57

Logs Search ←Prev date Next date→

Channels Documentation