IRCloggy #git 2008-07-16

Provider of IRC logs since 2005.

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

	2008-07-16
pasky	Spyro: do the fsck	00:00
rtyler	pasky: here's the brief transcript between me and the guy evaluationg 1.5.0	00:00
pasky	Spyro: and even before, check permissions	00:00
rtyler	http://pastebin.com/d582675cf	00:00
Spyro	I already did. it returned ok	00:00
	(well no text)	00:00
pasky	Spyro: the gitweb user has to have full access to the repo	00:00
Spyro	I've ensured a+r permissions across the repo	00:00
pasky	Spyro: maybe you're hitting the <8c6202> umask bug	00:00
Gitbot	[git 8c6202]: http://tinyurl.com/5snfbw -- Fix backwards-incompatible handling of core.sharedRepository	00:00
pasky	Spyro: and what about +x for directories?	00:01
	← sris left	00:01
pasky	(hint: a+rX)	00:01
Spyro	hm. just spotted this though: fatal: unable to read source tree (1d9d8fa8ce1a8481d0239948af1f512af703254b)	00:01
	← BabelO left	00:01
pasky	I bet on permissoin problem	00:01
Spyro	aha, a+rX. good call	00:01
	Thanks :-)	00:01
SRabbelier	!botsnack	00:02
	:P	00:02
pasky	#	00:02
	(12:24:57 AM) R. Tyler Ballance: like, pretend I wasn't in the meeting :p[M#y@	00:02
	lol	00:02
	(oops, sorry for the garbage; my terminal is going all wrong)	00:02
Spyro	is the umask issue fixed in git yet ?	00:02
rtyler	heh	00:02
pasky	Spyro: yes but not in any release	00:02
Spyro	right :)	00:02
pasky	Spyro: umask=0644 might be a workaround	00:02
	test it out	00:02
rtyler	I somehow got left out of the meeting about SVN, despite being the one looking into Git	00:02
pasky	er	00:02
	shared=0644	00:02
	sorry	00:02
	Spyro prays to the debian gods	00:02
pasky	rtyler: ah :)	00:03
Spyro	pasky: yeah I think Im gonna use that for now :)	00:03
pasky	well I mean I didn't test it ;)	00:03
ciaranm	the debian gods, eh? you need to inscribe your prayers on stone tablets	00:03
	→ igorgue joined	00:03
SRabbelier	ciaranm: nono, stone tablets are not in stable yet	00:04
Spyro	ciaranm: in triplicate no less... but then I have to say its been remarkably stable and easy to bend to my (unusual) setup.	00:04
ciaranm	SRabbelier: still? ouch.	00:04
SRabbelier	stable is still running version 1.4.4 of the Communication package	00:04
pasky	rtyler: uh - so I can repeatedly merge only if I do no local commits in my branch?	00:04
	is that really so?	00:04
SRabbelier	ciaranm: you'll have to use cave drawings instead, no words allowed!	00:04
Spyro	I have an unmodified debian booting from NFS on three machines, all sharing one fileserver.	00:04
ciaranm	SRabbelier: i guess that explains dselect	00:05
Spyro	my machines all run testing :)	00:05
SRabbelier	lol :P	00:05
	SRabbelier is running testing as well	00:05
SRabbelier	etch ftw	00:05
	pasky is actually pretty happy etch user	00:06
SRabbelier	:P	00:06
	Spyro uses the testing alias so I dont have to change it when they shuffle the names around ;-)	00:06
	→ cilly joined	00:06
pasky	oh yeah I used to do that for 'stable'	00:06
	learnt my lesson after few disasters	00:06
SRabbelier	Spyro: yeah, lenny, sarge, etch, whatever, I forget :P	00:06
	pasky: maybe specify the actual version instead then?	00:06
pasky	well, that's what I did	00:07
Spyro	well the filestore lost its NFS daemon yesterday. but one application crashing in over 400 days of use is pretty good I think	00:07
pasky	bearable ;)	00:07
Spyro	some things actually run faster than a local disk (others not but hey...)	00:07
	(gigE)	00:08
	SRabbelier yawns	00:08
pasky	stop yawning	00:08
SRabbelier	have fun :)	00:08
pasky	code!	00:08
SRabbelier	I'm off to bed :P	00:08
pasky	aaw	00:08
SRabbelier	code? tomorrow maybe!	00:08
	Spyro is gonna catch some darks too I think	00:08
pasky	hmm	00:08
	→ gitte joined	00:08
pasky	maybe I should do the same	00:08
SRabbelier	now is bed tiem :P	00:08
	← icwiener left	00:08
pasky	good night to you then	00:08
	Spyro just finished preparing the MFD patches for linux...	00:09
SRabbelier	pasky: check your crontab to see when suspend job runs :P	00:09
pasky	MFD?	00:09
Spyro	Multi Function Devices	00:09
SRabbelier	most f*ing dorky?	00:09
Spyro	currently queued in RMKs tree to go into mainline	00:09
SRabbelier	Spyro: DO NOT SEND IT IN YET!	00:09
pasky	Spyro: uh... like a mouse? I can both move and click!	00:09
SRabbelier	Spyro: wait till tomorrow ;)	00:09
	Spyro: ask pasky :P	00:09
Spyro	SRabbelier: of the code I did tonight is further updates, it needs review now ;-)	00:10
SRabbelier	don't do anything that could possibly go wrong after... say.. midgnith :P	00:10
pasky	bah	00:10
Spyro	but it builds, works, and passes through checkpatch without too many bruises...	00:10
SRabbelier	Spyro: if you send it in now you'll probably forget something	00:10
	Spyro: signing it off or such, ccing relevant people	00:10
Spyro	SRabbelier: Im not, dont worry	00:10
	thats why I wanted my gitweb working - so it can be reviewed!	00:10
SRabbelier	Spyro: you are guaranteed to forget something if you send in a patch this late :P	00:10
	anyway, I'm ranbling, better get going ^^	00:11
Spyro	I hear you, really :)	00:11
SRabbelier	cheers!	00:11
Spyro	night!	00:11
	← SRabbelier left	00:11
Spyro	Im off too I think :)	00:11
	TTFN!	00:11
rtyler	pasky: I'm not 100% certain	00:12
	I didn't play with it because I have enough brane damage after my git-svn fuckery	00:12
	← schacon left	00:13
	→ vmiklos joined	00:14
gitte	vmiklos: I thought you were on a bike...	00:15
	→ lu_zero joined	00:16
	← nbrunson left	00:16
pasky	rtyler: I see.. thanks anyway	00:19
rtyler	pasky: the merge tracking is sufficiently brain dead out of the gate though, it depends on svnprops and thus client and server must both be 1.5.0	00:20
	→ ketralnis joined	00:22
	← njal left	00:29
	→ doener_ joined	00:31
	→ foca joined	00:32
	→ heycarsten joined	00:33
	→ noecksit joined	00:34
pasky	i see	00:36
	hmm, renaming non-checkedout submodules is annoying	00:36
	→ elight joined	00:38
	← joydivider left	00:38
	pasky will do a horrible hack for now ;)	00:39
	→ hansengel joined	00:40
hansengel	Hi, I just downloaded the Git source, ran ./configure and then make.	00:40
	I got this error while compiling:	00:40
	CC daemon.o	00:40
	In file included from daemon.c:1:	00:40
	cache.h:9:18: error: zlib.h: No such file or directory	00:40
	make: *** [daemon.o] Error 1	00:40
spearce	gah. i just posted a message with no To: header and only Cc: header. stupid mutt.	00:40
	→ Leefmc joined	00:40
hansengel	Do I need zlib development files, or something?	00:41
spearce	hansengel: you appear to be missing the zlib dev package, yes	00:41
hansengel	ah	00:41
	yep, that fixed it so far :)	00:42
	→ brosner joined	00:43
	← doener left	00:44
	← RaceCondition left	00:45
nil	hi!	00:46
	← zawir left	00:46
	GeckoPie_ → GeckoPie	00:46
nil	I'm looking for the git equivalent to "darcs revert", ie cancel some changes i did but did not commit to the working files	00:46
	← hansengel left	00:47
ciaranm	nil: git checkout	00:47
	→ elliottc1ble joined	00:47
nil	ciaranm: it will let me choose which file to revert to the recorded state, and which to keep modified?	00:48
ciaranm	nil: you tell it the files you want	00:48
	git checkout iscrewedupthisfile.cc	00:48
	← rofrol left	00:49
nil	ok, the manpage has good exemples	00:49
	thanks	00:49
	← bdiego left	00:52
	→ mxey joined	00:54
	← elight left	00:55
	GeckoPie → GeckoPie\|Sleep	00:56
	→ crab joined	00:58
	← elliottcable left	00:59
crab	is there a shorter way to say "branch/name..origin/branch/name" when i'm doing git log?	00:59
gitte	No.	01:00
	But you can use tab completion if you're on bash or zsh.	01:00
	← markkalderon left	01:00
	← godlygeek left	01:00
	→ marc_vw joined	01:02
	← ketralnis left	01:03
Eridius\|work	crab: you could make an alias that manufactures that string	01:03
crab	it was more of a conceptual question.	01:04
	i'm glad i'm not missing anything obvious.	01:04
	(i have an alias already)	01:04
gitte	crab: I smell an XY problem.	01:07
	crab: if you only want to see if you're up-to-date, and you are on that branch, then a "git status" with the newest Git (next) will show you that.	01:07
	← rtomayko left	01:07
	→ rtomayko joined	01:08
gitster	You certainly mean "git checkout", not "git status", right? The latter one pages ;-)	01:10
crab	an "XY problem"? huh? i don't have a problem.	01:10
	→ elight joined	01:12
	→ markkalderon joined	01:14
	← ijcd_ left	01:19
	← lu_zero left	01:19
	← mvw left	01:19
gitte	gitster: I meant "git status". It pages alright, but it does not change the current working directory.	01:21
	crab: If I guessed right, and you wanted to know how many commits the upstream is ahead, then you have a problem: you asked the wrong question.	01:21
	← jonshea left	01:23
	← Fullmoon left	01:23
gitster	If you want to know how many commits the upstream is ahead, "git status" is also a wrong question, isn't it? It talks about your local changes and untracked files.	01:25
gitte	Didn't you recently change it so it has the same message as git checkout?	01:26
gitster	That is why I suspected you meant "git checkout" (no other parameters).	01:26
gitte	One of my repositories shows this:	01:26
	# On branch master	01:26
	# Your branch is ahead of the tracked remote branch 'origin/master' by 6 commits	01:26
gitster	it does not give you a lot of extra cruft as git-status does.	01:27
	s/it/checkout/	01:27
gitte	But it is a modifying operation.	01:27
	Oh, well. You're the gitster.	01:27
	← sgrimm left	01:27
	← warthog9 left	01:27
gitster	Not at all. "git checkout" is largely no-op if you do not give anything extra ;-)	01:27
gitte	... which is a usability wart, just as the name.	01:28
	Anyway, I have to close my eyes and get some letters from the end of the latin alphabet.	01:28
	IOW git night	01:29
	← gitte left	01:29
	→ VVSiz_ joined	01:29
gitster	The blame optimization is quite good, by the way ;-).	01:29
crab	gitte: anyway, i asked exactly what i meant to, and you guessed wrong. but good night.	01:29
	→ dramsay joined	01:30
crab	gitster: what blame optimisation?	01:30
pasky	yay	01:31
gitster	aya?	01:31
	yay?	01:32
	pasky finally caught hopefully all the corner cases of git mv submodule... though Junio was right to fear his hacks ;)	01:32
gitster	heh.	01:32
	← tmz left	01:32
pasky	git rm and conflicts at the morning	01:32
	'nite guys	01:32
	→ garrin joined	01:32
	crab browses the todo looking for something interesting to do.	01:35
pasky	crab: I could use git-gui interface for submodules ;)	01:37
	← rphillips left	01:37
crab	oh no, not more tcl.	01:38
	heh, "author refuses to respond to comments"	01:39
	→ elliottcable joined	01:40
	← elliottc1ble left	01:40
	← rdavila_ left	01:41
	← elliottcable left	01:43
	← Leefmc left	01:43
	→ hagabaka joined	01:44
hagabaka	if I push a commit and then later commit --amend it, is it possible to fix the remote?	01:44
	→ tmz joined	01:45
	→ jonshea joined	01:45
	← VVSiz left	01:46
	← carrus85 left	01:46
	→ elliottcable joined	01:48
hagabaka	or is it possible to get the "amendment" out as a separate commit so that I can push it?	01:48
uau	hagabaka: it is possible to destroy the existing commit on the remote and replace it, but that may cause problems for other people if they already fetched the original version	01:48
hagabaka	oh	01:49
	what about the second question?	01:49
	or, a way to warn me when I try to amend an already pushed commit	01:50
	← whee left	01:51
uau	if you do "git reset <the state you pushed>" while you have the amended version checked out	01:52
	then you should have the nonamended version as branch state but amended version as working directory state	01:53
	and could then commit it	01:53
	→ Leefmc joined	01:53
uau	creating an extra commit moving from original version to amended one	01:53
hagabaka	state is a commit?	01:54
	but doesn't amend change the commit, so I can't find the state before amend?	01:55
uau	you should be able to find it in the reflog	01:55
hagabaka	ah	01:55
uau	or since you pushed it	01:55
	you should be able to see it as the status of the remote branch	01:55
hagabaka	oh	01:55
	is there a script to automate that?	01:56
	→ joydivider joined	01:57
uau	no (and what exactly could it automate? hard to pick the right commit to reset to automatically etc)	01:57
crab	why not revert instead of amend?	01:57
	oh, sorry. i misunderstood.	01:58
	→ paltman joined	01:58
hagabaka	something like, find amended commits to the last pushed commit and change them into regular commits	01:59
	← lea_w left	01:59
hagabaka	preferably it would work even if there are more commits after the amend	01:59
	so i should be able to just look at remote HEAD and find it in the reflog, and look for amends starting from there...	02:01
	* the script should be able to ...	02:01
uau	you cannot have more than one such amended commit	02:02
hagabaka	why is that?	02:03
uau	since amend can only change the latest commit it's only possible to change the last commit you pushed	02:03
	and if you change the same commit multiple times the last amend just replaces the previous ones	02:03
hagabaka	does that mean the last amend is the sum of all the previous amends?	02:03
	← digitalhobbit left	02:03
crab	but you can amend multiple commits at separate times.	02:03
hagabaka	are amend commits actually stored as separated commits in the "reflog"?	02:04
uau	crab: but at most the first one has been pushed before amending	02:04
	hagabaka: amend removes the last commit from the branch and replaces it with a new one	02:04
	if you keep replacing the commit the result doesn't really differ	02:05
	if doesn't matter if you replace A with B, then B with C	02:05
	or A directly with C	02:05
hagabaka	but the state before amends are stored?	02:05
uau	i'm not sure what you're asking about	02:05
	reflog stores where a particular branch, or your checkout-out HEAD, points to at various times	02:06
	← Eridius\|work left	02:07
	→ godlygeek joined	02:07
	← geckosenator left	02:07
	→ piggybox joined	02:08
hagabaka	well, so would it work if a script finds the remote HEAD in reflog, then search starting from there and find the last amend to that commit, and convert that into a new commit?	02:08
uau	you'd at least need to give it a new commit message	02:11
	if instead of amended commit you'd have a separate fix commit	02:11
hagabaka	sure	02:12
uau	doing something with the same effect would probably be scriptable, but i doubt it would be worth it	02:12
	it shouldn't be that common	02:12
hagabaka	i've amended pushed commits a few times	02:13
crab	why?	02:13
hagabaka	because i pushed and forgot that I did	02:13
crab	if you're using amend regularly, you're doing something wrong.	02:14
ciaranm	or you like using git rebase -i to tidy up commits	02:15
hagabaka	it seems justified to me if I typo a line or forget to add a file before commiting	02:15
	→ litage joined	02:15
	← timocratic left	02:16
hagabaka	and the script doesn't really encourage amending, it's just to make it easier when you amend after pushing	02:16
	(if it existed)	02:16
	→ johnw joined	02:20
	← crab left	02:24
uau	if you have further commits after the amended one i think the following should work:	02:24
	git rebase -i pushed_branch	02:25
	← piggybox_ left	02:25
uau	mark the first commit (which should be the amended version) as "edit"	02:25
	← kelvie_work left	02:25
uau	when git tries to apply it it'll probably cause a conflict, but that doesn't matter	02:25
	then do "git checkout amended_commit_sha1 ."	02:26
	and "git commit" to create a fix commit	02:26
	then "git rebase --continue"	02:26
jackdempsey	dont' you want to be careful with rebasing something thats been pushed?	02:30
uau	the above was to undo an amend that was already accidentally made to a pushed branch	02:32
	→ kanru joined	02:32
uau	the "pushed_branch" in "git rebase -i pushed_branch" was meant to indicate the version that was actually pushed	02:33
	← dramsay left	02:39
hagabaka	how do I name pushed_branch? is it the SHA in reflog?	02:40
	or the remote?	02:40
uau	remote	02:41
	← david_koontz left	02:41
hagabaka	i used a local directory as test repository and pushed to it, then git rebase -i dir, and it says "fatal: Needed a single revision\nInvalid base"	02:43
	← jackdempsey left	02:43
uau	i meant a tracked remote branch	02:45
	whose state you'd have in the local repository	02:45
hagabaka	how would I type it in the command line?	02:45
uau	if you don't have such (just did a push to an explicitly named repository) then you should give the SHA1 you pushed	02:45
hagabaka	oh	02:45
	then checkout that same SHA1?	02:47
	oh, the one after amend?	02:48
	gives me "fatal: you need to resolve your current index first"	02:48
	→ chieh joined	02:49
uau	you should have the branch you've worked on checked out when you start that	02:50
	→ CodeWar joined	02:50
uau	(including all the further commits you've done)	02:50
	that error is because you have uncommitted modifications in your working directory	02:50
	→ paltman_ joined	02:52
	→ jackdempsey joined	02:53
	← paltman left	02:56
	← aroben left	02:56
hagabaka	hmm	02:57
	this is weird, this time it keeps saying "everything up to date" when I push	02:59
	but diffing the "remote" dir shows difference, and git diff in the working directory is empty	03:00
	don't directories work the same way as other repositories?	03:02
uau	what about "git status"? if you have the modifications in working directory AND index they don't show up in plain "git diff"	03:02
hagabaka	ah, it says "not currently on any branch"	03:03
	not sure why that happened	03:03
	oh because I tried to checkout before	03:03
	ok, i started your steps with clean working directory this time, and still got fatal: you need to resolve your current index first	03:07
	← chieh left	03:08
	← Sonderblade left	03:09
uau	at which step?	03:09
	when you tried to execute the rebase -i?	03:09
hagabaka	"git checkout amended_commit_sha1 ."	03:09
	rebase -i got a conflict, i edited the file to remove the part between <<<< and =====	03:10
	should I make a commit there?	03:11
uau	that should not have been necessary	03:11
hagabaka	it gave me the same error before I did that too	03:11
uau	just ignore the confict message if any and write the checkout command	03:11
	i'm not sure why you'd get that error though	03:11
	← cbreak left	03:11
uau	are you sure you included the last '.'?	03:11
	← \ask_ left	03:13
	→ bfrog joined	03:13
	→ \ask_ joined	03:13
hagabaka	yeah	03:15
	tried again and still that	03:15
	← cilly left	03:15
uau	which git version do you have?	03:15
hagabaka	1.5.4.3	03:15
uau	i don't get any warnings when using the "git checkout SHA1 file/directory" syntax whatever state the tree is in	03:16
hagabaka	oh	03:16
	...	03:16
	i didn't include the .	03:16
	← johnw left	03:18
	→ JaapSuter joined	03:18
JaapSuter	Hello...	03:18
	In more traditional SCM systems, I can submit ("push") changes I've made to some files, without having to sync to the latest for other files. In Git, if the push isn't a fast forward, I need to sync, merge, and then push again.	03:19
	Correct?	03:19
	In what sense is this a workflow decision (i.e., it's the right thing to do), and in what sense a technical decision?	03:20
	Sub-modules provide a way around it, I suspect. But sub-module layout is very much an a-priori thing, whereas in Perforce I can be more fluid in what I do and do not want to sync.	03:21
uau	you mean you intentionally want to keep an old version of a part of the tree while working on another part? or what?	03:22
JaapSuter	Yes. Let's say I have a tree with two sub-directories, A and B.	03:22
	I work in B, and want to submit my changes.	03:23
	Meanwhile, other people have submitted stuff to A, things I don't want yet.	03:23
	Submodules and branching can help here I suspect. But the former requires a-priori planning (A and B need to be made submodules) and the latter won't work for people who fear branching.	03:23
	One may argue that people who fear branching shouldn't be using Git, but since I have both artists and programmers use it, there's only so much I can do.	03:24
	→ carrus85 joined	03:24
JaapSuter	I suspect I just need to build some clever branching porcelain around A and B to make this work.	03:24
spearce	JaapSuter: its a workflow decision in some respects, yes.	03:25
	git grew up as a source code system, not as a "source and doc and art and video and audio and ..." system.	03:25
JaapSuter	Spearce: for source-code that decision makes a lot of sense. After all, how are we supposed to know that disparate versions of A and B would work together without having tested them...	03:25
	Yeah, precisely.	03:25
spearce	generally in source code the whole thing needs to compile and link in some way, such that it all runs and does something that is perhaps useful.	03:25
JaapSuter	Bummer...	03:26
spearce	so the result of a merge really should be at least compile-tested, if not unit tested, integration tested, etc.	03:26
JaapSuter	Anybody else here trying to shove art and other things into Git or a Git-like system?	03:26
spearce	most recommendations have been to move the art people into a submodule, so they only have to merge the files they are responsible for.	03:26
	we did that at day-job (they aren't artists, but they aren't programmers either so i don't want them merging my source, but their assets are several million lines that we have to track and use).	03:27
	it took a little bit of user training, but i was mostly able to make them happy with git-gui and a few branches that they commit to, then push almost immediately.	03:28
	once they understood the idea of needing to fetch+merge (aka pull) the new changes before pushing again they said "i like that, i should see what everyone else has done before i make additional changes"	03:28
	of course their work isn't exactly isolated either, their stuff "compiles" and "links" too and some changes can cause that to fail.	03:29
JaapSuter	One scenario that doesn't work is for a tree with, say, a thousand art files that are all independent of each other (think video data). I want to publish my latest tweak to one video, without having to sync to all other changes that other users have made...	03:29
	← madewokherd left	03:30
uau	is perforce like CVS where history is really per-file and larger projects are just weakly tied groups of files with individual histories?	03:32
JaapSuter	Yes.	03:33
	Fully agreed that a tree-based content versioning system is more suitable for code.	03:33
	But also trying to come up with a creative way to get the best of both models...	03:33
spearce	but yea, its not as well suited to say single file tracking. can you put each file on its own branch, and thus in its own tree, but park them all in the same repository anyway?	03:34
	or do you really need them all checked out at once?	03:34
JaapSuter	We may need them all at once. It's not so much a practical scenario I have on a specific project, but an investigation into Git's benefits for a large number of different projects.	03:36
	I'm trying to reason what would happen if I would allow a non-fastforward merge on the server. Mmm, I'll have to draw this out.	03:37
	← elight left	03:37
rtyler	man, 10.1G VM, 3.3G resident	03:37
	→ EiNZTEiN joined	03:37
rtyler	this is fun to watch	03:37
uau	JaapSuter: if by that you mean allowing non-fastforward pushes from the clients	03:39
spearce	JaapSuter: the server side of git doesn't support such a concept. though you may be able to perform an uncontested merge on the server side, but i'm not immediately sure where you would plug that into the existing server side support without it being either a custom hack for your site in Git, or making the mailing list puke over the idea.	03:39
JaapSuter	uau: yes.	03:39
uau	the server won't do any "merge"	03:39
spearce	rtyler: what are you doing now? git repack -a -f ?	03:39
uau	a non-fastforward push just overwrites the previous state with the pushed one	03:39
JaapSuter	Ah, thanks guys.	03:40
rtyler	spearce: git-http-push	03:40
	still	03:40
JaapSuter	So ignoring for a moment that I'll have to endure people puking all over me, or go into Git and make some local hack (possibly through the use of hooks, so it's not "too" ugly), can anybody see a reason why it wouldn't work?	03:40
	I have: A- > B	03:41
rtyler	I'm watching the apache logs, full of MOVEs and PUTs	03:41
JaapSuter	Server has: A	03:41
	Correction, server has A -> C	03:41
	← litage left	03:41
JaapSuter	I push, non-fastforward, and server does a merge. This creates a new commit D.	03:41
	I have: A -> B	03:41
spearce	JaapSuter: oh, it would work. its just the point where it has to go on the server side doesn't support this sort of "lets do work behind just updating a pointer".	03:41
JaapSuter	Server has: A -> C -> D	03:42
	Mmm, yes, that might work.	03:42
spearce	at some point though you have to fetch from the server, get D (the new merge of B and C) and update your working directory to it.	03:43
JaapSuter	I can verify that the merge the server does is always a merge that only involves changes on one side only (i.e., binary files of which one side is the same as it is on the base).	03:43
	Spearce: yes, but that's no different than Perforce.	03:43
spearce	otherwise you'll have hundreds of branches running forever because the clients keep making new commits on their side branches, forcing the server to merge.	03:43
JaapSuter	The real issue is that the server has a commit that never really existed anywhere on anybody's clients machine. That's not very elegant.	03:43
uau	JaapSuter: also if you want that to match the result of a git merge	03:43
spearce	rtyler: http-push has that much memory used itself? i know very little about how it is implemented, but clearly it doesn't do large trees like yours well.	03:44
uau	the server should have something like {{A->-C}, {A->B)}->D	03:44
rtyler	PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND	03:44
	14997 tyler 26 10 10.1g 3.1g 7996 S 50 80.9 108:54.59 git-http-push	03:44
uau	with two parents for the commit D	03:44
rtyler	spearce: clearly :p	03:44
uau	since it is a merge	03:44
rtyler	spearce: I've heard nobody wants to touch that code though ;)	03:44
	I've added abpout 2G of extra swap via swap files to make sure I have enough swap >_>	03:45
spearce	rtyler: do you want to touch it? you can have it. its yours for the taking. :-)	03:45
JaapSuter	uau -> Ooh, yeah. Two parents. Should have drawn the diagram that way. Thanks.	03:45
	← rtomayko left	03:45
rtyler	spearce: I might take a gander at it, but I've not had time for open source code in....well, since I joined Slide ;)	03:45
	→ imyousuf joined	03:45
spearce	jgit may someday support push over webdav. but only if we can do it like the current sftp:// and amazon-s3:// where its only a few hundred lines total to plugin the vfs layer.	03:45
rtyler	it	03:46
	it'd be nice if git hooked into gnome-vfs	03:46
JaapSuter	Speaking of two parents. A more-than-two parent merge is still done as a sequence of three-way merges behind the curtains, right?	03:46
	rtyler ducks	03:46
spearce	but these dumb (non git aware anyway) protocols are annoying to work with for git clients because you can't do intelligent things in them.	03:46
	→ litage joined	03:46
thiago	http-push could use some CGIs to receive data and do something intelligent	03:47
spearce	JaapSuter: yes, its still a sequence of 3 way merges. but there can't be any conflicts. and n-way merges (where n > 2) aren't very common. we don't encourage them.	03:47
thiago	or Apache module, like SVN does	03:47
rtyler	apache modules are ick	03:47
JaapSuter	Given the rarity of > 3 octopus merges, would it really have been detrimental if the successive merges showed up explicitly in the history, instead of being collapsed?	03:47
spearce	thiago: yea, a CGI fronting for a modified git-receive-pack, or some optional routine in gitweb.cgi would be nice. john hawley wants to do something like that for the fetch side, but never has time among his other duties.	03:47
thiago	SOAP	03:48
	thiago runs	03:48
spearce	no. but its easier to say "git merge A B C D E" then to say "git merge A; git merge B, ..."	03:48
	→ rphillips joined	03:48
uau	JaapSuter: that's how git does the automatic merge result	03:48
JaapSuter	Afk, brb. Thanks!	03:48
spearce	SOAP so long as we can use <bit><value>true</value></bit> encoding. :-)	03:48
uau	but there is no intermediate tree state recorded	03:48
imyousuf	since we are at communication over http how about REST :)	03:49
uau	if you have a 8-way merge there are no intermediate states recorded anywhere to show	03:49
	the user might have created the merge result with his custom algorithm any way he wants	03:50
	← CodeWar left	03:51
	→ ssb joined	03:52
thiago	imyousuf: for REST, you need CGI or Apache module	03:54
imyousuf	Why thiago?	03:55
thiago	I have no idea how CGIs operate with PUT and DELETE commands	03:55
spearce	you can send PUT/DELETE to a CGI. at least apache will certainly do it.	03:55
	→ ABCD joined	03:55
thiago	then it should work	03:55
spearce	PUT with a pack shouldn't be too hard to receive on the server side.	03:55
imyousuf	We could make the server component out of java servlets	03:56
spearce	what's a #@!(!@(!)!@ pita is GET to fetch a pack. :)	03:56
thiago	PROPFIND	03:56
	or send in the HTTP headers the meta-data that it needs to generate the pack	03:56
imyousuf	JGit can be used from the servlets to receive pack and unpack at the repo location	03:57
spearce	the problem is the handshake we do between the two sides during a fetch streams commits at each other until both sides realize they have found the common merge base.	03:57
	you can't do that sort of streaming in a single HTTP connection.	03:57
	but for push, typically the client knows everything.	03:58
thiago	multiple roundtrips?	03:58
	→ ben_h joined	03:58
spearce	yea, multiple round trips would be required for an efficient fetch.	03:58
	doable, but nobody has wanted to write the code. :)	03:58
	← paltman_ left	03:58
thiago	how about an "inverted push"?	03:58
spearce	imyousuf: yup. most (if not all) of the support needed to do that is already available in jgit. just no servlet wrapper, or client support to speak it.	03:58
thiago	it will be less efficient, but it should work	03:59
	less roundtrips	03:59
spearce	right, you could do it in just the one request, at the expense of possibly fetching more than you needed to.	03:59
thiago	that's the price you pay for using HTTP :-)	03:59
spearce	though in practice if you are using standard style remotes with tracking branches in .git/refs/remotes/ you may actually get a fairly optimal pack anyway.	03:59
imyousuf	I would be grateful to contribute in the HTTP wrapper part :)	04:00
spearce	it would only be non-optimal in a star workflow, where Alice sends to Bob, Bob sends to Charley, and then Alice pulls from Charley.	04:00
imyousuf	I am experimenting with Jersey so I can put it to practice :)	04:00
spearce	you are experimenting an island in the uk?	04:01
thiago	wouldn't a normal push under those conditions also be slightly inefficient?	04:01
imyousuf	hehe	04:01
spearce	yes. which is why the "tell-me-more" push extension was recently added to c git.	04:01
imyousuf	thats not a star network rather sounds like token ring :)	04:01
spearce	i can't remember if it was a new feature in 1.5.6, or will be in 1.6.0.	04:01
cehteh	i just wanted to say that	04:01
imyousuf	and Jersey is the RI for JAX-RS (JSR-311) (jersey.dev.java.net)	04:02
cehteh	or daisy chain .. but when at tokenring .. can you pass a token around whuch optimize the packs?	04:02
spearce	only if the token is small enough and runs fast enough to not get stuck in the tubes. :)	04:02
imyousuf	:)	04:03
cehteh	thats the question .. what would such a token be in case of git?	04:03
spearce	half of tux with the git logo on his belly, with the roadrunner's legs spinning widely underneath?	04:03
cehteh	haha	04:04
spearce	err, wildly. i need to go back to code and not bs on irc. :)	04:04
cehteh	well better you do the bs here than in code :)	04:04
	cehteh things the inverted push could be a brillant idea ...	04:05
cehteh	k	04:05
	← chrisb left	04:05
spearce	its entirely reasonable. its a better attempt at something optimal than current http fetch, and will work nearly optimally for _most_ users watching a single (or two) upstreams.	04:06
	which is where most http usage is i suspect.	04:06
rtyler	holy crap, it's done	04:07
thiago	rtyler: are you sure it just didn't crash? :-)	04:07
rtyler	no >_>	04:08
	← ben_h left	04:08
rtyler	updating 'refs/heads/master'	04:08
	from 0000000000000000000000000000000000000000	04:08
	to 590b10bf5d069b86a55516fa5322b37606788962	04:08
	sending 185628 objects	04:08
	done	04:08
	rtyler takes a deep breath	04:08
	→ ben_h joined	04:08
rtyler	time to try a clone	04:08
JaapSuter	Thanks everybody. Friendly people in here. Salut!	04:08
thiago	only 185k objects? Are you tracking very large files?	04:08
	← JaapSuter left	04:08
spearce	rtyler: did it unpack 185k files as loose objects? or upload pack files?	04:09
rtyler	I'm assuming pack files	04:09
	→ spb_ joined	04:09
	← EiNZTEiN left	04:11
	→ CodeWar joined	04:12
	← ben_h left	04:12
	→ ben_h joined	04:13
	← spb left	04:13
	→ asdfdsfsd joined	04:13
	spb_ → spb	04:13
	→ tiglionabbit joined	04:20
tiglionabbit	how do I commit the fact that I've deleted something if it shows up under "Changed but not updated: deleted:" ?	04:20
	→ pombreda joined	04:20
	← irotas left	04:20
Eridius	tiglionabbit: git rm it	04:20
	or git add its directory	04:21
tiglionabbit	I tried both	04:21
	git add didn't do anything, git rm said I need to use the -r option	04:21
	do I really want to git rm -r?	04:21
Eridius	git rm -r is for deleting the contents of directories	04:21
	and git add will work if you add the enclosing directory (since that picks up all changes)	04:21
tiglionabbit	yeah, I don't want to do that	04:21
thiago	just do a git commit -a	04:21
tiglionabbit	huh. I wanted to save some of these changes for a later commit...	04:22
Eridius	so git rm it	04:22
	elliottcable → ec	04:22
	← ajonat left	04:22
Eridius	you could say `git ls-files -d \| xargs git rm`	04:23
	← Leefmc left	04:25
	→ ajonat joined	04:28
	← tjafk1 left	04:30
	→ tjafk1 joined	04:30
	← CodeWar left	04:31
	→ agib joined	04:32
	→ sgrimm joined	04:37
	← queso left	04:44
	→ ketralnis joined	04:47
	← ben_h left	04:50
	→ ben_h joined	04:51
	→ tiglionabbit_ joined	05:01
tiglionabbit_	is it possible to add an empty directory to a repository?	05:02
	it's a log directory, and I don't want to version control anything inside of it, but would be nice if it existed as the logger expects it to	05:02
Eridius	I recommend just dropping a .gitignore inside	05:03
tiglionabbit_	empty one?	05:04
Eridius	yeah. or maybe one that matches the log files that your logger will make	05:04
tiglionabbit_	I'd like it to match everything that could be in that directory then	05:05
	other than itself	05:05
Eridius	doesn't matter if it matches self	05:06
	gitignore is ignored for tracked files	05:06
	just put a * in it	05:06
	→ lu_zero joined	05:06
	← ABCD left	05:06
tiglionabbit_	gitignore only affects subdirectories?	05:06
	(or, siblings)	05:07
Eridius	yes. And files are matched against the patterns in the .gitignore as if the root was right there	05:07
	so a sibling foo will match /foo but not */foo	05:07
	← thiago left	05:07
Eridius	actually maybe it will match /foo, since can match nothing	05:07
	in any case, it matches /foo	05:08
tiglionabbit_	cool	05:08
	marc_vw → mvw	05:09
	← krawek left	05:10
	→ floppyears joined	05:10
jackdempsey	why use / in that case	05:11
	why not just 'foo'	05:11
	any reason?	05:11
Eridius	'/foo' matches a sibling foo, 'foo' will match a file named foo in this directory or any subdirectories	05:12
	so 'foo' matches bar/foo, but '/foo' doesn't	05:12
	→ FunkeeMonk joined	05:13
	→ warthog9 joined	05:14
	← joydivider left	05:15
	→ Ademan joined	05:16
jackdempsey	oh man, any subdirs huh	05:19
	didn't realize that	05:19
	i thought it was more that if you said blah.txt that would only match ./blah.txt not foo/bar/blah.txt	05:19
	basically recursive by nature it seems.....good to know, thx!	05:19
Eridius	if it has a / in it, it matches the whole path. if it doesn't, it matches just the filename	05:21
jackdempsey	yea. kinda cool	05:21
	so foo/* will only match it at ./foo	05:22
	if you had bar/foo/baz	05:22
	baz won't be ignored	05:22
	→ daaku joined	05:26
	→ neynenmo joined	05:26
	← litage left	05:33
	← sbahra left	05:36
	← bryanray left	05:42
	← bobesponja left	05:43
	← Ademan left	05:44
	→ Ademan joined	05:45
	← igorgue left	05:50
	← neynenmo left	05:51
	→ sris joined	05:55
	→ paulttt joined	05:58
	→ BabelO joined	05:59
	← asdfdsfsd left	06:03
	→ deavid joined	06:13
	→ nud joined	06:15
	→ charon joined	06:15
	← sris left	06:22
	→ chieh joined	06:24
	← markkalderon left	06:28
	← morphir left	06:28
	jackdempsey → jack_sleep	06:33
	pilif → pilif\|away	06:35
	← jack_sleep left	06:35
	← chieh left	06:36
	← Ademan left	06:39
	→ Ademan joined	06:40
	→ reel_ joined	06:41
	→ crab joined	06:43
crab	did anything come of the proposal to make "git branch" able to switch branches too?	06:44
	→ icwiener joined	06:46
	← icwiener left	06:49
	← nud left	06:51
	→ Rustre joined	06:51
	← spearce left	06:52
	→ hein2 joined	06:52
	← Sho_ left	06:52
	hein2 → Sho_	06:52
	← bfrog left	06:55
	← Ademan left	06:57
	→ huyx joined	06:58
	→ jackdempsey joined	07:00
	VVSiz_ → VVSiz	07:11
	← daaku left	07:11
	→ Ademan joined	07:12
	→ tiglionabbit__ joined	07:12
	← jackdempsey left	07:13
	← ben_h left	07:17
	← SuttoL left	07:21
	→ nud joined	07:21
	→ barak joined	07:22
	← FunkeeMonk left	07:24
	← tiglionabbit_ left	07:24
	← CodeOfficer left	07:26
barak	I've used git-cvsimport to create my own git repo to deal with a project that uses CVS. Pushing my mods back from my git repo to the upstream CVS repo using git-cvsexportcommit. Which works fine unless there is a keyword mismatch, i.e., some file in git has $Id: foo$ while in CVS keyword expansion somehow changed it to $Id: bar$. In this case git-cvsexportcommit barfs with a patch failure, and I end up using git	07:26
	-format-patch to generate a patch file and integrate it manually. BLECH! Any advise?	07:26
	sed s/advise/advice/	07:27
	← Cope left	07:30
cebewee	barak: use -kk for the cvs checkout, -k for git-cvsimportcommit	07:30
	hm, we need an option for git-cvsexportcommit to do updates -kk, also	07:31
	otherwise, files added by cvsexportcommit wil not have -kk set. therefore i end up occasinally checking out a new cvs copy ...	07:33
	→ joevandyk_ joined	07:34
	← brosner left	07:35
	← Foriskak left	07:39
barak	Best would be if the patch-applier used by git-cvsexportcommit would do the dirty work of ignoring keyword mismatches for files that do not have keywords disabled in CVS. It seems like it is in the best position to solve the problem - everything else is hacking around to avoid it, and gets kind of brittle.	07:40
	So what I'm hoping is that there is an option to git-cvsexportcommit that does this, but it didn't make it into the documentation, so someone will tell me. Or if that fortuitous occurance does not come to pass, some other workaround. Best so far seems to be to use -kk and re-do my cvsimport to just avoid all keywords altogether. (There is a minor issue with this, in that the build system may barf with blank keywor	07:42
	ds, but that could be worked around.)	07:42
pasky	just as cebewee said?	07:42
	→ davidmccabe joined	07:43
davidmccabe	So I just got myself into a goofy situation:	07:43
barak	Yes exactly, that's what i meant.	07:43
	(THanks cebewee)	07:43
davidmccabe	At time t, two git repos have identical contents, but they are not related as far as git knows.	07:43
	← tiglionabbit__ left	07:44
davidmccabe	Several commits are made to one of them. Can these commits be remade into the other?	07:44
	← mithro left	07:45
crab	git-format-patch?	07:46
	← sverrej left	07:46
	→ CodeWar joined	07:46
	→ mithro joined	07:48
barak	Yes, when this problem pops up I use git-format-patch and apply the failing patch manually and manually respolve the keyword expansion mismatch. Blech.	07:48
	→ Mishakun joined	07:48
crab	barak: sorry, i meant to answer davidmccabe	07:48
davidmccabe	Thanks crab.	07:48
barak	sorry.	07:49
davidmccabe	Thanks barak.	07:49
	Is there a way to stage or commit just one patch band from a file?	07:50
	→ joydivider joined	07:51
pasky	patch band?	07:53
davidmccabe	insertion or deletion.	07:54
pasky	sax solo for sigsegv!	07:54
	hmm, well, I'm wondering if the staging features could be used for this somehow	07:55
	but I'd just manually edit the patch ;)	07:55
	you might find it perhaps easier to throw both repos into two branch of a single repo and then rebase	07:56
	→ joevandyk joined	07:59
	← huyx left	07:59
	→ Fullmoon joined	08:01
fujin	davidmccabe: can't you git-add -i now?	08:05
	interactive line by line	08:06
	or is that a to-be-added feature, i forget	08:06
	← joevandyk_ left	08:06
comp	nope, it works already	08:06
davidmccabe	Super.	08:07
	Ok, one last question for the night:	08:07
	How do I revert a file back to how it is in another branch?	08:07
	other than checking out the branch, copying, etc.	08:07
	→ redondos joined	08:08
davidmccabe	I guess I should just checkout the branch, copy, etc.... :)	08:10
crab	git checkout branch -- /path/to/file ?	08:11
davidmccabe	Thanks anyways.	08:13
	← davidmccabe left	08:13
	← crab left	08:13
	→ sverrej joined	08:14
Rustre	hello there	08:14
	do i need to do anything specific to "work" on the latest changes?	08:15
	say, i did a "git clone <url>"	08:15
	now, i want to update my local "clone", so i did a "git pull"	08:15
	if i do a "git show" i can see changes	08:15
	and the latest commits and comments and everything	08:16
	but the changes weren't made to my local files it seems	08:16
	do i need to do anything special? maybe a "git checkout" or something else?	08:16
	→ bentob0x joined	08:16
GarethAdams	Rustre: what does gitk --all show you?	08:17
Rustre	i can't use gitk, i'm on a remote linux box (using ssh)	08:17
	from a windows machine :P	08:17
	so there's no X export :P	08:18
	→ earcar joined	08:18
Rustre	any other equivalent for this command not using gitk?	08:20
barak	Just tried "cvs checkout -kk" when avoided all keyword expansion, then "git-cvsimport -k", and guess what? The resulting files in git had keywords expanded! That was with git 1.4.4.4, I'm retrying now with git 1.5.6 ... same thing, keywords are expanded. Perhaps the -k switch on git-cvsimport is just plain busted? Or am I doing something wrong?	08:20
Rustre	when i do a "git checkout master" i can see four lines (the four latest file changes from the url i'm pulling) with a "M" in front	08:22
	does that mean that i need to stage or commit or whatever?	08:22
jengelh	Rustre: git-forest	08:22
Rustre	?	08:23
jengelh	if you need the history but don't have gitk at hand	08:23
Rustre	oh ...	08:23
jengelh	M is modified	08:24
Rustre	but modified where?	08:24
jengelh	wc	08:24
Rustre	in my working tree? in the branch?	08:24
jengelh	It reminds you that you have modified these files.	08:26
Rustre	oh	08:26
	and if i want to "forget" those modifications?	08:27
jengelh	reset --hard	08:27
Rustre	and "squeeze" them with the latest changes from the url i'm pulling from?	08:27
	← sverrej left	08:27
Rustre	oh... so "git pull" isn't actually updating my local copy	08:27
jengelh	don't use pull if you don't know what's behind it	08:27
	→ reel__ joined	08:27
Rustre	i know what is :D	08:27
	→ loupgaroublond joined	08:28
Rustre	i modified it on my windows machine, then pushed to the linux box	08:28
	→ tiglionabbit_ joined	08:28
loupgaroublond	hi	08:28
Rustre	now i'd like my changes to be applied on the linux box	08:28
	but not quite sure how i should do that	08:28
loupgaroublond	is there a way to only partially stage changes to git-stash, like i can do with git-add --interactive ?	08:28
	→ mtodd__ joined	08:30
	→ _zack joined	08:30
	← macGerhard left	08:30
	← mtodd_ left	08:32
	← CodeWar left	08:32
	→ johnw joined	08:33
	→ fhobia joined	08:34
Rustre	is there a way to "pin" a file, and prevent it from being changed when pulling/reseting or whatever?	08:42
	i'm working on a web site, and i'd like one particular file (connect.php) to be unique to the web server, and of course, different than the one i'm working on (on the development computer)	08:43
	← reel_ left	08:43
Rustre	but still, i'd like this file to be in the repository (so when i clone the repo on the web server, i just have to manually modify the connect.php file once and for all)	08:43
johnw	why not modify connect.php in a branch	08:46
	and then just use git-rebase to bring in new changes from the master	08:46
	← Fullmoon left	08:46
johnw	I assume master will just contain the master template for connect.php	08:46
Rustre	i have no idea how and why to use git-rebase, but i'll check that straight away, thanks	08:47
johnw	if your branch contains only a single change (that to connect.php), then rebase moves that change forward in time, so that the branch only ever contains a single commit	08:47
	this is opposed to merging, which introduces "merge commits" every time you want to update to the latest code	08:48
	in other words, it's like a poor man's Stacked Git, or "quilt", if those mean anything to you	08:48
Rustre	(that doesn't mean anything to me, sorry :/)	08:48
	so, in other words	08:48
	the workflow would be :	08:48
johnw	in your branch, just type "git rebase master", to bring in changes from your master	08:48
Rustre	clone the repo from the development PC => work on the development PC => push to the webserver repo => connect on the webserver, and do a "git rebase master"	08:49
	is that all?	08:49
johnw	yep	08:49
	it's the mirror of "git merge master"	08:50
	but there are consequences to doing a rebase. one of the most important is: don't do it on any branch others can see	08:50
	→ _slacker_ joined	08:50
johnw	it's for private stuff that only you work with, in that case is perfectly legitimate	08:50
Rustre	ok	08:50
	i dont quite get why though	08:50
	→ mcella joined	08:50
Rustre	i'll read the man page on git rebase	08:50
johnw	because it rewrites history	08:51
Rustre	to understand what it means	08:51
johnw	i talk about it here: http://www.newartisans.com/blog_files/git.from.bottom.up.php	08:51
	one of the sections in that document is all about rebase vs. merge	08:51
Rustre	oh, thanks a lot, i'll read that then :D	08:52
	→ SuttoL joined	08:54
	→ markkalderon joined	08:57
Rustre	johnw: for my particular case, maybe i'd be better off adding a "excludesfile" in the ".git/config" file on the webserver once the clone has been done and the connect.php changed?	08:57
johnw	it might be much easier for you	08:58
	but i have not used that feature yet	08:58
Rustre	i'll give it a try then, thanks ;)	08:58
	← ben`afk left	08:59
	← paulttt left	09:00
	← carrus85 left	09:02
pasky	uh	09:03
	excludesfile?	09:03
Rustre	yup	09:03
	there's something written about that in the "gitignore" man page	09:03
pasky	that's just .git/info/exclude, no?	09:03
Rustre	it's a bit different it seems	09:03
	it does the same thing (maybe)	09:04
pasky	why would you think it's a bit different?	09:04
	(either way, that's useless for you - read the description carefully word by word:)	09:05
	A `gitignore` file specifies intentionally untracked files that	09:05
	git should ignore.	09:05
Rustre	then i'm back to the start :/	09:06
	dammit	09:06
pasky	what's wrong with going with johnw's suggestion?	09:06
Rustre	probably nothing	09:07
	i still need to go into it (because i didn't get a thing about rebase)	09:07
pasky	rebase takes commits you have extra on your branch compared to another branch and rebases them to the other branch - that is, just grabs the other and recommits them on top of it	09:08
	and there's a shorthand to call git pull --rebase	09:08
	so, in the webserver repo, you just commit your local change	09:08
	and then keep doing git pull --rebase in the future	09:08
loupgaroublond	does it keep those changes labelled as being on your local branch?	09:08
	→ jdq joined	09:09
pasky	labelled?	09:09
	I don't understand	09:09
jengelh	meh	09:09
	gitweb is broken	09:09
pasky	the usual state	09:09
Bombe	Good morning.	09:09
jengelh	entering a commit id in the serach field won't get me anywhere	09:09
pasky	you put that to the url right away ;)	09:09
loupgaroublond	say, i have 3 patches on origin:master in a branch foo	09:09
pasky	hello bombe	09:09
	loupgaroublond: sorry, you lost me right there	09:10
	loupgaroublond: or, you mean branch foo is origin:master + 3 patches?	09:10
loupgaroublond	yes	09:10
pasky	origin/master, that is	09:10
	ok	09:10
loupgaroublond	so then someone pushes other changes to origin/master	09:10
	→ cilly joined	09:10
	→ koke joined	09:11
loupgaroublond	you need to pull of course, does rebase keep your three patches in a local branch or does it end up combining the two?	09:11
pasky	uh	09:12
loupgaroublond	but i have to run now	09:12
pasky	hm	09:12
loupgaroublond	i'll play with it later	09:12
pasky	ok	09:12
	→ carrus85 joined	09:12
	← SuttoL left	09:14
	→ CodeWar joined	09:19
	→ ball_licker joined	09:21
ball_licker	I put on	09:22
	→ Fullmoon joined	09:22
ball_licker	East side, south side, west side....I put on	09:22
	jezzy baby	09:22
	jizzy	09:23
pasky	o.O	09:23
	← pflanze left	09:25
	→ pflanze joined	09:26
	→ priidu joined	09:27
	ball_licker → level2tech	09:28
Rustre	johnw: mmm i read part of your pdf, but i'm still a bit lost :/	09:28
	→ njal joined	09:30
	→ mstormo joined	09:31
	→ pombreda1 joined	09:31
	→ SuttoL joined	09:31
	← markkalderon left	09:33
johnw	rebase carries forward a set of commits as if it were a series of patches	09:33
	rewriting the history each time to make it appear as though those patches/commits had just been applied against the new HEAD	09:33
	you know, it's the kind of thing where you can say it a hundred times, but until you try it out, it won't ever make complete sense	09:33
Bombe	:)	09:35
	level2tech → level2tech_	09:35
	→ cardioid joined	09:36
	← johnw left	09:36
	level2tech_ → level2tech	09:36
	← level2tech left	09:36
	→ charon_ joined	09:41
	← charon left	09:41
	charon_ → charon	09:41
	← pombreda left	09:45
	→ chris2 joined	09:52
	← joydivider left	09:54
	← pombreda1 left	09:54
	← _zack left	09:59
	→ pombreda joined	09:59
	→ _zack joined	09:59
	← _zack left	10:00
	→ d0k joined	10:01
	→ SRabbelier joined	10:01
	→ _zack joined	10:03
	→ markkalderon joined	10:07
	→ dramsay joined	10:09
jengelh	git clone git://git.kernel.org/pub/scm/linux/kernel/torvalds/linux-2.6.git	10:09
	fatal: The remote end hung up unexpectedly	10:09
	something I dunno about?	10:09
	ah a missing git :p	10:09
UukGoblin	is there some nice way to cherry-pick only a few changes from a commit?	10:12
wildfire	UukGoblin, cherry-pick that commit as normal, then git reset --hard HEAD^ (i.e. go to the prior change)	10:13
	UukGoblin, then 'git add -i' might be what you are after	10:13
UukGoblin	wildfire, oh, cool, will check it out, thanks	10:13
loupgaroublond	don't you mean reset --soft ?	10:14
	--hard will modify the working tree	10:14
UukGoblin	yeah, probably	10:14
gitsky	but --soft on the other hand will touch only your HEAD, no?	10:15
	so just git reset HEAD^ (or git reset --mixed HEAD^, if you want) should work best ;)	10:16
	← CodeWar left	10:16
wildfire	loupgaroublond, gitsky, yes, of course - my bad	10:16
gitsky	and perhaps a simpler way is to	10:16
UukGoblin	I get the idea, get the changes into the working tree and then add them interactively to index	10:16
gitsky	git checkout commitid:filename	10:17
	instead of cherry-pick + reset	10:17
	er, make that s/:/ /	10:17
UukGoblin	hmmm, that sounds nice...	10:17
gitsky	oh no	10:17
	it will modify the index too -.-	10:17
UukGoblin	ah, yes	10:17
gitsky	so git cat-file blob commitid:filename >filename, I guess	10:18
wildfire	perhaps a 'git-cherry-pick -n <commitid>' solves the issue better?	10:18
	no need to reset	10:19
UukGoblin	wildfire, git-cherry-pick -n puts the change into index...	10:19
wildfire	ahh, the man page isn't clear on that	10:19
UukGoblin	cating the file seems ok, I just wonder what I'd do if I only wanted to take some of the changes to a file ;-)	10:20
gitsky	wait	10:20
	that's exactly what are we trying to solve	10:21
	modifying only your working tree so that you can git add -i	10:21
	and select what you want to put into the index	10:21
UukGoblin	correct	10:21
gitsky	if you are after the whole file, the git checkout commitid filename is the simplest way	10:21
UukGoblin	you mean git cat-file	10:22
gitsky	no	10:22
	the point is that git add -i interactively adds files from _working tree_ to _index_	10:22
UukGoblin	ah, because if I have the whole file I don't mind it will end up in index	10:22
gitsky	so that you can put only bits of the file into the index	10:23
	yes, exactly	10:23
	if you just want the whole file, might as well just throw it into the index right away	10:23
UukGoblin	yup	10:23
gitsky	anyway, at least we discovered that git-cherry-pick documentation is apparently buggy	10:24
	(did you test it, so that you know it puts the change into index?)	10:24
	the code looks like it does :(	10:24
UukGoblin	yeah, I just tried doing that, and according to gitk the change ended up in my index	10:24
gitsky	hmm, it pretty much has to, since cherry-pick in fact does a three-way merge	10:25
	I'll send a patch	10:25
UukGoblin	cool, thanks	10:25
	hm, actually, it seems that git checkout [<treeish>] <paths> also checks out into index, not only working tree	10:28
	the docs says "It updates the named paths in the working tree from the index file", I guess that can mean that it updates both	10:28
	→ lea_w joined	10:29
	→ thommey joined	10:30
gitsky	12:13 < gitsky> it will modify the index too -.-	10:30
	that was my point	10:30
	that's why you need to use git cat-file if you want to modify only the working tree	10:31
UukGoblin	yes, I was just talking about the manpage for that	10:31
	but I think it's better than cherry-pick	10:32
	→ whee joined	10:32
UukGoblin	and it says that it actually does git-checkout-index, so you can figure out things from that	10:32
gitsky	<tree-ish> argument can be	10:33
	used to specify a specific tree-ish (i.e. commit, tag or tree)	10:33
	to update the index for the given paths before updating the	10:33
	working tree.	10:33
UukGoblin	yup	10:33
gitsky	that's the "commitid" in my call	10:33
UukGoblin	it's okay then	10:33
gitsky	so it's hopefully fairly clear	10:33
	→ cbreak joined	10:35
	← fhobia left	10:37
telmich	the idea behind gitosis is to have just one shell account, but that has full shell access on the system, correct?	10:37
	→ mtkd joined	10:44
	← cbreak left	10:50
	→ cbreak joined	10:50
	→ powr-toc joined	10:52
	→ harinath joined	10:54
	← reel__ left	10:54
	→ reel__ joined	10:54
	← mstormo left	10:54
	GeckoPie\|Sleep → GeckoPie	10:55
	→ mstormo joined	10:57
	← dramsay left	11:06
gitsky	telmich: i don't think so...? i thought gitosis provided its own shell?	11:08
	i never used it, though	11:08
	← Fullmoon left	11:12
	← reel__ left	11:13
	← njal left	11:13
	→ reel__ joined	11:14
UukGoblin	git's awesome. I don't know how I could have lived without it ;-)	11:14
	I just split a line of development into two branches :-)	11:14
	bit by bit :-)	11:14
	← imyousuf left	11:15
	→ njal joined	11:15
	→ rlaing joined	11:24
telmich	gitsky: I read a howto that uses /bin/sh	11:24
	gitsky: http://scie.nti.st/2007/11/14/hosting-git-repositories-the-easy-and-secure-way	11:25
rlaing	hi. question: if i modify .gitignore after a first commit, how to update index to delete files now in ignore?	11:25
telmich	rlaing: do it manually	11:26
	rlaing: git-rm <files>	11:26
rlaing	git-update-index --delete filename?	11:26
	telmich: i don't want to remove file from working copy	11:26
	only from index	11:26
telmich	rlaing: git-rm --cached	11:27
rlaing	ok, thank you	11:27
	← koke left	11:28
	← rlaing left	11:28
	→ Fullmoon joined	11:31
	→ Leefmc joined	11:33
	→ janm joined	11:40
ndim	Oh my. How did I survive without "git rebase -i"?	11:41
cbreak	I don't even use branches... :)	11:41
telmich	ndim: hmm, I really tend to never use rebase, I must confess	11:42
spb	i tend to use rebase before push, just to avoid merges	11:43
	but that's about it	11:43
ndim	I find it awfully useful for developing private feature branches and nicely formatting them before pushing.	11:43
telmich	ndim: hmm, can you describle your workflow? as so many people seem to use it, I may want to have a deeper look at it	11:44
ndim	telmich: I have a master branch and add some feature consisting of a bunch of commits	11:45
	telmich: Then I create a feature-foo branch, and start committing commits A, B, C, D, E.	11:46
	telmich: Now I find that I forgot to cross a t in A, or something. I commit that now just like any patch.	11:47
	telmich: Now (or later), I can then "git rebase -i", and pick A, squash A-fix into it, and pick B,C,D,E after that.	11:47
	In case E happens to basically be a fix to B, I can squash E and B into one commit as well, while I am at it.	11:48
	→ SuttoL_ joined	11:49
ndim	Results in less clutter and thus better readability after publication.	11:49
telmich	ndim: ahh, ok	11:49
	hmpf, must go.. ndim: sounds interesting, will commen on it later	11:50
ndim	telmich: When the feature-foo branch is OK, I will then push those commits to the public master branch	11:50
	← raz left	11:53
	→ raz joined	11:53
	← Leefmc left	11:55
	→ Aikawa joined	12:01
	→ twicthy joined	12:03
twicthy	Hi. I got: permission denied (publickey). The remote end hung up unexpecdedly.. when I tried an: git push origin master to github.com	12:05
SRabbelier	gitsky: do you change your nick every 2 days or what? :P	12:06
	← SuttoL left	12:06
	→ emachado joined	12:07
gitsky	do you have to pick on me three times every day or what? :P	12:08
	;)	12:08
	actually my nick on freenode is usually quite stable	12:08
	but just was reading some very old #git logs tonight and got nostalgic ;)	12:08
SRabbelier	lol :D	12:09
	→ pdherbemont joined	12:12
	← X-Blade left	12:14
	← Fullmoon left	12:14
	→ gerke joined	12:17
	→ lorandi joined	12:17
	→ kukks joined	12:18
	→ X-Blade joined	12:19
	→ sbahra joined	12:19
	→ Yuuhi joined	12:21
	← sbahra left	12:22
	→ koke joined	12:22
	→ MarcWeber joined	12:22
Rustre	is there a way to ignore changes to a file that have previously been tracked?	12:24
	→ [raz] joined	12:24
	→ Xyri0 joined	12:24
	→ sbahra joined	12:25
Rustre	i'm having automatically generated files, and it seems i can't ignore them through the .gitignore file (they still show up as "to stage")	12:25
	and i'm guessing it's because i've added them previously	12:25
MarcWeber	Is there already a script which helps keeping a merge up to date? Say you have 3 branches not differing much. Now I'd like to create another branch merging them all. And whenever I add a patch to one of those 3 branches I'd like to merge those branches again. Or is there a better workflow?	12:25
Rustre	should i use git-rm ?	12:25
	→ dduncan joined	12:25
MarcWeber	Rustre: There is another command which let's you only remove them from index (thus beein something like undo of git-add).. but I can't recall it..	12:26
	Using git-gui it's easy to do as well	12:26
Rustre	how?	12:26
	i'm using git-gui, but couldn't spot this one	12:26
[raz]	hmm i'm confused	12:27
MarcWeber	Rustre: Just click on the icon on the left ni the lower window	12:27
[raz]	i just pushed from a shallow clone without problems.	12:27
	has the restriction (no push from shallow copy) been lifted recently?	12:28
Rustre	MarcWeber: which "icon on the left" ? i have no icons at all :P	12:28
	what version are you talking about? maybe i'm using a different one?	12:28
gitsky	[raz]: there was any restriction at all?	12:28
MarcWeber	git-gui version 0.10.2	12:29
gitsky	I wouldn't see any reason for a restriction at all as long as you have all the necessary objects to determine what to push	12:29
[raz]	gitsky, the man-page says: A shallow repository has a number of limitations (you cannot clone or fetch from it, nor push from nor into it)	12:29
Rustre	i'm using 0.10.1, maybe that's why	12:29
	← dduncan left	12:29
Rustre	i'll give it a go at upgrading my package then (if possible, i'm using msys git)	12:29
	→ ben_h joined	12:30
	← raz left	12:30
	[raz] → raz	12:30
	→ gitte joined	12:30
	← pygi left	12:30
raz	gitsky, well, i'm happy that it works. just a bit surprised	12:30
	→ pygi joined	12:30
MarcWeber	Rustre: http://mawercer.de/test.jpg	12:30
Rustre	url isn't working, is it?	12:31
MarcWeber	no it's not sry	12:31
Rustre	can't find mawercer.de	12:31
MarcWeber	I've stopped apache	12:31
	→ EmilMedve joined	12:31
MarcWeber	Rustre: retry	12:31
gitsky	what manpage says that?	12:31
	i see that only in the release notes for 1.5.0	12:31
	where it's just said that it is not expdeected to work, and the release notes are pretty old stuff	12:31
Rustre	yup	12:31
	i definitely don't have the same gui	12:31
	oh, wait MarcWeber, i'm getting what you're talking about	12:32
	i don't want to "unstage" the file	12:32
	i'd like to ignore it	12:32
	and not track it anymore	12:32
MarcWeber	Then echo a >> .gitignore should do the trick	12:32
	→ foca_ joined	12:33
telmich	re	12:33
MarcWeber	Rustre: Unstage it once, then it will not appear again	12:33
	→ lea_w1 joined	12:33
	← lea_w left	12:33
Rustre	well, it will	12:33
	of course	12:33
	← Xyrio left	12:34
	lea_w1 → lea_w	12:34
	Xyri0 → Xyrio	12:34
Rustre	if you unstage it, it just means "i modified it, but i'm not commiting it (yet)"	12:34
	but it'll still displayed as "unstaged changes"	12:34
telmich	ndim: ahh, ok.. I push all stuff out to the master branch, because I don't care if other people see my mistakes	12:34
MarcWeber	Rustre: Your problem is that .gitignore is ignored, right?	12:34
Rustre	MarcWeber: and the .gitignore doesn't seem to work for files that i've already commited previously	12:34
	yup	12:34
	it's not ignore, it's just not working for THAT file (which is in the repository, i'm guessing that's why)	12:35
	i guess i should use git-rm	12:35
MarcWeber	Then I can't help you. After adding a to .ignore git-gui still showd the file as staged. After unstaging it it vanished.	12:35
	git-rm removes the whole filed, doesn't it?	12:35
telmich	Rustre: sure, if it is in the index, gitignore will not ignore it	12:35
	Rustre: just do git-rm --cached <file>	12:36
Rustre	oh ok telmich	12:36
	that's what i should do then, thanks a lot MarcWeber and telmich	12:36
telmich	Rustre: no problem, have fun with git ;-)	12:36
	→ Arc_ joined	12:37
MarcWeber	I've missed --cached this time.. Will try to remember it	12:38
Rustre	i'm having fun for some time now already telmich	12:38
	it's just that i'm learning how to use it a bit more each day ;)	12:38
	(sometimes in pain :D)	12:38
	← sbahra left	12:39
	→ deavidsedice joined	12:39
ndim	telmich: Seeing mistakes is one aspect having to read two patches where the second one removes 80% of the first one is just wasting time, though.	12:42
	→ sbahra joined	12:42
Rustre	that's strange : i'm trying to push my changes to the "webserver repository", and i have a rejected : "master -> master (non-fast forward)"	12:43
	how come master to master isn't fast forward?	12:44
telmich	ndim: that's true, but I normally just do "git-diff mybranch other/branch" and I do not care about the history	12:44
	Rustre: did you use git-rebase?	12:44
Rustre	yup	12:45
	→ DAddYE joined	12:45
Rustre	not sure i did it correctly though	12:45
DAddYE	Hello big world	12:45
	any can help me? I need to make an hooks that send to specific emails the last commit message, any can help me to do this?	12:45
Rustre	i had a strange message (something like "needed single ref" or something like that)	12:45
	still, if i connect via ssh to the linux box (the webserver) and do a git-pull, it works perfect (it seems at least)	12:46
	← twicthy left	12:48
	← Arc left	12:48
	← foca left	12:50
	→ thiago joined	12:51
	→ dramsay joined	12:51
Lemurnomicon	anyone have any pointers / experience / "ZOMGDONTDOIT" tales on using git to keep a version history of executables?	12:56
Rustre	i think having rebased my repository is now preventing me from doing a few things (for example "git-rm" on a file i don't want to track anymore)	12:56
DAddYE	any can help me? I need to make an hooks that send to specific emails the last commit message, any can help me to do this?	12:59
	← tiglionabbit_ left	12:59
	← dramsay left	13:01
	→ dramsay joined	13:01
	← heycarsten left	13:03
ndim	DAddYE: I'd start with grep .git/hooks/*	13:04
	DAddYE: I'd start with "grep mail .git/hooks/*"	13:04
DAddYE	???	13:05
ndim	DAddYE: There is some mail hook in there	13:05
DAddYE	mmm	13:05
	→ LiamH joined	13:06
ndim	Rustre: When I am confused (and the repo has less than 10000 commits), "gitk --all" usually gives me back the orientation.	13:07
Rustre	can't use gitk, it's a remote linuxbox :s	13:08
DAddYE	ndim: there is nothing usefull for me	13:08
	→ Pistos joined	13:08
Pistos	I tried to STW for this, but I couldn't find it. I'd like to retrieve the last copy of a file before it was git rm'ed many commits ago. How to?	13:09
	I don't mean that I just locally rm'ed it and haven't checked it in, or any case like that.	13:09
	← gerke left	13:09
	→ FunkeeMonk joined	13:09
	→ pombreda1 joined	13:09
Pistos	Assume I don't have the SHA1 of the commit that rmed the file.	13:09
gitte	git show $(git log -1 --pretty=%H -- <file>)^:<file>	13:10
Pistos	gitte: Thanks; will try.	13:10
gitsky	lol, git show accepts this syntax? :)	13:11
DAddYE	Any know if there is an hooks for send only commit message to some email address????	13:11
	Any know if there is an hook for send only commit message to some email address????	13:11
Pistos	fatal: invalid --pretty format: %H	13:11
gitsky	these new user-friendly git tools...	13:11
	we didn't have those at my time!	13:11
gitte	gitsky: You mean the $() thing? No, that is bash, as you surely know...	13:11
gitsky	gitte: no, the :-form	13:11
gitte	gitsky: yep, <treeish>:<path>.	13:12
	Pistos: sorry, forgot: "--pretty=format:%H"	13:12
Pistos	Thanks. :)	13:12
gitsky	so, git show can show other objects than commits	13:12
	that's actually the news for me	13:12
gitte	gitsky: yes, we can	13:12
	;-)	13:13
	← lea_w left	13:13
Pistos	gitte: Stupendous. That worked marvelously, minus me having to specify the full path relative to the repo root.	13:13
	Next question: How is a git noob supposed to parse, understand or even remember a command like that? ;)	13:14
	→ brosner joined	13:14
ndim	Pistos: "git log -1 -- <file>" gives you the latest log entry touching <file>	13:14
Pistos	Okay.	13:15
ndim	And that should be the commit which removed the file.	13:15
	Pistos reads git log manpage.	13:15
ndim	--pretty=%H probably just prints the commit hash, then append a ^ to get the parent commit, i.e. the last commit where <file> existed.	13:15
	→ schacon joined	13:16
ndim	Then call "git show" on it.	13:16
gitte	Pistos: a noob is supposed to use "git log -- <file>", note the name, and then "git show" its parent's <file>.	13:16
Pistos	ndim: Great explanation. I grok.	13:16
gitte	ndim: except it is --pretty=format:%H	13:16
jengelh	except you could just use git-rev-parse :p	13:16
ndim	Pistos: I had no idea of that command, but that is how I would do it :)	13:16
Pistos	jengelh: Do tell. I was just at that man page, and couldn't easily surmise how to approach my problem with it.	13:17
	← pombreda left	13:17
	→ CareBear\ joined	13:17
CareBear\	hello friends	13:18
gitte	jengelh: except rev-parse is not meant for _humans_!!!	13:18
Pistos	So I pass "SHA1:filepath" to git show to make it give me the file for that snapshot?	13:18
	→ wereHamster joined	13:18
gitte	jengelh: except rev-parse does not find the commit which last touched the file!!!	13:18
wereHamster	does git export any functions that could be used from a C application?	13:18
gitte	wereHamster: yes.	13:19
wereHamster	that would the the 'plumbing' ..	13:19
Pistos	I'd be shocked if it couldn't. :)	13:19
wereHamster	is there a documentation?	13:19
jengelh	heh	13:19
gitte	wereHamster: no.	13:20
CareBear\	I used git-clone, made changes in the clone, committed, pushed, but then the origin (master?) thinks the reverse of those changes are modifications to HEAD. I don't want that.	13:20
jengelh	gitte: in that case, git-rev-list	13:20
gitte	jengelh: sorry, but you are the third guy in the last few days suggesting using plumbing to users. Unsurprisingly, users find Git "complicated".	13:20
Pistos	CareBear\: Non-bare repo?	13:20
CareBear\	What did I do wrong - or what have I not done yet?	13:20
wereHamster	gitte: intentionally or did noone write it yet?	13:20
	← garrin left	13:20
gitte	jengelh: same problem!!!!!!	13:21
jengelh	git-rev-list HEAD Makefile \| head -n1	13:21
gitte	wereHamster: patches welcome.	13:21
Pistos	CareBear\: I think it is a git sin to push to a "clothed" repo.	13:21
jengelh	it may be plumbing, but it's better than trying to grep in human output	13:21
Pistos	CareBear\: If the place you pushed to is a WC of someone, that place should pull from you instead.	13:21
gitte	jengelh: stop making it complicated for users.	13:21
CareBear\	Pistos : The master is bare, ie. just created with git init, added, committed, made some changes, committed again then the clone	13:21
gitte	jengelh: puhlease.	13:21
Pistos	But I'm just a git noob. :)	13:21
wereHamster	gitte: I'm interested in removing all the system() from http://pastey.net/91421	13:21
jengelh	gitte: it's the same reason you don't do rm -f $(ls *.jpg)	13:21
Pistos	CareBear\: Okay, then I dunno. :)	13:21
CareBear\	Pistos : And both are under my full control	13:22
	Pistos : Are push/pull completely analogous?	13:22
jengelh	push/pull are not analogous.	13:22
gitte	jengelh: no, it's not. You are making it harder for other people to sell Git as "easy". Because you make it "uneasy".	13:22
jengelh	you wrote the tool.	13:22
Pistos	Who wrote the what?	13:23
gitte	jengelh: git show? No.	13:23
jengelh	never mind, I always confuse gitte and gitster	13:23
CareBear\	Ok. So my error was to use push from the clone instead of pull in the master.	13:23
jengelh	gitte: anyway, see http://jengelh.medozas.de/articles/recurring-errors-sh.php	13:23
gitte	wereHamster: you need to read up on git first... you probably want to use read-tree anyway, and script what you want to do, rather than have it in a program.	13:24
	jengelh: no time.	13:24
jengelh	of course... who would have thought.	13:24
CareBear\	jengelh/Pistos : Can I fix that? :)	13:24
jengelh	CareBear\: send a bug report?	13:24
gitte	jengelh: hey, you suggested things to users which are not meant for them.	13:24
jengelh	git-log is not exactly fit for grepping.	13:25
Pistos	CareBear\: git stat should give instructions on how to un(rm\|do\|mv\|cache\|index) stuff.	13:25
CareBear\	jengelh : I don't think it's a bug though? The bug was me not understanding git push.	13:25
jengelh	CareBear\: i would still consider it as a bug. since push =transfer it, but pull=transfer+merge	13:25
	→ paltman joined	13:26
gitte	jengelh: if you use it incorrectly, right.	13:26
	jengelh: anyway, your link is nice, will have a look later.	13:26
	jengelh: and this discussion we are having is pointless: you should know that you are not exactly helpful when suggesting the use of rev-parse or rev-list.	13:26
CareBear\	Pistos : git reset HEAD doesn't work so well. :\ master doesn't include the commit from the clone in HEAD	13:26
gitte	jengelh: and I really have no time for this... I should learn to let things like this slip when I have to do proper work.	13:27
wereHamster	gitte: well, the C program uses lua scripts in its core. I already thought about spawning the git portion in a shell script or so, but I think it would be best if I could do it all inside C	13:27
Pistos	CareBear\: I've never found an instance where the git stat instructions failed to get me back to a clean state...	13:27
CareBear\	Pistos : clean, mh, but I want to merge :)	13:27
Pistos	CareBear\: Clean first, then pull. No? :)	13:28
CareBear\	git-merge doesn't play, but I will throw a few more commit ids at it.	13:28
	Pistos : ah yes, that's an idea.	13:28
	but no..	13:28
	after git reset HEAD, git diff shows the reverse of the clone commit :\	13:29
	git diff --cached is correct though.	13:29
wereHamster	gitte: also, I don't see how read-tree could help here. But then again, I don't know much about read-tree, I've never used it myself	13:29
	→ litage joined	13:30
gitte	wereHamster: read-tree can write to somewhere else. IIRC it is the --prefix option, but I am probably wrong.	13:30
CareBear\	git pull clone says Already up-to-date.	13:30
Pistos	CareBear\: Once you have everything unstaged, you can just git checkout ., I think.	13:30
wereHamster	gitte: you mean I could replace the git-archive \| tar -x with read-tree?	13:31
Pistos	That should get rid of all unstaged changes.	13:31
	(assuming that's what you want)	13:31
DAddYE	sorry with git how I can get the last commit message?????????	13:31
CareBear\	git log -1	13:32
	Pistos : ok, it looks almost right now.	13:33
Pistos	.. but? :)	13:33
CareBear\	Pistos : Fun thing is, I also had staged changes in master.	13:33
	→ shaftyy joined	13:33
Pistos	CareBear\: Ugh	13:33
CareBear\	Pistos : I still see them in git diff --cached	13:33
	but not in git diff	13:33
Pistos	CareBear\: Phew... I thought you lost them. :P	13:34
	Okay, so... isn't it better to merge only on a pristine WC? :P	13:34
	on/into	13:34
CareBear\	(I was careful not to reset the files that were changed in master)	13:34
	why doesn't the changes show up in git diff, but only in --cached ?	13:35
ndim	Because you have already staged them.	13:36
CareBear\	ndim : aha!	13:36
	and unstaging would be git-reset --soft	13:37
	hm needed a single revision	13:37
ndim	no idea :)	13:37
	→ queso joined	13:37
	→ vbabiy joined	13:37
CareBear\	Sorry I seem so confused, but git is so powerful that I'm not sure how to best wield it for my purposes. :)	13:38
ndim	I know the feeling, but what you have told us above was too complicated for me to grasp, so I'll leave that to other people to solve :)	13:39
CareBear\	git checkout . was the key piece of the puzzle.	13:41
	I pushed changes to the master, but not the master's working copy.	13:42
	→ Fullmoon joined	13:42
CareBear\	Pistos : Thanks for the help! :)	13:42
	← SuttoL_ left	13:43
	→ mountie joined	13:43
	← pflanze left	13:43
	gitsky just fixed totally broken styling in IE6 by inserting <span style="float: right"></span> at one precise spot of a page, yay!	13:43
gitte	gitsky: you fixed IE6? That's great. Does M$ give you any money for it? No? Then you just donated something to them...	13:45
CareBear\	IE6 is fun.	13:45
gitte	wereHamster: exactly	13:45
CareBear\	no wait.	13:45
gitsky	gitte: corporate environment forcing IE6 on us	13:46
	gitte: tough to do anything with that	13:46
	or about that	13:46
	gitte remembers gitsky having a suse.cz address...	13:46
gitsky	I'm on a break in another company in switzerland, working on a git project	13:47
	but I still need to figure out how much can I talk about it	13:47
	→ bryanray joined	13:48
CareBear\	the keyword situation	13:48
	gitte so loves NDAs	13:48
	gitte cuddles NDAs	13:48
	→ warlock_mza joined	13:49
	← schacon left	13:51
	→ juanjoc joined	13:51
	→ Aphelion joined	13:51
Aphelion	any reason for a tracked branch to push somewhere unexpected when using just 'git push'?	13:52
	← carrus85 left	13:53
Pistos	CareBear\: My pleasure. One noob to another. :)	13:56
	Anyway, I'm out [of this channel].	13:56
	← Sho_ left	13:56
	← Pistos left	13:57
	→ Sho_ joined	13:59
gitte	Aphelion: tracked branches are not pushed.	13:59
	Aphelion: but you probably mean a local forked branch with a _different_ name, right?	13:59
Aphelion	eh?	13:59
	nope, i mean a local forked branch and it has the same name	14:00
	i take it that's a problem?	14:00
gitte	Well, I guess that you have other local branches that are present at the remote site.	14:00
	But that is too much guessing; I should stop guessing without more information.	14:00
Aphelion	we're using an svn-style setup with multiple people having ssh access and git group access to a bare repository.	14:01
	Aphelion checked out a personal branch from origin to local, same name given to -b. git checkout --track -b name origin/name	14:02
Aphelion	git push, with no additional parameters, attempted to push to a different remote branch on the same remote (origin)	14:03
	it failed due to not being a fast-forward, which is good, but i have no idea why that would happen	14:03
Tv	Aphelion: have you read .git/config ?	14:04
Aphelion	double-checkin it	14:05
	← brosner left	14:06
Aphelion	Tv - the branch it tried to push to is higher in the config file, if that matters. the entry for the local branch has a remote config entry and a merge config entry.	14:07
	→ blowery joined	14:07
	← deavidsedice left	14:08
	→ heycarsten joined	14:09
	← zewerson left	14:10
	→ bryanl joined	14:11
dmq	is git-format-patch what i would use to take a change history for a file and push it into a different repository?	14:13
Tv	dmq: don't call it a push, but that is one way to achieve that goal	14:15
	dmq: most git users would 1) avoid needing to extract history of an individual file 2) use fetch+merge	14:15
	← lorandi left	14:15
	→ lorandi joined	14:15
dmq	well i wrote this script in one repository, and it really should live in another. these are totally unrelated repositories.	14:17
	so i was hoping there was an easy way to push just the changes to that file into a different repo.	14:18
	→ Cacheaway joined	14:18
Tv	dmq: git tracks whole trees; you could make a branch, then rewrite its history fully to remove all the other files from every commit; that would give you a history with just that single file	14:19
raz	dmq, i could imagine a very ugly script to automatically checkout each revision in order and commit that to the new repo.. but it's a kludge	14:19
	→ aroben joined	14:19
dmq	yeah i was hoping it would be possible to do it with git-format-patch but it doesnt seem to have a way to filter by files.	14:20
	→ pflanze joined	14:20
gebi	is git 1.5.4 useable for gitosis?	14:20
	← aroben left	14:20
gebi	just gitosis (nothing more)	14:20
raz	gebi, why should it not be?	14:20
	→ aroben joined	14:21
dmq	Tv: it seems like git log -p can do it.	14:22
gebi	raz: it's not that good for client side (slow (unrelated) pull, git-svn, ...)	14:23
	→ SuttoL joined	14:23
dmq	but can i pipe that into git-am.. i wonder...	14:23
raz	gebi, ok different question. why do you want to use that specific version then?	14:23
gebi	fedora 8 on server	14:23
raz	gebi, hmm ok i see. sorry i don't know an answer, just made me curious. i'd suggest to compile a stable version.	14:25
Tv	gebi: i have 1.5.4.3 on serverside myself	14:26
ndim	dmq: git filter-branch	14:30
	dwave → dwave\|w	14:30
dmq	ndim?	14:31
gebi	Tv: ah ok, thx :)	14:31
ndim	dmq: Create a new branch from HEAD, then do the filter-branch thing on that new branch, If you like the result, push that new branch to a different (possibly empty) repository.	14:32
	dmq: filter-branch can select just the commits affecting "this script".	14:32
	dmq: Do create a test branch to do that. Do check with "gitk --all" that it did what you expected.	14:33
	→ chrisb joined	14:34
dmq	ok.	14:35
	git-filter-branch has a frightening doc.	14:35
	and why do i see lots of .ft C things in it?	14:35
pygi	cnk, poke	14:36
	← bentob0x left	14:37
	← jonshea left	14:38
	→ G_SabinoMullane joined	14:40
dmq	uh ndim, could you suggest what filter-branch would look like?	14:40
	blowery → ben`afk	14:42
	→ cebewee_ joined	14:46
cebewee_	hm, is show-ref the right plumbing command to get a list of all branches?	14:46
	→ rdavila joined	14:48
ndim	dmq: the man page has the exact example you are looking for	14:48
	→ jackdempsey joined	14:49
ndim	dmq: Oh. Sorry. The man page only has it in reverse.	14:49
	→ fade-in joined	14:49
dmq	removing a filename you mean	14:49
ndim	Yupp.	14:49
	And you want "removing all filenames but the one", right?	14:50
dmq	yes	14:50
	i think i can use --subdirectory-filter	14:50
	since my script is the only thing in the subdirectory its in.	14:50
	but so far i hasnt worked as i would expect.	14:50
CareBear\	Can git-archive run twice internally?	14:51
	→ ckm joined	14:51
ndim	Sorry bout that. I don't have much experience with filter-branch, but it worked nicely for me in a different case the other day.	14:51
CareBear\	I want a zip with two copies of my tree, with different prefixes.	14:51
	→ brosner joined	14:52
thiago	CareBear\: concatenate the two zips together.	14:52
	or create a fake tree and run git archive on it	14:53
CareBear\	thiago : does that really work also for zip?	14:53
thiago	no, zips aren't cat-able. By concatenate, I meant with a theoretical tool that combines them together.	14:53
CareBear\	ya, no fun. :p	14:53
dmq	hmm i dont get it.	14:53
CareBear\	not as much fun as if git-archive could do it anyway. :)	14:54
dmq	i did git-filter-branch --subdirectory-filter=./my/path	14:54
	and i end up with a branch that only contains ./my/path but the files within are gone.	14:54
CareBear\	thiago : thanks for clearing it up!	14:54
	ok, something easy.	14:55
	hm maybe I already know the answer.	14:56
	testing	14:56
	← rdavila left	14:56
CareBear\	yep!	14:56
	→ thorny_sun joined	14:57
thiago	what?	14:58
	← jackdempsey left	14:58
Cacheaway	is there a problem running git commands from within hooks/post-update? I am trying to do cd ~/public_html ; git pull inside post-update, and I get "fatal: Not a git repository: '.'" when I push. I even added a pwd in the script to make sure I was getting to the right spot, i am. If i run the script manually it works, but it doens't work when invoked by git	14:58
CareBear\	want to add new things to an old commit	14:58
	=>git reset <commit before>	14:58
thorny_sun	glorious git gurus! how do tag all my added/rm-ed files to be commited on next commit without explicitly having to git rm / git add everything?	14:58
CareBear\	thiago : will the current HEAD commit cease to exist?	14:59
ciaranm	thorny_sun: add is easy. you just git add .	14:59
thorny_sun	but i have a lot of rm-ed stuff as well	14:59
	← ketralnis left	14:59
thorny_sun	ciaranm: is there an easy way for that?	14:59
ciaranm	don't think there's an easy way for rm, short of git status \| grep \| cut \| xargs git rm	15:00
litage	what's the difference between ``git init'' and ``git --bare init'' ?	15:00
	→ jonshea joined	15:00
thorny_sun	litage: --bare digs you down to just the .git folder-- it's used for storing a repo-- not actually using it	15:01
litage	thorny_sun: sorry, i'm not sure what you mean by "for storing a repo-- not actually using it"	15:01
thorny_sun	for example i use --bare to create all my remote repos and then locally push and pull to them	15:01
litage	ah	15:01
CareBear\	A repo without working copy	15:02
litage	makes sense	15:02
thorny_sun	exactly	15:02
	try it and you'll see	15:02
	← _zack left	15:03
	→ mib_h6p6q9yx joined	15:03
	← mib_h6p6q9yx left	15:04
litage	i created a repo on my local machine, and would like to provide others on my LAN with access to it. how would you guys recommend i accomplish that, without creating a local user account for each person to ssh into?	15:04
	← SuttoL left	15:04
	→ SuttoL joined	15:05
thorny_sun	you have to find a place to put it that everyone else can get to	15:05
	then you'll all have to push/pull to that repo	15:05
	ben`afk → blowery	15:05
ndim	litage: read/write or readonly?	15:05
litage	ndim: read only	15:05
thiago	thorny_sun: git add .; git commit -a	15:06
	rm'ed files get rm'ed	15:06
thorny_sun	thiago: really-- git status tells me otherwise	15:07
thiago	litage: git-daemon if it's read-only	15:07
thorny_sun	thorny_sun: what thiago said-- i thought everyone was gonna be writing	15:07
thiago	thorny_sun: git add .; then run git status -a to see the changes that would be committed with -a	15:07
	→ Rich_Morin joined	15:07
thorny_sun	oh cool!	15:07
	thiago: you're a genius-- thanks so much!	15:08
litage	should i use git-daemon or gitosis or something else to give others read-access to my repo?	15:08
Cacheaway	litage: you are worried about using ssh because you don't want people to have shells on your system	15:09
	?	15:09
thiago	thorny_sun: or git status . as well	15:09
litage	Cacheaway: correct	15:09
thiago	thorny_sun: . and -a perform the same thing if you're at the top level	15:09
	← pombreda1 left	15:09
Cacheaway	because there is a git-only shell available, set that as the default shell for the users, and there you go	15:09
DAddYE	Sorry, one question how I can get ONLY the last commit message?	15:10
CareBear\	how to export? is git-archive\|tar x the recommended method?	15:10
thorny_sun	thiago: how do i undo a merge i just did?	15:10
	→ jackdempsey joined	15:11
litage	Cacheaway: /bin/git-shell ?	15:11
DAddYE	Sorry, one question how I can get ONLY the last commit message?	15:11
tokkee	DAddYE: git show HEAD	15:11
Cacheaway	litage, i don't know anything about it other than what you can find out with man git-shell seems worth looking into	15:11
	→ chris2_ joined	15:12
DAddYE	tokkee: one last question, If I need this in a hook where I need to use it ? Update?	15:12
litage	Cacheaway: thanks for the suggestion. i'll have a gander	15:12
tokkee	DAddYE: I don't get that question, sorry...	15:12
	→ irotas joined	15:12
thiago	thorny_sun: the same way you undo a commit	15:12
	thorny_sun: git reset --hard HEAD^. Be careful! That erases the last commit from history.	15:13
	thorny_sun: always remember: do not go back in history if you've already pushed it	15:13
thorny_sun	k	15:13
DAddYE	tokkee: I want to send by email commit message through hooks... but update don't see the last commit message post-receive send me two mail!	15:13
thorny_sun	thanks	15:14
Rich_Morin	I need to interact with a slightly unorthodox git setup and need some clues - http://www.pastie.org/234825	15:14
	← chris2 left	15:14
Rustre	thanks a _lot_ for those who helped me this morning using the git rebase feature	15:14
	chris2_ → chris2	15:14
Rustre	it really rocks, and i feel i'm understanding it a bit better now	15:14
Cacheaway	does anyone have any insight into my question? (look back approx 17 minutes)	15:15
broonie	DAddYE: The default post-recieve-email hook might be helpful for you (look for teh bit that lists new revisions).	15:16
DAddYE	broonie: where I found it?	15:16
Rustre	Cacheaway: your ~/public_html is a git repo?	15:16
Cacheaway	yes	15:17
broonie	DAddYE: .git/hooks	15:17
	→ patco444 joined	15:17
DAddYE	there isn't!	15:17
thorny_sun	is there a way to do an interactive merge -- so i can say that i want this file from this branch instead of trying to merge it, but in another case i do want it to attempt a merge of the same file from the two branches?	15:17
patco444	The Best Online RPG / Strategy game ever : http://www.fallensword.com/?ref=1461055	15:18
	← patco444 left	15:18
thorny_sun	thiago: ?	15:18
Rustre	Cacheaway: no clue then, i'm a real git noob :s	15:18
	← jackdempsey left	15:19
DAddYE	broonie: ??	15:19
thiago	thorny_sun: git-merge-tool	15:19
broonie	DAddYE: That's where you can find it.	15:20
	← Aphelion left	15:21
DAddYE	broonie: there isn't for me	15:21
	→ schacon joined	15:27
Cacheaway	Rustre: thanks. Anyone else know why post-update would work when run manually, but fail when invoked by git?	15:27
	foca_ → foca	15:29
ciaran	can I set git-diff to not display ^M when showing a file with mixed line-endings?	15:31
	→ rdavila joined	15:31
DAddYE	post-receive are called two times if I do a push?	15:32
	→ davidfetter joined	15:32
	← markkalderon left	15:32
thiago	ciaran: you shouldn't have mixed line-endings	15:32
ciaran	I know	15:32
thiago	ciaran: CRLF is fine, LF is fine. Both isn't.	15:33
	so you should fix the file.	15:33
	← CareBear\ left	15:33
_slacker_	hi guys, how can i specify a certain revision of subversion to clone from when using git-svn? i read about this somewhere but i can't remember...	15:34
	→ digitalhobbit joined	15:34
	→ hobbes006 joined	15:34
	→ [raz] joined	15:35
	→ sverrej joined	15:36
bremner	_slacker_: you read the man page?	15:36
	← Rich_Morin left	15:36
_slacker_	bremner: yes, but don't know why the -r/--revision flag doesn't works for me	15:36
hobbes006	hi there, i have been reading about .gitignore and about .git/info/exclude but i am still not very sure how to use it. i am attempting to keep a "settings" file separate from branch to branch in my local machine. If I have branch A and branch B, the settings file in branch A will not be merged into the settings file in branch B. typing this filename into .git/info/exclude does not seem to prevent the settings file from being excluded fro	15:37
DAddYE	is possible to use git log -1 for a given revision?	15:37
_slacker_	I if do a clone without any option, it start to translate from the first svn revision, but i won't to do that because I have a lot of revisions, so i'm trying to find why the -r isn't working :s	15:37
bremner	_slacker_: ok, cool. All I can tell you is about the -r flag. git-svn fetch -r 99823 seems to be the documented option	15:38
thiago	_slacker_: the -r option has to point to a revision that does exist at the path you're trying to clone	15:38
	i.e., an actual commit that touches that path	15:38
Rustre	hobbes006: i might be able to help you out on this one, as i was struggling with it this morning :P	15:38
hobbes006	Rustre: Thanks! :)	15:38
Rustre	hobbes006: here's how i did it : i have a central repository, for convenience	15:39
	on a backuped server	15:39
	← cilly left	15:39
	hobbes006 is listening intently.	15:39
Rustre	and i push to this server from my "local" repository on my developent PC	15:39
_slacker_	thiago: yes yes, it exists	15:40
	→ lea_w joined	15:40
Rustre	i'm doing web development, and on my development PC, i have a settings file which will "connect" to my local mysql DB	15:40
thiago	_slacker_: can you svn info -r NNN URL?	15:40
hobbes006	right.	15:40
Rustre	i need that to be able to debug and play around and develop	15:40
	now, i need to "roll out" my website from time to time on the production web server	15:40
hobbes006	Rustre: do you happen to be using django ? :p	15:40
Rustre	nope, not yet, but i should/will	15:41
hobbes006	o ok. dont let me disrupt you. pls continue.	15:41
Rustre	(i'm new at this company, and everything is coded in php for the moment)	15:41
	← DAddYE left	15:41
Rustre	(and i'm new to django too, but i've been playing around with it for some time now)	15:41
	anyway	15:41
	on the production server, here's how i rolled out :	15:41
	→ koke_ joined	15:42
Rustre	i first did a "git clone" from my central repo, in a "local" repo (on the web server)	15:42
	ie wwwtemp	15:42
hobbes006	right	15:42
Rustre	i then copied everything from wwwtemp to the real www folder	15:42
	(included the .git folder and everything)	15:42
hobbes006	riht	15:42
Rustre	now i have a repository in www on the production server	15:43
	with the "wrong" settings file	15:43
hobbes006	right	15:43
Rustre	but all i need to do then is to modify locally the settings file	15:43
	then commit it	15:43
hobbes006	ok.	15:43
Rustre	to the local repo (i won't push it to the central repo)	15:43
hobbes006	but when you push?	15:43
	right	15:43
Rustre	now, each time i need to update my web production server :	15:43
	← thorny_sun left	15:44
hobbes006	?	15:44
Rustre	i first push from my local dev PC to the central repo	15:45
	and i pull from the prod server	15:45
	it'll merge the changes from the central repo	15:45
	into the local "prod" repo	15:45
	see what i mean?	15:45
hobbes006	hmm...	15:46
	cant we just use .gitignore and .git/info/exclude somehow.	15:46
	→ k0001 joined	15:46
Rustre	just a sec	15:47
	← raz left	15:48
	[raz] → raz	15:48
	→ Strogg joined	15:48
Rustre	sorry, talking with the boss	15:48
gitsky	hobbes006: .gitignore and .git/info/exclude concern only _untracked_ files	15:48
Rustre	gitsky: hobbes006 : exactly, that's what i figured the hard way this morning	15:49
gitsky	hobbes006: why not just commit your config file locally in each repository?	15:49
Rustre	you can't use it for this purpose, unless you just don't want _at all_ to track the settings file	15:49
hobbes006	but.... i dont want settings.py to be tracked.	15:49
gitsky	hobbes006: then don't add it to git?	15:49
Rustre	but then, you'd need to create it (copy it?) manually on each server/computer...	15:49
	← chrisb left	15:50
hobbes006	yes.	15:50
	but how do i exclude it ?	15:50
	→ zawir joined	15:50
hobbes006	when i type "git add ."	15:50
Rustre	hobbes006: if it is already tracked, then you can "untrack" it by using "git rm --cached settings.py"	15:50
	well	15:50
hobbes006	it automatically adds everything downwards from the $GIT_DIR.	15:50
Tv	hobbes006: echo /settings.py >>.gitignore	15:50
Rustre	first add it to the .gitignore file	15:50
hobbes006	ok	15:50
	← charon left	15:50
hobbes006	already did	15:50
	but it's still tracking :(	15:50
Rustre	but this will be useless if it's already tracked (already in the repo/index)	15:51
Tv	hobbes006: then remove it	15:51
Rustre	see what i mean?	15:51
	← samjam left	15:51
hobbes006	right	15:51
Rustre	the .gitignore and the others only prevent adding files to the repo	15:51
hobbes006	git rm --cached settings.py	15:51
Rustre	once they're added, they will always be tracked (until you untrack them with git rm --cached...)	15:51
hobbes006	AHHHHHHHHHH	15:51
Rustre	the --cached means "only remove it in the index/repo/whatever, not on the filesystem	15:52
	← redondos left	15:52
Rustre	if you don't put --cached, it'll remove it _also_ on the FS	15:52
hobbes006	i understand now.	15:52
Rustre	(from the working directoy)	15:52
hobbes006	so the other question is.	15:52
Rustre	well, all my speech was useless then, sorry about that :P	15:52
hobbes006	is it possible to specify that settings.py IS tracked in branch A	15:52
	but NOT tracked in branch B ?	15:52
	all on the same machine.	15:52
Tv	of course it is	15:53
Rustre	hobbes006: git rm it only on branch B i guess	15:53
Tv	branches aren't really related to each other in any way (apart from possibly sharing history)	15:53
hobbes006	i see	15:53
	so, does that mean that .gitignore file will be different from Branch to Branch ?	15:54
	when i am in branch A, my .gitignore file excludes settings.py	15:54
	← Strogg left	15:54
hobbes006	when i am in branch B, my .gitignore file is simply blank. nothing stated there.	15:54
	→ bdiego joined	15:54
hobbes006	do i "git rm --cached .gitignore" in order to achieve this ?	15:55
	→ Strogg joined	15:55
	→ ben_h_ joined	15:55
	→ DAddYE joined	15:55
	Arc_ → Arc	15:56
Rustre	errrr	15:56
	← koke_ left	15:56
Rustre	i wouldn't change the .gitignore file	15:57
	i would leave it as is (with the settings.py in it)	15:57
hobbes006	ok	15:57
DAddYE	Any can help me? Why this hook send me two mail? http://git.pastebin.com/d38583afb	15:57
Rustre	as it won't be taken into account if the settings.py file is already tracked, as said before	15:57
	OR, you could of course modify the .gitignore file differently on each branch	15:57
	← koke left	15:57
Rustre	i guess there's no problem at all with that, the .gitignore is just tracked as any other file	15:58
Cacheaway	what is the use of the GIT_DIR env variable? is it for using dirs other than .git ?	15:58
gitsky	I think it's cleaner to have settings.py in .gitignore only in branch A	15:58
ciaranm	Cacheaway: it's to make the baby jesus cry	15:58
gitsky	otherwise a year in the future, you'll get very confused about this for sure :)	15:58
	Cacheaway: yes	15:59
DAddYE	Any can help me? Why this hook send me two mail? http://git.pastebin.com/d38583afb	15:59
gitsky	though the Modern Shiny Way is to git --git-dir=...	15:59
Cacheaway	thanks	15:59
	i was running git commands in my post-update	16:00
	← lorandi left	16:01
gitsky	DAddYE: maybe once per each updated ref?	16:01
DAddYE	gitsky: I don't know... the only thing that I can say is that I receive two mail, one is correct, but one contains only the previous commit message	16:02
Cacheaway	basically, cd to a place where the same repo was cloned, and do a pull to get the new changes. it was failing because the bare repo i was pushing to set GIT_DIR=. (because it is a bare repo i guess) in the envoronment that post-update runs in	16:02
DAddYE	without my template	16:02
gitsky	UukGoblin: our great gitster guru spoke:	16:02
	True, and false ;-) Just "unstage" first in your "git add -i" session and	16:02
	then add only the necessary pieces back to the index (alternatively you	16:02
	can "reset" before entering "git add -i" session).	16:02
	should've said "thus spoke" :(	16:02
	DAddYE: well, debug it? this is pretty basic	16:03
	DAddYE: figure out with what parameters is it called each time, etc.	16:03
	← ben_h left	16:03
DAddYE	gitsky: mmmm	16:04
	resolved	16:04
	!	16:04
	Ive another ook... update!	16:04
gitsky	i think it will just behave like that randomly ;p	16:04
	based on how many branches you push, if you push any tags along or whatever	16:04
	but maybe not	16:04
	→ jamie_alm joined	16:05
UukGoblin	gitsky, cool :-)	16:05
	gitsky, did cherry-pick's manpage patch got approved? :-)	16:05
	→ markkalderon joined	16:05
	← jamie_alm left	16:06
	→ krawek joined	16:06
	→ jamie_alm_ joined	16:06
	→ jonshea_ joined	16:06
	blowery → ben`afk	16:07
jamie_alm_	bah, stupid internet	16:07
	→ lorandi joined	16:08
	→ koke joined	16:09
	→ yofel joined	16:10
	← sverrej left	16:11
	← cbreak left	16:13
	→ fhobia joined	16:13
gitsky	UukGoblin: yes	16:16
UukGoblin	nice	16:17
	aroben → aroben\|lunch	16:18
	← \ask_ left	16:20
	ben_h_ → ben_h	16:20
	← jonshea left	16:23
	→ xinming_ joined	16:23
xinming_	When I use git-push, I got '	16:24
	When I use git-push, I got ' ! [rejected] master -> master (non-fast forward)'	16:24
ciaran	one revision back I have a commit which includes a change I wanted to be in the commit before it. What’s the best way to fix that? Can rebase -i handle this?	16:24
xinming_	How will this happen?	16:24
Tv	xinming_: you need to merge the remote changes to your branch first (or, if you use rebase, stop rebasing published commits)	16:25
ciaranm	i'm making a patch for git rm --if-missing. poll: what exit status would you expect if you git rm --if-missing 'foo' and foo is not missing?	16:26
	→ ben_h_ joined	16:26
Tv	ciaranm: how's that different from --ignore-unmatch ?	16:27
xinming_	Tv: thanks, worked, another question, when I use git-pull, It won't pull from the repository. instead, It displays many messages.	16:27
ciaranm	Tv: it removes files only if they're ENOENT	16:27
xinming_	Tv: But git-remote show origin tells me the right path	16:27
	← cebewee_ left	16:27
Tv	ciaranm: how's that different from update-index --remove?	16:27
xinming_	Tv: Is there a way to make git pull from default repository?	16:27
ciaranm	Tv: it's part of 'git rm', which is where people expect it to be	16:28
Tv	xinming_: don't ask me, ask everyone	16:28
xinming_	Tv: sorry. :-)	16:28
Tv	ciaranm: to answer your question, shouldn't "git rm --if-missing ." be a valid thing to run, then?	16:29
xinming_	when I use git-pull, It won't pull from the repository. instead, It displays many messages.	16:29
Tv	ciaranm: so even if it is there, don't fail	16:29
	→ rtomayko joined	16:29
	→ ben_h__ joined	16:29
xinming_	So is there a way to make git pull from default repository? Or, I have to pull the remote by specify the "remote-name" ?	16:29
ciaranm	Tv: with -r, yes. the question is whether it should exit with success if it doesn't rm anything.	16:29
Tv	xinming_: maybe those messages actually say what is happening	16:29
	→ jerbear joined	16:29
	→ icwiener joined	16:30
	← reel__ left	16:30
ciaranm	i'd be inclined to say that it should exit with success so long as either stuff matches or you --ignore-unmatch, even if everything you list still exists	16:30
gitsky	hmm, so silly git newbie question	16:31
	how do i examine existing unmerged conflicts in my tree?	16:31
	is there anything better than git ls-files --stage for that?	16:31
	← earcar left	16:31
Tv	gitsky: git status?	16:31
gitsky	# unmerged: sub1	16:31
	that's not very informative	16:31
	I know that it's unmerged	16:31
	I want to know details	16:32
Tv	gitsky: like what?	16:32
	that's enough for me to open it in an editor..	16:32
gitsky	well, most importantly the competing sha1s	16:32
	the thing is, sub1 is nothing you could open in an editor	16:32
	→ timocratic joined	16:32
Tv	gitsky: then i think you do need ls-files	16:32
gitsky	hmf	16:32
	thanks	16:32
xinming_	Tv: hmm, I'll check that	16:33
Tv	gitsky: i didn't say text editor ;)	16:33
gitsky	but _what_ should I actually open in the editor?	16:33
Tv	gitsky: maybe you want git mergetool?	16:33
gitsky	if we're not talking about text :P	16:33
	git mergetool is only about text, no?	16:33
Tv	the things you hook into mergetool get all the possible alternatives	16:33
	nope	16:33
ciaran	ok, I tried doing a reset to the commit before it to edit it, but that seemed to go badly. Is that the right way?	16:33
	← halorgium left	16:33
gitsky	I see	16:34
	so this might work on binary files	16:34
	the only trouble now is that sub1 is not a file at all ;)	16:34
Tv	gitsky: well that's why tools run by mergetool get all kinds of input	16:35
	→ zed_DX joined	16:35
gitsky	docs say they just get the three files	16:36
Tv	gitsky: merge base, left, right, i believe -- how is that not sufficient?	16:36
	← icwiener left	16:39
	→ sergiodj joined	16:39
gitsky	Tv: as I said, sub1 is not a file	16:40
	which is something that makes mergetool profoundly unhappy ;)	16:40
Tv	gitsky: what, then?	16:40
gitsky	a submodule	16:40
Tv	ahhh	16:40
	→ icwiener joined	16:40
Tv	yeah i don't expect anything to merge submodule HEAD changes nicely	16:41
	mergetool could provide you more info; e.g. git diff gives you sha and mode etc	16:41
	would just need to be backwards compat	16:41
	← ben_h left	16:41
	→ markkald_ joined	16:41
	← markkalderon left	16:42
	→ earcar joined	16:42
	→ CodeWar joined	16:43
	← SuttoL left	16:44
	→ cbreak joined	16:45
	→ sverrej joined	16:46
	ben`afk → blowery	16:46
	← ben_h_ left	16:46
	blowery → blowery`afk	16:47
	blowery`afk → blowery	16:47
	cbreak → cbreak\|afk	16:47
	→ SuttoL joined	16:47
	→ asdfdsfsd joined	16:48
	← fhobia left	16:48
	← koke left	16:50
	→ koke joined	16:51
	← hobbes006 left	16:56
	← aroben\|lunch left	16:57
	← jonshea_ left	16:59
	← SRabbelier left	17:02
	← litage left	17:03
	← kanru left	17:04
	← CodeWar left	17:05
argonel	does git know about xattrs?	17:05
	← warlock_mza left	17:07
gitsky	no	17:07
argonel	pity	17:10
	→ aroben joined	17:10
gitsky	what about windows users? ;)	17:10
	and aren't they overally filesystem specific or something?	17:10
ToxicFrog	I believe the only attribute git tracks is executability.	17:11
	And yes, xattrs are filesystem specific. So are the rwx bits, for that matter.	17:11
	← nud left	17:12
ToxicFrog	NTFS has something similar but not the same (but you can fake it); FAT only has w.	17:12
argonel	gitsky: i have a deep lack of care for windows users :)	17:12
gitsky	:))	17:12
ToxicFrog	argonel: some of us have to release cross-platform software, you know.	17:12
	→ lbt joined	17:13
ToxicFrog	(or, well, ok, I don't have to, but it would be dickish not to)	17:13
	→ aroben_ joined	17:13
argonel	ToxicFrog: true enough, but you don't have to do a cross-platform release of my software, and even if you did "have to", i'd honestly prefer it if you didn't :)	17:13
lbt	hi - having a problem ... : git-fetch -v -t isn't reporting v2.6.26 pulling from a local mirror	17:13
ToxicFrog	True, but I do use git for version control on both platforms I support	17:14
	→ hawking joined	17:14
argonel	in fact, i rather like the idea of roadblocks making it hard to make my software work well on windows, aside from my unwillingness to provide a windows user with any support	17:14
	← kergoth`work left	17:15
lbt	I used git pull to get a copy of the kernel and then git-fetch -vt to update it periodically	17:15
	← aroben_ left	17:15
lbt	this latest fetch didn't work	17:15
	any ideas?	17:15
	→ aroben_ joined	17:15
ToxicFrog	argonel: I don't. Deliberately introducing incompatibility makes one just as bad as MS.	17:15
	gitster → paster	17:16
ToxicFrog	(well, ok, maybe not just as bad. But it's still not right, and very reminiscent of their tactics.)	17:16
argonel	ToxicFrog: sometimes the only way to teach someone not to punch you in the face is to punch them in the face	17:16
	gitsky doesn't understand what the problem is	17:17
	→ shingo joined	17:17
lbt	gitsky: my problem ? :)	17:18
argonel	ToxicFrog: i'm in the camp that believes that anything at all that makes it possible for you to stay in windows is inherently evil. but this is way off-topic for this channel :)	17:18
paster	pasky, "ls-files -u" is your friend ;-)	17:18
	→ bronson joined	17:18
	→ ijcd joined	17:19
	→ cilly joined	17:19
ToxicFrog	argonel: I write programs because I enjoy it. I release programs to make people's lives better. If they're currently using windows, then unless your program is so incredibly fantastic that it causes people's heads to spontaneously explode with joy, it probably isn't going to be the killer app that causes them to switch - assuming they can switch; not all windows software has linux equivalents, nor does it all	17:19
	run in wine.	17:19
gitsky	lbt: yes, but gotta run now	17:19
	paster: huh?	17:19
lbt	damn - OK	17:19
	http://pastebin.com/d484a124d	17:19
paster	"unmerged: sub1" ;-)	17:19
	← hawking left	17:19
	paster → gitster	17:20
gitsky	gitster: yes but ls-file sis plumbing	17:20
	I mentioned it in my question	17:20
	lbt: your reasons to believe the tag is on the server?	17:20
	really afk now :(	17:20
ToxicFrog	Basically, I view witholding releases just because you don't like an OS - not simply not making them because you don't have time or interest, but deliberately blocking attempts to create such ports - is a mean-spirited attempt to punish people for choosing the wrong OS, regardless of whether they actually have a choice, or if they do, whether they realized they had one.	17:20
gitsky	gitster: lol	17:20
lbt	err v2.6.26 was released :)	17:20
	→ hawking joined	17:21
gitsky	gitster: you got me there :))	17:21
	lbt: and mirrored out?	17:21
ToxicFrog	Furthermore, more ports -> wider user base -> more testing and bug reports -> better program.	17:21
lbt	I'm just putting together a pastebin to explain my setup	17:21
gitsky	lbt: first make sure it's on the server	17:21
	use git ls-remote	17:21
	← hawking left	17:21
argonel	ToxicFrog: i understand where you're coming from, but i have no time nor interest in trying to disambiguate whether the bug a windows user is reporting is because the software isn't working right, or because the OS isn't working right.	17:22
ToxicFrog	Fair enough. My contention here is that while "I don't want to support this OS" is reasonable, "I will spend extra time and effort making life difficult for people who do" is not.	17:23
	← tcoppi left	17:24
lbt	Here's my setup http://pastebin.com/d1cc1c7d0	17:24
	→ tcoppi joined	17:24
lbt	I have a local machine with a copy of Linus' git repo	17:24
	→ hawking joined	17:24
lbt	I then clone from that machine to other local machines	17:24
	normally (last 18 months+) this has been fine	17:25
argonel	ToxicFrog: the beginning of this discussion is whether or not git supports xattrs, which may or may not be portable to windows. not implementing a feature because it can't be done (or not easily) is, imo, insanity	17:25
ToxicFrog	Wait. Not implementing a feature because it can't be done is insanity?	17:25
uau	lbt: your paste has "From git://elm/linux-git"	17:25
lbt	git fetch -v -t on my local server gives: http://pastebin.com/d484a124d	17:26
	yep that's my local server, elm	17:26
uau	are you sure the tag is on THAT server?	17:26
	← ben_h__ left	17:26
uau	that 2.6.26 was released is obviously no reason to think that YOUR server would have the tag	17:26
argonel	ToxicFrog: i guess i left off end of the sentence, "not implementing a feature because it can't be done on windows is insanity"	17:26
	← loupgaroublond left	17:26
lbt	git tag \| grep 26 shows v2.6.26	17:26
	but	17:26
	git fetch -vt doesn't	17:27
	← gitte left	17:27
ToxicFrog	Aah.	17:27
lbt	http://pastebin.com/d484a124d	17:27
	← pdherbemont left	17:27
ToxicFrog	Can't be done on some linux filesystems, either.	17:27
	s/linux filesystems/filesystems supported by linux which one may very well have a git repo on/	17:27
argonel	well, and thats of course where the analogly falls apart :)	17:27
	s/ly/y	17:27
lbt	and I don't commit to my server (well, not this one)	17:28
	← markkald_ left	17:28
lbt	it's supposed to be a clean pull of Linus' and any stable releases (no merges should be needed)	17:28
	← aroben left	17:28
	→ kergoth`work joined	17:28
	← hawking left	17:29
lbt	I am on git --version 1.5.6.2	17:29
	and 1.5.6.1	17:29
	→ hawking joined	17:29
lbt	s/.1//	17:30
	→ kanru joined	17:30
	← DAddYE left	17:30
	← lbt left	17:31
	→ bobesponja joined	17:32
	→ p4tux joined	17:32
	→ _graham_ joined	17:32
	← Aikawa left	17:34
	→ joydivider joined	17:34
	← deavid left	17:36
	→ tommorris_ joined	17:37
	← hawking left	17:39
	→ sverrej_ joined	17:43
	← emachado left	17:43
	→ sergio__ joined	17:43
	← sergiodj left	17:43
	→ subtleGradient joined	17:45
	→ emachado joined	17:45
cnk	I needed to remove a password-containing file (database.yml) from my private github repository so I could open it up. I did a filter-branch command on a working copy and did a forced push to github and all looked well. Then I did something, not entirely clear what, from one of my other working copies and now 1) the database.yml file is back on github and 2) one of my working copies appears to have 2 copies of all commits. In looking into this, I am finding the me	17:45
subtleGradient	when sending people patches, how do you `git format-patch` multiple commits into a single patch?	17:45
	← Rustre left	17:46
cnk	Does someone have a best practices suggestion to using github to move the same code base to various developers?	17:46
jast	subtleGradient, one patch always equals one commit, so what you can do is use (for example) rebase -i to squash several commits into one, thereby altering history (so don't do it if you have already published the commits somewhere else)	17:47
subtleGradient	jast: ahhh. that explains a lot! Thanks	17:47
jast	cnk, your first line got cut off at "I am finding the me"	17:47
cnk	sorry - I am finding the merge history kind of confusing. I think this might be because my git workflow is suboptimal	17:48
	→ nud joined	17:48
jast	cnk, anyway, doing forced pushs can sometimes cause problems with other working copies that still have the old state of things. i guess what you may have done is merge the new state of the central repo into a local repo that had the old commits... that could result in having every commit twice	17:50
	→ _zack joined	17:50
cnk	that sounds right	17:50
subtleGradient	is there a good site for examples of recommended workflows?	17:51
cnk	so to fix it ??? I was thinking of creating a patch file of anything that is in that working ocpy that isn't in github, and then recloning github and applying the patch	17:51
	I would love to find just such a site	17:51
subtleGradient	cnk: the topic mentions http://www.gitcasts.com/	17:52
	I'm checking that out now	17:52
cnk	so far I think those are all just the peepcode screen cast recut into separate episodes	17:53
	I have watched those but still missing something	17:53
jast	cnk, yeah, that's probably the easiest way to go. you can also create a series of commits from your local changes, use format-patch and later apply the patches using git am. this way you don't have to squash everything together.	17:53
	← sverrej left	17:54
cnk	jast: thanks that might be nice. Though I am more concerned with just getting stuff over, less concerned about history and granularity	17:54
MarcWeber	! [rejected] uml -> desk/uml (non fast forward)	17:55
	Is there an easy way to compare the results of both "heads" ?	17:55
	If git would print both hashes you could diff them, couldn't you?	17:55
jast	you don't need the hashes, try git diff uml desk/uml	17:56
	← tommorris left	17:56
MarcWeber	jast: :) yay thanks	17:57
jast	what happened there is that in either the local or the remote repo, someone used something that altered history. you can pretty much forget about properly merging both together now, so let's hope one of the repos didn't add anything new. :}	17:57
cnk	But on subtleGradient and my quetions about workflow.... If you want to distribute the main line of development of a project via github, how do people approach the push? Do you have an explicit tracking branch for the github remote? pull to that branch? and then merge your local branch onto it? then push to github? Or should I be thinking 'rebase' somehow?	17:57
	→ Fullmoon_ joined	17:57
jast	i'll be back in a couple of minutes, then i'll try and answer that	17:58
cnk	jast: was your last reply to me? or to MarcWeber?	17:58
subtleGradient	cnk: it all depends on exactly what you're doing. But this is my current recommendation for MooTools: http://github.com/mootools/mootools-core/wikis/how-the-mootools-dev-team-contributes	17:59
MarcWeber	jast In any case it has only been me. As long as I haven't published the patch I amend-record some times..	17:59
	→ freaky[t] joined	17:59
subtleGradient	we all work on our own forks and push to them. a few of us also have access to push to the official fork	17:59
	→ johnevans joined	18:00
pygi	subtleGradient, I think first line on that page says "Dont do" what cnk did :p	18:00
subtleGradient	LD	18:00
	← koke left	18:00
subtleGradient	:D	18:00
	→ koke joined	18:00
uau	hmm running "git gc" on a repository that was cloned from a local one created 24000 loose objects	18:00
subtleGradient	but what is the recommended way to get rid of a password file or whatever from all history?	18:01
	→ Eridius\|work joined	18:01
meyering	anyone know why "git repack -adf" would leave 60k loose objects? I'm using git version 1.5.6.3.385.g9f9af and trying to clean up a 2.1GB repo produced by using cvsparse to convert the gdb cvs repo to git. Also tried git-1.5.5.1.	18:01
uau	there was a remote branch without a local one; the local mode of clone copied its objects even though they were unreferenced, then gc unpacked the unreferenced objects but didn't prune them because they were new	18:01
	← koke left	18:02
	→ jamessharp1 joined	18:02
uau	meyering: maybe they're unreferenced? those are never packed (compare to what i just wrote above)	18:03
	try running "git prune"	18:03
	← mcella left	18:04
meyering	uau: that did it. thanks. did something change in the last few months? istr I didn't need to do that before.	18:04
	← johnevans left	18:04
meyering	down to just 5MB, which sounds like it's too small to be valid	18:04
	→ deavid joined	18:05
pygi	subtleGradient, :	18:05
	git-filter-branch --tree-filter 'rm -f testme.txt' HEAD	18:05
	git reset --hard	18:05
	perhaps? :)	18:05
subtleGradient	pygi: you lost me :[	18:05
pygi	cnk, look above ^_^	18:05
subtleGradient	ahh, got it	18:06
cnk	so the first is what I did. But what is the reset --hard?	18:06
	← _slacker_ left	18:06
MarcWeber	cnk: reset branch to given hash --hard means also reset working directory	18:07
	→ DrFrasierCrane joined	18:07
	→ CodeOfficer joined	18:07
cnk	MarcWeber: but why is it needed in this circumstance? Isn't my working copy and index in sync after the filter-branch?	18:08
	git status at that point shows no changes	18:08
	→ ketralnis joined	18:08
	← ketralnis left	18:08
	→ ketralnis joined	18:09
	← Fullmoon_ left	18:09
	→ Fullmoon_ joined	18:09
cnk	subtleGradient: that MooTools doc looks good. I don't want multiple versions on github but isn't 'forking' in github is just creating a branch?	18:11
	← SuttoL left	18:11
gitsky	branch is an abstract and ambiguous term	18:11
	conceptually, fork might be a branch	18:12
MarcWeber	cnk: I guess it's to make sure it's also gone from your working directory.. But you shoul ask pygi .. I've never used git-filter-branch	18:12
gitsky	but not in the sense normally used in git	18:12
subtleGradient	cnk: github forking makes a copy of the entire project.	18:12
	in practice, a fork is like another repo with its own branches and everything.	18:12
	→ hawking joined	18:12
cnk	so is more like 'git clone'?	18:12
MarcWeber	Which is the way to compare cherry picked patches? (eg if they have the same comment they don't have to have the same contents..) ?	18:13
subtleGradient	A major advantage of forking is that you get a nice online backup of all your personal work-in-progress.	18:13
pygi	cnk, yes, you can consider github's forking as "git clone"	18:13
	→ pdherbemont joined	18:13
pygi	(sorry for being amazingly slow, am writing)	18:14
subtleGradient	we forgive you	18:14
	_this_ time	18:14
	← Fullmoon left	18:14
	→ bughunter2 joined	18:14
bughunter2	hey, can someone check this problem: http://marc.info/?l=git&m=121586372728355&w=2	18:15
	i've emailed about that a while ago, still couldn't figure it out	18:15
cnk	so if I want to adapt this workflow (http://github.com/mootools/mootools-core/wikis/how-the-mootools-dev-team-contributes) I should have a branch on my local repo that is exactly what is on github, and then a working branch that I use as the basis for feature branches	18:15
	then when I want stuff out on github, I pull to my github branch, then merge my working branch onto the github branch and then push to github	18:16
	would it be appropriate when I want my working branch merged into the github branch to use rebase?	18:17
	← joevandyk left	18:17
subtleGradient	cnk: never rebase a branch that has ever been pushed	18:18
	if you do then you lose the ability to push a "fast-forward"	18:18
bughunter2	subtleGradient: what's a "fast-forward" ?	18:18
	← jamessharpe left	18:18
subtleGradient	I may be wrong, but that's how I understand it	18:18
Eridius\|work	a fast-forward is when you try and do a merge but your HEAD is a direct ancestor of the remote branch, so instead of creating a merge commit it just "fast-forwards" your branch to point at the remote branch	18:19
subtleGradient	when you push to a remote. if there are any remote changes then it won't push, you have to pull first and then push again.	18:19
	→ Pistos joined	18:19
	← Bryson left	18:19
	← hawking left	18:19
cnk	I think I understand this - so in the workflow I described, I never push my working branch to github, only the github tracking branch I am using to sync up iwth my 'official' repo on github	18:19
	→ Bryson joined	18:20
subtleGradient	but if you rebase then you rewrite history that is already out there and you won't be able to push without forcing it and screwing everyone else up	18:20
	cnk: that's how I do it, yet	18:20
Eridius\|work	subtleGradient: actually, cnk's rebasing would work fine	18:20
Pistos	I'm trying to git clone ssh://.... but the ssh appears to be trying to use bash, which doesn't have a correct PATH to find the git executables. Besides fixing the bash PATH, can I get git to use zsh instead [which has the right path]?	18:20
subtleGradient	might not be the best wrkflow, but it seems to work for me	18:20
Eridius\|work	since he'd be rebasing his local working branch on top of the github-tracking branch	18:20
	Pistos: I would expect your shell to only set up the right PATH in an interactive shell anyhow	18:21
	→ hawking joined	18:21
Eridius\|work	you can use the --upload-pack command to pass the non-standard path	18:21
subtleGradient	Eridius\|work: I guess that makes good sense! As long as you don't rewrite history that has already been pushed, then you're good.	18:21
	I ususally avoid it just to keep from having to know if the changes were pushed before or not	18:22
	and since I've been pushing out both branches the whole time to different remotes	18:22
Pistos	Eridius\|work: Right, that does seem to be the case, because my interactive shell can find it.	18:23
	sergio__ → sergiodj	18:24
jdahm	when I clone a repo on another computer, it seems to only copy over the current branch, should the other branches be added with git remote add ?	18:24
subtleGradient	jdahm: you can use `git fetch remoteName` to fetch everything. Or `git pull remoteName branchName` to fetch and track a specific branch	18:26
	→ drizzd_ joined	18:27
	→ SuttoL joined	18:28
	← SuttoL left	18:28
	← Bryson left	18:29
	→ Bryson joined	18:29
	← jamie_alm_ left	18:29
	→ jamie_alm joined	18:29
	→ kelvie_work joined	18:30
	← rdavila left	18:30
	← powr-toc left	18:33
	← warthog9 left	18:33
	→ queso_ joined	18:34
	→ koke joined	18:36
	→ chrisb joined	18:37
	← drizzd left	18:37
	→ david_koontz joined	18:38
	queso_ → queso\|away	18:39
	← FH` left	18:40
	→ vintrepid joined	18:42
	→ FH` joined	18:43
	← bughunter2 left	18:43
Pistos	Eridius\|work: It doesn't seem to want to do anything with my -u or --upload-pack argument. It still says it can't find it.	18:43
	← floppyears left	18:43
Pistos	I guess I can just symlink in /usr/bin or something...	18:43
	← queso\|away left	18:43
	← queso left	18:43
Eridius\|work	Pistos: curious	18:44
Pistos	Hrm, still doesn't like it...	18:44
	Basically, I am running of the git version of git; I ahven't installed it with my distro's package manager.	18:45
	All fine with an interactive shell, because I set the PATH right.	18:45
	But now this is getting all tripped up...	18:45
	← chrisb left	18:49
	davidfetter waves to cody	18:49
davidfetter	oops. mischan. d'oh	18:50
Eridius\|work	Pistos: what if you edit your .bashrc?	18:51
	← chris2 left	18:53
Pistos	Eridius\|work: I did. Doesn't seem to help.	18:53
Eridius\|work	hmm	18:53
Pistos	Also .bash_profile.	18:53
Eridius\|work	what did you pass to --upload-pack?	18:53
Pistos	Eridius\|work: The full absolute path.	18:53
	I even tried symlinking in /usr/local/bin and passing that.	18:53
	I think I'd like a --verbose or --debug of some kind, at this point...	18:54
	So I can see what it's really trying...	18:54
Eridius\|work	oh wait.. upload-pack is for pull. you want --receive-pack for push (which should be the path to git-receive-pack)	18:54
Pistos	I'm trying a git clone.	18:54
	Oh	18:54
	I see.	18:54
Eridius\|work	oh clone	18:54
Pistos	Okay.	18:54
Eridius\|work	why'd I think you were pushing?	18:54
Pistos	Not sure. :)	18:54
Eridius\|work	so upload-pack was the path to git-fetch-pack?	18:54
Pistos	No, it was the path to git-upload-pack... should it be fetch-pack?	18:54
	Pistos tries.	18:54
Eridius\|work	yes	18:54
	→ kumbayo joined	18:55
Pistos	Eridius\|work: Docs here are not very clear on that, then. :P http://www.kernel.org/pub/software/scm/git/docs/git-clone.html	18:55
Eridius\|work	oh wait, I'm still confusing myself	18:55
	the manpage is confusing me, actually	18:55
	it is supposed to be the path to git-upload-pack	18:55
Pistos	Anyway, it still doesn't help.	18:55
	→ warthog9 joined	18:55
Eridius\|work	strange	18:55
Pistos	Hrm, alright.	18:55
	git clone --upload-pack /path/... right?	18:56
Eridius\|work	yes	18:56
	← asdfdsfsd left	18:56
Pistos	Bizarre.	18:56
	→ [tla] joined	18:56
Pistos	Maybe I need to change the PATH on the pulling side?	18:56
	Nope, nothing's helping.	18:57
	Puzzling.	18:57
	← zed_DX left	18:58
Pistos	Same error with git ls-remote, BTW.	18:59
ToxicFrog	Isn't it --upload-pack=/path/...?	18:59
Pistos	I'll try.	18:59
	man page doesn't indicate that, though. :P	18:59
ToxicFrog	Hmm. Which man page?	19:00
Pistos	http://www.kernel.org/pub/software/scm/git/docs/git-clone.html	19:00
ToxicFrog	Typically it's -x foo or --long-x=foo	19:00
Pistos	Anyway, that didn't help.	19:00
gitsky	lol, feaping creaturism	19:01
	→ SRabbelier joined	19:02
Pistos	This thread seems to discuss the issue What's cooking in git.git (topics)	19:06
	er	19:06
	http://thread.gmane.org/gmane.comp.version-control.git/85705	19:06
	→ ciaran joined	19:08
	← nud left	19:09
	→ tomas_ joined	19:10
	→ alfaaprimo joined	19:10
	→ redondos joined	19:10
	← krawek left	19:10
	tomas_ → tomascejka	19:11
	← tomascejka left	19:12
	→ krawek joined	19:12
	← subtleGradient left	19:12
	→ tomascejka joined	19:12
Pistos	Okay, I give up. :(	19:13
ciaranm	if i'm on a branch with a remote, is there a command that'll show me the name of that remote?	19:13
Pistos	I guess I can try something besides ssh://	19:13
	If I give a bogus --upload-pack executable, it still says the same thing.	19:15
	It's like it's completely ignoring my option. :P	19:15
henr_k	when doing a git commit, I'm put into my editor, with the list of the files being commited. Are there any ways of getting the diff (git diff --cached) there instead?	19:16
Pistos	Okay, git ls-remote takes my upload pack. But not clone... hmm...	19:16
Tv	ciaranm: git config branch.master.remote, supply branch name as appropriate...	19:16
Pistos	henr_k: git commit -v	19:16
henr_k	Pistos: excellent! thank you! \o/	19:17
	← CodeOfficer left	19:17
ciaranm	Tv: ah, no single command then? thanks	19:17
gitsky	yay, patchbomb'd!	19:17
	now off I go	19:17
	← tomascejka left	19:18
Pistos	This just feels like buggy behaviour. :(	19:19
gitsky	then complain on the list?	19:19
	→ EiNZTEiN joined	19:19
Pistos	k, got it, by sshing and sending remote command "echo $PATH" to see that I wasn't really putting symlinks in the path.	19:21
	← zawir left	19:22
	← bryanl left	19:24
	→ ennemy joined	19:25
	← ennemy left	19:26
	← [tla] left	19:29
	→ elight joined	19:34
	← Pistos left	19:39
	← Fullmoon_ left	19:41
	← johnjosephbachir left	19:42
	← Ademan left	19:43
	← vintrepid left	19:44
	→ Fullmoon joined	19:49
	→ Ademan joined	19:50
	← sgrimm left	19:50
	→ Galaxor joined	19:51
Galaxor	Hi. I added an svn remote by editing my .git/config. I did git fetch svn-project. Now, when I look at the branches that I've got (git branch -r), I can see all the tags and branches that were on the svn project, but I can't find the trunk (which is what I want).	19:52
	→ gcj joined	19:55
gcj	hi all, can anyone help me to download a single branch from a git tree without getting the whole thing?	19:55
uau	gcj: "git fetch repository branchname" should work	19:58
	that only downloads it so it's available with the name "FETCH_HEAD" without automatically creating any local branch for it	19:59
	if you instead want to track the remote branch in the future too then use "git remote add" with additional arguments to restrict tracked branches	20:00
	→ paulttt joined	20:00
gcj	yeah i think i'd like to track the remote branch	20:00
	ok i did: git remote add git://git.kernel.org/pub/scm/linux/kernel/git/djbw/xscaleiop.git misc	20:01
	that doesn't give me any files though	20:01
	and git fetch gives an error	20:01
	(i want to track the misc branch)	20:02
uau	that command seems to be missing the name you give to the remote	20:02
	← earcar left	20:02
uau	is "misc" the name of the branch you want to track?	20:02
gcj	yep	20:02
	reading man git-remote now	20:03
	← Galaxor left	20:05
	→ johnjosephbachir joined	20:08
	← drizzd_ left	20:08
	→ drizzd joined	20:09
	→ queso joined	20:09
	← Fullmoon left	20:12
	← eikonos left	20:14
	← koke left	20:16
	→ koke joined	20:17
	→ [raz] joined	20:19
	← DrFrasierCrane left	20:21
ciaranm	what's the best way to get rid of a remote? just remove it and then gc to get rid of all the things it had fetched that aren't used?	20:25
	telmich would guess yes	20:25
Eridius\|work	gc won't remove objects newer than 2 weeks, though	20:26
ciaranm	ach, i can wait two weeks	20:26
	i just don't want it lying around forever	20:26
Eridius\|work	`git prun`	20:27
	*prune	20:27
	→ sgrimm joined	20:27
	← krawek left	20:29
	← janm left	20:31
	← raz left	20:31
	[raz] → raz	20:31
	→ Lars_G joined	20:31
Lars_G	He.... HOLY F....	20:32
	Wow this channel is full!	20:32
	Lars_G cowers in a corner	20:32
Eridius\|work	full? how can a channel be full?	20:32
	Eridius\|work looks vaguely menacing	20:32
	→ krawek joined	20:32
gcj	uau, ok i think i want the clone model	20:33
	so i used this command: git remote add -f -t misc -m misc origin git://git.kernel.org/pub/scm/linux/kernel/git/djbw/xscaleiop.git	20:33
Lars_G	I want to archieve something, a little difficult.	20:33
gcj	and it's been running for about 20 minutes with no output now	20:33
	→ Eludias joined	20:34
	blowery → blowery`afk	20:36
Lars_G	I want to do something like this http://show.degopro.com/branches.png	20:36
ToxicFrog	gcj: how big is the repo it's fetching?	20:37
Lars_G	the trouble is, there will always be some differences between branch a and branch b	20:37
gcj	ToxicFrog, it's a kernel tree, it might be big	20:38
uau	gcj: running?	20:38
Lars_G	so there are commits I want to be able to merge into either direction, but there are differences that should never be merged	20:38
gcj	that's why i just want to fetch one branch	20:38
Eridius\|work	Lars_G: that branch graph looks like a fairly normal situation	20:38
ToxicFrog	That's probably it, then	20:38
Lars_G	is that doable?	20:38
uau	'git remote' should return immediately	20:38
ToxicFrog	uau: not with -f	20:38
	Which is "immediately fetch the branch"	20:38
Eridius\|work	Lars_G: you mean you want to keep a branch that has a permanent change that never gets merged back into master?	20:38
Tv	Lars_G: there's nothing special in that graph	20:38
gcj	ok it finished now, thanks :)	20:38
Lars_G	Eridius\|work: correct	20:39
Eridius\|work	just do a merge --no-commit, edit out that change, and commit the merge. Now unless you modify that change in the future, any more merges will just ignore it	20:39
gcj	doesn't git fetch normally give some progrss indication?	20:39
ToxicFrog	gcj: pull does, I'm not sure if fetch does	20:39
Lars_G	Eridius\|work: so after the edit it will ignore it forever until I do another --no-commit merge (where I can choose to add it in or not)	20:39
	Eridius\|work: beautyfull. thanks a lot!	20:39
	After the edit, will I always have to do a merge --no-commit or normal merges will work ignoring the change I edited out	20:40
Eridius\|work	normal merges will work	20:41
	unless you edit that hunk again	20:41
	Lars_G bows	20:41
Lars_G	It's hard to get your mind around it's working manners, but git rules	20:41
Eridius\|work	heh	20:42
Lars_G	Ah well I come from cvs and then svn, git is very different in some aspects from them	20:43
	:)	20:44
	← X-Blade left	20:44
Eridius\|work	indeed	20:45
	→ X-Blade joined	20:45
	→ Arc_ joined	20:48
	← davidfetter left	20:53
	← Ademan left	20:54
	→ Ademan joined	20:54
	← harinath left	20:54
	→ tusho joined	20:55
tusho	If I add a file to .gitignore that's already in the repo, is it removed?	20:55
Eridius\|work	no	20:55
tusho	What do I do to remove it without removing my actual copy of the file then	20:56
Eridius\|work	git rm --cached file	20:56
tusho	What if it's a directory?	20:56
	Will that still work?	20:56
	→ robinr joined	20:56
Eridius\|work	git rm --cached -r dir	20:57
tusho	OK	20:57
	Thanks	20:57
	← Arc left	21:01
	→ earcar joined	21:02
	→ vintrepid joined	21:02
	→ dduncan joined	21:03
	→ Fullmoon joined	21:03
	← dduncan left	21:04
	← X-Blade left	21:04
	← lorandi left	21:06
	→ ciaran_ joined	21:08
	← vintrepid left	21:08
	→ X-Blade joined	21:09
	← elight left	21:10
SRabbelier	Pieter\|Paris: ping?	21:14
	blowery`afk → blowery	21:15
	← mxey left	21:17
	← ciaran left	21:17
	← jamie_alm left	21:19
	← earcar left	21:21
	→ markkalderon joined	21:21
	← priidu left	21:22
lea_w	how come i have unpacked objects left in my object store after git repack and git gc?	21:23
	→ CodeOfficer joined	21:24
	→ zawir joined	21:24
Mikachu	unreachable objects younger than 2 weeks?	21:24
	→ elight joined	21:24
lea_w	hm... how can i get rid of them?	21:24
	without also expiring my reflogs i.e.?	21:25
	or i could also just pack them. just as long as they aren't loose ;)	21:25
	lea_w wonders...	21:25
Mikachu	git prune should remove them	21:25
drizzd	not being packed is the definition of loose	21:25
lea_w	Mikachu: huh, right. thanks!	21:26
Mikachu	i think gc invokes prune with the --expire switch set to 2 weeks, while the default is zero when you invoke it directly	21:26
	→ kallepersson joined	21:26
	← grahal left	21:26
lea_w	so why is the expiration time set to 2 weeks by default?	21:27
	would you ever need those unreachable objects?	21:27
Mikachu	safety measure?	21:27
kallepersson	Hello there, quite a newbie question as I have not used VCS before: I'm reading gittutorial and it suggest to do git add $file_i_just_edited. Do i really have to do that before each commit? Isn't that stuff automatied?	21:27
pygi	kallepersson, git commit -a will do that for you	21:27
kallepersson	-a for auto?	21:28
Mikachu	or git commit $file	21:28
pygi	unless it's a new file, then you have to do "git add file"	21:28
	→ evanlight joined	21:28
kallepersson	I see	21:28
	makes sense	21:28
pygi	kallepersson, no, for "add" :P	21:28
kallepersson	ah :)(	21:28
Mikachu	actually i think it's for "all"?	21:28
kallepersson	but will git commit -a check all files I have changed and commit them?	21:28
pygi	yes	21:28
Mikachu	yes	21:28
kallepersson	but I'd need to add them manually to be able to use git diff --cached	21:29
pygi	Mikachu, yes, it's "all" :)	21:29
	sorry :p	21:29
Mikachu	kallepersson: obviously	21:29
kallepersson	then it diffs my changes to the upstream version.	21:29
	so obviously	21:29
Mikachu	or you can not add it and just use "git diff"	21:30
lea_w	kallepersson: (no, to the last commit)	21:30
kallepersson	then it won't diff that file	21:30
	ah, yeah.	21:30
drizzd	or git diff HEAD	21:30
	kallepersson tries with -a	21:30
	← jamessharp1 left	21:30
kallepersson	commit, that is	21:30
	← deavid left	21:31
drizzd	kallepersson: btw., if you delete files using git rm, you don't have to tell git about it in a separate steps.	21:32
kallepersson	ah	21:32
	how would i tell it?	21:32
drizzd	git add or git rm --cached	21:32
	Arc_ → Arc	21:33
kallepersson	okay	21:33
drizzd	but I never actually use that unless I accidentally delete a tracked file using just rm instead of git-rm	21:33
kallepersson	good, just wanted to know	21:34
	obviously you'd use git-rm	21:34
	very nice that git-rm is a command	21:34
	instead of git rm	21:34
drizzd	hmm? git-rm and git rm is the same	21:34
kallepersson	so you'd go like git-r[TAB]. faster to use	21:34
	sorry, that's what I meant. I'll also stop using enter as punctuation.	21:35
drizzd	there is a script which tells bash about those tab completions, even without the hyphens	21:35
	it's in contrib/ if you compiled git from source.	21:35
gitsky	kallepersson: don't get used to that	21:36
	kallepersson: the dash-form is going away with new git release	21:37
	→ entie joined	21:37
kallepersson	oh. oh well. I guess that it won't hurt to type a few more letters.	21:37
gitsky	it may be faster to use but (i) there is a bash completion file for git (ii) there's about 150 of these ;)	21:37
kallepersson	yeah, bash said 136	21:37
	git is really awesome when you get used to it!	21:38
pygi	132	21:38
	:p	21:38
kallepersson	heh, oops.	21:38
drizzd	148	21:38
	drizzd wins	21:38
kallepersson	hah, watching the Linus Git video. he doesn't do backups :)	21:38
pygi	drizzd, patience :p	21:39
kallepersson	hm, I did git rm foo.txt and then git diff --cache	21:40
	nothing.	21:40
	→ eikonos joined	21:40
kallepersson	git commit did catch it though	21:40
	so, how often should\|do you commit with a regular project? after each big change or just randomly?	21:42
drizzd	you should get something like this from git diff --cached:	21:42
	diff --git a/csdf b/csdf	21:42
	deleted file mode 100644	21:42
	index e69de29..0000000	21:42
Gitbot	[git e69de29]: http://tinyurl.com/5lvq4r [blob]	21:42
kallepersson	didn't.	21:43
jast	gotta love Gitbot	21:43
kallepersson	created file, committed file. rm'd file, ran git diff --cache ... without d.	21:43
	← bongiojp left	21:43
drizzd	git diff --cache	21:43
	error: invalid option: --cache	21:43
	← elight left	21:44
	evanlight → elight	21:44
gitsky	kallepersson: I think that comes from experience	21:45
kallepersson	I did not see that error message due to less popping up on my console. too enthusiastic i guess :)	21:45
	indeed.	21:45
jast	as a rule of thumb, try to make each commit contain a logical unit of changes	21:45
gitsky	kallepersson: it's best to commit very often but there's certain overhead associated with committing for you (writing commit message, testing, etc.) and you should keep your eye bisectable	21:45
jast	if in doubt, err in favour of more commits	21:45
gitsky	kallepersson: but I'd say committing more often than optimal is better than committing less	21:45
kallepersson	hm, for instance	21:46
	after writing a random larger function, or even perhaps a class.	21:46
	and probably each time I'm about to do something else than code on that project.	21:46
	← shaftyy left	21:48
jast	sounds reasonable	21:48
	though you could argue for combining in a single commit a new feature, related tests and documentation	21:49
kallepersson	makes sense	21:49
gitsky	lea_w: (I'll try to reply to your last mail ASAP, but just in case, in case you decide to do Git::Command or whatever, please drop me a message so that we don't do duplicate work)	21:50
drizzd	I guess everybody develops his/her own style. Personally, I commit as soon as it compiles and runs. Usually I find out I broke something a few commits later.	21:50
gitsky	lea_w: (wrt. the gitweb changes, I'm waiting for splitted patches - as much as I'd like to, I don't currently have time to figure out the huge one :( )	21:50
kallepersson	ah, that's nice	21:50
Mikachu	it's easier to combine small commits into a big one than the opposite	21:51
kallepersson	I'm probably going to do mostly web stuff though	21:51
gitsky	lea_w: (and I'm not saying I'd do Git::Command today or tomorrow ;)	21:51
	← LiamH left	21:51
	← SRabbelier left	21:52
	← jerbear left	21:53
lea_w	gitsky: i'm not planning to add anything like Git::Command atm	21:53
	gitsky: yes, the next series will have the gitweb changes split	21:53
	gitsky: you have a lot of names btw ;-)	21:53
	→ seed joined	21:54
	→ SRabbelier joined	21:54
	← Eludias left	21:54
lea_w	does anyone know a good way to split a large patch? is there any better way than removing chunks from a patch file?	21:55
	i.e. anything better than using an editor?	21:55
drizzd	git rebase -i and then git add -p or gitk	21:57
	← Fullmoon left	21:57
	→ tiglionabbit_ joined	21:58
drizzd	s/gitk/git gui/	21:58
lea_w	drizzd: will i be able to conveniently remove and add chunks after the first pass though?	21:59
drizzd	I'm not sure about conveniently removing chunks with git add -p, probably have to use git add -i for that.	22:00
	You can remove chunks in git gui though	22:00
lea_w	ok... nods	22:00
	← Yuuhi left	22:00
lea_w	oh, hm, didn't know git gui did that. thanks!	22:00
	← juanjoc left	22:01
	→ Fullmoon joined	22:01
	→ juanjoc joined	22:02
	← krh left	22:02
	← kumbayo left	22:03
	← juanjoc left	22:03
	← sgrimm left	22:03
	← emachado left	22:05
	blowery → blowery`afk	22:08
	blowery`afk → blowery	22:10
	cbreak\|afk → cbreak	22:10
	← krawek left	22:10
	← paulttt left	22:10
	← FunkeeMonk left	22:12
	← paltman left	22:13
	← _graham_ left	22:15
	→ evan joined	22:16
evan	how does git detect that a blob is binary?	22:16
	← robinr left	22:19
lea_w	evan: where does it detect that?	22:20
	blobs are normally opaque, well, blobs of data	22:21
evan	well, "git format-patch" detects it to know whether to generate a binary patch or not	22:21
Eridius\|work	evan: xdiff-interface.c detects binary by looking for a 0 in the first 8000 bytes (dunno what xdiff-interface.c is for though)	22:21
evan	Eridius\|work: oh, interesting.	22:21
	ok	22:21
	i'll look in xdiff	22:22
Eridius\|work	convert.c detects binary by looking if the "stats" contains any NULs, or if the number of printable chars >> 7 is less than the number of nonprintable chars	22:22
	dunno what that's used for either though	22:22
evan	hm	22:22
	well, thanks	22:22
	those are 2 good places to start with.	22:22
cebewee	if I post a patch series, where some patches are from another author, do I need to add my SoB to these patches, too?	22:24
telmich	well, with SoB you just say that you trust your source	22:24
	→ BigCanOfTuna_ joined	22:25
	← dramsay left	22:25
jengelh	I trust myself, always :p	22:25
telmich	that's more than I do :-)	22:27
jast	xdiff is probably the right place to look	22:27
	cebewee, it's encouraged, yes (and it can't hurt, so what the heck :)). preferably there should by a SoB by the original author too, of course.	22:28
	← BabelO left	22:29
	→ johnw- joined	22:30
MarcWeber	Does git-config support listing all keys of a section?	22:31
Mikachu	you can -l \| grep	22:32
	← BigCanOfTuna left	22:32
MarcWeber	Mikachu: Ah fine.	22:32
Mikachu	(that is not a no as such, just a suggestion)	22:33
MarcWeber	It's ok for my little script	22:33
	→ CrnkyGeek joined	22:34
	→ BigCanOfTuna__ joined	22:36
	BigCanOfTuna__ → BigCanOfTuna	22:36
	← vbabiy left	22:37
Weasel[DK]	What is the correct way to edit the .git/description file ?	22:37
	← aroben_ left	22:38
jast	vim :)	22:38
gitster	emacs.	22:38
ciaranm	ed.	22:38
Weasel[DK]	he he funny... ;)	22:38
ciaranm	ed is the standard text editor.	22:38
	gitster says "holy war begins!"	22:38
Mikachu	$EDITOR .git/description	22:39
Gitbot	Mikachu: Unknown command or plugin 'EDITOR'.	22:39
gitster	sensible-editor ;-)?	22:39
Weasel[DK]	lol	22:39
jast	M-x butterfly	22:40
Weasel[DK]	what i mean is.. how does it get distributed... dont seem push is working here	22:40
Mikachu	it doesn't	22:40
	blowery → blowery`afk	22:40
Weasel[DK]	so it is only working downstream ?	22:41
jast	ah, clearly i don't know emacs. disregard my advice.	22:41
Mikachu	i think only gitweb reads that file?	22:41
jast	more like off-stream	22:41
Mikachu	i only know c-x c-c	22:42
Weasel[DK]	when you clone then	22:42
	← BigCanOfTuna_ left	22:45
Weasel[DK]	i just tried it out... the changes .git/description are not distributed when clone'ing.	22:46
	← SRabbelier left	22:47
	→ nud joined	22:48
	→ SRabbelier joined	22:48
uau	Weasel[DK]: it's only for local use	22:48
	→ ereslibre_laptop joined	22:49
ereslibre_laptop	hi there	22:49
Weasel[DK]	uau, so i see... a bit strange i think	22:49
ereslibre_laptop	after merging a branch, should I delete it, or is it good to keep it ? (no more work will be comitted on it). I have that branch both, local and remote	22:49
johnw-	even if you delete it, its commits will not be deleted	22:50
	it's just the name you're getting rid of	22:50
ereslibre_laptop	I know, but the history will identify that the branch happened ?	22:50
Mikachu	yes	22:50
ereslibre_laptop	the merge, sorry	22:50
johnw-	absolutely	22:50
ereslibre_laptop	great. thanks	22:50
Mikachu	the history doesn't care about branches really	22:50
ereslibre_laptop	I see	22:50
Mikachu	ie, about branch refs	22:50
johnw-	if you want to view the merged commits, you can use for example git log HEAD^2	22:51
Mikachu	the actual commits have info about the history	22:51
	→ amigrave joined	22:51
johnw-	while git log HEAD will just show the merge commit itself, not the commits that made it up	22:51
	← johnjosephbachir left	22:51
	→ mib_a7mj00 joined	22:51
Mikachu	johnw-: hm? pretty sure it will follow both legs	22:52
johnw-	git log will?	22:52
amigrave	how can I check a diff of only the last committed changes made to a particular file using command line ?	22:52
ereslibre_laptop	strange git-merge doesn't have an option for that, for instance git-merge -d foo	22:52
johnw-	i thought its output was linear, I could be wrong	22:52
ereslibre_laptop	(-d will delete foo after merging)	22:52
Mikachu	johnw-: git log --graph	22:52
johnw-	amigrave: git diff HEAD^ -- file?	22:52
	Mikachu: ah cool, thanks!	22:53
amigrave	johnw-: won't work. But there's something strange. When I type only "git diff", the output is empty !?	22:55
	← mib_a7mj00 left	22:55
johnw-	that just means you don't have uncommited changes	22:55
uau	amigrave: "git log -p -n 1 file" at least (it'll show the whole commit though, not just diff)	22:55
Mikachu	johnw-: it means he doesn't have unadded changse	22:55
uau	-n 1 means to show one commit only	22:55
Mikachu	-1 also works	22:55
johnw-	yes, I misspoke	22:56
Mikachu	maybe git show $(git rev-list -1 file) file	22:56
johnw-	why not just git show HEAD^ -- file?	22:56
	hmm git diff HEAD^ -- file works here, but show doesn't	22:56
Mikachu	the changes might be in HEAD~50	22:57
johnw-	ah	22:57
	→ jackdempsey joined	22:57
Mikachu	johnw-: git show HEAD -- file	22:57
johnw-	i'm just off my game today!	22:57
	← Mishakun left	22:59
amigrave	uau: Mikachu: johnw-: "git log -p -n 1 file" works by showing last commit but it doesn't seems to show only diff for specified file. If I do "git log -p -n 50 file" the outputted diff is showing modifications in other files than the specified file	23:01
Mikachu	amigrave: read what i wrote	23:01
	or for that matter, what uau wrote	23:01
	→ \ask_ joined	23:02
amigrave	Mikachu: ah yes ! sorry ! my fault	23:03
	← SRabbelier left	23:03
	→ EiNZTEiN_ joined	23:04
	← EiNZTEiN left	23:04
	← \ask_ left	23:05
MarcWeber	I've written my first git script.. Maybe its useful for you as well?	23:05
	→ \ask_ joined	23:05
MarcWeber	You can describe test merges and redo the merge using git-test-merge name..	23:05
	http://rafb.net/p/OiWgEU19.html	23:05
ereslibre_laptop	hmm, i am getting this error when trying to push	23:06
	! [rejected] master -> master (non-fast forward)	23:06
MarcWeber	ereslibre_laptop: This means someone has replaced one of the patches by another one..	23:06
	(which is called changing history and should be avoided for exactly this reason)	23:07
	← Ademan left	23:07
	← heycarsten left	23:07
MarcWeber	You can either push -f (override the target. this could loose commits) or fetch that branch and rebase your changes	23:08
ereslibre_laptop	hum, I just merged a branch locally (git checkout master; git merge foo; git push origin master; git branch -d foo; git push origin :heads/foo)	23:08
	MarcWeber: is that i did it wrongly ?	23:08
	now I did some commits locally on master and get that error when trying to git push origin again	23:09
uau	are you sure nobody else pushed commits in the meantime?	23:09
MarcWeber	ereslibre_laptop Create a new branch based on your master	23:09
	git-branch new master	23:10
	Then you can't loose your modifications.	23:10
ereslibre_laptop	uau: this is a testing thing in which the one that only commit is me ;)	23:10
MarcWeber	Than you can git-pull -f to reset your local master to the same hash than the remote one.	23:10
uau	if the commit you pushed is still in the history of your current master branch then you should not get that message	23:10
MarcWeber	After having done this you can use gitk --all to watch where both branches (new and master) diverged	23:11
uau	non-fast forward means that the remote state is not in the history of the commit you're trying to push	23:11
	← brosner left	23:11
ereslibre_laptop	MarcWeber: gitk == qgit ?	23:12
MarcWeber	No. Don't konw qgit. But I guess you can view multiple branches using qgit as well	23:12
	gitk is based on tcl/tk	23:12
ereslibre_laptop	yep	23:12
MarcWeber	ereslibre_laptop: If you do this there there is a common ancestor of new and master. Something has gone wrong here. It should have been commit - commit - master - commit - commit - new	23:13
ereslibre_laptop	MarcWeber: after creating the new branch and pulling, I see the old one again	23:14
	← d0k left	23:14
MarcWeber	ereslibre_laptop The old one you've pushed the first time?	23:14
ereslibre_laptop	yep	23:14
	now i think I can just merge newmaster with master	23:15
	and I am done ?	23:15
MarcWeber	If this is what you want /need then yes	23:15
	Is qgit more powerful than gitk ?	23:16
ereslibre_laptop	hvent used gitk	23:16
	MarcWeber: wait a sec, i have a small mess in here. lets see if taking a shot..	23:16
	MarcWeber: that way you can see how qgit is too	23:16
MarcWeber	I'm looking at the screenshots on sourceforge..	23:17
	I really like gitk though I haven't figured out how to find commits in history touching a path and removing the hash to enter a new one doesn't accept ctrl-backspace :(	23:17
	→ lolage0 joined	23:18
	← jackdempsey left	23:18
MarcWeber	And gitk can't filter history on selected code range, can it?	23:18
ereslibre_laptop	MarcWeber: what ihave is this http://media.ereslibre.es/2008/07/git.png	23:19
	MarcWeber: basically what I want is master and gitorious/master to be at the same commit level	23:19
	← EmilMedve left	23:20
ereslibre_laptop	it seems master is "after" gitorious master. I guess I only need to do a checkout on master and pull ?	23:20
Eridius\|work	"after"?	23:20
	master and gitorious/master have diverging histories	23:20
MarcWeber	If they were they would be on one line	23:20
	→ Ademan joined	23:20
ereslibre_laptop	how can that be if I did a git push gitorious master ?	23:21
MarcWeber	Good question.	23:21
Eridius\|work	does anybody else have commit access?	23:21
	err, push access?	23:21
ereslibre_laptop	nope	23:21
MarcWeber	Scroll down and see where the history differs	23:21
Eridius\|work	then where did the "Merge branch 'smooth'" commit come from?	23:21
ereslibre_laptop	only on the commit you see there	23:22
	"merge branch smooth"	23:22
Eridius\|work	I mean, if that wasn't in your master, and nobody else has push access, how'd that get into gitorious/master?	23:22
MarcWeber	ereslibre_laptop: Scrool down and watch the difference red and black line..	23:22
ereslibre_laptop	hehe... that is. dunno	23:22
	black line have some commits	23:23
	red one, only the "merge branch smooth"	23:23
	← Weasel[DK] left	23:23
ereslibre_laptop	is the only commit that black hasn't. and is the only commit on that branch actually	23:23
MarcWeber	And why did you merge red and green?	23:24
Eridius\|work	ereslibre_laptop: ya know, maybe you should just force-push master to gitorious and then merge in 'smooth' there	23:24
ereslibre_laptop	MarcWeber: that's a pretty good question. I just did the merge, and later did a git push gitorious master	23:25
MarcWeber	The "Add the codebase for".. patch should have been on master (black line) I guess?	23:25
Eridius\|work	it actually was on master as well	23:26
	same patch on both lines	23:26
	→ jengelh joined	23:26
ereslibre_laptop	MarcWeber: no, that ine is in 'smooth' and is fine there ;)	23:26
	ok, so now it seems it is working	23:27
	what i did is	23:27
	what you suggested ebfore	23:27
	create a branch	23:27
	newmaster	23:27
	merge newmaster on master	23:27
	git push gitorious master	23:27
	that id it	23:27
	did it*	23:27
lea_w	What's the difference between 'git checkout -v newbranch' and 'git branch newbranch'?	23:28
	s/-v/-b/	23:29
Eridius\|work	git branch doesn't checkout the branch	23:29
	it just makes it	23:29
	`git checkout -b newbranch` is the same as `git branch newbranch && git checkout newbranch`	23:29
lea_w	Eridius\|work: oh, i see. thanks!	23:29
	→ mxey joined	23:29
MarcWeber	ereslibre_laptop This doesn't make sense. The only commit on the red line is the merge?	23:30
ereslibre_laptop	yes	23:30
	¬¬	23:30
	that is what confuses me	23:30
	MarcWeber: maybe a qgit fault ?	23:30
	→ dramsay joined	23:30
	→ clsdaniel joined	23:31
	← p4tux left	23:31
MarcWeber	ereslibre_laptop Don't get it.. Why don't you just merge master and with "Add the codebase for a smooth .. "?	23:31
ereslibre_laptop	that is what i tried to do	23:31
	maybe when pushing it i did something wrong ?	23:32
	dunno	23:32
	¬¬	23:32
MarcWeber	ereslibre_laptop: Well.. That's the cause then..	23:32
Eridius\|work	a push cannot create a merge	23:32
	but whatever happened, we're not going to be able to figure it out	23:32
	as long as you got your history in a good state now, that's all that matters	23:32
ereslibre_laptop	it is very probably that i pushed when being on a different branch or something ? :/	23:32
	← _zack left	23:32
MarcWeber	ereslibre_laptop: You've got two masters: the remote one and the local one. And qgit did the merge with the older remote master..	23:33
ereslibre_laptop	qgit only reads, it doesnt do anything	23:33
Eridius\|work	MarcWeber: ahh, good thinking	23:33
ereslibre_laptop	MarcWeber: yes, that is what i have	23:33
	MarcWeber: master in there, and master in here	23:33
	i work here, commit	23:33
	and later, git push gitorious master. is that workflow wrong ?	23:34
MarcWeber	ereslibre_laptop: So create a new branch master2 and merge taht with the green commit.	23:34
cardioid	I read that as "git push glorious master"	23:34
	And I thought maybe you liked git a bit too much	23:34
ereslibre_laptop	xD	23:34
MarcWeber	ereslibre_laptop: No I'm talking rubbsih.. just tell qgit to use the correct branch when merging.. :)	23:34
	ereslibre_laptop installs gitk	23:35
MarcWeber	ereslibre_laptop gitk can't do merges	23:35
ereslibre_laptop	qgit can't either. qgit only reads ur repository	23:35
	→ GeckoPie_ joined	23:36
MarcWeber	Then I don't see why you want to get gitk? Maybe because it copies the hash to clipboard as soon as you mark a commit?	23:36
ereslibre_laptop	omg, this seems an scalextric	23:36
	← GeckoPie left	23:36
	→ Fullmoon_ joined	23:37
ereslibre_laptop	hehe cool	23:38
	http://media.ereslibre.es/2008/07/gitk.png < MarcWeber ;)	23:38
	→ jamie_alm joined	23:38
	ereslibre_laptop wonders wth did he did ;)	23:39
	← jamie_alm left	23:39
	→ amystrat joined	23:40
	← cilly left	23:40
	→ jackdempsey joined	23:41
	← ciaran_ left	23:41
amystrat	Having trouble with gitosis install	23:42
MarcWeber	What is gitosis?	23:42
amystrat	http://scie.nti.st/2007/11/14/hosting-git-repositories-the-easy-and-secure-way	23:42
Eridius\|work	ereslibre_laptop: you merged gitorious/master	23:42
	so it looks a bit odd	23:42
	cebewee has problems writing a test for the suite without stupid bugs =)	23:43
	← GeckoPie_ left	23:43
	← elight left	23:44
	← koke left	23:45
ereslibre_laptop	nah the problem seems to be git-svn	23:46
	git svn rebase ; git push gitorious master doesnt like it	23:46
Eridius\|work	well no, you've just rebased. if you rebase a branch you've published, you're going to have issues	23:47
ereslibre_laptop	Eridius\|work: so the problem is "git-svn rebase" ?	23:48
	← nud left	23:48
MarcWeber	ereslibre_laptop: Yes.	23:48
ereslibre_laptop	great, finally found	23:48
MarcWeber	It has only rebsed your local branch, but not the one one the server..	23:48
	ereslibre_laptop: Have a look at git-rerere --help	23:49
	→ SRabbelier joined	23:49
ereslibre_laptop	ok, thanks	23:49
MarcWeber	ereslibre_laptop: And if you want to keep a merge up to date merging the svn trunk and your fork for testing only you can use my script :-)	23:50
	gitster notices gitte also sounds like a poet ;-).	23:50
ereslibre_laptop	MarcWeber: hm, what I want is: get commits from kde svn, do my work locally, commit to gitorious in case my computer breaks, commit to kde	23:51
	what i need is the svn up that isn't git-svn rebase	23:51
	git svn fetch is probably the only thing i need	23:52
	→ davidfetter joined	23:54
	← Fullmoon left	23:55
	tommorris_ → tommorris	23:57
	← bronson left	23:59

Logs Search ←Prev date Next date→

Channels Documentation