IRCloggy #git 2007-09-05

Provider of IRC logs since 2005.

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

	2007-09-05
jlh	yes, that's what I'm wondering about. first I though it only sorts for size, so that similar objects are close together.	00:00
gitster	We sort by type (commits first, I think, but it does not really matter, as we never deltify across object types --- meaning a tree object is _never_ represented as a delta against a blob object).	00:00
	Now, then within the same type, we order by "name" but actually we do not use the full string.	00:00
	We give each name condensed "key" so that Makefile, sub1/Makefile, sub/dir/Makefile are sorted close to each other.	00:01
	which means we will sort so that contents that are expected to be similar to each other are grouped together, using the filename as hint.	00:02
	Essentially, that condensing "key" is a weighted sum of characters in the filename, and later part has more weight (so .c files sort together and .h files sort together).	00:02
	Also we sort by size, larger to smaller.	00:03
	So what typically happens is:	00:03
	(1) largest one is stored, as just gzip compressed.	00:03
	(2) other similar ones close by (defined by the above sort order) are compared with what we have already looked at, and we find the delta that is smallest to represent it.	00:04
	Now, our delta algorithm can express two operations:	00:04
	a. put byte range from here to there taken from the preimage (delta base) to this offset in the post image;	00:05
	b. put this literal byte sequence N-bytes long in the postimage at this offset.	00:05
jlh	ah yes, I think it uses xdelta for that, right?	00:05
gitster	Because we try to look at larger ones first, this tends to allow us expressing changes as removals --- which means we do not have to encode anything at all.	00:05
	Obviously we can not afford to look at _all_ objects we have looked at so far as candidates of delta base when we try to compute delta for a new object.	00:06
	So we have a sliding window of size N (that is what --window= parameter specifies), and check that many past objects.	00:07
jlh	I see. That phrase in the man page made me think it sorts only by type and then size, which I believe wouldn't scale well with lots of objects.	00:07
gitster	Another thing we do is even if an object _can_ be represented as a very small delta against something else,	00:07
jlh	and 'name' wasn't clear to me what it means, because blobs don't have names by themselves	00:07
gitster	we avoid making a delta against something that itself is a delta which in turn is a delta of something else --- and the chain to reach the final object is too long. That is controlled by --depth= parameter.	00:08
	← ferdy left	00:08
gitster	Now, after we know how each of the objects will be represented in the resulting pack (either "just compressed", or "expressed as a delta against something else"), we write things out, largely in the same order as rev-list gave us.	00:09
jlh	Is there a way to dump the content of a pack to see how it expresses objects?	00:10
gitster	with one twist: when it is turn to write out an object that is represented as a delta against something else, that something else must have been written out already. This rule hoists the base object in the order of final object stream.	00:10
	See documentation/technical/pack-format and then "verify-pack -v". There is a script to condense its output in contrib/stats/.	00:10
jlh	ok will do. thanks a lot for that explanation :) much clearer to me now	00:11
	→ mayz joined	00:13
	← mayz left	00:13
	← FunkeeMonk left	00:17
	→ jcollie joined	00:23
	← sinequanon left	00:27
	← kanru left	00:28
	gitte says good night.	00:29
	← gitte left	00:30
	← agoode left	00:30
	← devogon left	00:44
	→ flo joined	01:07
	← yorgen1512 left	01:11
	← rodserling__ left	01:17
	→ BadQuanta joined	01:21
	← rlb3 left	01:21
	← puzzles left	01:25
	→ puzzels joined	01:25
	→ DrNick joined	01:26
	← KirinDave left	01:27
	puzzels → puzzles	01:27
	→ rlb3 joined	01:30
flo	Hi ! I'm using git with a shared repo on a server (mainly for keeping my different machines in sync). Anyway i noticed that if i create a new branch on one machine, commit it, push it to the server then the new branch will now show up on a second machine that pulled from the server later	01:39
	is there a way to always push/pull all local branches ? (or is there something like a global branch ?)	01:39
jasam	flo, branches different from master are pulled as remote branches, you can see them with git branch -r	01:43
	← sgrimm left	01:45
jasam	gitster, I sent another try with a function for updating refs, generalizing the others	01:46
flo	jasam: great, thanks!	01:47
jasam	flo, :-) they also will remain unchanged, to have always a copy of the original pulled branch	01:48
	← QuercusMax left	01:50
	→ theCarpenter joined	01:57
	theCarpenter → Dragon45	01:57
	← lizac left	01:58
	← Yuvi left	02:08
Dragon45	what's the upper limit on the size of non-text files that git can handle?	02:11
	images, .exes, etc...	02:11
context	im guessing whatever your fs can handle	02:14
flo	My gitweb installation always says : "403 Forbidden - No project found." also i did chmod -R 777 on my repodir...	02:16
context	are you sure you gave it the right path to your project[s]	02:21
	← myrizio_ left	02:24
	→ dduncan joined	02:26
	← tongueroo left	02:27
	→ raeLLL joined	02:29
	→ QuercusMax joined	02:31
	← QuercusMax left	02:34
	← whee left	02:41
Dragon45	is git-daemon a requisite for git-web?	02:43
tcoppi	no	02:44
BadQuanta	Ok. I'm testing moving my project to git.. wish me luck.	02:55
	← HackyKid left	02:55
BadQuanta	Quick Question: I've got all my projects stored in one SVN repo.. I think I've read thats not the way you want to run GIT.. is that correct.	02:55
	← flo left	02:59
	→ flo joined	03:01
	← jasam left	03:02
Tv	BadQuanta: yes	03:02
BadQuanta	Ok. just checking.	03:03
	← aroben left	03:09
	← madewokherd left	03:17
	→ tongueroo joined	03:18
	→ KirinDave joined	03:18
fidodido	hey guys	03:21
	git rocks	03:21
	like amazing	03:21
	really imrpoved the way i work	03:21
	and can feel it	03:21
	← rlb3 left	03:21
fidodido	how do i ignore some files during a git merge	03:21
	is there like a .gitmergefile	03:21
	→ rlb3 joined	03:22
fidodido	or how to ignore binary files during a merge	03:22
	→ DrNick2 joined	03:23
	← flo left	03:23
fidodido	how do we ignore files from git merge	03:23
	← DrNick left	03:25
	DrNick2 → DrNick	03:25
	→ FauxFaux joined	03:26
	→ aroben joined	03:28
	Randal tries to figure out what "ignore files during a merge" might mean	03:28
Randal	you have C that's the result of A + B. And if parts of A and B conflict, you have to tell it which.	03:29
	that's the only reason you'd get involved.	03:29
context	randal: maybe re-read the whole paragraph over again	03:30
Randal	I did, three times.	03:30
	I still don't get it.	03:30
context	often times things dont sound correct out of context.. much like you just pasted to us	03:30
Randal	it's gonna require some context or motivation.	03:30
context	wheredoes it say tht	03:30
Randal	if you have a commit A and another commit B, you have to resolve every single file in both	03:31
	it doesn't make sense any other way.	03:31
context	wow, forget i even asked a question	03:31
Randal	I'm talking to fidodido	03:31
	not you	03:31
	→ fhobia joined	03:31
Randal	... <fidodido> how do i ignore some files during a git merge	03:32
context	oh ok. now i see	03:32
Randal	answering that	03:32
	trying to anyway	03:32
	but he's now apparently gone	03:32
	so I'm just wasting my breath. :)	03:32
context	randal: srry. ;)	03:32
Randal	heh no prob	03:32
	I do that sometimes too	03:32
	hard to tell which way you're facing in virtual space.	03:33
	FauxFaux peers at git-svnmirror.	03:33
	← cmarcelo left	03:33
gitster	often people mean "keep other person's version because what I produce is crap".	03:33
fidodido	randall	03:33
	Randal: i ve a few branches	03:33
	master	03:33
	slave	03:33
	and one more	03:34
	i want to pull some changes from master to slave	03:34
	but i dont want to pull all the changes	03:34
	i want to ignore some files	03:34
	so i jus want some code from the master	03:34
gitster	Sorry, you are not thinking git yet.	03:34
Randal	yeah - not thinking git	03:34
fidodido	but i want to ignore binary	03:34
Randal	that's why I was confused too	03:34
fidodido	and some other files	03:34
gitster	That's _not_ a merge.	03:34
fidodido	i do	03:34
	i do git checkout slave	03:34
	git merge master	03:34
	and i want to ignore certain .html files	03:35
Randal	do you understand what a "merge" is for, fidodido?	03:35
fidodido	during merge	03:35
gitster	with that command, you are telling git that you want whole tree merge.	03:35
Randal	I don't think you do.	03:35
fidodido	ok	03:35
Randal	at least not the git meaning of merge	03:35
fidodido	i want to merge the changes from the other branch	03:35
chowmeined	i dont understand.. how do i just fetch a remote branch and then use it?.. i cant find it anywhere	03:35
fidodido	ok	03:35
	so how do i pull some changes from a part branch	03:35
	master	03:35
	to slave	03:35
	mebe i dont need to do merge	03:36
	is there something else	03:36
	i can do	03:36
chowmeined	i fetched it with git fetch git://location/git/item.git core	03:36
fidodido	this is imperative	03:36
Randal	a merge is a third tree that results from mixing two other trees, representing a complete replacement for either of them.	03:36
gitster	fidodido: if you have three changes (three separate commits) on master,	03:36
fidodido	becuase	03:36
Randal	in other words, a "better" version than either of the parent versions.	03:36
fidodido	these are 2 projects	03:36
gitster	one touches file A B, another touches C (binary), and the other one touches A and D,	03:36
fidodido	master and slave	03:36
	and each share code	03:36
Randal	you want subprojects	03:36
fidodido	oh ok	03:36
Randal	not trees.	03:36
fidodido	how do i do that	03:36
gitster	you _COULD_ say "I want to cherry-pick the first and the third one".	03:36
Randal	(I think)	03:36
fidodido	randal	03:36
	→ aroben_ joined	03:36
fidodido	i think u got it	03:36
	bulls ey	03:36
	e	03:36
	:)	03:36
	so how to do subprojects	03:37
	in git	03:37
	← vdv left	03:37
gitster	Randal: I fail to see how this has anything to do with subprojects.	03:37
chowmeined	i fetched something.. but i cant find it anywhere	03:38
	what do i do to just fetch a branch from a remote repo?	03:38
aeruder	git branch -a	03:38
	→ yorgen1512 joined	03:38
aeruder	^^ chowmeined	03:38
Randal	it sounds like he has trees with independent histories	03:39
	that's what subprojects are for	03:39
fidodido	Randal: the current version of master	03:39
	← Sho_ left	03:39
fidodido	and current version of slave	03:39
	brance	03:39
	are 2 projects	03:39
	that are live	03:39
chowmeined	aeruder: it still only lists once branch.. but i fetched one	03:39
fidodido	but they share code	03:39
	but not all the code	03:39
	soem files are different	03:39
	for eg	03:39
	names of the proj	03:39
	logs	03:39
	← yorgen1512 left	03:39
fidodido	they are different	03:39
gitster	chowmeined: you earlier said "I fetched". How?	03:39
fidodido	but the backend	03:39
Randal	anyway... I'm leaving to meet friends (yes, I still have them) :)	03:39
aeruder	fidodido: man, could you please attempt to put more than 5 words on a line, you are destroying the backlog	03:39
fidodido	keeps improving	03:39
aeruder	because it	03:39
	gets	03:39
fidodido	oops	03:39
aeruder	so annoying :)	03:40
chowmeined	gitster: git fetch location:/git/test.git	03:40
fidodido	when i clone a repo	03:40
chowmeined	and also.. git fetch location:/git/test.git somebranch	03:40
fidodido	i do git clone /work/git/code	03:40
	and now i go to /tmp/code	03:40
	it only has master branch	03:40
	hw do i get all the branches	03:40
	and exact copy of the repo	03:40
aeruder	fidodido: git branch -a	03:40
fidodido	shud i jus copy the file in unix	03:40
	ok	03:40
aeruder	they're all there	03:40
gitster	chowmeined: Either does not specify tracking branch (see git-fetch manual and read description of refspec).	03:41
	So what you fetched is only found in .git/FETCH_HEAD.	03:41
fidodido	aeruder: git branch -a	03:41
	commentwall	03:41
	edviral	03:41
	* master	03:41
	pendingreg	03:41
	server	03:41
	today	03:41
	this wat i got	03:41
gitster	chowmeined: Is this location:/git/test.git/ repository what you will be interacting with repeatedly, or just is this just one shot deal to fetch from it?	03:42
chowmeined	so how do i fetch all the stuff?	03:42
	somebody cloned my repo.. and added some branches and stuff.. and that is their published work.. now i want to get it into my repo and merge it	03:42
gitster	So you heard from them that "here is _the_ branch that has updates on top of your work"?	03:43
chowmeined	yes.. their master branch is unchanged	03:43
gitster	And merging it back is one shot (although they may later come back with more updates)?	03:43
fidodido	aerude?	03:43
chowmeined	yes	03:43
gitster	If that is the case, you probably should have done "git pull location:/git/test.git/ that_branch_they_asked_you_to_merge"	03:44
	That's the SOP for one-shot merge from downstream.	03:44
chowmeined	cant i fetch it first so i can review what they did before merging it?	03:44
gitster	You could, but you can do the above pull, then "git log ORIG_HEAD.." (or "git diff ORIG_HEAD..") to inspect it _after_ the merge. If you do not like the result, you can always "git reset --hard ORIG_HEAD" to get rid of it.	03:45
	← KirinDave left	03:45
gitster	That's also the SOP for one-shot merge from downstream.	03:45
chowmeined	SOP?	03:45
gitster	standard-operating-procedure	03:45
chowmeined	standard operating procedure?	03:45
	ok	03:45
	thats what i thought	03:45
gitster	Having said that.	03:45
	You could "git fetch somewhere:somerepo.git somebranch" and then	03:46
chowmeined	but how would i fetch a remote branch so it becomes a branch in mine?	03:46
gitster	"git log ..FETCH_HEAD"	03:46
	wait a bit, I'll answer your earlier question first.	03:46
chowmeined	ok	03:46
gitster	If you did not do "git pull" but "git fetch", then you will have what you fetched in .git/FETCH_HEAD (file) and you can refer to it as FETCH_HEAD, like the above command. "git log ..FETCH_HEAD" means "Show me what my HEAD does not but what FETCH_HEAD has".	03:47
chowmeined	oh.. so i could git branch test .. then git merge FETCH_HEAD?	03:48
gitster	I did not say "could git branch test" part; I do not think you can.	03:48
	(continuing what I was saying) If you are satisfied with the output of "git log ..FETCH_HEAD", you should be able to say "git merge FETCH_HEAD" ("git pull" is "git fetch" + "git merge").	03:49
chowmeined	it just creates a new branch.. then i merge the updates with that branch	03:49
gitster	"creates a new branch" from where?	03:49
chowmeined	it branches from my master.. then i merge their updates with this new branch.. now i have their branch as a branch	03:50
gitster	So that's one way. Now, if you are _not_ doing a one-shot merge from downstream, iow, if the other guy is somebody you will be merging from repeatedly in the future, it is cumbersome to always type URL and branch name.	03:50
	It also is easier to keep track of what you fetched/merged from him last time.	03:50
	That's where remote tracking branches come to play.	03:50
	You can "git remote add somewhere:somerepo.git somename" once, and then	03:51
chowmeined	oh yea, thats what i was looking for	03:51
gitster	after setting that up once, you can: "git pull somename thatbranch"	03:51
chowmeined	excellent, thanks	03:52
gitster	where "somewhere:somerepo.git" is what the other side calls that repository, "somename" is just a name you privately use to refer to that (shorthand).	03:52
chowmeined	how would I handle something like.. they pulled from my repo, then i made some branches and added stuff.. and they made some branches and added stuff... do they just pull again? last time we tried that a bunch of errors came up	03:53
	← aroben left	03:53
gitster	If you are acting as the upstream to them, after you merge, I'd have them fetch from you and rebase what they did on top of your latest if I were you.	03:54
chowmeined	"fatal: needed a single revision".. and "The branch 'origin' is not a strict subset of your current HEAD"	03:55
	← federico1 left	03:55
gitster	what version of git are you using?	03:55
chowmeined	im using: git version 1.5.2.1	03:55
	they are using whatever comes with slackware.. i dont know...	03:55
gitster	was the repository cloned with ancient version of git (pre 1.5.0)?	03:56
chowmeined	possibly	03:56
gitster	Actually I know it is.	03:56
chowmeined	is that going to be an issue?	03:56
gitster	Let's fix it for you.	03:56
	"The branch 'origin' is not a strict subset". Who gets that message?	03:56
	You, or the other party?	03:57
	← raeLLL left	03:57
chowmeined	they do.. when they try to git pull to get updates	03:57
gitster	Two possibilities. (1) you rebased your 'master' after they fetched; (2) they committed on 'origin' after they fetched.	03:58
	Either is a grave mistake.	03:58
	Pre 1.5.0 git used 'origin' branch (local branch namespace, i.e. .git/refs/heads/origin) to keep track of 'master' obtained the other end.	03:59
	If (now I am talking about the latter possibility) they clone from you, your 'master' will be copied to their 'origin' in that layout.	03:59
chowmeined	i keep my public and development repo separate	04:00
gitster	If they make commit on top of 'origin' (I am not talking about making a different branch from 'origin' and start working on that different branch --- I am talking about committing while 'origin' branch is checked out), then	04:00
	next time when they try to fetch from you, you have something on top of your previous 'master', and fetch will try to use it to overwrite their 'origin', but it is not their origin's descendant anymore, iow, fetch will _lose_ their changes.	04:01
	That's what the error is about.	04:01
	← aroben_ left	04:01
chowmeined	they get that error when they try to merge origin into master	04:02
gitster	Modern git avoids this mistake by using separate namespace to keep track of remote branches' progress.	04:02
	"they try to merge origin into master" by doing what?	04:02
	"git checkout master && git merge origin" ?	04:03
chowmeined	yes	04:03
gitster	I do not think you can get "The branch 'origin' is not a strict subset" message from anywhere in that command sequence. The only place I know is possible is during "git fetch".	04:04
	← sbohrer left	04:04
chowmeined	hmm, well i wasnt there when they were doing all this.. so they might have been running other commands	04:04
gitster	"fatal: needed a single revision" is possible if they do not _have_ origin. Can they say "git show origin" and see a commit like you see when you type "git show master"?	04:04
chowmeined	i dont know what happened.. what we ended up doing was moving their files out.. and then just deleting it and cloning again.. then re-adding and committing their changes	04:06
Dragon45	is anyone logging these IRC chats and posting them somewhere online	04:06
	there's some useful info here :X	04:06
gitster	see /topic	04:06
	→ orospakr joined	04:06
Dragon45	:P	04:07
	ty	04:07
chowmeined	yea.. gitster is like a fountain of knowledge	04:07
Dragon45	i must wire this channel into my veins	04:07
	btw i started using git today to manage my class notes	04:08
	since classes started today :X	04:08
chowmeined	ive been using git for managing source code.. but i also want to use it to manage config files under /etc on servers	04:09
Dragon45	is there a way to get git to "parse" open office files properly like it does with regular text files?	04:10
gitster	managing /etc/ is a FAQ topic. I think they have an entry in git FAQ (I am not involved in maintaining it).	04:10
Dragon45	i saw something about an open office driver for it..	04:10
	for gitk anyway	04:10
gitster	What do you want to see happen with OOo files? What do you want out of git "parsing" OOo files?	04:10
Dragon45	line tracking across files	04:11
	git sees .odt as binary blobs right now	04:11
gitster	"git diff HEAD^ -- that.odp"?	04:11
Dragon45	but there's relevant textual information in them	04:11
	.odt is the suffix for the open office file format	04:11
gitster	Is something like the above command line what you want?	04:12
Dragon45	(at least the word processor, i havent used the others yet)	04:12
	oh	04:12
danieldg	iirc .odt is technically a .zip file with openoffice XML inside it	04:12
Dragon45	really?	04:12
	hm :X	04:12
gitster	There was a proof of concept patch posted on the mailing list recently.	04:12
danieldg	rename .odt to .zip and open it up, kinda interesting	04:12
gitster	Well, I should not say a "patch". You can do it without patching git at all.	04:12
	→ yorgen1512 joined	04:13
danieldg	so you'd have to get something that unzips and does diffs on the unzipped text	04:13
gitster	Yes, write a bit of script and point GIT_EXTERNAL_DIFF at it, and you are done.	04:13
	→ Insount joined	04:13
gitster	Search the list archive for articles during the past three weeks or so for the above environment variable.	04:13
Dragon45	cool - is this functionality going to be integrated into the main git branch?	04:15
gitster	Didn't I say you donot have to patch git at all?	04:15
	Even git v1.0 should be able to handle it.	04:15
Dragon45	k	04:15
chowmeined	hmm	04:15
Dragon45	still seems like it could be included by default though	04:16
chowmeined	well you could write your notes in text files	04:16
	and do like docbook or latex for papers	04:16
Dragon45	or maybe as part of a "useful git scripts" package or something ponders	04:16
	well, the thing is	04:16
Randal	odt files are essentially binaries	04:16
	do you want to do image-diff too?	04:16
Dragon45	there's always the risk of random diagrams that are important	04:16
gitster	There is an OOo to text filter, and GIT_EXTERNAL_DIFF will get two OOo files in temporary files; you run the OOo-to-text filter on both and run "diff -u" yourself.	04:17
	Randal rolls eyes at what people expect a source* manager to do*	04:17
Dragon45	how's this for a monkey wrench, randall	04:17
	source of what?	04:17
	im involved tangentially with an open source game engine project	04:17
gitster	source of confusion, apparently ;-).	04:17
danieldg	well, you could convince people to move to TeX, those files are nicely managable by SCMs :D	04:17
Dragon45	they're looking (still) for a robust repo system capable of dealing with the "source" of very large files	04:17
Randal	or POD :)	04:17
Dragon45	images, maps, etc	04:17
Randal	well - there won't be sensible "merges" of those... just allocating which version belongs in the final when there's a conflict	04:18
Dragon45	mhm	04:18
eMBee	good afternoon	04:19
Randal	for some meaning of afternoon. :)	04:19
Dragon45	midnight here ;)	04:20
eMBee	4pm here	04:20
	eMBee imported his companies svn repository into git last night	04:20
chowmeined	it doesnt have to be just source	04:20
gitster	http://marc.info/?l=git&m=118820836315602&w=2	04:21
chowmeined	id hope that it could handle pretty much any text files	04:21
eMBee	there is a gazillion of revisions, but just a few branches.	04:21
chowmeined	so isnt it the right thing to do is make lots of branches in git?	04:22
	like if i want to work on a new component.. make a branch for it until it can be merged and the result can be compiled	04:22
	eMBee is trying to tell gitk to reduce the revisions it shows to something managable	04:22
gitster	chowmeined: yes and no.	04:23
chowmeined	oh ok.. thanks	04:24
Dragon45	i was wondering who this enigmatic and knowledgable "gitser" fellow is - and then a simple /whois tells me all i need to know X-D	04:24
	← jcollie left	04:25
gitster	git may be able to handle many branches, but humans will have hard time keeping track of which branch is still relevant. So (1) make liberal use of topic branches (encouraged), (2) once your topic is in reasonable enough shape, merge it to your integration branch (or rebase onto the integration branch and then merge it into integration branch -- which will result in a fast-forward),	04:25
	and then (3) get rid of that topic once you merged it in the integration branch, unless there is a compelling reason to keep it around. (What I often do is "git branch -m a_topic done/a_topic" and forget about it)	04:26
	← rtmfd_icbm left	04:26
	← rkaway left	04:26
chowmeined	oh ok	04:26
	← jackbravo left	04:27
	→ rkaway joined	04:28
eMBee	how i can i select revisions only by date, no matter which branch they are in?	04:28
gitster	chowmeined: If you later find some problem with that topic, you can reopen that branch, either by resurrecting from done/a_topic, or making a new branch (if you deleted the topic after merging it into the integration branch, which I think is easier to manage) from the tip of the integration branch when the problem was discovered.	04:29
	eMBee: something like "git log --all --since=20.days.ago"?	04:29
eMBee	yeah, but i don't want since, but only that day or month	04:30
	and i have multiple days that i want	04:30
gitster	git commit set is a range operation. You cannot have 5 and 7 and 13 days ago without getting other dates in between.	04:31
eMBee	hmm	04:31
	gitster goes back to real work...	04:31
eMBee	:-)	04:31
	that's exactly what i want though	04:32
	Randal waits for imaginary work to come to him	04:32
	Dragon45 thinks Randal should get a job if he wants work	04:32
chowmeined	wait so like	04:32
	gister wrote bunches of git?	04:32
eMBee	to many commits, i only want to look at all the branch points (which ara about half a dozend)\	04:32
Randal	so like, yeah	04:32
	he's only like, you know, the chief maintainer. :)	04:33
Dragon45	yeh	04:33
	Linus -> gitster	04:33
Randal	you can like, hear him at the mp3 at, like http://www.twit.tv/floss19	04:33
Dragon45	that's pretty much the hierarchy	04:33
Randal	no. not linus	04:33
	Oh	04:33
	ok	04:33
	not mapping, but hierarchy	04:33
Dragon45	indeed	04:33
chowmeined	splendid.. git is really excellent	04:33
Dragon45	indeed it is :X	04:33
	Randal is trying to figure out why flickr uploads are being so slow.	04:34
gitster	do you have some interesting picture other than you drinking beer?	04:35
Randal	indeed.	04:35
	there are 4500 pictures there so far, and only 60 of them are of me. :)	04:35
	there are another 30,000 pictures on my other site, which I will select about 20% of to bring over to flickr	04:36
Dragon45	where is your site?	04:36
Randal	I just need... "more time"	04:36
	flickr.com/photos/randal-schwartz/	04:36
	although that will show the temporary pictures until they get moved into the right place.	04:36
Dragon45	http://en.wikipedia.org/wiki/Randal_L._Schwartz	04:37
	O_o	04:37
Randal	and?	04:37
Dragon45	same one?	04:37
Randal	tis me	04:37
Dragon45	nothing just wondering	04:38
	ah k =]	04:38
	Insount → Insount_away	04:55
chowmeined	wow	04:58
	famous people	04:58
Dragon45	um yes	05:00
	^_^	05:00
	not me though :(	05:00
	maybe one day ;)	05:00
	sleepztime	05:01
	gnight all =]	05:01
	← Dragon45 left	05:03
	→ aroben joined	05:03
	← tongueroo left	05:05
	← hap left	05:10
	→ hap joined	05:12
	← KiBi left	05:16
	→ masonj_ joined	05:18
	→ johan-s joined	05:23
EvanCarroll	Randal: why did you /nick Randal and not merlyn?	05:29
	you used to be merlyn	05:29
Randal	I'm still merlyn. Just not here. :)	05:29
	I'm doing this because most people aren't smart enough to type /whois, apparently	05:29
	gitster says /whois merlyn	05:30
EvanCarroll	Well I'm going to tell people you only let your special friends in the DND fetishist group call you "merlyn" now ;)	05:30
	OK merlyn?	05:30
Randal	what ever. {sigh}	05:30
	← johan-s left	05:33
	→ bitbang joined	05:33
	→ devogon joined	05:38
	→ HackyKid joined	05:44
	← aroben left	05:53
	→ aroben joined	05:54
	→ pdmef joined	06:00
	→ KirinDave joined	06:06
	← orospakr left	06:11
chowmeined	DND?	06:13
gitster	dungeons and dragons?	06:14
	→ bdowning_ joined	06:16
	← bdowning left	06:16
	← KirinDave left	06:21
	→ Gin-geR joined	06:22
	← bitbang left	06:23
	← polyonymous_ left	06:23
	→ EnryAngr joined	06:28
EnryAngr	Hello. Can anyone tell me how to copy files from one branch to another? Merge often fails and I just want the latest file from another branch.	06:31
	→ dkagedal joined	06:31
	← ricky left	06:31
	→ GyrosGeier joined	06:32
EvanCarroll	uh check it [check]out	06:38
	← masonj_ left	06:39
	→ masonj_ joined	06:39
EnryAngr	EvanC: Please be more specific	06:39
EvanCarroll	Sure, something like `git checkout HEAD awesome.php` where HEAD is the branch or tag. or tree-ishness	06:40
	or crap.pl, if you're a newblar. ;)	06:40
EnryAngr	That's what I needed, thanks	06:42
	→ lol_h joined	06:45
	→ MrPloddy29` joined	06:48
	→ Oejet joined	06:48
	→ elmex joined	06:51
	← MrPloddy29` left	06:51
	→ MrPloddy29` joined	06:53
EnryAngr	Ok, another question. How do I get a list of filenames which were changed for a commit? 'git log' shows commit messages and 'git log -p' show the contents which changed. I want something in between - just the filenames.	06:53
phx	git show ? or what was it	06:54
	git show --cached	06:54
	something similar iirc	06:54
EnryAngr	That looks about the same as git-log	06:55
gitster	"git show --stat", "git show --name-status", "git show --numstat", "git show --name-only", ... name your poison.	06:55
	← doublec left	06:56
	→ ricky joined	06:56
EnryAngr	gitster: Cool, thanks	06:58
	← MrPloddy29 left	07:01
thiago_home	EnryAngr: git what-changed	07:10
	→ pygi joined	07:11
thiago_home	erm, git whatchanged	07:11
	← pygi left	07:12
EnryAngr	Hmm.. so many options...	07:13
	← DrNick left	07:14
thiago_home	git whatchanged = git log --name-status -r -C -M (I think)	07:14
	← fhobia left	07:14
gitster	not at all.	07:15
	← lol_h left	07:16
gitster	whatchanged outlived its usefulness, unless you are an old timer.	07:16
	log is a lot shorter-and-sweeter to type ;-)	07:16
thiago_home	not if you want to get the file listing	07:17
	it's easier to type "whatchanged" than "log --name-status -r"	07:17
	especially since I get tab-completion for the former	07:17
EnryAngr	I wish you could abbreviate. 'git co' should work for checkout.	07:18
thiago_home	EnryAngr: your problems are solved!	07:18
gitster	[alias] co = checkout	07:19
	some people do "[alias] co = commit" though...	07:19
thiago_home	http://wincent.com/knowledge-base/Git_quickstart	07:19
	commit is "ci" (check-in)	07:19
gitster	I condier "add" a check-in.	07:19
	You check-in your changes piecemeal to the index to stage, and then make a commit out of it.	07:19
thiago_home	EnryAngr: if you're using the colouring options in the quickstart, don't forget to configure your less to run with -R	07:20
gitster	Or not to configure your less at all. We give that to you by default.	07:20
	→ fhobia joined	07:21
thiago_home	not working here	07:21
	must $LESS not be set at all?	07:21
	gotta go...	07:22
	← aroben left	07:24
	← dduncan left	07:27
	← EnryAngr left	07:30
	← fhobia left	07:39
	→ zmk joined	07:46
BadQuanta	Ok.. this is confusing me. I just did something to a file.. but I don't like what I did. I haven't commited I just want to rever the file. but git revert <file> doesn't seem to do that.. what do I need to do?	07:54
	s/rever/revert	07:54
siprbaum	git checkout HEAD file	07:58
BadQuanta	oh.. thanks .. I tried that but didn't have the file part.	08:00
	still learning :/	08:02
	← felipe left	08:05
	→ keinfarbton joined	08:11
thiago	git-revert reverts a commit	08:11
	svn revert is not equivalent to git-revert	08:12
keinfarbton	is there any known commercial git+trac project-hoster?	08:12
	← dkagedal left	08:13
gitster	how much are you willing to pay?	08:17
keinfarbton	svn+trac is about less 10$ per month	08:17
thiago	gitster: thinking of opening one? :-)	08:17
gitster	why not? we've been talking about a git company forever.	08:17
	gitster goes to bed.	08:18
keinfarbton	this one does svn, http://hosted-projects.com/	08:18
	→ fhobia joined	08:20
fidodido	git clone git://git.kernel.org/pub/scm/git/git.git	08:20
	i just got a folder	08:20
	with 25Mb	08:20
	how do i get the entire history	08:20
	every1 keeps talking about	08:20
	i did this for my project	08:20
	and jus got branch	08:20
	and lost my entire history	08:20
	when i deleted the older folder where i had the git history	08:21
	it is kinda stupid	08:21
	clone dont clone!!!!	08:21
	am i missing somehting?	08:21
thiago	yes: explaining the problem.	08:21
	what exactly did you run?	08:22
fidodido	how did i get the entire repo and histroy	08:22
	by clone	08:22
thiago	by cloning	08:22
	clone = download the entire history	08:22
fidodido	it doesnt lol	08:22
thiago	unless you tell it otherwise, that is	08:22
	sure it does. What exactly did you run?	08:23
fidodido	git clone git://git.kernel.org/pub/scm/git/git.git	08:23
	du -sh	08:23
	25M .	08:23
	gitster: how to have subprojects in git	08:23
	so i can have 2 projects	08:23
	one is called	08:23
	P1 n	08:23
	thiago tries that	08:23
fidodido	and 2nd one is P2	08:23
thiago	please stop flooding	08:23
	finish your sentences before you press Enter	08:23
fidodido	so how to share code between P1 And P2 and ignore a few files for sharing between P1 and P2	08:23
	but they get tracked by git	08:24
	is it possible to do it	08:24
	?	08:24
	i was hoping to see if merge ignore file would do it	08:24
	but there is no merge ignore	08:24
	so the project is a variation of the same basic code	08:24
thiago	I've just tried that clone: it creates a 25 MB directory	08:24
fidodido	such as reddit	08:24
	it clone lipstick	08:25
	and so on	08:25
	each has the same code	08:25
thiago	would you mind stop flooding?!	08:25
fidodido	but the front end is different	08:25
	thiago: it doesnt hve the entire repo	08:25
thiago	that clone operation you tried did get the entire history. Why do you think it didn't?	08:25
fidodido	it jus does wat svn checkout	08:25
	does	08:25
	so what is the big deal? in git	08:25
thiago	no, it got the entire history.	08:25
	I have just done that here and it got the entire history	08:26
	all 11362 commits	08:26
fidodido	ok	08:26
	it has only one branch master	08:26
thiago	those are your local branches	08:26
fidodido	the source dir that i cloned from had like 13 branches	08:26
thiago	run git branch -r or git branch -a to see the rest	08:26
fidodido	git branch -a	08:27
	* master	08:27
	origin/HEAD	08:27
	origin/html	08:27
	origin/maint	08:27
	origin/man	08:27
	origin/master	08:27
	origin/next	08:27
	origin/pu	08:27
	origin/todo	08:27
	thiago	08:27
thiago	what part of "stop flooding" did you not understand?	08:27
fidodido	so how do we acccess the other branches from here?	08:27
thiago	the "stop" or the "flooding"?	08:27
fidodido	sorry	08:28
	thiago	08:28
	no more	08:28
thiago	yes, no more. I'm not talking to you anymore.	08:28
fidodido	sorry man	08:28
	ok	08:28
	how to setup repos and share code with others	08:29
	that part is really hard?	08:29
	i dont get it	08:29
	elmarco\|food → elmarco	08:33
	← wart left	08:39
	→ wart joined	08:44
	→ meandtheshell joined	08:46
	← fhobia left	08:54
	← keinfarbton left	09:04
	→ whee joined	09:13
	→ Yuuhi joined	09:14
	→ kumbayo joined	09:16
	→ dkagedal joined	09:33
	→ cortilap joined	09:35
	→ Beket joined	09:37
	→ adante_ joined	09:59
	→ ferdy joined	10:07
	Gin-geR → polyonymous	10:12
	← adante left	10:14
	adante_ → adante	10:14
	→ janm joined	10:17
	→ felipe joined	10:19
	← masonj_ left	10:24
	→ Dentharg joined	10:47
Dentharg	Hi!	10:47
	need some guidance on git usage	10:49
	← fultilt left	10:54
vmiklos	did you read the 'Everyday GIT With 20 Commands Or So' document?	10:54
pasky	gitster: Can git-gc remove any unreferenced objects?	10:55
	gitster: Again the documentation sorta implies that it can't without --prune, but I suspect that it's again wrong and if the unreferenced object is packed it'll simply go away during the repack?	10:55
Dentharg	vmiklos, yes I did	10:55
	vmiklos, I just need some advice on branch layout and working	10:56
	vmiklos, is it better to have separate working directories when doing separate features	10:56
pasky	gitster: hmm, reading it more carefully again, it's hard to say, the description is confusing	10:56
Dentharg	vmiklos, or just pull into one place and do branch switching	10:56
	vmiklos, I am keen Subversion user that wants to switch	10:56
vmiklos	Dentharg use branches	10:56
	btw then read http://git.or.cz/course/svn.html	10:57
Dentharg	vmiklos, ok. thank you!	10:57
siprbaum	not so sure about the usage of brnaches	10:57
	it is very dependent on how you work. if you normally do only one thing then branches are fine (because you switch to/from them), but if you are doing several things (say multitasking) then seperate working dirs could be beneficial. especially if it takes a long time to compile	10:59
	but i agreee with vmiklos that branches are propably the best way to go, because humans tend to suck in multi-tasking.	11:00
	(at least I do ...)	11:00
gitster	pasky, let's talk about what happens and how it does first, then what the implications are, and finally what are the desired behaviour.	11:02
	1. what happens and how.	11:03
	gc is roughly "repack -a -d without prune" as long as objects are concerned.	11:03
Dentharg	siprbaum, currently I have several trunk checkouts to work on features and I am getting lost	11:03
pasky	yes	11:03
Dentharg	I only wish that git-svn was working ok on Windows :/	11:03
gitster	Now, "repack -a -d" is "pack everything reachable" followed by "remove all preexisting packs".	11:04
	That is "how" part -- so "what happens" is "objects in packs that are unreachable from refs will be gone".	11:04
	But we do not lose loose objects.	11:05
	2. Implications.	11:05
siprbaum	Dentharg: ok. then. I missunderstood you. branches is the way to go	11:05
pasky	alternates broken in this case too, but at least no race conditions	11:05
gitster	not really, but I'll come back to that later.	11:06
pasky	hmm, there are race conditions :(	11:06
Dentharg	so branches it is.	11:06
gitster	A fetch that produces loose objects (i.e. walker and fetch that runs unpack-objects) do not race with gc.	11:06
	But a fetch that leaves pack does (1) fetch and store pack (2) index pack and then (3) update ref. So if your gc runs REALLY FAST between the time (2) finishes and (3) finishes, then gc sees the just-built pack without noticing the new refs that want the objects in that pack, which repack -d will remove.	11:07
	So in theory there still is a race.	11:07
	That's the implications part.	11:08
	→ gitte joined	11:08
gitster	I am not sure about "3. Desired behaviour".	11:08
	As pack plus index does not really have the traversal order information, it is kind of hard to reconstruct the "tips that ought to be kept if we do not want to lose any objects from this pack".	11:09
hliusv561_	gitster: Aren't those packs protected by .keep file?	11:09
	← Dentharg left	11:09
	gitte wonders if gitster was already to bed...	11:09
gitster	That's right --- I forgot about that.	11:09
pasky	:)	11:09
	gitster couldn't sleep.	11:09
pasky	:(	11:10
gitte	gitster: my consolation.	11:10
gitster	.keep is there to close the race between (2) and (3).	11:10
	→ sirNetwerk joined	11:10
	gitster wonders if he knows hliusv561_ on the list. He never can connect IRC with the list...	11:10
hliusv561_	gitster: My name is ocassionally (at least twice) mentioned on list, but I haven't posted there myself.	11:11
gitte	gitster: yes, he is that long Finnish name ;-)	11:11
	Oh?	11:11
gitster	I _think_ the desired behaviour ought to be "we do not want duplicate objects across multiple packs, and that is why we are repacking to consolidate more than one pack into one, but we do want to pack _all_ objects in existing packs".	11:12
	It is rather hard to arrange, unfortunately. At least I do not think of a good way to do that offhand.	11:12
	On the other hand, if we literally follow that, we cannot easily get rid of objects belonging to rewound commits that are no longer needed from packs. I guess we could honor --prune option there, though...	11:14
pasky	yes	11:15
	that would be ideal	11:15
sirNetwerk	Concerting the post "unaware of the importance of git gc": If I never use any packs at all and keep everything with objects, will branch be terribly slow?	11:15
gitster	We probably could re-sort existing .idx file in the offset order to emulate the traversal order the original rev-list would have given to pack-objects.	11:15
hliusv561_	gitster: I was once thinking about option to pack all unreachable objects to second packfile which could then be copied to alternates. Probably has severe problems with race conditions though. :-(	11:16
	→ doublec joined	11:16
gitster	sirNetwerk: not "branch" but literally everything will be slow.	11:16
	You will be left wondering, literally, why everybody else would say "git is fast".	11:16
sirNetwerk	I see no reason why branch would be slow. But continue your current -more interesting- discussion.	11:17
	gitte did not yet start reading that monster thread.	11:17
gitster	If you've never repacked and have been working on a project of size of the kernel, you haven't seen git ;-)	11:17
	I suspect Linus was talking about "branch switching", not just "creating a branch".	11:18
sirNetwerk	Ahh right. For me the solution (thinking how big the repo will be in 10 years!), is to use "git clone --depth". Who wants to bisect 1000000 commits back?	11:18
	And keep everything as plain objects.	11:19
	→ pygi joined	11:19
gitster	hliusv561_: I think the issue is how you would create the list of those "unreachable" objects to pack in the secondary one.	11:19
doublec	how does git handle binary files when merging?	11:19
siprbaum	doublec: you have to decide which one to take	11:20
doublec	so you get a merge conflict then make a decision?	11:20
siprbaum	yes	11:20
gitster	Well, the initial prediction (I think it was after 2-3 weeks of use) was that in 2 years the kernel repository would be 3GB and that growth rate is acceptable. Shows that Linus is not perfect in predicting the future ;-).	11:20
doublec	how does it know what is a binary?	11:20
hliusv561_	gitster: Generate list of all loose objects and objects in non-kept packs and do set complement. Obiviously that option would only make sense only with -a.	11:20
doublec	This blog post mentions a problem so though I'd check: http://www.dekorte.com/blog/blog.cgi?do=item&id=2895	11:21
gitster	doublec: traditionally, it "guesses" exactly the same way as GNU diff does. Recent git you can tell git "this path is binary, do not touch".	11:21
doublec	great, thanks!	11:21
	what is the way to tell git that btw?	11:22
gitster	"man 5 gitattributes"	11:22
doublec	thanks gitster	11:22
madduck	so in migrating a package from debian dpatch to git feature branches, i am faced with the task to convert a patch-in-a-file to a branch. this means that I really want to remove the file from both branches after merging the patch into the second branch. this, however, requires two commits	11:23
	the file removal in the original branch	11:24
	and the patch application and file removal in the feature branch	11:24
	i want to logically tie these in with each other, but I guess that this is not possible, huh?	11:24
	← sirNetwerk left	11:26
siprbaum	madduck: why are there 2 branches involved?	11:26
	→ Lash\| joined	11:26
madduck	fair question; simply because it's a change from upstream i need to track separately	11:26
	the idea is that i have an upstream branch and a bunch of feature branches	11:26
	and an integration branch where i produce the final package	11:26
	some feature branches are intended to go upstream, some are debian specific	11:27
	so i'd like to keep them around	11:27
	but then on the other hand, i don't know how to really work with those.	11:27
	there are two models i can follow:	11:27
gitster	Currently you have a patchfile that contains both?	11:27
madduck	both?	11:27
gitster	both "for upstream" and "kept within debian"	11:28
madduck	no, i have separate patch files in both groups	11:28
	it's really a question about comverting patch files to branches	11:28
	orig branch has patch files, and i want to replace each of those files with a branch off orig	11:28
	the problem is that it seems impossible to do this with one commit	11:29
	← meandtheshell left	11:29
madduck	i have to commit rm patchfile to orig	11:29
gitster	you would start from upstream, then have a branch that forks from it that holds "for upstream" fixes.	11:29
madduck	and commit git-apply patchfile && rm patchfile to the feature branch	11:29
gitster	another fork, again from upstream, would hold "debianization".	11:29
	No patches, just maintain branches.	11:29
madduck	yes, i want to get rid of patchfiles	11:30
	which means i have to remove them from the repo after applying them to individual, new branches	11:30
gitster	When it comes to finally "package debianization", you can take diff between branches involved and produce the *.diff file, no?	11:30
madduck	there are two issues at hand and i think we're talking about different ones. :)	11:30
	but i think mine's a question that i've already answered, so let me respond to what you said...	11:30
	there are two models to follow when packaging with vcs: (a) one branch for each distro release and (b) one branch for each package version	11:31
	well, there is (c): one branch for everything, but...	11:31
	so for debian, i'd track upstream and branch off for sarge/etch/lenny	11:31
	i also maintain separate feature branches, all branched off upstream at one point.	11:32
	those feature branches are long-living branches, meaning they never get merged back into upstream	11:32
	wait, lemme draw this...	11:32
	→ Hacky joined	11:33
	gitster falls off the chair, dozing...	11:34
	madduck needs a proper tool for quick sketches. gimp and inkscape are just painful.	11:34
	madduck takes pen and paper and scans	11:35
	→ Sho_ joined	11:37
	gitster finally crashes...	11:37
	← jody left	11:37
madduck	:)	11:38
	almost done	11:38
gitte	Good night gitster	11:39
madduck	http://scratch.madduck.net/__tmp__branches.pdf	11:39
	so F1/F2 are debian-specific branches i merge into each release branch	11:40
	F3 is a feature branch/bug fix that upstream accepted between sarge and etch	11:40
	and F4 is a feature branch that i spawned out of the sarge branch and now merge into etch and lenny	11:41
gitte	madduck: you need a tablet ;-)	11:41
madduck	gitte: i really do.	11:41
	madduck writes it down on the "things to consider" list	11:41
madduck	that's one model.	11:41
siprbaum	madduck: ok, but whats the actual question on this organisation?	11:42
madduck	siprbaum: well, there is another model and i basically would love to get some input	11:42
hliusv561_	madduck: How git.git does it is to merge feature branches 'stablest first' and then merge these branches upwards towards more unstable branches.	11:42
madduck	the other model has the same feature branches and one debian branch and when sarge is released, then branch off for security fixes	11:43
	siprbaum: my original question was a different one, but this organisation has been in my head for days and i keep trying it and not knowing whether i am doing the right thing	11:44
	since it's hard to simulate in ways.	11:44
	hliusv561_: i am not sure i understand what you mean.	11:44
siprbaum	Ah. Now i think I understand	11:45
	madduck: in your drawing you always have to recreate the 'debian' folder from scratch, because you build your sarge, etch, sid whatever branch on top of upstream. Am I correct?	11:46
madduck	well, sort of. assume F1 is the "debiandir" branch, which adds it to each distro branch	11:46
	← Beket left	11:47
	← HackyKid left	11:49
siprbaum	and the other way to do it is ... ?	11:50
	← zmk left	11:50
	→ Beket_ joined	11:50
madduck	to branch off upstream once, branch debian feature branches of this debian branch and upustream feature branches off upstream (but merge them into debian)	11:51
	and when e.g. lenny is released, the branch a lenny-stable branch off the debian branch	11:51
Beket_	Hello people. Could anyone hint me regarding this ? http://rafb.net/p/f2EC2W97.html Thank you in advance	11:51
hliusv561_	madduck: If some feature branch goes to maint (most stable), then maint is merged to master, which is merged to next (least stable). This would go on to pu, but pu is rebased instead.	11:51
madduck	but keep developing in the debian branch (which is 'master', really, i guess)	11:51
	madduck reads http://git.or.cz/gitwiki/Notes	11:52
hliusv561_	Beket_: Don't try to merge with uncommitted changes in working tree...	11:53
madduck	Beket_: you need to commit local changes before you merge	11:53
	damn, hliusv561_ was faster. :_)	11:53
Beket_	In PC1: I commited those changes localy and the pushed them to REMOTE. From PC2: I want to pull those changes	11:53
hliusv561_	Beket_: Commit first on PC2.	11:54
	Beket_: Or if the changes are not ready, stash them.	11:54
Beket_	oh thanks people :)	11:54
	Yes, it worked now.. thanks again!	11:55
	← pygi left	11:55
madduck	hliusv561_: okay, so my second model seems to be pretty exactly what git does, master is the debian branch, maint is the *-stable branch.	11:56
hliusv561_	madduck: Instead of arrows from feature branches to all three release branches, I think there should be be arrows from sarge to etch and ecth to lenny...	11:56
madduck	i am a little unclear as to maint though, since it says that maint is branched off master at each featrue release	11:57
	i did not know you can branch a branch multiple times off master.	11:57
siprbaum	its just a merge of maint and master	11:57
	and they diverge again so the branched again from master :-)	11:57
madduck	siprbaum: so when 1.5.4 will be released, maint will be for 1.5.4 and there won't be a maint branch for 1.5.3?	11:58
	so if there's a bug in 1.5.3 after 1.5.4 is out, it won't get fixed?	11:58
hliusv561_	madduck: Besides requirement that history graph is DAG (so commits can be topologically sorted), the topology of repository history is unrestricted.	11:58
madduck	hliusv561_: well, yes. i am trying to decide whether the git model or the one in my drawing are more suitable for debian^W distro packaging.	11:59
	and since we have to support up to 6 different versions of each package, i think the model i drew makes most sense	11:59
	but then again, there are little functional differences.	11:59
hliusv561_	madduck: Save the earlier branch heads and treat earlier ones as more stable.	11:59
siprbaum	madduck: in your case you need a maint-sarge, maint-etch, maint-... branch. I think in git the maint branch is just for the actuall stable release	11:59
madduck	siprbaum: yes.	12:00
	hliusv561_: save them, as in maint-sarge, maint-etch branches?	12:00
	← jlh left	12:00
	→ jlh joined	12:00
hliusv561_	madduck: You pick the names. :-)	12:00
madduck	weeeh! :)	12:01
	so current git master is what will be in 1.5.4 and current git next is what might be in 1.5.5?	12:01
	as in stuff that's too much for 1.5.4 at this point, so it's kept for a future release?	12:01
hliusv561_	madduck: Nope. Next is more unstable changes that are not "proven".	12:02
madduck	'experimental' in debian parlance, then	12:02
hliusv561_	madduck: Or actually, changes that are still unstable, but seem likely not to have serious fundamental issues.	12:02
	madduck: pu contains changes even more unstable than that.	12:03
madduck	pu would be my scratch branch	12:05
	thanks guys, it'	12:07
	s been a good and most helpful discussion	12:07
	i think i'll just have to try it out now.	12:07
	and i guess the git model makes most sense, really.	12:08
hliusv561_	madduck: And if I noticed I made a mistake and merged some feature branch to more unstable branch than what was needed, I would merge it to proper branch, merge it up the chain, doing merges where it is already present using 'ours'.	12:10
	→ rodserling joined	12:16
	→ madewokherd joined	12:23
	→ jcollie joined	12:40
	→ lcapitulino joined	12:43
	→ WackyKid joined	12:46
madduck	i have a presentation template which i would like to store in a public repository. i also have a repository for e.g. my phd work. when i have to give a new talk, i would like to clone the presentation template into the phd repo. how would i do that? is it at all possible?	12:46
	← doublec left	12:47
	← Lash\| left	12:50
hliusv561_	madduck: fetch the branch containing presentation template into private repo.	12:51
madduck	hliusv561_: sure, i can fetch it alright, but what i want is to be able to use the branch as a template.	12:57
hliusv561_	madduck: Read it into some suitable subdirectory, branch it off and keep as independent branch or submodule?	13:00
madduck	sure. but how do i then clone it into a repo?	13:01
	let's make an example:	13:01
	git documentation is maintained in asciidoc	13:02
	there may be a template with basic stuff every document in git/Documentation should have	13:02
	← Hacky left	13:02
madduck	and because it's the same as in Linux (assume for now), it's a separate repo: repo.or.cz/docu.git	13:03
	now let's say I want to add git/Documentation/foo.txt:	13:03
hliusv561_	madduck: Sounds job for subprojects...	13:03
madduck	i know i can add a remote for repo.or.cz/docu.git or use a submodule	13:03
	in git/Documentation/templates and then I'd have git/Documentation/templates/template.txt	13:04
	but to create git/Documentation/foo.txt, I'd have to just copy the file	13:04
	and there would be no notion in the history that the file's ancestor is templates/template.txt	13:05
	am i making sense?	13:05
vmiklos	sure. the history is not shared between the two repos	13:06
	→ d0k joined	13:17
	→ cmarcelo joined	13:19
	→ weazzle joined	13:21
slaq	hrm	13:22
	anyone know if there's anything like trac or redmine for git?	13:23
	← rodserling left	13:24
gitte	slaq: not that I know of. IIRC someone wanted to look into porting Trac to use Git instead of Subversion, but I never heard of any progress there.	13:26
pickscrape	I've seen mention of a couple of git plugins for trac	13:27
	Just google for git trac plugin and see what you find.	13:28
slaq	i found a few patches and "experimental" stuff	13:30
	but nothing to warrant a testdrive	13:30
	i also found an interesting discussion on the git ml about a distributed bugtracke	13:31
	too bad this all doesnt seem to be ready yet.. i'm torn, want to stick with git but some of the svn tools are really useful :\	13:31
	→ bdiego joined	13:32
gitte	slaq: so you have to options: work on some solution with Git, or use Subversion instead.	13:34
	Pick one.	13:34
	← rlb3 left	13:38
vmiklos	hmm	13:40
	i have a file, say foo.c. there are 3 versions: #1 is HEAD, #2 is in index, #3 is in the working directory. is there any way to set the index version to HEAD without modifying the working dir?	13:41
	hap is impressed how fast the git commit goes	13:41
vmiklos	imho git reset HEAD foo.c will do what i want, but it will touch the working dir, too	13:42
gitte	vmiklos: AFAIK "git reset HEAD foo.c" (which is implicitly --mixed) will not touch the working dir.	13:43
	hap: hehe	13:44
vmiklos	and --soft is what will?	13:44
	ah, no	13:44
gitte	git reset with paths is not possible with --soft.	13:44
	vmiklos goes and reads the man page again	13:45
vmiklos	yes:)	13:45
gitte	There has been talk about adding a "--cached" option to git-checkout, though.	13:45
vmiklos	which would update the working dir from index or would update the index from HEAD?	13:45
	→ nikodemus` joined	13:45
hap	gitte: wait, i didnt try mercurial yet ;)	13:48
	← Beket_ left	13:55
	→ KirinDave joined	13:56
gitte	hap: hehe.	14:04
	← KirinDave left	14:07
	→ jackbravo joined	14:09
	→ aroben joined	14:11
	← lkjasa left	14:21
	← aroben left	14:25
	→ stouset joined	14:26
	→ KirinDave joined	14:28
madduck	http://git.madduck.net/v?p=tpl/s5/slides.git;a=summary	14:32
	the shortlog field is empty, probably because there is no default HEAD	14:32
	but how can i make it appear e.g. with the tip of the en branch?	14:33
	aaaah	14:33
	← pdmef left	14:33
madduck	HEAD was actually pointing at a nonexistent ref	14:33
	gosh i love how git is so filebased	14:33
	← gitte left	14:34
GyrosGeier	I argued against sections in the config file and suggested we use different files for that, but nooo, they wouldn't listen... :-(	14:34
hliusv561_	GyrosGeier: Very few filesystems support small files properly...	14:35
GyrosGeier	indeed	14:35
	→ myrizio joined	14:35
	← weazzle left	14:37
madduck	here is an idea about maintaining home in git across separate machines. of course i can just pull in repositories on each machine as i need them, but i somehow like the thought of just setting up a machine and having everything automatically go into its proper place.	14:38
	so i was thinking: make a directory hierarchy in ~/.gits and put the git config files in there	14:38
	then check in .gits into the machine repo	14:39
	now you can branch off that repo and invoke a shell script	14:39
	which then iterates ~/.gits and finds config files; e.g. for ~/.gits/.etc/mutt/config, it then git-inits ~/.etc/mutt	14:39
	and hardlinks (or symlinks) the config file to ~/.etc/mutt/.git/config	14:40
	the does a git pull and move on.	14:40
	the config files store everything that's important, such as origin location and identity to use for this repository.	14:40
	dwmw2_gone → dwmw2	14:41
	→ mankind_tweezer joined	14:43
mankind_tweezer	Is it possible to use the "filter" attribute to add something like a time-stamp to a checked-out file?	14:46
	→ krh joined	14:46
hliusv561_	mankind_tweezer: You could add timestamp of checkout, but not much else.	14:47
	mankind_tweezer: Since revision checked or the pathname of file are not passed into filter.	14:47
mankind_tweezer	hliusv561_: hmm, I see	14:47
	does it operate before or after "ident"?	14:48
hliusv561_	mankind_tweezer: IIRC, gitattributes(5) documents that... But I don't offhand remember.	14:49
mankind_tweezer	ah, it's after...	14:49
	so it can get the id of the blob	14:49
	but what can i do with that -- is there some way to take that id and find out the corresponding commit(s)?	14:50
hliusv561_	mankind_tweezer: No.	14:51
mankind_tweezer	d'oh	14:51
	i guess this is one of those things i'm not supposed to try to do	14:51
hliusv561_	mankind_tweezer: Exactly. You are supposed to use the repository data if you have it and add the data only on export.	14:52
mankind_tweezer	hmm. the use-case here is that I am using git to manage a bunch of LaTeX documents. when i print one, i'd like to have a stamp on it that tells me useful stuff like which branch it's from	14:53
hliusv561_	mankind_tweezer: There is patch that adds full keyword support to git-archive.	14:53
	mankind_tweezer: Branch commit is on is not well-defined in git.	14:53
mankind_tweezer	right, so i would settle for knowing which branch _i_ am on, i guess. maybe i should try to get LaTeX to call git directly	14:54
	→ orospakr joined	14:54
hliusv561_	mankind_tweezer: There's shell escapes in LaTeX?	14:54
mankind_tweezer	i believe it's possible to do that	14:55
	although it's not enabled by default...	14:55
	(for obvious reasons)	14:55
hliusv561_	mankind_tweezer: Principle of Maximal Authority...	14:55
mankind_tweezer	hmm, what's that?	14:56
	→ aroben joined	14:57
hliusv561_	mankind_tweezer: The general tendency of systems to assign truly excessive authority to executing code, instead of the apporiate authority it needs...	14:57
mankind_tweezer	ahh, i see. yes, indeed, that's the problem here	14:58
	it seems that if you run latex with "latex --shell-escape" then it's on... so i can have that call git, i guess.	14:58
	then the only problem is that sometimes i want to send these documents to other people, who aren't using git	14:59
hliusv561_	mankind_tweezer: For example, why does PNG image decoder have autority to directly write to filesystem?	14:59
	mankind_tweezer: git-archive extended keyword support should soon be added. There's patch for that already (altough it has performance problems).	15:00
mankind_tweezer	okay, maybe that's the way to go.	15:00
	thanks, this was extremely helpful	15:00
	it sounds like if i want to do this in a uniform way i should perhaps use git-archive even when i myself am printing a document, somehow	15:01
hliusv561_	mankind_tweezer: Or stuff makefile there that could print it...	15:02
	→ raw-bin joined	15:03
	→ QuercusMax joined	15:07
	← QuercusMax left	15:08
	→ QuercusMax joined	15:10
	← KirinDave left	15:17
	← aroben left	15:22
	→ aroben joined	15:26
	→ pygi joined	15:28
	← tchan left	15:31
	← aroben left	15:32
tsuna	Say I have a branch A in SVN. In git, I create a new branch B which I publish in SVN with git-svn dcommit. Later, some commits have been made in SVN on both A and B. I fetch the new changes. Now, can I easily know whether a commit in A has been merged in B?	15:37
	→ tchan joined	15:38
	→ KiBi joined	15:41
tokkee	tsuna: Have a look at git-cherry - is that what you want?	15:43
	→ p4tux joined	15:43
madduck	tsuna: you cannot.svn does not do merge tracking yet	15:43
	→ KirinDave joined	15:44
madduck	is there any way to be able to use $HOME or ~ in git-config? My homedirs on three machines are in different places and I'd like to keep some settings synchronised	15:47
	or relative paths?	15:47
	→ rgk joined	15:48
	← nashdj left	15:54
madduck	oh, .. is . since the paths are relative to $GIT_DIR	16:03
	so ../.. is .. :)	16:03
KiBi	heh :)	16:03
	→ vegard joined	16:07
	← jackbravo left	16:08
	→ jackbravo joined	16:08
	→ bentob0x joined	16:08
vegard	hello. I am having trouble compiling git with the sun C compiler. can somebody help me with this?	16:09
	→ robin joined	16:13
	← amfwrk left	16:13
madduck	so yeah, i admit i am trying crazy stuff, but what is this all about:	16:15
	fatal: unable to add ul/.git/config to index	16:15
	← mankind_tweezer left	16:15
madduck	yes, i am trying to add the config file of a lower $GIT_DIR to the current one	16:15
	→ amfwrk joined	16:16
	→ aroben joined	16:18
	← QuercusMax left	16:19
	← KirinDave left	16:19
	dwmw2 → dwmw2_gone	16:19
	→ QuercusMax joined	16:24
robin	QuercusMax: hi	16:27
QuercusMax	hey	16:27
robin	I've update my pu branch with fixes to the history view	16:28
QuercusMax	excellent, i'll take a look	16:28
	if I get a chance this evening, i'm going to see if I can finish up my ReadTree implementation, and probably merge the common parts with WorkdirCheckout	16:29
	I think I've got all the D/F conflict handling taken care of, which was actually surprisingly straightforward	16:29
	→ gnilor joined	16:32
	→ rtmfd_icbm joined	16:33
gnilor	if I have a git kernel tree (linville/wireless-dev.git) and i want to swich to another tree (davem/net-2.6.24) that is possible without cloning it all over again? how?	16:34
	i've been issueing varients of git pull commands, but always run into som "need update" or other error messages/warnings	16:34
QuercusMax	gnilor: you probably want fetch instead of pull	16:35
	then git-checkout -b mybranch davem/net-2.6.24	16:35
	that will create a new branch cloned from davem/net-2.6.24, and switch to it	16:36
	pull is the same as fetch + merge	16:36
gnilor	tnx, the fetch worked, now just the branch switching thing	16:37
	← pygi left	16:38
QuercusMax	gnilor: did you set up a remote? that's probably the simplest way to do it	16:39
	← kumbayo left	16:39
QuercusMax	gnilor: git-remote add somebody git://domain/path.git	16:39
	gnilor: git fetch somebody	16:39
	gnilor: that will set up a remote entry, so whenever you want to fetch, you can just use that remotename as an alias, instead of typing the whole url	16:40
gnilor	now when i do "git branch -r" i see a somebody/master , but checking out gives a warning/error saying some file would be overwritten on merge	16:42
	→ KirinDave joined	16:43
	← lcapitulino left	16:44
	→ workkevin joined	16:44
gnilor	aha -f did it. thanks a lot, saved me quite some time going through all the manuals .. guess i'm postponing the inevitable .. but thanks !	16:45
	← aroben left	16:47
	→ lcapitulino joined	16:50
	← dkagedal left	16:51
	→ fidodido_ joined	16:53
	← fidodido left	16:53
	→ kumbayo joined	16:55
	→ mario_ joined	16:57
	meyering → jim-away	16:58
	mario_ → pygi	16:58
	jim-away → meyering	16:58
	→ zeenix joined	17:01
zeenix	hi	17:01
	i've been using git-svn for more 2 months now and there hasn't been any major problem until now	17:02
	$ git-svn rebase	17:03
	Current branch master is up to date.	17:03
	$ git-svn dcommit	17:03
	Merge conflict during commit: Your file or directory 'ChangeLog' is probably out-of-date: The version resource does not correspond to the resource within the transaction. Either the requested version resource is out of date (needs to be updated), or the requested version resource is newer than the transaction root (restart the commit). at /usr/bin/git-svn line 405	17:03
	i'll appreciate if anyone can help here?	17:03
	the last commit was actually merge conflict resolution so that might have anything to do with it?	17:04
WackyKid	merge with another svn branch?	17:04
zeenix	no! my own git branch	17:05
WackyKid	ah	17:05
zeenix	elmarco: any clue?	17:07
WackyKid	i heard that git-svn gets its information about which svn branch to commit to/pull from from certain lines in the commit message...	17:07
	and sometimes this confuses it after merge	17:07
	dunno the details though...	17:07
zeenix	hm..	17:07
	← QuercusMax left	17:08
elmarco	zeenix: this has been change recently	17:11
	zeenix: it still yet not perfect, now I use set-tree to commit to our svn ;(	17:11
	I even do grafts manually.. so don't ask me	17:12
	it is a mess for me right now :)	17:12
	zeenix: conflicts merge might be like branch/merging	17:13
	zeenix: in that case, this you lost the parent relation (as far as I understand)	17:13
	and git-svn is confused, you can set a graft manually	17:13
WackyKid	i never had a problem with git-svn, but then i usually do rebase instead of merge heh	17:14
elmarco	rebase did not work for me, because I work with several branch comming from different repository (basically, debian, upstream and internal - all svn)	17:16
	and I work locally with git (happy world)	17:16
WackyKid	hehe, indeed :-)	17:16
	→ QuercusMax joined	17:18
	→ mog joined	17:18
WackyKid	for me git-svn already has many advantages over plain svn without even using git's merge support :-)	17:18
PerlJam	WackyKid: such as?	17:19
WackyKid	well, for one thing, after git-gc my git-svn repositories are smaller than svn checkouts :-)	17:19
	and the git-gui thingy with per-hunk committing is really really neat	17:20
	and of course all history browsing is blazingly fast compared to svn	17:20
	and then there's git-rebase to clean up a local commit series before i push it so the svn repo	17:20
PerlJam	I've been using git and svn as if they were in separate universes. I should probably checkout git-svn.	17:21
	(unintentional pun)	17:21
	:-)	17:21
WackyKid	:-p	17:22
elmarco	I miss git-svn tag though	17:22
	it could be possible to do it I think	17:22
WackyKid	and i didn't even mention being able to make local commits without publishing then to the server	17:23
	i'm already taking stuff for granted :-p	17:23
elmarco	(the same is true for branching, but I don't really need it since I use git ;)	17:23
	→ lyakh joined	17:23
	← elmex left	17:28
	→ elmex joined	17:29
	← Tv left	17:29
	→ dkagedal joined	17:33
WackyKid	cd #konversation	17:35
	oops :-p	17:35
capisce	/join	17:35
WackyKid	indeed	17:35
	my mind had only switched from shell to irc halfway :-p	17:36
	→ lichtblau joined	17:36
	← cortilap left	17:37
	← myrizio left	17:49
	← GyrosGeier left	17:51
	← QuercusMax left	17:53
zeenix	elmarco: sorry! was traveling to home :)	17:54
	elmarco: what is a 'graft'?	17:54
robin	zeenix: it is a mechanism to overide the commit relationships	17:56
	mostly intended for adding relations where none exists	17:56
elmarco	or when it is lost ;)	17:56
robin	no	17:56
	you cannot lose the parents	17:56
elmarco	robin: not my fault if git-svn lose the relation in 1.5.2	17:56
	and 1.5.3 cvs is weird when merging with branch not coming from svn	17:57
	← WackyKid left	18:02
zeenix	i just reset my master branch to the commit before the merge	18:03
	but i still get the same error from `git-svn dcommit`	18:03
	→ KirinDave_ joined	18:03
zeenix	ok! now it's good after i did a `git-rebase`	18:05
	← dkagedal left	18:10
	→ dorileo joined	18:12
	← KirinDave left	18:16
	KirinDave_ → KirinDave	18:16
	→ Beket_ joined	18:17
tsuna	I just updated to git.git master HEAD and git status doesn't seem to work anymore (git: 'status' is not a git-command) whereas git-status does, any idea?	18:18
Beket_	Hello people. Suppose I branch master at time A. Then I work at my branch. At time A' master branch has evolved and I want to incorporate those changes in my own branch. What do I need? git-rebase ? Thanks	18:19
	← yorgen1512 left	18:19
	→ Tv joined	18:20
	← Tv left	18:20
tsuna	Beket_: it depends. git-merge will create a new commit that incorporates the changes made in the original branch (which will lead to non-linear history) whereas rebase will pop your changes, apply the changes of the other branch and try to re-apply your changes on the top of the new HEAD (which is usually better for topic branches that contain unpublished changes)	18:20
	Hmm for some reason git help -a prints ``git commands available in '/Users/tsuna/bin'`` and then no commands are listed (whereas my git is installed under /usr/local/bin/git* that is prefix=/usr/local, the default prefix).	18:22
Beket_	Thank you tsuna	18:22
tsuna	Beket_: np, but remember that you shouldn't use rebase on a branch where commits have been published	18:23
	→ aroben joined	18:24
Beket_	sure, thanks for the hint	18:24
tsuna	ghee, I rebuilt everything with make clean all prefix=/usr/local and it's working now... I probably messed up my last compile or something oO	18:26
	→ HackyKid joined	18:27
	→ Lash\| joined	18:29
male	Does anyone here do work on gitk?	18:34
	← raw-bin left	18:34
	→ Eludias joined	18:34
	← devogon left	18:35
	elmarco → elmarco\|gym	18:35
	→ devogon joined	18:37
male	Nothing? Does anyone else have problems with gtik's interactive with window managers?	18:37
	Gitk inexplicably set the user-specified-location hint to 0,0. No matter what.	18:38
	It shouldn't touch this hint at all. It should use program-specified-location if it wants a certain placement, but I don't think it does because it's always 0,0.	18:39
tsuna	Can it harm to Ctrl-C git-gc?	18:45
robin	it should not	18:51
	if there is any damage it is a bug	18:51
workkevin	I'm sure tsuna finds that very comforting. ;)	18:56
robin	<std/disclaimer>	18:58
	→ ACSpike[Work] joined	19:02
ACSpike[Work]	is there a quick howto for msysgit?	19:03
tronical	ACSpike[Work]: download gitme.exe from http://code.google.com/p/msysgit/ ; run; wait; put $installpath/bin into your PATH; sit back and enjoy the ride :)	19:03
ACSpike[Work]	so running in cmd.exe? or the included sh.exe? or rxvt.exe?	19:04
	I've already got gitmed	19:04
tronical	ACSpike[Work]: it's a graphical installer, just click on it ;-)	19:04
ACSpike[Work]	right, I did	19:04
	it installed	19:05
tronical	cool :)	19:05
ACSpike[Work]	then I came back the next week	19:05
tronical	if all worked fine you should be set to use git then :)	19:05
ACSpike[Work]	right	19:05
	when I had tried to use git on windows in the past I ran into a number of troubles with cmd.exe not being bash	19:05
cortana	i run into these problems whenever i use windows :)	19:06
tronical	msysgit works fine for me in cmd.exe	19:06
	← bentob0x left	19:06
ACSpike[Work]	ok, thanks	19:07
	→ bentob0x joined	19:09
	→ nud joined	19:10
	← pygi left	19:10
workkevin	Is there a central location for the windows porting effort? I'd like to find out about what they are doing.	19:13
	← KirinDave left	19:15
vmiklos	http://code.google.com/p/msysgit/?	19:16
MadCoder	gitster: I just sent my patchset	19:19
	for whatever reason git-send-email didn't number them, there is 7 of them	19:19
	→ yorgen1512 joined	19:19
MadCoder	crap I sent a wrong one	19:20
ACSpike[Work]	cworth once told me how to amend the author name in the last commit but I can't remember now. I've set the name and email with git config now.	19:27
thiago_home	tronical: I came up with an "evil" plan to convert our infrastructure from p4 to git, little by little, from the ends. When people realise, we're using git :-)	19:27
tronical	thiago_home: add it to the current git transition plan on the wiki!	19:28
MadCoder	ACSpike[Work]: git commit --amend	19:28
pickscrape	Random idea: would it not be possible to create something like 'git-svk' which would be like git-svn but would be able to follow svk merges and therefore provide a more complete clone than git-svn does?	19:28
	(provided of course that the upstream repo's branches are managed by svk)	19:28
ACSpike[Work]	MadCoder: that's what I'm doing, but git log doesn't show the new name.	19:28
madduck	weird. core.worktree's description says: "The value will not be used in combination with repositories found automatically in a .git directory (i.e. $GIT_DIR is not set)" but actually it seems as if it is being used.	19:31
	which is good for my case, but I wonder what's wrong: docs or implementation...	19:31
	or whether i am misunderstanding something.	19:31
jaalto	This is real basic, I know, but how Do I get a remote branch to current repository? (the repo is empty initially)	19:32
	http	19:32
Beket_	git checkout --track -b <local name> origin/<remote name> ?	19:34
	This command creates a local branch based on the remote branch and switches your working copy to that branch	19:35
jaalto	Beket_, that that --track, thanks	19:35
Beket_	your welcome jaalto	19:36
jaalto	Beket_, Erm where do I put that htpp://..... URL in that command?	19:38
	Beket_, I explained porly: I want to fetch from http:// server a specific branch to local repo	19:39
	→ strangy joined	19:40
ACSpike[Work]	tronical: have you had success cloning over ssh?	19:42
	→ flo joined	19:43
flo	Hi! My gitweb just shows "403 Forbidden - No project found." Any suggestions ?	19:43
tronical	ACSpike[Work]: haven't tried, only used git protocol sofar	19:43
Beket_	jaalto, I don't know that, I am a noob too :)	19:43
jaalto	Beket_, It must be git-fetch, but I get error "Cannot get the repository state from ..."	19:44
	flo: I would check the file system permissions	19:45
	flo: http://www.kernel.org/pub/software/scm/git/docs/everyday.html	19:46
flo	jaalto: i set the whole repodir to 777 to make sure but that doesn't help	19:46
jaalto	flo: I understand 777 for testing, but you can see them permissions under text "CVS-style shared repository."	19:47
	← d0k left	19:48
csc`	can you push over ftp?	19:49
tsuna	Hmm git-grep keeps annoying me because I often do VPATH builds (that is, project/ is the source tree and project/_build is my build tree and my cwd) and `git-grep foo ..' (or `-- ..') won't work ("fatal: git-grep: cannot generate relative filenames containing '..'"), any suggestion?	19:49
jaalto	Beket_, I figured the error. I forgot to run "git update-server-info" at REMOTE	19:50
Beket_	oh thanks jaalto for the feedback	19:51
jaalto	For the record: To fetch from remote, simply: git fetch <URL> <branch>	19:51
	← strangy left	19:52
	elmarco\|gym → elmarco	19:52
	→ strangy joined	19:52
flo	jaalto: ok but that that can't be my problem right ? I mean it should work with 77...	19:52
jaalto	flo: Correct. Can you configure Apache to just popint to that Git share and try to generate the index listing? I would rules out the obvious reasons first	19:58
	point	19:58
ACSpike[Work]	tronical: I'm struggling. trying to see if ssh will work from inside of rxvt.exe but I'm not getting that to work either	19:59
	← strangy left	20:03
	← ACSpike[Work] left	20:04
	→ KirinDave joined	20:05
	← Lash\| left	20:07
amfwrk	I have a bunch of WIP commits, if I want them to look like a single commit pushed upstream (to svn via git-svn) do I need to git-reset --soft HEAD^^^ or how ever many commits I made then recommit, then push?	20:09
	→ ACSpike[Work] joined	20:09
	→ mnemoc joined	20:11
gitster	The easiest way to create a single commit out of them would be "git checkout -b temp HEAD^^^" (or however many) then "git merge --squash your-original-branch" followed by "git commit". Another possibility is "git rebase -i" and mark them for "squash", which is in 1.5.3 and later.	20:15
flo	jaalto: hmm no i can't right now - any other way to check ?	20:16
ACSpike[Work]	oh, running from the little msys.bat in the root works remarkably better	20:16
jaalto	flo: Unfortunately it's late here, so I must leave. I hope you get further answers	20:18
	jaalto goes sleeping	20:18
amfwrk	gitster: Neat, thanks!	20:18
	jaalto away at sleep	20:18
flo	jaalto: ok thanks!	20:18
	← nud left	20:18
flo	So... anyone else here who can assist me with setting up gitweb ?	20:18
ACSpike[Work]	tronical: yay, it works if the remote server is correctly configured. now if only I could remember what I had changed on the working box	20:19
	→ nud joined	20:20
	← nikodemus` left	20:26
	→ sgrimm joined	20:27
	← elmex left	20:32
	→ elmex joined	20:32
	← elmex left	20:38
	elmarco → elmarco\|away	20:39
xkr47	hi	20:39
	← bentob0x left	20:39
	→ bentob0x joined	20:40
	→ elmex joined	20:40
vmiklos	hm	20:43
	← bentob0x left	20:44
vmiklos	say i have foo.c, bar.c and baz.c added to index. foo.c and bar.c was already tracked, baz.c is a new file	20:45
	is it possible to undo 'git add baz.c' without unstaging foo.c and bar.c?	20:45
gitster	"git reset HEAD -- baz.c"?	20:45
hliusv561_	vmiklos: 'git rm --cached baz.c'	20:45
gitster	Reads "reset my index entry at baz.c".	20:45
vmiklos	let's try with -- :)	20:46
	→ bentob0x joined	20:46
vmiklos	fatal: pathspec 'baz.c' did not match any files	20:47
	for git rm --cached	20:47
	gitster git status still says 'new file baz.c'	20:47
gitster	is baz.c a new file?	20:47
vmiklos	yes	20:47
gitster	Then what hliusv561_ gave you plus -- should work.	20:47
vmiklos	here is the actual problem:	20:48
	# new file: po/pootle/da/jToolkit.po	20:48
	from git status	20:48
	git rm --cached -- po/pootle/da/jToolkit.po	20:48
	fatal: pathspec 'po/po/pootle/da/jToolkit.po' did not match any files	20:48
	doh	20:48
	sorry, pebkac:)	20:49
	i was not in the repo root	20:49
	→ QuercusMax joined	20:49
xkr47	:)	20:49
vmiklos	git rm -f --cached worked fine	20:50
gitster	why did you even need "-f"?	20:50
vmiklos	git rm --cached -- pootle/da/jToolkit.po	20:51
	error: 'po/pootle/da/jToolkit.po' has changes staged in the index (hint: try -f)	20:51
	i think because it was a new file	20:51
gitster	It was because you had modifications since the staged version, which makes sense.	20:51
vmiklos	hm, possible :)	20:52
	← bentob0x left	20:53
xkr47	hmm, if that was the case I think the error message is slighly misleading	20:53
gitster	error: 'foo' has staged content different from both the file and the HEAD	20:54
	(use -f to force removal)	20:54
vmiklos	hm yes	20:55
	s/possible/that was the situation/	20:55
MadCoder	oh about this "root" issue, why does git commit ../foo refuses to work ?	20:58
	it's pretty irritating actually	20:58
	OTOH, git diff and other commands works fine with ../	20:58
gitster	heh, don't make partial commits.	20:59
	patches welcome.	20:59
MadCoder	gitster: well It's a usual thing I do, when I'm twidling with a "subsystem" of our work repo	21:00
	but that I have to also add some macro or interface in our core libraries	21:00
	I usually want to bind them all in the same commit, and I'm in the subsystem directory and want to:	21:00
	git commit . ../path/to/the/lib/file.c	21:00
	gitster: actually it's unclear to me wich component of git refuses the ..	21:01
	but well, I'm not iritated enough about that yet to be bothered to trace it :)	21:01
	it was just vmiklos remarks about not being at the root of his repository that made me remember about it	21:02
	← rtmfd_icbm left	21:02
gitster	as "extra" and "sb->len" are size_t, I do not think the check in strbuf_grow() is all that portable...	21:02
	oops, wrong buffer.	21:03
	heh, but that's ok, I am talking with you anyway about your patch ;-)	21:03
	ACSpike[Work] does a little dance after completing his first two clones with msysgit	21:04
MadCoder	hehe	21:05
gitster	I hate this "= STRBUF_INIT" thing, though.	21:06
MadCoder	well, actually it's not really needed as a memset works for it. It's nice if we change the interface some day, for the static variable in modules	21:06
	maybe I should use that less in functions, and use the _init() function instead	21:07
	→ bfields joined	21:08
	← lcapitulino left	21:09
	→ etnt joined	21:09
	← QuercusMax left	21:09
	← etnt left	21:09
	← robfitz_ left	21:10
	← janm left	21:10
hliusv561_	MadCoder: AFAIK, You can't add three variables and just check overflow once, unless one of them is fixed to 1, and then you need to use '<=' instead of '<'.	21:11
	← ACSpike[Work] left	21:11
workkevin	Is there an easy way to find your repository root?	21:12
MadCoder	hliusv561_: the point is, one of the "variables' is fixed to 1	21:12
	and maybe indeed I should have used <=	21:12
	the actual test is: if (sb->len + extra + 1 < sb->len)	21:13
	:)	21:13
hliusv561_	MadCoder: And the point was that equality in that case also means overflow.	21:13
robin	lots of talk about gc. must try it someday	21:13
MadCoder	hliusv561_: and I agree like said	21:13
	→ Pistahh joined	21:13
gitster	robin: do you like the two-patch series from today?	21:14
hliusv561_	MadCoder: If extra is 2^32 - 1 (or 2^64-1 for 64-bit), the condition will be false, but overflow will occur in that calculation...	21:15
	→ robfitz joined	21:15
MadCoder	hliusv561_: you're a bit blind right ?	21:15
	23:12 MadCoder \| and maybe indeed I should have used <=	21:15
	23:13 MadCoder \| hliusv561_: and I agree like said	21:15
robin	gitster: the auto-gc	21:15
MadCoder	should I say it a third time ? :)	21:15
robin	I have looked yer	21:16
	generally, I think git should do its own homework	21:16
	→ QuercusMax joined	21:17
MadCoder	gitster: as a matter of a fact, the thing that is the less portable is the va_copy, I think there is none on AIX e.g.	21:17
	though there is compatibility code for that around	21:17
robin	i.e. I like the idea at least	21:18
MadCoder	(and IIRC if gcc is used, it provides -- or at least it did -- a __va_copy on those platforms)	21:18
robin	MadCoder: the result of signed integer overflows are undefined....	21:19
MadCoder	last time I checked, size_t was unsigned	21:19
robin	then it's fine as long as the others are unsigned too	21:21
	or..	21:21
	it's ok	21:21
MadCoder	or git won't work on some arcane DSP or imaginary architecture nobody cares about ? :)	21:22
	cough	21:22
robin	or is it.. if the result is undefined no comparison makes sense	21:22
	:)	21:23
	→ Eludias__ joined	21:23
MadCoder	I could also argue that sb->len cannot be ULONG_MAX as it's odd	21:24
	and that the allocation strategy cannot reach taht number :P	21:24
	but well	21:24
gitster	ah, I missted that size_t (somehow I read it as ssize_t), my fault.	21:24
MadCoder	size1 + size2 + 1 >= size1 is the proper test	21:24
	instead of my > and all is fine	21:25
hliusv561_	MadCoder: And aren't extensible arrays usually grown exponentially so they have O(N) amortized running time?	21:25
	← elmex left	21:26
	→ bronson joined	21:26
MadCoder	hliusv561_: I was just making silly arguments, you're not really supposed to answer to them ;)	21:26
	→ elmex joined	21:26
MadCoder	hmm gitster: there is a small issue in strbuf.c now that I reread the code again	21:28
	(it's despairing sometimes to read 200 lines of code and always come with tiny little holes ...)	21:28
	gitster: strbuf_reset should set ->eof to 0	21:28
	it breaks read_line not to do so, at least on the paper, no in the reality because we never reuse a strbuf on multiple files with this API (yet)	21:29
	← sgrimm left	21:29
	→ rtmfd_icbm joined	21:33
	→ dirak1 joined	21:35
HackyKid	isnt it a bit weird storing an eof flag in something thats supposed to be a string? or is it just me :O	21:36
MadCoder	HackyKid: well, I'd like to get rid of it as well, it's not pretty, and it's there because code using read_line are sloppy in many ways (IMHO)	21:38
	← lichtblau left	21:38
MadCoder	but I've not wrapped my mind about it yet, so I've left the current logic untouched	21:38
HackyKid	ah, i see	21:39
	← jcollie left	21:40
HackyKid	should be a seperate patch in any case i think, don't change too much at once if you dont have too :-)	21:40
MadCoder	that's kind of the point	21:40
HackyKid	hmm	21:41
MadCoder	the real reason it not really the eof thing, but that read_line tries to work with strbufs the way fgets works with buffers	21:41
	but not exactly like it, and it has some shortcomings	21:41
	e.g. it always drop the terminator we are looking for	21:41
	sometimes we want to, but I've seen at least one place where it does not makese sense, and we want to keep the final '\n' (like fgets does)	21:42
	and it's way too cumbersome to add it again	21:42
HackyKid	sounds tricky to get it all right....	21:43
MadCoder	not really, it just needs to look at git's code a bit more and understand the need	21:43
	once you know the need, the correct API is often pretty obvious	21:43
	→ sgrimm joined	21:44
HackyKid	true	21:44
MadCoder	and also it's not critical IMHO. current strbufs have IMHO grave flaws (and like Junios said it wasn't really "designed" after all)	21:44
	and that's the issue I try to address	21:44
	the read_line "shortcomings" are rather cosmetical and taste issues	21:45
	not really the same	21:45
HackyKid	yeah	21:46
MadCoder	damn, I have to buy gerit pape a beer at some point	21:47
	he uploads new git versions to debian less than 24 hours after it's released	21:47
robin	QuercusMax: ping	21:47
	← Beket_ left	21:48
	→ doublec joined	21:49
	← dirak1 left	21:50
	← dorileo left	21:51
KiBi	Gerrit, even.	21:52
hliusv561_	Concerning pack merging, where are tags sorted in packs? Mixed with commits?	21:55
	← Kenjin left	21:57
gitster	Roughly speaking, wherever "rev-list --objects" output gives them.	22:00
hliusv561_	Ah, they go into extreme end.	22:02
flo	So far i've been using svn to keep my code in sync between my main machine and my notebook. However since branches are local in git - if i start a new branch on one machine it will not be available on the second right ? (Even so i'm syncing using a shared repo)	22:03
hliusv561_	flo: You need push the branch into the shared repo and pull it on other. This does not help with reflogs through.	22:04
	← fidodido_ left	22:04
	→ pygi joined	22:04
hliusv561_	flo: s/pull/fetch/.	22:05
	→ nashdj joined	22:05
flo	hliusv561_: Hmm so i have to keep track of branches all the time ? There is no way to always push all branches and always pull all branches ?	22:07
	← amfwrk left	22:07
hliusv561_	flo: Git 1.5 and newer supports wildcards in refspecs. Also, push has option '--all'.	22:08
flo	hliusv561_: ok i will try that.. thanks	22:08
	← aroben left	22:09
	→ achiang joined	22:15
	← Eludias__ left	22:15
	← Eludias left	22:15
	← Oejet left	22:16
achiang	if i am working in a branch foo and make changes; create a new branch (git branch bar origin); and switch to that branch (git checkout bar), why do i see changes from foo?	22:16
	i would have expected that bar would match origin	22:17
gitster	When you switch branches, you are taking your local changes with you.	22:17
HackyKid	why does my git-gui look better in windows than in linux :O	22:17
gitster	The alternative you are suggesting is to effectively "lose" your local changes that are not committed to your current branch.	22:18
flo	hliusv561_: Ok seems i can push branches that way to the server - but if i pull from the server the branch will be a remote branch (origin/branchname) - Can i checkout a remote branch, work on it, commit it and push it to the server again ?	22:18
gitster	HackyKid: probably because you are too used to Windows.	22:18
HackyKid	hmm, maybe... thats part of it	22:19
	→ DrFrasierCrane joined	22:19
achiang	gitster: ok, but what if my changes in foo are temporary and i don't really want to commit them before playing around in a different branch?	22:19
DrFrasierCrane	how do I remove a file from commit ?	22:19
	← gnilor left	22:19
	← bdiego left	22:19
achiang	gitster: for example, adding debug printk statements, etc	22:19
gitster	A short answer: don't. Another answer: stash it away (see git-stash(1)).	22:19
	→ zmk joined	22:19
hliusv561_	flo: No, remote branches don't work like locals. 'git checkout' has -b option to create local branch and check it out. Also see option --track.	22:20
HackyKid	i actually set the win98(or something) theme in linux, but git-gui doesnt listen :-)	22:20
gitster	Yet another answer: if you do not need that temporary change, "reset --hard" away after or before switching branches.	22:20
	It all depends on what you want to do, and there are suitable tools to allow you do so. Learn to use them ;-)	22:21
achiang	gitster: ah, cool! that is exactly what i want... wonder why that's not called out more prominently on http://www.kernel.org/pub/software/scm/git/docs/tutorial.html	22:21
	gitster: i do notice it here: http://www.kernel.org/pub/software/scm/git/docs/user-manual.html	22:21
	gitster: yep, i don't mind reading docs/man pages, but i didn't even know git-stash existed. thanks	22:22
	fwiw, i don't think "temporarily setting aside work in progress" should really fall into the "Fixing Mistakes" category	22:22
	but thanks for the help	22:23
HackyKid	argh! the font list popupin git-gui is too large to fit on the screen, and there is no way to scroll it	22:23
achiang	ah, nuts. debian etch (4.0) only has git 1.4.4.4. :(	22:24
HackyKid	ok.... this is weird, the windows version does have arrows at top/bottom to scroll it....	22:24
	← bfields left	22:24
	← DrFrasierCrane left	22:24
HackyKid	but in linux i can only select fonts upto 'g' as starting letter :-(	22:25
gitster	annoying, isn't it? Bug spearce.	22:26
	← nud left	22:26
hliusv561_	HackyKid: And using arrow keys probably results selected entry going out of screen? :-\|	22:26
cortana	achiang: backports.org	22:26
	you can get 1.5.2 from there :)	22:26
	→ agoode joined	22:26
HackyKid	arrow keys do nothing	22:27
cortana	in farct, according to http://packages.debian.org/git-core you can get 1.5.2.4	22:27
hliusv561_	cortana: Sadly 1.5.2.4 does not have 'stash' or 'rebase -i'.	22:27
HackyKid	scroll wheels close the popup	22:27
	but its known bug i hear :-p	22:28
cortana	ah, though it should be easy to build the package of 1.5.3 from unstable on etch	22:28
achiang	cortana: ah, thanks for the backports pointer. installing it now, thanks.	22:29
	achiang discovers what hliusv561_ points out	22:30
hliusv561_	HackyKid: I got arrow keys to do something. I can't reproduce it through. And yes, the selection goes off the screen. BTW: I got list only up to 'd'.	22:31
MadCoder	cortana: as soon as 1.5.3 is in testing I'll do a backport	22:32
	but I can't do that earlier	22:32
cortana	MadCoder: sweet	22:32
	fair enough	22:32
MadCoder	and it seems there is issues with gcc on hppa again so I'm not sure when it will be	22:32
cortana	ugh	22:32
	← elmex left	22:32
MadCoder	http://buildd.debian.org/~jeroen/status/package.php?p=git-core&suite=unstable	22:32
	it happened before, I suppose building with -O0 will do the trick	22:33
	← cmarcelo left	22:34
cortana	hm another buildd status page :)	22:34
	→ lichtblau joined	22:35
MadCoder	that'sthe sole useful one :P	22:35
cortana	that makes three i have bookmarked now :D	22:35
hliusv561_	HackyKid: I figured out how to reproduce that. Pick some long font name (so the combobox will be wide). Then open the list and drag to combobox. Now up and down arrow keys work.	22:35
MadCoder	cortana: you need 2	22:36
	this one and experimental.ftbfs.de/new/ for the bpo, volatile and experimental networks	22:36
cortana	ah nice	22:36
MadCoder	the latter is the same as the former, it just that it deals with other sets	22:36
cortana	yup	22:36
hliusv561_	HackyKid: Using that trick, I managed to select font name beginning with 'v' from the font list.	22:39
MadCoder	or you can close git-gui, edit its config file and use xfontsel and use vi	22:40
csc`	so can git push over ftp :s?	22:41
HackyKid	ah, arrows work now yeah, not too usefull when you cant see what you select though :-)	22:41
	← zmk left	22:44
	→ aroben joined	22:45
	→ jasam joined	22:49
	← devogon left	22:51
	← ferdy left	22:55
	← p4tux left	22:56
hliusv561_	HackyKid: I think the problem is in TK libraries, not in Git-gui.	23:12
	← _aleph left	23:14
	← kblin left	23:14
	→ kblin joined	23:15
	← kblin left	23:15
	→ kblin joined	23:15
	← kblin left	23:15
	→ kblin joined	23:15
	← kblin left	23:16
	← orospakr left	23:16
	→ dash__ joined	23:16
	→ cdcarter joined	23:17
	← cdcarter left	23:17
	← flo left	23:17
	← lyakh left	23:20
	→ Nemoish joined	23:26
	← Nemoish left	23:27
	← dash_ left	23:28
HackyKid	hliusv561_: yeah, looks that way	23:29
	especially because the same script does allow scrolling on windows :O	23:29
	← BadQuanta left	23:37
	→ BadQuanta joined	23:37
	→ bdiego joined	23:37
	→ reelman_ joined	23:39
	← achiang left	23:39
	← reelman_ left	23:39
	→ reelman joined	23:40
gitster	I thought that was known and you guys were talking about a workaround...	23:43
	← robin left	23:44
reelman	hi, if i clone a git repo, the branches dont get cloned... how do i clone with all the branches?	23:44
gitster	what makes you think branches are not cloned?	23:44
	(hint "git branch -a")	23:44
	← krh left	23:47
reelman	so when i check out a branch with origin/branchname what happens?	23:47
	like...	23:47
	am little lost!	23:47
	its says am not on a local branch!	23:47
hliusv561_	reelman: 'git checkout --track -b branchname origin/branchname' probably (if you intend to commit on top of it)...	23:48
reelman	cool.. so if i make changes to this branchname branch and commit and later push	23:50
	the remote branch will get updated?	23:50
hliusv561_	reelman: Yes.	23:50
reelman	hliusv561_: OK.. thanks.. i will try this!	23:50
hliusv561_	reelman: The '--track' sets defaults such that 'git pull' will try to merge origin/branchname (after fetching it from remote repo).	23:52
reelman	if i do this , then master branch wont get pulled?	23:52
	when i do git pull?	23:52
hliusv561_	reelman: Yes.	23:53
	reelman: If you really want to merge origin/master for some odd reason, you can first fetch and then 'git merge origin/master', or 'git pull origin master' (or what the exact syntax was).	23:54
reelman	so is it possible to track more than one branch?	23:54
	→ FunkeeMonk joined	23:54
reelman	ok	23:55
	i think i get it	23:55
	hliusv561_ doesn't use 'git pull' much. He likes more to use 'git fetch' first and then 'git merge'.	23:56
	← Yuuhi left	23:56
	→ standel joined	23:56
	← standel left	23:56

Logs Search ←Prev date Next date→

Channels Documentation