IRCloggy #git 2008-12-26

Provider of IRC logs since 2005.

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

	2008-12-26
	← dasil003 left	00:00
gitster	"git diff $commit...origin -- manifest" would show lines that they did not change prefixed with " ", lines they changed prefixed with "-" and "+"; - are what the file had before their change, + are what the file contains after their change.	00:00
	→ akitada joined	00:00
	← akitada left	00:02
	→ akitada_ joined	00:03
	→ oudeis joined	00:03
	→ Jacolyte joined	00:04
	← akitada_ left	00:06
	← Ademan left	00:07
	→ aoper joined	00:09
	← charon left	00:09
	← aoper left	00:13
	→ aoper joined	00:14
	← TanEk left	00:16
	← zerok_ left	00:16
	→ akitada_ joined	00:20
	→ dreiss joined	00:21
	← akitada_ left	00:22
	← kumbayo left	00:29
	← _ajonat left	00:33
	→ johnw joined	00:35
	← jmatthews left	00:37
	← gitster left	00:37
	← johnw left	00:38
	← foutrelis left	00:38
	← Tr3B left	00:41
	→ jmatthews joined	00:41
	← dwmw2_gone left	00:43
	← LotR left	00:45
	→ gitster joined	00:50
	→ pie joined	00:52
	← d0k left	00:52
	→ __david joined	00:56
__david	How do I combine several commits into a single commit?	00:56
gitster	"merge --squash" or "rebase -i"	00:56
__david	can you be a bit more specific?	00:56
	Like, if I want to combine the past N commits?	00:57
	And how does it deal with their commit messages?	00:57
	Does it append them, or discard them and let me specify a new one?	00:57
	← scrambledegg left	00:59
	→ uva joined	01:03
	← CodeOfficer left	01:10
	← greyface left	01:11
	← chris2 left	01:14
	← joydivider left	01:17
	← name left	01:19
	→ Strogg joined	01:23
	kbiddlecomb → kbiddlecomb_	01:29
azuk	I wonder what's the logic with the Branch: field gitk displays in the diff, it is empty in some branch tip commits but not in all	01:30
	→ quoin joined	01:32
	→ stouset joined	01:32
	→ tanek_ joined	01:36
	← priidu left	01:36
	← markelikalderon left	01:36
	← stouset left	01:41
	→ markelikalderon joined	01:41
	→ cartucho joined	01:42
Ilari	azuk: Are they even local branches? Not remote branches or even tags?	01:43
azuk	yes	01:43
	← pygi left	01:50
	← bobesponja left	01:51
azuk	while I'm trying to come up with a script to create such a repository, here's another case. does this look correct to you: <http://www.iki.fi/azure/tmp/gitk-test1.png> (test script to recreate that repository: <http://www.iki.fi/azure/tmp/gitk-test-script>)	01:52
Gitbot	azuk: I'm sorry, there's no such object: http://www.iki.fi/azure/tmp/gitk-test1.png.	01:52
	→ bobesponja joined	01:52
j_engelh	azuk: repost with --date-order	01:52
	simply commit the entires from bottom to top and you recreated it	01:53
	← quoin left	01:53
j_engelh	azuk: you're good when you can recreate something like http://jengelh.medozas.de/images/git-forest.png	01:55
	← uva left	01:55
azuk	ok, here's with --date-order, not that understand what difference and why that should make: <http://www.iki.fi/azure/tmp/gitk-test1-date-order.png>	01:56
Gitbot	azuk: I'm sorry, there's no such object: http://www.iki.fi/azure/tmp/gitk-test1-date-order.png.	01:56
azuk	...and the point is that I'd expect it to display both debug2 and branch as branches	01:57
	→ uva joined	02:00
azuk	should I adjust my expectations? :)	02:01
j_engelh	and?	02:02
azuk	and what?	02:02
j_engelh	well think about it	02:03
azuk	I have been thinking about it and it doesn't make sense to me, as far as I can see that commit is reachable from both debug2 and branch and so should be part of both branches, just like its parent	02:05
	← aoper left	02:06
j_engelh	you have exactly 6 branches in there, of which 3 are accessible my a green name	02:06
	s/my/by/	02:06
	→ moconnor joined	02:07
	→ aoper joined	02:07
	← markelikalderon left	02:09
azuk	I can agree with that	02:09
	← chuck left	02:11
azuk	I don't see how that answers my question, though	02:12
	→ k776 joined	02:13
	← kbiddlecomb_ left	02:15
	→ markelikalderon joined	02:16
	← k776 left	02:18
	← aoper left	02:19
j_engelh	I really don't know where the problem is.	02:22
	If you want to recreate the history tree like that, just commit in the right order.	02:22
azuk	I know how to create a history like that, I just want to know why, in that screenshot, gitk displays "Branch: debug2" instead of "Branch: debug2, branch"	02:24
	→ litage joined	02:24
	← mbroeker left	02:28
	→ QaDeS_ joined	02:28
j_engelh	azuk: it shows both branches for me	02:29
	azuk: http://tinyurl.com/82wyd8	02:31
azuk	ok, I suppose I should make a bug report then	02:33
	← oudeis left	02:33
j_engelh	no, why? there is no bug.	02:33
	at least not in the version I have, whcih is the most recent.	02:33
azuk	hm, mine should be the most recent as well, let me verify	02:34
	← Yuuhi left	02:34
azuk	ok, I'm missing a couple of commits, but they all look completely unrelated. but let me build the latest version just in case	02:36
j_engelh	and be sure to run gitk --all	02:37
azuk	yep, that's what I did	02:37
j_engelh	winner? wtf. debian is falling apart due to their politics, and ubuntu does not have the technical know-how to get anything goin.	02:37
	→ j_engelh joined	02:38
j_engelh	screw this irc client to hell and back	02:38
offby1	er?	02:39
j_engelh	or maybe my keyboard is giving up on me	02:39
	(when trying to use ^X to switch channels)	02:40
offby1	dried eggnog stuck between the contacts, most likely	02:40
	→ faithfulgeek joined	02:40
j_engelh	i don't drink that crap	02:40
	← faithfulgeek left	02:44
	← QaDeS left	02:45
	→ metajack joined	02:47
	← bobesponja left	02:47
	← bryanl left	02:49
	→ julian81 joined	02:51
	← drewr left	02:53
	→ bobesponja joined	02:58
	← bobesponja left	03:05
	← julian81_ left	03:05
	→ dasil003 joined	03:09
	→ _slacker_ joined	03:11
	← thiago left	03:15
	← cartucho left	03:16
	→ ari-_-e joined	03:17
ari-_-e	hey guys, what command would I use if I wanted get the tree that a tree-ish resolves to?	03:18
j_engelh	well	03:18
	a tree-ish already is a tree	03:18
	you mean display contents?	03:18
deskin	ari-_-e: git rev-parse $treeish^{tree}	03:18
ari-_-e	erm no, commits and tags are also tree-ishes	03:19
	deskin: thanks	03:19
j_engelh	oh right	03:19
	→ mbroeker joined	03:20
	→ aoper joined	03:22
ari-_-e	another question: is it possible to request a single object and all additional objects that would be required for "git fsck" to report no problems from remote?	03:26
	rather than fetching a ref?	03:26
j_engelh	that would be all reachable committish,treeishs and blobishs	03:27
ari-_-e	hm?	03:27
j_engelh	git rev-list $object	03:28
ari-_-e	well it'd be weird for commits because it would fetch the entire history, but for a tree (which is what I want) you'd just have to get maybe some more trees and blobs	03:28
j_engelh	what's weird with not fetching all commits..	03:28
Tv	ari-_-e: IIRC git-daemon only serves things from refs so it won't accidentally serve "deleted" or "just sharing storage" objects you're not supposed to see	03:28
j_engelh	a missing object is not necessarily a bad thing	03:28
ari-_-e	hmm, damn	03:29
j_engelh	you just can't play with it if it's not there	03:29
Tv	ari-_-e: asking it for a blob or tree would force it to walk the DAG to see if that's allowed; as that's costly, it chooses not to allow that	03:29
ari-_-e	j_engelh: hmm, well if something points to it it's bad :-/	03:29
j_engelh	ari-_-e: not necessarily	03:29
Tv	ari-_-e: you can grab a full clone and feed those in from that, though	03:29
j_engelh	there is this thing called shallow clones	03:29
ari-_-e	oh, right	03:29
j_engelh	I could imagine - but it does not seem implemented of course - that ther emight also be shallow trees	03:29
ari-_-e	j_engelh: so it doesn't just do shallow clones by having the last commit in the chain not have a parent?	03:30
	I guess it would have to change the hash...	03:30
j_engelh	hm?	03:30
	forget what I mumbled; trees should be complete or you can't easily verify them	03:30
Tv	ari-_-e: well yes but not by mutating anything, by writing down "i don't have the parents of X"	03:30
ari-_-e	Tv: ok	03:31
	Tv: what does DAG stand for?	03:31
Ilari	ari-_-e: Directed Acyclic Graph.	03:32
j_engelh	Dag is a nordic name	03:32
	:p	03:32
ari-_-e	ok, so that's just the commit chain complete with branches and all that	03:32
	right?	03:32
j_engelh	-ETOOMANYQUESTIONSATONCE	03:32
ari-_-e	:P	03:32
	← nanodeath left	03:33
Tv	ari-_-e: http://eagain.net/articles/git-for-computer-scientists/	03:33
Ilari	ari-_-e: Branches are just pointers to nodes of that DAG. And the DAG can be highly nonlinear.	03:34
ari-_-e	hey, that looks cool	03:34
	Tv: thanks for the link	03:34
j_engelh	ze internet is on ze fritz again	03:37
	← markelikalderon left	03:37
	← flaguy left	03:38
	← da1l6 left	03:39
	→ da1l6 joined	03:40
	← dasil003 left	03:42
	→ jeckel joined	03:46
	← jbellis left	03:48
	← mbroeker left	03:49
	← jmatthews left	03:49
	← QaDeS_ left	03:49
	→ tanek__ joined	03:50
	← scientes left	03:52
	← jeckel left	03:52
	→ jeckel joined	03:52
	→ jmatthews joined	03:55
	← Shilin2 left	03:59
	← elmex left	04:00
	← jeckel left	04:01
	→ jeckel joined	04:01
	→ elmex joined	04:03
	→ muthu_ joined	04:03
	← aoper left	04:04
	→ mithro joined	04:05
	→ aoper joined	04:05
	→ jeckel_ joined	04:06
	← jeckel left	04:06
	← jeckel_ left	04:06
	← tanek_ left	04:07
	← kukks left	04:10
	→ kevwil joined	04:11
	← parasti left	04:11
	→ j416 joined	04:11
	← kevwil left	04:12
	→ cartucho joined	04:17
	→ bobesponja joined	04:34
	← aziz left	04:34
	← mithro left	04:34
	→ psankar joined	04:37
	→ bieneff joined	04:39
	← bieneff left	04:44
	← ari-_-e left	04:44
	← GyrosGei1r left	04:47
	→ GyrosGeier joined	04:47
	← schlort left	04:52
	← moconnor left	05:01
	→ rubydiamond joined	05:07
	→ sammcd joined	05:12
	→ xinming joined	05:16
	→ tjafk1 joined	05:27
	← tjafk left	05:28
	→ jackowayed joined	05:30
	→ gottesmm joined	05:39
	→ qrush joined	05:41
	← ijcd left	05:41
	← j416 left	05:47
	← gottesmm left	05:48
	→ lholden joined	05:51
	← branstrom left	05:52
	→ jstad joined	05:55
	→ j416 joined	05:56
	→ jackdempsey joined	06:01
	→ carllerche joined	06:02
	← jackdempsey left	06:06
	→ lresende joined	06:07
	← lresende left	06:10
	→ etienne9 joined	06:13
	← ankit9 left	06:13
	← sammcd left	06:15
	→ cedricv joined	06:15
	← etienne9 left	06:16
	← lholden left	06:18
	← jackowayed left	06:19
	← rubydiamond left	06:19
	→ rubydiamond joined	06:20
	← rubydiamond left	06:20
	→ lresende joined	06:24
	→ quoin joined	06:25
	→ oudeis joined	06:26
	← cartucho left	06:28
	→ Shilin2 joined	06:37
offby1	alors	06:38
	→ dasil003 joined	06:46
	→ PloenK joined	06:56
	→ jackdempsey joined	06:59
	→ __doc__ joined	07:04
	← dasil003 left	07:10
	← doener left	07:11
	← univate left	07:11
	→ doener joined	07:11
	→ joydivider joined	07:15
	← bobesponja left	07:29
	← carllerche left	07:33
	→ pyalot joined	07:35
	→ priidu joined	07:39
	← __doc__ left	07:40
	← jstad left	07:41
	← jackdempsey left	07:54
	← dreiss left	07:56
	→ temoto joined	07:56
temoto	Hello. We're 3 devs pushing to one repo and pulling from that one repo. Like SVN. It was once that pull created a merge commit. I reset that back and did rebase, resulting in conflict and again i did a merge commit. Does it mean we can generally use pull to synchronize with center-repo or better use rebase?	07:59
	i mean if merge commit was like unevitable	08:00
godlygeek	as far as i know, git pull --rebase is a reasonable thing to do in that situation, to try to rebase upon the stuff that's pulled down, rather than merge it into your branch.	08:01
	← litage left	08:02
temoto	godlygeek: ya and there is branch.name.rebase = true config to do that for each pull.	08:02
	→ davik joined	08:03
godlygeek	yeah.	08:03
temoto	I'm trying to get the feeling which one is appropriate in which situation.	08:03
	Like i'm sure where to use tuple, where to use list in python.	08:04
godlygeek	yeah. i have a general feeling for when to use each, but since i'm not exactly an expert on the matter, i won't muddy the waters with my limited understanding...	08:05
	← julian81 left	08:05
	→ ankit9 joined	08:05
	← ankit9 left	08:06
	→ ankit9 joined	08:06
temoto	godlygeek: thank you :)	08:07
	← ankit9 left	08:10
	→ ankit9 joined	08:11
	← aoper left	08:19
	← pyalot left	08:20
	→ __doc__ joined	08:25
	→ aoper joined	08:27
	→ muthu joined	08:28
	← __doc__ left	08:38
	← xinming left	08:41
	← oudeis left	08:42
	← muthu_ left	08:43
	→ __doc__ joined	08:47
	→ Ademan joined	09:04
	← Jacolyte left	09:05
	← ankit9 left	09:06
	→ ankit9 joined	09:06
	← cao left	09:07
	← temoto left	09:07
	← PloenK left	09:08
	→ scientes joined	09:16
	→ oudeis joined	09:27
	→ daa joined	09:34
	← ia left	09:40
	→ ia joined	09:41
	← unreal left	09:43
	← daa left	09:44
	→ wwwald joined	09:44
	→ unreal joined	09:49
	← ankit9 left	09:53
	→ ankit9 joined	09:54
	→ Tr3B joined	09:58
	← joydivider left	10:04
	→ chris2_ joined	10:14
	← wwwald left	10:16
	chris2_ → chris2	10:16
	→ Voker57 joined	10:16
	→ Flathead joined	10:17
	→ mbroeker joined	10:18
	← __doc__ left	10:22
	chris2 → chris2_	10:27
	chris2_ → chris2__	10:27
	chris2__ → chris2	10:28
	→ joydivider joined	10:31
	→ schlort joined	10:31
	← mbroeker left	10:36
	→ mbroeker joined	10:36
	→ greyface joined	10:37
	← dermoth left	10:43
	← unreal left	10:46
	→ dermoth joined	10:46
	→ drizzd joined	10:47
	← chris2 left	11:02
	← ndim left	11:08
	→ pygi joined	11:09
	← joydivider left	11:10
	← Shilin2 left	11:10
	→ not-xjjk joined	11:11
	← pygi left	11:14
	→ ph^ joined	11:14
	→ Shilin2 joined	11:16
	→ rtomayko_ joined	11:17
	← rtomayko left	11:17
	→ carllerche joined	11:21
	→ k776 joined	11:22
	← xjjk left	11:23
	→ QaDeS_ joined	11:23
	← k776 left	11:25
	← muthu left	11:26
	→ muthu_ joined	11:30
	→ masterzen joined	11:31
	← carllerche left	11:34
	→ bentob0x joined	11:34
	← Voker57 left	11:35
	→ Voker57 joined	11:36
dmq	i dont "get" git describe	11:39
	it warns about stuff that you cant control.	11:39
	which doesnt make sense to me.	11:39
	← j416 left	11:42
	← schlort left	11:42
	← psankar left	11:43
	← davik left	11:43
Ilari	dmq: What warning?	11:48
dmq	the problem is with git describe --all	11:49
	just mailed the list about it.	11:50
	but let me make a paste	11:51
Ilari	dmq: That tag foo is really tags/foo message?	11:51
dmq	http://pasta.test-smoke.org/122	11:51
	yes exactly.	11:52
	theres no way to turn it off	11:52
	or to force git-describe to just output "tags/blah blah" instead	11:52
	and i cant not use --all as if i dont use it some of my commits dont resolve.	11:52
Ilari	Apparently trying to describe anything that it uses tags as base gives that warning. It even pops up in git.git clone I have. And AFAICT those tags it warns about are correctly named.	11:54
dmq	yeah	11:55
	the warning is one of those silly warnings that we all stick in code without thinking it through.	11:55
	→ thiago joined	11:55
	→ [TuxCM] joined	11:55
dmq	never warn about something you cant control, and if you do warn include explanation for how to turn off the warning.	11:55
Ilari	dmq: It seems that conditions to display it contain logic error. It seems to pop up in cases where it shouldn't pop up.	11:57
dmq	that would explain why its so annoying :-)	11:57
	and yes, it almost seems random.	11:57
	see, ive been running git-describe on about 1000 commits from the perl history.	11:57
	everywhere that the version of perl was changed.	11:58
	so that i can ensure we have all the right version tags.	11:58
	and part of it is git-describing each.	11:58
	and i get that error about 30% of the time or so...	11:58
	→ peritus_ joined	11:58
	→ kumbayo joined	11:58
dmq	that warning	11:58
	and its really irritating	11:58
	as a) id like all the git-describes to use the same qualified format, and b) i cant turn it off.	11:59
Ilari	dmq: I'd say there's not much need to turn that off if it would be fixed to only warn when it should...	12:00
dmq	well, id like to turn it off by saying "always use the qualified form" and then it should have nothing to warn about right?	12:01
Sonderblade	How do you view the diff of a specific commit?	12:02
andreaja	git show <hash>	12:02
Sonderblade	thanks	12:03
dmq	or if you feel like typing more git log -p -1 SHA!	12:03
Sonderblade	it seems git svn rebase has made it so that some commits have become duplicated, how can that happen?	12:04
dmq	ilari, id be really interested to see how such a change/new switch would be handled. ive looked at patching git a couple of times, but i didnt get a lot of traction.	12:04
	it would be nice to see an example of adding a well defined and small feature.	12:05
	→ temoto joined	12:09
Ilari	dmq: IMO, its more like bugfix. If it's indeed logic error, then fixing that is a bugfix.	12:12
	← kumbayo left	12:13
	→ d0k joined	12:13
temoto	Hello. We're moved from SVN, 3 devs pushing to one centralized repo. Now which one of pull&rebase to prefer and when?	12:14
dmq	Ilari: sure, but adding aswitch to force the qualified format always would be a teeny feature?	12:15
thiago	temoto: to do it "the git way", you want rebase	12:15
	git checkout master; git pull; git checkout mybranch; git rebase master; git checkout master; git merge mybranch; git push	12:15
	or, more simply: git checkout mybranch; git pull --rebase; git push	12:16
temoto	hmm	12:17
	do i have to have separate branch?	12:17
	we're working on master	12:18
Ilari	dmq: Hmm... It seems that my first guess at what causes that warning was right...	12:18
temoto	and we're pushing our master to server master	12:18
bremner	temoto: you can also do this with git. Don't be afraid of branches in git though	12:19
	→ tomsw joined	12:19
temoto	bremner: i'm not, i'm using few branches for my own projects.	12:19
tomsw	happy Boxing Day everyone	12:20
	bremner boxes tomsw's ears :-)	12:20
dmq	ilari: bug?	12:20
temoto	It's just a question of is there any good reason to have mybranch and keep rebasing it to master?	12:20
tomsw	Is it possible to move the last n commits from one branch to another?	12:20
dmq	keeps the commit history clean.	12:20
tomsw	I did some work on the wrong branch	12:21
	← [TuxCM] left	12:21
temoto	dmq: i see. That way i'll have to make "merge commits" that glue several commits of mybranch into one on master branch?	12:22
bremner	temoto: not really. Also "merge commits" are something else. You probably mean squashed commits in git lingo	12:24
Ilari	temoto: If you keep mybranch and keep it rebased to master, don't push it anywhere without warnings about it being rebased.	12:24
dmq	tomsw: my comment was more to temoto.	12:25
	← ankit9 left	12:25
dmq	temoto: im not really sure if i follow you. the reason some people like rebase is because it keeps the commit history clean on master.	12:26
	→ ankit9 joined	12:26
dmq	as in "clean of merges".	12:27
	so it looks like a straight line in things like gitk.	12:27
Ilari	And rebasing together with push and pull can create nasty surprises if one isn't careful...	12:27
dmq	yes	12:27
thiago	temoto: you asked what was the recommended way	12:28
temoto	ya	12:28
thiago	the recommended way is not to work on master, but to work on topic branches	12:28
temoto	topic?	12:28
thiago	check out, rebase, work	12:28
	when you're done, you merge it into master, then push	12:28
dmq	yeah	12:29
	but it doesnt always quite work out that way.	12:29
temoto	I'm sorry what is "topic branch"?	12:29
dmq	you create a temporary named branch that is for working on a given "topic"	12:29
Ilari	That leaves out the step for deleting that branch (comes after merging it to master)... And to avoid nonliearities, one should do one final rebase before merge (to ensure merge will be trivial).	12:29
temoto	omg	12:30
dmq	you do your changes there, and then rebase it into master and throw away the temporary branch when you are done	12:30
	tomsw figures it out	12:30
Ilari	temoto: For example branch 'bugfix-225' (for fixing bug #225 in bugtracker). Or branch 'ECDSA' for implementing ECDSA (a digital signature algorithm)...	12:30
	← ankit9 left	12:31
dmq	at least thats the idea.	12:31
	imo it doesnt work out that way in practice.	12:31
	→ ankit9 joined	12:31
temoto	to rebase master, i should do git checkout master ; git rebase bugfix-225 ?	12:31
	← muthu_ left	12:31
Sonderblade	i'm trying to rebase a branch, i get a conflict, i solve it by selecting the second version, then the changeset becomes empty and rebase can't continue	12:31
	git is broken	12:31
dmq	not with a large number of committers with varying skills working out the same repo anyway.	12:31
thiago	in practice, my work consists of making small changes in master	12:32
	if I don't finish it with one commit or in a couple of hours, I make a branch out of it for later	12:32
Ilari	temoto: More like 'git checkout bugfix-225', 'git rebase master'.	12:32
bremner	dmq: "working out of the same repo" is not "the git way"	12:32
thiago	since every commit, before being pushed, needs to be reviewed by a peer, sometimes I need to create a branch in order to keep commits that aren't reviewed yet	12:32
Ilari	Sonderblade: 'git rebase --skip'?	12:32
dmq	bremner: i mean sharing a single master repo.	12:33
temoto	Ilari: i meant before pushing	12:33
dmq	and yes i know.	12:33
thiago	bremner: actually, it is. Just because the kernel uses pulling doesn't mean everyone has to.	12:33
Sonderblade	Ilari: that worked, how strange	12:33
dmq	bremner: IOW, a bare shared repo as a master repository.	12:33
bremner	thiago, dmq: I thought dmq meant one local repo. my misunderstanding	12:34
dmq	to use the pull model you need release managers afaict.	12:34
Ilari	temoto: The command you gave would rebase master on top of bugfix-225 (yielding likely some very wrong results).	12:34
bremner	dmq: so what breaks about topic branches in this setting?	12:34
Ilari	temoto: The usual syntax of rebase is 'git rebase <branch on which current branch is based on>'.	12:35
dmq	bremner: its not that anything breaks, its that it requires a reasonably clueful set of devs.	12:36
Ilari	temoto: It would be possible to configure 'git pull' to perform the correct rebase when on those branches...	12:36
bremner	dmq: maybe you need the "idiot branch", which is merged back in by someone clueful	12:36
dmq	if you have a mixed team, maybe web designers too, businessy people, etc. merging starts becomming much more common just because of the skill levels.	12:37
	bremner: thats what i mean by a release manager	12:37
	→ girishr joined	12:38
dmq	anyway, in a team where they try not to have such things, rebasing starts causing a lot of problems, and so people end up merging a lot.	12:38
	at least thats what ive seen so far.	12:38
	← ankit9 left	12:38
bremner	dmq: ok, I understand your point now. Merging is not the end of the world, after all	12:38
	← girishr left	12:38
dmq	right, my point was mostly the objective of a clean tree is hard to achieve with a mixed crew of committers. So unless you are willing to have branch managers doing the pull/cherry model (expensive for a company, but superior, and required in open source normally), the end result will be the push/merge model and a "messy tree".	12:40
	← QaDeS_ left	12:43
thiago	what we're doing is that we're developing a tool to do that for us	12:44
	instead of pushing, each person uses a separate command	12:44
	that sends the commit(s) into a change-tracking system	12:44
	each changeset needs to be reviewed by someone else and it needs to pass the tests	12:45
temoto	thiago: like mondrian?	12:45
thiago	once it does, the system automatically rebases and pushes	12:45
	yes	12:45
temoto	thiago: i've heard there's an opensource analog.	12:45
dmq	one of our teams "things" is not having beurocratic barriers to dev.	12:45
	they wont go for any kind of formal review process at all.	12:46
thiago	it's not a bureaucratic barrier per se	12:46
temoto	dmq: good one.	12:46
thiago	it's just making a tool for what we already do: reviews	12:46
	nothing goes in unless it's reviewed	12:46
temoto	thiago: how's it called?	12:46
thiago	"the tool we're developing"	12:46
dmq	thats a nice system, but its slow, and its expensive.	12:46
	:-)	12:46
	personally i like the "two man rule", i think that works well. but i cant imagine even that being put in place at my shop.	12:47
thiago	when you have such a huge codebase as we do (even with paying customers), you can't afford silly mistakes to go unchecked.	12:47
	→ bryanl joined	12:47
petteri	I, know that this is not a strictly git question, but here it is anyway. I have installed git to /opt/, so /opt/bin/ has git binaries. When I try to use git clone to clone repository over ssh from the host where git is in /opt I get bash: git-upload-pack: command not found. If I execute ssh petteri@host_where_the_git_is_in_opt "echo $PATH" i can see that my path is set correctly, bit I still can't execute any git commands over ssh. Any pointers what I am doing	12:47
thiago	besides, it improves code quality	12:47
dmq	yeah. i understand.	12:47
	btw, whats a "huge codebase" to you?	12:48
thiago	right now, what we do is very simple: we have an internal pastebin website	12:48
	once you're done with a change, you paste it there, then ping someone on IRC to review it	12:48
	usually, that happens within 10 minutes. Once you've got it, you push.	12:48
dmq	are they release managers? or is it just a formal version of two-man?	12:48
thiago	27000 files, way over a million lines of code.	12:48
	any one (who knows the code)	12:49
	the release manager has his own branch of what he'll release	12:49
dmq	ah, that sounds pretty reasonable.	12:49
	we use one-man change-and-rollout, and suffer the consequences if you break our systems.	12:50
	→ branstrom joined	12:50
thiago	as in "push and run" ? :-)	12:50
dmq	heh.	12:51
	something like that.	12:51
thiago	we had that a few weeks ago. It was even reviewed...	12:51
	but he went on vacation and, next Tuesday, we had to revert everything.	12:51
dmq	push and then sync to whacks of servers and then wait for the fireworks.	12:51
	which will come almost instantly. :-)	12:51
thiago	it was supposed to be optimisations, but it introduced a ton of regressions	12:52
dmq	yeah, but that always happens, no matter what you do, if you put code live before a holiday.	12:52
thiago	of course, the worst part is that he's the boss :-)	12:53
	→ dato joined	12:53
dmq	Thats the golden rule. Never put code live before the weekend or a vacation. Regardless of how many regression tests you have. The Gods of computing are unyealding on this rule, they Shall Punish All Who Transgress.	12:53
	:-)	12:53
	the boss eh.	12:54
	nice.	12:54
	← Ademan left	12:56
thiago	and, looking at his code, you know why he's the boss	12:56
	I mean, I'm nowhere near inexperienced, but when he asked me to review a set of patches, I kept thinking to myself, "cool, that's very ingenious"	12:56
dmq	heh, that one could go both ways. either its peter principle, or hes really really good. :-)	12:56
Ilari	dmq: The reason for warning seems to be a bug in display_name() of builtin-describe.c. It doesn't take into account that in --all mode, normal tags in refs/tags/ have their names passed as 'tags/foo', while tag object internal name is 'foo'.	12:56
	dmq: In non-all mode, tag 'foo' is passed in as 'foo'.	12:57
	→ Ademan joined	12:57
	← d0k left	12:57
dmq	thiago: theres the old line: debugging is harder than coding, so if you code at the maximum of your ability you are by definition not qualified to debug it.	12:57
thiago	yep	12:57
dmq	ilari: so it then strips off the "tags/" and warns about it too?	12:59
	neat.	12:59
petteri	found out what was wrong. I had to enable PermitUserEnvironment from sshd_config and add path to ~/.ssh/environment	13:00
Ilari	dmq: I tested patch which strips off 'tags/' prefix in all-refs mode. The test outputs some warnings and errors but they all seem genuine.	13:00
	petteri: Bit of warning about 'PermitUserEnvironment'. Some environment variables can be used to cause dynamic linker to transfer control to arbitiary code. So if any user has restricted shell, ensure that they can't modify user environment...	13:02
	← drizzd left	13:02
petteri	ok, thanks. This machine is only for my private use, but good to know	13:03
Ilari	petteri: It gets worse. Even /bin/false can be hijacked with those env vars...	13:04
	→ charon joined	13:04
Ilari	petteri: That is, if "shell" program is not statically linked, setuid, setgid or setsid, it can be hijacked.	13:05
	And that setsid thing requires dynamic linker that is aware of AT_SECURE flag.	13:09
	setsid = Set Security ID.	13:09
petteri	other option seems to recompile sshd so that its paths include /opt	13:10
tomsw	shouldn't a hard reset delete any local changes?	13:10
	if I do "git reset --hard" followed by "git status" there's a changed file	13:11
	which I don't understand	13:11
	→ chalky joined	13:13
j_engelh	Ilari: absolutely not	13:13
	s is for session here, but read the manpage.	13:14
Ilari	j_engelh: sid is also security ID. So like uid => setuid, sid => setsid. Obiviously, setsid() is also function used for something completely different.	13:18
j_engelh	Ilari: only on Windows. And Windows is not quite relevant here :p	13:19
	→ d0k joined	13:25
	→ oudeis_ joined	13:25
Ilari	j_engelh: Or on Linux with pretty much any LSM more complex than the standard capabilities active...	13:25
j_engelh	fact is, setsid, on Linux, it's session.	13:26
	→ flaguy joined	13:26
	→ markelikalderon joined	13:26
	→ aziz joined	13:27
Ilari	j_engelh: SID is also used to mean 'Security ID'. And not just for Windows...	13:30
	← bryanl left	13:30
	→ unreal joined	13:32
	← oudeis left	13:34
	← Flathead left	13:34
	← oudeis_ left	13:37
dato	why there are only 3 %C<color>?	13:38
	(in --pretty, I mean)	13:38
Ilari	dato: Because there aren't more of them implemented? Altough due to limitations of teletypes, there are really only 6 usable colors.	13:41
	← bentob0x left	13:41
dato	I was hoping for the 7 or 8, plus their bold counterparts	13:42
deskin	dato: you can always inject ascii color sequences using literal bytes, %x00 etc.	13:42
dato	right	13:42
Ilari	dato: The missing three are 'brown', 'cyan' and 'magenta'...	13:42
dato	I also use black sometimes (for grey)	13:43
	→ pygi joined	13:44
	← ph^ left	13:46
	← masterzen left	13:48
Ilari	There's Xterm 256 colors extension but it is nonstandard...	13:50
	→ xenoterracide joined	13:50
j_engelh	7 usable.	13:51
	red,green,brown,blue,magenta,cyan,gray.	13:51
xenoterracide	if I want to change the default pull of git pull how would I do that? read the man page but not sure I get it. (to early in the morn)	13:51
j_engelh	but who cares about 7	13:51
	What one needs to know: VGA has 16 foreground/8 background; standard Xterm has 16/8 by default too plus 3 markups; the 256 are extended indeed.	13:52
deskin	xenoterracide: you mean, the remote repository where pull comes from when you don't give it any arguments?	13:52
xenoterracide	deskin yes	13:52
deskin	look in .git/config for [remote "origin"] section; change the url and fetch parameters to whatever you want	13:53
	→ wwwald joined	13:53
deskin	or can be done with git remote, or git config if you prefer	13:53
xenoterracide	do I have to change the origin remote? can I just tell it to use a different remote/branch?	13:54
	as default	13:55
deskin	xenoterracide: yes, but then you're giving it an argument :) origin is the default remote by definition, I don't think there's a way to change that	13:56
temoto	Is it possible to commit symlink?	13:56
j_engelh	it is possible to commit murder too.	13:56
xenoterracide	temoto: yes	13:57
temoto	xenoterracide: thank you.	13:57
xenoterracide	deskin: hmm... that sucks	13:57
dato	j_engelh: well, having 3 now, moving to 7 + bolds would be nice. do you think a patch for that would be accepted?	13:57
j_engelh	patch? what? where?	13:58
	what has 3?	13:58
dato	sorry, previous conversation	13:59
deskin	xenoterracide: it's just a name; if you like your origin remote, change its name and create a new origin. If your origin really should be origin, well, then make another remote, and when you want to pull from it, give its name explicitly	13:59
dato	3 == %C{red,green,blue}	13:59
j_engelh	ah git log --pretty..	13:59
	→ CodeOfficer joined	13:59
dato	yes	13:59
j_engelh	that's the worst place to have color in the first place	14:00
	dato: you can easily use format:'^[[31m%an' to get red...	14:01
dato	well, I didn't put it there in the firt place	14:03
	and since I suspect the existing ones are not going to be removed...	14:04
	→ faithfulgeek joined	14:05
xenoterracide	deskin right, just a complication with a patch to a piece of software that can call git, but it's only capable of doing git pull	14:07
deskin	xenoterracide: it must be too early, because I've been misleading you; actually one can have git pull by default use different remotes depending on which branch is checked out; look at 'Default Behaviour' section of man git-pull	14:09
Gitbot	xenoterracide: the git-pull manpage can be found at http://git.or.cz/man/git-pull	14:09
deskin	basically you set git config branch.<branchname>.remote myremote	14:09
	and when <branchname> is checked out, git pull uses myremote by default	14:10
	→ mw joined	14:11
xenoterracide	deskin see that syntax sounds confusing... I'm getting the impression that's not exactly what I have to put in my .git/config	14:13
	→ oudeis joined	14:13
	← d0k left	14:13
deskin	well, there's more to it (which I elided because I don't know precisely, and don't want to read the docs now); git checkout -t -b mybranch remotes/myremote/remotebranch will set up the config for pulling from myremote, and merging remotebranch, perhaps you can use that as a guide to get started	14:15
	← scsiraider left	14:15
xenoterracide	ok	14:15
	← markelikalderon left	14:23
	← pie left	14:23
	← Shilin2 left	14:24
	← quoin left	14:28
	→ markelikalderon joined	14:29
	← Aides left	14:30
	→ Aides joined	14:31
	→ muthu_ joined	14:32
	→ rubydiamond joined	14:34
	← unreal left	14:34
	← markelikalderon left	14:37
	← faithfulgeek left	14:37
	→ LiamH joined	14:38
	→ parasti joined	14:43
	→ __doc__ joined	14:43
	→ bosie joined	14:49
	→ foutrelis joined	14:51
	→ chris2 joined	14:52
	→ gasbag joined	14:53
Ilari	dmq: 'git ls-files \| grep -v /'? Since filenames can't have '/' in them...	14:53
	→ kukks joined	14:53
dmq	very convenient.	14:54
	→ Tr3B_ joined	14:54
Ilari	dmq: Well, except that they might have '\n' in them...	14:54
dmq	as in not very conenient.	14:54
	and difficult to use in a portable manner.	14:54
	← Tr3B left	14:54
	→ unreal joined	14:55
Ilari	dmq: Not very poritable in what sense? You thinking msysgit or something? Or is '-v' some nonstandard extension?	14:55
dmq	requiring grep can be problematic.	14:56
	getting the pattern for grep right can be problematic.	14:56
	← greyface left	14:56
	→ ankit9 joined	14:57
dmq	and yeah, i was thinking windows/linux	14:57
	← gasbag left	14:58
dmq	i mean, obviously its surmountable, but it would be easier if there was a --no-recurse option. especially if that means doing less work overall.	14:58
Ilari	dmq: Git requires grep anyway.	14:58
dmq	still.	14:58
	← foutrelis left	14:59
Ilari	dmq: It doesn't mean that much less work... Trees are hierarchial. The index is not.	14:59
	dmq: Also, not just filenames can't have '/' in them, any sort of recursive path MUST have '/' in it.	15:00
bremner	so if I wedged HEAD, and now git things it isn't a repo anymore, should I just edit .git/HEAD	15:01
Ilari	bremner: Yup. What it contains now?	15:01
	← pygi left	15:01
dmq	ilari: well, ok. but, well, i still a think a --no-recurse option makes sense.	15:02
bremner	Ilari: ref: checkout	15:02
	bremner observes one should type carefully when using git-symbolic-ref	15:03
Ilari	bremner: Huh... If its not detached, it should contain "ref: refs/heads/foo'...	15:03
bremner	Ilari: I was trying to make a detached worktree, but I think I made a typo	15:05
	bremner: now I guess I want reset --hard to restore sanity	15:06
	→ pygi joined	15:06
bremner	footgun back in the holster. Is there a nice description of using detached worktrees somewhere?	15:08
	→ rubydiam_ joined	15:10
	← rubydiamond left	15:10
	→ Acry_ler joined	15:10
	→ jm joined	15:12
	→ foutrelis joined	15:15
	← Acryler left	15:15
Ilari	bremner: You mean Detached HEAD or creating two working directories for the same repository?	15:16
bremner	Ilari: two working directories for same repo (second will be read only)	15:16
Ilari	bremner: It doesn't have to be read only (just don't check out the same branch twice)...	15:17
bremner	Ilari: ahh, there was my mistake	15:17
	→ ph^ joined	15:18
bremner	bremner: I can make a temporary branch I guess	15:18
	rubydiam_ → rubydiamond	15:18
	→ marvil07 joined	15:21
bremner	'git --git-dir=$PWD/.git --work-tree=/tmp/foo checkout tmp' doesn't work by itself because everything is deleted. Do I need git-read-tree first?	15:23
	or git archive to get the files?	15:23
	(/tmp/foo is empty to start)	15:24
	→ bobmcw joined	15:24
	← Illocution left	15:24
	← bobmcw left	15:25
	→ Illocution joined	15:25
	→ bobmcw joined	15:25
bremner	bremner, ok git reset seems to do the trick. sorry for the noise	15:26
	← rubydiamond left	15:29
	→ bryanl joined	15:30
	→ earcar joined	15:32
bremner	one more question (for now :-). Is there a good way to find where the .git dir is in a normal worktree? I guess git does this all the time.	15:33
	→ greyface joined	15:34
deskin	bremner: git rev-parse --git-dir, as long as you're not in a subdir of .git itself	15:34
bremner	deskin, thanks	15:35
	← Khisanth left	15:35
deskin	--show	15:35
	er --show-cdup is similar, in that it gives the argument to cd which will put you at the top of the worktree	15:35
Ilari	bremner: There's script for making new working trees. And it makes new repo which has some stuff as symlinks (refs/, logs/ and objects/) and some as seperate stuff.	15:38
bremner	Ilari: sounds useful. Actually, I figured out I can use git-archive, then set GIT_DIR the topgit export operation. That seems cleaner.	15:39
	← __doc__ left	15:41
bremner	deskin: ref-parse --git-dir seems slightly inconsistent, since it gives absolute paths unless in the root of the worktree	15:41
	s/ref-parse/rev-parse	15:42
	→ aoper_ joined	15:42
	← aoper left	15:42
	→ danten joined	15:43
bremner	deskin, anyway cdup is certainly usable, thanks again	15:44
	→ gebi_ joined	15:45
	← bosie left	15:48
	← gebi_ left	15:50
	→ gebi_ joined	15:50
	← bobmcw left	15:52
	← gebi left	15:56
	← lresende left	15:56
	← temoto left	16:01
	← Tr3B_ left	16:03
	← tanek__ left	16:07
	→ tanek_ joined	16:09
	← charon left	16:12
	→ metaskills joined	16:14
	→ brennandunn joined	16:15
metaskills	Whoa... first time visiter to this channel. Lot of people :) Quick question if anyone can help.	16:15
jast	there's always a good chance someone can	16:15
metaskills	I am looking to share a topic branch that I'm pushing to a local server both me and a co-worker have access too.	16:16
	This topic branch is very divergent from the master and every now and then when I rebase it I have to alter some past commits in that topic branch to remove a file, where it was previously edited.	16:17
	The problem is when I go to push that topic branch up to the server I am sharing it over, I get [rejected] messages.	16:17
	← Simetrical left	16:18
jast	the problem is that rebased branches and pull don't like each other much	16:18
	you can force the push (e.g. using -f) but the next pull by the other guy will be icky, and so on for every rebase either of you does	16:18
	same for pull --rebase, I guess	16:19
metaskills	That makes sense... let me try the -f option.	16:19
	That did work... I	16:20
	→ Simetrical joined	16:20
metaskills	http://pastie.org/346999	16:21
	Is this a no win scenerio?	16:21
	← mndoci_ left	16:21
jast	collaborating on rebased branches always involves a fair amount of manual work	16:21
metaskills	That sounds fair.	16:22
jast	probably the easiest way to work with rebased branches is to create a series of patches from your local unpublished commits in that branch	16:22
	← eMBee left	16:22
mnemoc	metaskills: rebase to the remote branch before pushing	16:22
	gebi_ → gebi	16:22
	→ eMBee joined	16:22
jast	and whenever you get an update from the other side, you just reset the topic branch to whatever went upstream and reapply those patches	16:22
	once you start rewriting older commits using pull --rebase or similar might be icky too	16:23
	it's worth a try, i suppose. you can undo pretty much anything anyway (if you know how).	16:23
metaskills	This is great info. I never knew you could rebase to a remote branch. Likewise the --rebase option to pull.	16:25
	← peritus_ left	16:25
jast	you can rebase against pretty much anything (though it doesn't always make sense, of course)	16:25
	← thiago left	16:26
	→ nice_burger joined	16:26
	→ stephens joined	16:28
	← jmatthews left	16:30
	→ markkalderon joined	16:33
	→ QaDeS_ joined	16:33
	← earcar left	16:33
	→ ckm joined	16:34
	→ jmatthews joined	16:36
Sonderblade	when using git-svn, is there a way to pull from the svn repository so that all new changes from it comes before all local branches?	16:37
j_engelh	eh?	16:38
	← markkalderon left	16:38
nice_burger	really doubt it	16:38
j_engelh	sounds like trying to solve a nonexistent problem	16:38
Sonderblade	well i have x feature branches all based on the svn master	16:39
nice_burger	why would the order matter though	16:39
j_engelh	that's my word	16:39
Sonderblade	when someone commits to the svn repos, i want those commits to go before my local ones, so that my feature branches will still apply cleanly to svn trunk	16:39
metaskills	@jast and @mnemoc Was just chatting with my co-worker. It seems this is not so bad. I'll champion the rebase of our TB to master after we both reach certain milestones. If there is a conflict that causes me to alter a commit in the shared TB, then I'll use the -f option and he will delete and retrack the TB. Thanks for the help!!!	16:40
jast	sounds good. have fun. :)	16:40
metaskills	Thanks.	16:41
	← bryanl left	16:41
	← raz left	16:41
	← omakase left	16:42
	← EgS left	16:42
	← Sput left	16:42
metaskills	Sonderblad: You should be working in a topic branch.	16:43
	→ EgS joined	16:43
	→ Sput joined	16:43
Sonderblade	metaskills: i am, thats what i meant with feature branch	16:43
metaskills	And just rebase from master. That way when you are ready to move those commits from your topic branch it will replay them all on top of master... they are ordered correctly.	16:44
	Oh... I might not understand the issue then.	16:44
	Because when I work in that flow, my TB commits are always after svn trunk ones.	16:45
Sonderblade	it is hard to explain	16:46
	← rtomayko_ left	16:46
	← dmq left	16:47
parasti	Sonderblade: how are you updating your git-svn repo? the only way I know of is by using "git svn rebase", and it does exactly what you're asking for	16:49
	→ peritus_ joined	16:49
Sonderblade	my repo tree looks like this: first master then 10 random commits, then head A. on top of head A there are 3 head branches B, C, D.	16:51
	→ akitada joined	16:51
Sonderblade	when i stand in head A and type "git svn rebase" new changes from svn is placed after head A	16:51
	which is not what i wanted. i wanted all new svn changes to be placed before the first commit to the head A branch	16:52
	→ Yuuhi joined	16:52
da1l6	hi	16:53
metaskills	Sounds like your using git-svn to sync to more than one svn branch? If so that is far above what I do.	16:54
da1l6	Tried to use git filter-branch to replace some invalid email addreses in older commit messages.	16:54
	But it keeps on frying the repository.	16:54
	Even simplifying a manpage example to the point where it essentially should do nothing has this problem:	16:54
	→ bobmcw joined	16:55
da1l6	$ git filter-branch --commit-filter `git commit-tree "$@";` HEAD	16:55
	Outputs lots of usage help messages and end with:	16:55
	Ref 'refs/heads/master' was deleted	16:55
	fatal: Not a valid object name HEAD	16:55
	any hints what i am doing wrong?	16:55
Sonderblade	metaskills: nope	16:55
	← wwwald left	16:56
da1l6	git version 1.5.6.3	16:56
	→ aspotashev joined	17:01
	→ nothinghappens joined	17:02
	← bobmcw left	17:03
parasti	Sonderblade: git svn rebase works by taking all the commits since the last SVN commit, fetching new commits from SVN, then applying the non-SVN commits on top of the new commits	17:03
	→ RayMorgan joined	17:03
parasti	Sonderblade: if that's not what happens in your case, either it's a bug or there's something you're not telling us	17:03
	→ bobmcw joined	17:04
Sonderblade	parasti: does the "non-SVN commits" include all the local commits in all local branches?	17:04
parasti	Sonderblade: only the branch you are on when running git svn rebase	17:04
Sonderblade	parasti: ah so that is why it messes up my repository and creates duplicate commits	17:06
	→ dasil003 joined	17:06
parasti	Sonderblade: when you work with topic branches, usually you git svn rebase the main branch (like master), and then rebase each topic branch against master	17:06
	err, against the main branch	17:07
	→ josephwilk joined	17:07
Sonderblade	parasti: that's what i'm doing. it just seems unoptimal	17:07
	← bobmcw left	17:09
	→ icwiener joined	17:09
	→ cgardner joined	17:11
	← josephwilk left	17:12
	→ bobmcw joined	17:17
	→ rubydiamond joined	17:18
	← metaskills left	17:20
	← __david left	17:21
	← dasil003 left	17:21
	QaDeS_ → QaDeS	17:24
	→ ceej joined	17:24
	→ ceej_ joined	17:26
	← brennandunn left	17:27
	← jepler left	17:28
	← rubydiamond left	17:30
	← akitada left	17:31
	→ RayMorgan_ joined	17:34
	→ jackdempsey joined	17:35
	→ bdiego joined	17:37
	← ckm left	17:38
	→ ashH joined	17:39
	← RayMorgan left	17:40
	→ hobodave joined	17:45
	→ litage joined	17:46
	← oudeis left	17:48
	→ oudeis joined	17:49
	← ceej left	17:49
	← aoper_ left	17:52
	→ aoper joined	17:53
	← Simetrical left	17:53
	ceej_ → ceej	17:54
	← aoper left	17:54
	→ aoper joined	17:55
	← muthu_ left	17:55
	→ omakase joined	17:59
	→ Simetrical joined	18:00
	← foutrelis left	18:04
	→ pedro_mg joined	18:06
	← unreal left	18:10
	→ unreal joined	18:13
	→ d0k joined	18:18
	← QaDeS left	18:18
	→ davidfetter joined	18:21
	→ abbe joined	18:22
	← peritus_ left	18:23
	→ octernion joined	18:23
octernion	hi, is there an easy way to apply all the commits from one branch in a specific directory to another branch?	18:23
mnemoc	octernion: add it as a remote and cherry-pick the range (or rebase)	18:27
bremner	mnemoc: "in a specific directory" ?	18:28
octernion	mnemoc: oh, i didn't know i could add arbitrary directories as remotes?	18:28
jast	arbitrary repositories, to be precise	18:28
	← jackdempsey left	18:29
mnemoc	I assume it's another working directory (with it's corresponding .git)	18:29
bremner	octernion: the only way I know off hand uses git-filter-branch and git-format-patch.	18:29
	→ mndoci joined	18:29
bremner	octernion: I assume you just want changes to a subdir?	18:29
octernion	well, the directory itself doesn't have a .git	18:29
aspotashev	he probably meant "parts of commits" regarding particular subdirectories in the repository	18:29
octernion	bremner: yes, exactly	18:29
	aspotashev: yes, sorry for not being clear	18:29
mnemoc	ic	18:29
octernion	i will take a look at filter-branch and format-patch	18:30
mnemoc	if it's not a working directory, he doesn't have commits to filter	18:30
bremner	octernion: basically use (I think) subtree filter in a clone of the repo to get rid of everything else, format-patch to make patches, and git am to apply them	18:31
	less messy suggestions welcome	18:31
octernion	bremner: dang, that sounds pretty complicated :)	18:31
	→ patrikf joined	18:31
octernion	the problem arose because i was very stupid and thought i was working on master, when i was in fact working on a branch, and i want the specific changes (all in one subdirectory) to be available on master	18:32
bremner	octernion: if the changes really only touch that directory, you can use rebase	18:33
	→ lresende joined	18:33
octernion	bremner: yep, they are all there. how do i do that for the subdirectory?	18:33
	← lresende left	18:33
octernion	(sorry, i've never actually used rebase :/)	18:33
	→ etienne9 joined	18:34
bremner	octernion: basically git checkout master && git rebase other-branch. see man git-rebase for some examples	18:35
Gitbot	octernion: the git-rebase manpage can be found at http://git.or.cz/man/git-rebase	18:35
	→ QaDeS joined	18:35
octernion	ok. the manpage is pretty helpful for that, i just didn't see anything pertaining to subdirectories	18:36
	← ankit9 left	18:36
Tv	octernion: i don't see how subdirectories make any difference, in this case	18:37
bremner	octernion: you have to identify the commits somehow (git log is probably enough)	18:37
octernion	bremner: ahh, ok, i gotcha. thanks very much!	18:37
	→ lolage0 joined	18:38
parasti	bremner: that rebase example seems strange. what would it do? reset master to other-branch, then apply other-branch..master commits on top of it (basically a no-op)?	18:39
bremner	parasti: well, in the optimistic case that other-branch was branched from master, it should bring the other commits back. But yeah, in general more is needed	18:40
	→ markelikalderon joined	18:41
bremner	parasti: or maybe I am confused here.	18:42
parasti	bremner: I think the short version of that command sequence is git reset --hard other-branch :P	18:42
	← aspotashev left	18:44
bremner	parasti: ok, the goal is git checkout other-branch && git format-patch since && git-checkout master && git am 000*	18:44
	→ aspotashev joined	18:45
	← offby1 left	18:45
bremner	Can't rebase do that? I guess --onto is the trick, see the example in the man page	18:46
	→ faithfulgeek joined	18:47
parasti	bremner: yeah, that might work	18:47
	← faithfulgeek left	18:47
	→ faithfulgeek joined	18:48
bremner	octernion: you follow that?	18:48
octernion	bremner: hmm, nope. i ended up just formatting the patch like you suggested earlier	18:48
bremner	octernion: OK :-)	18:48
octernion	it really wasn't that bad, since i realized it was only one commit :)	18:48
bremner	octernion: ahh, next time git-cherry-pick	18:49
octernion	but thanks for the comprehensive answer; i knew git would be able to do it	18:49
	bremner: yeah, actually i just saw it after i formatted it :(	18:49
	er, by it i mean i saw git-cherry-pick	18:49
parasti	"git log --pretty=email --reverse -p ..other-branch -- $subdir \| git am	18:49
	" might do the trick	18:49
octernion	parasti: wow, nifty. thanks :)	18:50
	← faithfulgeek left	18:50
	← bobmcw left	18:53
	→ Mo joined	18:53
Mo	hi.how are you	18:55
j_engelh	better than you :p	18:55
	→ dreiss joined	18:57
mnemoc	woh, long time without seeing someone on irc with such a possive view of his life	18:58
	← Mo left	18:59
	→ Jacolyte joined	19:00
	← tomsw left	19:03
	→ markoa joined	19:05
	→ name joined	19:07
	→ jackdempsey joined	19:07
	→ offby1 joined	19:09
	← name left	19:10
	→ perezd88 joined	19:12
	← octernion left	19:14
	← ceej left	19:16
	→ jbellis joined	19:20
	← jbellis left	19:21
	→ earcar joined	19:21
	→ bosie joined	19:21
	← aoper left	19:22
	← earcar left	19:24
	→ jschoolcraft joined	19:25
jschoolcraft	can you git clone a specific branch or do you have to clone the whole repo and checkout the branch in question?	19:25
offby1	excellent question	19:26
	dunno	19:26
	my guess is: yes, but you don't use "git clone"	19:26
	← QaDeS left	19:27
Tv	jschoolcraft: you could init, remote add, fetch just the branch you're interested in, checkout -b	19:27
offby1	instead I'd do "git init" and then "git remote add", passing a refspec that includes only that branch	19:27
Tv	jschoolcraft: most likely that isn't worth the trouble	19:27
offby1	depends: if the source repository has a bunch of unrelated stuff in it, it might save a lot of download time	19:27
Tv	jschoolcraft: and if you weren't interested in minimizing download, just affecting what branch is checked out after clone, clone just checks out the remote HEAD, no way to affect that	19:27
	offby1: and i'm guessing 90% of the time typing in all that is slower than the download ;)	19:28
offby1	could be	19:28
Tv	jschoolcraft: git clone foo bar && cd bar && git checkout -b foo origin/foo	19:29
jschoolcraft	Tv: offby1: thanks	19:29
Tv	(maybe with less collisions of meaning of foo)	19:29
	← greyface left	19:29
j_engelh	offby1: -> offbydefault	19:30
	→ rubydiamond joined	19:30
	→ cemerick joined	19:36
offby1	dude, don't be hatin' on my nick	19:41
	← metajack left	19:42
	→ metajack joined	19:42
davidfetter	happy boxing day, Tv :)	19:43
	← markoa left	19:43
	→ k776 joined	19:45
	← ph^ left	19:48
	← marvil07 left	19:48
	→ kumbayo joined	19:48
	→ ph^ joined	19:49
	→ charon joined	19:49
	→ sgrimm joined	19:54
	→ faithfulgeek joined	19:55
	→ dwmw2_gone joined	20:00
	← jschoolcraft left	20:01
	→ leveille joined	20:04
	← iulian left	20:09
	← k776 left	20:18
	← Tsion left	20:20
	→ jstad joined	20:23
	← cemerick left	20:25
	→ advany joined	20:27
	→ lresende joined	20:28
	→ Eulex joined	20:28
Eulex	hello	20:28
	offby1 glances around nervously.	20:28
me22	uhoh	20:29
Eulex	is there a way to properly clone a git-svn repo (so that I can commit) other than copying the files themselves?	20:29
	← LiamH left	20:30
Tv	Eulex: cloning it and git svn rebuild used to work, but that command doesn't seem to exist anymore	20:31
charon	Eulex: run the same 'git svn init' command as for the original repo (or copy the relevant bits of .git/config), copy over the refs used by git-svn (possibly 'git-svn' or 'trunk'), then run 'git svn fetch' to rebuild its cache	20:32
	← scientes left	20:32
	→ scientes joined	20:32
offby1	charon: will that require re-downloading all the svn commits?	20:33
Tv	offby1: at least with older versions that was a "no"	20:33
charon	offby1: no, git-svn will just regenerate its internal cache from the git-svn-id: lines in the commit messages	20:33
offby1	good.	20:33
charon	of course that won't work if you said --no-metadata, but you obviously shouldn't.	20:33
offby1	I too have thought about this	20:33
Tv	git svn changes incompatibly often enough that it's kinda hard to know how it behaves	20:33
Eulex	charon: let's try that	20:36
charon	note that some tweaking of the fetch lines in .git/config may make this task much easier if you intend to run a central git-svn mirror that the other repos can fetch from	20:38
	← perezd88 left	20:38
	→ ashleyw joined	20:40
Eulex	charon: interesting, which would those be?	20:43
offby1	yeah, do tell	20:46
Tv	probably make git-svn write refs outside the remotes/ namespace, IIRC it does that by default	20:47
	(and that default makes perfect sense, they are refs controlled by outsiders, you shouldn't checkout & commit to them, etc..)	20:47
charon	that, or fetch them in the clone	20:50
Tv	well that's a per-clone config change	20:51
charon	true, but it'll put them in the place where git-svn expects it	20:51
	e.g. you could add a line to the remote saying 'fetch = +refs/remotes/git-svn:refs/remotes/git-svn' (if you use a branch-less git-svn setup)	20:51
Tv	yeah.. git-svn isn't really designed to share the gittifying work	20:51
offby1	that sounds awfully sensible	20:52
	wonder why I didn't think of that	20:52
	→ bobesponja joined	20:54
jast	there's an example in the git-svn manpage for the centralized git-svn mirror scenario	20:54
	→ drizzd joined	20:54
	← leveille left	20:56
offby1	there is?!	20:57
	offby1 never noticed that	20:57
Eulex	jast: taking a look at that	20:58
offby1	Oh yeah: "If multiple people (or one person with multiple machines) want to use ..."	20:58
	← metajack left	20:58
jast	yeah, that	20:58
	recently fixed so it actually works	20:59
offby1	never even thought to try that.	20:59
	offby1 administers self-dope-slap	20:59
	← chalky left	20:59
	→ metajack joined	20:59
offby1	The question is: can those different people share patches without committing them to svn?	21:00
jast	that's where it gets hairy	21:00
offby1	I managed somehow. Not sure what I did.	21:00
jast	it's not that bad if you have a single integrator	21:01
offby1	actually, in my case, "those different people" were all me, just working on different machines.	21:01
	← icwiener left	21:01
offby1	My company uses svn, but I prefer git ... and I sometimes do my work on my work box, and sometimes on my laptop at home.	21:01
jast	if the patches in question don't change during the rebase (and neither does their order), it generally works	21:02
	→ chipdude joined	21:03
	→ bobesponja_ joined	21:06
offby1	I suspect I fell into that category	21:06
	← exg left	21:07
	← oudeis left	21:07
Eulex	jast: I suspect I've got such instructions that don't work; I never get any contents in the directory itself, and the last command to run, git-svn rebase, fails with "fatal: ambiguous argument 'HEAD': unknown revision or path not in the working tree."	21:07
	← bdiego left	21:07
Eulex	so where do I find the new ones? :)	21:08
	nvm, found it	21:09
	← bobesponja_ left	21:11
	→ dduncan joined	21:12
	← jstad left	21:12
	← bobesponja left	21:13
jast	great :)	21:13
offby1	I love questions that answer themselves	21:14
	← dduncan left	21:14
	← pygi left	21:15
	← abbe left	21:15
	→ clairvy joined	21:18
	→ altfive joined	21:19
	→ ijcd joined	21:19
Eulex	now I've got a directory full of files, and all three branches (master, git-svn and git-svn@1081), I ran git-svn init to set the right upstream svn repo, but when I run git-svn rebase I get "Unable to determine upstream SVN information from working tree history"	21:19
	→ exg joined	21:20
	→ name joined	21:20
	→ rtomayko joined	21:21
	→ pygi joined	21:25
	→ spearce joined	21:26
	→ marvil07 joined	21:26
	→ clairvy__ joined	21:28
	→ oudeis joined	21:31
	→ Tr3B joined	21:32
	← cgardner left	21:32
	→ cemerick joined	21:32
	← faithfulgeek left	21:36
Sonderblade	i have a gray "stash" branch, how do i reach it?	21:37
	→ DrNick2 joined	21:37
j_engelh	gray!?	21:37
	it's either blue or it's your terminal color	21:37
	← DrNick left	21:38
	DrNick2 → DrNick	21:38
drizzd	in gitk it's grey I believe	21:38
	← Voker57 left	21:39
j_engelh	absolutely not.	21:39
	For believing, please visit Church.	21:39
vmiklos	:)	21:39
drizzd	j_engelh: the "stash" label in gitk is grey (the shade between black and white)	21:40
Sonderblade	whatever	21:41
	i still need to reach it	21:41
	← cemerick left	21:41
doener	it's not a branch (like master), it's a ref referencing your top-most stash entry	21:41
drizzd	Sonderblade: the stash is not really a branch, see "git stash list"	21:42
doener	and the stash is accessed using "git stash"	21:42
Sonderblade	is it deletable? i dont want it anymore	21:42
doener	git help stash	21:43
	← clairvy left	21:44
j_engelh	your display must really suck then	21:44
	← mndoci left	21:44
j_engelh	stash tag is not gray: http://tinyurl.com/a5wpea	21:45
drizzd	j_engelh: what color do you think it is? your picture still looks grey to me	21:46
Sonderblade	me too	21:47
	it is your display that suck!	21:47
j_engelh	drizzd: pale blue. aka #8080FF	21:47
	mixed by using 1/2 #0000FF + 1/2 #FFFFFF	21:48
	← ashleyw left	21:49
doener	hm, gimp tells me that it's #ddddff	21:49
j_engelh	that's still not gray	21:50
	(albeit a bit more of the 'pale' in 'pale blue')	21:50
doener	I didn't say it's grey, just that it's #ddddff, not #8080ff	21:50
j_engelh	point taken	21:51
	→ cemerick joined	21:52
	→ mndoci joined	21:55
	← cemerick left	21:57
	→ knobo joined	21:57
Sonderblade	i have a TB on top of my local branch which is branched from master. is there a way to diff TB against master so that the changes in the local branch is somehow excluded?	21:59
gitster	rebase TB on top of master?	22:00
Sonderblade	then i will get lots of conflicts to resolve	22:01
doener	hm, do you really want to diff against master, or do you want a diff of the changes that "TB" introduced?	22:01
	for the latter, "git diff local...TB" should do	22:01
	that's a diff of TB against the merge base of TB and "local"	22:02
	← offby1 left	22:03
	→ joshdavey joined	22:03
doener	short of a better description: a cumulative diff of the commits that are in TB but not in local	22:03
Sonderblade	i want the changes TB would have introduced had it been based on master	22:05
	→ gottesmm joined	22:05
Sonderblade	the TB to master diff - the local to mater diff	22:06
drizzd	git diff master..TB	22:06
Sonderblade	drizzd: no, that includes changesets from local	22:07
drizzd	Sonderblade: if git rebase cannot rebase without conflicts, how do you think git diff should be able to find the solution?	22:09
Sonderblade	by subtracting the local branch changesets	22:12
	mw → mw\|out	22:12
	← scientes left	22:13
	→ perezd88 joined	22:14
	← perezd88 left	22:14
	→ alikins_ joined	22:17
	← spearce left	22:18
	→ greyface joined	22:18
	← alikins left	22:18
	← mw\|out left	22:19
	→ mw\|out joined	22:20
	← pedro_mg left	22:23
	← nice_burger left	22:24
	← davidfetter left	22:26
knobo	If someone pushes changes to my repository, I can see the changes with git-log, but how do I update my branch(current files) with the canges?	22:27
	← Acry_ler left	22:29
knobo	git-fetch says: fatal: 'origin': unable to chdir or not a git archive	22:29
Eulex	you could fetch directly from the url to the repo	22:29
	or set origin correctly, then retry	22:30
	(git remote add origin blahblah)	22:30
knobo	so git fetch ./git should do it?	22:30
gitster	knobo: is this about the "don't push into a non-bare repository" faq?	22:30
	→ ari-_-e joined	22:31
knobo	gitster: I don't know. I'll have to check.	22:31
	→ gasbag joined	22:31
Eulex	uh, so someone pushed into your repository on your computer?	22:31
gitster	you earlier said "someone pushes changes to my repository".	22:31
ari-_-e	how might I go about completely removing a couple of commits from my history, i.e. without using git revert?	22:31
	and no, they're not the two most current ones	22:31
Eulex	or to the repository on a server?	22:31
parasti	ari-_-e: git rebase -i $commitbeforeallofthem	22:32
gitster	that is not exact enough description. "to your repository's _which_ branch" and "if that branch is the one you have checked out and is working on" are necessary to answer your question.	22:32
parasti	ari-_-e: it might not work if you have merges after that commit, have a look at the manpage	22:33
ari-_-e	parasti: no merges, this is a very simple linear history	22:33
gitster	If the answers are "pushed to the branch I have checked out", and "no, luckily I do not have any local changes in my working tree", then the answer is "git reset --hard". But look for "mothership/satellite" to find a better arrangement to allow pushing into a non-bare repository in the FAQ.	22:34
ari-_-e	parasti: NICE, I tried using git-rebase, but I was missing the -i	22:34
	worked like a charm, thanks	22:34
parasti	ari-_-e: you've been missing out. ;) have fun	22:35
knobo	gitster: it's the "Why won't I see changes in the remote repo after "git push"?" faq.	22:35
	So, I'll just read the faq.	22:35
Eulex	knobo: as I said, it appears that you don't have an origin set.	22:36
	gitster has given answers in advance ;-)	22:36
	← joshdavey left	22:40
	← altfive left	22:40
	→ __doc__ joined	22:41
	← gasbag left	22:42
knobo	Eulex: it appears that I did not read the faq ;)	22:43
Eulex	:)	22:44
	← danten left	22:45
ari-_-e	so if I have some random string, how would I go about finding out if it's the name of a branch?	22:45
Eulex	well, is i listed in git branch? :)	22:46
ari-_-e	right, but programmatically	22:47
	I assume git branch\|grep stuff isn't the right way to do it :-/	22:47
Eulex	you could \|grep "^[* ] stuff$" I guess	22:49
	→ git_ joined	22:51
git_	hi, anyone know where i can get a static binary of git?	22:51
	← rtomayko left	22:52
	→ faithfulgeek joined	22:52
drizzd	ari-_-e: git show-ref --heads <branch> (no output if <branch> is not a branch)	22:52
ari-_-e	drizzd: cool, thanks	22:52
drizzd	git_: your preferred distribution probably	22:53
	← gottesmm left	22:54
	→ cemerick joined	22:56
	→ gottesmm joined	22:56
	← cemerick left	22:57
	← marvil07 left	22:58
	→ akitada joined	23:00
	← Ademan left	23:09
	→ peritus_ joined	23:10
	← j_engelh left	23:10
	← daonb left	23:15
	← gottesmm left	23:17
	→ Turl1 joined	23:20
Turl1	hello1	23:20
	← kukks left	23:20
Turl1	how can I revert any changes to my git clone and leave it as it is in the server?	23:20
	← selckin left	23:20
	→ spearce joined	23:24
	← kumbayo left	23:25
	← peritus_ left	23:26
	→ spearce` joined	23:27
	← spearce left	23:29
	← hobodave left	23:32
	← bosie left	23:33
	→ j_engelh joined	23:33
	← j_engelh left	23:33
charon	Turl1: depends what kind of changes you made. if you are still on the 'master' branch (check 'git branch') and the 'origin' remote is still there, you can say 'git reset --hard origin/master'. obviously that's a rather destructive operation.	23:35
	← ph^ left	23:37
	← markelikalderon left	23:37
	→ markelikalderon joined	23:38
	← markelikalderon left	23:38
	→ lholden joined	23:40
	← Tr3B left	23:45
	← Turl1 left	23:46
	→ hobodave joined	23:46
	→ k776 joined	23:52
	← mbroeker left	23:58
	← hobodave left	23:59

Logs Search ←Prev date Next date→

Channels Documentation