IRCloggy #git 2008-06-30

Provider of IRC logs since 2005.

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

	2008-06-30
	← ajonat left	00:01
	→ CrnkyGee1 joined	00:03
	← CrnkyGeek left	00:03
	← DrNick left	00:05
	→ DrNick joined	00:06
	← d0k_ left	00:08
	→ elight joined	00:09
gitte	vmiklos: hey, sorry... GF.	00:12
	vmiklos: I did not have time to read Junio's answer...	00:12
	vmiklos: but maybe I still can help?	00:12
davidfetter	GFs take precedence	00:14
vmiklos	exactly :)	00:14
	especially on weekends.	00:14
davidfetter	my favorite toast:	00:14
vmiklos	gitte: i just wanted to ask if it's better to reply off-list (to you) for Junio's mail to make his load lower.	00:14
davidfetter	To wives and sweethearts...	00:14
	may they never meet ;)	00:14
gitte	vmiklos: no, I think by answering he asked for some response on your part.	00:15
vmiklos	ok :)	00:15
	→ mordaunt joined	00:15
	← scroogemcduck left	00:16
	← Eridius_ left	00:16
	→ evanlight_ joined	00:18
mordaunt	is there a way to get git to only track some file extensions? i.e. instead explicitly ignoring stuff. explicitly include stuff and ignore eveyrthing else?	00:18
gitte	Sure.	00:18
mordaunt	how?	00:18
	o_O	00:18
gitte	There are rules to exclude certain patterns with "!" in .gitignore.	00:18
	Also, AFAICR later patterns take precedence over earlier ones.	00:19
	So, "echo '' > .gitignore; echo '!.blub' >> .gitignore" should work.	00:19
mordaunt	hmm I'll try it!	00:22
	mordaunt hops in the gitmobile and drives off	00:22
vmiklos	gitte: i just noticed i can avoid a work() when calling rerere, there is already a rerere() in commit.h ;)	00:22
gitte	Heh.	00:23
vmiklos	fork()*	00:23
gitte	(Yes, I guessed so much ;-))	00:23
	← mordaunt left	00:23
	← evanlight left	00:24
	← evanlight__ left	00:27
	→ doener joined	00:28
	→ huyx joined	00:29
	→ Ariella joined	00:30
	→ sfullenwider joined	00:31
sfullenwider	anybody remembere where that cool git picture is with the four repositories showing how git workflow works?	00:31
	→ marc_vw joined	00:33
	→ spearce joined	00:35
gitte	Hey gitney!	00:35
spearce	morning	00:35
gitte	Already in California?	00:35
spearce	no. still in ny	00:35
	i move on the 23rd of july	00:36
	blargh. jgit can't fetch junio's git tree from repo.or.cz over git://	00:37
gitte	Hm.	00:37
spearce	"fatal: Packfile corruption detected: unknown compression method"	00:37
	gitte wanted to get involved into jgit for a long time now.	00:37
spearce	never a message you want to see from your git implementation	00:37
gitte	Whic compression method might that be?	00:37
	s/Whic/Which/	00:37
	gitte curses her keyboard.	00:37
	← vbabiy left	00:38
spearce	my guess is we got confused reading the pack at some point (not sure how?) and handed data to zlib that isn't zlib data	00:38
	hah. at least its reproducible right now so i can debug it	00:39
	← doener_ left	00:39
spearce	wtf. a 7 byte annotated tag?	00:41
	that's so not right	00:41
	well, i guess that just goes to show you how good the git pack is at dealing with "corruption". even a perfectly valid pack read under buggy code is likely to claim its corrupt.	00:42
	oooh. "fatal: pack has bad object at offset 194489: inflate returned -3"	00:44
	that's c git.	00:44
	_wtf_	00:44
	← sfullenwider left	00:44
gitte	Where does that pack come from?	00:46
spearce	repo.or.cz's git:// service	00:47
	→ evanlight joined	00:47
spearce	from gitster's alt-git.git tree	00:47
	in theory, none of this should be happening, even to c git	00:47
	← huyx left	00:48
	elight → Guest25428	00:48
	← Guest25428 left	00:48
	evanlight → elight	00:48
	← evanlight_ left	00:48
vmiklos	unless there is a hdd corruption on repo.or.cz..	00:50
	→ huyx joined	00:50
	gitte tries to clone alt-git.	00:50
spearce	it gets better. the SHA-1 signature of the pack matches its footer.	00:50
	and repo.or.cz should be running a new enough git that it validates the data before it writes it into the pack, to avoid bit errors but still have a valid SHA-1 footer.	00:51
	← mvw left	00:51
	gitte does not get any error, neither from the clone nor from fsck.	00:52
	← elight left	00:54
	→ _GuidoKackwurst_ joined	00:54
	← _GuidoKackwurst_ left	00:54
	→ elight joined	00:55
	→ jmspeex joined	00:56
	← jordi_bunster left	00:57
spearce	hmmph.	01:00
	→ jordi_bunster joined	01:00
spearce	c git gets a different object count, and a different pack, when it fetches	01:00
	the c git pack ain't corrupt either	01:01
	gitte sighs	01:01
	gitte sighs at something different	01:01
spearce	i'm sure jgit asks for objects a little different then fetch-pack does during the commit negotiation so we may come up needing a few extra objects than c git wanted, but asking for too much (by not listing enough common commits or something) should not have caused a corrupt pack to be delivered.	01:02
gitte	Right.	01:03
	Because the pack that is delivered is still generated with pack-objects.	01:03
	spearce wonders if the jgit sideband multiplexer f'd up	01:03
spearce	err, de-multiplexer	01:03
	marc_vw → mvw	01:05
	← scrooloose left	01:07
	← volante left	01:11
	← geckosenator left	01:15
	← SuttoL_ left	01:15
	→ albertito joined	01:18
spearce	haha.	01:18
	bug in jgit index-pack.	01:18
	we completed a thin pack by writing a whole object at any old offset. like on top of a delta that wanted the whole object as a base.	01:19
	so then when we re-read the delta, we read garbage. it was the middle of the whole object.	01:19
	→ pombreda1 joined	01:19
	spearce grumbles something about the stupid author of jgit's index pack, and the moron who last touched it.	01:19
	spearce then remembers it was him	01:20
	← Leefmc left	01:21
spearce	i found it. moved code around in a stupid way when i added CRC32 support. blargh.	01:22
	→ SuttoL joined	01:23
gitte	Happens to me all the time. Fix something, break something else.	01:25
	→ bryanray joined	01:26
	→ VVSiz_ joined	01:27
spearce	sheesh. that was a f'ing dumb thing to have done	01:29
	gitte would not be that harsh.	01:29
spearce	i am on myself. :)	01:29
gitte	;-)	01:30
spearce	jgit has too much cooking in pu and not enough merging.	01:30
	on the other hand, the breakage was in pu, so its fixable	01:30
	← drizzd left	01:30
gitte	I have to admit that I was intimidated by the sheer size of the _single_ patches in the recent JGIT patch serieses, let alone the _number_ of patches.	01:31
spearce	yea, we had a couple of very large ones lately	01:31
	both in patch and series	01:31
gitte	Several versions, even.	01:32
	← pombreda left	01:32
spearce	hmm. did i build this bundle wrong (and don't have the base objects it needs), or is there a bug in jgit's unbundle code?	01:33
	sigh	01:33
gitte	Check with "git bundle verify"?	01:34
spearce	says its ok	01:35
gitte	Then it should be okay, and it is jgit's unbundle code...	01:35
	← _VVSiz_ left	01:35
gitte	(Unless it is some really internal corruption)	01:35
spearce	jgit wants 0e948be65d164c8f4cff78417b1138f644364dfc as a delta base for an object, and i don't have that.	01:35
	hmm.	01:35
gitte	Oops.	01:37
	bundle verify does not verify the pack.	01:37
	It only verifies the prerequisites.	01:37
spearce	damnit, then its useless, useless i say! ;-)	01:37
gitte	It is.	01:37
	gitte is sorry	01:37
	gitte forgot to squash that TODO	01:38
gitte	I think it should call verify-pack on the pack in the bundle...	01:38
	But that would involve quite some hackery on add_packed_git(), to allow packs that do not start at the beginning of a file...	01:39
spearce	you can't call verify-pack without an index.	01:39
gitte	Oh, that's right.	01:39
	← gass left	01:39
spearce	running index-pack to create it is just as good as running verify-pack.	01:39
	← kukks left	01:39
gitte	Yeah, but for "git bundle verify" I would not want to write anything.	01:40
	So maybe I'll have to pipe it to "git index-pack -o /dev/null"?	01:40
spearce	c git "git bundle unbundle foo.bundle" ran just fine and produced the pack and its .idx. and that's based on feeding the pack to index-pack, right?	01:40
	gitte is too tired for that.	01:40
spearce	heh. ok	01:40
gitte	Yes, index-pack is called.	01:41
	← proxie left	01:41
gitte	(It has to, since the bundle only contains a pack, not its index)	01:42
spearce	well, it could call unpack-objects. :)	01:42
gitte	Also, the contained pack is typically thin.	01:42
spearce	yea, jgit is barfing in its make-it-non-thin method	01:42
gitte	As I guessed from you saying it looks for a delta base...	01:43
	← SuttoL left	01:43
gitte	But verify-pack -v should give you an idea which delta bases are needed, right?	01:43
spearce	true, yes, it would	01:43
	spearce runs that	01:43
gitte	And your jgit code is probably already instrumented to show you which objects are processed, right?	01:44
	→ blowery joined	01:44
spearce	eh, not so much. :)	01:44
	oh hell.	01:44
	the missing object is in this pack.	01:44
	no wonder its "missing" but is a valid delta base.	01:45
	:)	01:45
gitte	;-)	01:45
	gitte remembers vaguely that there were issues as to the position of delta bases in the pack.	01:45
gitte	Did Nico not say at some stage that they can be later in the pack?	01:45
spearce	yes, they can appear anywhere.	01:45
gitte	IOW that you could not incrementally build the fat pack?	01:46
spearce	unless its an OBJ_OFS_DELTA, which must have the base appear first	01:46
	this is an OBJ_REF_DELTA, so the base can be anywhere.	01:46
	→ jordi_bunste joined	01:46
	← blowery left	01:46
gitte	Ah.	01:47
spearce	we should have code in here to handle this case; but its not getting the delta and resolving it, so we wind up with this unresolved delta at the end and go looking in the repository for the base. only the repository doesn't have it, because its in the pack.	01:47
	we should have paired up this delta to the base much earlier. :-\|	01:47
	vmiklos wonders for what the IMITCNVHO acronym stands for. google gaves 0 results ;)	01:53
spearce	oh, i see it.	01:55
	this is a chain of A-B-C. A references B by name; B references C by name; C is not in the pack and must be appended to the end.	01:56
	only we try to resolve the A-B relationship first, and cannot find B, as the B-C one hasn't been resolved yet.	01:56
doener	vmiklos: it's an extended version of a rather well known acronym :-)	01:56
vmiklos	which is? :)	01:56
doener	vmiklos: try dropping some characters from the middle	01:57
	→ foca joined	01:57
vmiklos	if i should drop 'ITCNV', then i have no idea what 'ITC' means ;)	01:58
	← jordi_bunster left	01:58
doener	I guessed for "in this case"	01:58
vmiklos	ah :)	01:58
	tricky;)	01:58
doener	yeah :-) never seen that before either, but was fun to decipher	01:59
	← jscoot left	01:59
	→ agenteo joined	02:00
agenteo	can I leave uncommited changes in a branch, switch to another branch and working with HEAD? I am pretty sure git checkout master will port my changes there...	02:01
	← wagle left	02:01
doener	agenteo: you can stash your changes, or just leave a temporary commit on your branch and amend that later	02:02
agenteo	doener: if I commit, I can jump to another branch and merge at anytime in the future right? I'll check stash now	02:05
spearce	yay. jgit bundle works. again.	02:05
	→ gcv joined	02:08
	→ sr_ joined	02:10
	→ jrdavid joined	02:10
	→ garrin joined	02:11
gitte	spearce: congratulations!	02:13
spearce	yea, i've really started to eat my own dogfood. and pay the price for it when its gross. :)	02:13
	sigh. too bad jgit doesn't recognize ~/.ssh/config and the host aliasing thing it does	02:16
gitte	doener, vmiklos: congratulations!	02:19
	→ wagle joined	02:21
	← wagle left	02:23
	→ wagle joined	02:23
	→ krawek joined	02:24
	→ Leefmc joined	02:25
	gitte gives up in despair on trying to convince Daniel that his scenario with GIT_CONFIG and clone does not make any sense at all.	02:25
	→ imyousuf joined	02:34
	← flaguy left	02:38
spearce	god i love spam bots just probing for open relays.	02:40
	so much fun in your mail logs	02:40
	← jelmer left	02:45
	→ jelmer joined	02:46
	← gcv left	02:53
	← kanru left	02:58
	← eternaleye left	03:02
	→ daaku joined	03:05
	← lea_w left	03:07
	→ arun joined	03:12
	← catdude left	03:16
	← jordi_bunste left	03:24
	← saintdev left	03:27
	→ Ademan joined	03:29
	→ mtodd_ joined	03:30
	→ kanru joined	03:30
	→ spool joined	03:32
	→ lea_w joined	03:32
	→ vbabiy joined	03:33
	← gitte left	03:33
spool	hi, i'm trying to use 'git stash apply', but I get the error 'Cannot restore on top of a dirty state'	03:34
	← elight left	03:34
DrNick	spool: what's git status say?	03:34
spool	nothing special	03:35
	I've got 1 unadded file	03:36
	and one file I've modified	03:36
Eridius	so you have a dirty tree	03:36
spool	mmm	03:36
	ok	03:36
	so I need to commit my modification?	03:36
	← lea_w left	03:36
spool	can I ignore the unadded file?	03:37
Eridius	$note allan inner_six's working tree isn't on any branch. She also has a locally-modified tree. I recommend committing and rebasing against master. I also recommend updating the version of git	03:37
	spool: I believe so	03:37
	ack, wrong channel for that $note	03:37
spool	Eridius: :) ok I'll try	03:37
	Eridius: :) sweet	03:38
	Eridius: errr... sort of	03:38
thiago	spool: the problem isn't the untracked file	03:39
spool	the reason I hadn't commited that file is it's a blob, so now I've got a merge error :(	03:39
thiago	it's the modified file	03:39
	you can just git add the file	03:39
	then git stash apply will work	03:40
	← peritus__ left	03:40
spool	yeah I did that, that's why I said sweet ;)	03:40
	thiago: git stash then automerged(?) the binary (a sqlite db)	03:41
	thiago: which is a longwinded way of saying thank you for your advice :)	03:42
	Eridius: thank you too	03:42
	← krawek left	03:44
	→ joydivider joined	03:45
	← mtodd left	03:46
	→ foca_ joined	03:49
	← albertito left	03:50
	← igorgue left	03:52
	← madewokherd left	03:56
	→ albertito joined	03:58
	← brosner left	04:00
	← foca left	04:06
	→ brynary joined	04:11
	→ tjafk1 joined	04:12
	← Ademan left	04:12
	→ saintdev joined	04:14
	→ peritus__ joined	04:15
	← adante left	04:25
	← freelock left	04:26
	→ Ademan joined	04:28
	← tjafk2 left	04:29
	→ alenoosh joined	04:31
	→ freelock joined	04:31
	← thiago left	04:32
	← Leefmc left	04:32
	← albertito left	04:34
	→ mtodd joined	04:36
	→ swtaarrs joined	04:39
	← joydivider left	04:40
	→ joydivider_ joined	04:40
	← swtaarrs left	04:43
	→ krawek joined	04:48
	← krawek left	04:50
	→ krawek joined	04:50
	→ mbuf joined	04:51
	← mtodd_ left	04:56
	Arme[0] → Arme[N]	05:01
	← joydivider_ left	05:11
	← Modius__ left	05:17
	← ciaran left	05:17
	→ ciaran joined	05:17
	← EiNZTEiN left	05:18
	→ albertito joined	05:24
	foca_ → foca	05:24
	← davidfetter left	05:24
	← bryanray left	05:27
	← ricky left	05:29
	← vbabiy left	05:32
	→ srid joined	05:33
	← srid left	05:36
	→ adante joined	05:37
	→ ricky joined	05:37
	→ sohail joined	05:41
sohail	hey guys, if I wanted to track another repository within my own repository, how would I do that with git?	05:41
	← krawek left	05:43
	← CodeOfficer left	05:45
	→ CodeOfficer joined	05:45
	← Ademan left	05:46
	→ Ademan joined	05:47
robinr	sohail: git remote add <somenameforit> <otherrepo>; git fetch <somenameforit>	05:48
	← CodeOfficer left	05:48
sohail	robinr, thanks!	05:49
	→ cncfanatics joined	05:54
	→ segy joined	05:55
segy	I just commited to code when I was on "no branch" and then switched to a branch	05:55
	any way to find out where that commit actually went?	05:55
	← albertito left	05:56
segy	ahh, nm. git-lost-found	05:57
pigeon	can you do something like a git remote show but without having it to connect to the remote host?	06:00
	the info shown with git remote show doesn't seem to need an actual connection (or i'm wrong?)	06:02
	→ albertito joined	06:05
	← robinr left	06:10
Ilari	segy: Or 'git reflog'.	06:11
	← brynary left	06:14
	← bobesponja left	06:17
	→ deavid joined	06:18
	← sohail left	06:20
	→ priidu joined	06:20
	→ hashif joined	06:28
hashif	I made a git repo on machineOne. I pulled everything onto machineTwo and made changes to the code on both machineOne and machineTwo. Then I did a git push from machineTwo back to machineOne. But I can't figure out which command I need to type on machineOne for my code to reflect the changes. I'm sure it's something simple.	06:29
	→ l0verb0y joined	06:31
	→ _zack joined	06:31
hashif	(typing "git log" on machineOne shows the history of changes, but none are actually applied to the code files. There must be a way to update the files to match.)	06:31
	← Ademan left	06:31
Arjen	hashif: see http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	06:34
	← ciaran left	06:35
BeerSerc	OK maybe this is a really stupid question: How can I create "sub branches"? I want to create a branch like origin/bob/new_version. I am only able to create origin/bob or origin/new_version, but not origin/bob/new_version. how would I do that or whats the correct keyword for finding documentation?	06:36
	← johan-s left	06:37
	→ bentob0x joined	06:39
Pieter	just use the slash in the branchname	06:40
Arjen	BeerSerc: You can create origin/bob/new_version if origin/bob doesn't exist yet	06:40
	← foca left	06:40
	→ pygi joined	06:41
Arjen	BeerSerc: Also, they're not sub-branches. They're just a name	06:41
hashif	Arjen, Okay, thanks for the info. It sounds like the answer is "never do that". I'll poke around and see what the easiest way is to fix it from here.	06:43
Arjen	There's a link at the end of the answer with the 'correct' way	06:43
hashif	Arjen, well, after I used "git log" to show me that the changes were successfully back onto machineOne, machineTwo was reformatted :-)	06:44
Arjen	Ah :-)	06:44
	← pflanze left	06:45
hashif	Arjen, but I think I can abandon the files on machineOne and just use the index, and rebuild the files from the index into another directory.	06:45
	← bentob0x left	06:45
	→ bentob0x joined	06:45
Arjen	If you don't have any uncommitted changes, I think you can use git checkout	06:47
	→ Fullmoon joined	06:49
	→ Ademan joined	06:52
hashif	Arjen, I did, but I'm all straight now. Thx for your help!	06:53
	← ash_ left	06:53
Arjen	np	06:53
hashif	Arjen, (I cloned into another directory, and am just keeping the .git in the original directory to be a central server kind of)	06:53
	→ ash_ joined	06:53
	← _zack left	06:55
BeerSerc	Arjen: thx	06:59
	← Ademan left	07:03
	metze_away → metze	07:10
	→ johan-s_ joined	07:13
	→ Ademan joined	07:14
Ilari	BeerSerc: Also, if you have branch 'foo', you can't have 'foo/bar' (and vice versa).	07:17
	← hashif left	07:17
	← Ademan left	07:19
	→ gberdyshev joined	07:19
	← gberdyshev left	07:20
	→ gberdyshev joined	07:20
	→ Ademan joined	07:21
	gberdyshev → gberdyshev_	07:21
	→ BabelO joined	07:22
	← gberdyshev_ left	07:25
is_null	hello everybody, how to git clone --bare a particular directory please? I checked man git-clone but i couldn't find	07:31
	although i'm convinced that it's possible	07:31
jengelh	got git 1.5.6?	07:32
is_null	jengelh, 1.5.4.5 does that mean i have to upgrade?	07:32
jengelh	right..	07:33
	→ gberdyshev joined	07:34
	← pergesu left	07:34
	→ reel_ joined	07:34
	gberdyshev → gberdyshev_	07:35
	gberdyshev_ → gberdyshev	07:35
	gberdyshev → gberdyshev_	07:36
	gberdyshev_ → gberdyshev	07:36
	→ CodeWar joined	07:36
	→ ph^ joined	07:36
	gberdyshev → gberdyshev_	07:36
	→ rraasch joined	07:38
	← l0verb0y left	07:38
	← sverrej left	07:39
	→ hashif joined	07:39
	→ mcella joined	07:44
	← peritus__ left	07:49
	← alegscogs left	07:53
	→ charon joined	07:56
	→ nud joined	07:56
	← CodeWar left	07:56
ehamberg	can i make git diff show the differences in an extern viewer like kompare?	08:02
is_null	jengelh, i got it but i don't see any revelant change in the manual, am i missing anything?	08:03
	→ _eric joined	08:03
	← hashif left	08:03
	→ CodeWar joined	08:04
	→ hashif joined	08:04
_eric	if I do a "git pull" and there are merge conflicts, is there a command I can run to take "their" version of a file?	08:04
CodeWar	I ve been searching for debates about branching vs patching. So far most google searches dont point me to somethign constructive. Do folks here have any fav thread or opinions	08:04
jengelh	is_null: strange	08:05
bremner	CodeWar: not understanding what patching means in a git context	08:05
jengelh	I see a git-clone --bare there	08:05
agenteo	I did a git pull, and I have some unmerged changes in a file that I want to get rid of, if I use git checkout file the file still contains the diff	08:06
CodeWar	mercurial style patch queues	08:06
jengelh	_eric: start off their branch and merge yours in, with --strategy ours or so	08:06
_eric	so if I've already done the pull, there isn't a simple way to fix it for a file?	08:07
bremner	CodeWar: do you know about stg (stacked git?)	08:08
	← daaku left	08:08
CodeWar	sorry I meant to think of the philosophy not a particular SCM's implementation. perhaps I should ask in the linux kernel groups and I m not from the Linux kernel background. at first it sounded pretty apalling that people maintain a branch and then there are queues of patches being carried around. Surely there must be a justification	08:10
	→ charon_ joined	08:11
	→ anmoch joined	08:14
	← charon left	08:14
	anmoch → charon	08:14
	← charon_ left	08:14
agenteo	cool, git checkout HEAD file	08:15
janneb	CodeWar: My impression is that with the advent of SCM:s with decent branching and merging such as git or mercurial, the "feature branch" style of development has become more popular. Whereas with, say, CVS or SVN where merging sucks the "head that is branched for releases" seems more popular..	08:15
sbahra	janneb, branching is so much more useful when you can feature branch :D	08:16
	easily	08:16
	→ _zack joined	08:17
jdq	janneb: I do both... I have a current developing head, feature branches and branches for special releases	08:17
	cherry picking is such a nice feature for those release branches... I dont want to miss it	08:18
CodeWar	jdq: by cherry picking you mean pick a subset of changes and apply it to the release branch	08:19
sbahra	jdq, nod	08:19
jdq	CodeWar: yeah	08:19
	→ sverrej joined	08:21
janneb	sbahra, jdq: agreed. My point being, if your SCM makes branching painful, you tend to avoid them. Hence with SVN the usual process is to do all development on head, and branch only for releases.	08:21
sbahra	Yes, this is what I'm using it for at the moment for any projects I work on.	08:22
CodeWar	So soudns like the general air is that branch if your SCM allows it for beggars use patches on the main branch	08:22
sbahra	Unfortunately, since branching over there is implemented the way it is, you also make the mistake of branching "sub-projects" too.	08:22
jdq	absolutly... two years ago, we developed a software with like 10 people and in the end more then 4000 commits to the svn... but not a single branch being made	08:22
	→ trochala joined	08:22
sbahra	Which really doesn't reflect the ontology, at all.	08:22
	Also because SVN isn't distributed, I was forced to give commit access to developers who should not have had commit access.	08:23
	I spent many hours being a maid.	08:23
	I really really like git. :-P	08:23
	dwmw2_gone → dwmw2	08:24
	→ ciaran29d joined	08:24
	→ koke joined	08:24
	→ peritus_ joined	08:29
	← sverrej left	08:33
charon	so it seems gitk persistently refuses to restore my sash positions in the history window beyond 1190px... anyone else had this problem?	08:34
	it saves the correct coordinates to ~/.gitk, but the 'sash place' command apparently never gets beyond 1190	08:35
Bombe	“sash”?	08:36
	→ sverrej joined	08:36
	→ charon_ joined	08:37
	← charon_ left	08:38
jengelh	vertical divider.	08:39
	(or horizontal)	08:39
	→ charon_ joined	08:39
	← charon left	08:41
	charon_ → charon	08:41
Bombe	Ah.	08:43
	Yeah, I have the same problem... at home with 1280px width the dividers are fine, at work with 1680px they are not.	08:43
charon	(sorry, my wlan here is terrible)	08:45
	i'll toy some more, but i'm more or less tcl illiterate	08:45
	→ SuttoL joined	08:47
	→ sbodo_w joined	08:52
	← snailrails left	08:52
	→ chris2 joined	08:53
	→ sbeyer joined	08:58
	→ x-oweb joined	09:02
	← clairvy left	09:03
	→ kraymer joined	09:03
	→ douthat joined	09:03
	← CodeWar left	09:06
	→ snailrails joined	09:06
	→ _slacker_ joined	09:07
	← felipec left	09:11
douthat	question: I have some legacy code that I imported into git. There were two installations of the same original code base that became divergent, and no history was kept. Lets call these versionA and versionB. When I initially imported these into git, I branched versionB from versionA, and worked for a few weeks without problem. However, now that I'm trying to merge versionA and versionB. I realize now that I should have imported each into "empt	09:12
	→ emdete joined	09:12
	→ d0k joined	09:13
emdete	hi, after a `git reset --hard HEAD` i get the message "warning: refname 'HEAD' is ambiguous." when doing `git status` and `git branch` says "* (no branch)" - whats wrong?	09:14
Ilari	emdete: You have branch named 'HEAD'?	09:14
	→ ebel joined	09:15
douthat	that should be: I realize now that I should have imported each into "empty nodes" instead of branching. How can I go back in time and tell versionB to not have a parent?	09:15
	→ fisons joined	09:16
Ilari	douthat: Figure out full 40 hex digit id for versionB, add it to .git/info/grafts and use 'git filter-branch versionB'?	09:18
douthat	Ilari: thanks, I'll give that a try	09:19
	← reel_ left	09:20
	→ reel_ joined	09:21
	← douthat left	09:21
	← spearce left	09:22
	pilif\|away → pilif	09:24
emdete	Ilari: isn't that some default?	09:25
	Ilari: and: no i havent	09:26
	→ georgyberdyshev joined	09:32
	← gberdyshev_ left	09:32
	georgyberdyshev → gberdyshev_	09:32
ciaran29d	does git have a feature similar to the new svn “changelists”?	09:37
bremner	ciaran29d: maybe you could tell us what they are?	09:38
ciaran29d	bremner: basically it’s a way that I can set up a shortcut to perform operations on set lists of files. E.g. I can say that a.txt and b.txt are part of changelist "text", and then svn status --changelist text, or svn commit --changelist text	09:39
mxey	File aliases, so to speak?	09:40
ciaran29d	a group of file aliases, yes	09:42
	→ Man-erg joined	09:42
paakku	Does git cherry do anything with the "(cherry picked from ...)" lines added by git cherry-pick -x?	09:46
	← Fullmoon left	09:53
	→ markkalderon joined	09:54
	← peritus_ left	09:54
	→ highandwild joined	09:55
highandwild	hi all	09:55
	how do I create a branch so that if a file is modified in master, the change also reflects in this branch ?	09:55
	or do I have to manually merge them ?	09:55
	← agenteo left	09:56
	→ Mikachu joined	10:01
	→ janm joined	10:02
Ilari	emdete: 'git for-each-ref \| grep HEAD' might be helpful in locating the second (conflicting) HEAD reference...	10:02
bremner	ciaran29d: sounds nice. the only thing that occurs to me is git-stash, but that is about snapshots of contents, not names	10:04
Ilari	highandwild: Yes, either manually merge them or rebase the brnach (assuming the new branch is private).	10:04
	← fhobia left	10:04
Ilari	ciaran29d: Something similar would be to script 'git update-index' (or even 'git add', altough that isn't 'scripting grade').	10:05
	← hashif left	10:05
	→ hashif joined	10:05
	← hashif left	10:05
ciaran29d	I suppose I could fiddle something with aliases but then I’d have to do one for each command…	10:05
	→ hashif joined	10:06
bremner	ciaran29d: or shell variables, if you are desperate. list="foo/bar fubar"	10:06
ciaran29d	yeah	10:06
	not repo-specific then though	10:07
Ilari	Turn extended globbing on, and one gets glob patterns as powerful as regular expressions...	10:07
highandwild	Ilari: but wouldn't rebase overwrite any conflicts ?	10:07
Ilari	highandwild: Rebase essentially makes patches out of the branch, resets the branch away and reappiles patches to recreate the branch...	10:08
	highandwild: And yes, rebase can run into conflicts...	10:08
highandwild	Ilari: ok, so I think merge is the way to go for me	10:08
	Ilari: thanks	10:08
Ilari	highandwild: Merge leads to sightly worse history, but doesn't mess with existing history (important on published branches).	10:09
	highandwild: Which is way to go actually more depends on what that branch is for...	10:15
bremner	ciaran29d: you could make a wrapper script for git that expands changelists and then calls git (changelists could be stored as git config)	10:15
	bremner hopes that last suggestion was awful enough that someone will think of something good :-)	10:16
highandwild	Ilari: well working on an version 2.0 of a web app	10:17
	Ilari: specifically working on the UI only	10:17
	Ilari: so I want any non-UI changes in master to be merged into the UI branch while I'm working on it	10:17
Ilari	highandwild: That sounds like it will be 'published', so merging is the right way...	10:19
	← emdete left	10:19
highandwild	Ilari: yes, i expect that UI will be finished and published at some point	10:20
	its not a throwaway branch	10:20
	← hashif left	10:20
Ilari	highandwild: 'Published' here refers to branch being available for fetch/pull by others...	10:21
highandwild	Ilari: ah sorry	10:22
	Ilari: its not an open-source project, so no it won't be published like that	10:22
bremner	highandwild: open source is not the point.	10:22
Ilari	highandwild: Well, if there are multiple devs working on it, then it likely will be 'published' in sense I descibed.	10:22
highandwild	well, I'm a single-user on this project	10:23
	single-dev rather :)	10:23
Ilari	highandwild: So 2.0-UI is topic branch?	10:23
highandwild	well, sorry I'm not familiar with the topic term	10:24
	but basically, once 2.0 is finished, I'd merge it to master, since its a web app and I wouldn't really have anyone using 1.0	10:25
	but my initial concern was that _while_ I was working on the branch, I'd also be making changes in 1.0 that would need to be reflected in 2.0	10:25
Ilari	highandwild: Based on what you have described, updating by rebase could be better than update by merge...	10:26
highandwild	But then note that some changes in 2.0 might also affect the master - that I would not want to merge until I finish work on 2.0	10:26
	So no its not a strict topic branch. It would affect some of the non-UI code	10:27
	Am I making sense ? I'm sorry to be dense, but I'm relatively new to git	10:27
Ilari	There's actually another point: If stuff in master changes radically, you could need to do lots of fixing up in topic branch also (since interfaces have changed).	10:28
highandwild	ah	10:28
Ilari	With merge, you only need to fix up stuff once per breakage...	10:28
highandwild	no I would just be adding functions and views, not _changing_ the existing api	10:28
Ilari	...And not having to fix older stuff...	10:28
highandwild	i.e in master	10:28
	Another point - does a tracking branch help my situation ?	10:29
Ilari	highandwild: If app is properly seperated, i.e. UI code is in seperate files, then both rebases and merges should autoresolve conflicts...	10:29
highandwild	Ilari: its a rails app, they're in separate files, but changing the views that I way intend to, will affect the controller code as well	10:30
	← sr left	10:32
	sr_ → sr	10:33
highandwild	but definitely, there won't be many changes to master, just a few. The bulk of the work will be in UI	10:33
	→ peritus_ joined	10:37
	→ hashif joined	10:38
	← peritus_ left	10:38
Ilari	As long as changes aren't very extensive, the conflicts should remain fairly rare...	10:39
highandwild	right - I am beginning to understand rebase now	10:40
	how do I check what the conflict in a merge would be without actually doing a merge. Doing a diff, right ?	10:40
	so I can do a git diff, and if I think there's no conflict, I can git rebase	10:41
	if there's a complicated conflict, I can use merge	10:41
bremner	highandwild: git merge --no-commit ?	10:42
highandwild	bremner: thanks	10:42
Ilari	Or just attempt the merge and rebase and if it goes sour, abort/undo it?	10:43
highandwild	right	10:43
	ok, I think this is good	10:43
	Appreciate your patience, guys	10:43
	→ hendry joined	10:46
	→ MichaelRaskin_ joined	10:47
hendry	i accidentally added a file in git. how do i unadd? git reset or git update-index --force-remove ?	10:47
Mikachu	the first	10:48
doener	git reset -- file or git rm --cached file	10:48
hendry	Mikachu: thanks	10:48
	→ yofel joined	10:51
hendry	doener: thanks for pointing out that odd '--' syntax.	10:54
	→ nothingmuch joined	10:54
	→ Xyri0 joined	10:54
nothingmuch	i want to migrate from SVK to git-svn for a project	10:55
	← huyx left	10:55
nothingmuch	my problem is that it has lots of top level sub projects, each with it's own trunk/branches/tags	10:55
	each of these needs it's own git checkout	10:55
	and mirrorring the whole thing will take me forever	10:55
	← mithro left	10:55
nothingmuch	is there a better approach than just git svn clone for every subproject?	10:56
avar	I don't know but you should only have to clone the repository once to your machine	10:56
bremner	nothingmuch: maybe http://kitenet.net/~joey/code/mr/ is interesting to you	10:57
	nothingmuch: there are also submodules, but I'm against them this week :-)	10:57
nothingmuch	okies =)	10:58
	avar: git svn clone is only copying the subset of the revs applying to the tree, so i think actually not	10:58
	it's once per branches/tags/trunk tuple	10:58
	but not once per SVN repository	10:58
doener	nothingmuch: how many revisions does that repo have?	10:58
nothingmuch	several thousands	10:59
	i think about 10k	10:59
	not sure	10:59
	← hendry left	10:59
nothingmuch	ideally i could abuse ~/.svk which already has a mirror of the whole thing	10:59
	aha! http://utsl.gen.nz/talks/git-svn/intro.html#howto-fetch-convert	11:00
	mugwump++	11:00
doener	I cloned a 11k repo two days ago, using 1.5.6.1 + git-svn fixes, that only took maybe 2 hours?	11:00
nothingmuch	it took me about an hour to git svn clone just one of the subprojects	11:01
	about 1500 revs	11:01
doener	which git version?	11:01
nothingmuch	1.5.6.1	11:01
	→ roham[rt] joined	11:01
nothingmuch	what I linked to should be OK	11:01
doener	note that that versions two bugs, one that leaves tmp files around (eventually slowing down things due to lots and lots of files in /tmp)	11:02
	if /tmp runs out of space, you also get a corrupted cloe	11:02
	s/cloe/clone/	11:02
	and the other is a memory leak, if you're short on memory, that might cause your box to swap... (although for 1500k revs, it should "only" eat about 100-150MB, I guess, depending on the contents of the svn repo)	11:03
	ehrm s/1500k/1500/	11:03
	for 1500k it would likely eat a PB or so :-)	11:03
Pieter	1500k should be doable though	11:04
	oh well, better use something else then	11:04
doener	nothingmuch: there are fixes for both bugs in git.git master	11:04
	← Xyrio left	11:05
	Xyri0 → Xyrio	11:05
doener	oh, and in maint, of couse	11:05
	s/couse/course/	11:05
nothingmuch	doener: good to know	11:05
doener	so if possible use one of those	11:05
nothingmuch	will do	11:05
	i compiled git from a checkout	11:05
	i just checked out the tag for 1.5.6.1	11:05
	i guess checking out maint is prefferred?	11:06
doener	ok, re-reading what I wrote, all those stupid typos tell me I and some coffee... brb	11:06
	nothingmuch: I'm running master here all the time, but if you're the cautious type of user, use maint	11:06
	master is already supposed to be reliable (most of the time)	11:07
nothingmuch	i guess i could use master	11:07
	← MichaelRaskin_ left	11:07
nothingmuch	i'm not a heavy user	11:07
	but I suppose strange new devs will be done in branches	11:07
	so that master should be OK	11:07
	btw, asciidoc seems utterly broken on OSX	11:08
	xmlto is too	11:08
	(none of the deps installed without futzing, and the xslt is still not working)	11:08
	so I can't build the manuals	11:08
	is this a known issue?	11:08
doener	might be the asciidoc version, IIRC there's a Makefile variable for certain versions	11:09
nothingmuch	asciidoc needed --unsafe to be passed	11:09
Pieter	it's only broken with macports I think	11:09
nothingmuch	i installed asciidoc, xmlto and getopt from their respective tarballs	11:09
Arjen	I just get the built docs from the man and html branches :-)	11:09
doener	nothingmuch: and in git.git, there are branched with the docs pre-built	11:09
	← adante left	11:09
nothingmuch	doener: i think I tried to use that	11:10
	but didn't see them	11:10
	← highandwild left	11:10
nothingmuch	maybe I did it wrong	11:10
	nothingmuch is still a n00b	11:10
	← roham[rt] left	11:11
doener	nothingmuch: there's a quick-install-doc make target	11:11
nothingmuch	ah	11:12
doener	nothingmuch: and you can see the branches with "git branch -r" (remote branches) or "git branch -a" (all branches)	11:12
	← nud left	11:12
	→ kukks joined	11:16
	← imyousuf left	11:18
	→ emachado joined	11:20
	← priidu left	11:23
	→ priidu joined	11:24
	→ peritus_ joined	11:28
	x-oweb → X-Blade	11:29
	← tmz left	11:30
	→ georgyberdyshev joined	11:32
	← gberdyshev_ left	11:32
	georgyberdyshev → gberdyshev_	11:32
paakku	Does git have some command to show how many years, months and days passed between two tags?	11:33
	→ mithro joined	11:34
	→ lamont joined	11:34
	← lamont left	11:34
	→ lamont joined	11:35
	→ gitte joined	11:36
nothingmuch	bremner: where is the extrension for git-svn support in mr?	11:37
	i can't find it on the page	11:37
	← koke left	11:37
	→ jordi_bunster joined	11:40
	→ radarek joined	11:43
	→ guardian joined	11:44
guardian	hi	11:44
	does rebase automatically fetch ?	11:44
	→ warlock_mza joined	11:46
paakku	no but there is git pull --rebase	11:46
	← mithro left	11:46
	← alenoosh left	11:47
bremner	nothingmuch: you are right. I was just thinking about the "many git repos" issue.	11:50
nothingmuch	bremner: actually the mr page mentions an extension	11:50
	oh	11:51
	found it	11:51
	mr.git/lib	11:51
	→ jordi_bunste joined	11:51
bremner	nothingmuch: cool, I love when I am helpful by mistake	11:51
	→ jordi_bunst joined	11:52
	→ felipec joined	11:54
	→ sids joined	11:57
	→ betagan joined	11:57
betagan	hi	11:58
	← charon left	11:59
	→ lorandi joined	12:00
	← mxey left	12:01
	→ cannonball joined	12:03
	→ madewokherd joined	12:07
	← guardian left	12:07
	← Mikachu left	12:08
	← jordi_bunster left	12:09
	← jordi_bunste left	12:09
	→ guardian joined	12:09
	← data`_ left	12:09
	→ pflanze joined	12:11
	→ Helios joined	12:12
	← felipec left	12:12
	VVSiz_ → VVSiz	12:12
	← metze left	12:12
	→ metze joined	12:15
	→ data` joined	12:15
kraymer	hi there! i have a problem with git-svn: i mirrored part of a svn repo (~400 revisions) then made some (~10) local git commits. when i did the dcommit, the first three commits went fine, then the dsl connection got interrupted	12:16
	i thought i could just restart dcommit but my local repo is in an 'unexpected' state	12:17
	it has all the modification from all missing commits as a local, uncommited change	12:17
	i guess i could just git-commit, then git-svn dcommit it but this isn't what i intended to do in the first place	12:17
	anyone experience with that? or ideas on how to solve?	12:17
	..or both? ;)	12:17
	← Helios` left	12:18
	← hashif left	12:20
	→ hashif joined	12:20
jast	here's what probably happened: dcommit did a rebase and that got interrupted. see if git rebase --abort helps (though i'm not sure whether git-svn actually uses that part of rebase).	12:22
	otherwise, find the right state of your branch in git reflog and then git reset --hard <the name of that reflog entry, e.g. HEAD@{123}>	12:22
	(but do make a backup before doing reset --hard)	12:23
	← Man-erg left	12:24
	→ jordi_bunster joined	12:25
kraymer	hey jast. git rebase doesn't work. it says there was no rebase in progess	12:25
	i'll have a look at the revlog thing, thanks	12:26
jast	i suspected as much, to be honest	12:26
	git-svn uses rebase in an interesting way	12:26
	→ alegscogs joined	12:26
	← peritus_ left	12:28
	→ koke joined	12:28
kraymer	mokay, so i found thath head@{12} is the newest revision that got into the svn repo	12:30
jast	that's not what matters	12:30
	← is_null left	12:30
jast	what you want to do is find the newest commit of those that are not in svn yet	12:30
kraymer	ok, that would be head@11	12:31
	so I'll hard-reset there	12:31
doener	kraymer: you can also look at the reflog for your branch: git reflog show $branchname	12:31
jast	so what sort of things do 10..0 point to?	12:31
doener	that's sometime easie	12:31
	r	12:31
	oops	12:31
	← jordi_bunster left	12:35
	→ grahal joined	12:35
	→ EmilMedve joined	12:36
kraymer	hm, 10..0 are all rebase actions	12:36
	doener: thanks for the hint	12:37
doener	kraymer: you should now reset --hard back to the state that you were at when you started your dcommit, then svn rebase and then try to svn dcommit again	12:38
	(doesn't hurt to verify the history after the svn rebase ;-)	12:38
kraymer	ok, will do	12:38
	i guess so ;)	12:38
doener	the rebase step should cause those commits that already got dcommitted to be replaced by the new commits (those with the git-svn-id lines in the commit messages)	12:39
	otherwise, you _might_ end up with those commits duplicated. git-svn might be smart enough to avoid that, but I never tried, and don't intend to g	12:40
	← hashif left	12:40
	← jordi_bunst left	12:42
kraymer	wow. thanks guys, it worked	12:44
	so i went back to that ref via reset, git-svn fetched and rebased again (because it complained my revision was out of date), then dcommited	12:44
doener	yeah, sounds good	12:45
kraymer	i would never have got there!	12:45
	doener fortunately didn't have to do that yet	12:45
kraymer	git has steep learning curve ;)	12:45
	but well, I'm to blame for using it with svn i guess :)	12:46
doener	I'd probably go and look at the svn repo for hours to check for any corruption on that end...	12:46
	kraymer: right!	12:46
	:-)	12:46
kraymer	well, it's for mplayer. i just wanted to try git on something 'real' and not just locally	12:47
doener	heh	12:47
kraymer	diego will kick my butt if i screwed with mplayer repo now :)	12:48
	but i think it all worked	12:48
	← CIA-50 left	12:48
kraymer	anyway. thanks again!	12:49
nothingmuch	is there any way of setting up aliases for svk heads, like st, br etc?	12:49
	i keep getting an error from git st when I mean to type git status	12:49
	← csc` left	12:49
nothingmuch	it's in muscle memory by now =P	12:50
doener	nothingmuch: git config alias.st status	12:50
nothingmuch	ta =)	12:50
doener	nothingmuch: or better add --global	12:50
nothingmuch	hai	12:50
doener	otherwise, it's only for your current repo	12:50
nothingmuch	nod	12:51
	← trochala left	12:51
	→ CIA-2 joined	12:57
	→ peritus_ joined	12:57
	→ csc` joined	12:58
	→ sergiodj joined	12:59
	→ wagle_ joined	13:02
	← wagle_ left	13:02
	→ cmarcelo joined	13:03
	← spool left	13:11
	→ brynary joined	13:11
	→ Beket joined	13:13
	→ charon joined	13:14
	→ flashwolf joined	13:15
flashwolf	hi there	13:16
	→ zorglu__ joined	13:16
	Arme[N] → Arme[0]	13:18
	→ bartocc joined	13:19
bartocc	hi	13:19
	I've rebased and push on another branch something that I had already commited	13:19
	how can I move the previously pushed branch forward ?	13:20
	← alegscogs left	13:21
	→ vbabiy joined	13:22
bartocc	ok, I figured it out	13:22
	← sids left	13:22
	← markkalderon left	13:23
	← brynary left	13:23
	← zorglu__ left	13:25
	← seezer left	13:30
	→ seezer joined	13:31
	→ georgyberdyshev joined	13:31
	← gberdyshev_ left	13:32
	georgyberdyshev → gberdyshev_	13:32
	→ bryanray joined	13:35
	→ joydivider joined	13:35
	→ alegscogs joined	13:36
	→ aroben joined	13:36
	→ joydivider_ joined	13:38
	← joydivider left	13:38
	→ lea_w joined	13:39
	← rraasch left	13:41
	← joydivider_ left	13:44
	← _zack left	13:45
	→ foca joined	13:48
	→ clairvy joined	13:48
	← scel left	13:48
	→ zedr0n joined	13:51
zedr0n	hi	13:51
	does git handle hard links?	13:51
Tv	it will consider them just regular files	13:52
tchan	Has something changed in git that would break the permissions needed for gitweb.cgi to run correctly? Updated to git-1.5.6.1 and now after every commit to a git repo, gitweb says "No Commits" until I manually run "git gc"	13:52
zedr0n	ah, that's why I have problems	13:52
	← flashwolf left	13:53
zedr0n	I store the headers to my library in separate directory using hard links	13:53
Tv	zedr0n: sounds like you wouldn't want to version control those, as they aren't really source..	13:53
zedr0n	yes	13:53
	how do I untrack them?	13:53
Tv	git rm --cached	13:53
zedr0n	thx	13:53
	I was debugging it for several hours before it occured to me	13:54
	the class definitions were different in the linked object file from the library and the source file where I used the library	13:55
	→ deavidsedice joined	13:59
	→ Yuuhi joined	14:00
	→ _zack joined	14:01
Tv	zedr0n: most people would have used symlinks or /usr/bin/install	14:04
zedr0n	hm, I think I needed hard links because CMake doesn't handle symlinks	14:04
	→ dotsintacks_ joined	14:05
	→ spearce joined	14:08
	→ thiago joined	14:10
	← spearce left	14:12
zedr0n	hm, can I clone a repository from the local machine so that the clone is an exact copy and isn't connected with the original in any way?	14:12
	will just copying do?	14:13
	i want to test some merging but don't want to mess up the repository	14:13
	→ blowery joined	14:15
charon	copying is fine	14:15
	→ adante joined	14:15
Ilari	zedr0n: After copying, you might want to run 'git status' in the copy to fix up cached working tree status data...	14:16
zedr0n	ok, I'll try that	14:16
	← madewokherd left	14:18
	← dotsintacks left	14:22
	← ebel left	14:22
	→ glommer joined	14:24
	→ davidfetter joined	14:25
	→ kevwil joined	14:26
	→ ebel joined	14:26
	← _zack left	14:26
	→ douthat joined	14:34
	← douthat left	14:34
	→ tmz joined	14:35
tchan	Does anyone have an idea on what to look for in a git repo that was working fine with gitweb but after the recent bump to 1.5.6.1 from 1.5.5.4, that same repo has problems being viewed by gitweb - "No Commits". Problem goes away temporarily if I run "git gc" on the repo but the problem comes right back after the next commit to the repo.	14:35
	→ tmz_ joined	14:35
	← EmilMedve left	14:37
	← tmz_ left	14:37
	→ lamont` joined	14:38
	→ ^Beket\| joined	14:41
	cncfanatics → cnc\|away	14:42
paakku	tchan: you might try deleting .git/gitk.cache; although I don't know if it is possible to corrupt that in such a way that your symptoms would appear.	14:45
	→ _zack joined	14:45
	→ mxey joined	14:45
	→ hendry joined	14:46
	← lamont left	14:47
	lamont` → lamont	14:48
	→ Fullmoon joined	14:50
Ilari	tchan: Also, file permissions might be messed up...	14:50
	tchan: Especially modes of objects/<2 hex digits> and objects/pack directories.	14:50
	blowery → blowery`afk	14:53
	← Beket left	14:57
tchan	I checked and there is no .git/gitk.cache file.	14:59
	→ Ryback_ joined	15:00
tchan	What should the file permissions of objects and objects/pack directories be ?	15:00
	← _zack left	15:00
Ilari	tchan: What are they currently?	15:01
tchan	.git/objects is drwxrwxr-x+	15:02
	.git/objects/pack is drwxrwxr-x+	15:02
paakku	the + indicates an extra ACL, right?	15:02
Ilari	tchan: That looks fairly sane (except the .git part)... There are no dirs like objects/52?	15:03
paakku	To find out which part of gc temporarily fixes the problem I suggest you run git pack-refs separately.	15:04
	→ queso joined	15:04
tchan	nothing like objects/52 because the git gc finished okay	15:04
	→ bobesponja joined	15:04
	← Ryback_ left	15:05
tchan	its not really .git, as I left off the extraneous path stuff like /home/tchan/moonbase.git/.git/objects/pack	15:05
Ilari	tchan: Shouldn't it be something like '/home/tchan/moonbase.git/objects/pack'?	15:06
	Arme[0] → Arme[N]	15:06
	← johan-s_ left	15:07
	→ l0verb0y joined	15:07
	→ Modius joined	15:07
tchan	Ilari: It isn't here. /home/tchan/moonbase.git is my working repo and it does have a .git subdir that contains objects/pack. So /home/tchan/moonbase.git/.git/object/pack is working here.	15:08
Ilari	tchan: Usually working repos are named something like 'foo' instead of 'foo.git'...	15:09
	→ foucist joined	15:10
paakku	Possibly some such assumption in gitk then?	15:10
foucist	i just did a git pull and it scrolled off my screen buffer, how can i see all the changed files etc again	15:11
	← mbuf left	15:11
tchan	I didn't think gitweb uses gitk. Does it ?	15:11
jengelh	not at all	15:11
Ilari	foucist: 'git diff HEAD@{1} HEAD'?	15:11
	→ nachof joined	15:11
paakku	tchan: oh sorry, I misread and thought you had trouble with gitk	15:12
	didn't see the gitweb part	15:12
Ilari	foucist: (assuming that the merge was actually successful).	15:12
foucist	Ilari: doesn't look like that	15:12
	git diff HEAD@{1} HEAD just shows all the content differences	15:13
	→ bdiego joined	15:13
foucist	i just want to see the files not the content	15:13
	i want to see roughly what git pull shows	15:13
paakku	diff --stat perhaps	15:13
nachof	is there a way for git to preserve the timestamp of files? I have a bunch of files and their creation time is an important piece of information; how can I preserve that?	15:14
Ilari	nachof: There's no space to put that information. :-/	15:15
	nachof: (Well, outside working tree itself).	15:15
	← hendry left	15:15
nachof	Ilari, I suspected that, but I wanted to confirm it; well, I guess I'll take the time to copy that info inside the files themselves; thanks!	15:16
Ilari	nachof: Why is creation time important information?	15:16
	← zedr0n left	15:16
foucist	git diff HEAD^..HEAD --stat is showing changes to files that were changed a long time ago.. i.e. git diff HEAD^..HEAD <specific file> ain't showing anything at all	15:17
chris2	nachof: there is some too to preserve metadata in a dotfile	15:17
	*tool	15:17
	→ zedr0n joined	15:18
nachof	Ilari: I want to use git to manage a bunch of short stories I wrote; many of them a while ago; the newer ones have the date info at the bottom, but the older ones don't (I didn't think at the time that it would be somehting I might want to know later)	15:18
	chris2: that could be useful too	15:18
chris2	i think it's called etckeeper	15:18
	→ mtodd_ joined	15:19
nachof	chris2: thanks, I'll google it	15:20
	← clairvy left	15:20
	← davidfetter left	15:21
tchan	paakku: Okay, added a commit to my git repo, gitweb again says "No commits". Running "git pack-refs" does not fix the problem. If I run "git gc" then gitweb correctly displays the repo.	15:26
	→ krh joined	15:28
Tv	nachof: why don't you add the times at the bottom now?	15:31
	← foucist left	15:31
	→ georgyberdyshev joined	15:31
	← gberdyshev_ left	15:31
	georgyberdyshev → gberdyshev_	15:32
	→ _zack joined	15:32
nachof	Tv: I will, I just wanted to save the work :)	15:32
Tv	nachof: etckeeper is going to be more work	15:32
	→ digitalhobbit joined	15:33
Tv	nachof: you can always for for a in *.txt; do stat --format=%z "$a" >"${a%%.txt.timestamp}"; done	15:34
	err	15:34
	← bgerber left	15:34
Tv	for for a in *.txt; do stat --format=%z "$a" >"${a%%.txt}.timestamp"; done	15:34
	cnc\|away → cncfanatics	15:34
	← mtodd left	15:35
	→ bgerber joined	15:35
cardioid	Hope you haven't messed with the files since the stories were written	15:35
nachof	Tv, you're right; I didn't think of automating it, thanks :D	15:36
cardioid	Isn't %z the file size?	15:36
	→ xet7 joined	15:37
paakku	tchan: well, try renaming /home/tchan/moonbase.git to /home/tchan/moonbase; perhaps it'll work then	15:38
	queso → queso\|away	15:40
	→ rdavila joined	15:42
	→ kumbayo joined	15:42
	← lea_w left	15:44
	queso\|away → queso	15:44
paakku	gitweb does $git_dir = "$projectroot/$project" in several places, and sub git_cmd then uses '--git-dir='.$git_dir;	15:45
	→ jordi_bunster joined	15:45
	→ brosner joined	15:45
paakku	so if $project is the name of a directory that contains a .git subdirectory, I don't think git will know to look there.	15:46
	→ krawek joined	15:46
paakku	yet git-instaweb is supposed to work with a non-bare repository	15:47
	← nachof left	15:48
	→ nachof joined	15:49
	→ foca_ joined	15:49
	← Ademan left	15:50
	→ cedricv joined	15:50
	→ Ademan joined	15:51
	← krawek left	15:53
	→ tvachon joined	15:55
	← charon left	15:57
	→ EmilMedve joined	15:59
	→ krawek joined	15:59
	→ matled- joined	16:00
	← krawek left	16:01
	← matled left	16:04
	matled- → matled	16:04
	→ jargan joined	16:04
	→ gebi_ joined	16:04
	← mtodd_ left	16:04
	← ^Beket\| left	16:04
	← ricky left	16:04
	← Gitzilla left	16:04
	← Strogg left	16:04
	← ferdy left	16:04
	← Arjen left	16:04
	← hacim left	16:04
	← jamesjb left	16:04
	← cebewee left	16:04
	← reuss left	16:04
	← Tali left	16:04
	← ftehw left	16:04
	← djpig left	16:04
	← udoprog left	16:04
	← Leonidas left	16:04
	← chrizz left	16:04
	← gk left	16:04
	← Pieter left	16:04
	← gebi left	16:04
	← up_the_irons left	16:04
	← rphillips left	16:04
	← slack left	16:04
	← ehamberg left	16:04
	← dionoea left	16:04
	← jast left	16:04
	← mountie`` left	16:04
	← kanru left	16:06
	→ hacim joined	16:06
	→ ehamberg joined	16:06
	← foca left	16:07
	nachof → foca	16:07
	foca → nachof	16:07
	dwmw2 → dwmw2_gone	16:07
	← jordi_bunster left	16:08
	→ Arjen joined	16:08
	→ jamesjb joined	16:08
	→ mtodd_ joined	16:08
	→ ^Beket\| joined	16:08
	→ ricky joined	16:08
	→ Gitzilla joined	16:08
	→ Strogg joined	16:08
	→ ferdy joined	16:08
	→ mountie`` joined	16:08
	→ Tali joined	16:08
	→ djpig joined	16:08
	→ reuss joined	16:08
	→ dionoea joined	16:08
	→ chrizz joined	16:08
	→ rphillips joined	16:08
	→ udoprog joined	16:08
	→ gebi joined	16:08
	→ Leonidas joined	16:08
	→ cebewee joined	16:08
	→ gk joined	16:08
	→ Pieter joined	16:08
	→ ftehw joined	16:08
	→ up_the_irons joined	16:08
	→ slack joined	16:08
	← ftehw left	16:08
	← gk left	16:08
	← ferdy left	16:08
	→ gk joined	16:08
	→ ftehw joined	16:08
	← ricky left	16:08
	→ ferdy joined	16:08
	→ ricky joined	16:08
	← udoprog left	16:09
	→ udoprog joined	16:09
	blowery`afk → blowery	16:09
	→ byronvf joined	16:09
	→ krawek joined	16:10
	← blowery left	16:10
byronvf	I believe there is a tool that displays a branch view of git's log at the command line, anyone know the name of it?	16:10
RandalSchwartz	git log ? :)	16:10
	git log branchnamehere	16:11
	or was this some sort of trick question	16:11
byronvf	But also displays a ascii art like graph	16:11
RandalSchwartz	ahh!	16:11
	like gitk	16:11
byronvf	yes	16:11
RandalSchwartz	without all the fancy graphics	16:11
	→ blowery joined	16:11
RandalSchwartz	hold on let me find it	16:11
	foca_ → foca	16:12
RandalSchwartz	git show-branch	16:12
	← guardian left	16:13
	← krawek left	16:14
RandalSchwartz	did that help?	16:14
	→ k0001 joined	16:14
RandalSchwartz	byronvf?	16:14
byronvf	Hmm, well, sorta, I remember seeing something on the web that was a little better in the ascii art department :) if that makes sense	16:15
	But I forgot the name.. I'm looking at git-show-branch now	16:15
Ilari	byronvf: git-forest?	16:16
	byronvf: (not included in standard distribution AFAIK).	16:16
byronvf	Ahh, yes, I think it's git-forest	16:16
RandalSchwartz	ahh	16:18
	that's why I hadn't seen it	16:18
	→ nbrunson joined	16:19
	← gebi left	16:20
byronvf	Randal, Ilari: thanks for the help	16:20
	← cncfanatics left	16:20
	← k0001 left	16:21
	→ fhobia joined	16:21
ReinH	Damn you OS X case insensitive filesystem	16:23
	← nachof left	16:23
RandalSchwartz	blame english for that.	16:23
	not OSX. :)	16:23
ReinH	I have a branch with a file FOO. In another branch that's been changed to foo	16:23
RandalSchwartz	Yeah, that's gonna hurt	16:23
	← sbodo_w left	16:23
ReinH	I can't git checkout another branch because of this	16:23
RandalSchwartz	if you have to maintain that, create a case-sensitive HFS+ DMG	16:24
ReinH	the fail will be of enormous magnitude... there is no escaping </mooninites>	16:24
RandalSchwartz	or even a UFS DMG	16:24
	and then unload the project in there	16:24
ReinH	RandalSchwartz: I just want it to not break this one time.	16:24
RandalSchwartz	that's what I do	16:24
ReinH	yeah, fun eh	16:24
	I basically want to make this change once and then not maintain it again	16:24
RandalSchwartz	Blame NeXT. :)	16:24
	although that may also be an homage to OS9 too	16:25
	queso → queso\|away	16:25
	→ sbodo_w joined	16:25
RandalSchwartz	at least on OSX, you can work around it	16:25
	that'd be harder on windows	16:25
Ilari	Actually, case insensitivity isn't the worst part about some Mac OS X filesystems...	16:26
RandalSchwartz	so thank the maker you're not on windows	16:26
	Yeah - the UTF8 handling is atrocious	16:26
ReinH	RandalSchwartz: I'll thank me, thank you very much ;)	16:26
RandalSchwartz	if you made yourself, you're a better man than I am. :)	16:26
ReinH	RandalSchwartz: I made myself chose OS X ;)	16:26
RandalSchwartz	OK - I'll give you that one. :)	16:27
ReinH	RandalSchwartz: but back to the issue... what's the least painful way to correct this? git update-index or etc?	16:27
	I basically want this file to be foo everywhere insetead of FOO, where everywhere is a small number of braches	16:27
RandalSchwartz	you'll need to make that change on a case-sensitive filesystem	16:28
ReinH	oh joy.	16:28
	heh	16:28
	← cedricv left	16:28
ReinH	ok, so I can do it on my server	16:28
RandalSchwartz	hence my suggestion of creating a DMG	16:28
ReinH	right, got it	16:28
	makes sense	16:28
RandalSchwartz	I have a 8GB UFS DMG that I mount after every reboot	16:28
ReinH	smart	16:28
RandalSchwartz	just to handle possible conflicts with "classic" unix	16:28
	→ lea_w joined	16:28
ReinH	oh hey, tpope is on a case sensitive filesystem. joy.	16:28
tpope	yes	16:29
RandalSchwartz	since UFS works "normal". No alias files, etc.	16:29
ReinH	RandalSchwartz: smart :)	16:29
RandalSchwartz	And most of my git mirrors mirror there	16:29
ReinH	word	16:29
paakku	How do people usually import a tarball into an existing branch?	16:29
ReinH	It still amazes me that anyone ever thought a case insensitive filesystem is the Right Way	16:29
RandalSchwartz	don't go there	16:30
tpope	I join you in laughing at ReinH's filesystem, randal	16:30
ReinH	:(	16:30
RandalSchwartz	that's "emacs" vs "vi" by proxy	16:30
ReinH	RandalSchwartz: ha, I go there pretty much daily	16:30
RandalSchwartz	or "bsd" vs "gpl"	16:30
ReinH	heh yes	16:30
	← byronvf left	16:30
RandalSchwartz	or "gnome" vs "kde"	16:30
	no winners, anywhere. :)	16:30
	perl v python v ruby v haskell v ...	16:31
ReinH	I retract my statement in the interest of not starting ragnarok	16:31
	(Ruby)	16:31
RandalSchwartz	I clearly prefer smalltalk to ruby	16:31
ReinH	RandalSchwartz: heh :) I am a big fan of st	16:31
RandalSchwartz	have you seen my blog?	16:31
ReinH	RandalSchwartz: no sir, but I would like to	16:31
	← garrin left	16:32
RandalSchwartz	MethodsAndMessages.vox.com	16:32
ReinH	heh word	16:32
	RandalSchwartz: You're familiar with SmallTalk Best Practice Patterns?	16:32
RandalSchwartz	of course	16:32
	I've quoted it	16:32
	built courseware around it	16:32
	even exchanged email with Beck about it	16:32
ReinH	RandalSchwartz: I'm co-authoring Ruby Best Practice Patterns with Beck :)	16:32
RandalSchwartz	Ahh! you're the crazy guy!	16:32
	I heard about that	16:33
ReinH	:/	16:33
	crazy guy :(	16:33
RandalSchwartz	he also did Java patterns recently too	16:33
tpope	I can confirm	16:33
RandalSchwartz	seaside will probably make rails irrelevant in a few years	16:33
ReinH	RandalSchwartz: Or MagLev	16:33
	→ maks_ joined	16:33
RandalSchwartz	that too	16:33
maks_	how can i edit author and email aftwards of an commit?	16:33
RandalSchwartz	but once people discover that maglev can still run smalltalk...	16:33
	← mikearr left	16:34
maks_	git commit --amend drops me in for the commit msg	16:34
RandalSchwartz	and even interoperate...	16:34
	the ruby people will come back into the smalltalk fold. )	16:34
ReinH	RandalSchwartz: We're quite excited about MagLev at Hashrocket	16:34
RandalSchwartz	"come back. we forgive you."	16:34
ReinH	RandalSchwartz: and I personally would love to do more paid SmallTalk work	16:34
	and by "more" I mean "any"	16:34
RandalSchwartz	well - help me push Seaside then	16:34
	→ kanru joined	16:35
RandalSchwartz	have you seen smalltalkjobs.dabbledb.com	16:35
	lots more jobs, every week	16:35
ReinH	:)	16:35
RandalSchwartz	and first learn, it's not SmallTalk	16:35
	ugh!	16:35
	← zedr0n left	16:35
ReinH	woops.	16:35
RandalSchwartz	that's like "Postgre"	16:35
ReinH	heh.	16:35
RandalSchwartz	it's a clue you're not "in the know"	16:35
	→ cncfanatics joined	16:35
RandalSchwartz	or "PERL" for that matter. :)	16:35
ReinH	or HashRocket	16:35
	:p	16:35
RandalSchwartz	see	16:35
ReinH	consider it a typo	16:35
RandalSchwartz	we all have our secret handshakes	16:35
ReinH	;)	16:35
RandalSchwartz	anyway - read my blog from the first post (reverse order)	16:36
	and see what I've been doing with smalltalk	16:36
	and promoting seaside	16:36
	seaside is the next rails	16:36
	← hacim left	16:36
comp	maks_: well .. you can change the author via --author	16:37
maks_	saw	16:38
	doing together with rebase -i and edit	16:38
	but not practical for a bunch of commits ;) comp	16:38
comp	hehe	16:38
	→ k0001 joined	16:38
Pieter	you can't spell "postgresql" without "perl"	16:39
	← bartocc left	16:39
ReinH	RandalSchwartz: I've got a lot of joy out of smalltalk and translating st best practices into the ruby idiom has been very rewarding	16:42
	*gotten	16:42
	There are some significant differences, though, that lead me to prefer Ruby on a day to day basis.	16:43
	→ charon joined	16:44
Pieter	speed! ;)	16:44
ReinH	And I certainly wouldn't ever advance the claim that Rails is an exemplar of the kind of idiomatic Ruby I talk about in my book	16:45
	In some ways it's very much not idiomatic ruby	16:45
	← Ian00 left	16:45
	← mcella left	16:49
	→ krawek joined	16:50
	← chris2 left	16:51
	← sbodo_w left	16:51
	← fhobia left	16:51
	→ robinr joined	16:53
	→ sbodo_w joined	16:54
	→ foucist joined	16:55
	← eikonos left	16:57
	→ rubydiamond joined	16:58
	→ eternaleye joined	16:58
	← foucist left	17:00
	← rubydiamond left	17:00
	→ rubydiamond joined	17:01
	← pombreda1 left	17:01
	→ _slacker__ joined	17:04
	Arme[N] → Arme[0]	17:04
	← warlock_mza left	17:06
	← pygi left	17:06
	metze → metze_away	17:06
	→ ketralnis joined	17:07
	→ timocratic joined	17:07
	← sverrej left	17:07
	← reel_ left	17:07
	← ketralnis left	17:08
	→ bronson_ joined	17:08
	queso\|away → queso	17:11
	→ eikonos joined	17:12
	← statim left	17:12
	→ p4tux joined	17:12
	← _slacker_ left	17:13
	← _zack left	17:13
	→ douthat joined	17:15
	→ CodeOfficer joined	17:16
	← kanru left	17:17
	→ kanru joined	17:19
	→ joydivider joined	17:19
	← douthat left	17:19
	← alegscogs left	17:22
	← Fullmoon left	17:23
	→ juanjoc joined	17:24
	← joydivider left	17:25
	← tvachon left	17:25
	← ebel left	17:28
	→ joydivider joined	17:28
	← joydivider left	17:29
	→ joydivider joined	17:29
	→ georgyberdyshev joined	17:31
	betagan → \b	17:31
	← gberdyshev_ left	17:31
	georgyberdyshev → gberdyshev_	17:32
	\b → betagan	17:32
	← joydivider left	17:33
	→ tvachon joined	17:34
	→ joydivider joined	17:34
	→ ketralnis joined	17:34
	→ SuttoL_ joined	17:35
	← joydivider left	17:35
	← shd left	17:36
	→ joydivider joined	17:36
	→ Eridius\|work joined	17:37
	→ nud joined	17:37
	→ Eridius\|work_ joined	17:38
	← deavid left	17:39
	← joydivider left	17:39
	→ joydivider joined	17:40
	→ cedricv joined	17:41
	← cedricv left	17:41
	← sbodo_w left	17:43
	→ ckoehler joined	17:43
	← joydivider left	17:43
	← _slacker__ left	17:43
	← CIA-2 left	17:44
	→ joydivider joined	17:45
	→ les joined	17:45
	← joydivider left	17:45
	→ joydivider joined	17:46
les	I have a question about setting up someone else's github repo as a remote for an existing local repo.	17:46
	← quabbles left	17:46
les	Basically, I was wondering if anyone could help me with the correct syntax for my .git/config?	17:46
	→ quabbles joined	17:47
doener	les: just "git remote add somename repo-url"	17:48
tpope	git remote add myfriend git://github.com/myfriend/repo.git	17:48
Eridius\|work_	les: `sudo gem install defunkt-github --source http://gems.github.com && github track myfriend/repo`	17:48
	← joydivider left	17:48
	→ joydivider joined	17:49
	← joydivider left	17:50
	→ lolage0 joined	17:51
	← SuttoL left	17:52
	→ lwhacker joined	17:53
	→ sbodo_w joined	17:53
	← Eridius\|work left	17:53
les	Thanks guys... trying it out	17:53
	→ pygi joined	17:53
	→ joydivider joined	17:54
	← ketralnis left	17:55
	→ Fullmoon joined	17:55
les	tpope: thanks, that worked wonderfully!	17:56
	→ drizzd joined	17:57
	← joydivider left	18:01
	→ joydivider joined	18:02
	← sbodo_w left	18:02
	→ pombreda joined	18:03
	→ kdubois joined	18:05
	Eridius\|work_ → Eridius\|work	18:09
	→ davidfetter joined	18:12
	→ ketralnis joined	18:12
	→ statim joined	18:16
	→ alley_cat joined	18:20
	← joydivider left	18:22
	→ jerbear joined	18:23
	→ CIA-2 joined	18:27
	→ douthat joined	18:27
	← douthat left	18:28
	← bentob0x left	18:28
	← mxey left	18:29
	→ mxey joined	18:29
	← warthog9 left	18:31
	← lea_w left	18:31
	← kdubois left	18:32
	→ catdude joined	18:32
	→ david_koontz joined	18:33
	← cworth left	18:33
	→ cworth- joined	18:34
	← mxey left	18:34
	→ mxey joined	18:34
	→ kdubois joined	18:36
	← foca left	18:36
	→ ciaran joined	18:38
	→ redinger joined	18:39
	← redinger left	18:39
	← Tigerchen left	18:40
	→ BugeyeD joined	18:41
	→ is_null joined	18:43
davidfetter	apart from hosted git and gitosis, what are some options for making writeable repos available to a developer community?	18:45
is_null	hello everybody, how to clone --bare a sub-dir of a project managed with git 1.5.6.1? I did not found it in man git-clone	18:45
BugeyeD	hi all. following gittutorial, i cd into my project directory and issue 'git init ; git add . ; git commit' ... the commit fails with "fatal: exec <snip>/.git/hooks/pre-commit failed" ... this is on openbsd 4.3. any pointers? i'm very new to git.	18:45
	→ mxey_ joined	18:46
	← mxey left	18:46
bremner	is_null: AFAIK, you need some voodoo like git filter-branch to clone a subdir	18:46
	→ paltman joined	18:47
paakku	How would one go about getting a tag signed by multiple persons?	18:48
	→ felipec joined	18:48
paakku	There doesn't seem to be git tag --amend.	18:49
felipec	paakku: git tag --force	18:49
paakku	felipec: Doesn't that lose the previous GPG signature?	18:49
	When multiple persons endorse a commit, are they supposed to create different tags each?	18:50
felipec	paakku: I don't know abut GPG signatures, but isn't the purpose of the signature to avoid modifications?	18:50
paakku	felipec: Well yeah, but there ought to be a way to have multiple signatures for the same signed content.	18:51
	→ donomo joined	18:51
donomo	whats the equivalent for 'svn info', i want to see what repo this git repo was cloned from.	18:51
felipec	paakku: so you want git tag --amend so you can add another signature?	18:51
paakku	felipec: That was my thought.	18:52
felipec	donomo: git remote -v is the most similar to that I guess	18:52
	← koke left	18:52
	← kdubois left	18:52
donomo	felipec: thx	18:52
felipec	paakku: ah, then somebody else should answer that	18:52
	→ bentob0x joined	18:52
	← eikonos left	18:53
paakku	Of course if the Tagger in the tag object is supposed to match the name and email address listed in the key that matches the signature, then it'll get tricky.	18:54
	Is there any such assumption now?	18:54
	→ murmel joined	18:54
	← donomo left	18:54
	→ warthog9 joined	18:54
	→ thannoy joined	18:54
cnk	Not clear if this is a git or a github issue, but I have a project that uses submodules. I am trying to clone to a new machine and getting an error saying "ERROR: Permission to cnk/empcms denied to cnk/restful-authentication." empcms is a private github repository and restful-authentication is a public fork of the normal repo. I can use the url from the submodule config file to checkout my restful-authentication fork on its own. Any guesses on what is going on?	18:54
Pieter	that's a github issue	18:55
	looks like you used deploy keys	18:55
	those should be unique	18:55
	you used the same on two repositories	18:55
murmel	I have a noobish question	18:55
	← rubydiamond left	18:56
murmel	I need to clone a public repo and publish my own copy of it and am not sure how to do it with remote branches etc.. should I use github, gitorious or repo.or.cz?	18:56
	and what is the best way to track upstream changes, via "fetch remote-branch"?	18:57
cnk	Pieter: thanks I don't think I used any deploy keys - which might be my problem	18:58
	← lwhacker left	19:03
	→ Wraithan joined	19:04
	← alley_cat left	19:04
felipec	a branch that has 'master' + one patch can be called a floating branch, right?	19:05
	and I'll have to rebase it all the time	19:05
	← tzafrir_laptop left	19:05
RandalSchwartz	a branch doesn't really "have" master. :)	19:06
	you have a branch that is one commit away from the branch you call master	19:06
	and yes, you can rebase that commit to a different commit, such as when you've updated master	19:06
	I don't recall the term "floating branch" being used within git before	19:07
	perhaps that's a term from a different SCM	19:07
felipec	RandalSchwartz: a branch points to a commit, right? so I think it's reasonable to assume a branch has in it all the commits parent to that commit	19:07
	→ johan-s joined	19:07
RandalSchwartz	well - that's not the terminology we use	19:08
	a branch is a name for a commit	19:08
	not the parentage of it	19:08
	the commit knows its parentage	19:08
	but that's nameless at that point	19:08
	it's just "the parent(s) of a commit"	19:08
	so in git a branch is an endpoint, not a line	19:09
	→ koke joined	19:09
RandalSchwartz	simply a name for a commit of interest	19:09
	← bdiego left	19:09
RandalSchwartz	that's why it's cheap and frequent to create and delete "branches"	19:09
	those are simply "commits of interest at the moment"	19:09
	← cardioid left	19:10
	→ cardioid joined	19:10
	→ aggiejy joined	19:11
felipec	RandalSchwartz: yeah, I know that a branch is a ref, but a "branch" in english means something that deviates from the trunk	19:12
	→ adamm joined	19:12
	← snailrails left	19:13
	→ bdiego joined	19:13
felipec	RandalSchwartz: in the design it makes sense to threat a branch just like a ref to a commit	19:13
	→ _zack joined	19:13
RandalSchwartz	well - you're also presuming that there's a "trunk" then too	19:13
	ckoehler → wecbot	19:13
RandalSchwartz	but there's nothing special about the letters "m a s t e r"	19:13
	at all	19:14
	they're all completely interchangable as far as git cares	19:14
	jengelh shows RandalSchwartz the trunk	19:14
RandalSchwartz	it's just a bunch of named commits, ready to accept a new commit on top of them	19:14
pasky	(though actually, "master" should've been called "trunk" ;))	19:14
RandalSchwartz	perhaps	19:14
jengelh	it should not	19:14
felipec	RandalSchwartz: I know that, it can be "trunk" or "main", or whatever	19:15
jengelh	pasky: the git inquisition will squash any svn	19:15
pasky	it's just that it's nicer word :)	19:15
felipec	but if branches are not meant to be treated as master then why call them that way?	19:15
	why not heads?	19:15
jengelh	they are, actually, heads.	19:15
	see refs/heads/	19:15
RandalSchwartz	this is why I'm telling you this	19:15
	it's just a bunch of heads	19:15
jengelh	a hydra!!	19:16
	← kevwil left	19:16
RandalSchwartz	"branch" in git doesn't mean what "branch" means in other SCM	19:16
	I think I started with that statement	19:16
felipec	jengelh: yeap, exactly	19:16
	wecbot → ckoehler	19:16
RandalSchwartz	why is it taking you so long to understand. :)	19:16
jengelh	then	19:16
	should not we s/branch/head/g all over the git documentation, to avoid this confusino?	19:16
RandalSchwartz	no - we just need to beat people up who come into #git :)	19:16
jengelh	RandalSchwartz: Confucious says violence is not a solution	19:17
pasky	I think the branch concept is actually reasonably good approximation of the head concept	19:17
felipec	RandalSchwartz: I understand that a branch is a head, so I'm asking: why name it head instead of branch?	19:17
	er, the other way around	19:17
RandalSchwartz	Confuscious say - "I'm dead! Why people still listen?"	19:17
	← koke left	19:18
pasky	(or rather, you can define something that's intuitively a branch if you have 2+ heads ;)	19:18
	but it's all just play with words	19:18
	← aggiejy left	19:18
jengelh	but we ain't playing scrabble	19:18
RandalSchwartz	I'm not sure what your question is felipec?	19:18
	I'm describing how it works	19:18
	using the terminology within git	19:18
pasky	felipec: are you proposing some change in the documentation? or something else?	19:19
RandalSchwartz	if you want to cloud those names with other notions, feel free to be confused at your leisure.	19:19
felipec	pasky: no	19:19
RandalSchwartz	I will not share responsibility for your confusion. :)	19:19
pasky	felipec: then I'm lost too	19:20
RandalSchwartz	local branch = a name under .git/refs/heads naming a given commit as a place that will accept new commits	19:20
	← Wraithan left	19:20
	← murmel left	19:20
RandalSchwartz	remote branch = a name under .git/refs/remotes tracking commits named on remote repos	19:20
felipec	all I'm saying is; internally a branch is a head, right? So why name it branch? A branch implies a series of commits, well, because from a head you can find the branch, so it doesn't matter how it's handled internally, you can find the branch (the series of commits)	19:20
Ilari	Branches in Git can be nonlinear.	19:21
RandalSchwartz	felipec - it's because it almost acts that way	19:21
	but if you think of it as only acting that way, you limit your understanding	19:21
	git maintains a directed graph of commits	19:21
	some of those commits are interesting	19:21
	we call those "branches"	19:21
	some of those branch heads have roughly linear parents	19:22
	you could casually call that a "branch", but that's not precise	19:22
	→ lea_w joined	19:22
RandalSchwartz	and limiting your thinking to that will confuse you when we talk about "lightweight" or "topic" branches.	19:22
	or rebasing. :)	19:22
	does that help?	19:22
	the parent of a commit that we call "master" is not on any "branch" necessarily	19:23
	it's just the parent of a commit. :)	19:23
	so the names really do apply only to the heads.	19:23
felipec	Ilari: right, s/series/bunch/	19:23
	→ johnw joined	19:23
	← quabbles left	19:24
felipec	RandalSchwartz: I know how things work internally, I know what a ref is, and what a symbolic ref is, why branches are cheap, and why topic branches are a good idea	19:25
RandalSchwartz	then you should easily accept that "branch" means "head", and move on. :)	19:25
felipec	RandalSchwartz: when somebody is introduced to git, he doesn't need to how what a branch really _is_, just what a branch _means_	19:27
	← xkr47 left	19:27
felipec	RandalSchwartz: sure, it helps to know about refs and all that stuff, but that's not required	19:27
pasky	when somebody is introduced to version control, I believe "alternative line of development" is sufficient definition	19:28
	for a layman :)	19:28
	→ Sonderblade joined	19:28
Ilari	And looking at POV of "light cones" from SR, it can also be defined as "alternative history".	19:29
RandalSchwartz	SR?	19:29
	→ ElAngelo joined	19:29
ElAngelo	! [rejected] master -> master (non-fast forward)	19:29
	when i get this...	19:29
	what does it mean?	19:30
Ilari	ElAngelo: Did you rebase, amend or filter-branch carelessly?	19:30
ElAngelo	no	19:30
Ilari	ElAngelo: Maybe do 'git fetch oriign	19:30
felipec	RandalSchwartz: for example, if you use monotone, do you really need to know that a branch is a revision cert, and what is a revision cert? that multiple developers can set different key values for the same cert and so on?	19:30
Ilari	ElAngelo: Maybe do 'git fetch origin' and 'gitk master...origin/master' to investigate?	19:30
RandalSchwartz	I can't answer that, not being a monotone user	19:30
	but if it's important to the model, then yes, I need to know	19:31
	and I think coming into git thinking "branch" means "CVS branch" destroys good practices	19:31
Ilari	At least CVS actually had "branches"... :-)	19:31
	→ georgyberdyshev joined	19:31
	← gberdyshev_ left	19:31
	georgyberdyshev → gberdyshev_	19:32
RandalSchwartz	what tool generated that message, ElAngelo?	19:32
	I don't recognize it	19:32
	is that the output of git show-branch?	19:33
	felipec googles for git jargon	19:33
RandalSchwartz	google for gogopuffs!	19:33
felipec	I can't find a glossary of git jargon, just that the "GitSurvey2007" said "Less jargon", "Try to avoid jargon"	19:35
RandalSchwartz	yeah - always a good goal. rarely achievable in any discipline	19:36
ElAngelo	RandalSchwartz: just a git-push	19:36
	git 1.5.6	19:36
RandalSchwartz	imagine if two doctors said "you know, the bone that's just above the knee, ahh - push	19:37
ElAngelo	git push ssh://elangelo@foo-projects.org/var/git/lunar/moonbase.git master:master	19:37
RandalSchwartz	yeah - pushes must be fast forward	19:37
ElAngelo	?	19:37
RandalSchwartz	you need to update your local branch first	19:37
pasky	felipec: Documentation/glossary.txt could give a good general idea?	19:37
ElAngelo	i just did a git pull	19:37
RandalSchwartz	so that your commit is based on the current head of the remote	19:37
ElAngelo	there is nothing	19:37
	hmmm	19:37
RandalSchwartz	ignore that	19:37
ElAngelo	now it craps out on git-pull as well	19:37
RandalSchwartz	is your current commit you want to push a fast forward of the remote?	19:38
ElAngelo	i don't understand what a "fast forward" is	19:38
Ilari	ElAngelo: 'git pull ssh://elangelo@foo-projects.org/var/git/lunar/moonbase.git master'?	19:38
ElAngelo	i'm coding... not git'ing	19:38
RandalSchwartz	when you're adding commits to the end of a commit	19:38
felipec	branch: A "branch" is an active line of development. The most recent commit on a branch is referred to as the tip of that branch.	19:38
RandalSchwartz	ElAngelo . if the remote has commits a-b-c	19:39
	and you have a-b-c-d-e-f	19:39
	then it's a fast forward	19:39
ElAngelo	ah	19:39
	ok	19:39
	i get it	19:39
RandalSchwartz	but if the remote has a-b-c-1-2-3	19:39
	and you have a-b-c-d-e-f	19:39
	then f is not a descendant of 3	19:39
felipec	pasky: thanks ;)	19:39
	← _zack left	19:39
RandalSchwartz	so you can't push	19:39
	you need to fetch and rebase	19:39
	or merge	19:39
ElAngelo	ok	19:40
	now it worked	19:40
	thnx	19:40
	at least i understand why now :)	19:40
	RandalSchwartz bows	19:40
	RandalSchwartz hits head on table	19:40
RandalSchwartz	ouch!	19:40
jengelh	pasky: where to dump my files?	19:40
	→ stillLotR joined	19:43
	→ mikearr joined	19:47
	→ daaku joined	19:48
	← emachado left	19:48
	→ vasudeva joined	19:49
	← LotR left	19:49
vasudeva	i've just done an accidental 'git reset --hard' from my bash history. can i undo that?	19:49
Eridius\|work	no	19:50
vasudeva	damnit	19:50
	→ emachado joined	19:51
thiago	you only lost your uncommitted changes	19:52
	ckoehler → wecbot	19:53
	→ snailrails joined	19:53
	→ xkr47 joined	19:54
	wecbot → ckoehler	19:55
Bombe	Hmm... git-svn is segfaulting on one machine but not on another. Both machines have identical versions of git, svn, and neon installed. The only difference I can think of is the 32/64 bit issue.	19:56
	Is there any other, more sensible, explanation for that? :)	19:57
thiago	MacOS X?	19:57
Ilari	Bombe: Segfaulting on clone? Git configuration?	19:57
	→ rtomayko joined	19:57
Bombe	Both machines run Gentoo, segfaulting on fetch.	19:57
thiago	build everything in debug mode and valgrind	19:58
Bombe	Well, only one machine segfaults on fetch, the other runs perfectly fine.	19:58
	thiago, that’s the “I don’t really care” answer, isn’t it? :) I remember a thread on the git-list about that problem and even people that were a lot more proficient in perl than me could not find the problem, as it seems.	19:59
	So I’m a bit hesitant.	19:59
nbrunson	vasudeva: read http://www.kernel.org/pub/software/scm/git/docs/user-manual.html#recovering-lost-changes, I don't know if it will really help in your case, but you could play around with what they talk about in this section.	19:59
thiago	Bombe: valgrind will tell you where the problem is	20:00
	it could be perl or svn or git	20:00
freelock	ok, now I'm really confused...	20:00
thiago	more likely svn or neon	20:00
Bombe	So I need to rebuild them all, I guess?	20:00
	Hmm hmm...	20:00
freelock	I've created some bare repositories using git svn to clone the source repositories...	20:00
	Got 3 different projects I'm mirroring, but only one is actually getting new updates with git --bare svn fetch	20:01
	the other two have up to ~20 revisions that haven't been fetched, but git --bare svn fetch doesn't get them...	20:01
Bombe	Okay, crap.	20:02
	It’s something in my config, as it seems.	20:02
	← SRabbelier left	20:02
freelock	If I go to a repo with a working copy, git svn rebase gets them just fine	20:02
Bombe	git-svn fetch runs fine with a freshly created user.	20:02
	Damn.	20:02
	→ sverrej joined	20:02
Ilari	It could even be something in SVN config... :-/	20:02
Eridius\|work	perl version? version of libraries? maybe it's the svn libs and the swig bindings?	20:03
Bombe	Okay, removing ~/.subversion did not change a thing...	20:04
	← vasudeva left	20:04
Ilari	And hopefully there aren't any libs installed in home directory of that user... Library version incompatiblities can cause all sorts of trouble...	20:04
	← k0001 left	20:06
	→ k0001 joined	20:06
thiago	Bombe: try stracing	20:06
	maybe it will hint to where the problem is	20:06
Bombe	No, strace did not help, just a bunch of munmaps before the SIGSEGV.	20:07
	Also, ~/lib does not contain anything of neon, svn, or git.	20:07
	← zacharyc left	20:09
	← aroben left	20:09
	→ aroben joined	20:10
	→ SRabbelier joined	20:10
thiago	gdb or valgrind then	20:10
	try valgrinding anyways, maybe it shows something useful	20:10
	→ eikonos joined	20:11
pasky	jengelh: huh?	20:12
	→ lloop joined	20:14
	← eikonos left	20:14
lloop	how do i import a local repo to a remote repo?	20:15
	I'm trying to convert from svn. git-svn pulled down changes and I have a git repo setup	20:15
	→ eikonos joined	20:15
Pieter	Ah, I found something better than git	20:17
	Dioblo 3 \o/	20:17
	← CodeOfficer left	20:18
	← seanius left	20:19
	← snailrails left	20:20
	→ elight joined	20:20
	→ seanius joined	20:20
felipec	Pieter: better? you should integrate them :)	20:21
Pieter	git-whirlhi	20:22
	git-whirlwind :)	20:22
	→ sharp15 joined	20:22
Bombe	Hmm, now it doesn’t even segfault on a fresh clone of another project so I guess the repository I’m trying to git-svn-fetch into is somehow b0rked.	20:23
	← Fullmoon left	20:23
	stillLotR → LotR	20:23
ElAngelo	Your branch is ahead of the tracked remote branch 'origin/master' by 9 commits.	20:23
	← emachado left	20:23
sharp15	i accidentally added more files than i wanted to (using git-add). i wanted to put a commit message for each file. no commit has been done. is it simple to undo this or not?	20:23
ElAngelo	what does this mean?	20:24
Ilari	Not just Neon, SVN or Git related stuff... Perl related stuff might break it also (actually pretty much if it loads any badly built library from there)...	20:24
	ElAngelo: You have 9 commits you haven't pushed into origin yet...	20:25
ElAngelo	hmmm	20:26
	awkward	20:26
	then why doesn't git-push reveal that to me?	20:27
Ilari	ElAngelo: Was that one push one-off as you put full URL there?	20:27
ElAngelo	$ git-push-master	20:27
	Everything up-to-date	20:27
	$ cat ~/bin/git-push-master	20:27
	git push ssh://elangelo@foo-projects.org/var/git/lunar/moonbase.git master:master	20:27
felipec	sharp15: git status tells you what to do	20:27
Ilari	ElAngelo: 'git push origin'?	20:27
ElAngelo	$ git push origin	20:27
	Everything up-to-date	20:27
	→ ketralnis_ joined	20:28
felipec	sharp15: (use "git reset HEAD <file>..." to unstage)	20:28
Ilari	ElAngelo: Maybe try 'git fetch origin'?	20:28
	→ Cacher joined	20:28
Ilari	ElAngelo: Using URLs to refer to repositories which also have nicknames (like 'origin') is not recommended.	20:28
Cacher	so I am on master branch, I make a couple of changes, don't add or commit yet, and I decide that these changes I just made could go on a new branch. . . how is this done?	20:29
ElAngelo	normally i only use full urls	20:29
	← Sho_ left	20:29
	→ thannoy_ joined	20:29
Ilari	ElAngelo: Full URLs are pretty much only for one-off operations and for registering remote repo nicks...	20:30
	ElAngelo: And also for communicating to others...	20:30
ElAngelo	i still prefer it honestly	20:30
sharp15	felipec: thank you. i hadn't really understood what that did.	20:30
ElAngelo	i just have a bunch of scripts that i should use	20:30
	→ jordi_bunster joined	20:30
felipec	sharp15: perhaps a git-unstage command would make sense, I might do write it sometime :)	20:31
	← Yuuhi left	20:31
Bombe	Cacher, if you haven’t committed them yet, just create a new branch with "git branch <new-branch>", your changes will be left in the working tree and you can commit them to the new branch.	20:31
Ilari	ElAngelo: That it displayed that message about being 9 commits ahead was apparently git getting confused because one pushed by URL instead of by nick...	20:32
Cacher	thanks Bombe, i thought that would error	20:33
	→ Sho_ joined	20:33
	← thannoy left	20:33
Bombe	Cacher, it would only refuse to switch branches if the new branch already exists and already contains the files you changed.	20:33
	Cacher, oh, you need to "git checkout <new-branch>" after creating it, sorry. Forgot that. :)	20:34
	→ jacobat joined	20:35
	← l0verb0y left	20:36
	→ noTV joined	20:37
	→ dasil003 joined	20:38
paakku	Bombe: there's also git checkout -b which creates the branch and checks it out right away.	20:39
Bombe	paakku, indeed there is.	20:40
	Didn’t think of that.	20:40
dasil003	Can anyone tell me the proper way to `git reset --hard -- path/to/file`?	20:41
Ilari	dasil003: 'git checkout HEAD -- path/to/file'.	20:41
dasil003	thanks	20:42
	← ketralnis left	20:43
	← sergiodj left	20:45
	→ xkr47^ joined	20:45
	← xkr47^ left	20:46
	← thannoy_ left	20:46
	→ \|Ryan52 joined	20:49
	← janm left	20:49
	← vbgunz left	20:49
\|Ryan52	I have a directory mirrors/debian that is under revision control using git. I want all of mirrors/ to be. How do I do that?	20:50
	blowery → blowery`afk	20:52
thiago	keeping the same repository?	20:53
\|Ryan52	ya.	20:53
thiago	cd mirrors/debian; mkdir debian; mv * debian; mv .git ..; cd ..; git add *; git commit	20:54
	← ^Beket\| left	20:54
thiago	oh, you will want to move debian/debian back to debian before doing the git add	20:54
\|Ryan52	ya, I get it. I was just wondering if there was some cool git way.	20:55
thiago	this is it	20:55
	there's a more complex way but it's unnecessary pain	20:55
\|Ryan52	oh.	20:56
	← p4tux left	20:59
	→ p4tux joined	21:00
	← CIA-2 left	21:00
	← cannonball left	21:04
	→ Fullmoon joined	21:05
	→ chrisb joined	21:05
	→ CIA-2 joined	21:06
	→ raalkml joined	21:08
	← felipec left	21:08
	← lorandi left	21:09
	← Eridius\|work left	21:09
	blowery`afk → blowery	21:09
	→ lwhacker joined	21:10
	← csc` left	21:12
	← d0k left	21:13
	→ ijcd joined	21:14
	← elight left	21:14
	→ shd joined	21:15
	→ csc` joined	21:15
	→ d0k joined	21:15
	→ scrooloose joined	21:18
	→ Eridius\|work joined	21:18
	← Sonderblade left	21:20
	← lloop left	21:21
	← Eridius\|work left	21:22
	→ Eridius\|work joined	21:23
	← cncfanatics left	21:24
	← sharp15 left	21:25
	← peritus_ left	21:25
	← SRabbelier left	21:25
	→ peritus_ joined	21:26
	→ clairvy joined	21:26
	→ SRabbelier1 joined	21:26
	SRabbelier1 → SRabbelier	21:26
	→ foca joined	21:28
	→ artisonian joined	21:28
	← rdavila left	21:29
	← SRabbelier left	21:30
	← ketralnis_ left	21:30
	→ rdavila joined	21:30
	→ SRabbelier1 joined	21:30
	SRabbelier1 → SRabbelier	21:30
	→ ketralnis joined	21:30
	← jordi_bunster left	21:31
	← elliottcable left	21:32
	→ Eridius\|work_ joined	21:33
	→ kraymer_ joined	21:34
	→ trochala joined	21:34
	→ yatiohi joined	21:34
	← yatiohi left	21:35
	← lea_w left	21:35
	← \|Ryan52 left	21:37
	→ peritus__ joined	21:40
	← Eridius\|work left	21:41
	Eridius\|work_ → Eridius\|work	21:41
	← rdavila left	21:41
	← kraymer left	21:41
	→ alegscogs joined	21:42
	← krawek left	21:45
	→ peeja joined	21:45
plopix	was "git push --mirror" broken in 1.5.4.3?	21:45
	← vbabiy left	21:45
	← peritus_ left	21:48
Ilari	plopix: No idea. What kind of 'breakage' are you seeing?	21:48
	← SuttoL_ left	21:48
	→ krawek joined	21:49
	← trochala left	21:51
	← nbrunson left	21:52
	→ SuttoL joined	21:52
	→ ebel joined	21:53
	← kumbayo left	21:55
eMBee	good morning	21:56
	→ peeja_ joined	21:57
	← peeja left	21:57
	← raalkml left	21:58
plopix	Ilari: I am getting the help page if I add -v	21:59
	← aroben left	22:01
	→ aroben joined	22:01
eMBee	git clone --no-hardlinks does not do what is expected if the source is itself a shared clone.	22:03
	← krh left	22:03
eMBee	how can i make a clone that is not shared from a shared repo?	22:03
	eMBee is now faking it by running ssh through localhost	22:03
vmiklos	why not file:///?	22:03
eMBee	hmm?	22:04
	file:/// as opposed to just the path?	22:05
Pieter	yes	22:05
eMBee	ok, i didn't know about that option	22:05
	and even if i knew i would not have thought that this would make a difference	22:08
	← Ademan left	22:09
	← kraymer_ left	22:10
eMBee	how much of git depends on perl?	22:10
	i just installed git on the xo which does not have perl installed (and i forced rpm to ignore the depencendy on perl)	22:11
RandalSchwartz	the XO?	22:11
	git-svn needs perl	22:11
eMBee	the olpc project	22:11
RandalSchwartz	I think that's the only part of git that left	22:11
	I know what XO is :)	22:12
	just amazed you're using one	22:12
eMBee	ah, heh	22:12
RandalSchwartz	the XO has Squeak too	22:12
	← queso left	22:12
	eMBee nods	22:12
RandalSchwartz	so 100K new kids now have exposure to Smalltalk	22:12
	this is good	22:12
	eMBee has been playing with squeak already a few years ago, teaching some kids. unfortunately i could not keep up with them so they lost interest	22:13
eMBee	git on the xo is 1.5.3 and it has a perl dependency on git-core	22:14
	← ph^ left	22:14
paakku	git-relink and git-add--interactive are in Perl	22:14
eMBee	but git-svn is extra, so maybe that dependency is a bug	22:14
paakku	gitweb too	22:16
eMBee	ok, but gitweb should also be an extra package	22:16
paakku	and many tests and contribs and steps in the build system	22:16
	← bentob0x left	22:16
	← peritus__ left	22:17
paakku	hm, there is @@PERL@@ in git-bisect.sh too	22:17
Pieter	who pays $59 for a subversion client :/	22:17
	← Arrowmaster left	22:17
	→ Arrowmaster joined	22:18
eMBee	ok, the xo is prepared to do some offline development while traveling. time to pack and leave	22:20
	← Arrowmaster left	22:21
	→ njal joined	22:24
njal	when I do "git reset --hard HEAD" in a branch, git status is supposed to be empty?	22:24
	← deavidsedice left	22:24
paakku	there might be files that are not in the index; IIRC git reset --hard does not delete those, but you can get them with git clean	22:25
cardioid	That might not affect untracked files	22:25
	What paakku said	22:25
	gebi_ → gebi	22:25
njal	hmm.. ok, because I changed branches from my branch to master and back again, and got a ton of files listed under "# Changed but not updated:" when using git status, but nothing shows up for a "git diff"	22:27
	"git clean -f" didnt seem to do anything either, so now Im completely fumbled and dont know what to do. git diff shows no output at all, but git status gives a huuuge list of all files I've touched in my branch + a few more that has been changed in master during my work on my branch	22:29
	← BabelO left	22:30
paakku	might that be caused by the clock skipping backwards? possibly git add -u would help?	22:31
	is that in a local directory or something like NFS?	22:32
	← peeja_ left	22:32
njal	its over nfs	22:33
	→ peeja joined	22:33
njal	doing a git add -p -u now	22:33
	← csc` left	22:34
njal	and its basically printing a ton of "diff --git a/.... b/....", but nothing more	22:34
	and git still puts the files under "changed, but not updated"	22:34
gitster	try "git update-index --refresh" first. It is never needed but I am suggesting this only for diagnosing what is broken in your environment.	22:35
njal	seems like its NFS that is borking it for git.. ssh'd to the server I'm mounting it from, and it seems to be ok there	22:35
	huh.. "git update-index --refresh" ran instantly and it's back to where I was hoping it would be	22:36
	→ csc` joined	22:36
	← eikke left	22:36
	blowery → blowery`afk	22:38
	→ _zack joined	22:38
njal	on second thought, it doesnt seem that update-index had anything to do with it, but going into the server with the nfs share and writing "git status" seems to fix whatever is broken	22:40
	→ powr-toc joined	22:41
	← lolage0 left	22:43
	← _zack left	22:43
	← p4tux left	22:49
	← artisonian left	22:50
	→ daniel_b joined	22:54
	→ insin joined	22:55
	→ cousin_it joined	22:56
	← peeja left	22:58
	← juanjoc left	22:59
	← SRabbelier left	22:59
	← clairvy left	23:00
	→ slink joined	23:01
	← slink left	23:02
njal	next (and hopefully last) question of the night: when I've done a successful git rebase, my branch and the remote tracked branche have diverged.. do I just do a git push -f, or is there something more graceful?	23:02
	← pygi left	23:03
	→ pygi joined	23:03
gitte	njal: you avoid pushing rewritten history.	23:03
njal	so my error was pushing before the rebase?	23:04
gitte	(IOW: you neither push -f, nor try to play cute games)	23:04
	njal: well, keep in mind: everybody pulling from you, who pulled what you have now rewritten, will either have a totally bogus merge and eternally have a divergent local copy of your branch, or will have to do a lot of cleanup to do herself.	23:05
njal	hmm.. well, I was told I could do whatever with my branch by my workmate, and he wanted me to push stuff so he could see what I was up to	23:06
gitte	But yes, if you tell people that you rebased that branch, and that they should make sure that they rebase their stuff on top of your new version, you can "push -f".	23:06
	← cousin_it left	23:06
njal	so in this case, the branch is supposed to be just that: complete work in progress	23:06
gitte	Well, then.	23:06
	push -f is as graceful as it gets.	23:07
	→ johnw_ joined	23:08
	← Fullmoon left	23:09
	← johnw left	23:10
	→ johnw__ joined	23:10
sbahra	"I was wondering if version 1.5.6 would be alright on this. I took a look at 1.5.6.1 and it seems to have introduced a number of bugs in the configure script, causing the relocation of the install to do some odd things. 1.5.6 appears to work correctly though. Please let me know if that sounds reasonable and I'll go ahead with that version."	23:11
	Is this a known issue?	23:11
	→ hpa_ joined	23:12
gitster	heh, gitte writing in Python --- that's a change ;-)	23:13
	→ vbabiy joined	23:13
gitte	Hey, I have a hard time already ;-)	23:13
	s/already/&, give me a break/	23:13
	→ spearce joined	23:14
gitster	I am not complaining. Looking forward to another importer from you, this time in Ruby ;-)	23:14
hpa_	A feature request: an option to git-apply/git-am to apply a patch to a specified subtree	23:14
	← eternaleye left	23:14
hpa_	with patch(1) one just cd's to the subdirectory before applying	23:14
	I don't think that works with git	23:14
	← ijcd left	23:14
gitster	Yeah, I've heard that one before. What should be the interface?	23:15
gitte	gitster: don't tempt me, or you'll end up with an importer in PostScript.	23:15
	→ ijcd joined	23:15
gitster	"git apply -p$n" can chomp leading paths after a/ and b/; but...	23:15
hpa_	gitster: How about git {am\|apply} --root=foo/bar	23:15
	← charon left	23:15
gitster	So the assumption/limitation is that everything in that patch is rooted at the same "elsewhere"?	23:15
hpa_	Yes	23:15
	That's the common case	23:16
sbahra	hpa_, I am not yet this experienced with git. Noting this, how is that different than cherry-pick?	23:16
gitte	Sort of a "-s subtree"?	23:16
	sbahra: hpa is working with patches.	23:16
	sbahra: they are no commits yet.	23:16
sbahra	ahh	23:16
	ok	23:16
gitster	Sounds good. "sed -e '/^diff --git/s\|[ab]/\|&foo/bar/\|' patch \| git apply"?	23:16
	← aroben left	23:17
hpa_	So in the case of "it apply --root=foo/bar -p2" a hunk with a/baz/quux/xyzzy.c" would be applied to foo/bar/quux/xyzzy.c	23:17
	gitster: Yup, pretty much; except you have to change the +++ and --- lines and so on	23:18
gitster	Sounds sane enough and simple enough to explain to the users.	23:18
hpa_	:)	23:20
gitte	Actually, I think that this "-p" takes precedence over "--root" makes the patch harder.	23:23
pygi	hey spearce :)	23:23
spearce	hey pygi	23:23
	← nud left	23:24
hpa_	gitte: Actually you need both	23:24
robinr	spearce: pls double check that I didn't screw up master (I don't think I did , but...)	23:24
spearce	robinr: i'll take a peek in a second.	23:25
gitte	hpa_: sorry, I did not mean precedence.	23:25
robinr	there were more branches than I'm used to	23:25
spearce	i've been busy with relocating my domain between servers and half the net caught up today so i just found half my email spread around. yay.	23:25
gitte	hpa_: I meant "is applied before", but as I am not a native speaker, I cannot express myself.	23:25
hpa_	gitte: Ah yes	23:25
	Yes, and it really does need to work that way	23:26
robinr	or maybe, not come think of how many I have at work, but then I don't rebase tgere	23:26
gitte	hpa_: and I am full of sh*t... find_name() must be patched, and you're done.	23:26
hpa_	Heh	23:26
	→ zenspider joined	23:26
	← ckoehler left	23:26
spearce	robinr: at least you have my index pack data corruption fix in there. :-)	23:27
zenspider	what is the git equivalent to: `svn up -r PREV; ...; svn up` ? meaning, I want to back up at least 1 revision, do stuff, and then move forward as normal. The move forward as normal is `git pull`, but that doesn't work if I use `git reset` to do the backing up.	23:28
	moving by date would be fine too	23:28
robinr	I've been using jgit to push sometimes now.	23:28
	← johnw_ left	23:28
zenspider	(all I want to do is simulate having an out of date clone)	23:29
spearce	robinr: me too actually. been using it a lot at day-job and missing your console password prompt thing. you are right, using awt is damn annoying.	23:29
RandalSchwartz	you mean you want to commit based on the parent commit of the current upstream head?	23:29
robinr	impressive work shown	23:29
spearce	robinr: i think master looks good. i'm diffing it against what i had locally (and expected to see) and it looks good	23:29
robinr	s/a/	23:29
	→ decostop joined	23:29
	→ eternaleye joined	23:29
RandalSchwartz	git checkout -b mywork origin/master^	23:30
	not sure why you'd want to base on downrev stuff though	23:30
spearce	most of the push code is zawir's, i just patched in dumb transport support.	23:30
	the s3 stuff has been nice lately. i'm backing up my server's /etc/ through git to s3, encrypted. yay.	23:30
robinr	spearce: yes, true	23:30
zenspider	RandalSchwartz: I'm not committing anything... I'm just rolling back in time to simulate work to be done in a CI system	23:30
RandalSchwartz	I don't know what CI is	23:31
robinr	would push over smb be hard?	23:31
RandalSchwartz	so I can't tel if that would help	23:31
	→ mithro joined	23:31
spearce	robinr: not if we had an SMB client in java, no	23:31
robinr	i.e. without mounting shares	23:31
zenspider	RandalSchwartz: continuous integration build/test system	23:31
	I just want to back up in time N commits	23:31
spearce	robinr: we have two very different protocols already supported. its only about 200 lines of code for both push and fetch, if we have a library to read/write files for us.	23:31
RandalSchwartz	N of your comits	23:31
	or upstream commits	23:31
spearce	robinr: the S3 code took more because i had to do a full s3 client	23:32
	→ georgyberdyshev joined	23:32
	← gberdyshev_ left	23:32
zenspider	again, I'm not committing ANYTHING, ever... so upstream sounds right	23:32
RandalSchwartz	since commits aren't linear, N doesn't make sense then	23:32
	georgyberdyshev → gberdyshev_	23:32
robinr	spearce: jcifs is LGPL	23:32
RandalSchwartz	given a directed graph of parent commits, what do you want to see?	23:32
spearce	robinr: so can we link to it with our bsd code? :)	23:32
RandalSchwartz	since the top-level commit remotely could in fact have N parents	23:32
robinr	spearce: yes	23:33
zenspider	they're linear in git log... I don't really care how it is described... I just need to back up	23:33
RandalSchwartz	they have the illusion of being linear in git-log	23:33
zenspider	RandalSchwartz: ok. let me descibe it differently... I need to execute SOMETHING such that my checkout will guarantee to do SOMETHING when I do 'git pull'	23:33
	→ elight joined	23:34
	→ Bryson joined	23:34
RandalSchwartz	other than create an artificial merge?	23:34
robinr	we should be able to get the push and pull into IRC or something too :)	23:34
RandalSchwartz	because that's what pull will do if you have something old checked out	23:34
spearce	pack files in irc aren't user friendly. base 64 encode them first? :)	23:34
RandalSchwartz	and that might even conflict	23:34
	I guess I need to udnerstand your goals better	23:34
robinr	spearce: bundles?	23:35
zenspider	RandalSchwartz: what's artificial about it? I thought that was the point of git pull?	23:35
RandalSchwartz	git-pull is not always a safe operation	23:35
	no	23:35
zenspider	how can it conflict if I've NEVER committed anything?	23:35
RandalSchwartz	in fact, I suggest people stay away from git-pull	23:35
spearce	robinr: yes, we can support a bundle now. we just can't do it in push, for the same reason C Git doesn't do push to a bundle.	23:35
RandalSchwartz	pull = fetch + merge	23:35
gitte	robinr, spearce: why not fast-export format?	23:35
RandalSchwartz	do you really want to merge?	23:35
	or just fetch, and then check out the latest?	23:35
spearce	gitte: because we don't need one right now? and we have bigger fish to fry?	23:35
	← insin left	23:36
gitte	spearce: I thought you wanted to push into IRC... /me shuts up.	23:36
zenspider	in the case of a clone with zero local commits, does it even matter?	23:36
RandalSchwartz	Yes	23:36
paakku	I'm tweaking a script that makes elinks-*.tar.gz. I'd like to give users an easy way to find the commit ID from which the archive was made. I see git-archive can put this info in a pax header; however the tar file needs to contain also generated files that are not in git, so I can't use git-archive directly. I tried adding a similar header with GNU tar but --format=pax makes it also store unnecessary atime info for every file. Then	23:36
	I tried adding the commit id as .git/HEAD but that broke Debian's git-import-orig script. So I'm thinking the commit ID needs to go in a regular file not in .git/. Is there a conventional name for such a file?	23:36
RandalSchwartz	again = pull = fetch + merge	23:36
spearce	gitte: i won't stop someone from contributing the code (so long as its reasonably well written), but i want to focus on things like merge that impact most users. i did S3 support because _I_ wanted to use it as a backup solution, and it looked really easy once I had sftp push done, which I did as a proof of concept to show that marek's work was extensible to other protocols.	23:36
robinr	spearce: we need some negotiations. Maybe easier to do in jabber	23:37
spearce	(and thus validate his design)	23:37
zenspider	RandalSchwartz: goal: have some work for the continuous build system to do. At this point, I don't care how.	23:37
RandalSchwartz	if you create a merge commit, you'll have to commit it	23:37
zenspider	where a unit of work == a commit	23:37
RandalSchwartz	imagine the following scenario	23:37
	(and now to come up with it)	23:37
	a b c d e	23:37
gitte	spearce: I think it is nice. But why not use Google App? ;-)	23:37
RandalSchwartz	where c is actually a merge with x y	23:37
	so c = y + b	23:37
	ok	23:37
spearce	gitte: google app isn't bulk data storage like s3 is. :-)	23:37
RandalSchwartz	hard to draw linear	23:37
	now you check out a	23:38
	build with it	23:38
	and it's time to pull new work	23:38
gitte	spearce: but you get some space for free, no?	23:38
RandalSchwartz	that would be f	23:38
	but you'll end up with a merge between a + f	23:38
gitte	spearce: I dream of a gmail backend...	23:38
RandalSchwartz	because it's not a "fast forward" from a to f	23:38
vmiklos	gitte: what's the recommended way to check if an sha1 (unsigned char[20]) is set? it's a global variable so i know it's zeroed out till it's not set, but maybe it's still better to use an int is_foo_set helper variable?	23:38
	→ proxie joined	23:38
	← agib left	23:38
spearce	gitte: true, 500M, but you must write python code to talk to the storage layer and bridge the java/c client into it. s3 is a straight http service, with no additional smarts.	23:38
gitte	vmiklos: is_null_sha() I believe.	23:38
RandalSchwartz	anyway, what you want is "fetch + checkout latest"	23:38
gitte	vmiklos: is_null_sha1().	23:39
spearce	gitte: i actually am waiting for google app engine to support java. then we can run jgit there, and have smart git hosting on their cluster hardware. even if code.google.com doesn't.	23:39
robinr	well, I'm in another time zone. good night.	23:39
RandalSchwartz	git fetch; git checkout origin/master	23:39
	like that	23:39
spearce	night robinr.	23:39
vmiklos	gitte: aah, great! :)	23:39
spearce	thanks for updating master.	23:39
RandalSchwartz	no PULL	23:39
	ever	23:39
gitte	spearce: right.	23:39
	vmiklos: np	23:39
Ilari	Pull is handy thing if configured correctly...	23:39
zenspider	RandalSchwartz: origin/master is the latest, right?	23:39
RandalSchwartz	it might be	23:39
	depends on your upstream	23:39
	and how they are setting things up	23:40
	← ebel left	23:40
RandalSchwartz	the names are all arbitrary	23:40
	just saying the default names	23:40
	go ask your upstream :)	23:40
zenspider	RandalSchwartz: so how do I checkout 3 previous to origin/master?	23:40
	← elight left	23:40
robinr	spearce: you might want to try the update site now	23:40
zenspider	will HEAD~3 work there?	23:40
RandalSchwartz	"3 previous" doesn't make sense	23:40
	it's not linear	23:40
	suppose origin/master is a merge	23:40
	which parent do you use?	23:40
	→ elight joined	23:40
RandalSchwartz	git maintains directed graphs	23:40
	← robinr left	23:41
RandalSchwartz	not linear lines	23:41
	it's like asking "I want the third child directory of /"	23:41
	doesn't make sense	23:41
	since / can contain more than one dir	23:41
	and each of those can contain more than one dir	23:41
	→ Arrowmaster joined	23:41
	← drizzd left	23:41
RandalSchwartz	think "DAG"	23:41
	← k0001 left	23:42
	← EmilMedve left	23:42
paakku	will it necessarily be dAg if there are grafts?	23:42
	RandalSchwartz wonders why this is such a hard concept for people to get	23:42
	→ CodeOfficer joined	23:42
zenspider	RandalSchwartz: I'm fine with DAGs, I just think you're mentarbating a scenario I don't need to deal with. The reality is that there IS a linear flow here even if it is collected into a DAG. Hypothetically, yes... Realistically, no.	23:43
RandalSchwartz	paakku - troublemaker. :)	23:43
zenspider	To pretend there isn't a flow of time in this repro is useless. I want to check out every commit upstream and test it and report the results... this isn't rocket science.	23:44
RandalSchwartz	zenspider - i think you have to think of the general though, to keep from making presumptions that sometimes will not be true	23:44
	so that you can come up with a pattern that will always work	23:44
paakku	oh well, I'll put the commit id in a "git-commit-id" file; I couldn't decide whether the file should be hidden or in all caps, so I guess it should be neither.	23:44
	← d0k left	23:44
RandalSchwartz	"every" upstream will require doing some sort of node traversal	23:44
	so work out a strategy for that	23:45
	depth-first? likely not	23:45
	→ pergesu joined	23:45
RandalSchwartz	so breadth-first, probably	23:45
zenspider	RandalSchwartz: I'd rather not, honestly. I'd rather have something I can get done today that is good enough. We don't have a clusterfuck development model, so I think it is fine to treat this as linear.	23:45
RandalSchwartz	then don't ask me. work it out using your own presumptions.	23:45
zenspider	If we did have branching and merging going on all over the place, I'd understand the need for the extra effort, but we don't	23:45
RandalSchwartz	I can only speak for how to do it properly	23:46
	not for your own narrow assumptions	23:46
Eridius\|work	zenspider: by "we don't have a clusterfuck development model" do you mean "we don't use branches"?	23:46
RandalSchwartz	since I don't know your particular flow	23:46
	you asked me how to do something with git	23:46
	I'm answering how to do it with git	23:46
zenspider	Eridius\|work: no, I mean we've got a lot of discipline in how our tree is structured. and how it is merged	23:47
RandalSchwartz	if you have a particular way of misusing git, then you need to address that within your own thinking.	23:47
	← radarek left	23:47
RandalSchwartz	but in general, presuming some sort of "linear development" with git for a general tool is likely to smack you full speed into a wall at some point	23:47
Eridius\|work	zenspider: but you are using merges?	23:47
RandalSchwartz	and I won't help you with that.	23:47
zenspider	RandalSchwartz: that's a bit presumptuous... I asked the channel what the equivalent of a very basic svn command was.	23:47
	RandalSchwartz: that's fine. I understand. thanks.	23:48
Eridius\|work	zenspider: but there's no 1-to-1 equivalence	23:48
RandalSchwartz	it may be that there's no equivalent	23:48
zenspider	nod and I suspected that might be the case	23:48
RandalSchwartz	it's like asking "what's the equivalent to 'pull back on the stick in my airplane' for a car?"	23:48
	it's a non-starting question	23:48
zenspider	the cheapest way for me to do this is to copy off the repo I have and wait a day	23:49
RandalSchwartz	and presumes a broken view of the world	23:49
	I was trying to correct that view.	23:49
	← cmarcelo left	23:49
RandalSchwartz	perhaps you are uncorrectable. :)	23:49
	← foca left	23:50
zenspider	Eridius\|work: I'm doing pulls off of our master currently, but need to be able to "back up" in time so I can test out the latest changes to my script against some units of work (commits)	23:50
	RandalSchwartz: given your broken metaphor, quite possibly.	23:51
Eridius\|work	zenspider: what exactly do you want to do?	23:51
	← vbabiy left	23:51
Eridius\|work	HEAD~3 will give you the commit 3 down the tree from HEAD, following only the first parent	23:51
	if you want the 3rd commit down the tree, going purely in date order, try `git rev-list HEAD --date-order -n4 \| tail -1`	23:52
zenspider	Eridius\|work: I'd like to back up my repo to a previous state. N commits in git log for example. but do so in such a way that the rest of my script can go unchanged. I'm currently using `git pull` to grab updates from upstream	23:52
RandalSchwartz	keeping in mind "date order" is expensive	23:52
zenspider	RandalSchwartz hinted that might be a bad thing to do, but given that this is a read-only depot, I'm guessing that it is still ok	23:52
	← priidu left	23:52
Eridius\|work	N commits in git log? git rev-list -n# \| tail -1	23:53
zenspider	expensive is fine... this is bound by CPU on the test side	23:53
	← powr-toc left	23:53
	→ powr-toc joined	23:53
Eridius\|work	the real question here is how you want to handle merges when going backwards	23:53
zenspider	Eridius\|work: yes, that identifies it, but I want it to be the current state in the depot...	23:54
	← bryanray left	23:54
powr-toc	I get an error when I try to do a merge from a remote branch: remote: error: remote: unable to find 9a8d3...remote:	23:54
	any ideas?	23:54
Eridius\|work	git reset --hard `command to get sha1 here`	23:54
zenspider	at this stage, I think holding off for a day and doing rsync's might be easier...	23:54
	← Bryson left	23:54
zenspider	Eridius\|work: I've got that currently, but that totally borks a subsequent git pull	23:54
	→ Bryson joined	23:55
Eridius\|work	well yeah, because the pull will try and merge	23:55
RandalSchwartz	and that's why I said "do not do a pull"	23:55
Eridius\|work	well, probably fast-forward	23:55
RandalSchwartz	and I'll repeat that again	23:55
Eridius\|work	why the hell are you trying to back up and then pull? it doesn't make sense	23:55
RandalSchwartz	git fetch; git checkout $SHA1	23:55
	your job is to compute SHA1	23:55
	you really are untrainable.	23:55
	since you don't pay attention to the things I'm saying, even when I'm trying to make sense and not just tell you you're off in the deep end.	23:56
	RandalSchwartz stops trying to teach the pig to sing	23:56
zenspider	RandalSchwartz: actually, what you missed telling me until now was the git checkout $SHA1 part... since that isn't in the help output it wasn't obvious.	23:57
	so that'll help, thanks.	23:57
RandalSchwartz	do I need to scroll back	23:57
VVSiz	zenspider: don't listen to RandalSchwartz, he's definitely over-complicating things!	23:57
RandalSchwartz	.... git fetch; git checkout origin/master	23:57
	dammit said it	23:57
	pay f'ing attention	23:57
	read the scroll	23:57
	it's in there	23:57
zenspider	RandalSchwartz: quit being a bitch. I can search backwards just fine.	23:58
RandalSchwartz	so don't say "you didn't say it"	23:58
VVSiz	if git bisect can effectively bisect between two revisions, then why not check out any revision between those two points	23:58
zenspider	origin/master is not the same, is it? I even asked and you gave me another non-answer.	23:58
RandalSchwartz	because it might not be correct	23:58
	you'll have to ask upstream	23:58
	and I said THAT TOO	23:58
	ASK YOUR UPSTREAM WHAT THE LATEST IS	23:58
	how much clearer do I need to make that?	23:58
	I've now gone from generic, to a sample specific, with qualifications for why the specific might be wrong	23:59
	and I've done that twice	23:59
powr-toc	Any idea why I get the following error when doing a fetch on a remote branch? remote: fatal: remote: unable to get type of object 9a8d39...remote: remote: aborting due to possible repository corruption on the remote side.	23:59

Logs Search ←Prev date Next date→

Channels Documentation