IRCloggy #git 2008-05-31

Provider of IRC logs since 2005.

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

	2008-05-31
	→ juanjoc joined	00:04
	← piclez left	00:04
	→ spearce` joined	00:05
	← juanjoc left	00:07
Eridius\|work	wtf? git-read-tree -m isn't merging	00:07
	I'm trying to construct a very specific object graph in a temp repo, and so I'm trying to build a merge by hand	00:08
	I have two parentless commits which each create a different file	00:08
	git read-tree A; git read-tree -m B	00:08
	I would expect the index to now have both files, but no	00:08
vmiklos	why do you use read-tree for that?	00:08
	→ david_koontz joined	00:08
vmiklos	git-merge handles this	00:09
Eridius\|work	oh wait, I must be misunderstanding read-tree. apparently I can pass multiple tree-ishs	00:09
	vmiklos: I'm trying to ensure I understand every detail of the commit obj construction, so I'm using plumbing	00:09
vmiklos	then see git-merge's source on what read-tree parameters it uses :)	00:09
DrNick	skip down to the MERGING section of the read-tree manual	00:10
Eridius\|work	I'm reading that now	00:10
	→ rwanderley joined	00:11
	→ ekidd joined	00:12
	← devogon left	00:12
	← devogon__ left	00:12
	← ben_h left	00:13
	→ brosner joined	00:20
	← spearce left	00:22
	← mattf left	00:24
	pasky is maybe for the first time ever writing some gui application for real	00:24
mutex	ummm	00:25
pasky	GTK such so much, but still it's strangely enjoyable :)	00:25
	*sucks	00:25
mutex	can I convince git apply to patch even if there is a conflict in the index ?	00:25
	err git stash apply	00:25
	← pombreda left	00:26
vmiklos	that would be bad.	00:27
	you should not have conflicts when you apply anything, as that could lead to a new conflict	00:27
mutex	well	00:27
	i cherry picked a fix from another branch	00:28
dotsintacks	how do I "checkout" a remote branch?	00:28
mutex	after stashing my work	00:28
	then tried to reapply teh stash	00:28
dotsintacks	so that I can push changes back to the branch	00:28
vmiklos	git checkout -b localbranch remote/branch; hack hack hack; commit; push	00:28
mutex	my 'fix' is to go back to previous commit-id, apply stash, reset index, stash, change to cherry commit, apply stash(no index)	00:29
vmiklos	or maybe just commit after cherry-pick; stash apply; commit; and git rebase -i HEAD~2 and squash	00:29
	← ekidd left	00:29
	mutex head explodes	00:30
mutex	heh	00:31
doener	git doesn't support that, can only detach heads, not make them explode	00:31
	mutex snickers	00:31
	→ rubbish joined	00:32
	← redondos left	00:33
	→ mattf joined	00:35
	→ pombreda joined	00:35
mutex	can one reset a particular file in the index ?	00:41
	...	00:41
	mutex sees it on the man page the second after he asks	00:41
	← BabelO left	00:42
	← datafirm left	00:42
	← pombreda left	00:43
	→ pombreda joined	00:44
vmiklos	:)	00:46
	← tvachon left	00:47
mutex	so what does '--' stand for	00:47
Mikachu	end of revisions, start of files	00:48
mutex	ah	00:48
Mikachu	or if you prefer, "everything to the left is a revision, everything to the right is a file"	00:48
mutex	makes sense	00:48
Mikachu	sometimes just a trailing -- makes a command work where it wouldn't without it	00:48
	← madewokherd left	00:49
Mikachu	touch master, git checkout master doesn't work, but git checkout master -- does work	00:49
	→ a-priori_ joined	00:50
Mikachu	actually that might work, but there's some command that breaks when you touch a file with the same name as a branch :)	00:51
	i think show breaks	00:51
	→ threeve joined	00:59
	→ madewokherd joined	01:02
	→ Ramune joined	01:05
	← jamessharpe left	01:05
	← a-priori left	01:05
	→ doener_ joined	01:07
	→ Leefmc joined	01:07
	→ gcarrier joined	01:11
gcarrier	hi! i love the git -v feature, but i'd love to be able to edit the diff directly...	01:12
	is it somehow possible to do this fast?	01:12
	→ pombred1 joined	01:13
	→ datafirm joined	01:14
Mikachu	the what feature?	01:14
doener_	guess he means status -v	01:17
gcarrier	doener_: sorry, git commit -v	01:17
	i'd like to tweak my commit directly using some diff version	01:18
	i speak diff :)	01:18
	→ capisce joined	01:18
	← igorgue left	01:18
	← doener left	01:19
	← robin_ left	01:22
	← capisce_ left	01:22
	→ sverrej joined	01:24
	← Yuuhi left	01:26
	← pombreda left	01:26
	← strangy left	01:26
	→ pflanze joined	01:27
pflanze	Hello. I'm trying to find the sources of the lightweight git library which has been mentioned a few times (on the list, iirc).	01:27
	not the big unfinished libification thing, but something which already works for linking into apps and does the basic stuff.	01:28
Mikachu	that sounds like a work of fiction to me	01:30
Ilari	IIRC, there was some stuff, but it does not do a lot...	01:33
	http://repo.or.cz/w/git/libgit-gsoc.git	01:34
	← sgrimm left	01:37
pflanze	libgit-thin, yep, I think was the name.	01:37
	← Leefmc left	01:38
Ilari	pflanze: Yup libgit-thin. See the link above.	01:39
pflanze	I'm trying to find out how to check it out,	01:39
	← datafirm left	01:39
pflanze	git://repo.or.cz/w/git/libgit-gsoc.git or http://repo.or.cz/w/git/libgit-gsoc.git won't clone.	01:39
	← tongueroo left	01:41
pflanze	"fatal: The remote end hung up unexpectedly", is repo.or.cz's git service down?	01:41
Ilari	pflanze: Both of those URLs are incorrect.	01:43
	→ aroben joined	01:43
pflanze	so how to get the correct one? libgit isn't even listed on the repo.* homepage.	01:43
doener_	pflanze: feed the url to your browser	01:44
pflanze	yes, I see the gitweb instance but I don't see any hint how to clone	01:44
doener_	"Mirror URL"	01:44
pflanze	ah, Mirror url	01:44
	d'oh	01:44
	(I somehow expected an icon or something at the top)	01:46
	← ruphy left	01:49
Ilari	pflanze: There's no browser integration in Git... :-)	01:53
pflanze	well, ajaxy gitweb, why not. Or a Firefox plugin. But I didn't expect that, just didn't parse the relevant info out of the page.	01:54
	← david_koontz left	01:55
	← Eridius\|work left	01:57
	→ Eridius\|work joined	01:57
	→ Antrover joined	01:59
	→ destruct_ joined	02:03
	← mattf left	02:08
	← timocratic left	02:13
pflanze	Hm.	02:13
	← rubbish left	02:14
pflanze	Maybe it would be better to go with the 'real' libgit and somehow automatically convert it to be used embedded.	02:14
	← destruct left	02:14
Ilari	pflanze: That would be quite difficult...	02:14
vmiklos	hehe, the key word is automatucally	02:14
	i don't think that can be automated	02:15
pflanze	I guess two things are missing: (1) global variables (turn to a super struct or such, or maybe at least just thread-local storage), (2) memory deallocation, could be solved by using the BoehmGC.	02:15
Ilari	pflanze: Converting libgit suitable to be used as library is difficult and tedious thing to do...	02:16
pflanze	If C++ had inner classes, one could probably just wrap everything with a class definition :)	02:16
Ilari	pflanze: BoehmGC. Haha... There's thing called 'Speed'.	02:16
gcarrier	so, not possible to change an incoming commit directly by editing its diff?	02:16
	that would rooock	02:16
pflanze	Ilari: why is speed a problem with the BoehmGC? From what I read and experienced it's almost as fast as malloc.	02:17
Ilari	gcarrier: Write tool that lets to do that?	02:17
	pflanze: "Almost as fast as malloc" falls far short.	02:18
gcarrier	git-diff --cached > ~/.newdiff;git co HEAD;$EDITOR ~/.newdiff;git apply ~/.newdiff;git commit	02:18
	but i need to track which files were in the index	02:18
	else, seems good	02:18
	i mean, at least the general idea	02:19
	but not efficient at all	02:19
pflanze	Ilari: GC'd languages usually only take 20% or less of their runtime in GC. And so even if the BoehmGC is 2 times slower than malloc, that'd make it maybe 40%, so what, 1.2 times slower.	02:19
gcarrier	stupid, doesn't deal at all with multiple repos, etc. i'm use someone better than me would like this feature	02:20
	→ cored joined	02:21
cored	hi guys	02:21
	i just made some changes to a file inside my new repo	02:21
Ilari	pflanze: Doesn't matter. There are structures within git that have N+1 objects allocated, and use their own memory management (they have to).	02:21
	→ gregf joined	02:22
cored	everytime i make a change to a file i have to do a git add first ?	02:22
vmiklos	that's in the faq	02:22
gregf	is there a way to totally remove a file from git's history if you push something you really didn't want going public?	02:22
vmiklos	http://git.or.cz/gitwiki/GitFaq#head-3aa45c7d75d40068e07231a5bf8a1a0db9a8b717	02:23
	gregf: use git-filter-branch?	02:23
cored	vmiklos: that's for me ?	02:23
pflanze	Ilari: so have you, or has anyone, actually tried to compile git with the BoehmGC and see how it performs?	02:23
gregf	vmiklos: thanks	02:24
vmiklos	cored: yes, the link is for you	02:24
cored	vmiklos: thanks	02:24
vmiklos	pflanze: and you haven't reached the die() problem yet ;)	02:24
pflanze	vmiklos: ok. Thanks for noting. Well, replacing with longjmp or C++ exceptions could be a way.	02:25
Ilari	pflanze: I'm saying that just compiling git with BohemGC would not help much...	02:26
vmiklos	huh, but such a patch will be never merged ;)	02:26
	(the one that uses c++ exceptions in die() ;) )	02:26
pflanze	vmiklos: I'm talking about a way to automatically turn libgit into an embeddable state.	02:27
	Nothing to be merged into the normal one.	02:27
	Could either be a merge, or a tool who does it.	02:27
	a merge to a separate tree.	02:27
	← Lucine2 left	02:27
pflanze	Ilari: you probably want to say that the memory handling it does internally wouldn't be seen by the GC and thus not free'd; but after the one-shot operation, the whole pool could be free'd, ok?	02:28
	← warthog9 left	02:28
pflanze	even by the then (if the pool was malloc'ed).	02:28
	by the *GC	02:28
	→ dcrawford_ joined	02:29
	→ igorgue joined	02:29
Ilari	pflanze: The memory management and die() are not all of the problems. Also, there is just one copy of flags table (that needs to be shared). Also, some of the die()s might be in points where state of "library" is inconsistent...	02:29
	← gregf left	02:29
dcrawford_	I'm not sure if my from-source git is correct ... anyone seen this before:	02:30
Ilari	pflanze: Making possible to have have multiple flags tables would enlarge per-object overhead -> speed penalty.	02:30
dcrawford_	I can clone fine from a git: repo, but from the equiv http: repo I get a valid .git but the dir is otherwise empty	02:30
pflanze	I'm thinking that if it works for one-shot commands, doing one-shots in the library by initializing and then afterwards dropping all state would solve that.	02:30
	Would require initialization code to be cleaned up maybe.	02:31
Ilari	pflanze: Reinitializing the needed state after dropping is not one of the fastest operations...	02:32
pflanze	Faster than starting a process?	02:32
	well; mmap'ing would be a thing to look at, too.	02:32
	if it really doesn't unmap itself.	02:33
Ilari	pflanze: One could write pooled version of malloc() and use that...	02:33
	pflanze: Except that there is stuff like pack mmap windows.	02:33
vmiklos	btw what's the problem with fork+exec when you use git from an external application?	02:34
	← cored left	02:34
pflanze	yeah speed	02:34
	I'm looking at using Git as storage engine for (web) applications,	02:35
	→ ekidd joined	02:35
pflanze	which also needs indizes and such, which requires fast access to individual objects.	02:35
	yes still tearing up and down mmaps and such stuff will not do good for performance.	02:36
Ilari	pflanze: Just reading the object database is not that hard (wonder how many times has that been implemented, I say at least 4).	02:36
pflanze	Some of that could be overridden by cached/pooled ops possibly.	02:37
Ilari	pflanze: Depends on OS. Mmaps are quite slow on systems such as Mac OS X, but Linux VFS tends to be really fast.	02:38
pflanze	C, java (Eclipse), C# (ongoing mono gsoc project), and?	02:38
	← Antrover left	02:38
pflanze	libgit-thin? I just fear that it won't work as Git progresses.	02:38
Ilari	pflanze: C, Java (Jgit) and two ruby reimplementations...	02:38
	← krawek left	02:39
	← ekidd left	02:39
pflanze	Stuff like locking rules etc., I wouldn't like if upstream git changes and then libgit-thin or my reimplementation/whatever won't behave anymore.	02:40
Ilari	The running joke is that Git is the most reimplemented SCM... :-)	02:40
	→ cored joined	02:40
cored	vmiklos: sorry for bothering you again	02:40
	vmiklos: after doing the commit	02:40
Ilari	pflanze: Only thing that would really break it would be index or pack formats changing...	02:40
cored	vmiklos: i just check out the repo again and i did not see my changes in there	02:40
Ilari	cored: You did another clone of the repository?	02:41
cored	yes	02:41
vmiklos	http://git.or.cz/gitwiki/GitFaq#head-b96f48bc9c925074be9f95c0fce69bcece5f6e73	02:41
	please, read over the faq	02:41
	pflanze just realized that C++ actually does* offer inner classes as it seems*	02:41
cored	vmiklos: ok	02:42
	← shoe left	02:42
Ilari	cored: Was the repo you cloned the one you committed? If not, did you push the change to that repo?	02:43
	s/committed/committed to/	02:43
cored	wait	02:43
Ilari	cored: Each clone of Git repository is complete stand-alone repository itself.	02:46
cored	ok	02:46
	i think the problem is because i did a git remote add another repo	02:46
	which i don't have permissions to write there	02:47
	maybe that's why i can't make the push what do you think ?	02:47
DrNick	pflanze: not exactly	02:47
Ilari	cored: Yes, you need write permissions to remote repo to make a push.	02:47
DrNick	pflanze: at least, not in the Java sense	02:47
	→ tongueroo joined	02:48
pflanze	yeah, I just wondered, "A nested class may inherit from private members of its enclosing class.", just inheriting is not the same;	02:48
Ilari	cored: Maybe go to that another clone, register the other as remote, and pull the changes?	02:48
pflanze	but then what is different from just normal interiting class declarations?	02:48
	← SuttoL left	02:48
cored	Ilari: how can i go back to use my repo and not any ref to the remote	02:48
	?	02:48
	→ SuttoL joined	02:49
DrNick	pflanze: Java inner classes get a hidden reference to their containing class and are allowed to access it's private members, while C++'s "inner" classes are just a namespace thing	02:49
	→ duck1123 joined	02:49
	→ imyousuf joined	02:49
Ilari	cored: Well, you can just ignore the remote reference...	02:49
cored	hm	02:51
	well it's looks like is another thing	02:51
	because after the push command i'm getting an error	02:51
	and the url is pointing is to my own repo	02:51
	fatal: The remote end hung up unexpectedly	02:51
Ilari	cored: Is there any other errors (such as "...can't chdir...") before it?	02:52
cored	nop	02:53
	after i have this one	02:53
	error: failed to push to 'git://github.com/cored/merb-core.git'	02:53
Ilari	cored: You can't push over git:// (well, if you care about security, that is).	02:54
	→ Ken69267 joined	02:55
	← Eridius\|work left	02:55
Ken69267	did git-svn change behavior when dcommitting?	02:55
cored	Ilari: ok, so where can i do the push ?	02:55
Ilari	cored: Pushes to remote hosts are normally done using ssh://... URLs.	02:56
cored	i see	02:57
	→ loops joined	02:57
cored	so i have to re clone it using the ssh url	02:57
Ilari	cored: Maybe edit '.git/config'. The URL is there...	02:57
	← duck1123 left	02:58
Ilari	cored: Or 'git config remote.origin.urk <new URL>'.	03:00
	*'git config remote.origin.url <new URL>'.	03:00
Ken69267	indeed	03:02
	← Ken69267 left	03:02
	← kanru left	03:03
	→ rubbish joined	03:03
	→ spearce joined	03:05
	← rwanderley left	03:06
	← flaguy left	03:14
	→ aroben_ joined	03:17
	← aroben left	03:17
	→ capisce_ joined	03:19
	← loops left	03:21
	← spearce` left	03:21
	← doener_ left	03:22
	← dcrawford_ left	03:22
	← capisce left	03:23
	← gcarrier left	03:24
	← a-priori_ left	03:27
	← Aughey left	03:41
	→ warthog9 joined	03:41
	→ ekidd joined	03:42
	→ threeve_ joined	03:45
cored	Ilari: thanks for all your help	03:47
	← cored left	03:47
	→ mvanholstyn joined	03:49
	← ekidd left	03:53
	← mvanholstyn left	03:54
	→ theperitus joined	03:55
	← threeve_ left	03:57
	← SuttoL left	03:57
	→ SuttoL joined	03:57
	← aroben_ left	03:59
	← threeve left	04:02
	→ aroben joined	04:02
	← rubbish left	04:04
	→ eternaleye joined	04:08
	← jeffpc left	04:10
	→ Surfer15 joined	04:10
	Surfer15 Visit http://www.FakeMagazineCover.com (upload pic make mag) - http://www.SillyWebcam.com (play with webcam online) - http://www.Is-A-Jerk.com (insulter/anon email) - http://www.ComedySearchEngine.com (fun) - http://www.BodySwitcher.com (put your face on funny body) - http://www.MedChecker.com (health) - http://www.Canuckster.com (Canada eh) - http://www.Nerdful.com (geeks)	04:10
	← Surfer15 left	04:10
	→ tjafk1 joined	04:10
	← daperitus left	04:11
	→ jeffpc joined	04:11
Eridius	I wish freenode would kline this Surfer## spammer	04:11
	← madewokherd left	04:16
	→ spearce` joined	04:17
	← SuttoL left	04:18
	→ jwasson joined	04:19
	← pflanze left	04:20
	← tjafk2 left	04:27
	← jroes left	04:27
	← eikonos left	04:29
	→ kanru joined	04:31
	← spearce left	04:34
	→ cmo-0 joined	04:35
	→ bootie joined	04:37
	→ ckoehler joined	04:40
	← spearce` left	04:44
	← Eridius left	04:46
	← ckoehler left	04:49
	→ sgrimm joined	04:53
	← cmo-0 left	04:54
	← ggeecko left	04:59
	→ ggeecko joined	04:59
	← jeckel left	05:05
	→ ggeecko_ joined	05:05
	→ cmo-00 joined	05:09
cmo-00	i'm trying to clone a git repo. through git protocol on github.com, when doing 'git clone git://github.com/bruce/rtex.git' the program hangs there, only it maked an empty rtex folder, but nothing more. i tried telnet github.com 9418, which logged in and logged out when i typed 'junk'	05:11
	if i do 'git clone git://github.com/bruce/rte' the remote end hung up unexpectedly	05:12
tpope	cmo-00: the first is cloning fine for me	05:13
	still in progress but it's a slow hotel connection	05:13
cmo-00	tpope: this is on my office pc, at home it works perfectly	05:15
halorgium	cmo-00: maybe the port is blocked	05:17
	dropping your request packets	05:17
	→ ruphy joined	05:19
tpope	is it an old git version?	05:20
cmo-00	two things i know for sure. 1) the iptables policy for input, output and forward are accept. 2) the router has a port-forwarding rule on 9418/tcp	05:20
	1.5.4	05:20
tpope	you shouldn't need 9418 forwarded	05:20
cmo-00	i have a git-daemon on the office server also	05:20
	→ gnufied joined	05:23
gnufied	correct me, if I am wrong, git checkout foo, switches to branch foo, but also updates branch foo with changes from current working branch, right?	05:24
RandalSchwartz	no	05:24
	but if you have dirty files in your tree, it brings them along	05:24
	on the theory that you really wanted to check them into a different branch, but forgot to switch to that first	05:24
gnufied	whats dirty files? modified files, which aren't checked in?	05:25
RandalSchwartz	to move work from one branch to another, you need to cherry pick, merge, or rebase, or some combinatin of those	05:25
	dirty = anything "git status" complains about	05:25
gnufied	hmm, is there way to not carry those dirty files when switch branch ?	05:26
RandalSchwartz	in general, you don't want to be using git checkout unless git status shows clean, unless you really understand what you're doing	05:26
	well - why do you have dirty files?	05:26
	is it work in progress?	05:26
	or things you don't want to track at all?	05:26
	→ loops joined	05:26
gnufied	yeah, something, experimental.	05:27
RandalSchwartz	yeah what?	05:27
	stuff you never want back again?	05:27
	or stuff you might want back later?	05:27
	← gmcinnes left	05:27
gnufied	yeah, I understand the flow now.	05:27
RandalSchwartz	eh?	05:28
	you're not answering me?	05:28
	which was it	05:28
gnufied	like, i could have commited those changes to local branch. and then switched. that branch was anyway, local so it woulnd't have affected the branch to which i was switching	05:28
RandalSchwartz	right	05:29
cmo-00	Strange!!! the following works perfectly git clone git://git.savannah.nongnu.org/stumpwm.git .	05:29
RandalSchwartz	or just simply use git clean first	05:29
gnufied	those changes are something, i think obviously useful, but yet ready for use by other third party libraries.	05:29
RandalSchwartz	that throws away things	05:29
	oh - then yes, commit first	05:29
	you can always unwind the commit later	05:30
	essentially a managed git-stash	05:30
gnufied	yeah. right o	05:30
RandalSchwartz	cmo?	05:30
	RandalSchwartz hangs out "the doctor is IN" sign	05:30
cmo-00	RandalSchwartz: i'm trying to clone a git repo. through git protocol on github.com, when doing 'git clone git://github.com/bruce/rtex.git' the program hangs there, only it maked an empty rtex folder, but nothing more. i tried telnet github.com 9418, which logged in and logged out when i typed 'junk'	05:31
	gnufied is reading too many PG wodehouse books.	05:31
tpope	cmo-00: can you clone another repo?	05:31
cmo-00	yes and this is one ... Strange!!! the following works perfectly git clone git://git.savannah.nongnu.org/stumpwm.git .	05:31
	← loops left	05:31
RandalSchwartz	that clone is working for me	05:31
	... git clone git://github.com/bruce/rtex.git	05:31
	yes. just finished.	05:32
cmo-00	tpope: git clone git://git.savannah.nongnu.org/stumpwm.git	05:32
	tpope: but note git clone git://github.com/bruce/rtex.git	05:32
RandalSchwartz	cmo-00 - that github clone works for me	05:32
	→ rubydiamond joined	05:32
cmo-00	but not for me	05:32
RandalSchwartz	what version of git do you have?	05:32
cmo-00	1.5.4	05:33
	→ ckoehler joined	05:33
RandalSchwartz	ooooh	05:33
	that version is so, like, two months ago	05:33
	please upgrade	05:33
cmo-00	even with 1.5.5 from sources the same happend	05:33
gitster	bs. anything remotely recent such as 1.3.0 should be Ok.	05:33
RandalSchwartz	we're on 1.5.6 though :)	05:33
	heh	05:33
rubydiamond	How do I uninstall Git from my macbook... which I installed from source.... I want to upgrade it	05:33
RandalSchwartz	did you install it into something cool like /opt/git ?	05:34
	or just let it default and spray all over your default dirs	05:34
rubydiamond	RandalSchwartz: no..	05:34
	its in default dirs	05:34
RandalSchwartz	good luck cleaning then	05:34
rubydiamond	RandalSchwartz: hmm	05:34
gitster	isn't the default /home/$user/bin?	05:34
rubydiamond	is there any uninstall script?	05:35
RandalSchwartz	not that I've seen	05:35
gitster	I do not think so.	05:35
rubydiamond	its there in usr/local/bin/git	05:35
RandalSchwartz	default what?	05:35
	this is why I use PREFIX=/opt/$project for everything I install locally	05:35
	so I can undo it.	05:36
rubydiamond	I hope there is some uninstall script and I will just run it to install it	05:36
RandalSchwartz	rubydiamond - as we have just said, not likely	05:36
rubydiamond	RandalSchwartz: how do u undo it	05:36
	hmm	05:36
gitster	well you can install it again and see what's installed ;-)	05:36
RandalSchwartz	"u"?	05:36
	→ loops joined	05:36
RandalSchwartz	are you paying by the character for IRC?	05:36
	if so, must suck to be "u"	05:36
rubydiamond	RandalSchwartz: I mean... what is difference having it in usr/local/bin/git and /opt/git	05:36
cmo-00	/usr/bin/git	05:36
RandalSchwartz	if it's in /opt/git, I can tell everything that is installed	05:37
	it's in /opt/git/bin/*	05:37
	and /opt/git/lib/*	05:37
	etc	05:37
rubydiamond	RandalSchwartz: oh sorry sir	05:37
RandalSchwartz	I want to remove git...	05:37
	I say rm -rf /opt/git	05:37
	and it's all gone	05:37
	easy	05:37
rubydiamond	oh..	05:37
RandalSchwartz	30 years of Unix	05:37
	this is one thing I've learned. :)	05:37
rubydiamond	so now.. what is the way to upgrade it on macbook..	05:37
RandalSchwartz	never use the default install dirs. :)	05:37
rubydiamond	RandalSchwartz: oh	05:37
RandalSchwartz	the new install will overwrite the old though	05:38
	so you upgrade just by... upgrading. :)	05:38
rubydiamond	yeah... seems good way to upgrade it..	05:38
	will try it	05:38
jwasson	Funny, it only took me 5 years to learn not to use Unix. ;)	05:38
RandalSchwartz	but in the future, either use a package manager (fink or macports) or install things in /opt	05:39
gnufied	cmo-00, i have observed that sometimes, if you have slow internet connection using git protocol causes cloning process to hang.	05:39
rubydiamond	thanks RandalSchwartz	05:39
	RandalSchwartz bows	05:39
	RandalSchwartz hits head on desk	05:39
RandalSchwartz	ouch!	05:39
rubydiamond	jwasson: so.. you use Git on unix machines	05:39
RandalSchwartz	hi gitster, by the way	05:39
rubydiamond	I mean Windoze machines	05:39
jwasson	I actually run it on both.	05:39
	Er, I mean I run it on Unix and Windows. I don't use Windoze.	05:39
rubydiamond	jwasson: so do you use Cygwin or any other installer for it?	05:40
jwasson	I used msysgit.	05:40
rubydiamond	so.. you just had said that you learned in 5 years that you learned not to used Unix, thats is why I asked	05:40
jwasson	I was joking.	05:40
rubydiamond	jwasson: msysgit looks nice	05:41
	jwasson: oh	05:41
RandalSchwartz	And yes, I indeed typed "ls" for the first time in 1977.	05:41
	I wasn't kidding about 30 years.	05:41
jwasson	I stopped running Linux for a couple of years.	05:41
RandalSchwartz	the entire unix kernel fit in 32k back then	05:41
	yes. 32 K	05:41
	not meg	05:41
jwasson	You've got me beat, I only started using Linux 15 years ago.	05:41
RandalSchwartz	22 system calls	05:41
	Linux isn't unix	05:41
	it's sorta unix like	05:42
	except where it's wildly different	05:42
jwasson	I used Xenix a little back then as well.	05:42
RandalSchwartz	I won't run linux on anything that matters	05:42
gnufied	RandalSchwartz, ouch, then what?	05:42
RandalSchwartz	osx for the laptop. openbsd on the server so I can sleep at night.	05:43
jwasson	Yeah, I like *BSD.	05:43
RandalSchwartz	theo's paranoia is my ambien. :)	05:43
	→ datafirm joined	05:43
gnufied	ubuntu 8.04 runs like champ on notebook. everything works.	05:43
jwasson	I managed to install Debian during it's "What is this 'entropy' you speak of?" phase.	05:43
	*** cdm (n=cdm@pool-71-116-66-101.snfcca.dsl-w.verizon.net) has joined channel	05:44
	Oops.	05:44
	Wrong button.	05:44
	gnufied Linus soul would be twitching in his body, if he would have heard what RandalSchwartz got to say.	05:44
RandalSchwartz	no. I've told him that to his face. :)	05:44
cmo-00	gnufied: my home connection is 3 times slower than my office connection,but it still works	05:45
	→ gmcinnes joined	05:45
RandalSchwartz	take a look at the security alerts against linux vs those against openbsd	05:45
	and you'll see what I mean	05:45
	← datafirm left	05:45
gnufied	he he, i smell of openssl.	05:45
RandalSchwartz	oh yeah, wow, that debian mistake	05:46
	what a fiasco	05:46
	had they only submitted the patch upstream, theo's boys would have caught it	05:46
	instead, they sat on it	05:46
	dumb	05:46
gnufied	yeah, they have. my boxen are patched.	05:46
	but I do believe, for poor guys like me Ubuntu/Linux is a good choice for laptops.	05:48
	→ sikanda joined	05:48
gnufied	OS X is good, but aren't you again handling everything to apple, again? i mean, philosophically, how is it better than windoze?	05:48
	*handing over	05:49
RandalSchwartz	it's far better than windoze, because I can compile all my unix toys on it.	05:49
	and because it's just not windows. :)	05:49
	and I do trust unix-heads to get security a bit better than redmond, given the relative trackrecord.	05:50
Arrowmaster	debian did ask for help on that patch from upstream	05:50
	got ignored	05:50
RandalSchwartz	they didn't submit an actual patch	05:50
	they just talked about it	05:50
	I've seen the report	05:50
	it would be easy for that to not get noticed	05:51
	actual patch would have been examined and rejected	05:51
Arrowmaster	got ignored by openssh though since it was related to how openssh got its random numbers	05:51
RandalSchwartz	very likely	05:51
gnufied	yeah, i think it was OpenSource and public and hence issue got wide coverage with xkcd publishing a cartoon with /* */, if it would have anything, they would have it under carpet in a minute.	05:53
Arrowmaster	nah it got a lot of coverage even before xkcd	05:54
	but a lot of people didnt even bother reading about it and just assumed the dev did something stupid without talking to anybody	05:54
	→ Lucine2 joined	05:55
jwasson	It just shows that a "million eyes" can all blink at the same time.	05:57
	panagos_ → panagos	05:57
	→ Xyri0 joined	06:01
	← igorgue left	06:02
	← imyousuf left	06:03
	→ datafirm joined	06:04
	← Xyrio left	06:11
	Xyri0 → Xyrio	06:11
	← bobesponja left	06:14
	→ adante_ joined	06:18
	→ Antrover joined	06:24
	← Antrover left	06:25
	← adante left	06:34
	adante_ → adante	06:34
	← sikanda left	06:34
	→ Xyri0 joined	06:42
	→ mvanholstyn joined	06:44
	→ mvanholstyn_ joined	06:44
	→ drizzd_ joined	06:46
rubydiamond	hi	06:48
	where is the git_autocompletion.sh is installed when you install git install from source /usr/local	06:48
	← bootie left	06:50
	→ deeno joined	06:51
	← Xyrio left	06:52
	Xyri0 → Xyrio	06:52
	← mvanholstyn_ left	06:54
	→ Eludias joined	06:55
	→ thiago_home joined	06:56
	← drizzd left	06:57
	→ Beket joined	06:58
	← tongueroo left	06:59
	← tenpaiyomi left	07:01
	← mvanholstyn left	07:01
	→ xilef joined	07:01
xilef	can anyone tell me how git handles symbolic links? i particularly don't want them stored as links but rather de-referenced	07:02
	← cmo-00 left	07:03
	← aroben left	07:04
	← Modius left	07:05
	→ Modius joined	07:06
	→ _graham_ joined	07:07
	→ tzafrir_home joined	07:10
thiago_home	xilef: it stores them as links	07:11
xilef	thiago_home, can i turn that off?	07:11
	on a per file basis?	07:11
thiago_home	you can't	07:12
	symlinks are stored as special files	07:12
xilef	righto, thanks for your help	07:12
	→ d0k joined	07:15
	→ mvanholstyn joined	07:16
	← mvanholstyn left	07:19
	← xilef left	07:21
	→ ph^ joined	07:22
	→ holizz joined	07:22
	→ _zack joined	07:23
	← Beket left	07:23
	→ mrCrowley joined	07:24
	→ priidu joined	07:28
	← _graham_ left	07:31
	← roylee17 left	07:36
	→ \|kukks\| joined	07:37
	→ robin_ joined	07:46
	→ david_koontz joined	07:47
	→ FunkeeMonk joined	07:48
	← Lucine2 left	07:49
	← mrCrowley left	07:49
	→ LotR joined	07:50
	← datafirm left	07:50
	← ckoehler left	07:53
	← kukks left	07:55
	→ mvanholstyn joined	08:08
	← d0k left	08:09
	→ shaftyy joined	08:15
tzafrir_home	bah, this greylisting on the list server is annoying	08:16
	I thought my own server was bad	08:16
	← ajonat left	08:18
rubydiamond	tzafrir_home: what is your server	08:22
tzafrir_home	I send from local.xorcom.com	08:23
	← david_koontz left	08:23
tzafrir_home	It was plain annoying. It means I can't just "subscribe" to the list, and just "post to the list. There is an inherent delay for newcomers	08:24
	(and I know, I'm not the first one saying this, and I'm not the admin of the list server who actually has to deal with the spam )	08:24
	→ twb joined	08:27
twb	etckeeper has resulted in a 40MB /etc/.git for a 45MB /etc, after a few months of use. Can I somehow optimize the .git directory to use less space?	08:29
Pieter	git gc	08:29
twb	Thank you.	08:29
tzafrir_home	twb, does it record changes automatically?	08:30
Pieter	and if you really want to spice it up, do a git repack -adf --window=200	08:30
tzafrir_home	also: maybe exclude some files?	08:30
twb	tzafrir_home: yeah, I should probably ignore resolv.conf and stuff.	08:31
tzafrir_home	depends where	08:31
twb	Pieter: impressive, 40MB -> 7.4MB	08:31
tzafrir_home	with the resolveconf package on Debian resolve.conf changes reside in /var anyway	08:31
twb	I do not like resolveconf.	08:32
	← FunkeeMonk left	08:32
Pieter	twb: nice :)	08:33
twb	Pieter: I'll ask upstream to make a cron job do that	08:33
	→ hein2 joined	08:37
	← Sho_ left	08:37
	hein2 → Sho_	08:37
	→ deavid joined	08:37
	← twb left	08:43
	dwmw2_HEL → dwmw2_gone	08:44
	→ kumbayo joined	08:46
	→ ph^_ joined	08:58
	→ gitte joined	09:04
	→ imyousuf joined	09:06
	← pygi left	09:11
	← deavid left	09:22
	← statim left	09:24
	→ pygi joined	09:25
	← imyousuf left	09:26
	→ imyousuf joined	09:29
	← imyousuf left	09:29
	→ chris2 joined	09:31
	→ strangy joined	09:38
	→ zachinglis joined	09:39
	→ statim joined	09:43
	→ cedricv joined	09:45
	← ph^ left	09:49
	ph^_ → ph^	09:49
	→ charon joined	09:51
	← priidu left	09:53
	← gmcinnes left	09:53
	→ shaftyy_ joined	09:55
	dwmw2_gone → dwmw2_HEL	09:55
	→ priidu joined	09:59
	← Ademan left	10:00
	→ Ademan joined	10:01
	→ dirk1296 joined	10:02
	← shaftyy_ left	10:04
	← shaftyy left	10:06
	← csc` left	10:06
	← holizz left	10:10
	→ stillLotR joined	10:10
	→ redondos joined	10:11
	→ tko_ joined	10:12
	→ ben_h joined	10:13
tko_	can anyone perhaps suggest a patch tracker that could monitor mailing list posts of patches sent for review and keep track which patches have been reviewed/committed/etc. ?	10:14
	← LotR left	10:16
	← ben_h left	10:17
	← zachinglis left	10:17
	→ johan-s joined	10:18
	→ shaftyy joined	10:19
	← shaftyy left	10:20
	← jelmer left	10:30
	→ tommorris joined	10:30
	→ crab joined	10:32
drizzd_	tko_: maybe you should ask junio, I was wondering how he does it.	10:33
	→ shaftyy joined	10:36
	→ lea_w joined	10:37
	→ radarek joined	10:38
	→ nud joined	10:38
	tko_ → tko	10:39
tko	maybe he's just better at it :)	10:39
	← tommorris left	10:40
drizzd_	maybe :-). he must have some kind of scripting support though. see for example his what's cooking posts.	10:41
	I doubt he does that manually	10:42
	→ Xyri0 joined	10:43
	→ tommorris joined	10:46
Ilari	drizzd_: Aren't those scripts available in some branch of git.git?	10:48
drizzd_	I wouldn't know	10:48
tko	I suppose we could put everything in 'review-needed' branch or something instead	10:49
drizzd_	there is git-topic.perl in the todo branch	10:49
	← Xyrio left	10:53
	Xyri0 → Xyrio	10:53
	→ devogon joined	10:54
	← brosner left	10:54
	→ shaftyyy joined	10:59
	→ dmq joined	11:06
	← tommorris left	11:08
	← lea_w left	11:10
	← Eludias left	11:12
	← dirk1296 left	11:15
	← shaftyy left	11:15
	← mvanholstyn left	11:16
	← rubydiamond left	11:16
	→ rubydiamond joined	11:16
	← shaftyyy left	11:17
	→ BabelO joined	11:19
rubydiamond	I just did rm file	11:21
	git status shows "# deleted: fetcher.rb"	11:22
	I want it back	11:22
	how do I do that?	11:22
	anybody know how do I do that?	11:22
	→ swombat joined	11:24
swombat	is it possible to merge two git repositories together? Basically, I have a project that's split across 2 repos and I'd like to scrunch them together without losing the revision history	11:24
	→ rubbish joined	11:24
jengelh	fetch and merge, simple as that	11:25
swombat	these are two separate sets of files	11:25
jengelh	well you don't even need to merge; you can have many trees, since git actually stores a forest	11:25
swombat	one is the rails backend, the other the flex frontend	11:25
	we made a bad decision in splitting them up	11:25
	can i be in repo 1 and "fetch" repo 2 then?	11:26
	(is that what you're saying?)	11:26
jengelh	yes	11:27
	→ bentob0x joined	11:27
swombat	incredible. that's with the fetch command? (I'll go look at the manual page if you just point me to the right command :-) )	11:27
jengelh	git remote add; git fetch	11:28
swombat	unbelievable.	11:28
	thank you :-)	11:28
	← ggeecko left	11:29
swombat	corrolary question: a large part of the space used by the repo is hundreds of copies of the 500k-1.5M SWF built by Flex... is there a way to nuke a specific file (in a specific directory) across all the revision trees to reduce the size of the repo?	11:29
	→ pflanze joined	11:29
	← ggeecko_ left	11:29
cedricv	hmm doesn't git save binary files changes as deltas too?	11:31
jengelh	because they are hard to compress?	11:31
cedricv	oh the delta is after compression?	11:32
swombat	cedricv: it might, but this binary has changed hundreds or even thousands of times, and it makes our repo about 1.5G atm :-)	11:32
	and I'd say 98% of that is that SWF file	11:32
Arrowmaster	you would have to do revision rewriting to remove them all i think	11:32
jengelh	binary files can often change so that the delta would be bigger than storing a raw copy;	11:32
Arrowmaster	and probably want to add that SWF file to your .gitignore	11:32
cedricv	swombat: yeah sure anyway it doesn't make it a good idea to put in a rep something that is generated but i was wondering about how git store binary changes?	11:32
swombat	Arrowmaster: can't... we need it for deployments	11:33
	Arrowmaster: the thing is, we only really need the current copy, since it can be rebuilt from sources	11:33
cedricv	swombat: doesn't git gc or git repack -a -d would dramatically reduce the size of your rapo anyway?	11:33
swombat	git repack??	11:34
	man i need to learn more about git.	11:34
Arrowmaster	try git gc and see if the size changes	11:34
Ilari	cedricv: Git can ineed attempt to delta any type of file, but binary files (especially compressed or encrypted files) may not delta-compress very well (if at all)...	11:35
swombat	running..	11:35
	← Sho_ left	11:35
Arrowmaster	you can use filters to try and uncompress files automaticly before commiting them and recompress them on checkout	11:35
	i think	11:36
swombat	will that gc reduce the size of the remote repo too or just the local one?	11:36
jengelh	whichever repo it is run in	11:37
swombat	oh... so it wouldn't help reduce the size of the github repo then	11:37
jengelh	you have to run gc on the remote side to squeeze it there	11:37
swombat	lol... it's gone from 1.5G.... to 1.5G	11:37
	not a huge reduction.	11:37
	Arrowmaster: ^	11:38
	(from git gc)	11:38
Arrowmaster	it was worth a shot	11:38
	→ lea_w joined	11:38
swombat	Arrowmaster: yep :)	11:38
jengelh	git repack -a -d --depth=250	11:38
	→ Sho_ joined	11:38
jengelh	try	11:38
swombat	what will that do?	11:38
jengelh	compress-me-harder	11:38
Arrowmaster	i dont know enough to help you try doing history rewriting to remove all the old files	11:38
swombat	i'm not really concerned about the local repo though	11:38
jengelh	Arrowmaster: you set a graft, and then let rewrite the revisions, then remove the graft, then gc.	11:39
pflanze	The problem will be to keep the sha1 id's	11:40
	unless you don't care about those.	11:40
swombat	I don't care about them	11:40
jengelh	it's probably a necessary evil	11:40
Arrowmaster	if your going to be doing a lot of rewriting because you need the file to in the recent version but not in the older ones would moving it to a submodule help with the problem of sharing rewriten repos?	11:40
swombat	I just want to nuke the swf file	11:40
jengelh	though it screws with tags and all that	11:41
swombat	Arrowmaster: it can be deleted completely including current version if necessary	11:41
	takes about 10 seconds to recreate it :-)	11:41
jengelh	mh	11:41
	don't add files to the repository that are autogenerated	11:41
swombat	jengelh: i need to, for deployment purposes	11:41
jengelh	nah	11:41
	things like "configure" for example don't belong into a repo	11:41
swombat	hmm	11:42
	I suppose capistrano could just scp the file over	11:42
jengelh	for things that are not easily autogenerated, like .swfs it may be ok	11:42
pflanze	swombat: you only ever want one version in the repo -- but you say you need it versioned for deployment. That sounds like a contradiction.	11:42
jengelh	but for configure, people just run autogen.sh before	11:42
Arrowmaster	if you absolutely have to i bet a submodule would atleast be less hassle then constantly rewriting the history on the main repo	11:42
swombat	no, i need it in the repo for deployment to get it deployed	11:42
jengelh	mh	11:43
pflanze	swombat: just add a script which fetches it from a web url?	11:43
swombat	Arrowmaster: that sounds like a potential solution too	11:43
jengelh	swombat: use a new history tree and use amend to update it	11:43
swombat	pflanze: i think i might just write a step that scp's it actually	11:43
jengelh	that should work, keep exactly 1 revision of it, and always transfer it with git fetch	11:43
swombat	you guys are a gold mine of ideas, btw	11:43
	thanks	11:43
jengelh	well like I said earlier, git stores not just a history tree, but a history forest	11:44
drizzd_	can I make git citool break lines in commit messages?	11:44
jengelh	drizzd_: use git commit instead?	11:44
drizzd_	very funny	11:44
jengelh	i'm serious	11:44
swombat	drizzd_: i think he's serious	11:44
Arrowmaster	im lazy, git-gui &	11:44
drizzd_	well that costs me an extra 10 keys of typing at least	11:44
jengelh	moving the mouse costs time too :p	11:45
swombat	drizzd_: alias?	11:45
	swombat does: gca -m "Some commit messagenwith line breaks n stuff"	11:45
jengelh	what I am saying is that with the console-based git-commit, you can set an $EDITOR of your choice, and -- surprise surprise -- that editor can do automatic line breaking on request	11:45
pflanze	swombat: btw you can even add that script as a hook to git, so upon git pull the swf is fetched automatically, too.	11:45
drizzd_	look, sometimes a type the message, stage hunks, change the message, unstage hunks again etc.	11:46
	I can't do that with git add/git commit	11:46
jengelh	use stgit.	11:46
swombat	right. other git question. We've done a lot of stuff in our master branch with submodules and such, and now when I try to checkout the production branch it complains about some untracked files and such in some of the directories which are now submodules:	11:47
	git checkout production	11:47
	fatal: Untracked working tree file 'vendor/plugins/rspec/.autotest' would be overwritten by merge.	11:47
	I think it's related to the submodule stuff	11:47
	not quite sure what to do. happy to nuke the production branch after I've switched to it successfully and had one last look	11:47
	← SRabbelier left	11:48
pflanze	well you seem to have checked in that .autotest file into the repo	11:48
	→ SRabbelier1 joined	11:48
drizzd_	jengelh: that seems like overkill	11:48
swombat	pflanze: it's part of the (external) submodule	11:48
drizzd_	but i'll have a look at it	11:48
swombat	pflanze: it's an rspec file, i mean	11:48
jengelh	drizzd_: I am not sure why you would unstage things anyway	11:48
	← rubbish left	11:48
	SRabbelier1 → SRabbelier	11:48
drizzd_	jengelh: because I don't want to commit all my changes	11:48
jengelh	oh...	11:49
drizzd_	I often do something, fix something else in the process	11:49
jengelh	use branches	11:49
drizzd_	and so I make two commits for completely different reasons	11:49
jengelh	that's good	11:49
drizzd_	oh come on that's ridiculous	11:49
jengelh	that's how people do it	11:49
swombat	drizzd_: what's ridiculous? branches?	11:49
jengelh	as soon as they think "hey I can change that because it looks worth fixing", they open a new branch.	11:49
drizzd_	i'm not going to make a branch for 5 seconds	11:50
swombat	drizzd_: why not?	11:50
	it takes like 0.5s	11:50
pflanze	swombat: I haven't used submodules, but you must have this same file in several modules, and/or locally generated; so go check where the doppelgänger is	11:50
jengelh	As Linus puts it, because branches are so cheap, people have 3, 5 or 10 of theem	11:50
	and you do not have to merge em, you can rebase it.	11:50
drizzd_	guys, branches won't help my problem at all	11:50
jengelh	which makes for a nice linear history	11:50
swombat	drizzd_: trust us... with git, you can branch like crazy really easily (especially locally). That's one of git's huge advantages over svn.	11:51
	try it	11:51
jengelh	if, during editing for a change, you notice that you have to fix something else first, you use stgit (or stash in rare cases) to back out your changes and do the required fix first	11:51
drizzd_	ok, I understand how branches work, and I use them extensively. That's not the issue. My problem is totally unrelated to branches.	11:51
	I have a dirty working directory	11:51
	jengelh listens	11:52
drizzd_	One changed file, for example	11:52
	two hunks	11:52
	I want to make two commits	11:52
	one for the first hunk	11:52
jengelh	git add -i	11:52
drizzd_	another one for the other	11:52
	oh, have you ever used that?	11:52
jengelh	yes	11:52
	solemn, but yes, I know how it works	11:52
drizzd_	well maybe I don't get it but it just doesn't work for me	11:52
jengelh	I rather take the branch approach	11:52
	but that's just me.	11:52
drizzd_	how does a branch help me staging hunks?	11:52
Arrowmaster	everybody has their own development method	11:52
jengelh	drizzd_: it does not, you have to edit it again :)	11:53
pflanze	drizzd_: you can also just "git commit first/file", then "git commit other/file"	11:53
drizzd_	well, git citool solves that problem for me, but not having line breaks in the commit messages sucks.	11:53
	pflanze: it's two hunks in the same file	11:54
pflanze	ah	11:54
	then undo one of the hunks, maybe through editor undo, commit, redo, commit	11:54
swombat	drizzd_: 1) decide to work on defect 123; create branch 123; make change to file X; commit to branch 123; Oh shit, need to work on 124; create branch 123; make change to file X; commit; checkout master branch; merge 124; push to origin; checkout 123 again; rebase; resume work on 123	11:55
drizzd_	this was just an example, usually it's a few files and quite a few hunks	11:55
swombat	er... -1)	11:55
	and the second create branch should be 124.. slaps self	11:55
pflanze	drizzd_: commit often as you work, and afterwards fix up the history if you think it is looking ugly	11:56
drizzd_	I often break my executable during work, and I really don't like commiting something that doesn't even compile	11:56
pflanze	why not?	11:56
swombat	you can amend the commit later	11:57
pflanze	you don't need to push/publish to others	11:57
drizzd_	because it will break bisect, for example	11:57
swombat	and that... it stays private	11:57
pflanze	drizzd_: that's why I said fix the history afterwards	11:57
swombat	nobody needs to know you broke the executables :-)	11:57
	← Xyrio left	11:57
pflanze	before you publish it.	11:57
jengelh	fix it before you commit it :P	11:58
drizzd_	I guess the problem is that I'm just to lazy to commit every time I decide to do something different	11:58
	because whenever I fix something I end up fixing 5 other things as well.	11:58
swombat	drizzd_: once you're used to it it's so quick you don't even spend any brainpower on it	11:58
	it becomes second nature to create branches as you need them	11:59
	(for even the tiniest change)	11:59
pflanze	drizzd_: how would the tool you imagine (and which doesn't exist afaik) find out which hunks belong together, and worse, if two independent changes happen to be in the same hunk, how to take those apart?	12:00
	You'll have to do just as much work as you'd do when committing early, I suppose.	12:00
drizzd_	pflanze: as I said, git gui is the solution to my problem. I simply can't work with git add -i, so I don't have any other decent tool to stage hunks. That is all.	12:01
	But you're wrong. I have to commit less often, so to speak, because in the end I won't have to amend anything.	12:02
pflanze	so, think what you want, if you've got a solution I'm not sure what you're after then.	12:02
drizzd_	If it's about bigger changes, I do exactly the thing you suggested. But for small changes it's just not worth it, IMO.	12:03
	→ tommorris joined	12:03
drizzd_	pflanze: I just asked how to do line breaks in the commit message with git gui, because that's the only thing bothering me.	12:03
	And everyone told me not to use it...	12:03
	So I made my case	12:03
TGEN	there's an official git gui?	12:04
pflanze	ah -- well I don't use it either :), so I'll leave that open to whoever uses it.,	12:04
	→ Yuuhi joined	12:04
pflanze	iirc it's in tcl/tk, so you might change it easily enough yourself.	12:04
drizzd_	yeah, it's already on my todo list :-)	12:05
	TGEN: yes, git-gui is in official git	12:06
TGEN	hmm, well, of it's written in Tcl I won't bother with it anyway	12:06
	(though gitk is useful)	12:06
drizzd_	yeah, tcl is a pita	12:07
pasky	it's usable	12:08
	but hacking it is no-go for me either :)	12:08
swombat	bah	12:09
TGEN	if I were to use a gui for e.g. committing you'd get stuff like :wq at the end of commit messages all the time ;)	12:09
swombat	i managed to switch to the production branch with git checkout -f production	12:09
	now trying to merge master in and getting:	12:09
	git merge master	12:09
	fatal: Entry 'vendor/plugins/rspec/.autotest' would be overwritten by merge. Cannot merge.	12:09
	i've already removed that whole directory and committed the remove	12:09
	(the rspec directory i mean)	12:10
	(which, in master, is a submodule)	12:10
	← lea_w left	12:11
swombat	anyone know how i can just squash whatever's in the production branch with whatever's in the master branch?	12:14
	I'm not too bothered about keeping the history in the production branch at this point	12:14
	maybe with a special merge strategy?	12:15
jengelh	git rebase -i	12:15
	filter-branch also does it, with grafts	12:15
swombat	git rebase -i isn't doing anything	12:15
jengelh	well	12:16
	git rebase -i --onto <here> <here>	12:16
	and then let it squash	12:16
swombat	<here>?	12:16
jengelh	an appropraite ID	12:16
swombat	would git rebase -i --onto HEAD do it?	12:17
	k, got it	12:18
	i think	12:18
	thanks again	12:18
pflanze	drizzd_: I've played with git-gui and can write multi-line commit messages just fine (version 1.5.5.1)	12:20
drizzd_	Yes, but I have to break lines manually. You don't?	12:21
pflanze	I'm entering the commit message into the text box; do you mean from the commandline?	12:21
	dwmw2_HEL → dwmw2_gone	12:22
pflanze	or, what do you mean by having to break lines manually.	12:22
drizzd_	no, I mean the text box. When I write a long line, it doesn't wrap around but it makes a scroll bar at the bottom...	12:22
pflanze	ah. well.	12:23
drizzd_	and it starts scrolling to the right indefinitely	12:23
pflanze	I see.	12:23
drizzd_	actually, I don't even have a scrollbar	12:23
	→ lea_w joined	12:23
drizzd_	it sucks even more than I realized	12:23
	but I think it used to be fine a long time ago.	12:24
pflanze	Maybe someone thought that it sucked to not be able to write overly long lines?..	12:24
	you could check it's history (the git repo history) of course.	12:25
drizzd_	It kinda defeats the purpose of the box.	12:25
	I think it's probably some change to tcl/tk, I'll check it out later.	12:26
pflanze	you really want something like emacs'es reindent to selectively reformat paragraphs.	12:26
drizzd_	But the fact that nobody bothers to change this tells me that there's actually very few people using it seriously.	12:26
pflanze	ehr not reindent -- whatever. Meta-q	12:26
drizzd_	Well, it would be a great improvement if it would wrap the lines just visually, So I can actually _read_ what I wrote.	12:26
pflanze	You do see that sometimes you want to put source code fragments into a commit, which could have long lines which you don't want to have wrapped.	12:27
	→ LotR joined	12:27
pflanze	So the thing to go for is, hit meta-q or something similar after done writing a paragraph which you know you want to have broken into short lines. And soft wrap while writing, yes.	12:28
	(Depending on your preferences, you might even want to embed git-gui into the editor, so you've got all the text handling functionality you'd ever want.)	12:29
drizzd_	I don't know, I seriously doubt the current behaviour is reasonable for anyone. Even if you don't want the lines to wrap (visually or in the actual commit message), you'd at least need a scrollbar. And we can make the rest configurable.	12:29
	→ ggeecko joined	12:30
drizzd_	yeah, git-vi would be cool	12:30
pflanze	git-gui is actually kinda neat, but I'd prefer a good emacs mode (but I haven't even checked if one exists now).	12:30
	← stillLotR left	12:32
	← Yuuhi left	12:34
	← ggeecko left	12:34
Pieter	whoo	12:37
	→ Yuuhi joined	12:37
jengelh	export EDITOR=joe; git-commit ... ^K^J = autoalign	12:37
Pieter	repository shrunk from 4.1GB to 120MB	12:37
	\|kukks\| → kukks	12:37
jengelh	Pieter: did what?	12:38
pflanze	(Anyway, git-gui doesn't seem to be able to select individual chunks for commit either.)	12:39
Ilari	Pieter: How big it was in the original SCM?	12:39
reaVer	what was the original SCM?	12:40
Pieter	Ilari: 300MB	12:40
	bazaar	12:40
reaVer	ah	12:40
	btw, how can I make git rebase autofetch?	12:41
Pieter	git pull --rebase	12:41
reaVer	ah ok:P	12:42
	tnx	12:42
Pieter	or, if you do a git config branch.branchname.rebase true, you can do just git pull	12:42
Ilari	Setting that is handy for private topic branches.	12:47
chris2	pflanze: if you like emacs, try gitsum http://github.com/chneukirchen/gitsum/tree/master but beware it's alpha	12:48
	→ VitaRara_ joined	12:52
	← VitaRara_ left	12:52
	→ jelmer joined	12:57
pflanze	chris2: thanks; although I've never felt the need for selective commits (drizzd_ did), but I'll check it out some time.	12:59
	(selective on the level of thunks)	12:59
chris2	i'm very used to it :)	12:59
	due to darcs mainly	13:00
drizzd_	Unfortunately, I don't like emacs. But maybe I can copy the idea for a vi mode. ;-)	13:01
chris2	hehe	13:01
	it's just a mode to edit git diff and then apply it	13:01
	→ holizz joined	13:02
drizzd_	I was thinking, it would be cool to have a mode which marks the modified and staged lines in the editor. And then you can stage/unstage individual lines or a selected number of lines.	13:04
	→ Ademan_ joined	13:10
	← chris2 left	13:11
	← Ademan left	13:15
	→ madewokherd joined	13:18
	← gitte left	13:22
	← tommorris left	13:23
lea_w	Is a tab width of 8 correct for git's perl modules?	13:24
jengelh	a tab is a tab, nothing else	13:24
lea_w	yes it is ;-)	13:24
jengelh	what your editor makes out of it is its own business	13:24
Pieter	nah	13:24
	that's not really true	13:24
thiago_home	as long as you use tab for indention and spaces for alignment, it works	13:24
	otherwise, it looks just plain ugly	13:24
jengelh	and that's the supposed way	13:24
	a number of people fail to understand that sigh	13:25
thiago_home	so, to make things practical, don't use tabs :-)	13:25
lea_w	this_is_a_long_function_name(42,	13:25
	<indentation here>	13:25
jengelh	actually <alignment here>, not <indent>	13:25
lea_w	Right.	13:25
Pieter	that's not what always happens in git's code	13:25
lea_w	But my editor doesn't know the difference. So are those 28 spaces equal to 3 tabs + 4 blanks or 7 tabs?	13:26
jengelh	that's because it's an imperfect world	13:26
	lea_w: 28 spaces are 28 spaces.	13:26
	→ pombreda joined	13:26
jengelh	some editors reformat code behind your back, they are the worst.	13:26
lea_w	jengelh: How do I do this with Emacs? I.e., getting indentation as tabs and alignment as blanks?	13:26
jengelh	I don't use emacs, thankfully.	13:26
thiago_home	emacs uses, by default, a mix of tabs and spaces	13:26
lea_w	jengelh: What do you use?	13:26
jengelh	joe and mcedit	13:27
Pieter	vim \o/	13:27
thiago_home	especially since the default indentation is 2 spaces	13:27
Pieter	oh :)	13:27
jengelh	yeah and vim	13:27
	as a fallback	13:27
thiago_home	however, Emacs with spaces-only indentation works much better	13:27
	M-x set-variable indent-tabs-mode nil	13:27
lea_w	Good, so say my editor messes up and replaces alignment blanks with tabs, and I need to edit a file that uses indentation with, what is the most reasonable default? 4 blanks = 1 tab or 1 blanks = 1 tab?	13:28
	uses indentation with tabs*	13:28
jengelh	8 blanks to 1 tab	13:28
lea_w	jengelh: Thanks!	13:28
jengelh	but i'd rather turn that stupid setting of automatic tab<->space transformation off, now.	13:28
	→ tommorris joined	13:29
lea_w	there's no way to get it right with emacs, i believe	13:29
thiago_home	if your editor messes up, don't edit the file	13:29
	use another editor	13:29
	with Emacs, you can probably make it work right, but if it's not one of the easy modes, I don't know how to do it	13:30
jengelh	then again, I would have expected emacs be at least a bit saner than those windows editors	13:30
thiago_home	Emacs can do anything	13:30
jengelh	but perhaps it does the wrong thing _by default_ :)	13:30
lea_w	"can do anything given infinite time to write Lisp code"	13:30
	which I don't have	13:30
jengelh	haha	13:30
	so don't use it, simple as that	13:31
lea_w	hates-emacs-and-needs-to-switch-at-some-point	13:31
jengelh	"Like it or lump it", as the saying goes.	13:31
thiago_home	by default, it uses default indentation of 2 spaces and a mix of tabs and spaces	13:31
	← Ademan_ left	13:31
jengelh	the GNU "style", if you can call it that	13:31
thiago_home	editing a text file with Emacs without setting up the indentation mode properly is a pain in the a**	13:31
	it'll reindent your line at semi-colons, parentheses, etc.	13:31
lea_w	thiago_home: Majorly insane indeed. When you use global-whitespace-mode (or so), it stops using tabs, which is fine most of the time. Until you edit files with tabs.	13:32
thiago_home	which is why we (Qt) converted to all-spaces a while ago	13:32
lea_w	Thanksfully, the oh-so-powerful Emacs is unable to automatically recognize that a file uses tabs and keeps inserting rows of blanks. Yuppee. shakes-head</rant>	13:32
thiago_home	no issue with any editor	13:32
jengelh	just use vim or nano, will ya	13:33
	so as to get started with some sane tab settings	13:33
lea_w	Yup, I suppose I'll move to vim...	13:33
	→ Ademan_ joined	13:34
lea_w	pasky: You say I should move minor cleanups to separate patches.	13:35
	What's the best way to do minor cleanups then?	13:36
	I don't exactly want to flood the mailing list with patches that just move lines around, add two comments, or fix a typo.	13:36
	Should I wait and then submit a larger cleanup patch?	13:36
jengelh	you should not do any cleanup	13:36
lea_w	Why not?	13:37
jengelh	to which project?	13:37
lea_w	git (perl/Git.pm)	13:37
jengelh	because then everyone starts sending "cleanups", the linux list is flooded with that already regularly	13:37
	it gets much worse in projects which do not have a CodingStyle set into stone	13:37
	because people then change it forth and back	13:38
	you can introduce a proper style when you write new lines, that's the best way of introducing them	13:38
	gradually over time improving the code	13:38
lea_w	That's why I just sent my cleanup patch as part of another patch (which is how I've been doing it in other projects), but pasky objected: mid.gmane.org/20080531130450.GI18781@machine.or.cz	13:38
pasky	I'm personally fine with small cleanup patches, but I don't know the currently prevailing mileage at git@	13:41
jengelh	lea_w: See, readability is not functionality.	13:41
	← pombred1 left	13:41
jengelh	for example, when you removed the lines from get_head_hash (to write new ones), you can write the new ones in the preferred style without flagging everyone's "split this" mentality	13:41
pasky	lea_w: I don't personally have anything against cleanup like this, but it'd be again greatly preferred to be consistent - are there no other places in gitweb where you could fix this?	13:43
	if not, I'm fine with the patch; if yes, you should fix all the places when you're cleaning up already	13:43
lea_w	okay...	13:44
	i'll remove it from the patch for now	13:44
	← GyrosGeier left	13:47
	← hipertracker left	13:48
	→ mattf joined	13:52
	→ vbabiy-laptop joined	13:54
	→ VitaRara joined	13:54
vbabiy-laptop	Hey guys if I tag a local branch and I push it a remote branch will it carry over the tags. Or do tag only reside locally?	13:54
jengelh	only the branch	13:55
vbabiy-laptop	jengelh: so is there any way to push a tag to a remote branch?	13:55
jengelh	git push <remote> <tagname>	13:55
vbabiy-laptop	jengelh: but will that push the tag or the code at that stage?	13:57
jengelh	the tag, and everything it depends upon.	13:57
vbabiy-laptop	okay	13:57
VitaRara	jengelh: Thanks	13:59
	I'm working with vbabiy-laptop	13:59
	Still trying to git it.	13:59
qknight	tags are used for releases?	13:59
jengelh	primarily yes	13:59
qknight	ok, then i know them from svn. good to know	13:59
	that git has them as well	13:59
vbabiy-laptop	qknight: that is how we are using them	14:00
jengelh	svn does not really have tags.	14:00
	neither has it branches; it just has directories	14:00
qknight	jengelh: yes don't to into detail too much. i just used it that way	14:00
	← tommorris left	14:00
	← mstormo left	14:02
	→ eilartt joined	14:03
	← rubydiamond left	14:03
	← thiago left	14:05
eilartt	how can i retrieve the files of a certain commit?	14:06
	← bentob0x left	14:06
Ilari	Tags are not associated with branches. They are branchlike entites themselves.	14:06
	eilartt: 'git checkout <revision> -- file1 file2'	14:06
eilartt	revision can be a commit as well?	14:06
	→ tommorris joined	14:07
Ilari	eilartt: It can be any valid commit or expression resolving to commit.	14:07
jengelh	<? extends committish>	14:08
eilartt	ok, thanks	14:08
	→ koke joined	14:08
	→ Lucine2 joined	14:11
	← priidu left	14:11
Ilari	eilartt: Examples: 'master~5' (five commits before tip of branch master, using first parents), 01234567 (commit with numeric ID starting with 01234567).	14:11
pflanze	Numeric id's?	14:12
jengelh	and?	14:12
Ilari	pflanze: Every commit has 40 hex digit ID.	14:13
pflanze	but it's not numeric	14:13
	well	14:13
	if you count hex numbers as numbers, then yes	14:14
jengelh	it is counted as hex numbers	14:14
pflanze	but it's not a number in the sense of numbering something, still	14:14
	it's just binary data.	14:14
	a hash.	14:14
holizz	So ... I did git reset --hard HEAD^, and then git commit -ac ORIG_HEAD. But now it complains when I try to push to other repositories.	14:16
	! [rejected] master -> master (non-fast forward)	14:17
Ilari	holizz: It is considered bad practice to try to edit what is already pushed.	14:18
holizz	I see. So what _should_ I be doing when I want to revert the tree to a previous revision? (And how do I unbreak what I have?)	14:20
pflanze	holizz: you can either undo your reset by using git-reflog; or if you want to risk making other people unhappy, there's a force flag to git-push	14:22
Ilari	holizz: Maybe revert the reset and then do 'git revert HEAD'?	14:23
	holizz: And if you really wanted to get rid of last commit in chain, it would be just 'git reset --hard HEAD^'. None of that SVN branches-are-not-first-class nonsense.	14:25
holizz	Thanks. But how do I reset it if the original commit has disappeared (at least as far as gitk is concerned)?	14:27
	→ Sonderblade joined	14:28
pflanze	holizz: as I said, use git-reflog and you'll see the old commits	14:28
	then you can git reset back to the right one.	14:28
pygi	or just use git checkout sha	14:28
pflanze	yeah, would work, too	14:28
	but to get the sha, you use git-reflog. Or, if you know already that it's 'HEAD@{1}' then I think you can use just that string.	14:29
	→ a-priori joined	14:29
pygi	pflanze, true =)	14:29
holizz	Aha, thank you. I think I'm back to a sensible state.	14:30
	← a-priori left	14:30
	→ a-priori joined	14:31
Ilari	holizz: After you have undone that reset, maybe 'git revert HEAD' would work to revert the change in newest commit (it should commit the revert automatically).	14:32
holizz	Thank you very much. It all works happily now.	14:37
SRabbelier	What's the proper plumbing way of showing a commit, but not the patch? ('git show <commit>' does what I want except that it also shows the patch)	14:43
	← strangy left	14:43
thiago_home	git cat-file -p <id>	14:43
	← eilartt left	14:43
	→ strangy joined	14:44
SRabbelier	thiago_home: almost what I want, it doesn't do the indentation of the log msg?	14:44
	and it skips the colons	14:44
	← strangy left	14:45
Pieter	yeah, that's the raw message	14:45
SRabbelier	mhh, I say plumbing, but what I want is to just show one commit	14:45
Pieter	otherwise use git log or git rev-list	14:45
SRabbelier	in the regular format :-/	14:45
Pieter	git log -1 commit	14:45
thiago_home	SRabbelier: the indentation comes from git log	14:45
	git log commit^!	14:46
SRabbelier	thiago_home: Ah, in that case I should just go for git log :)	14:46
	will that also work with multiple refs?	14:46
thiago_home	no	14:46
SRabbelier	how could I do that?	14:46
	do I have to call git log for each ref? 0.o	14:46
thiago_home	git log only works in ranges	14:46
SRabbelier	drat	14:46
	that's gonna be expensive :(	14:47
thiago_home	if the commits you want are sequential, then it can show	14:47
	otherwise, no	14:47
SRabbelier	ah well	14:47
lea_w	I'm trying to create a test suite for Git.pm, and I'd like to add tests for it under the t/ directory. t/README tells me the meaning of the starting digits 0-9, but none of them seem to match for Git.pm (a Perl API).	14:49
	→ GyrosGeier joined	14:50
SRabbelier	lea_w: how about 9?	14:50
lea_w	The closest would be using 9 ("the git tools") as a catch-al	14:50
SRabbelier	lea_w: I think the numbering is mostly for grouping/ordering, I don't think anyone will freak out over the Git.pm test suite being under 9?	14:51
lea_w	SRabbelier: one problem is that we might run out of digits for the second digit (the "command" we are testing) -- there are only 3 left ^^	14:51
SRabbelier	lea_w: heh, yeah, that's what I noticed too when reading that README	14:52
	just lump them all under t97... ? :P	14:52
lea_w	SRabbelier: Perhaps add hex digits a-f?	14:53
SRabbelier	hehe, nice hack :P, wonder if the ordering still works though	14:53
lea_w	a could be used for API's	14:53
Ilari	It should still work then.	14:53
SRabbelier	Ilari: the 1-9 get executed before a-f?	14:53
	btw, why limit just to hex?	14:54
lea_w	SRabbelier: They should be	14:54
	SRabbelier shrugs	14:54
SRabbelier	in that case	14:54
	if we need to, just extend to a-z?	14:54
	→ Lodorius joined	14:54
SRabbelier	or we'll run into the same problem in ~a year	14:54
Ilari	... Since 0-9 are ASCII codes 48-57 and a-z are codes 97...122...	14:54
jengelh	what problem?	14:54
	sha ids are eternal :p	14:55
SRabbelier	Ilari: +1	14:55
	jengelh: we're not wtalking about sha's though :(	14:55
	jengelh: about 4 digit numbers for the tests	14:55
lea_w	SRabbelier: The problem with a-z is that it might accidentally match	14:55
jengelh	heh	14:55
	go 5 numbers	14:55
	→ hagabaka` joined	14:56
lea_w	test2-lib.sh would match for instance	14:56
jengelh	since you count upwards anyway	14:56
SRabbelier	lea_w: I claim trofl.sh	14:56
	← tommorris left	14:56
SRabbelier	jengelh: mhhh, not sure, it's more like a grouped approach	14:56
lea_w	bah, lousy ordering system.	14:56
SRabbelier	jengelh: that wouldn't allow for more commands/groups, only for more subtests per command	14:56
	very much so :P	14:56
	it doesn't scale well at all	14:57
jengelh	use a huffmann code	14:57
SRabbelier	jengelh: ... lol	14:57
	← Lodorius left	14:57
lea_w	okay... so move it to a-f for now?	14:57
	i'll submit a patch...	14:57
SRabbelier	lea_w: bring it up with the list?	14:58
vmiklos	SRabbelier: did you ever get to rewrite your py script to sh, which summarized tests results?	14:58
SRabbelier	vmiklos: eh, no, mhhh, I should consider doing that if I have some time	14:58
	vmiklos: thanks for asking though ^__^	14:58
vmiklos	:)	14:58
	SRabbelier puts a +1 on his 'someone read my proposal' board	14:59
vmiklos	i liked that patch, even i was thinking about adding such a rewrite item to my todo ;)	14:59
	→ tommorris joined	14:59
SRabbelier	vmiklos: really? awesome!	14:59
	SRabbelier cheers	15:00
SRabbelier	It works :)	15:00
	./stats.py	15:00
	lists all the commits that touched the same files as any of my staged changes	15:00
jengelh	uhm	15:00
	git log --author="me" thisfile.c --pretty=format:%H ?	15:00
SRabbelier	jengelh: well, yes, but that's not what it does	15:01
	erm, skip the --author filter for one	15:01
jengelh	what does it do?	15:01
	and more interestingly, what does it output	15:01
SRabbelier	and replace thisfile.c with all the files that are staged	15:01
jengelh	hm	15:01
SRabbelier	also, format is the regular commit list format from git log	15:02
	→ mvanholstyn joined	15:02
vmiklos	SRabbelier: i like python as well, but git is written in sh/perl/C, so it's probably better if you don't use python for your scripts if you ever want it to be merged..	15:02
SRabbelier	vmiklos: I'm doing it in python first, and then if I have time spare I'll rewrite to C	15:02
jengelh	{ Python. Regex. Slow. Repeat. }	15:02
SRabbelier	I'm happy with living in /contrib till doing the rewrite	15:02
	← mvanholstyn left	15:02
	→ bobesponja joined	15:03
vmiklos	ah, k	15:03
SRabbelier	vmiklos: mostly for ease of development	15:03
vmiklos	well, as you want, it was just a suggestion	15:04
	SRabbelier nods	15:04
vmiklos	if you write it directly in perl or sh, you can avoid a rewrite	15:04
SRabbelier	jengelh: http://rafb.net/p/rL8pu293.html	15:04
	vmiklos: I don't know perl, and I want to do it in C for speed	15:05
	vmiklos: since some parts are likely to a real cpu-burner on big repos	15:05
vmiklos	hm yes, that makes sense	15:06
SRabbelier	but c doesn't allow the ease of change python does	15:06
vmiklos	sure, writing a prototype should not happen in C	15:06
SRabbelier	I mean, I'm rather good in C even if I say so myself, but python is just better for prototyping	15:06
	exactly :)	15:06
	→ dirk1296 joined	15:08
	← hagabaka left	15:10
	→ ekidd joined	15:14
	← tommorris left	15:15
	→ mvanholstyn joined	15:15
	→ aroben joined	15:16
	← mvanholstyn left	15:18
	→ lea_w1 joined	15:19
	← vbabiy-laptop left	15:24
	→ doener joined	15:25
	← lea_w left	15:28
	→ pombred1 joined	15:32
	← crab left	15:39
	→ sanxiyn joined	15:40
	→ tommorris joined	15:42
	← tommorris left	15:42
sanxiyn	Hi, progress.c seems to use VLA, and it seems to be the only place VLA is used. Would you want this to be "fixed"?	15:44
jengelh	No.	15:44
sanxiyn	(To make it compile on compilers not supporting VLA)	15:44
	jengelh: Why not?	15:44
jengelh	VLA is C99.	15:44
sanxiyn	I know.	15:44
pasky	gah, great	15:44
jengelh	I would not take such a retrograde step.	15:45
pasky	now git log is outputing spurious empty lines	15:45
	→ hramrach joined	15:45
pasky	:/	15:45
hramrach	hello	15:45
sanxiyn	jengelh: Then why is VLA used in this single file only? (It could simplify code elsewhere.)	15:45
jengelh	because no occassions to use VLA in other places have come up? :)	15:45
sanxiyn	jengelh: Well, I can find several easily.	15:45
hramrach	How do you revert a commit? I want to edit the commit message basically but I guess it's not possible that simply	15:46
jengelh	sanxiyn: send a patch :)\	15:46
	make it use vla	15:46
pygi	hramrach, git checkout sha	15:46
RandalSchwartz	hramrach - see "git commit --amend"	15:46
vmiklos	pasky: hey, is the git-mirror script available somewhere? the one you use in repo.git. it seems it's not part of git.git or repo.git at the moment	15:46
hramrach	thanks, I'll try	15:46
doener	hramrach: or for older commits, look into rebase -i	15:47
sanxiyn	jengelh: That would be useless to me. So I would send a patch to remove VLA usage instead.	15:47
pasky	vmiklos: it's in git/repo.git	15:47
	vmiklos: but it's very legacy	15:47
sanxiyn	jengelh: Thank you for an opinion.	15:47
pasky	you should be able to do the same stuff with normal git	15:47
	if you setup the remotes right	15:47
jengelh	sanxiyn: it's a real pity that some compilers are not even c99. I'd wish for them to rot.	15:47
vmiklos	pasky: git remote --mirror does not seem to remove branches, but i may be wrong, i haven't checked recently	15:47
pasky	jengelh: and is gcc finalyl fully c99-conformant itself? ;)	15:48
	afaik it wasn't for long time :)	15:48
jengelh	pasky: at least it supported VLAs for long	15:48
pasky	well that's true	15:48
	but c99 has many hairy stuff	15:48
hramrach	It looks like amend is what I want. Thanks	15:49
jengelh	such as	15:49
pasky	such as what's listed as "Missing" at http://gcc.gnu.org/c99status.html ;)	15:49
	← _zack left	15:50
	→ _zack joined	15:50
pasky	(or Broken - like "variable-length arrays" :P)	15:50
	(no clue what the breakage is, in case you want to ask)	15:50
vmiklos	yes, just checked; git fetch won't remove branches, even if you added the remote with --mirror	15:51
	→ pfs0 joined	15:51
pasky	vmiklos: hmm, strange... there were some discussions around this long time ago but I don't remember the outcome	15:51
	← pombreda left	15:51
pasky	vmiklos: git remote prune?	15:51
vmiklos	let me check	15:52
	← lamont left	15:52
vmiklos	aha.	15:53
	so git fetch origin; git remote prune origin seems to do the trick :)	15:53
	thanks	15:53
	→ lamont joined	15:54
	→ hacim joined	15:54
hacim	I did a git checkout remote/branch -b mybranch ; then I did a git pull, which pulled in a lot of stuff, some conflicting	15:55
	→ igorgue joined	15:55
hacim	now I can't get back to my master branch, because it wants to merge and I've got unresolved conflicts that I do not know how to resolve	15:56
	← sanxiyn left	15:56
doener	you can "git reset --hard" to abort the merge	15:56
	afterwards, you should have no problems checking out master	15:56
hacim	you are right	15:56
	i'm surprised that the git pull failed, because its on a remote repo that I have made no changes to locally	15:57
RandalSchwartz	that's because pull is fetch + merge	15:57
	did you really want to merge?	15:57
doener	RandalSchwartz: well, should result in a fast-forward then though	15:57
hacim	RandalSchwartz: i wanted the latest changes from the remote in that branch	15:57
doener	hacim: which git version is that?	15:58
hacim	(a branch I have not changed, except for the last time I yanked it from upstream)	15:58
	RandalSchwartz still laments that merge has no "--fast-forward-only" switch	15:58
hacim	doener: 1.5.5.1	15:58
RandalSchwartz	in spite of it always being the solution to this stuff	15:58
	→ mohbana joined	15:58
doener	hm, that should have setup the tracking stuff correctly... weird	15:58
mohbana	is there a git book?	15:58
hacim	maybe I setup the remote branch incorrectly?	15:59
	← bryanl left	15:59
doener	hacim: unlikely... I'd check four things: Am I on the right branch? Did I really do no commits locally? Is branch.$name.{remote/merge} correct? Did the remote history get rewritten?	15:59
jengelh	pasky: most of it is done	16:00
	pasky: the rest probably does not have that much value	16:00
hacim	doener: i did a git-checkout -b remote/branch mybranch, so I know my branch is brand new	16:00
	i'm not sure how to check what branch.$name.{remote/merge} is?	16:01
doener	git config --get-regexp branch.master	16:01
	replace master with your branch	16:01
hacim	ok, how do I tell if it is correct?	16:03
	→ mvanholstyn joined	16:04
doener	remote should be the right remote and merge should be the right branch from which you want to merge	16:04
	eg. refs/head/some_cool_branch	16:04
	s/head/heads/	16:04
hacim	doener: http://pastebin.com/m50e48785	16:05
doener	hacim: rewritten history it seems, complain to your upstream (unless they said upfront that that branch is going to be rewritten)	16:06
hacim	thanks	16:07
doener	hacim: IIRC "git remote update" also does non-fast-forward fetches	16:07
hacim	but in my case, I want the latest, so I want it to be fast-forwarded, don't I?	16:07
doener	so you could just do that and hard reset your branch to the new remote state	16:07
jengelh	git remote update also makes forced updates	16:07
	→ datafirm joined	16:07
doener	hacim: so: git remote update && git checkout emacs-mode && git reset --hard emacs-mode/emacs-mode	16:08
	← Tv left	16:08
hacim	doener: that gets me to where the repo was before the update	16:09
	hagabaka` → hagabaka	16:09
hacim	but how do I get to the latest? I suppose i can do: git remote rm emacs-mode/emacs-mode; git add remote ...	16:10
doener	hacim: http://archives.eyrie.org/software/git/puppet -- seems broken, can't clone	16:10
	maybe someone forgot to git-update-server-info?	16:10
hacim	doener: http://archives.eyrie.org/software/git/puppet.git	16:10
	← jeffpc left	16:10
doener	ok, that works	16:11
	→ jeffpc joined	16:11
doener	no	16:11
	http://pastebin.com/m200a0c24	16:11
	→ rubydiamond joined	16:12
hacim	doener: maybe try again? I just did it in a new location and it worked	16:12
	→ rwanderley joined	16:13
	← lea_w1 left	16:13
doener	hacim: failed three times now	16:14
	→ lea_w joined	16:14
hacim	doener: maybe its your version of git? because I've succeeded 2x and not had an error	16:15
	→ Tv joined	16:15
	→ zawir joined	16:16
	→ brosner joined	16:16
rubydiamond	hey...	16:18
	how to use git submodule thing	16:19
doener	hacim: haha, latest master now, and a new error	16:19
hacim	is the only way to fast-forward by merging?	16:19
	doener: weird	16:19
doener	http://git.pastebin.com/m4366771a	16:19
jengelh	no, you can also FF by reset --harding.	16:19
	but it's the destructive way	16:20
hacim	doener: thats really strange	16:20
jengelh	http is strange.	16:21
	called update-server-info?	16:21
	← rubydiamond left	16:22
	→ rubydiamond joined	16:22
hacim	jengelh: its not my repo	16:23
	→ gmcinnes joined	16:25
	→ trochala joined	16:28
	← SRabbelier left	16:28
	→ SRabbelier1 joined	16:28
	SRabbelier1 → SRabbelier	16:29
	← kanru left	16:29
	→ agib_ joined	16:31
	→ strangy joined	16:32
hacim	if I do cherry-picking from a remote, should I cherry-pick the merge commit?	16:35
	→ sr joined	16:36
	→ cmo-0 joined	16:36
Ilari	hacim: cherry-picking is diff-level operation, so cherry-picking merges is special. OTOH, merges shouldn't contain anything important to cherry-pick...	16:37
	hacim: *changeset-level operation, that is...	16:37
thiago_home	they'll contain the merge-result	16:37
	you'd be squashing the merge if you cherry-pick it	16:37
hacim	what would the result of squashing the merge be?	16:38
thiago_home	the merge applied, but without the second parent	16:39
doener	it's not even necessarily equal to a squashed merged. The merge bases between your branch and the branch from which you pick might be different WRT the branch that got merge	16:40
	d	16:40
Ilari	'shouldn't contain anything important to cherry-pick' means that those commits only contain the merge resolution, but it needs to be done again anyway...	16:41
thiago_home	right, good point	16:41
cmo-0	what does git-upload-pack do ...?	16:41
thiago_home	cmo-0: it uploads a pack file to git-receive-pack	16:41
	← rwanderley left	16:41
	→ rwanderl` joined	16:41
cmo-0	do you mean git-daemon uses it behinde the sence	16:41
	sence=scene	16:42
	← aroben left	16:42
thiago_home	yes	16:42
Ilari	send-pack/receive-pack from a pair. As do upload-pack/fetch-pack.	16:43
	send-pack and fetch-pack run on client, receive-pack and upload-pack run on server. send/receive pair is used for push, upload/fetch pair for fetch...	16:44
	thiago_home didn't know that	16:44
cmo-0	ok. yesterday and the day before i was complaining in this channel that the git-daemon i have on my office serever hangs around 37% to 50% percentage. while pooling around with 'ps aux \| grep git' I saw ' git-upload-pack --strict --timeout=0 .' and thought that timeout=0 is the clue print!	16:44
	cmiiw	16:45
thiago_home	0 timeouts usually mean "no timeout". Whether that means "immediate" or "forever" depends on the code.	16:45
	you cannot infer anything just by reading the command line.	16:45
	← agib left	16:45
	→ tenpaiyomi joined	16:46
	← tenpaiyomi left	16:46
Ilari	cmo-0: Does the upload-pack process stick around after it hangs on client?	16:47
	→ pfs0_ joined	16:47
	→ panagos_ joined	16:47
	→ _Ramblurr joined	16:47
	→ seebyr joined	16:47
	→ dmitry_afanasiev joined	16:48
	← jeffpc left	16:48
	← joevandyk left	16:48
	← solars left	16:48
	← Vantage left	16:48
	← wolf^ left	16:48
	← aazar left	16:48
	← jengelh left	16:48
	← pasky left	16:48
	← cschneid_ left	16:48
	← aporetic left	16:48
	← slack left	16:48
	← thijso left	16:48
	← lydgate left	16:48
	← datafirm left	16:48
	← igorgue left	16:48
	← pfs0 left	16:48
	← bobesponja left	16:48
	← dirk1296 left	16:48
	← a-priori left	16:48
	← Sonderblade left	16:48
	← johan-s left	16:48
	← cedricv left	16:48
	← robin_ left	16:48
	← loops left	16:48
	← tjafk1 left	16:48
	← jwasson left	16:48
	← Ramune left	16:48
	← manveru left	16:48
	← pdherbemont_ left	16:48
	← gnufied left	16:48
	← tcoppi left	16:48
	← panagos left	16:48
	← sbeyer left	16:48
	← jarod left	16:48
	← mtodd_ left	16:48
	← christastrophe left	16:48
	← comp left	16:48
	← iabervon left	16:48
	← mugwump left	16:48
	← gitster left	16:48
	← microm left	16:48
	← jpieper left	16:48
	← Ramblurr left	16:48
	← albertito left	16:48
	← rikur left	16:48
	← tchan left	16:48
	← gcl left	16:48
	← Zycon left	16:48
	← grasshopp left	16:48
	← Mikachu left	16:48
	← eikke left	16:48
	← up_the_irons left	16:48
	← efaistos_ left	16:48
	← halfline left	16:48
	← tsuna left	16:48
	← mutex left	16:48
	← Husio left	16:48
	← cehteh left	16:48
	← normalperson left	16:48
	← reaVer left	16:48
	← ThunderChicken left	16:48
	→ Husio joined	16:49
	→ bobesponja joined	16:49
	→ comp joined	16:49
	→ reaVer joined	16:49
	→ rikur joined	16:49
cmo-0	yes for a while ... i found using ps a couple (approx. 6) half are defunct	16:49
	→ normalperson joined	16:50
	→ jarod joined	16:50
	→ tsuna joined	16:50
	→ robin_ joined	16:50
	→ kanru joined	16:50
	← com4 left	16:52
cmo-0	Ilari: yes for a while ... i found using ps a couple (approx. 6) half are defunct	16:52
	→ gcl joined	16:52
	← zawir left	16:52
	← jacobat left	16:53
	→ ThunderChicken joined	16:53
Ilari	cmo-0: Is there anything unusual in the SSH connections, like using 'ProxyCommand'?	16:53
	→ johan-s joined	16:53
cmo-0	i don't use ssh. only git:// not ssh+git://	16:53
	→ microm joined	16:54
	→ up_the_irons joined	16:54
	→ efaistos joined	16:54
	→ albertito joined	16:54
	→ jeffpc joined	16:54
	→ joevandyk joined	16:54
	→ solars joined	16:54
	→ Vantage joined	16:54
	→ wolf^ joined	16:54
	→ cschneid_ joined	16:54
	→ aazar joined	16:54
	→ lydgate joined	16:54
	→ slack joined	16:54
	→ jengelh joined	16:54
	→ thijso joined	16:54
	→ pasky joined	16:54
	→ pdherbemont_ joined	16:54
	→ a-priori joined	16:56
	→ mtodd_ joined	16:56
Ilari	cmo-0: What state are those (non-defunct) upload-packs in? S?	16:56
	→ cyclist joined	16:56
cyclist	jgit is git reimplemented entirely in java?	16:57
Ilari	cyclist: Right.	16:57
cyclist	by the core git devs? i don't understand	16:58
	← ForsakenSoul left	16:58
cmo-0	they disappeared!	16:59
Ilari	cyclist: Main authors of Jgit are Shawn O. Pearce and Robin Rosenberg.	17:00
	→ gitster joined	17:00
	← simmel left	17:00
	→ simmel joined	17:00
	→ samgranieri joined	17:01
	→ jpieper joined	17:01
	→ Sonderblade joined	17:01
	→ cedricv joined	17:02
	→ tjafk1 joined	17:02
	← rwanderl` left	17:02
	→ gnufied joined	17:02
	→ halfline joined	17:02
	→ tcoppi joined	17:03
	→ igorgue joined	17:03
cyclist	ahh the new generation who learned Java in university	17:04
	→ rwanderley joined	17:05
	← ekidd left	17:05
	→ lirit joined	17:05
	→ loops joined	17:05
	→ grasshopp joined	17:05
Ilari	cyclist: Shawn is more known in Git circles for Git-gui.	17:05
	← rwanderley left	17:06
	→ rwanderley joined	17:06
	← lirit left	17:07
	→ lirit joined	17:07
Ilari	cyclist: And Jgit uses various advanced (at least for Java programs) memory management tricks to get good performance.	17:07
	→ Mikachu joined	17:08
	← mvanholstyn left	17:08
	→ cehteh joined	17:10
Ilari	cyclist: Jgit is mainly used by Egit (Git Eclipse plugin). As it was probably the easiest to reimplement Git in Java (so one doesn't have to mess with using non-Java code).	17:10
cyclist	makes -some- sense	17:11
Ilari	cyclist: I wouldn't classify Shawn as 'new-generation university-churned Java programmer'... :-)	17:12
cyclist	how old is he?	17:16
Ilari	cyclist: No idea.	17:16
	← lirit left	17:18
	← gmcinnes left	17:19
	→ lirit joined	17:19
pasky	i think there were some hairy licence reasons for the reimplementation too	17:21
	but i may be misremembering	17:21
cmo-0	Ilari: should i paste the output of 'ps aux \| grep git' on the office server (7 lines of output).	17:21
	← _zack left	17:23
	→ _zack joined	17:23
	← cyclist left	17:25
	→ ckoehler joined	17:25
Ilari	cmo-0: Yea, it could be useful...	17:27
	cmo-0: Paste it to some pastebin...	17:28
	→ ckoehler_ joined	17:30
	← nuge left	17:30
	← SRabbelier left	17:30
	← ckoehler left	17:30
	→ chris2 joined	17:30
	← gram left	17:31
mohbana	is there a git book?	17:31
pasky	there go 15 minutes of my life wasted by debugging code that had my ($self, $minihunk, $line); instead of my ($self, $minihunk, $line) = @_;	17:32
cmo-0	Ilari: http://pastie.caboo.se/206439	17:34
Bombe	And that is reason #11758291 why Perl sucks... ;)	17:34
	→ Paf joined	17:34
	→ SRabbelier joined	17:35
	→ nuge joined	17:35
hramrach	hmm, now I would like to change a deeply buried commit	17:36
	Not that I could not revert the later ones .. but It might be nice to know how it's done	17:36
Ilari	hramrach: 'git rebase -i <commit>^'?	17:37
hramrach	Ilari: but I do not get what that does	17:37
lea_w	mohbana: Search Amazon. ;-) There's one coming out in September.	17:37
hramrach	Ilari: for me it created an anonymous branch that is incompatible with master now	17:38
	← pfs0_ left	17:38
Ilari	cmo-0: Both upload-pack and pack-objects are waiting to become runnable (state S). Are all git processes on local side in state S as well?	17:38
	ckoehler_ → ckoehler	17:38
Ilari	hramrach: Yes, rewriting a commit also rewrites everything coming after it...	17:38
hramrach	Ilari: yes, but I do not have the stuff after that	17:39
cmo-0	Ilari: all are in S+	17:39
hramrach	does that mean I will have to create a new branch an cherry-pick master to skip the broken commit?	17:40
Ilari	cmo-0: That sounds like both ends are waiting for each other... And you have seen Timeout errors from server?	17:40
cmo-0	i do not wait that much. i just C-c it when i see it hangs for more than 10 sec	17:41
	or so	17:41
Ilari	cmo-0: Timeout errors on server end, that is?	17:41
	→ christastrophe joined	17:41
	← samgranieri left	17:41
	← lirit left	17:42
	← radarek left	17:42
	→ pfs0 joined	17:43
Ilari	cmo-0: Also, what OS home PC runs?	17:44
cmo-0	here is a tail of the output git-daemon	17:44
	Ilari: http://pastie.caboo.se/206443	17:44
	I run ubuntu server edition 8.04 on officePC and ubuntu 7.04 on my HomePC	17:44
Ilari	cmo-0: 2008-05-31_17:33:59.27767 fatal: write error (Connection timed out)	17:45
	cmo-0: Linux, that is. Linux is not known for broken TCP stack...	17:45
	→ nuge__ joined	17:45
cmo-0	?	17:46
Ilari	cmo-0: On Linux, killing the task(s) holding the connection kills the connection instantly. Some broken TCP stacks may not do so.	17:47
mohbana	lea_w: what's the name of it?	17:47
	→ mvanholstyn joined	17:48
	← chris2 left	17:51
	_Ramblurr → Ramblurr	17:51
cmo-0	Ilari: on the server i can make 'git clone git://localhost/as-epanel' worked	17:51
	is it a problem in the router?	17:52
lea_w	mohbana: Goodness, http://www.amazon.com/s?ie=UTF8&keywords=git&index=blended	17:52
	→ felipec joined	17:53
Ilari	cmo-0: I'm thinking problem with the connection between the two machines...	17:53
cmo-0	i had some silly thought that may be the isp is tracking and droping packets?	17:54
Ilari	cmo-0: Or bad router at home...	17:54
	← igorgue left	17:54
cmo-0	Ilari: the following works: 'git clone git://github.com/bruce/rtex.git' on either home or server	17:55
	← nuge left	17:58
	→ jeckel joined	17:59
jettero	is git-svn reasonably usable? or am I better off just using svn (for this project that's already in svn)?	18:01
Ilari	jettero: Git-svn is way more usable than SVN...	18:02
	← mohbana left	18:03
jettero	he	18:04
	... heh rather	18:04
	well, they have svk over there	18:04
	I doubt it's that bad with local commits...	18:04
	→ zachinglis joined	18:04
Ilari	jettero: Not that Git-svn is good (in absoulte terms). SVN is just way worse.	18:05
jettero	I just wondered if git-svn was unsupported like my poor beloved cogito	18:05
jengelh	cogito is obsolete..	18:05
jettero	yeah, I didn't know that until it disappaered from ubuntu	18:05
Ilari	jettero: Git-svn is supported. Cognito is dead.	18:05
jengelh	svn is also obsolete ;-)	18:05
jettero	I didn't realize how sucky cogito was until I got tused to pure git	18:05
cmo-0	Ilari: one thing strange. when i do 'git clone git://github.com/bruce/rtex.git' on server through ssh, but from within the office LAN it worked (before 3 hours) now doing it through ssh but from home hangs???	18:06
jettero	svn isn't obsolete, it's tne text crappy lowest common demonitnator	18:06
	at least cvs is finally leaving	18:06
	wow, the git svn clone takes a while	18:07
jengelh	it's svn...	18:07
Ilari	cmo-0: Break out Wireshark?	18:07
doener	yeah, it fetches each and every revision, so you get the full svn slowness	18:07
cmo-0	?	18:07
	sorry ?	18:08
doener	jettero: latest git.git master has some patches that improve performance	18:08
jengelh	you can't improve upon the slowness of svn's side	18:08
	← Tv left	18:08
doener	not sure how much of a difference they make though	18:08
Ilari	cmo-0: Wireshark is packet sniffer. Turn off Promisc mode, as this reduces clutter when debugging from endpoint of connection...	18:09
jettero	jengelh: svn imports much faster than git imports from svn since it has to replay each revision	18:09
	→ gmcinnes joined	18:09
jettero	not that you can do anything about it	18:09
	I think I'm going to be much happier with git-svn than pure svn though...	18:09
	this feels nice	18:09
Ilari	jettero: In fact, git-svnimport was removed from git.git (with designated replacement of git-svn).	18:10
jettero	I'm nervous about trying a dcommit, but I imagine it works reliably	18:11
Ilari	jettero: Just double-check if the commits look sane...	18:12
doener	jettero: check the commits and dcommit -n first to check what and where it's going	18:12
	→ xenoterracide joined	18:13
jettero	-n is dry run?	18:13
jengelh	often	18:14
doener	yep	18:14
jengelh	other times, it's -n(umeric)	18:14
jettero	does dcommit do an svn commit for each of my commits? or does it make one big patch?	18:14
jengelh	jettero: if it made one big patch git would be no better than svn	18:14
	→ gram joined	18:15
jettero	jengelh: please expand that thought... I'm curious about your thinking	18:16
jengelh	gram: just when did you register that domain? :)	18:16
jettero	that's a cool domain name	18:17
sr	huum, is there any way to know who and when a line was deleted? ala git-blame - i know that's a stupid question, but still, i can't figure how to do that :p	18:17
jengelh	jettero: you spend lots of time keeping bits in smaller commits, and then dcommit makes one big pile out of it. If that were the case, you could just use the regular svn client where you also need to commit before making the next eidt.	18:17
	gram: istr that this domain was free a few months ago	18:17
doener	sr: try "git log -p -Ssomething" where "something" is (part of) the stuff that got deleted	18:18
xenoterracide	anyone having problems reaching github via ssh?	18:18
jettero	are you guys talking about mknod.org? cuz that doesn't seem new to me	18:19
	→ Beket joined	18:20
	← zachinglis left	18:20
xenoterracide	does anyone have any idea what this means? http://vim.pastey.net/88734	18:22
	← gitster left	18:23
	→ a-atwood joined	18:23
	→ chrisrb joined	18:24
cmo-0	Ilair: should i look for specific thing? (hint or link or doc) ;)	18:25
	→ putrum joined	18:25
cmo-0	Ilari: should i look for specific thing? (hint or link or doc) ;)	18:26
	← mvanholstyn left	18:27
Ilari	cmo-0: One thing to look for would be what happens immediately before and after the connection cuts off...	18:28
	← pfs0 left	18:32
	← chrisrb left	18:33
	← hramrach left	18:35
	→ csc` joined	18:35
	← cmo-0 left	18:47
	→ Seldaek joined	18:48
	→ NewMonarch joined	18:49
NewMonarch	Hey guys, I'm having a problem. Git keeps removing a frozen copy of rails from my application with every other commit. I have to keep reissuing the add command. What gives?	18:49
	→ aroben joined	18:50
	→ mvanholstyn joined	18:50
	→ _graham_ joined	18:52
	← Seldaek left	18:53
	← Sho_ left	18:55
	→ zachinglis joined	18:57
	→ Angelic13 joined	18:58
	→ Sho_ joined	18:59
	→ hipertracker joined	19:03
	← Sho_ left	19:03
	← zachinglis left	19:04
	← mvanholstyn left	19:07
	→ Sho_ joined	19:07
	← Sho_ left	19:08
	← Ramblurr left	19:08
	→ mvanholstyn joined	19:09
	→ Sho_ joined	19:09
	← Sho_ left	19:09
	← mvanholstyn left	19:10
	← jlh left	19:11
	→ Sho_ joined	19:11
	→ jlh_ joined	19:11
	→ priidu joined	19:11
	← kumbayo left	19:11
	← Ingmar left	19:14
	→ Ingmar joined	19:14
pasky	Someone might find http://pasky.or.cz/~pasky/dev/git/giddy.perl interesting.	19:17
	Basically, a GUI pickaxe frontend.	19:17
	Still pretty rudmental.	19:17
	pasky -> Indiana Jones	19:17
	→ zachinglis joined	19:18
Mikachu	it seems a bit tricky to find the actual change in a commit ,you have to scroll and loo kfor the green?	19:19
	← Angelic13 left	19:19
pasky	it's constructed differently than say gitk	19:19
	it's not meant to browse commits	19:19
	→ bigfleet joined	19:19
pasky	it's meant to get history of certain content	19:19
	(if, after all, everyone says git is meant to track content - this thing is built to _show_ the tracks :)	19:20
	→ capisce joined	19:20
pasky	right-click somewhere inside the file	19:20
	(or, make some selection first)	19:20
Mikachu	yeah i did that, and it limited commits to those that changed that string i think	19:20
pasky	yes... that's the basic idea :)	19:21
Mikachu	but presumably one would then be interested in the diffs of those commits :)	19:21
pasky	I can spawn xterm with git show... ;)	19:21
Mikachu	heh	19:21
bigfleet	Hello, all. I am trying to convert an SVN repo to git	19:21
	→ peeja joined	19:21
Mikachu	well, it looks like it might be nice	19:22
bigfleet	But at some point a git repo was committed to the SVN repo	19:22
	So now the svn fetch does this	19:22
	http://pastie.textmate.org/private/zdqpygfc3is26pjfcuyx2g	19:22
	Any tips for how I can get past this?	19:22
	I don't need any commit history for the git subrepo	19:22
	but would like to be able to keep history for everything	19:22
jengelh	then use amending.	19:22
	then don't use amending :p	19:22
bigfleet	jengelh: I am a total n00b, so I may need more help than that	19:23
jengelh	you're contradicting	19:23
bigfleet	The element that is causing the conflict when doing the fetch	19:23
	← Ingmar left	19:23
bigfleet	I wouldn't mind if that never existed	19:23
jengelh	fetch itself does not cause any conflicts generally	19:23
bigfleet	but I do need the rest of the repo converted	19:23
	→ Ingmar joined	19:24
bigfleet	problem, then	19:24
	What would amending allow me to do? I cannot pull the change that I would need to amend?	19:25
Mikachu	jengelh: i think th eproblem is the svn repo has a .git dir which git refuses to import?	19:27
jengelh	nice :)	19:27
bigfleet	That is the problem, yes	19:27
	← putrum left	19:28
	← capisce_ left	19:32
	← strangy left	19:34
	→ strangy joined	19:34
	← rubydiamond left	19:35
solars	how do I remove a file from the index after I added it to gitignore, and it has been already in the index? man git-remove says "removes a file from the index OR from the working dir" but there is no option to distinguish that	19:38
	→ jwasson joined	19:39
	→ iabervon joined	19:39
	→ Tv joined	19:39
	→ tchan joined	19:39
	→ manveru_ joined	19:39
	→ mugwump_ joined	19:39
	→ eikke joined	19:39
RandalSchwartz	git rm --cached path/to/file	19:39
	what do you mean "no option"?	19:40
solars	gnah, sorry	19:40
	overlooked that :)	19:40
	thanks	19:40
	RandalSchwartz bows	19:40
	RandalSchwartz hits head on desk	19:41
RandalSchwartz	ouch!	19:41
jengelh	RandalSchwartz: doing it repeatedly strengths your skull	19:41
	← bigfleet left	19:42
	← eternaleye left	19:42
	← gmcinnes left	19:43
	pombred1 → pombreda	19:43
	← robin_ left	19:44
	→ capisce_ joined	19:45
NewMonarch	Hey guys, I'm having a problem. Git keeps removing a frozen copy of rails from my application with every other commit. I have to keep reissuing the add command. What gives?	19:49
	← capisce left	19:49
	→ Kagemaru joined	19:51
Kagemaru	hi	19:51
pasky	RandalSchwartz: Speak o' guru, why do I have to to pilgrim to the holy site of CPAN to get Algorithm::Diff on my humble systems?	19:52
RandalSchwartz	Not sure what you mean by "pilgrim"	19:54
	← brosner left	19:55
RandalSchwartz	you mean, why are some things in the Perl core, and others on a network of widely geolocated mirror sites?	19:55
	that answer is simple... would you want the perl.tar.gz to be 100GB? :)	19:55
	RandalSchwartz waits for pasky to respond	19:56
RandalSchwartz	(not to mention, instantly out of date, since there are about 20 updates to the CPAN daily)	19:57
	→ ajonat joined	19:57
	← ajonat left	19:57
pasky	well, but commonly used modules are part of perl.tar.gz ;)	19:57
	pasky feels that he somehow ends up using algorithm::diff for something in half of all scripts he writes ;)	19:58
RandalSchwartz	ok - that's you	19:58
	I haven't used it except for a magazine article	19:58
	→ ajonat joined	19:58
RandalSchwartz	everyone's locality of modules is different	19:58
	→ mstormo joined	19:58
	← vbgunz left	19:59
	→ gmcinnes joined	20:00
pasky	everyone, use algorithm::dif more!	20:00
RandalSchwartz	well, it's pretty unlikely that anything new will be added to the core	20:00
	the usual talk is "how can we remove things"	20:01
	since the CPAN is so easy to install from now	20:01
pasky	hmm, pity :)	20:01
RandalSchwartz	why?	20:02
pasky	people are afraid of scripts with dependencies	20:02
	→ gitster joined	20:02
pasky	it's psychology :)	20:02
	→ capisce joined	20:02
madewokherd	I'm not afraid of scripts with dependencies	20:02
	I just expect sometime to magically handle them for me	20:03
	*something	20:03
	without any work	20:03
RandalSchwartz	people are also afraid of commitment :)	20:03
	mostly guys though	20:03
	and yet, "git-commit" exists	20:03
	← gmcinnes left	20:03
RandalSchwartz	as I tell people, if you're not allowed to install from the CPAN, you're not using Perl.	20:03
pasky	you can always amend ;)	20:03
	→ igorgue joined	20:04
pasky	RandalSchwartz: is it seamless for non-root users to install from CPAN by now?	20:04
	→ pfs0 joined	20:04
RandalSchwartz	it has been for years.	20:05
	perldoc perlmodinstall	20:05
	set a prefix, set a envar	20:05
	and you're done	20:05
madewokherd	so no	20:05
RandalSchwartz	how could it be any easier?	20:05
	perl has to know where to get non-standard modules	20:05
Kagemaru	I'm trying to figure out how to change the default url "git push" uses for my project... I tried remote.rubybot.url without any luck. Could anyone help?	20:05
pasky	my perlmodinstall says just	20:06
	Make sure you have the appropriate permissions to install the module in your Perl 5 library	20:06
RandalSchwartz	and also that it should be installed in there	20:06
pasky	directory. Often, you'll need to be root.	20:06
madewokherd	it requires modifying the environment?	20:06
RandalSchwartz	you expect it to be psychic instead?	20:06
	how would it know?	20:06
madewokherd	know what?	20:06
RandalSchwartz	where to find non-root modules	20:06
pasky	something like ~/.perl ought to be in default @INC and perl -MCPAN should automatically install there if it notices it is not root	20:06
madewokherd	it can't pick a default place?	20:06
pasky	that's how I imagine "seamless"	20:06
RandalSchwartz	and where would it get that path?	20:06
	you have a bootstrapping problem here	20:07
pasky	well, it's, you know - default	20:07
RandalSchwartz	it can't be anythign to do with $HOME	20:07
pasky	why?	20:07
RandalSchwartz	because what happens when YOU run MY code?	20:07
	it'd break	20:07
pasky	how?	20:07
jengelh	RandalSchwartz: it works but I'd have no clue what you wrote :p	20:07
Ilari	Kagemaru: What does 'git remote' give?	20:07
RandalSchwartz	so that's why there's lots of options about that	20:07
	you can put it into the file	20:07
	or you can put it into your environment	20:07
	or some combination of both	20:07
Kagemaru	Ilari: origin and rubybot	20:08
RandalSchwartz	different advantages and disadvantages	20:08
	so it's seamless, when you choose the stitching pattern. :)	20:08
Ilari	Kagemaru: The default remote might be 'origin'.	20:08
pasky	RandalSchwartz: umm, okay :) not satisfyingly seamless for me	20:09
	→ brosner joined	20:09
RandalSchwartz	I still wonder how you would do it better	20:10
Kagemaru	Ilari: push rubybot works now... but it says "Everything up to date", although there's a new file locally but not remote :\	20:10
	→ hein2 joined	20:10
	→ aikidev joined	20:10
	← Sho_ left	20:10
	hein2 → Sho_	20:10
pasky	RandalSchwartz: I already explained once, and I still don't understand what the problem would be	20:10
RandalSchwartz	because what happens when I run your code?	20:11
	← Sho_ left	20:11
pasky	I don't understand	20:11
RandalSchwartz	it won't have access to YOUR installed modules	20:11
pasky	what special should happen?	20:11
	why not?	20:11
RandalSchwartz	because it'll look in MY home not yours	20:11
	since ~ is per-user	20:11
pasky	why would it?	20:11
	ah, you mean it this way	20:11
	well, then you gotta install the modules too	20:11
RandalSchwartz	ok - if you're gonna be that clueless, I won't argue it with you	20:11
madewokherd	I would do it better by using a real package manager :p	20:11
	→ gmcinnes joined	20:11
	← gmcinnes left	20:12
RandalSchwartz	and you can get that behavior right now with envars	20:12
	you set an envar in your .login	20:12
pasky	yes, but you have to set them up manually	20:12
RandalSchwartz	and then you tell perl to install there	20:12
pasky	and that's my point	20:12
RandalSchwartz	so you have that BROKEN behavior already	20:12
pasky	again, manually	20:12
RandalSchwartz	broken in one way	20:12
	the other way that works better is to edit the file to include a "use lib"	20:12
	then it works for both you and me	20:12
	but it's a different thing you have ot do	20:12
	Perl lets you choose which kind of problem	20:12
	there's no way to do both TRANSPARENTLY	20:13
	that's the point	20:13
	you choose which you want to work	20:14
	not sure why that's so hard to reason out	20:14
	you just want a default that will BREAK for some class of people	20:14
	→ Sho_ joined	20:14
RandalSchwartz	I don't think that's a sane default	20:14
pasky	but I don't want to do both, and I never wanted; if I don't have module installed locally in my ~ or system-wide, I don't expect Perl to magically find it elsewhere	20:14
	← aikidev left	20:14
RandalSchwartz	it's also a security risk	20:14
pasky	my problem is that Perl doesn't _let_ me choose, it _forces_ me to choose, and does not pick sensible default	20:14
	← capisce_ left	20:14
pasky	so now it will BREAK of _everyone_	20:14
RandalSchwartz	that's because the default is wrong for most people	20:14
	← ajonat left	20:14
RandalSchwartz	the ~ approach is wrong for most	20:15
pasky	how so?	20:15
RandalSchwartz	even though you would like it as a default	20:15
	dammit - see scrollback	20:15
	I don't like repeating myself	20:15
pasky	I've read all you wrote	20:15
RandalSchwartz	it breaks when you want to use my scripts	20:15
	or when we install things into a shared lib	20:15
	it's not sufficient for most users	20:16
	so why make it the default when it breaks things for most people	20:16
	leads them to a false sense of hope	20:16
pasky	but I almost _never_ care about running scripts of different users on the same system, and I don't see how would almost anyone care	20:16
RandalSchwartz	yeah - you're NOT THE NORMAL PERL USER	20:16
	get it?	20:16
pasky	no	20:16
RandalSchwartz	too bad	20:16
pasky	I think I'm pretty normal perl user	20:16
RandalSchwartz	I think you aren't	20:16
pasky	but why?	20:17
RandalSchwartz	and I would probably know better than you	20:17
	since I've been dealing with crazy perl users for 15 years	20:17
madewokherd	technically, normal perl users don't write or read perl	20:17
RandalSchwartz	but if you won't accept my broad view on this, then be that way.	20:17
pasky	okay, I guess...	20:17
madewokherd	they just have it installed so perl software works	20:17
RandalSchwartz	go ahead, discount my experience	20:17
	I don't care now.	20:17
pasky	tough to argue if you don't state any arguments :)	20:17
RandalSchwartz	MY EXPERIENCE	20:18
	WHAT I HAVE SEEN	20:18
	how do I recount that in IRC?	20:18
	danderson thinks everyone could stand to calm down, it's the weekend, and this is irc, hardly worth getting annoyed over	20:18
pasky	you could give some concrete example or something	20:18
	pasky is not getting annoyed, personally	20:18
RandalSchwartz	yes. MOST installations I've seen PEOPLE are using OTHER PEOPLE'S SCRIPTS	20:18
	for which ~/perl would MOST CERTAINLY BREAK	20:18
	I gave that example now three times	20:19
	maybe even four	20:19
	or five	20:19
Mikachu	RandalSchwartz: how is forcing them to set an env var and then installing the module better than only having them installing the module to the default ~/perl location before they run the script? your argument makes no sense	20:19
RandalSchwartz	how many more times do you want me to type that	20:19
pasky	well, most users I've seen actually use a single-user machine so there are no other people's scripts to use	20:19
	that's my experience	20:19
	and they almost certainly use some perl scripts	20:19
RandalSchwartz	Mikachu - for that type of usage, "use lib" is the proper solution	20:19
	which again doesn't rely on ~/perl	20:19
	or an envarr	20:20
pasky	you're creating some false dichotomy there, that solving one of the problems by default prevents solving the other one	20:20
	RandalSchwartz rolls eyes	20:20
RandalSchwartz	ok - triage time	20:20
gram	jengelh: at least a few years now	20:20
RandalSchwartz	you're not instructable here	20:20
	you won't accept my 15 years of teaching Perl as valid experience.	20:20
	so there's no point in me wasting my time	20:20
pasky	I'm accepting it as valid experience	20:20
RandalSchwartz	you. have. options.	20:21
pasky	but I don't accept blanket statement "I have more experience as you so I'm automatically right and you're wrong"	20:21
RandalSchwartz	if it's just for you, set an envar	20:21
	and that envar will break for others	20:21
	hence you have to "use lib" for others	20:21
	it's a dicotomy BECAUSE IT IS	20:21
	either you have the envar or you don't	20:21
	→ d0k joined	20:21
RandalSchwartz	how many f'ing times do I need to say that?	20:21
Mikachu	whether there's a default to ~/perl or not, you still have to set up a path to use modules from other users, i think that is quite obvious	20:22
RandalSchwartz	so the envar is the WRONG choice if you're sharing	20:22
	and it's the RIGHT choice if you're not	20:22
Mikachu	but why would setting it break other users more than not doing so?	20:22
RandalSchwartz	but in either case, why make it the DEFAULT choice?	20:22
	← zachinglis left	20:22
	RandalSchwartz wanders off	20:22
	← Ademan_ left	20:23
pasky	hm... I guess my argument did not sink in :(	20:24
jengelh	Each method introduces user-defined behavior.	20:24
Kagemaru	Ilari: still there?	20:26
Ilari	...	20:26
Kagemaru	I'll take that as a yes >.>	20:27
jengelh	it's neither	20:27
	← gnufied left	20:27
Kagemaru	push rubybot works now... but it says "Everything up to date", although there's a new file locally but not remote :\	20:27
jengelh	still need to commit...	20:27
Kagemaru	I did	20:28
jengelh	push the right branch	20:28
Kagemaru	uhm	20:28
	I think so?	20:28
jengelh	I don't?	20:28
Kagemaru	there's only master	20:28
jengelh	fire up gitk --all, see where master is, and see where origin/master is	20:28
Ilari	Kagemaru: How did you get that rubybot remote? Via 'git remote add'?	20:29
Kagemaru	git clone	20:29
Mikachu	Kagemaru: if you're expecting the working dir of the remote to update when you push, read the faq	20:29
Ilari	Kagemaru: Uh... Isn't the default remote from clone called 'origin'.	20:29
	← koke left	20:30
Kagemaru	if I do 'git branch' it gives me *master	20:30
	if I do 'gitk --all' I see master and origin/master	20:31
	→ DJH__ joined	20:31
	pasky always wishes he'd called it 'trunk' instead of the silly 'master' name when reading these discussions ;)	20:31
jengelh	trunk is svn is forbidden	20:31
pasky	it makes more sense :)	20:31
	or even 'main'	20:31
	but, what's done is done ;)	20:32
Ilari	Kagemaru: If you do 'less .git/config', which settings you see in section '[remote rubybot]'? There should be 'fetch' and 'url' there...	20:32
	pasky: It is just a default in 'git init'.	20:32
pasky	well, its unfortunate legacy of cogito ;)	20:33
Kagemaru	just [remote "origin"] and [branch "master"]	20:34
	and [core] of course	20:34
	→ Angelic13 joined	20:34
Kagemaru	Mikachu: found it in the FAQ, thanks	20:39
	but what would be the expected way to do it?	20:39
	→ mtodd__ joined	20:39
Mikachu	you found it, but didn't read it? ;)	20:39
Kagemaru	I did	20:39
	it just.... goes a bit over my head	20:39
	u.u	20:39
	something about a hook	20:39
	← Angelic13 left	20:41
holizz	Is there a way to make git-status greppable?	20:42
	What I want to do is something like "svn st \| grep '?' \| awk '{print $2}' \| xargs rm" (deleting all files that the VCS doesn't know about).	20:43
Mikachu	holizz: man git-clean	20:43
holizz	Aha. Thank you.	20:44
Mikachu	also related, git-ls-files	20:44
holizz	git could be the new emacs, it has so many features.	20:45
Kagemaru	oh... I've seen it -_-... I stopped reading there cause I thought the link is only for people with firewall problems :P	20:46
	thanks for the help =)	20:46
	→ WALoeIII joined	20:46
	← mtodd left	20:47
Ilari	Kagemaru: The stuff in [] is section headers. There are options after those section headers.	20:47
	← albertito left	20:47
	→ poildur joined	20:47
	cbrake → cbrake_away	20:47
	→ aroben_ joined	20:48
	→ mtodd joined	20:50
	← poildur left	20:51
drizzd_	how can I push only annotated tags?	20:53
Kagemaru	Ilari: thanks for the help, 'git push' works as expected now =)	20:54
	→ com4 joined	20:54
	→ Ademan joined	20:55
	→ pior joined	20:55
	← pior left	20:55
Ilari	drizzd_: I don't know any built-in way, but use git-for-each-ref and some text manipulation utils to get list of objects of type 'tag' in refs/tags and pass those to git push...	20:56
	← mtodd__ left	20:57
	← aroben left	20:57
drizzd_	ok, because non-annotated tags have no objects associated with them, only refs. right?	20:57
	→ hipertracker_ joined	20:57
	← ToxicFrog left	20:57
jengelh	well they have the commit associated	20:57
	but it's a ref basically	20:58
	→ mvanholstyn joined	20:59
	→ robin_ joined	21:00
	→ d2dchat joined	21:01
drizzd_	ok, so git-for-each-ref refs/tags \| sed 's/.*tag\t//' seems to do the trick	21:01
	and then i can just do \| xargs git push origin	21:02
Ilari	drizzd_: Or 'git push origin `...`' where ... prints the list of those tags...	21:02
	drizzd_: Hmm... That sed looks bit odd...	21:03
	WALoeIII → WALoeIII_em	21:03
drizzd_	maybe it's too minimalistic	21:04
	→ bigfleet joined	21:05
Ilari	drizzd_: It should rewrite all matching lines and remove all nonmatching ones...	21:06
drizzd_	yeah, it does	21:07
Ilari	drizzd_: Also, using the `` operator won't work, since the command should not be executed if nothing matches...	21:07
	← brosner left	21:07
	→ PrOmEte joined	21:07
	manveru_ → manveru	21:07
	← PrOmEte left	21:08
	← aroben_ left	21:08
Ilari	drizzd_: It doesn't appear to remove nonmatching ones...	21:10
drizzd_	Ilari: yes, of course. you're right. mea culpa	21:10
	→ kyevan joined	21:10
drizzd_	git-for-each-ref refs/tags \| sed -n 's/.*tag\t//p' is what we want	21:10
Ilari	drizzd_: You have multiple annotated tags and at least one non-annotated one for testing?	21:11
	← hipertracker left	21:13
	← GyrosGeier left	21:24
jdavis	If I have modifications to some file and I have not added it to the index, how do I revert those changes?	21:27
Ilari	jdavis: 'git checkout -- file'.	21:29
	← priidu left	21:29
jdavis	what do the "--" mean before the file?	21:30
	← Ademan left	21:30
	→ priidu joined	21:32
	→ Ademan joined	21:32
jdavis	Ilari: I'm assuming it means the current branch... anyways, it worked perfectly. Thanks.	21:32
Ilari	jdavis: It means 'filenames follow'.	21:34
	→ pfs0_ joined	21:35
Ilari	jdavis: Current commit/branch is called 'HEAD'. 'git checkout HEAD -- file' "revert file to version in last commit in this branch". Without any commit specified, it defaults to staged version.	21:35
	← mvanholstyn left	21:36
	← Beket left	21:36
jdavis	Ilari: I see, thanks.	21:37
	← _zack left	21:39
	→ Michelangelo joined	21:40
	← d2dchat left	21:41
	← Michelangelo left	21:41
	→ brosner joined	21:41
	→ tommorris joined	21:41
	← strangy left	21:45
	← WALoeIII_em left	21:45
	→ radarek joined	21:46
	→ hipertracker joined	21:47
	← pfs0 left	21:48
	→ WALoeIII joined	21:50
	WALoeIII → WALoeIII_small_t	21:51
	← tommorris left	21:52
	← holizz left	21:53
	← priidu left	21:53
	→ tommorris joined	21:56
	← lucs left	21:57
kyevan	Hmm.	21:59
	Er, wrong window	22:00
	→ d2dchat joined	22:00
	← tko left	22:01
	← ndim left	22:01
	← felipec left	22:07
	→ peeja_ joined	22:09
	← d0k left	22:09
	← _graham_ left	22:10
	→ mithro joined	22:12
	← hipertracker_ left	22:14
	→ zachinglis joined	22:15
	← mattf left	22:16
	← BabelO left	22:18
	jlh_ → jlh	22:18
	pilif → pilif\|away	22:19
	→ gitte joined	22:20
	→ bootie joined	22:20
	→ Ramune joined	22:22
	← peeja left	22:23
	→ igorgue_ joined	22:26
lea_w	Is there some reason the Git test suite uses /bin/sh rather than bash?	22:26
	← igorgue left	22:27
lea_w	I'm trying to implement some a new feature in test-lib.sh, and keeping Bourne Shell compatibility is just painful.	22:27
thiago_home	does it use anything bash specific?	22:27
lea_w	Yup.	22:27
gitte	lea_w: some platforms do not have bash installed by default. For example Ubuntu.	22:28
lea_w	gitte: I'm pretty sure Ubuntu has bash ;-)	22:28
gitte	What bash specifics do you think you cannot live without?	22:28
	lea_w: not by default. That is what I am saying.	22:28
	lea_w: it uses _dash_.	22:28
	So again, what bashism do you absolutely want?	22:28
vmiklos	gitte: hey :) are you before or after the travel?	22:29
gitte	Before.	22:29
lea_w	gitte: thiago_home: I'm trying to get the return value of the first command in a pipe.	22:29
vmiklos	ah.	22:29
gitte	vmiklos: In 7 hours...	22:29
lea_w	So I need pipefail or process substitution.	22:29
gitte	lea_w: why don't you use temporary files for that?	22:29
lea_w	Because I need continuous output.	22:30
	→ ychat joined	22:30
lea_w	It's test output from external test scripts that get piped through sed/.	22:30
	sed.	22:30
	I'd much rather use pipefail than create a FIFO.	22:30
gitte	And "(xyz \|\| do_something_drastic) \| sed" does not do what you need?	22:31
lea_w	(Piping things through FIFO is asking for trouble, I think.)	22:31
gitte	I think you are trying to do something too cute.	22:31
	Test scripts should be simple, first and foremost.	22:31
	→ redondos_ joined	22:31
lea_w	Yes, but you want the output to be readable.	22:31
gitster	I won't accept any script that does not run under /bin/dash; period.	22:31
lea_w	Btw, Ubuntu has bash, only /bin/sh is dash. So how about changing everything to #!/bin/bash?	22:32
gitster	One obvious exception is contrib/completion/git-completion.bash	22:32
	← pombreda left	22:32
gitte	lea_w: I want the output to be readable, but if that takes too cute constructs, I'd rather live without something readable. As long as you can find out _easily_ if it is a bug in the tested program or the test script.	22:32
	lea_w: why are you so married to breaking compatibility?	22:32
thiago_home	lea_w: if your program is too complex, maybe it shouldn't be a shell script in the first place	22:33
	← pygi left	22:33
lea_w	gitte: I'm not married to breaking compatibility! I just want to indent the output of a test script, because otherwise the output will be pretty unreadable.	22:33
	thiago_home: It needs to be in test-lib.sh. :(	22:33
gitte	lea_w: I doubt that indenting changes much, as far as readability is concerned.	22:34
	lea_w: besides, the output will _only_ be read if the test fails.	22:34
lea_w	gitte: Yes it does, trust me. I've seen it.	22:34
	No, the output is continuous "ok" output. I don't care about indenting failure output.	22:34
gitte	lea_w: so wasting time on making it pretty, breaking compatibility with non-bash-yet-posix shells, is, uhm, not desirable.	22:34
lea_w	So what's wrong with turning /bin/sh into /bin/bash?	22:35
	→ VitaRara_ joined	22:35
gitster	there are platforms without bash.	22:35
lea_w	gitte, gitster?	22:35
	gitster: What platform?	22:35
gitte	lea_w: why don't you see that? Other people do prefer other shells, and just _not_ want bash!	22:35
lea_w	confused	22:35
gitster	keep number of files that begins with "#!/bin/bash" to the minimum.	22:36
gitte	lea_w: that's what posix is for: it is a lowest common denominator.	22:36
lea_w	Every platform I know has bash installed	22:36
	Well, okay. shrugs	22:36
gitte	lea_w: anyway, arguing for bash when everybody tells you not to, is not only a waste of time for you.	22:36
	→ pombreda joined	22:36
gitte	lea_w: Windows?	22:36
	lea_w: VMS?	22:36
gitster	Maybe you do not know enough platforms. I have debian and ubuntu installations without bash.	22:36
jast	windows doesn't have posix sh either ;)	22:37
lea_w	gitster: Wow. Okay, never mind then. :)	22:37
gitte	lea_w: small Linux installations with BusyBox?	22:37
jast	oh, yeah. one of those is actually running in my apartment	22:37
gitster	I am not crazy enough to do so on my primary machine, but that is not a good reason to force people to install /bin/bash	22:37
	→ jbunster joined	22:37
jast	i always force people to install zsh ;)	22:38
gitte	Anybody who ever looked at the source code of bash will try to avoid it, anyway.	22:38
jast	just out of spite	22:38
gitte	jast: that is spiteful, alright.	22:38
jast	(in fact i don't. i don't release my shell scripts.)	22:38
	→ raster_burn joined	22:38
	← d2dchat left	22:39
gitte	lea_w: seriously, coming back to your problem: if your output consists only of "ok"s, then I think you might need to rethink the test a bit.	22:39
	lea_w: for example, there is nothing wrong with using separate "test_expect_success" clauses for different tests.	22:39
	← bigfleet left	22:41
	← zachinglis left	22:41
	← WALoeIII_small_t left	22:41
	→ adante_ joined	22:41
	igorgue_ → igorgue	22:42
	← DJH__ left	22:42
	← christastrophe left	22:42
	← ckoehler left	22:42
	← cehteh left	22:42
	← Mikachu left	22:42
	← tjafk1 left	22:42
	← jpieper left	22:42
	← panagos_ left	22:42
	← lea_w left	22:42
	← doener left	22:42
	← adante left	22:42
	← AlexanderS left	22:42
	← ft left	22:42
	← shad left	22:42
	← G_SabinoMullane left	22:42
	← ruskie left	22:42
	← dvorak_ left	22:42
	← bdowning left	22:42
	← ska-fan left	22:42
	← kyevan left	22:42
	← solatis left	22:42
	← jezdez left	22:42
	← tpope left	22:42
	← mika left	22:42
	← pigeon left	22:42
	← halorgium left	22:42
	← sr left	22:42
	← Strogg left	22:42
	← Cap_J_L_Picard left	22:42
	← nutrimatt left	22:42
	← Bombe left	22:42
	← tmm1 left	22:42
	← metze_away left	22:42
	← [dmp] left	22:42
	adante_ → adante	22:42
	← VitaRara left	22:42
	→ bdowning joined	22:43
	← redondos left	22:43
	redondos_ → redondos	22:43
	→ Strogg joined	22:43
	→ kyevan joined	22:44
	→ DJH__ joined	22:44
	→ christastrophe joined	22:44
	→ ckoehler joined	22:44
	→ cehteh joined	22:44
	→ Mikachu joined	22:44
	→ tjafk1 joined	22:44
	→ jpieper joined	22:44
	→ panagos_ joined	22:44
	→ sr joined	22:44
	→ lea_w joined	22:44
	→ doener joined	22:44
	→ AlexanderS joined	22:44
	→ ft joined	22:44
	→ shad joined	22:44
	→ G_SabinoMullane joined	22:44
	→ ruskie joined	22:44
	→ Bombe joined	22:44
	→ pigeon joined	22:44
	→ jezdez joined	22:44
	→ ska-fan joined	22:44
	→ tmm1 joined	22:44
	→ nutrimatt joined	22:44
	→ tpope joined	22:44
	→ halorgium joined	22:44
	→ dvorak_ joined	22:44
	→ solatis joined	22:44
	→ mika joined	22:44
	→ [dmp] joined	22:44
	→ Cap_J_L_Picard joined	22:44
	→ metze_away joined	22:44
	← ska-fan left	22:44
	← metze_away left	22:44
	→ metze_away joined	22:44
	→ ska-fan joined	22:44
	← solatis left	22:44
	→ solatis joined	22:44
	← Cap_J_L_Picard left	22:44
	← sr left	22:44
	→ sr_ joined	22:44
	← mika left	22:44
	→ pygi joined	22:44
	→ mika joined	22:44
	← ft left	22:44
	→ ft joined	22:45
	← ychat left	22:45
	sr_ → sr	22:45
SRabbelier	how could I go about listing files changed in a commit?	22:47
	← jbunster left	22:47
	→ jbunster joined	22:47
SRabbelier	e.g., I want --name-status	22:47
	but for a specific commit	22:47
Pieter	git log --name-status -1 --pretty="format:" HEAD^^	22:48
SRabbelier	Pieter: it gives a bonus newline, lol :P	22:49
	cos of the format:?	22:49
	also, is it ok to use git log in a script like that?	22:50
Pieter	I think so.. you might want to look at git whatchanged too though	22:51
Mikachu	there's git diff --name-only sha1^..sha1 too	22:51
SRabbelier	Pieter: whatchanged is nice, bug I don't want the log msg -mutters-	22:52
	→ Surfer58 joined	22:52
	Surfer58 Visit http://www.FakeMagazineCover.com (upload pic make mag) - http://www.SillyWebcam.com (play with webcam online) - http://www.Is-A-Jerk.com (insulter/anon email) - http://www.ComedySearchEngine.com (fun) - http://www.BodySwitcher.com (put your face on funny body) - http://www.MedChecker.com (health) - http://www.Canuckster.com (Canada eh) - http://www.Nerdful.com (geeks)	22:52
SRabbelier	Mikachu: mhhh, sounds a bit contrieved, but that would be what I want	22:52
	← Surfer58 left	22:52
	SRabbelier sigh	22:52
SRabbelier	spam on IRC now too? sheesh	22:53
Pieter	SRabbelier: whatchanged takes the --pretty option too :)	22:53
jengelh	SRabbelier: it's older than smtp-spam.	22:53
SRabbelier	jengelh: I thought they gave up on IRC spam	22:53
jengelh	SRabbelier: %-ally, yes. But that does not rule out there are still some idiots around.	22:53
Pieter	fake magazine covers? cool	22:53
jengelh	SRabbelier: also, it only really works when channels are +R or +m	22:54
	→ deavid joined	22:54
SRabbelier	jengelh: yeah, I guess that's true... mhh	22:54
	Mikachu: it doesn't eat the commit	22:54
	Mikachu: complaining about it beign an unknown revision	22:55
Pieter	SRabbelier: it won't work on the first commit in a repo	22:55
	the git log thing will though	22:55
SRabbelier	oh, gah, such a silly mistake	22:55
	ok ok :P	22:55
	→ goibhniu joined	22:56
SRabbelier	I'll go for git log and take the free newline as an added bonus	22:56
	(we need a --pretty=none)	22:56
	→ strangy joined	22:56
Pieter	send in a patch :)	22:56
SRabbelier	why don't we have one?	22:56
	where's pretty parsed?	22:57
	builtin-pretty.c? :P	22:57
gitte	pretty.c	22:57
	--pretty=raw?	22:57
	SRabbelier crawls under a rock	22:57
SRabbelier	gitte: no, not raw, none!	22:58
	like, don't show anything	22:58
gitte	GIT_PAGER="cat > /dev/null" git log	22:58
	;-)	22:58
SRabbelier	gitte: erm, but, I do want the --name-stat part :P	22:59
	what I want is a nice hack to get git log to only give me the --name-stat :)	22:59
gitte	git diff <commit>^..<commit> --name-stat	22:59
SRabbelier	yy, was suggested, but doesn't work for first commit	22:59
gitte	Oh, okay.	22:59
	How about "git log --pretty=format:" then?	23:00
SRabbelier	also	23:00
	it feels kinda.. dirty :(	23:00
	gitte: yeah, that was what we were thinking off	23:00
	gitte: but it gives a bonus newline	23:00
	gitte: which I don't need, hence I came up with --pretty=none :)	23:00
gitte	I think the extra newline comes from the --name-stat part.	23:00
SRabbelier	hmmm, no	23:00
	well, that is	23:01
Pieter	I wonder how much faster a --pretty=none would be compared to --pretty=format:	23:01
SRabbelier	it didn't when I used it with git-diff-index	23:01
	Pieter: possibly not a lot? I just don't want the newline! :P	23:01
	ok, pretty.c is unreadable to the untrained eye :P	23:02
	← thiago_home left	23:02
	→ albertito joined	23:02
Pieter	I think it's okay	23:02
SRabbelier	I think... I need to edit somewhere in the bottom most function	23:02
Pieter	I tried to figure out the insert/get_marks thing in fast-export.c today	23:02
	← DJH__ left	23:02
Pieter	took me some time to figure out what happened there	23:03
	← jbunster left	23:03
	→ rubbish joined	23:03
SRabbelier	/* Make sure there is an EOLN for the non-oneline case */	23:03
	if (fmt != CMIT_FMT_ONELINE)	23:03
	strbuf_addch(sb, '\n');	23:03
	I think that's the offending bit	23:04
	I suspect changing it to if(fmt != CMIT_FMT_ONELINE && fmt != CMT_FMT_NONE)	23:04
gitte	From a simple gdb run, I know that the newline comes from line 315 in log-tree.c.	23:04
	Function show_log()	23:05
	→ d2dchat joined	23:05
gitte	putchar(opt->diffopt.line_termination);	23:05
SRabbelier	gitte: whow, skills	23:05
	← Yuuhi left	23:05
gitte	It's in an if() clause that does trigger when !opt->use_terminator.	23:05
	← panagos_ left	23:06
	← trochala left	23:07
SRabbelier	you're using a different version I think	23:07
	if (opt->use_terminator)	23:07
	putchar('\n');	23:07
	gitte is on 'next' + patches	23:07
SRabbelier	ehe	23:08
	mhh, so I'll just have to parse away those newlines, yes?	23:08
	no easy way around it?	23:08
	→ ironfroggy joined	23:15
	→ simosx joined	23:16
gitte	I think the easiest would be to detect a root commit (via git rev-parse --verify), and replace <commit>^.. by --root, and go with git-diff.	23:17
	→ bigfleet joined	23:18
Mikachu	if you want easiest, just do blabla sha1^..sha1 2> /dev/null \|\| blabla --root sha1 :)	23:18
SRabbelier	you think that is easy? :P	23:19
	gitte: mhhh... I did it with git log, I'll take the "if len(line.lstrip()) == 0: continue" and go with git log	23:19
	no need to check for root	23:19
	← ckoehler left	23:20
gitte	But making the source ugly.	23:20
SRabbelier	and no 'complex' stuff with redirecting output from python :P	23:20
	ugly? why is it ugly?	23:20
	← hipertracker left	23:21
SRabbelier	http://rafb.net/p/jxnojI78.html	23:21
	← brosner left	23:22
	→ nice_burger joined	23:22
	→ sudocoup joined	23:22
	→ mvanholstyn joined	23:22
	sudocoup → gitmachine	23:22
gitmachine	evening all	23:23
	gitmachine → gitdrunk	23:23
SRabbelier	oh dear...	23:23
	gitte: ping? :D	23:23
	gitdrunk → gitonup	23:23
	gitonup → sudocoup	23:23
gitte	SRabbelier: because you are not really logging commits, are you?	23:24
sudocoup	cant have a git like nick in here really	23:24
gitte	You are really interested in the names of the diff.	23:24
SRabbelier	gitte: ah, mhhh, you mean it's "abusing" git log?	23:24
gitte	Yes. Anyway, gotta go.	23:25
SRabbelier	gitte: mhhh, point taken, thanks for the advice :)	23:25
	→ WALoeIII joined	23:27
	← johan-s left	23:28
	→ johan-s joined	23:28
	WALoeIII → WALoeIII_skynet	23:30
Pieter	SRabbelier: http://git.pastebin.com/m7f2d7999 ;)	23:35
	anyway, it's not much quicker than format: :)	23:35
SRabbelier	Pieter: hehe, nice, it still prints the newline though?	23:36
Pieter	I guess.. just add another if ;)	23:37
	→ mithro_ joined	23:37
SRabbelier	Pieter: also, gitte's point still stands, I'm using git log to do something it was "not intended for" (OTOH, there isn't really something to do what I want.. the sha^...sha construct is a bit ugly imho :-/)	23:37
	it sorta violates DRY	23:37
Pieter	there was a debate about that on the mailing list.. there are things that rev-list can do that log can't, and vice versa	23:38
	like --name-status	23:38
	← mithro left	23:38
SRabbelier	ah, yes, I read that :)	23:38
	mithro_ → mithro	23:38
SRabbelier	(well, I read all the mails, but I read that one with slightly more attention :P)	23:38
	what I want is git-showmewahtchangedincommit sha	23:39
	sudocoup → invalidrecord	23:40
	← ironfroggy left	23:41
	invalidrecord → gitlifo	23:44
	gitlifo → gitrsl	23:44
	gitrsl → gitlaid	23:44
	gitlaid → sudocoup	23:45
	→ zachinglis joined	23:46
	sudocoup → invalidrecord	23:47
	← devogon left	23:49
	← zachinglis left	23:54
	→ mithro_ joined	23:57
	← mithro left	23:58
	← gitte left	23:59
	← raster_burn left	23:59

Logs Search ←Prev date Next date→

Channels Documentation