IRCloggy #git 2007-09-25

Provider of IRC logs since 2005.

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

	2007-09-25
aeruder	so git checkout pixman-0.9.5	00:00
underplay	it finds them both	00:00
	HEAD is now at 8ff7213... rasterize traps that extend outside of the image bounds correctly	00:00
	← branstrom left	00:00
underplay	now do git-pull?	00:00
Randal	why pull?	00:00
	you're already checked out	00:00
underplay	Thats downloads all oof 0.9.5?	00:00
	that*	00:01
Randal	download?	00:01
Mikachu	you did that when you cloned	00:01
Randal	you said you git-clone already	00:01
Mikachu	you have all of all versions	00:01
	← rlb3_ left	00:01
underplay	I did, but i wanted to use a specific version	00:01
Mikachu	you are	00:01
underplay	i wanted to pull 0.9.5 specificly	00:01
Randal	and a specfic version is a subset of all versions. :)	00:01
Mikachu	(why didn't you just download the tarball?)	00:01
underplay	couldnt find one	00:01
	← apol left	00:02
aeruder	underplay: you download the ful history, when you do the git checkout you are checking out that particular version	00:02
	you are done	00:02
Mikachu	underplay: "pull" means to get the latest development changes from upstream	00:02
	which isn't what you want	00:02
underplay	oh ok	00:03
	Already, thanks guys, i appreicate your patience and help	00:03
	→ coxy joined	00:06
	← jtoy left	00:07
coxy	how do I get all the commits for a single file and apply them in another repository?	00:07
	→ mithro joined	00:08
	→ theCarpenter joined	00:09
	→ doublec joined	00:09
	→ jtoy joined	00:12
	← fultilt left	00:14
	→ agoode joined	00:15
	→ fultilt joined	00:18
	→ amfwrk joined	00:18
	← jtoy left	00:21
	→ bife-the-wise joined	00:22
bife-the-wise	hello	00:22
Tene	Hi!	00:22
bife-the-wise	I have a gitk question, I'm wondering if this is the right channel to ask	00:23
Tene	This is a pretty good channel for it, I believe.	00:23
bife-the-wise	ok, I'm actually using hgk for Mercurial, which I'm told it's an older version of gitk adapted for hg	00:24
	→ stick joined	00:24
bife-the-wise	I got it to work fine, but I've been struggling to find good docs	00:24
	mainly I'm curious about the meaning of the colors of the lines that connect revisions	00:25
	are they random or is there a method to them?	00:25
Mikachu	i believe they are random	00:25
coxy	is it possible to use git-svnimport to import just a single file?	00:25
bife-the-wise	I see	00:25
	← coxy left	00:37
	→ lyoungz joined	00:37
	← yorgen15 left	00:40
	← tongueroo left	00:48
	→ spearce joined	00:57
	→ mankind_tweezer joined	01:01
	→ jtoy joined	01:03
	→ branstrom joined	01:07
	← mankind_tweezer left	01:13
	← jtoy left	01:14
	→ telmich_ joined	01:17
	← kanru2 left	01:19
	← KirinDave left	01:23
	→ saintdev_ joined	01:26
	← saintdev left	01:27
	saintdev_ → saintdev	01:27
	← telmich left	01:29
	← lyoungz left	01:39
	→ yorgen15 joined	01:39
	→ factor joined	01:40
factor	can I pull just a latest version number with git?	01:40
aeruder	you can grab the latest version generally with the gitweb	01:43
	and i think git can do shallow clones	01:43
Mikachu	--depth 1	01:43
	but that is most likely not a released version	01:43
aeruder	there you go	01:43
	→ stouset joined	01:44
	metze_away → metze	01:45
	→ DrNick joined	01:46
Mikachu	DrNick: aren't you supposed to say "HELLO EVERYBODY!"?	01:49
DrNick	no	01:49
Mikachu	and then we say HI DR NICK!	01:49
spearce	why are we shouting at DrNick ?	01:49
Mikachu	simpsons references	01:49
spearce	isn't that just a reference to Cheers ?	01:50
Mikachu	too young for that :)	01:50
factor	what does --depth 1 do	01:51
	← aroben left	01:51
Mikachu	it fetches only the objects you need to see the last two commits on every branch	01:51
factor	k	01:51
	would try it off hand but GITing kernel tree so will take a bit	01:52
	because I would like to have a few git repositories on my server but would like to have its web front end grab the version number from its git tree after a mirror	01:53
	→ FunkeeMonk joined	01:59
	← metze left	02:09
	→ metze joined	02:10
	← glguy left	02:18
	← metafollic left	02:19
	← jasam left	02:19
	→ metafollic joined	02:21
	← underplay left	02:25
	→ jtoy joined	02:26
	← yorgen15 left	02:30
	→ aroben joined	02:34
	→ branstro1 joined	02:38
	← Ingmar left	02:43
	→ Ingmar joined	02:45
	← leonardop left	02:50
	← branstrom left	02:56
	→ glguy joined	02:59
	→ elmex_ joined	03:05
	← bdiego left	03:05
	← rtmfd_icbm left	03:10
	← brandonz left	03:10
	→ rtmfd_icbm joined	03:19
	← FunkeeMonk left	03:19
	← elmex left	03:19
	← jtoy left	03:21
	→ yorgen15 joined	03:24
	→ Teln1100A joined	03:25
Teln1100A	I am trying to install Git, I used the command make prefix=/usr all doc info here is teh output: http://pastebin.com/m458c78e9	03:26
	→ fhobia joined	03:28
spearce	you don't have libcurl on your system. either install it, or build git without curl by doing `make NO_CURL=1 prefix=....`. building without curl means you cannot clone/fetch/push over http. which may still be acceptable for your needs.	03:28
Teln1100A	I am using centos 5 is there a way to yum install libcurl	03:30
spearce	`yum install libcurl-devel` ? i don't know, i stay away from things named yum...	03:30
Teln1100A	what bout rpm, what do you think of those	03:31
eMBee	having two workdirectories (using git-new-workdir) confuses svn fetch	03:33
	if i try svn-fetch from both	03:34
	→ thiago_h1me joined	03:38
	→ jtoy joined	03:39
	← jtoy left	03:40
spearce	heh: "git-send-pack died with strange error". its not a strange error. i got a sigsegv you moron.	03:46
Randal	you need a stranger error. :)	03:47
spearce	indeed. unfortunately the address it was using was 0x34. not the most interesting of numbers.	03:47
	← thiago_home left	03:51
	← krawek left	04:04
gitster	spearce: were you talking to me?	04:06
spearce	not recently, no.	04:06
	i have a patch for send-pack though.	04:06
gitster	I thought you were saying "you moron" some time ago...	04:06
spearce	yea. to git. not you.	04:06
	unless you think git is yourself. in which case i think you have some issues. :)	04:06
glguy	so either he has issues for thinking he's git, or you have issues for conversing with a program	04:09
	;)	04:09
	→ rodserling joined	04:09
	spearce thinks he's the one with issues	04:09
Randal	zen master to hot dog vendor: Make me One with Everything.	04:10
gitster	Has anybody who's using 'next' got burned by the auto gc yet?	04:11
spearce	i got slammed by a badly timed gc --auto on commit early this morning. really pissed me off.	04:12
	not burned (no damage done) just wasted a lot of my time.	04:12
gitster	which means --auto is not lightweight enough.	04:13
	which part can and should we shave?	04:13
spearce	this morning --auto turned into `repack -a -d` for some unknown reason. when it was done i had 1 packfile and 40 loose objects. not what i expected from an auto.	04:13
	← aroben left	04:14
	spearce sends gitster 3 patches for send-pack segfault issue	04:14
cehteh	auto gc'ing shall never be dangerous? ... why not fork it into the back?	04:14
	(by config variable, not default)	04:14
gitster	The tip of jc/auto which is in next does "repack -A -d -l" when there are more than 20 packs that are not marked with *.keep (notice capital A not -a).	04:15
	maybe "there are too many packs" check went berzerk?	04:15
spearce	that's probably what it was. its my git.git mirror at day-job, fetched over http. probably had a ton of packfiles in there.	04:15
	it was very poorly timed as my system was also busy virus scanning the entire drive (ran all day and never finished), i couldn't abort the $@#(@!(!@( virus scanner, and i had a build running in another window, and my email was downloading... my system was way overtaxed. i didn't expect git-commit to launch into a full repack.	04:16
cehteh	nice it :)	04:17
eMBee	heh	04:17
	maybe the command can check the system load	04:18
mugwump	cehteh: backgrounding gc is dangerous	04:18
cehteh	mugwump: it doesnt --prune or?	04:19
gitster	mugwump: maybe you are not aware that we've been devising undangerous version for this very purpose.	04:19
cehteh	iirc it should (actually must) be not dangerous	04:19
mugwump	I looked over some of the new auto gc stuff	04:19
cehteh	with the cfq scheduler nice even affects io allocation	04:19
	fork in back, run nice .. if thats dangerous the whole auto-gc design is flawed anyways imo	04:20
mugwump	the repack would need to lock the objects and packs it's repacking so no other auto-gc tries to repack them as well	04:20
cehteh	(while i like the idea of auto-gcing .. it just shouldnt spit into my face)	04:20
mugwump	s/'// :-}	04:20
	← branstro1 left	04:21
cehteh	mugwump: we could create a lockfile for that .git/autogc_running or whatever	04:21
mugwump	yes that would do	04:21
	well, apart from when it jams ;)	04:21
	→ branstrom joined	04:21
cehteh	there is no guarantee either that no autogc happens in parallel .. imagine serveral user accessing it over different logins ...	04:22
	jams as in process get killed, stale lockfile left?	04:23
	thats fixable	04:23
mugwump	shared filesystems..	04:23
DrNick	there's an autogc now?	04:23
cehteh	in the works	04:23
spearce	`gc --auto` in next	04:23
DrNick	does it have a "Compacting object database (Press Ctrl-C to safely interrupt)... 30%" ?	04:24
cehteh	even shared filesystem give some guarantees about atomic operations	04:24
DrNick	and when you press Ctrl-C, does it remember for, oh, a half hour or so?	04:24
cehteh	gnu-arch did a lot tricks in that area :P	04:24
mugwump	right, but you can't detect a stale lock that way, unless you say "any lockfile not touched in 30min is stale"	04:24
cehteh	thats sufficient for this purpose, i would even write the gc's hostname/pid every once a while to that lockfile then you can check if the gc process is still running with more or less efforts	04:26
mugwump	thing is, by default it doesn't pack until you hit 6700 objects or so. I think a better figure is about 100-200	04:27
	that way it's a really quick repack	04:27
cehteh	having a auto-gc once left out because of a stale lock isnt dangerous at all	04:27
	← rkaway left	04:27
mugwump	so long as it eventually resumes	04:27
cehteh	yes	04:28
	say repacking shall write a host/pid tuple every minute into the lockfile .. and if the lockfile is older than 30 minutes then consider to test if the packer still running ...	04:29
	it might even ring an alarm bell if it thinks there is a stale lockfile and not remove it automatically	04:30
	→ rkaway joined	04:30
mugwump	so, conglomerating small packs should hopefully be less work than making one mighty pack in the first place ... hence the original generational idea	04:31
cehteh	automatic packing should be a) solid and b) optional ... anything else is a bug isnt it?	04:32
mugwump	of course I have no hard figures to base that guess off	04:32
	by building smaller packs more often, and only collating the small ones into larger chunks, you pay the price of a large repack in small instalments	04:32
gitster	and probably end up with way suboptimal packs, I am afraid.	04:33
	That's where the 6700 figure comes from.	04:33
spearce	and the only way to get a better pack is to do a --aggressive/-f which takes a heckva long time on a sizeable project. :)	04:34
gitster	The whole "auto-gc" thing was just to make sure people do not go forever without running _any_ gc.	04:34
mugwump	agreed on that, for sure, but the main point was to avoid horror repositories, not make every repository as small as possible by default	04:34
	I'd think that no auto-GC should ever try to collate more than, say X MB of packfile together	04:35
cehteh	can you do small repacks frequently and big complete repacks infrequently?	04:35
	→ Pieter joined	04:36
gitster	Oh, the main point is to avoid horror repositories, _AND_ not make repositories of people who diligently do their own repack suffer from auto-gc. That is why setting the threshold too low is not a good idea.	04:36
mugwump	right, because the deltas won't be recomputed without -f	04:36
	can deltas be extended without -f?	04:36
	or is any deltified object considered "finished" and not used for deltifying others?	04:37
gitster	if you are already a delta and your base is being repacked then you won't get recomputed.	04:37
eMBee	how can i get rid of all untracked files from my workdir?	04:38
cehteh	git clean	04:38
eMBee	ah, thanks	04:38
gitster	also I think if you are already in a pack and you are not a delta, you won't get recomputed either.	04:38
cehteh	gitster: so what speaks against going into background?	04:39
gitster	that it does not buy you much?	04:39
	you can send the stuff to the background with "^Z; bg". I do not think backgrounding is an issue.	04:40
cehteh	huh? .. it can continue working cant i?	04:40
mugwump	or you can put git-gc --auto & into a hook	04:40
cehteh	well i mean in a smart way .. writing this lock/pidfile and nicing	04:41
eMBee	hmm, is there a way to search the changes? ie, some change removed some code, now i am trying to find where that was removed	04:41
gitster	git log	04:41
mugwump	eMBee: git log -S, git log -p \| grep	04:41
eMBee	thanks	04:41
mugwump	or use gitk and enter in box "highlight commits adding/removing string"	04:41
cehteh	well ... i admit i didnt looked. .. but why is autogcing not implemented in a hook altogether?	04:42
eMBee	ahh, that's even better, thank you!	04:42
mugwump	cehteh: I sent one to the list that worked like that a while abck	04:42
	gitster: ok, so it seems like repack would want to be able to treat "short" delta chains as possibly able to be lengthened. which sounds like a feature with limited use	04:44
	picking the right deltas can really make a huge difference	04:46
	in the Perl repository import, getting the deltas right, plus allowing very deep deltas on the initial import, made my git-fast-import output pack smaller than one with badly ordered objects that was later repacked with --depth=100 --window=100	04:47
	← fhobia left	04:51
	→ orospakr joined	04:52
	← stick left	04:59
	← Unleashed left	05:07
	← rodserling left	05:07
	thiago_h1me → thiago_home	05:17
	→ rodserling joined	05:28
	→ fhobia joined	05:34
	← DrNick left	05:34
eMBee	yay, cherry-pick handles subtree commits, i am impressed, i expected it to bail out	05:38
	← madewokherd left	05:39
	← plexq left	05:40
	→ _hdl_ joined	05:49
	← spearce left	05:50
	← \|hdl\| left	05:50
	← fhobia left	05:52
	← janm left	05:54
	← doublec left	05:54
	→ phmelak joined	05:56
	→ aroben joined	05:57
eMBee	if i want to replace a bunch of remote branches with a submodule, do i just delete the branches and add the module?	06:01
	← rodserling left	06:08
mugwump	eMBee: yep	06:09
	you don't need to delete the remote branches of course	06:09
eMBee	i don't? ok	06:09
glguy	are submodules new?	06:09
	-ish	06:09
eMBee	new in 1.5.3	06:10
glguy	looks like I need to catch up!	06:10
eMBee	what about the data that is already in the directory where the submodule is to be cloned?	06:13
	→ saintdev_ joined	06:18
	← saintdev left	06:18
	saintdev_ → saintdev	06:19
	← polyonymous left	06:22
	→ polyonymous joined	06:23
	← Sho_ left	06:23
	→ FunkeeMonk joined	06:30
Ilari	eMBee: You need to get rid of it or move it elsewhere in revisions that have submodule there...	06:30
eMBee	ok	06:30
	← saintdev left	06:31
	→ engla joined	06:31
Ilari	eMBee: Because submodules work bit like mount points. You can't modify what is under unless you modify what is linked.	06:31
	eMBee nods	06:32
eMBee	i understand the why	06:32
	just truing to figure out the actual steps i need to take	06:32
glguy	so the idea is you could depend on a library at a specific version	06:32
	and library development can continue	06:32
	without breaking the project?	06:32
eMBee	no, the idea is that a subdirectory of your code lives in a different repository	06:35
	you would still use it the same way as if it were in the same repo	06:35
mugwump	right, but it needs its own url	06:35
Ilari	Probably easiest way would be to remove the remote remote and reclone it where you want it and add that directory. There are tricks to recover most of object database entries so to dramatically reduce amount of data needed for recloning.	06:36
eMBee	as in rebuild my repo from scratch more or less	06:36
	thought about that too	06:37
	trying both	06:37
	problem is one of the repos is svn	06:37
	so the recover trick may not work	06:38
	→ mgrimes joined	06:38
Ilari	eMBee: I'm sure sightly more complicated version of it would work with git-svn. But I haven't played with it...	06:39
eMBee	and another issue is, that one the submodule branches needs to be tracked in both repos	06:39
	well, i'll see. thanks so far	06:40
	→ saintdev joined	06:41
	← saintdev left	06:41
glguy	is a new blob created every time a file is edited and committed?	06:41
	(I suppose that it would have to be..)	06:41
	→ saintdev joined	06:42
gitster	every time a content that your git repository has never seen is "git add"ed.	06:45
	→ pdmef joined	06:49
	→ Pistahhh joined	06:51
	→ devogon joined	06:52
	← vmiklos left	06:52
	→ vmiklos joined	06:58
	→ dkagedal joined	06:58
	← jlh left	07:07
	← Pistahh left	07:07
	← mithro left	07:07
	→ jlh joined	07:08
	Pistahhh → Pistahh	07:09
	← tronical left	07:16
	← rtmfd_icbm left	07:21
	→ tronical joined	07:23
	← mgrimes left	07:27
	→ fowlduck joined	07:28
fowlduck	unexpected behavior here, using the example from git-svn: git-svn clone http://svn.foo.org/project -T trunk -b branches -t tags	07:28
	inside of the project directory is the contents of trunk	07:29
	granted tags and branches are empty in the svn repository	07:29
	if a svn branch were to be added, what behavior could I expect?	07:29
	ahh, fooey, dead in here, methinks	07:32
	← FunkeeMonk left	07:33
	← dkagedal left	07:38
	← devogon left	07:42
	← phmelak left	07:42
	← telmich_ left	07:42
	← loops left	07:42
	← pa-ching left	07:42
	← lazyatom left	07:42
	← Thumper_ left	07:42
	← jtraill left	07:42
	← ojn left	07:42
	← jepler left	07:42
	← ruskie left	07:42
	← tcoppi left	07:42
	← jeremyb left	07:42
	← eMBee left	07:42
fowlduck	spliiiit	07:42
	so can anyone answer my question?	07:43
	→ Thumper_ joined	07:44
	→ devogon joined	07:45
	→ phmelak joined	07:45
	→ telmich joined	07:45
	→ loops joined	07:45
	→ pa-ching joined	07:45
	→ lazyatom joined	07:45
	→ tcoppi joined	07:45
	→ ruskie joined	07:45
	→ eMBee joined	07:45
	→ jepler joined	07:45
	→ jeremyb joined	07:45
	→ jtraill joined	07:45
	→ ojn joined	07:45
	← bcr left	07:48
	→ FunkeeMonk joined	07:49
	→ tango_ joined	08:02
tango_	hm I did a git pull of the git repo today and got this error: error: Object a0e7d36193b96f552073558acf5fcc1f10528917 is a blob, not a commit	08:02
	however, the merge completed successfully. any indication on what could the problem be and what I can do to solve it?	08:02
	← aroben left	08:06
	→ standel joined	08:11
	← standel left	08:11
	→ Bertl joined	08:19
	← glguy left	08:20
Bertl	greetings folks! I have a git question:	08:20
	→ glguy joined	08:21
	← FunkeeMonk left	08:21
Bertl	when I pull/merge from a branch, and get conflicts	08:21
	is there a simple way to say: take all versions from the 'new' branch?	08:21
MadCoder	git merge -s theirs iirc	08:25
glguy	there's a theirs merge strategy?	08:26
MadCoder	okay I'mwrong there is only ours	08:26
Bertl	my git says no :)	08:26
glguy	ours isn't really a "merge" as much as a symbolic dependency	08:26
	as far as I can tell at least, I'm not a merge expert	08:27
	← cortana left	08:27
Bertl	so any ideas how I can avoid going through N files and resolving merge conflicts manually when I _know_ I want the version from the merged in branch?	08:27
MadCoder	well I would use a git macro for that	08:28
tokkee	Bertl: Switch to the other branch and use "ours"?	08:28
	→ markoa joined	08:29
glguy	That strategy instructs git that the merged result should be the same	08:29
	as the current HEAD. Any other branches are recorded as parents, but	08:29
	their contents are ignored.	08:29
	I don't think that that is what he wants	08:29
Bertl	hum, besides that this would make it very confusing because then I'm on the wrong branch :)	08:29
	→ dkagedal joined	08:29
tokkee	True...	08:30
Bertl	is this really such an unusual situation?	08:30
MadCoder	no I've done that many times before	08:31
	but like said I use a macro from my editor for that	08:31
Bertl	so manually then or what?	08:31
MadCoder	because I always want to check it does the right thing	08:31
	it's manually automatic	08:31
	:)	08:31
	you also can use git mergetool	08:32
	with kdiff3 you can tell, always j'ai	08:32
	with kdiff3 you can tell, always keep from left or right side	08:32
Bertl	hmm, I'm kind of sure that I do not want to go through 1000+ files manually :)	08:32
tokkee	Bertl: A hackish way might be to create a diff yours..theirs, apply that to your branch and then merge the other branch...	08:32
MadCoder	1000+ ? heh	08:32
fowlduck	hmm	08:32
MadCoder	Bertl: sed is your friend	08:32
fowlduck	I was hoping it automagically detected new files and added them	08:33
	that sucks	08:33
Bertl	yes, I thought about that ... and I think that should work, but I couldn't believe that there is no simple way	08:33
	after all, git _shows_ me the three versions as hashes	08:33
MadCoder	sed -i -e '/^<<<<<<<$/,/^=======$/d;/^>>>>>>>$/d' yourfile.c	08:33
	that would keep the "second" part always	08:33
	(I think)	08:34
	→ tuxos joined	08:34
MadCoder	yeah it works	08:34
	of course if it's the first you want it would be	08:35
markoa	Is it possible to undo a commit, ie keep the changes it introduced in the working tree?	08:35
MadCoder	sed -i -e '/^=======$,/^>>>>>>>$/d;/^<<<<<<<$/d' yourfile.c	08:35
	markoa: git reset HEAD~1	08:35
	(not --hard !)	08:35
tuxos	is there a general way to show the chronological changes for the last, say 3-5 commits that affect a particular file ?	08:35
markoa	MadCoder, is that a minus or a tilda?	08:36
MadCoder	tuxos: git log -- your-file	08:36
	markoa: a tilde	08:36
	tuxos: of course it would show all the commits	08:36
	but it's fast	08:36
	you can also use git rev-list -- your-file \| head 5 \| git log --stdin or sth like that if you only want 5	08:36
tuxos	no, MadCoder: I want the changeset, not the log	08:37
MadCoder	gitk -- your-file	08:37
	and walk through it	08:37
tuxos	yes exactly - but on the command line?	08:37
MadCoder	yes	08:37
tuxos	I mean: how to on the console?	08:37
	(without X)	08:37
MadCoder	git rev-list -- your-file \| head 5 \| git log -c --stdin	08:37
	should probably work	08:38
	or sth similar	08:38
fowlduck	hmm, any good frontends to git? I never found one I liked for svn, wondering if anything is better on the git side in this regards	08:38
tuxos	rev-list takes a commit, so that doesn't do it	08:38
MadCoder	okay git log doesn't take --stdin	08:38
	it does --all	08:39
	git log -c -- your-file	08:39
markoa	MadCoder, thanks	08:39
MadCoder	that should do it, as it's streamed in less, you can read the result before it's 100% done	08:39
	err -p	08:39
	not -c	08:39
	I'm really confused today morning	08:39
tuxos	MadCoder: thanks a lot, that's it!	08:40
MadCoder	markoa, tuxos: 'welcome	08:40
	fowlduck: how would it do that ? adding any new file in the respository is almost never what you want	08:41
	and I don't think there is a lot of front-ends yet, except git-gui and the eclipse plugins	08:41
tuxos	and another question: when do I ever need reset --soft ?	08:41
fowlduck	MadCoder: that's what I mean, it would be hard to implement, I just hate the old svn move svn copy svn this svn that commands	08:41
MadCoder	tuxos: --soft is the default	08:42
fowlduck	or in this case git	08:42
tuxos	--mixed is	08:42
MadCoder	oh I don't know then	08:42
	I never used it	08:42
fowlduck	svn always broke on me when I did that crap slightly incorrectly, is git so unforgiving? (I'm investigating, presently)	08:42
MadCoder	fowlduck: git does not do mv/cp like svn at all	08:43
fowlduck	oh, it has mv and cp though, so what's the diff?	08:43
MadCoder	for real git does not records anything on move or copy	08:43
	← markoa left	08:43
MadCoder	when you ask for a diff, git cleverly guess which files have been moved or not	08:44
	`git mv a b` is merely the same as doing:	08:44
fowlduck	I'm looking at git --help though and there's mv and cp and rm	08:44
MadCoder	mv a b	08:44
	git add b	08:44
	git rm a	08:44
fowlduck	oh, i see	08:44
MadCoder	it's just a convenient method	08:44
fowlduck	so it doesn't penalize you like svn does	08:44
	good	08:44
MadCoder	but what I just wrote will work _exactly_ the same	08:44
	no	08:44
	you can mv 3 4 10 times before commiting, it will work	08:45
fowlduck	yeah, svn is pretty stupid about all that	08:45
	so frustrating	08:45
MadCoder	yes because it registers the fact that it's a copy (as mv is merly a copy + rm in svn)	08:45
	git doesn't, it guess it each time you ask for it	08:45
fowlduck	kk	08:45
	thanks	08:45
MadCoder	that's the reason why there is a --find-copies-harder to many commands	08:46
	← glguy left	08:47
fowlduck	I'm so new, I didn't even know that	08:48
	← saintdev left	08:50
	→ adante joined	08:51
	← tuxos left	08:51
	→ saintdev joined	08:52
	← yorgen15 left	09:03
	→ cortilap joined	09:09
	dwmw2_gone → dwmw2	09:15
	← laj left	09:18
	→ dooglus joined	09:21
dooglus	where can I get an icon for my 'gitk' launcher?	09:21
	→ zecke joined	09:24
	→ david__ joined	09:26
fonseca	dooglus: Maybe you can use the git.git/gitweb/git-logo.png	09:32
dooglus	it's a bit wide, but thanks :)	09:40
	I cropped out a 24x24 icon: http://dooglus.rincevent.net/random/git-logo-24x24.png	09:46
	and http://dooglus.rincevent.net/random/git-logo2-24x24.png	09:59
	→ chris2 joined	10:01
fonseca	logo2 looks nice.	10:16
dooglus	I like the simplicity of the first one	10:17
simmel	I also like theese http://henrik.nyh.se/2007/06/alternative-git-logo-and-favicon	10:17
fonseca	Oh, yeah, it is used by WinGit ...	10:21
simmel	Oh really?	10:22
	I must tell Malesca this because I don't think he knows = )	10:22
fonseca	For the icon of the downloaded package and the installed program link to the git shell.	10:23
simmel	Got an URL for WinGit? Was hard to search	10:23
	Found it	10:24
fonseca	It is one of the downloads on http://code.google.com/p/msysgit/	10:24
	→ doublec joined	10:27
	← fultilt left	10:28
	→ Yuuhi joined	10:32
	→ Roomster joined	10:53
	→ ferdy joined	10:56
	← Romster left	11:03
	→ GyrosGeier joined	11:12
	← orospakr left	11:19
	→ aruiz joined	11:35
	→ duncanmv joined	11:39
duncanmv	hi	11:39
	→ laj joined	11:39
duncanmv	is the same for git to diff and merge between two directories (different checkouts/repos) than to do it in the same repo across branches?	11:39
MadCoder	you can't merge/diff between two different repositories	11:40
	you have to fetch one in the other, and do the merge in the checkout of the latter	11:41
	if it's a merge, once done, you can push it to the former if appropriate	11:41
duncanmv	ahh ok	11:42
	← doublec left	11:48
	← duncanmv left	11:50
	← factor left	11:50
	jwb_gone → jwb	11:52
	→ plokier joined	11:54
	← plokier left	11:57
	elmarco → elmarco\|away	11:58
	→ kanru joined	12:03
	→ Sho_ joined	12:05
	→ factor joined	12:06
	→ lcapitulino joined	12:08
	← jlh left	12:14
	→ jlh joined	12:14
	→ jtoy joined	12:15
	← jtoy left	12:18
	← factor left	12:35
	→ cmarcelo joined	12:44
Mikachu	hm, i have a shared clone of linux in /tmp, and it has some extra objects in it that isn't in the main repo	12:45
	and i fetched into the /tmp clone and pushed into the main repo a couple of times	12:45
	then i ran git-repack -a -d in the /tmp one, and the size went from 4MB to 84MB	12:45
	the only way i managed to get it down was to move the pack file out of .git and run git-unpack-objects on it, and then rerun git-repack -d (without -a)	12:46
	is there another way?	12:46
	→ yorgen15 joined	12:58
	← jeffpc left	12:58
	→ Ryback_ joined	12:59
	→ jtoy joined	13:01
	→ madewokherd joined	13:08
	← phmelak left	13:11
telmich	someone an idea, why I've merge problems with a newly created .git? = >http://home.schottelius.org/~nico/unix/progs/git/git-pull1	13:27
	← stouset left	13:27
telmich	man git-config says " Without this option, git pull defaults to merge the first refspec fetched."	13:28
	(below branch.<name>.merge)	13:28
	→ bdiego joined	13:30
	← _spuk left	13:31
	→ rodserling joined	13:38
	robfitz_ → robfitz	13:44
	← theCarpenter left	13:54
	← engla left	13:55
hliusv561	Mikachu: Alternates?	13:58
Mikachu	yes, as i said it is a -s clone	13:59
thiago	I'd like to hear some opinions on what would be the "git way" on this issue:	14:00
	there's this project I am developing on top of the master branch of a repository	14:01
	so, I am usually using "git rebase" to get up-to-date with master	14:01
	now, I have people working with me so we want to have a shared repository	14:01
	should I keep on using rebase + push -f?	14:02
Mikachu	i would maybe try and not rebase until there were any conflicts with upstream that needs resolving	14:03
hliusv561	thiago: Rebase and shared repositories do not mix.	14:05
thiago	so, I should merge instead of rebase from master from time to time?	14:06
	→ engla joined	14:06
hliusv561	thiago: Yes. Then after the feature is more or less compelete, rebase it (making sure everybody you are working with that part understands what you are doing.)	14:07
thiago	hliusv561: ok, I see	14:07
	thanks	14:07
Mikachu	(ie, don't keep the merges in history?)	14:07
thiago	yep, I can't keep the merges in history in this particular case	14:07
	I'll probably squash most of the commits anyways	14:07
	→ stick joined	14:10
	→ rtmfd_icbm joined	14:16
	→ p4tux joined	14:17
	→ PerlJam joined	14:20
	→ jcollie joined	14:20
	elmarco\|away → elmarco	14:23
	→ orospakr joined	14:25
	→ plexv joined	14:33
	← plexv left	14:34
	← dkagedal left	14:35
	→ bentob0x joined	14:36
	← rafl left	14:39
	→ rafl joined	14:43
	→ stouset joined	14:50
	→ dkagedal joined	14:59
	← tango_ left	15:00
	← aruiz left	15:04
	→ krh joined	15:06
	← fdr- left	15:15
	→ fdr joined	15:18
	fdr → fdr-	15:18
robewald\|work	hello, I want to use git-format-patch -p HEAD^ with GIT_EXTERNAL_DIFF set, but i don't get a custom diff. just the usual one. When I do git-diff HEAD^ i get what I expect. What am I doing wrong?	15:19
Mikachu	--ext-diff?	15:20
robewald\|work	yes, thats it, thanks :-)	15:22
	→ gitte joined	15:26
	← tcoppi left	15:28
	→ sudoer joined	15:29
	→ tcoppi joined	15:29
	→ leonardop joined	15:31
engla	hm how do I get the name of the latest tag (from a script)	15:31
gitte	I suspect you want to say "git describe HEAD"	15:31
Mikachu	possibly git describe --tags HEAD	15:33
	→ kumbayo joined	15:34
gitte	Mikachu: that will include the light-weight tags. Not sure when you would ever want that.	15:36
Mikachu	that depends on how much you use light-weight tags	15:36
	← kanru left	15:38
	← jtoy left	15:38
	← dkagedal left	15:39
	→ jtoy joined	15:39
engla	describe works well enough	15:46
	← Fullmoon left	15:47
engla	I suppose many people thought of using git describe and git archive to package source tarballs	15:50
	→ bronson joined	15:50
	→ kanru joined	15:50
	→ elmex joined	15:53
	← sudoer left	15:55
kelvie_	is it safe to continue work while git-gc is running?	15:58
	it's at repack.. and it doesn't look like it's going to stop soon (I just checked out our 8 gig or so svn repo and converting it to git :/)	15:58
gitte	Usually, it is.	15:58
	(Safe)	15:58
Mikachu	unless you said --prune, i think	15:58
kelvie_	just git-gc ..	15:59
gitte	Even with --prune, it is usually safe.	15:59
	(It does not prune objects referenced from the index)	15:59
engla	kelvie_: tell us how big it is after repacking compared to the original svn repo	16:01
kelvie_	haha	16:01
	is there a way to tell on the svn web interface?	16:02
	I do _not_ want to make clean my tree	16:02
engla	I don't know how to do that with svn	16:09
	← elmex_ left	16:10
	← anders_ left	16:13
Mikachu	kelvie_: are you importing full history or only the latest revision?	16:15
	← adante left	16:15
	→ adante joined	16:15
kelvie_	full history, and already imported :(	16:15
	it's just repacking now	16:16
	Mikachu: ^	16:16
Mikachu	ah	16:16
	← david__ left	16:19
kelvie_	anyone here use git-svn?	16:24
engla	I used it only to import. so not much	16:24
kelvie_	I just want to know if you can set dcommit to not commit every single change as a new revision	16:24
	but rather one large one	16:24
gitte	Use, yes. Expert, no.	16:24
kelvie_	or should I just use svn commit :P	16:25
	where change = git commit	16:25
MadCoder	kelvie_: use git rebase -i	16:25
	and then dcommit	16:25
engla	kelvie_: I know you can do a squish merge or similar. So you merge your git branch back to the svn tracking branch as one commit. And then dcommit it. But that is what I have heard	16:25
	I wonder if this git alias is sane, for packing up a tarball	16:26
	dist = !git archive --format=tar --prefix="`basename $PWD`-`git describe HEAD`/" `git describe HEAD` \| gzip > "../`basename $PWD`-`git describe HEAD`.tar.gz"	16:26
	← fowlduck left	16:27
	→ robinr joined	16:27
kelvie_	MadCoder: doesn't that retain all the separate git commits?	16:28
	→ strangy joined	16:30
MadCoder	kelvie_: git rebase -i allow you to squash git commits together	16:31
	interactively	16:31
kelvie_	oh	16:31
MadCoder	you can even reorder them, or edit them	16:32
	then you dcommit sth that makes sense and has a nice history	16:32
	← p4tux left	16:32
kelvie_	MadCoder: oh wow, that is nice :)	16:35
	you can even change the commit messages (which was something I had to do)	16:35
robinr	who posted git onto freshmeat? maybe time for an update. 1.3.1 may not be the optimal version for a newcomer.	16:38
Theory	hehe	16:38
robinr	oh, it say there. Nguyen.	16:39
	← bife-the-wise left	16:47
	→ aroben joined	16:48
	→ HackyKid joined	16:50
	→ nud joined	16:51
	→ bife-the-wise joined	16:53
Mikachu	anyone can update packages on freshmeat	16:56
cehteh	robinr: at least it isnt 1.4.x	16:58
	← sgrimm left	17:00
gitte	Mikachu: nope. Not git. Maybe an error by pclouds...	17:05
	← amfwrk left	17:06
Mikachu	Your new release (ID 262484) has been stored for verification by the freshmeat.net crew. Please note that submissions may be edited for brevity, grammar and in some cases content.	17:08
	← gitte left	17:13
	→ praka joined	17:13
	← rtmfd_icbm left	17:14
	→ amfwrk joined	17:20
Ilari	engla: It doesn't store the version number into tar file (which you might want to do, since that is no longer available from VCS).	17:23
	→ drizzd joined	17:24
	→ markoa joined	17:28
markoa	hi folks	17:29
	← Yuuhi left	17:29
markoa	I've created a local branch, done some work	17:29
	now if I push to my repo, from this new branch...	17:30
	will it end up in the repo?	17:30
	← praka left	17:30
markoa	(master untouched)	17:30
kelvie_	well then	17:32
	is it normal to have "failed to unpack tree object" errors while git-gc'ing?	17:32
	← drizzd left	17:32
	→ drizzd joined	17:33
markoa	hmm (regarding my q) it won't, that easily at least...	17:35
	→ alley_cat joined	17:35
workkevin	markoa: Do you mean "will it end up in the working tree"?	17:37
	If so, the answer is "no, and you probably shouldn't push to a repo with a working tree"	17:37
markoa	workkevin, I'm not sure I know what you mean by working tree here. My local working tree? or the repository's current tree?	17:38
workkevin	"Working tree" refers to the checked out source files. In this case, I'm talking about the working tree on the remote repository	17:39
Mikachu	markoa: it depends on what the push line says for the remote in .git/config	17:41
	i think if it says "push = refs/heads/:refs/heads/" you'll push all your branches	17:41
markoa	Mikachu, heh, it doesn't say anything	17:41
	I only have the url and fetch lines	17:42
Mikachu	then it will probably only push the current branch	17:42
	note that i'm not 100% sure about this, i usually fiddle around with it until it works :)	17:42
	← Sho_ left	17:43
	→ Sho_ joined	17:43
markoa	so far I'd always merge work into my local master, then push, which would always go into the remote repo's master	17:43
Mikachu	if you want to push the new branch just this one time you can say "git push remotename branchname:refs/heads/branchname"	17:44
	you need the refs/heads/ prefix only the first time in order to create the branch	17:44
	→ rtmfd_icbm joined	17:45
markoa	and eg if I want to push from that branch (to that branch on remote) in future as well? It seems like I need to configure the remote a bit.	17:45
engla	Ilari: good point	17:46
	→ KirinDave joined	17:49
Mikachu	i wonder if that should be chiisai (small) instead of chisai (district court)? :)	17:50
KirinDave	Mikachu: Both work. ;)	17:53
Mikachu	hehe	17:53
	→ bfields joined	17:56
markoa	so I did 'git push remotename newbranchname'	17:56
	how does someone cloning the repo get that branch ?	17:57
	ok, there is git branch -r to see it...	18:02
	→ Eludias joined	18:07
Ilari	kelvie_: Definitely not normal...	18:08
	← rtmfd_icbm left	18:08
markoa	ok, I did "git fetch origin refs/heads/branch-name:refs/heads/origin/branch-name"	18:09
	hopefully that's up to the convention	18:09
kelvie_	Ilari: yep -- I figured out what it was.. apparently if you use a symlinked subdirectory of a repo	18:09
	it doesn't work too well with git	18:09
	alias cd='cd -P' :P	18:10
	→ rtmfd_icbm joined	18:11
	← puzzles left	18:16
	→ puzzles joined	18:16
	← loops left	18:18
	← pdmef left	18:19
	← yorgen15 left	18:19
	← strangy left	18:21
	→ loops joined	18:21
	← Husio left	18:23
	dwmw2 → dwmw2_gone	18:28
drizzd	is the 'Pull:' line in .git/branches/origin equivalent to the value of remote.origin.fetch?	18:30
	→ yorgen15 joined	18:31
Ilari	drizzd: No. It was earlier version of IIRC "branch.orgin.merge"...	18:34
drizzd	oh I see, but it's still in the git-pull manual	18:34
Mikachu	i think they still work	18:34
drizzd	I like the config version better in any case	18:35
gitster	both are supported. in .git/remotes files there is no concept of branch.<name>.merge; the Pull: lines correspond to remote.<name>.fetch.	18:36
	→ sgrimm joined	18:36
	→ tarrybone joined	18:37
	→ glguy joined	18:37
	← glguy left	18:38
gitster	The config version was introduced later, because people wanted to get extended semantics that was cumbersome to express in .git/remotes version, and by now the former is much more developed. So we teach new people the former, but the latter is still supported and I do not see them to be deprecated anytime soon.	18:38
	→ glguy joined	18:38
drizzd	ok, so when the manual says: add foo to the Pull: line I can equivalently say git config remote.<name>.fetch foo?	18:39
	← loops left	18:39
	→ loops joined	18:42
Mikachu	that section of the manual is probably outdated then	18:45
	← cortilap left	18:47
kergoth	Any idea why a git push to my apache2/webdav git repo isn't working? i can use cadaver to push a file to that webdav path manually fine	18:51
	seems to just hang	18:54
cehteh	is Lars Hjemli here on irc?	18:54
drizzd	you're probably lacking a slash in our url	18:54
	if you don't have that slash git get's a redirect from apache which it can't handle -> infinite loop	18:55
kergoth	hmm, it says "no DAV locking support on remote repo".. would i get that if its the redirect issue?	18:57
	took a while for it to error out though	18:57
	kergoth tries pushing with trailing / :)	18:58
kergoth	aha, thanks :)	18:58
drizzd	:-)	18:58
kergoth	course it still failed, but its probably just a permissions issue now, stupid git-update-server-info	18:58
gitster	how hard would it be to update http-push to understand redirects, I wonder...	18:59
	So that "no DAV locking" was a bogus diagnostic message?	18:59
kergoth	yep, i can push to dav fine with cadaver, so locking works	19:00
	i think..	19:00
	← markoa left	19:00
	kergoth debugs the permissions issue first	19:00
	← Ingmar left	19:00
kelvie_	is there a way to do a git-svn rebase on all branches at once?	19:00
	→ markoa joined	19:01
amfwrk	I just did a git-svn rebase (with my changes commited to a WIP since I had to switch brances) and had a bunch of conflicts, now that the conflicts are fixed and in the index do I need to commit to fix the conflict (I usually commit -a -m 'WIP'; git-svn rebase; git-reset --soft HEAD^	19:01
kelvie_	amfwrk: git-svn rebase --continue	19:01
	I'd think	19:01
gitster	I've seen that error message mentioned here a few times, written them off as server configuration error, but if that is the case at least could somebody put it in the FAQ? Teaching http-push about redirect (or even easier would be to add trailing slash) would also be nice.	19:01
	← rtmfd_icbm left	19:04
	← alley_cat left	19:04
	→ alley_cat joined	19:04
amfwrk	kelvie_: Unknown option: continue; The docs say that it accepts all options that git-rebase accepts too	19:05
kelvie_	yeah that's weird :P	19:05
	→ Ingmar joined	19:05
kelvie_	just use rebase --continue I guess?	19:05
drizzd	actually, the howto at http://www.kernel.org/pub/software/scm/git/docs/howto/setup-git-server-over-http.txt says that 'no DAV locking support' is usually an authentication problem	19:05
kelvie_	actaully don't	19:05
	I wouldn't know the repercussions of that :/	19:05
	so maybe you should wait for someone more knowledgeable to chime in first :P	19:05
kergoth	suggestion: should probably unset GREP_OPTIONS in git scripts. I expect that it'd be bad if the user changed the behavior to something git doesn't expect	19:06
	← metafollic left	19:06
kergoth	also, i doubt you'd want color output if you're grepping for something in git, and the user could be setting --color=always in gnu grep :)	19:06
	← yorgen15 left	19:06
amfwrk	kelvie_: Everything looks fine from git-status, but I'm sure git thinks it is in a conflict/merge state (correct me if I'm wrong)	19:07
kelvie_	amfwrk: that's what I'd assume too	19:08
	but hey, I just started using git for real yesterday :P	19:08
kergoth	ah, thats why i keep running into permissions issues with my http git repositories.. needed to set the umask to 0002 in my post-update's, so everyone in the git group can write to the files (i have some repositories owned by me, some by the git user)	19:09
Ilari	kergoth: You should set the shared flag instead...	19:10
kergoth	oh, right, dangit, i knew i was forgetting something :)	19:11
	thanks	19:11
	→ meandtheshell joined	19:13
	→ metafollic joined	19:13
	elmarco → elmarco\|awayyy	19:16
	up_the_irons → irons\|lunch	19:28
engla	kelvie_: did the repack finish?	19:29
kelvie_	engla: 93% :D	19:29
	it's going really fast now	19:30
engla	:)	19:30
	← rodserling left	19:31
	← stouset left	19:33
	← metafollic left	19:33
	→ stouset joined	19:34
	→ metafollic joined	19:36
meandtheshell	hi folks, is this statement http://rafb.net/p/s3GNO938.nln.html correct?	19:36
Mikachu	not entirely	19:37
	← moh left	19:37
meandtheshell	how so?	19:37
kelvie_	engla: or not .. now it's writing objects.. it's just done deltifying them :O	19:37
meandtheshell	Mikachu: how would you rephrase it?	19:37
Randal	"other than the current one" is clearly wrong	19:38
	I can have 30 local branches, and zero or more remote branches	19:39
	→ H4ckyK1d joined	19:39
	← H4ckyK1d left	19:39
Randal	Oh. "other than the currrent one" doesn't apply to branch	19:39
	well, that's ambiguous then. :)	19:39
Mikachu	and the remote branches are all in the repository as well	19:39
Randal	a remote branch indeed	19:39
meandtheshell	Randal: right - it refers to repos	19:39
Randal	yeah, was going there.	19:39
	a "remote branch" is a branch that is tracking a remote repo, and is generally updated by "fetching" that repo.	19:40
	→ dduncan joined	19:41
Randal	as opposed to local branches that are updated either by local commits or local merges or merges with commits from remote branches	19:41
meandtheshell	sure but a branch is considered remote it it resides in a repository other than the current one - yes?	19:42
Randal	"resides" is the crazy word there	19:42
meandtheshell	it it*	19:42
Randal	there's no "one" home for a branch	19:42
	a branch can be everywhere	19:42
	this is what distributed systems are	19:42
kelvie_	wow I really should have niced the git-gc :/	19:42
Randal	my local branch is your remote branch	19:42
	my remote branch is your local branch, or maybe another of your remote branches	19:42
	but the branch "resides" on both repos	19:43
	so that's a bad word	19:43
	there is no difference between commit DEADBABE on my machine and commit DEADBABE on yours	19:43
	they both comtain identical information	19:43
Tene	kelvie_: renice	19:45
Mikachu	kelvie_: renice 20 `pidof git-gc` ?	19:45
	and possibly child-processes	19:45
	`pgrep -t pts/n` where n is the number of the terminal	19:45
kelvie_	doh	19:46
	right I forgot I could do that :P	19:46
meandtheshell	Randal: ok, so that was ambiguous - however, from the repository point of view I can say "A branch is considered a remote branch if it gets fetched from a repository other than the current one" ... - yes? Do we agree on that?	19:46
Randal	that's closer	19:47
Mikachu	a branch is a remote branch if it's in the remotes/ namespace instead of refs/	19:47
Randal	there may still be exceptions	19:47
meandtheshell	:)	19:47
Randal	yeah Mikachu has it	19:47
Mikachu	that doesn't tell you a lot about what it is though :)	19:47
Randal	but I think meandtheshell wants a more useful view, not an implementaion view	19:47
meandtheshell	right - something a novice can understand better ... however, Mikachu statement is not bad	19:48
Mikachu	'these branch tips are automatically updated when git fetch is run, and shouldn't be modified locally' or something?	19:48
	← CIA-11 left	19:48
meandtheshell	hm ... that's with regards to the workflow ... not what I need	19:48
Mikachu	doesn't the tutorial already say something clever about them?	19:48
meandtheshell	nope	19:48
	that's why - the wiki is also quiet about "what is a remote branch" - sure, how to create and deal with them is there	19:49
	also why they are useful etc.	19:50
Mikachu	it's the same as a local branch in most other regards	19:50
	ie just a reference to a commit sha1	19:50
	→ moh joined	19:50
meandtheshell	yep, we know that but would you put that "as is" on a nice howto :)	19:51
Mikachu	i'm not writing a howto :P	19:51
meandtheshell	hehe	19:51
	← stouset left	19:52
meandtheshell	we should find two or three appropriate sentences to put them on the wiki ...	19:52
	→ CIA-11 joined	19:55
	← HackyKid left	19:57
	← tcoppi left	20:00
Randal	and maybe a few inappropriate ones, just for fun. Hey, it's a wiki! What's a wiki without wiki spam and revert wars!	20:02
	← bentob0x left	20:02
LotR	Randal: a useful resource? ;-p	20:03
meandtheshell	Randal: don't do that please	20:04
	→ stouset joined	20:06
kergoth	man, i hate how easy it is to end up with a stale webdav lock if you interrupt a push over it :\	20:12
Ilari	kergoth: Webdav sucks, doesn' it?	20:14
	→ HackyKid joined	20:20
kumbayo	is there a easy way to delete a packed ref beside git-update-ref -d <ref> <oldvalue> ?	20:22
Mikachu	is that hard?	20:22
kumbayo	yes i have to search refname and old hash value :-)	20:23
Mikachu	can't you just say ref ref?	20:23
Ilari	kumbayo: See also 'git branch -d' and 'git branch -D'.	20:23
MadCoder	and git tag -d	20:23
Mikachu	that is true, that is a lot easier	20:23
MadCoder	kumbayo: else git update-ref -d ref $(git parse-rev ref) works great you know	20:24
	rev-parse actually	20:24
kumbayo	hmm it is in refs/patches/obd_development/thread_safety_8 so it is not a branch and not a tag	20:25
MadCoder	kumbayo: see what I just typed	20:25
kumbayo	it is a remanent of a deleted stg patch	20:25
MadCoder	git update-ref refs/patches/obd_development/thread_safety_8 $(git rev-parse refs/patches/obd_development/thread_safety_8)	20:25
kumbayo	i understood :-)	20:26
	already	20:26
MadCoder	:P	20:26
kelvie_	what's the most ideal way to commit to an svn tree, to commit your working branch and git-svn rebase, or to merge into your master branch and commit?	20:26
MadCoder	git svn rebase is better	20:26
	I mean no it sucks hard	20:26
	but you have to talk to svn in the end	20:27
	so you'd better behave more like svn than like git sadly	20:27
	irons\|lunch → up_the_irons	20:31
thiago_home	you don't need rev-parse	20:32
	this works: git update-ref refs/patches/obd_development/thread_safety_8 refs/patches/obd_development/thread_safety_8	20:32
kumbayo	hmm nice, i used magic shell middle button paste + ^W^W :-)	20:35
glguy	refs/patches? is there a list somewhere of all the things that are refs?	20:35
Mikachu	anything in refs/ :)	20:38
	thiago_home: that doesn't delete the ref though	20:38
thiago_home	add -d	20:40
Mikachu	that most likely doesn't accept an -ish	20:41
thiago_home	git update-ref -d refs/patches/obd_development/thread_safety_8 refs/patches/obd_development/thread_safety_8	20:41
	sure it does	20:41
Mikachu	then what is the point? :)	20:41
kumbayo	i tried here	20:41
	jwb → jwb_gone	20:42
	→ H4ckyK1d joined	20:42
	← HackyKid left	20:42
	H4ckyK1d → HackyKid	20:42
shoe	I found it surprising that "git-init --bare --shared" doesn't work, while "git --bare init --shared" does.	20:44
Mikachu	git-init --bare doesn't work either, it seems	20:45
shoe	really? git-clone --bare work, I think.	20:45
Mikachu	yes	20:45
shoe	that's definitely strange, then.	20:46
Mikachu	maybe you're the first to want to init a bare repo	20:46
shoe	well, I really did the init for the --shared, so it's not that odd.	20:46
	→ rtmfd_icbm joined	20:46
Randal	yeah, there is a gap there	20:47
	you can git-init, and then config it to be bare.	20:47
Mikachu	(--shared to git-init isn't the same as -s for git-clone)	20:47
	→ mmp joined	20:47
	← rtmfd_icbm left	20:48
	→ rtmfd_icbm joined	20:48
mmp	hello -- any idea how to manually deinitialize (or forcibly reinitialize) branch on which 'stg init' was used?	20:49
	← csc` left	20:50
	→ rodserling joined	20:51
	→ troyt joined	20:53
	← polyonymous left	20:54
	→ kristoffer joined	20:54
kristoffer	Any git guru got a sec?	20:54
	I've downloaded a repository which consists of 3 branches. I can see them doing 'git branch -r'	20:54
	but Im lost howto move to another branch, since in my heads/ I only got the main branch	20:55
	Im guessing the heads wasn't downloaded properly or something, or am I doing something silly?	20:55
Ilari	kristoffer: 'git checkout -b foo origin/foo'...	20:55
kristoffer	Ilari, oki, but doesn't that create the local repository foo?	20:56
Randal	not repo	20:56
	branch	20:56
Ilari	That creates branch foo, starting from orgin/foo and then checks it out.	20:56
Randal	just like you ask for	20:56
	→ polyonymous joined	20:56
kristoffer	aaah, oki. Big thanx	20:56
	← troyt left	21:00
	← lcapitulino left	21:01
	← G_SabinoMullane left	21:04
	→ nud_ joined	21:04
	← bife-the-wise left	21:07
kelvie_	hrm the git repo ended up larger than the svn repo	21:09
MadCoder	because you didn't repacked	21:10
thiago_home	repack now	21:10
MadCoder	git gc --aggressive	21:10
thiago_home	did you import with git-svn?	21:10
kelvie_	yep	21:10
	← devogon left	21:10
MadCoder	kelvie_: git gc --aggressive	21:11
thiago_home	how many SVN revisions are there in your repo?	21:11
kelvie_	39k at the moment	21:11
thiago_home	or, more to the point, what is the highest SVN revision number you imported?	21:11
kelvie_	39 thousand something :P	21:11
MadCoder	or git repack -a -d -l -f --window=50	21:11
	it will be slow	21:11
	but it's definitely a thing to be done after a big import	21:11
thiago_home	ok, you have then 1.6 MB of git-svn database	21:12
MadCoder	there is that too indeed	21:12
	git-svn is a pig wrt that	21:12
kelvie_	my svn (old) repo is 7.7 gigs	21:12
	and that's with object files	21:12
	→ devogon joined	21:12
kelvie_	right now, my "clean" svn import after a git-gc is 11 gigs	21:12
thiago_home	have you repacked now?	21:12
kelvie_	yes	21:12
MadCoder	with the options I proposed ?	21:12
thiago_home	how big is the pack file?	21:12
kelvie_	er h rm	21:12
	packed-refs?	21:13
MadCoder	no packed-refs will gain almost nothing	21:13
	git repack -a -d -l -f --window=50	21:14
kelvie_	yeah I'm doing that	21:14
MadCoder	this will be slow, but should achieve a way better compression ratio	21:14
kelvie_	looks like it'll be another many hours	21:14
MadCoder	it will hog your machine	21:14
	probably	21:14
kelvie_	well either way it's 11 gigs right now	21:14
	→ matteo joined	21:14
matteo	hi all	21:14
MadCoder	and like thiago_home asked, what does du .git/objects says ?	21:14
kelvie_	and my svn repo from before was 7.7 gigs.. with object files :O	21:14
matteo	quick question, how to revert my kernel to the 2.6.22 tag?	21:14
MadCoder	git-svn is a pig and has many many things in .git/svn or sth	21:14
kelvie_	objects are 8.8 gigs	21:14
MadCoder	okay, I wouldn't be surprised that after the repack it'll be better	21:15
	though it sounds like you should split the repo in many smaller	21:15
kelvie_	.git/svn is only 14 megs :P	21:15
	yeah can you do that? :/	21:15
MadCoder	or is there a lot of big big binary files in it ?	21:15
kelvie_	very possible	21:15
MadCoder	of course you can	21:15
kelvie_	but wouldn't explain why the svn is still smaller	21:15
	→ winter-m1te joined	21:15
MadCoder	though it's not very easy	21:15
kelvie_	even after "make"	21:16
MadCoder	heh	21:16
	otoh you have a full history since rev 0	21:16
kelvie_	or well we use jam	21:16
MadCoder	it has to be a bit bigger	21:16
kelvie_	oh svn doesn't store all the revisions?	21:16
	← winter-mute left	21:16
kelvie_	by default	21:16
	← winter-m1te left	21:16
MadCoder	not in the checkout	21:16
	→ winter-mute joined	21:16
kelvie_	ooh	21:16
MadCoder	in your .git there is _all_ the revisions	21:16
	← meandtheshell left	21:17
kelvie_	how many does it checkout by default?	21:17
MadCoder	one	21:17
kelvie_	oh	21:17
	really o.O	21:17
MadCoder	I mean, svn when you svn co has only informations about the last revision	21:17
	all the rest needs network access to the server	21:17
kelvie_	aah	21:17
MadCoder	git works 100% offline	21:17
kelvie_	well either way I had to make a git import for my coworkers	21:17
	poor guys are using cygwin :P	21:17
	4 days importing and counting...	21:18
MadCoder	heh	21:18
	anyways I'm completely sick, hence heading to bed right now	21:18
	cu	21:18
kelvie_	Thanks	21:18
matteo	how can I revert to a tag?	21:18
Ilari	matteo: If you don't have any uncommitted changes, 'git reset --hard v2.6.22'.	21:19
matteo	ok tnx Ilari	21:19
	← kumbayo left	21:20
shoe	Is there a convenient way to clone a repo while preserving the git-svn bits?	21:21
	← matteo left	21:21
kelvie_	how do you only clone one branch?	21:23
	I imported an svn tree, and want to clone and share it, but I'd like to take out my private branches :P	21:24
vmiklos	git fetch url branch	21:24
	that won't fetch object with are not reachable from that branch i think	21:24
	→ anonuser joined	21:25
kelvie_	mm	21:26
	← nud left	21:26
anonuser	Is there ever going to be a libgit or something like that? Or is current methodology just to delegate to git?	21:28
vmiklos	there is a libgit actually	21:30
	see also the related gsoc project:)	21:31
	← glguy left	21:34
	← jcollie left	21:35
anonuser	vmiklos, oh really.	21:37
	I thought it was more of an internal thing.	21:38
	It does't actually produce a shared object?	21:38
	← stick left	21:39
vmiklos	the one in git.git is a static lib so yes, it's mostly internal	21:43
	i don't think they really care about api changes:)	21:43
anonuser	vmiklos, well i'm looking at the libgit stuff and I have no idea what was accomplished :-\	21:43
vmiklos	i have no idea about what the gsoc changed	21:43
anonuser	ahh i found it	21:44
vmiklos	http://repo.or.cz/w/git/libgit-gsoc.git	21:45
anonuser	vmiklos, http://permalink.gmane.org/gmane.comp.version-control.git/53433	21:45
	that's what I was looking for	21:45
vmiklos	yes, that's the same:)	21:45
anonuser	Okay neat	21:45
vmiklos	but be aware that this is not yet merged afaik	21:45
anonuser	It's not merged as far as I can tell.	21:46
	← Roomster left	21:46
	→ matteo joined	21:47
anonuser	gitweb isn't allowing me to search so I can't really search :-\	21:47
matteo	hi	21:47
corecode	dammit	21:47
anonuser	If it doesn't change I guess I can work on the language bindings with the thin lib	21:48
corecode	how do i pull one file from a different revision?	21:48
	git reset doesn't work	21:48
gitster	you do not pull. Do you want to see the file from a different commit?	21:48
	Is the commit in the same repository?	21:48
	does "git show that_commit:path/to/the/file" ring bell?	21:48
	is "git checkout that_commit path/to/the/file" what you want?	21:49
shoe	I guess not too many people try to clone a git-svn managed repo.	21:50
	← devogon left	21:50
gitster	I had an impression that git-svn author was against that whole notion.	21:51
	→ devogon joined	21:51
shoe	But it seems like that might be a gentle way to transition from git-svn + svn to just git.	21:51
gitster	... but I may be mistaken.	21:51
	Yes using git-svn seems like so. Cloning such a repository is a separate issue.	21:52
	Actually, cloning and expect that cloned one to interoperate with the same SVN repository _directly_ is a separate issue.	21:52
	I do not know if normalperson is around...	21:53
shoe	Well, I'd be willing to tell people "hey, add this [git-svn] block to your config"	21:53
	But I'd rather they not have to wait for git-svn to rebuild the revdb.	21:54
gitster	That's what I meant. In that workflow, everybody interoperates with svn individually.	21:54
shoe	I guess I could just use 'tar' instead of git-clone.	21:55
gitster	I think alternatively you could designate one git repository to interoperate with git-svn.	21:55
	← nud_ left	21:55
gitster	That repository would need to have local branches that correspond to git-svn managed "remote" tracking branch, though. Then your git-only users would clone/pull/develop/rebase/push into it as if it is just a regular git repository.	21:56
shoe	hmm, that might work... like a gateway... maybe with a post-commit hook with git-svn dcommit?	21:56
gitster	The designated git-svn repository that interact with SVN would need to handle dcommit though.	21:56
	Ask git-svn people; unfortunately I am not one of them.	21:57
shoe	oh... but I'd want to preserve author, of course.	21:57
gitster	Isn't author preserved by dcommit?	21:57
shoe	Not really.	21:57
	→ Romster joined	21:57
gitster	dcommit propagates committer identity to SVN?	21:58
shoe	The dcommitted commits are thrown away, and the svn-commits determine author.	21:58
gitster	Ah, but that is because git-svn's mental model is still "SVN is the king".	21:58
	IOW, as far as the project as the whole is concerned, what's committed to SVN is the official history.	21:59
shoe	well, I think it's more a function of the fact that git-svn can't control who svn thinks the author is.	21:59
	only the svn-server does the user-mapping.	22:00
gitster	That is something you cannot fix with git-svn without really migrating out and makign the git side the official repository, I am afraid.	22:00
	But I am not sure if git-svn can still be used after such a migration happens to a project.	22:01
shoe	there'd be no point.	22:01
gitster	Because the whole notion of "git rebase and then commit to svn" or "dcommit" comes from the basic attitude of treating the SVN side the official history.	22:01
	Why there is no point? After you manage to migrate some or even most people it would be nice to keep the legacy SCM around during the transition period.	22:02
shoe	perhaps.	22:04
	Anyway, I guess I'll just pass around a tarball of the git-svn repo, and let people cut their teeth there.	22:05
	← jtoy left	22:06
	← orospakr left	22:06
	← alley_cat left	22:07
	→ yorgen15 joined	22:07
	→ doublec joined	22:08
	← chris2 left	22:12
	← tarrybone left	22:14
	→ tarrybone joined	22:14
	← bfields left	22:15
Randal	what's the simplest way to have a laptop and an online version of my personal work repo	22:18
	treat it as if I was two people pushing and pulling?	22:18
gitster	I would say so but I probably would do pushing and fetching+rebasing for my own stuff.	22:25
	← bdiego left	22:25
gitster	... and using the "mothership-satellite" aka "push into tracking branch" layout.	22:26
	← elmex left	22:28
	← markoa left	22:31
	← yorgen15 left	22:31
Randal	I think I can do it with a bare repo on the server, and a working dir on the laptop	22:37
	then it's just a matter of the right git-remote. :)	22:37
mugwump	that's the use case that made me throw svk away	22:37
Randal	What is "Can't locate Git.pm in @INC" when invoking git-remote?	22:37
mugwump	Randal, surely... :)	22:38
	← anonuser left	22:38
Randal	I know what causes it. are there things that add to PERL5LIB?	22:38
	or how else does it find a weird location for Git.pm?	22:38
	ahh - second line of git-remote	22:39
	the path there must've come from configuration	22:40
	see - leave me alone and I come up with my own worldview. :)	22:40
	← krh left	22:42
HackyKid	finally! i did it! the pern finally compiles in msys!	22:42
	s/pern/perl	22:43
mugwump	HackyKid, did Strawberry Perl not work for you?	22:43
	→ gitte joined	22:43
HackyKid	hey gitte, you just missed it :-p	22:44
gitte	What?	22:44
HackyKid	<HackyKid> finally! i did it! the pern finally compiles in msys!	22:44
gitte	perl?	22:44
gitster	you said pern again ;-)	22:44
HackyKid	yeah	22:44
	heh	22:44
gitte	Okay!	22:44
HackyKid	lazy copy/paste :-)	22:44
Randal	on my git-remote, I have use lib (split(/:/, $ENV{GITPERLLIB} \|\| "/opt/git/lib/perl5/site_perl"));	22:44
gitte	Now, not to spoil your party, but does it do MSys paths?	22:44
Randal	but on brian's, he says no path there, just the gitperllib	22:44
	did that change recently?	22:44
HackyKid	mugwump: well. i needed a perl that understood the msys unix-style paths	22:45
	gitte: hmm, yeah, still need to test that :-)	22:45
gitster	Randal: "how do I find out what got changed in a directory recently?"	22:45
Randal	heh	22:45
	ok	22:45
gitte	git log directory/?	22:45
	gitte is watching the Mondrian talk.	22:45
Randal	git-blame git-remote.perl :)	22:45
gitster	"perl/" directory Aug 20 is the last commit I see.	22:46
Randal	oh wait - git-remote.perl doesn't have that line at all	22:46
	must be part of some build process	22:46
gitte	HackyKid: was the recipe not sufficient?	22:46
HackyKid	gitte: it does link against the msys dll :-)	22:46
gitte	HackyKid: oh, oh!	22:47
gitster	Makefile has a funky rule.	22:47
cmarcelo	gitster: what is that "mothership-satellite" layout you talked about?	22:47
Randal	Funky!	22:47
HackyKid	wasnt the recipe for a dmake based build?	22:47
gitster	We had a patch posted to that part this morning.	22:47
	cmarcelo: http://git.or.cz/gitwiki/GitFaq#head-46de97a5ac549d3adf406c22674b3325ae25d09c	22:48
gitte	HackyKid: try 'perl -e "open IN, "/bin/awk"; while (<>) { print $_; }', please	22:48
	HackyKid: yes, it was.	22:48
cmarcelo	tks	22:48
Randal	eww. /bin/awk	22:49
	you're reading a binary?	22:49
Tene	gitte: check your quotes.	22:49
	← branstrom left	22:49
gitte	Tene: very funny.	22:49
HackyKid	gitte: hmm, misquoting?	22:49
	→ branstrom joined	22:49
gitte	HackyKid: you know whaddimean.	22:50
HackyKid	i do? i dont actually know perl, you know :-)	22:50
	→ yorgen15 joined	22:50
gitte	HackyKid: okay: open an editor, please.	22:50
	→ tcoppi joined	22:51
Randal	"perl -perl /bin/awk" would have worked as well :)	22:51
	lot less typing	22:51
	"perl -perl" is the new "/bin/cat" :)	22:51
gitte	HackyKid: and then paste http://paste.debian.net/38033	22:51
	Randal: won't work.	22:52
Randal	in what sense of "won't"?	22:52
gitte	MSys' bash will mangle the command line before you can say "oh?"	22:52
Randal	mangle how?	22:52
gitte	So perl will see C:\bla\blub\msysGit\bin\awk	22:52
	Not /bin/awk.	22:52
Randal	oh weird.	22:52
gitte	Yes, it's tricky.	22:52
Randal	then quote it	22:52
	perl -perl "/bin/awk"	22:53
gitte	Nope.	22:53
	AFAIK won't work either.	22:53
	The safest way _is_ to write a script.	22:53
Randal	perl -perl "</bin/awk" then	22:53
	it is a script	22:53
	just a short one	22:53
	perl -perl "date\|" - the hard way to get a date	22:53
HackyKid	gitte: it prints /bin/awk	22:53
gitte	Oh, I seem to have my nobody-understands-what-I-mean day!	22:53
	HackyKid: very well done!	22:53
	So what is the correct recipe?	22:53
Randal	I've made a living from those	22:53
	← ferdy left	22:54
gitte	Randal: of those days?	22:54
HackyKid	heh, work around a lot of problems :-p	22:54
gitte	Randal: you'll have to show me how.	22:54
HackyKid	but i'll write it all up tomorrow or so	22:54
gitte	HackyKid: I hope you wrote all down.	22:54
HackyKid	yeah most of it	22:54
Randal	indeed	22:54
gitte	HackyKid: just for my curiosity: what are the essential tricks?	22:55
	gitster wonders if "git -git" also meows like a cat...	22:55
HackyKid	wrestling the configure system into submission :O	22:55
	for some reason it sometimes deletes the makefiles after it is done generating them	22:56
gitte	But then you got it compiled _and_ linked to msys-1.0.dll?	22:56
HackyKid	yeah	22:56
gitte	Ah, I know why.	22:56
	Case insensitivity.	22:56
	It removes "makefile".	22:56
	Not "Makefile".	22:56
HackyKid	yeah, i suspected as much	22:56
gitte	Alas, Windows is dumb as a chair.	22:56
HackyKid	:O	22:56
gitte	Chair like in "flying" ;-)	22:56
	← wycats left	22:56
HackyKid	also there was an error in one of the msys headers :O	22:56
	← GyrosGeier left	22:57
gitte	Yeah, I think I edited one too.	22:57
HackyKid	oh, and i tries perl 5.6.1 this time instead of the newest version	22:57
gitte	Just before giving up on configure.	22:57
HackyKid	s/tries/tried	22:57
gitster	randal: what does command line switch -r do?	22:58
gitte	HackyKid: (But I used "git init" on the perl sources just before playing with it, so I'd not have to write it all down ;-))	22:58
Randal	I don't know	22:58
HackyKid	i did too this time :-)	22:58
Randal	that's not what that's doing	22:58
	perl -perl is actually perl -p -e 'rl'	22:58
HackyKid	so my config settings are at least still in there	22:58
Randal	where 'rl;' is the "program" I'm running	22:58
	→ bobesponja joined	22:59
Randal	so perl -peatmoss would also work	22:59
gitster	what does rl; do?	22:59
Randal	it's just the bareword "rl"	22:59
gitte	HackyKid: as long as you did not edit intermediate files ;-)	22:59
bobesponja	what's the best way to create a patch file from a commit so I can send it by mail?	22:59
gitster	ok.	22:59
Randal	perl -p -e '"rl";'	22:59
	that'd be the equivalent fully dressed	22:59
gitte	HackyKid: since I am so curious, could you send me the diff?	22:59
gitster	"perl -pwerl" wouldn't have worked, in other words?	23:00
HackyKid	heh, and i had to rename INSTALL, otherwise 'make install' would not work :-)	23:01
gitte	yep, I remember deleting it.	23:02
Randal	that would, but with an error, probably	23:03
	I never use warnings, dunno	23:03
bobesponja	anyone about best practice for patch creation?	23:03
HackyKid	gitte: hmm, only modification to the perl sources needed was uncommenting a line in pp_sys.c ("extern int h_errno;")	23:03
	eh	23:03
gitster	Sure, write a perfect code ;-)	23:04
HackyKid	s/uncommenting/commenting	23:04
	commenting out, even :O	23:04
gitte	HackyKid: makes sense...	23:04
	bobesponja: Look at Documentation/git-format-patch.txt	23:05
	→ robfitz_ joined	23:06
	← stouset left	23:06
jepler	HackyKid: if you're using GNU make, see: info make "Phony Targets"	23:06
	HackyKid: '.PHONY: install' should make the install target work even when a file named 'install' exists	23:06
	← Eludias left	23:07
	← mmp left	23:07
bobesponja	gitte: thanx, I'm trying but it doesn't seem to create any file or at least I can't find it, any idea?	23:10
	→ gitte_ joined	23:12
	← gitte left	23:12
jepler	bobesponja: how did you run it -- did you give a revision? On my tree, this prints a diff on stdout: git-format-patch --stdout HEAD^	23:12
	gitte_ → gitte	23:12
jepler	without --stdout, it would go in a file the name of which is printed on stdout	23:12
	e.g., 0001-work-properly-for-negative-scales.txt	23:13
gitte	bobesponja: I guess you did not provide _any_ arguments, right?	23:13
	You have to give git a chance to know _what_ patches you want...	23:13
bobesponja	gitte: I did: "git-format-patch 2d9d04eec20d -o some_path/"	23:13
jepler	I'm not sure why git-format-patch doesn't print an error when there are no revisions given, since it's never useful without at least one revision (right?)	23:13
gitte	bobesponja: where 2d9d04 is HEAD?	23:14
bobesponja	gitte: yes that's HEAD	23:14
jepler	% git-format-patch HEAD^ -o some_path/	23:14
	Not a valid rev -o^ (-o)	23:14
gitte	For historical reasons, giving _one_ ref means that you want to print all patches _since_ that ref.	23:14
	In your case... none.	23:14
jepler	on my git this isn't accepted, but maybe that's changed since ancient times?	23:14
	← Ryback_ left	23:14
gitte	Oh yeah, and I'd _definitely_ write the options before the parameters.	23:15
	(Like in most Unix programs)	23:15
bobesponja	ok thanks a lot to both	23:15
gitte	yw	23:16
	Mikachu has a keybind in zsh that moves options before parameters for you :)	23:18
	← robfitz left	23:18
gitte	Mikachu: so what do people have with this newfangled zsh? Is it so fast or what?	23:18
Mikachu	i've actually never used bash in the first place	23:19
gitte	So what system installs it by default?	23:19
Mikachu	i don't know if any do	23:19
	→ dash__ joined	23:20
gitte	Then you had to have a reason to install it yourself.	23:20
Mikachu	yeah, the name sounded cool	23:20
	(this was a few years ago :)	23:20
gitte	Ah, a very sound reason then ;-)	23:21
Mikachu	i'd say it's about as good as using a shell because it's the default	23:21
gitte	Umm. It's the default?	23:22
Mikachu	no i meant my reason is as good as yours is for using bash	23:22
	→ jtoy joined	23:23
gitte	I actually played a little with busybox' builtin shell.	23:23
	And with tcsh, since Mac OS X 10.2.8. had _that_ as a default shell.	23:23
	As well as the Alpha boxen I was forced to use a few years ago.	23:23
Mikachu	zsh's globbing is so good i hardly even know how to use 'find' :)	23:24
	and there are lots of crazy shorthand syntax for common operations	23:25
	← cmarcelo left	23:25
Mikachu	like if foo=dir/name.ext, then $foo:h is dir, $foo:r is dir/name, $foo:t is name.ext, and so on :)	23:26
	which is handy in for loops when converting png to jpg or whatever, you can say $i:r.jpg instead of some crazy ${i%.png}.jpg	23:27
gitte	Oh? perl gone shell?	23:27
Mikachu	i usually joke that zsh is the only language more unreadable than perl	23:27
gitte	You don't know brainfuck.	23:27
	Or Malbolge.	23:28
Mikachu	apart from those :)	23:28
gitte	Whereas a colleague of mine said Cobol would beat Malbolge hands down.	23:28
Mikachu	*(@e:'[[ ! -e `zstat +link $REPLY` ]]':) expands to all broken symlinks	23:31
gitte	Now, that is a good reason to switch ;-)	23:31
robinr	COBOL is brainfck, except you don't know you're being fcked	23:31
	← dash_ left	23:31
gitte	BTW am I the only one who thinks that the Mondrian talk was a little cough dry?	23:31
	→ jtoy_ joined	23:32
	← branstrom left	23:32
robinr	link?	23:32
Mikachu	gitte: show me how to do it in bash then :)	23:33
gitte	robinr: http://video.google.com/videoplay?docid=-8502904076440714866	23:33
	Mikachu: I'll show you next time I need it, okay?	23:33
Mikachu	sure	23:33
gitte	But don't hold your breath for it ;-)	23:33
	Mikachu creates a function so (e:brokensym:) also works*	23:34
gitte	(I never, ever had the need to find all broken symlinks)	23:34
Mikachu	some usually accumulate in /usr/bin, and it's boring to copy paste red text from ls output	23:34
robinr	gitte: nowhere near as boring a Marc Andreesen at oopsla '95 (or somewhere near that)	23:36
	→ branstrom joined	23:37
gitte	robinr: somebody should have told Guido _before_ minute 30 or so that it helps to drink, and then more than a drop.	23:37
	Mikachu: find -maxdepth 1 -type l \| while read line; do test -e "$(readlink "$line")" \|\| echo "$line"; done	23:37
	Mikachu: Scrap the "-maxdepth 1" if you do not want to test, as I did.	23:37
	Mikachu: I find that utterly more readable, thankyouverymuch	23:37
Mikachu	sure :)	23:38
	mine forks less	23:38
Randal	find2perl . -eval '-l and not -e' -ls \| perl	23:38
	that forks even less. :)	23:38
	as in, not at all	23:39
	two processes total	23:39
Mikachu	that is two forks then, mine does zero	23:39
gitte	Hah. Perl is even _less_ obfuscated!	23:39
Randal	Yeah, but you have to use bash.	23:39
	I won't.	23:39
Mikachu	gitte: i think i agree with your assessment of that video	23:40
	heh, the guy in the "Dtrace Review" video reminds me of daniel in sg-1	23:42
	← robinr left	23:45
matteo	why a git tree has many heads?	23:45
	how to select one?	23:45
Mikachu	git-checkout	23:45
gitte	Those are branches, IIUYC.	23:45
matteo	also, can I get infos about such branches?	23:45
gitte	What kind of info?	23:45
matteo	a comment or so	23:46
	http://git.kernel.org/?p=linux/kernel/git/linville/wireless-2.6.git;a=summary	23:46
	→ robinr joined	23:46
matteo	there are almost 10 branches	23:46
gitte	Sure, I can give you a comment about ath5k.	23:47
	This is the branch for the Atheros 5k.	23:47
	Or z1211: the branch developing the driver for z1211.	23:47
	Should I go on? ;-)	23:47
	→ adante_ joined	23:55
	← devogon left	23:58

Logs Search ←Prev date Next date→

Channels Documentation