IRCloggy #git 2007-11-25

Provider of IRC logs since 2005.

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

	2007-11-25
anttih	Mikachu: that is wrong?	00:00
Ilari	robinr: And there are lots of crap books about C++ that get things wrong...	00:00
Randal	it's what I told you earlier, I think	00:01
robinr	dsop: my guess is master is not initialied unless you import HEAD	00:01
Randal	never push to a non-bare repo	00:01
robinr	Ilari: plenty	00:01
Randal	oh - it's what I was supposing but someone else said your problem had already been answered	00:01
	but now I see that it was what I preduicted before. :)	00:01
anttih	Randal: I have many problems :-D	00:01
	this is a new issue	00:01
	I suck	00:01
	← aroben_ left	00:02
Randal	no, you just don't blow enough. :)	00:02
robinr	I've been trying to tell C++ educators not to start with C. ... you're not supposed to do "C" programming in C++ unless you are and STL implementor	00:02
anttih	but you say never push to a non-bare repo	00:02
Ilari	Or if you really need to, force the push go to tracking branches via configuration (perferably to tracking branches of that repository).	00:02
Randal	don't confuse anttih... :)	00:02
robinr	i've even seen supposedly "good" c++ programmers do new char...	00:03
anttih	you bet	00:03
Randal	C++ is evil. OO should be taught with Smalltalk	00:03
robinr	or LISP :)	00:03
anttih	ok, I think I understood something	00:03
Randal	no - it should be a pure OO language	00:03
	to keep complications out of there.	00:04
semi_	pure OO is hard to define	00:04
Randal	which rules out Java, C++, Python, and Perl.	00:04
	Not at all	00:04
anttih	have you tried D?	00:04
Randal	"everything is an object"	00:04
	"all classes can be subclassed"	00:04
	"all classes can be extended"	00:04
	that's pure	00:04
anttih	that's JS	00:04
Randal	No. JS has primitive data types	00:04
vmiklos	heh	00:05
	this is true for python, too i think ;)	00:05
Randal	No - python has primitive data types	00:05
semi_	OO is overestimated	00:05
Randal	and so does java	00:05
	Smalltalk has no "primitive" types	00:05
robinr	Object?	00:06
vmiklos	if you say p = 1 then you have an int object	00:06
	that's not a primitive type	00:06
robinr	atoms	00:06
vmiklos	(like the java one)	00:06
robinr	can you subclass atoms?	00:06
Ilari	I say the language should also feature 'strong component isolation' as standard part. Those also tend also be 'cleaner' OO languages (excluding pathologies like capfuck)...	00:06
tpope	javascript doesn't have primitive datatypes does it?	00:06
Randal	what are you talking about?	00:06
tpope	String, Number, Function, these can all be extended	00:07
anttih	has someone tried D?	00:07
vmiklos	s = -1; then i can say s.__abs__()	00:07
iabervon	I thought python primitive data types weren't distinguishably not objects as far as you can tell within the language.	00:07
Ilari	One example of such language is E. I don't know how pure OO it is...	00:07
Randal	you can't subclass python primitive data types	00:07
	if I recall correctly	00:08
Mikachu	you can even say (-1).__abs__()	00:08
robinr	iabervon: maybe they use what java calls autoboxing?	00:08
	i.e. the compiler wraps primitives with things like new Integer() if an Object is required	00:08
	but that is not what Smalltalk does	00:09
	I think smalltalk internally tries top optimize using primitives if possible	00:09
iabervon	I've never tried subclassing them, but I'd be a bit surprised if you can't, given how much you can muck with everything in python.	00:09
Randal	robinr - not the smalltalks I've seen	00:10
	→ tongueroo joined	00:10
Mikachu	so if you say 1+1 it runs some hundred operations? :)	00:10
robinr	Randal: no wonder it's slow	00:11
Randal	it does a simple dispatch to the SmallInteger class	00:11
	looking for a + method, yes	00:11
	"slow"? when's the last time you played with smalltalk?	00:11
robinr	....a while ago	00:11
	when 200 MHz was a fast processor	00:11
Randal	the web framework Seaside beats Ruby On Rails by a factor of 10-to-1	00:12
robinr	and, Oh, I run a couple of swikis	00:12
	now	00:12
Mikachu	afaiu, ruby isn't really hard to beat in speed	00:12
Randal	java's relatively slow too	00:12
	some modern smalltalks do some JITting	00:12
robinr	the fastest cvs-git importer is largely written in ruby	00:12
	ths RCS parser is C though	00:12
	Randal: relatively slow compared to C yes, not not much else	00:13
theCarpenter	RoR is garbage for speed	00:13
Randal	anyway - the basic point is that Smalltalk is pure OO and should be used to teach OO	00:13
	I didn't make any claims for speed	00:13
Mikachu	but then everyone will just know smalltalk	00:14
robinr	It's fast enough for that purpose	00:14
Randal	they'll know real OO	00:14
	→ besonen_pidgi1 joined	00:15
semi_	Randal: suggest some reading material for that?	00:16
robinr	semi_: OO or smalltalk?	00:17
DrNick	Ilari: you still shouldn't design your own crypto algorithms	00:17
semi_	well, not for me, but for those that "should learn OO with smalltalk first"	00:18
robinr	I worked for a company that ran smalltalk courses for newcomers to OO	00:19
semi_	that's pretty cool	00:19
Ilari	DrNick: Of course not.	00:19
	DrNick: But encrypting data is the easy part...	00:19
	→ yorgen15 joined	00:20
robinr	"we" (not me) did smalltalk implementations of systems. Unfortunately they had those smalltalkish GUI disease, which I think is the reason Smalltalk never got big.	00:22
	users didn't care about the implementation language, but they see the user interface	00:23
	→ nzkoz joined	00:23
Randal	there are a few books on squeak smalltalk at your local bookstore	00:24
	there's a few free books online as well	00:24
	← jkimball4 left	00:24
Randal	the trouble is that I learned before those books were out, so I don't know which ones teach well	00:24
	what is a "smalltalkish GUI disease"	00:25
robinr	the user interfaces from hell that most smalltalks apps used to have	00:25
Mikachu	can't you make c-smalltalk bindings for any lib(toolkit) you want to use?	00:26
robinr	the look and fell from Xerox Palo Alto of 1980 is not that nice	00:26
seanh	theCarpenter -- yeah the XO's filesystem looks like the hammer that Randal was referring to. It'd be great if it's available for general use one day	00:26
robinr	Mikachu: sure, there were even a couple of smalltalk implementers that tries to make the platform compliant	00:26
Randal	everything in the XO is open source	00:26
	go get it and install it. :)	00:27
seanh	Randal -- yeah I know, so I guess it will be one day	00:27
Mikachu	open source doesn't mean you can use it with no effort :)	00:27
	→ bdiego joined	00:27
seanh	Well for now I don't think they've finished it, and I don't think it runs on anything but the xo operating system and sugar interface	00:27
Randal	smalltalk's UI works "well enough", which is why people focussed on more important things	00:27
robinr	Digital Smalltalk was one, but they dropped portability a lot, creating much more lock-in than one wants	00:28
Randal	it's also completely in smalltalk, which means it can be examined and customized as needed.	00:28
robinr	no, it did not work well enough,	00:28
Randal	"native" widgets raise lots of trouble	00:28
robinr	they tried to implemented the native look but forgot the feel	00:29
Randal	faulting ST for that is like immediately ruling out python because of the stupid indentation. :)	00:30
robinr	so standard things like TAB, shortcuts etc typically never appeared	00:30
	← robewald\|home left	00:30
robinr	maybe good enough for 3270 users, but other..	00:30
Randal	you really ought to look at modern Squeak	00:30
robinr	yes, my ST experience essentially stopped ten years ago	00:31
	I like strong typing though, so I think there's nothing that can make me switch to ST anyway	00:32
	I guess there are different kinds of people	00:32
Randal	Oh yeah, I think we had this conversation	00:32
	smalltalk is strong typing	00:32
	it's just that values are typed, not variables.	00:33
robinr	ok, stop it. I use the term as defined in a book by Horowitz "Fundamentals of programming languages"	00:33
Randal	I don't have that book, therefore I'm not bound by that. :)	00:33
robinr	where strong typing mean that the compiler know the type	00:33
Randal	well - if you want to reduce programmer productivity using compile-time bindings, go right ahead.	00:34
	"strong typing" using your definition is neither sufficient, nor necessary	00:34
robinr	it's one of the classics on programming languages	00:34
Randal	I wrote up an article on that somewhere.	00:34
robinr	yes, they didn't think of languages like java where the type is partially known	00:35
Randal	duck typing is so much more practical, but you have to be out in the field quite a long time to realize why	00:36
robinr	never heard that term	00:36
	ah	00:36
Randal	should I point out the book on my shelf that defines it? :)	00:37
robinr	no, I googled it	00:37
Randal	heh. I'm having a discussion about smalltalk on the git channel, and ubuntu on the smalltalk channel. :)	00:37
	damn channels, what are they for?	00:37
robinr	reducing the number of concurrent users	00:38
	→ jjore-m joined	00:38
	jjore-m peeks.	00:38
robinr	I like way strong typing (horowitz defn) simplifies massive refactoring (in the field)	00:39
Randal	I like the way the Squeak Refactoring Browser simplifies massive refactoring. :)	00:39
	and test driven development means I can sleep at night	00:40
	and since there are tests, there's no need for the compiler to "get the types right" because the tests would have failed	00:40
robinr	second that	00:40
Randal	that's the short version of "strong typing is neither sufficient nor necessary"	00:40
robinr	unfortunately a lot of code in the field is not equipped with unit tests...	00:41
Randal	that's the fault of management	00:41
	not the fault of a compiler	00:41
robinr	management have no clue	00:41
Randal	adding a compiler won't force the other half of the tests to be added	00:41
	hence, a compiler is deceptive	00:41
robinr	the programmers are lying to them	00:41
Randal	by compiler I mean "compile-time type checking"	00:41
rhalff_	Randal: you have any examples of nice seaside sites online ?	00:41
Randal	dabbledb.com is all seaside	00:41
	seaside.st is seaside. :)	00:41
	seaside-hosting.st lets you host a smalltalk image for free!	00:42
rhalff_	oh ok, yeah ever have been there	00:42
robinr	"I'm unit testing" = I'm desperately trying to make the program run one manual test case on production data	00:42
Randal	rhalff_? no comprende	00:42
	jjore-m hasn't really liked DabbleDb. It's pokey.	00:43
robinr	someone start doing unit tests, and the rest adopt the term	00:43
Randal	it's pokey because it's a money decision	00:43
	he throws more money at it, it gets faster.	00:43
iabervon	I personally think that explicit typing is most helpful as comments whose accuracy the compiler enforces.	00:43
Randal	but as long as it's "fast enough", no big deal	00:43
	the problem with explicit typing is that almost every program will want/need to break it at some point	00:44
	consider "generics"	00:44
	and while that should theoretically never be needed, in practice, it occurs a lot more often than theory says	00:44
	hence, duck typing works better than compiler-enforced typing in practice	00:44
	sadly, kids are coming out of school brainwashed into believing compile-time typing solves all.	00:45
robinr	if programmers write their unit tests	00:45
Randal	well, that's why we should teach OO with smalltalk, not Java	00:46
	or C++	00:46
theCarpenter	Randal: What schools are you talking about?	00:46
Randal	then unit testing becomes the norm, not the exception	00:46
iabervon	If there were a "Whatever" type, that did duck typing, but the rest of the type system were explicit, I think that would be ideal.	00:46
theCarpenter	Or is this just one of those "back in my day" rants? ;)	00:46
Randal	well, consider whatever school brainwashed robinr. :)	00:46
jjore-m	Randal, shrug I thought it was pokey so I stopped using it. It wasn't fast enough.	00:46
robinr	lol	00:47
theCarpenter	well robinr, what was this evil school? ;)	00:47
robinr	I don't want to give them a bad reputation	00:47
theCarpenter	haha	00:47
robinr	it's not their fault	00:47
	it's from my experience. the compiler's typing helps my memory	00:48
theCarpenter	Randal: I say "what schools are you talking about" because the very first "introductory programming" class you take here, the first thing they say when discussing typing is that "compile time type checking does not resolve every issue"	00:48
robinr	noone here said that	00:49
theCarpenter	thanks for that little duck typing mention, I've never heard of it before =]	00:52
	Hey cool, duck typing is used Python... 8-)	00:52
Randal	duck typing is what we've come to call the type of strong typing that "dynamic" languages use, like Perl, Python, Smalltalk, etc.	00:53
	where the value is typed, not the variable	00:53
	and any value that responds to whatever methods we happen to send is OK	00:53
	"if it walks like a duck, it's a duck"	00:53
iabervon	Are there Smalltalk IDEs that annotate variables with the methods they need to have? That's what I'd like to see.	00:55
Randal	"need to have"?	00:55
robinr	iabervon: you don't need that	00:55
Randal	you mean, show all uses of this variable?	00:55
	for a local variable, that'd be relatively easy to spot. :)	00:56
	for an instance variable, yeah, you can browse all accesses to an instance	00:56
	either in the class, or class and all subclasses	00:56
robinr	there are tools that do some static type checking, I've been told	00:56
	← bdiego left	00:56
iabervon	Yeah, but for method arguments.	00:56
nzkoz	I personally like steve yegge's take on non inferring static type systems	00:56
	http://photos1.blogger.com/blogger2/1715/1669/1600/larson-oct-1987.gif	00:56
Randal	again, method arguments are all within a few lines of code	00:56
	if your method is longer than 20 lines, you've misfactored	00:56
pedro_mg	like in ruby 3.methods	00:57
iabervon	Well, the call site isn't going to be nearby.	00:57
robinr	pedro_mg: ?	00:57
nzkoz	robinr: in ruby you can ask an object what methods it has. by calling .methods	00:57
iabervon	If you see that foo gets used in bar.baz(foo), that has implications about what methods of foo will be used by that call that aren't easy to see locally.	00:58
pedro_mg	like 3.respond_to?"hash"	00:58
robinr	ok, I think all dynamic languages have that	00:58
nzkoz	robinr: don't believe 'all' of them can. but yeah, it's pretty common	00:59
robinr	nzkoz: just can't think of anyone that doesn't but then I don't know em all	01:00
	metadata comes natural in such environments	01:01
	→ tmokros joined	01:03
robinr	a problem with the "quacks like" analogy is that quack may mean entirely different things "foo :quack". Take a [ anScm :reset "Foo" ] as an example	01:03
	it's not just that different scm's implement reset differently, it's not even the same concept	01:04
	robinr was thinking of revert actually	01:04
nzkoz	sure, but similarly I could have public void reset() { doSomethingUnexpected();}	01:05
iabervon	Obviously, you need "foo :duck.quacks". Otherwise, it just quacks, not necessarily like a duck.	01:05
robinr	but in that case you are breaking the contract	01:05
nzkoz	I dunno, I'm always skeptical of attempts to solve code review problems with technology.... but I don't feel particularly religious about dynamic typing. It's just that anything else seems 'dumb' to me. but I don't claim to know everything	01:07
robinr	that's sensible	01:08
iabervon	In order for code review to work, you need the language to make it harder than not to write something that looks right and is wrong.	01:09
	→ jkimball4 joined	01:09
seanh	If I run git checkout when the working copy has been modified, it seems to print out a list of the modifications in the working copy. Is this just to tell me why I can't checkout right now?	01:10
iabervon	I wonder how much it would hurt to try to make a version of git-am that would search history for somewhere that a patch applies.	01:10
	seanh: if it didn't say it couldn't do the checkout, then it did it, and those changes are floated over to the branch you've checked out.	01:11
robinr	seanh: you can, but you need to tell git more about what you want to happen; like do you want to throw away your changes or merge them into branch you are switching to	01:13
Mikachu	if none of the changed files need to be touched by the checkout, it'll do the checkout, otherwise you have to say -m	01:13
seanh	Oh ... so it does the checkout and the prints a status report? I am now on the branch I checked out, but my working copy hasn't changed, and it now differs from the branch I'm on in the ways that git printed out	01:14
iabervon	Other parts of your working copy may have changed; other than that, yes.	01:15
seanh	what d'you mean other parts of my working copy?	01:15
	← tongueroo left	01:16
iabervon	If there are files that you hadn't touched in your working copy, and those are different between the two branches, those will have changed to the new branch's versions.	01:16
seanh	Oh right, thanks	01:16
	So is checkout the right command to use if you want to throw away what's in your working copy and go back to some commit?	01:17
	(or is it reset?)	01:18
	← KirinDave left	01:18
Mikachu	reset will change a branch tip, while checkout will just move HEAD around	01:18
	→ glommer joined	01:18
iabervon	Checkout is what you want if you want to make individual files in your working copy match a commit.	01:18
seanh	it seems that either git checkout -f or git reset --hard would do it	01:18
	→ consolers joined	01:19
iabervon	"git checkout HEAD file..." does those files. "git reset --hard HEAD" is the right thing for discarding all of your local changes.	01:19
	→ tongueroo joined	01:20
iabervon	Without filenames, "git checkout branch" retains your changes while moving to the other branch.	01:20
consolers	hmm, doesnt it fail if you had changes ?	01:22
seanh	Bug what about "git checkout -f branch"? That deletes your changes. right? Sounds the same as git reset --hard	01:22
robinr	reset --hard plus checkout branch	01:22
iabervon	Yeah, but it has some other issues; IIRC, if you have tracked files that aren't in the branch you're checking out, "git checkout -f branch" won't remove them.	01:24
Mikachu	seanh: and like i said, checkout doesn't move your branch tip	01:24
seanh	So if I want to move a git repository, but minimise the amount of data that has to be transferred, I could just move the .git file to the new location, an empty folder, then run "git reset --hard HEAD" in that folder and it'll checkout all the files	01:24
	→ Luniv joined	01:26
seanh	that seems to be what it did anyway	01:26
iabervon	Yeah, that should work.	01:27
seanh	I really like that about git, that if you want to move a repository around you just move the folder. You can burn it to a disc or something. That combined with every repository having a full signed history and git-fsck, is really good	01:28
consolers	can anyone recommend pointers to other fast-import scripts that i can read for my edification	01:29
	(i've seen git-import.sh and those in contrib/)	01:30
	→ bdiego joined	01:35
robinr	consolers: http://ww2.fs.ei.tum.de/~corecode/hg/fromcvs/	01:37
consolers	[any way to get the tip from hg as a tar.gz by http] ?	01:41
	(without hg, i.e)	01:41
robinr	doesn't clicking the links at the top work?	01:41
Randal	how many examples do you need? :)	01:42
consolers	sorry i was blind	01:42
Randal	well I mean even before that last link?	01:43
	were you looking for a particular importer	01:43
	or more ways to write the code?	01:43
	fred gets fedup with msysgit	01:43
	fred exports his directory on windows, mounts it with smbfs on the laptop next to the machine currently in windows, and uses a decent environment for it :)	01:44
consolers	just checking prior art	01:46
	← mjc left	01:46
	← anttih left	01:51
Randal	"art" used loosely. :)	01:52
	→ pd joined	01:54
	→ GyrosGeier joined	01:55
	← dpc left	01:58
	← tongueroo left	02:04
	← seanh left	02:05
theCarpenter	Any Russians in here?	02:07
	← tarrybon1 left	02:16
	← glommer left	02:16
	→ samsarin joined	02:18
	← jjore-m left	02:27
	→ spuk joined	02:28
	← ekim\|linux left	02:38
iabervon	-mm has a lot of patches that don't have subject lines.	02:44
nzkoz	Anyone around who knows how git-svn works with the git 'plumbing'? I have a few questions about why sharing is difficult with git-svn repositories	02:47
	→ rjent joined	02:49
	→ tarrybon1 joined	02:53
	→ elmex joined	03:06
	← jkimball4 left	03:10
mugwump	nzkoz: its ref layout isn't great	03:11
	if you're sharing you have to write push: lines that look like refs/remotes/:refs/heads/	03:11
nzkoz	I'm just wondering (assuming) it's some fundamental problem that makes it hard to share git-svn repositories	03:11
	when you dcommit it rewrites the commits right?	03:11
mugwump	yes	03:12
	so that someone pulling via git-svn ends up with the same commit ids	03:12
nzkoz	but it means that someone pulling from my git svn repository will see my own commits, then get merge issues if they rebase again after I dcommit right?	03:13
	which is an artifact of the fact that commits are signed so you can't change them ? (obviously that's a feature ;))	03:13
	← meandtheshell left	03:13
	← consolers left	03:13
	→ eatnumber1 joined	03:13
eatnumber1	hey, I used git reset HEAD^^^ to move back in the commits, but I realized I did something wrong... how do I get back to the original head	03:14
	?	03:14
iabervon	git reflog will show you.	03:15
	← nashdj left	03:15
eatnumber1	what do I do with reflog's output?	03:15
mugwump	nzkoz: essentially	03:15
nzkoz	mugwump: so without some kind of git core 'this changeset is equivalent to that changeset' you'll never be able to share the repositories?	03:16
mugwump	they won't get merge issues in many simple cases	03:16
	there is the patch ID	03:16
iabervon	If you find the point you want on the right side, look at the short hash on the left side, and do "git reset <hash>"	03:16
	→ dsaxena joined	03:17
	→ _spuk joined	03:18
mugwump	nzkoz: it would be possible in principle to store the extra information in svn properties when dcommitting, so that the git patch didn't have to be rebased	03:19
eatnumber1	ok, I got that fixed, now I want to fix the Author of a change made three commits ago... how would I do that?	03:19
mugwump	but I think nobody cares enough	03:19
iabervon	You can only really do it if you haven't published the changes yet.	03:20
eatnumber1	iabervon: darn, I published it	03:20
mugwump	eatnumber1: lots of ways, but to start off with, make a new branch before the commit you made, cherry pick it, fix it up with git commit --amend, and then cherry pick the later ones	03:21
eatnumber1	mugwump: how do I make a new branch before the bad commit?	03:21
iabervon	In that case, people have the history with the author wrong and there's no way to make them forget it.	03:21
mugwump	eg, git checkout -b tmpbranch HEAD~3	03:21
eatnumber1	iabervon: can I fix it in the repo?	03:22
mugwump	rebasing public history is considered antisocial on many projects, but it is fine - people tracking your repo will have to rebase though	03:22
nzkoz	mugwump: cheers, I thought as much. it'd be a neat feature, but definitely can see how my perspective is a little niche here	03:23
mugwump	yeah	03:23
eatnumber1	mugwump: how do I cherry pick the changes?	03:23
mugwump	Did you see the git-svnserver GSoC project?	03:23
	eatnumber1: in gitk, right-click and select "cherry-pick this commit"	03:23
	or run git-cherry-pick	03:23
	actually you can do everything I've said in git-gui and gitk	03:24
	nzkoz: http://utsl.gen.nz/git/git-svnserver.txt	03:24
	The student didn't get the funding and tailed off	03:24
nzkoz	mugwump: are you sam or pointing me at his stuff?	03:24
	← samsarin left	03:24
mugwump	sam i am	03:24
nzkoz	hey	03:25
iabervon	The easy way to generate the new history is "git rebase -i HEAD^^^^", say you want to edit the one you want to edit.	03:25
mugwump	sunny day out huh? :)	03:25
nzkoz	yeah, got a server to migrate :/ dog to walk soon though	03:25
	exercise will be mine!	03:25
eatnumber1	iabervon: is that directed at me?	03:25
iabervon	eatnumber1: yeah.	03:25
	← tarrybon1 left	03:26
iabervon	There isn't really a central location in which you could change the history, and everything in git is checksummed such that you can't change it silently anyway, but if you can convince people to jump onto the new history from the old history, it's possible.	03:27
eatnumber1	iabervon: eh, it's not really worth it... oh well, root is now a contributor to zen-sources =)	03:28
iabervon	IIRC, there's a file of "this address really belongs to this other person", if it's root@somewhere distinctive.	03:29
eatnumber1	ah, ok	03:30
	that works then	03:30
	← eatnumber1 left	03:32
	→ tongueroo joined	03:35
	→ winter-mute joined	03:36
iabervon	Stuff going into -mm gets kind of shredded, it seems.	03:42
	→ KirinDave joined	03:45
	→ tarrybon1 joined	03:47
	→ dsaxena_ joined	03:47
	← dsaxena left	03:47
	← bdiego left	03:50
	← sgrimm left	03:59
	← KirinDave left	04:01
	← tarrybone left	04:03
	tarrybon1 → tarrybone	04:03
	→ tongueroo_ joined	04:05
	← tongueroo left	04:06
	→ z3ro joined	04:06
	← tongueroo_ left	04:06
	← rjent left	04:08
	→ rjent joined	04:08
	← rjent left	04:32
	← tsc_ left	04:37
	→ tchan joined	04:44
	→ priidu_ joined	04:52
	← pedro_mg left	05:12
	→ CIA-5 joined	05:12
	← iabervon left	05:23
	← kukks left	05:24
	→ rkaway joined	05:29
	→ tsc_ joined	05:35
	← tsc_ left	05:36
	← _spuk left	05:38
	← Luniv left	05:55
	→ tongueroo joined	06:03
	→ tarrybon1 joined	06:12
	→ eatnumber1 joined	06:23
	← tarrybone left	06:23
eatnumber1	how do I find out what commits modify a particular file?	06:23
DrNick	git log -- path/to/file	06:26
eatnumber1	thanks	06:26
	→ mithro joined	06:33
	→ orospakr_ joined	06:34
	← orospakr_ left	06:34
	→ Ademan joined	06:35
	→ janm joined	06:46
	← GyrosGeier left	06:47
	← priidu_ left	06:50
z3ro	if a merge failed, and I don't want to resolve it right now, whats the correct way to "abort" the merge?	07:00
	I think last time I used git reset --hard HEAD~1 it messed up some other commits too	07:00
	and I lost some history.	07:00
	← tchan left	07:01
z3ro	hmm nevermind; "git reset; git checkout -f" did it.	07:03
	← theCarpenter left	07:04
	→ kahmalo joined	07:04
	→ tchan joined	07:05
kahmalo	With what precision does Git store commit times? Is it one second?	07:05
	and authorship times	07:05
	Just wondering as I got 9:00:00 exact and am curious of whether there is a nonzero milliseconds field.	07:07
	→ sgrimm joined	07:09
	← sgrimm left	07:31
	→ kmap joined	07:34
	← tarrybon1 left	07:39
	→ tarrybone joined	07:42
	→ shahid21st joined	07:56
	→ aroben joined	07:59
	← tarrybone left	08:01
	→ jerbear joined	08:10
	→ tarrybone joined	08:13
	← Ryback_ left	08:14
	← tongueroo left	08:20
	→ tongueroo joined	08:20
	← tongueroo left	08:21
	→ anttih joined	08:22
shahid21st	hi everybody....	08:27
	I am facing a silly problem..hope you people can help.....	08:27
	i have done some changes in two fiels......and then added by git add	08:28
	so the git status shows to modified files to commit....	08:28
	now I want to revert the changes of one file while keeping the changes of the other intact....	08:28
	How can I do it?	08:28
gitster	"git checkout HEAD the-file-with-unwanted-changes"	08:30
kahmalo	git gui also has a way to do this	08:30
	← jerbear left	08:31
shahid21st	"git checkout HEAD the-file-with-unwanted-changes" - does it mean to revert the-file-with-unwanted-changes to HEAD?	08:31
gitster	"check out the-file-with-unwanted-changes from the HEAD", which means it lets you to discard the changes you made to the work tree and staged to the index with the previous "git add".	08:34
	If it is a new file, then "git rm the-file-I-did-not-want-to-have" will remove it from the work tree and the index.	08:34
	← madewokherd left	08:35
	← kmap left	08:36
kahmalo	man git-config says that branch.autosetupmerge defaults to false; but builtin-branch.c has branch_track = 1 by default. Bug?	08:38
shahid21st	<gister> i changed in my test repo: a file named b.txt then add it via git add. then running - "git checkout head b.txt" shows this error message:	08:40
kahmalo	I don't have that option in .git/config but git branch creates a [branch "foo"] section there even without --track. Not without --no-track though.	08:40
shahid21st	oh sorry ..its working	08:40
kahmalo	I mean, not with --no-track.	08:40
shahid21st	<gitster> thanks	08:40
	I typed head instead of HEAD....	08:41
	→ kmap joined	08:42
kahmalo	And actually I meant git checkout rather than git branch....	08:42
	← shahid21st left	08:44
kahmalo	Although it works the same with git branch too.	08:48
	← Ademan left	08:49
	← kmap left	08:57
	→ kmap joined	08:57
	→ diacritical joined	09:05
	← yorgen15 left	09:06
	→ mgrimes joined	09:06
diacritical	hg-to-git.py fails on corecode's fromcvs hg repo atm	09:06
	← cmarcelo left	09:14
	→ nashdj joined	09:16
	← nzkoz left	09:18
	→ yorgen15 joined	09:20
	→ xaiki joined	09:38
	← diacritical left	09:40
	→ Leonidas joined	09:42
	→ robbat2 joined	09:46
robbat2	if I have a repo that uses 'alternates', how can I convert it to one that doesn't use alternates (I want to get rid of the old one)	09:47
Pieter	perhaps just clone it?	09:49
robbat2	cloneing preserves the alternate file, and it still references the old location	09:49
kahmalo	I suppose you could copy all the objects from the alternate repository and then prune the unneeded ones. If the file names match then the contents should match as well.	09:50
	Packing the alternate repository first might make this easier.	09:51
	→ Dodji joined	10:01
	→ Ademan joined	10:04
	→ alley_cat joined	10:04
	→ Luniv joined	10:09
	← xaiki left	10:10
	→ xaiki_ joined	10:10
	→ ponch_ joined	10:22
	← mgrimes left	10:37
	→ aroben_ joined	10:39
	→ aroben__ joined	10:44
	← aroben left	10:45
	aroben__ → aroben	10:45
	← aroben_ left	10:46
	← Luniv left	10:48
	→ siprbaum joined	10:55
	→ Oejet joined	10:57
	→ Eludias joined	10:59
	→ Sho_ joined	11:01
	→ devogon joined	11:03
	iXce` → iXce	11:10
	← anttih left	11:18
	→ aroben_ joined	11:20
	→ Eludias_ joined	11:20
	← Eludias left	11:22
	← aroben left	11:32
	→ aroben joined	11:35
	→ Yuuhi joined	11:36
	→ Leonidas_ joined	11:42
	← Leonidas left	11:42
	Leonidas_ → Leonidas	11:42
	← aroben_ left	11:50
	→ anttih joined	11:50
	← aroben left	11:56
	→ xTs joined	11:58
xTs	Hi	11:58
	I did a pull on a repo and i see that some files get downloaded but nothing gets updated	11:58
	i just get: foo/bar.c: needs update	11:59
	and 'foo/foobar.c' not uptodate. Cannot merge	11:59
	what am I doing wrong?	11:59
capisce	looks like you have modified files in your working directory	12:01
	either commit them before pulling, or revert them	12:02
xTs	oh, yes, forget that i had changed some things.	12:02
	thanks	12:02
	and how can i solve conflicts? git merge?	12:03
	→ aroben joined	12:04
capisce	pull alredy does a merge	12:05
xTs	yes, but i got a conflict	12:05
capisce	if you have a conflict you can reset to HEAD to undo the merge, or edit the conflicting files and add and commit them	12:05
xTs	ah, ok	12:05
	robbat2 → robbat2\|na	12:09
	← anttih left	12:37
	→ ferdy joined	12:40
	→ Mikachu joined	12:45
	← ferdy left	12:54
	→ bentob0x joined	12:54
	→ ferdy joined	12:55
	← yorgen15 left	13:03
	→ Phoop joined	13:04
	→ nessundorma joined	13:15
	→ robewald\|home joined	13:15
	→ yann joined	13:21
	← Phoop left	13:29
	← nashdj left	13:37
	→ meandtheshell joined	13:55
	→ bobesponja joined	14:00
	← spuk left	14:01
	← ruphy left	14:14
	→ ricky joined	14:18
	→ kukks joined	14:29
	→ nashdj joined	14:31
	→ bdiego joined	15:04
	→ priidu joined	15:05
	← aroben left	15:06
	→ chris2 joined	15:10
heipei	bc	15:15
	oops ;)	15:15
	← xTs left	15:17
	← nashdj left	15:21
	← fobos left	15:22
	→ nud joined	15:29
	← chris2 left	15:30
	→ chris2 joined	15:31
	→ ruphy joined	15:34
	→ _spuk joined	15:59
	← chris2 left	16:01
	→ telmich joined	16:02
telmich	good evening	16:02
Strogg	you have a strange definition of 'good'	16:03
heipei	hi telmich	16:03
	→ bdiego__ joined	16:15
	← kukks left	16:16
	→ madewokherd joined	16:25
	→ kukks joined	16:25
	← bdiego left	16:29
	bdiego__ → bdiego	16:29
	→ ruskie joined	16:42
	→ sgbirch joined	16:42
	→ pcapriotti_ joined	16:43
	→ KillerFox joined	16:43
KillerFox	hello	16:44
	pcapriotti_ → pcapriotti	16:44
Mikachu	hi	16:44
	→ ruphy-lapdog joined	16:44
KillerFox	is there a way to merge two different git repositories together?	16:45
	i've to two projects which startet at it's own repos and now I like to merge them together with all commits	16:45
	s/startet/started/	16:45
kahmalo	I should think it's possible	16:48
	first you should fetch both series of commits to the same repository	16:49
Strogg	you know.. git-svn is pretty damn hot.	16:49
KillerFox	Strogg: well it still is, but it's a pain if you _must_ set Id fields (as a policy)	16:50
	→ Luniv joined	16:50
Strogg	I could understand that.	16:50
KillerFox	kahmalo: so I must start over a new repo and fetch all commits into that repo?	16:51
corecode	KillerFox: you can simply merge them and create a repo with two roots	16:52
	KillerFox: if that is what you want	16:52
kahmalo	git-merge seems happy to merge without a common parent, at least when there are no conflicting files	16:52
	KillerFox: do you have many branches in those two repositories?	16:53
	→ Ryback_ joined	16:54
KillerFox	kahmalo: no, they are all master branches, currently	16:55
kahmalo	KillerFox: I tried this here with two repositories	16:56
	fetched the head of one into another: git fetch ../joae HEAD	16:56
	then made a branch for that: git branch hips FETCH_HEAD	16:56
	and merged it in: git merge hips	16:56
	though it'd work without a branch as well, I suppose	16:56
	git fetch displayed "warning: no common commits"	16:57
Mikachu	KillerFox: ot, urxvt 8.7 is out :)	16:57
KillerFox	kahmalo: thank you, I'll try that.	16:59
	Mikachu: heh	16:59
	← ruphy left	17:01
	← priidu left	17:05
	→ chris2 joined	17:06
	→ yashi joined	17:09
	→ lu_zero joined	17:28
	→ cmarcelo joined	17:30
	← kmap left	17:39
	→ kmap joined	17:40
	← kmap left	17:41
	→ kmap joined	17:41
	→ ruphy-lapdoog joined	17:43
	→ pcapriotti_ joined	17:44
	← nessundorma left	17:45
	robbat2\|na → robbat2	17:46
	→ tango_ joined	17:46
tango_	hello all. is there a git command to copy a file, so that git knows that the next commit of the file can delta from the other one? something like git mv, but actually a git cp	17:47
Ilari	tango_: Just copy the file and add the copy.	17:47
albertito	tango_: git doesn't record the cp, but uses heuristics to detect it. Just use cp and add, and then git log -p -C and you'll see git noticed the copy	17:48
Ilari	tango_: git mv is actually pretty much is 'git rm --cached', 'mv' and 'git add' combo...	17:48
tango_	ah ok. even if the first checkin is a modified version?	17:48
albertito	tango_: it's an important design choice, you can find information about the rationale behind it on the wiki, if you're interested	17:48
tango_	or should I cp, git add, commit, edit, commit again?	17:49
	I mean, can I do cp, edit, git add & commit, or should I do cp, git add & commit, edit, commit ?	17:49
	← eatnumber1 left	17:49
albertito	tango_: if it's not completely modified, it will notice a close-to-100% similarity between the files, and notice it too	17:49
tango_	ok	17:50
albertito	tango_: it actually tells you the "similarity index" of the file. Try doing it and see what git log -p -C shows you. If you're unhappy, you can always use git reset to go back as nothing happened =)	17:50
tango_	ok thanks 8)	17:51
doener	tango_: git doesn't care at all _how_ the contents came to be. It just looks at the content itself.	17:51
	thus it doesn't have to record a cp/mv at all, but just figures it out from the content	17:51
	← kmap left	17:52
	→ kmap joined	17:52
	← yashi left	17:53
	← kmap left	17:54
	→ kmap joined	17:55
	→ pombreda joined	17:57
	← robbat2 left	17:57
tango_	ah, seems to think it's a new file 8P	17:57
	← ruphy-lapdog left	17:59
	← pcapriotti left	17:59
albertito	tango_: use -C	18:07
	tango_: that enables the "cp detection"	18:07
	tango_: for mv, use -M, which enables "mv detection"	18:07
tango_	I did use -C	18:08
	it still diffs affine_prepare.cpp against /dev/null instead of affine_knee.cpp as it should be	18:08
	← Eludias_ left	18:15
albertito	tango_: using git diff or git log?	18:16
tango_	git log -1 -p -C	18:16
albertito	tango_: because you're limiting it to show only one commit, git doesn't compare the files against the previous commit. Use -2	18:17
tango_	albertito, same thing	18:17
albertito	tango_: (btw, "git log -1 -p" is quite like "git show")	18:18
	tango_: ok, let me try	18:18
	tango_: sorry, it's my fault. It should have been "git log -p -C -C"	18:20
tango_	ah	18:20
	bingo	18:20
	8)	18:20
	← davi left	18:20
tango_	why two?	18:20
albertito	tango_: see git-diff-tree --help: "For performance reasons, by default, -C option finds copies only if the original file of the copy was modified"	18:20
tango_	'try harder' ?	18:20
	ah ok	18:21
	→ fhobia joined	18:21
albertito	tango_: using two -Cs makes it try harder	18:21
tango_	ok	18:21
albertito	(it's "was modified in the same commit", forgot to copy the last part)	18:22
tango_	I see. -C only checks against other modified files, whereas -C -C checks against all other files in the tree, which is way more expensive. makes snese	18:22
albertito	tango_: exactly =)	18:22
tango_	thanks 8)	18:22
albertito	tango_: it's not _that_ expensive to check for 100% similarity	18:23
tango_	well depends how many files you have 8)	18:23
albertito	tango_: it's more expensive to check for cp+edit	18:23
	tango_: yes, but, as you said, it depends on that. As git already has sha1s, it's pretty much comparing the (already calculated) sha1 of the files in the tree	18:24
tango_	ah right	18:24
albertito	tango_: which is not as expensive as doing a (tricky and very optimized, but expensive still) content compare to extract a similarity index	18:25
	→ jerbear joined	18:28
	← jerbear left	18:33
	← ferdy left	18:33
	→ jkimball4 joined	18:44
	→ dfa_ joined	18:47
	← dfa_ left	18:49
	← kmap left	18:54
	→ drizzd joined	18:54
	→ robinr joined	19:04
	← drizzd_ left	19:07
	→ Helios- joined	19:07
	→ _spuk_ joined	19:15
	← meandtheshell left	19:20
	→ sgrimm joined	19:21
	← ponch_ left	19:23
	← _spuk left	19:29
	→ ferdy joined	19:30
	→ aroben joined	19:45
	← ruphy-lapdoog left	20:04
	← janm left	20:08
	→ janm joined	20:14
	→ Ingmar joined	20:15
	→ FransE joined	20:15
	→ nessundorma joined	20:16
	← Oejet left	20:27
Randal	Hmm. git.git's Makefile uses $(SCRIPT_PERL), but Documentation/Makefile uses just a PATH-based "perl"	20:38
	that's not good	20:38
	Randal submits a but report to the mailing list	20:40
Randal	or even a bug report :)	20:41
	→ pigeon joined	20:46
	→ FransE_ joined	20:47
	← tango_ left	20:48
	→ jartz_ joined	20:49
	→ jartz__ joined	20:50
	← jartz left	20:56
	→ csc` joined	20:57
	← jartz_ left	20:58
	→ pigeon_ joined	20:59
	← FransE left	21:04
	→ meandtheshell joined	21:05
	→ rtmfd joined	21:06
	← sgbirch left	21:08
	→ pigeon__ joined	21:09
	← pigeon left	21:09
	pigeon__ → pigeon	21:09
ibid	a but report seems to me an applicable term	21:11
	→ male joined	21:16
	→ pigeon__ joined	21:19
Randal	is the git.git maint branch generally safer?	21:20
	how does it differ from master?	21:20
	← mithraic left	21:21
	← pigeon_ left	21:21
chrizz-	Randal: check http://marc.info/?l=git&m=119569175608108&w=2	21:22
Randal	[reading	21:22
	Hmm. so maybe I want to track maint on my more conservative machine	21:23
	← pigeon left	21:24
	← bdiego left	21:24
	→ pigeon joined	21:29
	→ ribas joined	21:30
	← aroben left	21:32
	← robewald\|home left	21:35
	← FransE_ left	21:38
	← Leonidas left	21:40
	← pigeon__ left	21:41
	→ redondos_ joined	21:44
	→ aroben joined	21:45
	redondos_ → redondos	21:49
	← pigeon left	21:54
	→ pigeon joined	21:57
	→ ebel joined	22:02
gitster	I am not so sure about that --- I do not know of anybody who actively fixes bugs that follow 'maint', but they tend to be running something closer to 'master' (and their fix are more applicable to 'master').	22:15
	'maint' is more for people who for some unknown reason prefer to have named releases ;-)	22:15
	← sdh left	22:17
	← devogon left	22:17
ndim	$ git status	22:19
	fatal: Not a git repository	22:19
	$ git diff	22:20
	$ echo $?	22:20
	0	22:20
Mikachu	is this a bug?	22:20
	cd git; touch v1.5.3.6; git-show v1.5.3.6	22:20
ndim	Is this intentional?	22:20
Mikachu	fatal: ambiguous argument 'v1.5.3.6': both revision and filename	22:20
	Use '--' to separate filenames from revisions	22:21
	← jkimball4 left	22:21
Mikachu	oh okay, you can just say -- after without any filenames and it'll know it's a tag	22:21
	that is sort of a crazy syntax though	22:21
ndim	Mikachu: Unusual.	22:21
Mikachu	you usually only need -- for the things that come after it	22:21
mugwump	-- denotes that all switches came before it	22:25
	← pigeon left	22:25
mugwump	and arguments follow	22:25
	← nessundorma left	22:25
Mikachu	i always thought of it as "only filenames follow"	22:25
	for example you can say ls file1 -l -- file2 -h	22:25
	and it will give you a detailed listing of file1 file2 and -h	22:26
mugwump	yeah. you're right, but I've also seen it used to denote the end of a list as well	22:28
Mikachu	i guess it doesn't happen very often that someone creates a tag with the name of a file	22:28
mugwump	there was a patch for that today actually :)	22:29
Randal	or a file with the name of a tag. :)	22:29
Mikachu	i like to put my list of configure options in a tag for some programs though	22:29
	Randal: i guess one advantage of using maint over next is that git-cherry-pick doesn't crash and git-tag -s works	22:32
	← csc` left	22:33
	→ FransE_ joined	22:34
	← FransE_ left	22:35
	← kahmalo left	22:38
gitster	Huh?	22:43
	"--" denotes all revs came before it; pathspecs may follow but do not have to (i.e. can be empty).	22:45
Mikachu	still, it's weird that -- changes the meaning of what came before it, i've never seen that in any other app	22:51
	just saying it's a bit surprising	22:52
gitster	Well, get used to it and thank that Linus did not pick anything more esoteric, say "----" ;-)	22:55
	← KillerFox left	22:55
Mikachu	heh	22:55
Randal	-[]- :)	22:56
gitster	Sounds good and much easier to type than "--" ;-)	22:56
	← aroben left	22:57
Ilari	Maybe not on keyboard layout I have... That would be -, ALTGR+8, ALTGR+9, -...	22:58
Mikachu	in a way ---- would be better since then i wouldn't confuse it with the usual meaning of -- :)	22:58
gitster	Mikachu: about the "cherry-pick" thing, my understanding is that it happens only when you revert/cherry-pick an annotated tag. Is that correct?	22:59
	IOW, you would not see the issue if you did "git cherry-pick foo^0" instead of just "foo".	22:59
Mikachu	i'm not sure, i only saved the test case that crashed for me	22:59
	← bentob0x left	22:59
Mikachu	i will compile the version without my patch and try	23:00
	is that what you're seeing?	23:00
gitster	I think a proper fix is to just remove the whole "if (needed_deref) { ... }" thing, and I vaguely recall suggesting that to you already.	23:01
Mikachu	well, i'm not so comfortable editing the code as i have no idea what most of it does	23:01
gitster	Will prepare a patch to 'master'.	23:02
Mikachu	btw does 'git-tag -s foobar' work for you? here it just creates an unannotated tag	23:03
	→ GyrosGeier joined	23:03
Mikachu	-a does pop up the editor though	23:03
	aha, -a -s works	23:03
	(but surely that's not intended?)	23:04
	← alley_cat left	23:04
gitster	Gaah, thanks for telling me offline. This should have been brought up on the list much earlier.	23:05
Mikachu	i just cherry picked an annotated, unsigned tag and it didn't crash	23:05
	was that ironic or not? :)	23:05
gitster	It is not ironic nor non-ironic because I do not know what you are talking about, sorry.	23:06
	The "crashing" example you had, what was the command line?	23:06
Mikachu	okay, we are probably equally confused then	23:06
gitster	"git cherry-pick XXXX" what was XXXX?	23:06
Mikachu	override_events_2	23:06
gitster	which is? A tag? The name of a working tree file? A branch?	23:07
mugwump	gitk frequently gives me an apparently harmess dialogue when I cherry pick	23:09
Mikachu	sigh, i'm so stupid	23:09
	i forgot the ./ when i ran git-tag	23:09
	i mean cherry pick	23:09
gitster	The parse-options conversion surely look wrong.	23:10
	It lost that "-s implies -a" bit. The fix should be trivial.	23:10
	I wonder why our testsuite did not catch it.	23:10
Mikachu	git cherry-pick indeed crashes on my test created annotated tag	23:10
gitster	I know. That's what I've been saying.	23:10
Mikachu	yes, i just wanted to confirm it	23:11
	because i mistakenly said 5 minutes ago that it didn't :)	23:11
	→ pigeon joined	23:11
gitster	So override_events_2 is an annotated tag?	23:11
Mikachu	yes	23:11
gitster	ok, thanks. Then, with a vanilla git, "git cherry-pick override_events_2^0" would not crash.	23:11
Mikachu	also confirmed	23:12
gitster	Thanks. I am sending a fix to the list ;-)	23:12
Mikachu	for some reason i always start making silly mistakes when trying to report bugs, i don't know why, i usually don't :)	23:12
	→ aroben joined	23:19
gitster	Another patch for "git tag -s" sent. Thanks.	23:22
	→ janm_ joined	23:24
	← Dodji left	23:29
	→ mithraic joined	23:34
	→ flaguy joined	23:36
	← janm left	23:36
kelvie	are git-repacks supposed to be CPU heavy?	23:36
	it doesn't get really high on top	23:36
	so I'm assuming the fs syscalls are taking up the cpu	23:37
mugwump	you can increase --window if you want it to demonstrate its CPU swallowing capabilities	23:38
Ilari	kelvie: AFAIK, system time is included in process time...	23:39
Mikachu	i don't think CPU% includes iowait though	23:39
kelvie	oh hrm	23:39
	I just don't recall git-repack slowing down my computer at work	23:39
	this much	23:40
	so I want to see if I need to switch my filesystem :P	23:40
Ilari	Mikachu: It does not.	23:40
cehteh	not enough ram, swapping?	23:40
kelvie	2g ram 8g swap?	23:40
	I think I'm OK :/	23:40
mugwump	yeah, it can be slow if you've got loads of loose objects and a FS which doesn't handle that	23:40
cehteh	bit much swap :)	23:40
mugwump	eg xfs	23:40
kelvie	hehe I don't wanna repartition in case	23:40
Ilari	Actually, if system is set up wrong, any kind of actual I/O makes the system really crawl...	23:40
kelvie	yes I just recently decided to try xfs	23:41
mugwump	turn on the auto-GC	23:41
Mikachu	speaking of git-repack, is there a way to exclude duplicated objects from packs in a shared clone? right now i mv the .pack file away, then run git-unpack-objects, and finally git-repack -d	23:41
kelvie	all my other boxes with reiser/ext3 are fine	23:41
cehteh	i have xfs too .. no problems with it	23:41
kelvie	but whenever I write/unlink/read with XFS my computer slows down :/	23:41
mugwump	cehteh: which kernel version? or on irix?	23:41
cehteh	x86	23:41
	→ priidu_ joined	23:42
cehteh	laptop dmcrypt partition	23:42
	not really fast but usually no big performance degrade either	23:42
	← ferdy left	23:42
mugwump	strange	23:42
Ilari	kelvie: I have the same problem, but with ReiserFS...	23:42
kelvie	mm I've never seen this with reiserfs.. perhaps my problem is elsewhere	23:42
mugwump	Because as far as I've noticed, xfs performance has totally sucked for operations involving lots of files	23:42
kelvie	same :/	23:43
cehteh	i dont say that it is blazing fast .. but performance is far from being annoying slow	23:43
kelvie	well performance was OK	23:43
	but it slows down my computer :/	23:43
Mikachu	checking out linux takes some 5 minutes on xfs here	23:43
	← aroben left	23:43
	cehteh tries	23:43
Ilari	When the system slows down, iowait is real high...	23:43
mugwump	yeah I think it must be silly directory locks or something like that	23:44
	← priidu_ left	23:44
cehteh	mhm git clone --no-hardlinks /usr/src/linux-2.6 ... would a local clone suffice?	23:44
kelvie	arg next weekend I'll switch back to reiser and see how it goes I guess	23:44
	→ priidu_ joined	23:44
gitster	It was reported some time ago (and Linus was in the discussion) that xfs was slow.	23:44
Mikachu	my workaround is to have a bare clone of linux on xfs and the workingdir on tmpfs in /tmp	23:45
	where it takes 5 seconds to checkout all 35000 files	23:45
cehteh	outch	23:45
Mikachu	then when i'm done fiddling branches around and merging, i copy a tree back to disk to compile	23:45
Bartman007	Mikachu: hooray for kludges!	23:46
Mikachu	not that i'll be doing that a lot, tried .23 for a month and am back on .19-ck2 now :)	23:46
cehteh	mhm ok cloning linux is not really fast on my laptop	23:46
Mikachu	cloning is okay, it's just the checking out bit that's slow	23:47
kelvie	because xfs is _really_ bad at unlinking?	23:47
cehteh	thats what i meant .. checkout is part of clone	23:47
Mikachu	that's true	23:47
cehteh	yes thats a bad spot of xfs	23:47
	← nud left	23:47
Mikachu	kelvie: someone told me the other day xfs is slow at all metadata operations	23:47
kelvie	ugh	23:47
	I want to find the guy that told me xfs was mature and fast	23:48
	and choke him	23:48
Mikachu	essentially reiserfs for many small files and xfs for big files	23:48
cehteh	how about ext4?	23:48
kelvie	I can't even shrink it to switch	23:48
Mikachu	is what he told me	23:48
telmich	kelvie: xfs is quite mature and fast	23:48
	→ engla joined	23:48
kelvie	except you can't shrink it :/	23:48
	and unlinking is painfully slow	23:48
telmich	though it has some quota problems	23:48
	← fhobia left	23:48
Mikachu	unlinking big files on xfs is much faster than on ext3	23:48
cehteh	real 2m52.753s	23:48
telmich	kelvie: well, most fs cannot shrink, only grow	23:48
	compare it to reiser, jfs, ufs, ...	23:49
kelvie	looking at the gparted list.. most of them can except jfs and xfs	23:49
	not online of course	23:49
telmich	well, only ext3 and perhaps ext4 can do, afaik	23:49
robinr	ntfs, fat	23:49
kelvie	I'm pretty sure reiser can?	23:50
telmich	kelvie: well, ext3 is quite robust and is the only fs I would use when wanting to shrink	23:50
cehteh	heh .. how about reiser4? :) whenever i tried it, i manged to crash it .. but performance before that crash was impressive	23:50
telmich	kelvie: reiser cannot	23:50
	→ tongueroo joined	23:50
telmich	and ntfs and fat are no posix-fs; and I really wonder, how shrinking fat should work, though it's perhaps just defrag + rewrite partition table	23:50
	robinr always needs more* disk space, never less*	23:50
tongueroo	git status \| grep 'deleted:' \| sed 's/.*deleted:\s+//' , the \s+ regular expression in sed doesnt seem to work...?	23:50
telmich	I mean, in general all fs _could_ be shrinked, when doing it righn	23:51
tongueroo	git status \| grep 'deleted:' \| sed 's/.*deleted:\s//' , \s works though	23:51
cehteh	with concurrent load reiser4 was up to 2.5 times faster than any other fs ... but it chews lots of CPU too	23:51
	telmich is a big jfs fan - almost no cpu usage	23:51
gitster	"git ls-files -d" perhaps?	23:51
cehteh	maybe an alternative for laptops where cpu saving is good	23:52
tongueroo	gister: ahhh.. much easier	23:52
	thanks	23:52
Mikachu	tongueroo: -r or \+	23:52
	it defaults to basic rexe	23:53
	regex*	23:53
	← rtmfd left	23:53
robinr	telmich: you just mean there is no known method of shrinking an XFS fs?	23:53
gitster	"dump; mkfs; restore"?	23:53
	← flaguy left	23:53
robinr	no really the same	23:54
gitster	you get the defragmentation for free ;-)	23:54
robinr	xfs has a defragmenter	23:54
cehteh	.. and you should use it, because it fragments	23:55
robinr	works on mountes fs too	23:55
	well all file systems fragment their files	23:55
kelvie	telmich: man resize_reiserfs; you can shrink it :/	23:55
geckosenator	I would rather have a filesystem with compression	23:55
cehteh	well more than you would expect	23:55
robinr	xfs seems better than most at not doing it, but it does happen	23:55
cehteh	geckosenator: reiser4 \o/	23:55
gitster	and too slow to be usable to extract linux source? what good is it with bells-and-whistles?	23:55
	→ aroben joined	23:55
geckosenator	cehteh: heh yeah	23:56
	gitster admits he is not in his best mood.	23:56
tongueroo	Mikachu: thanks! have been wondering why my sed regex dont work	23:56
cehteh	anyone knows anything about the fate of Daniel Phillips tux2 (..3?)	23:56
	tongueroo: locale problem?	23:57
	Ilari came to conclusion that this system is severly misconfigured... And fixing those practically means reinstalling the whole thing... :-(	23:57
cehteh	eh .. posix/sed regex dont know \s iirc	23:57
tongueroo	cehteh: nah, im not to escaping the sed regex properly, 's/\s\+//' instead of 's/\s+//'	23:57
telmich	robinr: yes	23:58
	kelvie: uhh, that's new to me	23:58
cehteh	tongueroo: [::space::]+ (or was it [[:space:]]?)	23:58
tongueroo	the first one	23:58

Logs Search ←Prev date Next date→

Channels Documentation