IRC log for #darcs, 2007-03-06

All times shown according to UTC.

Time	Nick	Message
09:48	CosmicRay	drat, an "impossible case" trying to tailorize tailor.
09:48	CosmicRay	how ironic is that
14:31	pbx	When the FreeBSD port for darcs asks me if I want to "install server", is it talking about this? http://www.equational.org/darcs-server/
14:31	lambdabot	Title: Darcs-server
14:36	Igloo	Probably; you certainly ought to be able to answer no and end up with a working darcs
14:36	pbx	Igloo: Alas, no -- for other reasons.
14:36	pbx	===> ghc-6.6 unsupported at the moment.
14:36	pbx	*** Error code 1
14:48	^self	heh.
14:54	pbx	It's a bummer! I got it to install on my 6.1 box, but not on 4.11 (work).
14:55	Igloo	Wow, being 2 major OS releases behind sounds scary...but then, I am a Debian user
14:55	pbx	5.x doesn't really count :)
15:02	^self	wait, i'm updating my ports in the other window
15:02	^self	why do you need 6.6? oh, are you on amd64?
15:03	^self	oh...4.11
15:03	^self	.if ${OSVERSION} < 500000
15:03	^self	#DISTFILES+= ${BOOT_DIST4}
15:03	^self	IGNORE= unsupported at the moment
15:03	^self	try uncommenting DISTFILES and commenting out IGNORE?
15:04	^self	4.11 really is obsolete
15:04	^self	no updates, no more ports support, nothing
15:07	^self	sad that my freebsd box is a 2ghz celeron
15:07	^self	so. slow.
15:07	^self	good thing i mostly use it just for mail
15:07	pbx	^self: I'll probably just hold off until we upgrade the box to 6.x.
15:09	^self	i was fooling around with haskell (and using hugs) back in 1996/1997 -- my last years in college
15:09	^self	but then, i dunno, i stopped somewhere
15:09	^self	never got back into it
15:12	pbx	Haskell is very cool. I decided to learn a bit of it this year just for mind-stretching.
15:14	^self	i had some water damage near my bookshelf late last year
15:14	^self	nothing too terrible
15:14	^self	but it made me go through my books again
15:15	^self	and i decided to finally finish the little/seasoned schemers
15:15	^self	i figure, i'll complete the seasoned schemer, then do ullman's ml book
15:50	joelr1	good day folks! how do i unapply recorded patch to my current repo?
15:50	joelr1	i would like to remove patches one by one until i get back to a working version of a file
15:51	Igloo	unpull
15:52	joelr1	does it matter that i recorded patches locally and pushed them to the remote repo? is it still unpull?
15:52	joelr1	Igloo: thanks, trying...
15:53	Igloo	Still unpull (also called obliterate if you prefer)
15:54	joelr1	Igloo: thanks again!
16:15	kc5tja	^self: Dude, my box at home is an 800MHz Athlon, SLOT A (yes, you read that right), which I use for everything from web browsing to playing DVDs, to ... Just what kind of mail are you reading that it makes it so slow?!
16:29	pbx	I solved my darcs problem -- my FreeBSD server has linux-binary-compatability turned on, so the static binary (http://evan.martin.googlepages.com/home) works for me.
16:29	lambdabot	Title: evan.martin - darcs binaries
16:31	pbx	is happy not to have to use something else...
19:03	CosmicRay	arggh.. it should not take 1GB of ram to pull the very first patch on http://darcs.complete.org/debian/bacula under tailor
19:03	lambdabot	Title: Index of /debian/bacula
19:05	CosmicRay	I don't understand. I can darcs get the first 250 patches in the repo and that works fine.
19:06	CosmicRay	but darcs init, and then darcs pull the first patch only, and it uses gobs of ram.
19:06	kc5tja	As I understand things, darcs pull on a specific patch will (1) grab ALL patches, then (2) roll back all patches after the specified patch.
19:09	CosmicRay	yipe.
19:10	CosmicRay	I think that is what get does, actually
19:11	kowey	darcs get works that way, but i don't think this is the case for darcs pull
19:11	kowey	you can try darcs init'ing a blank repository and just pulling one patch
19:11	kowey	is guessing
19:11	kc5tja	11:07 < CosmicRay> but darcs init, and then darcs pull the first patch only, and it uses gobs of ram.
19:11	kc5tja	Looks like that's already been tried.
19:12	kowey	oops! sorry, not reading carefully enough
19:12	kc5tja	:)
19:12	kc5tja	isn't awake yet either. I don't start becoming fully conscious until after lunch.
19:12	kc5tja	(Breakfast provides just enough fuel to get me through til then. :) )
19:13	CosmicRay	shows what frame of mind I'm in; I first read that as "fully conscious until after branching."
19:13	kc5tja	Haha -- to the unaided eye, that would be the type of thing that a machine language coder would say.
19:14	kc5tja	<-- loves assembly language coding.
19:14	CosmicRay	heh, I hadn't even thought of that possibility. I was thinking of branching in the vcs sense
19:14	kc5tja	is reading up on SequenceL, APL, and J and so forth now. VERY fascinating, and for the first time, I understand how fully implicit iteration over vectors and matrices work.
19:14	CosmicRay	<-- hates assembly language coding
19:14	kc5tja	is so bad when it comes to math, which is odd, since I got straight As in college and highschool math classes.
19:15	kc5tja	And, of course, to understand any functional language, you need a solid background in math.
19:15	kowey	that's a rather large patch...
19:15	CosmicRay	kc5tja: I say that to understand math, you need a solid background in a functional language ;-)
19:16	kowey	i have a pretty poor background in math, i can sort of use a functional language... maybe not understand it
19:16	kc5tja	CosmicRay: Funny you mention that -- I'd LOVE to find a book on (Linear) algebra that uses J as a language to help teach it.
19:16	CosmicRay	heh
19:16	kc5tja	That would give me incentive to download and play with it.
19:17	kc5tja	I heard that there was a book on teaching Calculus using J, which introduced the concept of fractional derivatives (not partial derivatives).
19:17	kc5tja	(e.g., it's possible to take the 1.5th derivative of a function, for example)
19:17	kc5tja	Don't ask me to explain -- I haven't read the book, much less find a way to visualize it.
19:17	kc5tja	Or, for that matter, to even consider applications for it.
19:18	sieni	well, they are used in various places in analysis
19:18	CosmicRay	kc5tja: you may find this of some interest: http://changelog.complete.org/[…]mple-Example.html
19:18	kc5tja	sieni: Interesting; I actually didn't know such a thing could even exist. The whole idea of a fractional derivative is totally alien to me.
19:18	lambdabot	Title: Why I Love Haskell In One Simple Example - The Changelog, http://tinyurl.com/yo57sn
19:19	kc5tja	CosmicRay: Checking.
19:19	CosmicRay	this is more down-to-earth, but a nice demonstration of haskell's numeric typeclasses
19:19	kc5tja	Oh, I'm already familiar with typeclasses in general, and I definitely love them.
19:19	kowey	CosmicRay: it might interest you to know that eps files are treated as binary by default
19:19	kowey	i thought i submitted a patch for that...
19:19	CosmicRay	that is probably a good thing
19:19	CosmicRay	in this case, actually.
19:19	CosmicRay	in general, maybe not
19:20	kowey	oh, ok... i saw some huge hunks, and eps, and just assumed they were related
19:20	CosmicRay	kc5tja: yeah, I just thought it was very sweet that it's so easy to define new numbers that carry units with them
19:21	kc5tja	CosmicRay: Just think of how much that would have helped the developers of the mars probe that we lost. :)
19:22	sieni	kc5tja: well, if you consider that taking n derivatives is basically multiplication by x^n (modulo constants) on the Fourier sides, then you might wonder, what happens when n is not an integer.
19:22	kc5tja	Does it perform compile-time checking of units? E.g., attempting to add meters to inches produces a compiler error?
19:22	kc5tja	sieni: What you just said means nothing to me, sorry.
19:23	kc5tja	sieni: I managed to go through Calculus II in college, before having to drop out. I barely grok power series. :) Concepts like "Fourier sides" are not known to me.
19:24	sieni	well, I apparently expressed myself in a very confusing manner
19:25	sieni	kc5tja: http://en.wikipedia.org/wiki/F[…]rential_equations
19:25	lambdabot	http://tinyurl.com/27n3g
19:26	kc5tja	sieni: I'll put that on my list of things to read; I am still not quite finished with the paper I'm reading now.
19:26	kc5tja	(about 15 more pages).
19:26	CosmicRay	kc5tja: hehe
19:30	kc5tja	It's funny reading up on the comparison between SequenceL and APL.
19:30	kc5tja	In the description, they give a one-liner APL function to find prime numbers.
19:30	kc5tja	THen they explain how to do the same thing in SequenceL, but it requires two definitions of code, and the introduction of free variables.
19:31	kc5tja	Then they write things like, "We can easily define things in SequenceL that exist in APL."
19:31	kc5tja	I thought the whole purpose of a language report is to promote your own language, and not someone else's. :)
22:31	malsyned	So this article, http://bryan-murdock.blogspot.[…]sion-control.html which made it to the front page of reddit compares four RC systems and darcs comes out looking really good. He then says "Darcs still reportedly has a deep, serious bug. Don't use it (though it is nice)." There's no further context. Any ideas what deep, serious bug he's referring to?
22:31	lambdabot	Title: Cyclopedia Square: Cutting Edge Revision Control, http://tinyurl.com/yvgeyp
22:32	kc5tja	malsyned: I'm assuming it's the "bug" where darcs takes forever on mammothly huge patches.
22:33	kc5tja	However, just because it sucks RAM and takes forever, doesn't necessarily make it "buggy" in the sense of incorrectness; huge patches take forever to apply, but they DO apply, and they do so correctly.
22:33	malsyned	geez. I was expecting something like "the theory of patches guarantees that your most important change will be silently destroyed". That may be deep, but it doesn't strike me as nearly serious enough to warrant a "don't use it" warning.
22:34	kc5tja	manages a 170MB website with darcs, which contains sub-repositories as well, and darcs has yet to fail me in any way.
22:34	malsyned	sub-repos? how does that work?
22:34	kc5tja	mkdir foo
22:34	kc5tja	darcs init foo
22:34	kc5tja	err
22:34	kc5tja	let's try again.
22:34	kc5tja	mkdir foo; cd foo; darcs init foo
22:34	kc5tja	mkdir bar; cd bar; darcs init
22:35	kc5tja	echo "test" >blah
22:35	kc5tja	darcs add blah; darcs record -a -m "testing added"
22:35	kc5tja	( at this point, the "inner" repository is recorded, but the outer repository has been touched. )
22:35	kc5tja	cd ..
22:35	kc5tja	darcs record -a --look-for-adds -m "Recording the inner repo changes"
22:35	kc5tja	That's basically it.
22:36	kc5tja	So, my website on my local box is located in /v/sites/falvotech.com, which is itself a Darcs repository.
22:37	malsyned	so, changes to the inner repositories also get recorded as changes to the outer repository, as long as you look-for-adds?
22:37	kc5tja	Within that, I have /v/sites/falvotech.com/content/cut, content/kestrel/2/firmware, content/kestrel/2/emulator, etc.
22:37	kc5tja	Each of these are separate Darcs repositories.
22:37	kc5tja	Yes. The outer repository is never updated automatically; you have to manually "record" in that repo as well. This is because, as far as Darcs is concerned, the inner repos are just normal data files like any other.
22:39	malsyned	seems like that would interact poorly with 1) allowing gets from both inner and outer repos, and 2) subrepos with all kinds of non-rc'd files in them.
22:41	kc5tja	It works fine for me.
22:42	kc5tja	I make no claims that this is suitable for everyone though.
22:42	kc5tja	It works for me because:
22:42	malsyned	Yeah, I can see it working great for some use cases, and being unworkable for others.
22:42	kc5tja	1) Inner repos don't track non-RCed files. This is good, because it keeps things like .o files and so forth out of the source distribution.
22:43	kc5tja	2) Outer repo, because I use --look-for-adds, DOES record the .o files, which is good, because if I make a major fubar, I can recover by checking the site out again. :)
22:43	malsyned	Oh, in the comments for that article he provides a link to the problems with darcs he was talking about. He links to http://zooko.com/darcs_demystified.html and points to the section "Darcs in practice", subsection "Troublesome Corner Cases"
22:43	lambdabot	Title: Darcs Demystified
22:44	kc5tja	I'll have to read that article as I get the time.
22:44	kc5tja	Currently at work, so can't spend too much attention on reading at the moment.
22:44	kc5tja	Too busy writing. :)
22:45	malsyned	no problem. it's two corner cases where two different people commit a sequence of patches that result in exponential runtime behavior on any repo with those patches.
22:49	malsyned	hey, what is a partial get, anyway?
22:51	kc5tja	Partial gets are used in conjunction with "checkpoints."
22:51	kc5tja	Basically, a checkpoint is a "snapshot" of a repository with all patches up to that checkpoint already applied.
22:52	malsyned	ah. gotcha. thanks.
22:53	kc5tja	So instead of grabbing 1000 patches, you get only the last 250. :)
23:02	sockmonk	kc5tja: in your 'parent' repo that you were talking about, does it treat the sub-repo _darcs directories like normal directories, or do you have those flagged as boring in the parent repo?
23:08	kc5tja	For my purposes, I treat them like normal, every-day directories.
23:08	kc5tja	I use the outer repository as an archival mechanism.