2008-08-07 05:07 -!- tux3bot(~tux3bot@yzf.shapor.com) has joined #tux3
2008-08-07 05:56 <shapor> http://web.archive.org/web/20060904185736/http://www.complang.tuwien.ac.at/anton/lfs/
2008-08-07 05:57 <shapor> similar thought process on versioning
2008-08-07 06:56 <tim_dimm> wow- that's pretty cool.
2008-08-07 06:57 <tim_dimm> very similar method of managing free space
2008-08-07 07:01 -!- pgquiles(~pgquiles@224.Red-81-39-154.dynamicIP.rima-tde.net) has joined #tux3
2008-08-07 09:23 <shapor> more purdy: http://shapor.com/tux3/
2008-08-07 09:50 <tim_dimm> bitchin'
2008-08-07 09:58 <tim_dimm> that really looks nice shapor
2008-08-07 09:58 <tim_dimm> is this now the "official site" or the mirror site?
2008-08-07 09:59 <tim_dimm> and wtf on zumastor?
2008-08-07 10:07 <shapor> wtf?
2008-08-07 10:11 <shapor> wtf wtf i mean 
2008-08-07 10:11 <tim_dimm> :-0
2008-08-07 10:12 <shapor> i'm thinking of organizing daniels mailing list posts in to a design document
2008-08-07 10:12 <flips> morning
2008-08-07 10:16 <tim_dimm> morning flips
2008-08-07 10:28 <flips> shapor, that would be most wonderful
2008-08-07 10:29 <tim_dimm> u guys heard of MogileFS?
2008-08-07 10:29 <tim_dimm> http://www.danga.com/mogilefs/
2008-08-07 10:29 <flips> never
2008-08-07 10:29 <shapor> its a selfish effort, to absorb it all ;)
2008-08-07 10:29 <tim_dimm> distributed fs that allows you to run any type of fs locally
2008-08-07 10:30 <shapor> Application level -- no special kernel modules required.
2008-08-07 10:30 <tim_dimm> veoh networks uses it.  was developed for live journal
2008-08-07 10:30 <shapor> sounds like not a filesystem really
2008-08-07 10:30 <shapor> actually sounds a lot like google's gfs
2008-08-07 10:31 <shapor> MogileFS is not:
2008-08-07 10:31 <shapor> * POSIX Compliant
2008-08-07 10:31 <flips> kerneltrap picked up the matt dillon dialogue
2008-08-07 10:31 <shapor> so yeah, much like google's gfs, not a real filesystem ;)
2008-08-07 10:31 <flips> useful thing maybe though
2008-08-07 10:32 <flips> who knows
2008-08-07 10:32 <flips> cloud?
2008-08-07 10:32 <shapor> its vrey useful that its open source
2008-08-07 10:32 <tim_dimm> yeah, saw the *not* posix note
2008-08-07 10:32 <shapor> like hadoop and memcached
2008-08-07 10:32 <shapor> nice tools for building clusters for specific applications
2008-08-07 10:33 <shapor> web 2.0 stuff
2008-08-07 10:33 <shapor> speaking of which
2008-08-07 10:33 <flips> danga = livejournal guy
2008-08-07 10:33 <shapor> someone at a startup i was talking to yesterday
2008-08-07 10:33 <shapor> is running redhat's gfs
2008-08-07 10:33 <flips> heh
2008-08-07 10:33 <flips> fun?
2008-08-07 10:33 <shapor> in a backend cluster
2008-08-07 10:33 <shapor> and it keeps crashing
2008-08-07 10:33 <shapor> when load gets high
2008-08-07 10:33 <shapor> cascading failure
2008-08-07 10:33 <shapor> all nodes die
2008-08-07 10:33 <flips> never woulda thunkit
2008-08-07 10:34 <tim_dimm> i'm lurking on the hadoop irc now
2008-08-07 10:34 <tim_dimm> that's how i heard of mogilefs
2008-08-07 10:35 <flips> the lfs article is a good read
2008-08-07 10:35 <flips> thinking way back, it was _really_ dumb of me to try to put the free space bitmaps inside the snapshot
2008-08-07 10:36 <flips> nobody questioned it at the time
2008-08-07 10:36 <tim_dimm> good thing you can question your own work, huh?
2008-08-07 10:37 <flips> if you live long enough
2008-08-07 10:37 <flips> I'm thinking about that because the lfs guy seems to be busy making the same mistake
2008-08-07 10:38 <flips> "At each snapshot we have a set of free blocks (and complementary, a set of allocated blocks) for the files reachable through this snapshot."
2008-08-07 10:39 <flips> btrfs uses per-block refcounts for that (ouch) and zfs uses dead block lists (complexity and weirdness)
2008-08-07 10:39 <flips> tux3 just has a conventional allocator and can figure out what blocks can actually be freed by looking at its version information
2008-08-07 10:40 <flips> this is a huge advantage to keeping all the version information for a given block together in one place
2008-08-07 10:40 <shapor> flips:  i found that site linked off comments on lwn from last year
2008-08-07 10:40 <shapor> related to zumastor i think
2008-08-07 10:40 <flips> ah
2008-08-07 10:41 <shapor> http://lwn.net/Articles/170346/
2008-08-07 10:41 <shapor> hm thats not the one
2008-08-07 10:42 <shapor> http://lwn.net/Articles/239369/
2008-08-07 10:42 <shapor> that one
2008-08-07 10:42 <shapor> ah yeah stumbled across it reading about gplv3
2008-08-07 10:43 <flips> that comment is not very accurate re wafl
2008-08-07 10:44 <shapor> I like linus's comment
2008-08-07 10:44 <shapor> "Umm. You are making the fundamental mistake of thinking that Sun is in 
2008-08-07 10:44 <shapor> this to actually further some open-source agenda."
2008-08-07 10:46 <flips> remember scott mcneally wearing the penguin suit?
2008-08-07 10:51 <shapor> lol that was a long time ago
2008-08-07 10:51 <flips> seems like only yestaday
2008-08-07 10:55 <flips> see "Free-space management and clones" in http://www.complang.tuwien.ac.at/anton/lfs/ and you will see diagrams much like versioned pointers
2008-08-07 10:55 <shapor> yes!
2008-08-07 10:56 <flips> but he didn't go on to realize you could actually represent the version data that way
2008-08-07 10:56 <flips> only has rather similar freeable block algorithm
2008-08-07 10:57 <flips> also, you don't need .killed, on .born
2008-08-07 10:57 <flips> only .born
2008-08-07 10:57 <shapor> you should email him
2008-08-07 10:57 <shapor> invite him to the list
2008-08-07 10:57 <flips> sure
2008-08-07 10:57 <shapor> he is a professor i believe
2008-08-07 10:57 <shapor> he was talking about potentialy putting some students on lfs
2008-08-07 10:57 <shapor> might be useful for tux3
2008-08-07 10:58 <flips> very pretty site shapor
2008-08-07 10:58 <flips> hand coded html?
2008-08-07 10:58 <tim_dimm> looks like you speak his language
2008-08-07 10:58 <flips> I think so
2008-08-07 10:58 <flips> yes I will email
2008-08-07 10:58 <tim_dimm> fartenpoopenspakenziepissin
2008-08-07 10:58 <tim_dimm> :-)
2008-08-07 10:58 <flips> genau
2008-08-07 10:59 <shapor> yes hand coded html
2008-08-07 10:59 <shapor> looked like crap until i added the css
2008-08-07 10:59 <shapor> its very simple just 1996-ish h1, h2, and li tags
2008-08-07 10:59 <tim_dimm> I have issues with this post http://www.complang.tuwien.ac.at/anton/memory-wall.html
2008-08-07 11:00 <tim_dimm> I think its a few years old
2008-08-07 11:01 <tim_dimm> there's no such thing as an infinitely fast CPU 
2008-08-07 11:52 <flips> mind if I plagiarize your html?
2008-08-07 11:53 <flips> tim_dimm, what about the disk wall?
2008-08-07 11:53 <flips> or more properly, disk chasm
2008-08-07 12:08 <tim_dimm> disk wall?  you mean wall-o-disk ?
2008-08-07 12:09 <tim_dimm> the big leap?  the humongoid gap?
2008-08-07 12:49 <flips> was playing lego indiana jones last night, why I thought about chasms
2008-08-07 13:32 <shapor> why:
2008-08-07 13:33 <shapor> -if (!tuxread(&sb, 6, buf, 11))
2008-08-07 13:33 <shapor> +if (tuxread(&sb, 6, buf, 11))
2008-08-07 13:34 <shapor> wow 711 diggs!
2008-08-07 13:36 <shapor> "But will Tux3 be the ReiserFS killer?"
2008-08-07 13:36 <shapor> "The ReiserFS murderer"
2008-08-07 13:36 <flips> lol
2008-08-07 13:37 <shapor> "Are you saying that Tux3 will be made from the blood, sweat and horror filled tears of the developers Wives..........probably :P"
2008-08-07 13:37 <flips> http://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices <- this is really helping me think about lvm3
2008-08-07 13:37 <flips> url?
2008-08-07 13:37 <shapor> linked from http://shapor.com/tux3/
2008-08-07 13:37 <shapor> the "digg" link at the bottom
2008-08-07 13:38 <flips> ah, I looked at that line lots of times without noticing the digg link
2008-08-07 13:39 <shapor> the reiser jokes seem to keep the conversation going
2008-08-07 13:39 <shapor> 70 comments
2008-08-07 13:39 <flips> http://oss.oracle.com/projects/btrfs/dist/documentation/btrfs-volumes.html <- actually, this is the interesting one
2008-08-07 13:39 <shapor> most of the good ones re "below the threshold"
2008-08-07 13:39 <shapor> heh
2008-08-07 13:44 <flips> wow. css is a completely different syntax than html/xml
2008-08-07 13:48 <shapor> yes it is metadata for the html to use
2008-08-07 13:48 <shapor> the one i put up is in no way minimalistic
2008-08-07 13:49 <shapor> it is originally from doxygen
2008-08-07 13:49 <shapor> so it has lots of crap in it
2008-08-07 13:49 <shapor> could be trimmed down to 10 lines to do the same thing
2008-08-07 13:54 <flips> mistake in your html
2008-08-07 13:55 <flips> </td><tr></table>
2008-08-07 13:55 <flips> looking forward to the 10 line version
2008-08-07 13:56 <shapor> i wasn't planning on making it 10 lines
2008-08-07 13:57 <shapor> its also probably not w3c compliant
2008-08-07 13:58 <shapor> http://validator.w3.org/check?uri=http%3A%2F%2Fshapor.com%2Ftux3
2008-08-07 14:01 <shapor> flips: i just quickly pulled all the css out for classes
2008-08-07 14:01 <shapor> since i dont set any, they never get used
2008-08-07 14:01 <shapor> so it is equivalent
2008-08-07 14:09 <flips> created a mercurial tree for tux3 kernel
2008-08-07 14:17 <flips> linus knocked us off the top of the lkml.org hot list with his 2.6.27-rc2 announcement
2008-08-07 14:17 <flips> now #2
2008-08-07 14:33 <shapor> heh
2008-08-07 15:17 <tim_dimm> only 
2008-08-07 15:17 <tim_dimm> only #?
2008-08-07 15:17 <tim_dimm> wtf good is that?
2008-08-07 15:17 <tim_dimm> ;-)
2008-08-07 15:49 <shapor> flips: bitbucket has a wiki feature, its a nicer way of storing notes/links than editing html
2008-08-07 15:49 <shapor> best part is, its all tracked in a mercurial repo
2008-08-07 15:49 <shapor> http://www.bitbucket.org/shapor/tux3/wiki/OtherFilesystems
2008-08-07 15:50 <flips> where can I download the source?
2008-08-07 15:50 <shapor> http://www.wikicreole.org/ i think
2008-08-07 15:51 <shapor> although i'm not sure what they are using to integrate with hg
2008-08-07 15:54 <flips> sudo apt-get install ikiwiki
2008-08-07 15:55 <flips> I like this: http://www.selenic.com/mercurial/wiki/
2008-08-07 15:55 <shapor> oh neat
2008-08-07 15:55 <shapor> yeah thats better than using some 3rd party thing
2008-08-07 15:56 <shapor> little more work to setup though 
2008-08-07 15:56 <flips> no doubt
2008-08-07 15:56 <flips> there was also one with an issue tracker somewhere
2008-08-07 15:57 <shapor> people seem to be using rcs as a data backend for more and more things these days
2008-08-07 15:57 <shapor> it just makes sense
2008-08-07 15:58 <shapor> however it is quite slow
2008-08-07 15:58 <flips> http://moinmoin.wikiwikiweb.de/MoinMoin 500 - Internal Server Error
2008-08-07 16:05 <flips> http://www.mantisbt.org/bugs/view_all_bug_page.php
2008-08-07 16:10 <flips> http://moinmo.in/ <- works
2008-08-07 16:13 <flips> tim_dimm, want to try the dcc again?
2008-08-07 16:13 <flips> I keep failing to notice the popup
2008-08-07 16:13 <flips> not that xchat dcc has ever worked for me
2008-08-07 16:13 <tim_dimm> hm
2008-08-07 16:13 <tim_dimm> call me then
2008-08-07 16:13 <flips> something abourt nat and xchat being lame
2008-08-07 16:57 -!- MaZe(~MaZe@64.173.151.3) has joined #tux3
2008-08-07 16:59 <tim_dimm> http://www.physorg.com/news137325794.html
2008-08-07 17:13 <flips> hey maze
2008-08-07 17:13 <flips> ok, let's consider a serious issue: how to get people to send us beer
2008-08-07 17:13 <flips> I'll put a link on the tux3.org page when we figure out what the link should say
2008-08-07 17:17 <flips> http://tux3.org/ <- send beer here
2008-08-07 17:33 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3
2008-08-07 17:52 <flips> struct btree_ops {
2008-08-07 17:52 <flips> 	int (*leaf_verify)(SB, void *leaf);
2008-08-07 17:52 <flips> };
2008-08-07 17:52 <flips> struct btree_ops ftree_ops = {
2008-08-07 17:52 <flips> 	.leaf_verify = leaf_verify,
2008-08-07 17:52 <flips> };
2008-08-07 17:53 <flips> btree.c on its way to crappy-c-style object oriented
2008-08-07 18:06 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3
2008-08-07 18:37 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3
2008-08-07 19:13 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3
2008-08-07 21:44 <flips> shapor, a starting point: http://tux3.org/design.html
2008-08-07 22:11 <shapor> oh thats nice
2008-08-07 22:12 <shapor> i'm going to be pretty busy the next couple days, probably wont get a chance to do much until sunday
2008-08-07 22:13 <shapor> docs should go in vcs too though
2008-08-07 22:15 <shapor> flips: http://lwn.net/Articles/234441/
2008-08-07 22:16 <shapor> mentions "versioned pointers"
2008-08-07 22:16 <flips> only to the root inode
2008-08-07 22:16 <flips> that was one of the few pre-existing hits on versioned pointers before my post
2008-08-07 22:17 <flips> now 10 out of first 10 google hits point at me and/or tux3
2008-08-07 22:18 <flips> hit number 22 is finally on some msft concept
2008-08-07 22:19 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3
2008-08-07 22:20 <flips> there are a lot of similarities coming from the logfs people
2008-08-07 22:20 <flips> but basically they are another tree versioned concept
2008-08-07 22:20 <flips> hamer is closer
2008-08-07 22:20 <flips> hammer
2008-08-07 22:20 <flips> see the beer link yet?
2008-08-07 22:20 <tim_dimm> i did
2008-08-07 22:20 <flips> so far no beer
2008-08-07 22:21 <tim_dimm> heheh
2008-08-07 22:21 <tim_dimm> nice
2008-08-07 22:21 <shapor> yeah i saw it too ;)
2008-08-07 22:21 <shapor> flips: only because you can't email beer ;)
2008-08-07 22:21 <tim_dimm> I don't have any beer here, only...
2008-08-07 22:21 <shapor> Results 1 - 10 of about 308,000 for tux3.
2008-08-07 22:21 <flips> if only you could
2008-08-07 22:21 <shapor> flips: i gave you one the other day!!
2008-08-07 22:21 <flips> I'm still subsisting on ramback beer
2008-08-07 22:21 <shapor> so i'm the first beer contributor
2008-08-07 22:21 <shapor> :-)
2008-08-07 22:21 <flips> true
2008-08-07 22:22 <tim_dimm> and what does that make me?
2008-08-07 22:22 <shapor> although i think that may have hindered more than it helped ;)
2008-08-07 22:22 <flips> a contributor
2008-08-07 22:22 <flips> to juvenile delinquency
2008-08-07 22:22 <flips> I was pretty useless that night
2008-08-07 22:22 <flips> crashed just in front of my door
2008-08-07 22:22 <shapor> good beer :-D
2008-08-07 22:22 <flips> hung out on 3rd st on the way back
2008-08-07 22:22 <flips> that was great
2008-08-07 22:22 <shapor> hehe
2008-08-07 22:23 <tim_dimm> shapor, you caused me to almost break my wrist. you know that, right/
2008-08-07 22:23 <tim_dimm> ?
2008-08-07 22:23 <shapor> 
ACTION is a bad influence

2008-08-07 22:23 <tim_dimm> showing off for drunks
2008-08-07 22:23 <tim_dimm> that's not right
2008-08-07 22:23 <flips> juvenile delinquency is a bad influence on us respectable citizens
2008-08-07 22:25 <shapor> "It's good that we have projects like this I think. Sun and ZFS really need some competition, and maybe soon we'll have no need to envy ZFS."
2008-08-07 22:25 <tim_dimm> on the zfs list?
2008-08-07 22:25 <flips> "do you have ZFS envy" -> tux3.org poll question
2008-08-07 22:26 <tim_dimm> nice
2008-08-07 22:26 <tim_dimm> nice hook
2008-08-07 22:26 <flips> tuxopen(&sb, 123, buf, sizeof(buf), 0);
2008-08-07 22:27 <flips> Segmentation fault
2008-08-07 22:28 <flips> left out the init_buffers and tux3_init
2008-08-07 22:28 <flips> new node
2008-08-07 22:28 <flips> new leaf blocksize = 4096
2008-08-07 22:28 <flips> root at 1
2008-08-07 22:28 <flips> leaf at 2
2008-08-07 22:28 <flips> Thu Aug  7 22:28:01 2008: [10463] probe: Failed assertion "(ops->leaf_sniff)(sb, buffer->data)"
2008-08-07 22:28 <tim_dimm> http://www.scdsource.com/article.php?id=296
2008-08-07 22:28 <flips> Trace/breakpoint trap
2008-08-07 22:29 <tim_dimm> Looks like sicortex interconnect design
2008-08-07 22:29 <flips> somebody should push forward the i/o pin tech
2008-08-07 22:30 <flips> would not be that hard
2008-08-07 22:30 <tim_dimm> 3D stacking for pins
2008-08-07 22:30 <flips> micro pins
2008-08-07 22:30 <flips> get way more that way
2008-08-07 22:30 <tim_dimm> actually, 3D switches with pins on the ends of the switches
2008-08-07 22:31 <tim_dimm> pin->switch->pin
2008-08-07 22:31 <flips> you know pins are currently .1 mm right? how stupid is that?
2008-08-07 22:31 <tim_dimm> ball pens
2008-08-07 22:31 <flips> they could easily be 1 um
2008-08-07 22:31 <tim_dimm> yeah, its a problem for the memory channel too
2008-08-07 22:31 <tim_dimm> problem isn't so much the pin, its the link
2008-08-07 22:31 <tim_dimm> on the mobo
2008-08-07 22:31 <flips> problem for most packages
2008-08-07 22:32 <shapor> tim_dimm: from http://www.linux.com/feed/142781 comments
2008-08-07 22:32 <flips> idiotic that it hasn't been fixed reasonably
2008-08-07 22:32 <tim_dimm> nano-technology to the rescue
2008-08-07 22:33 <tim_dimm> "slew-rate requirements"
2008-08-07 22:33 <tim_dimm> data skew
2008-08-07 22:34 <shapor> flips: wheres tuxopen ?
2008-08-07 22:34 <flips> not checked in yet
2008-08-07 22:34 <flips> pukes
2008-08-07 22:34 <shapor> ah thats why its hard to find
2008-08-07 22:34 <flips> coming in about .5 hr, depending on whether I get into another ramback beer or not
2008-08-07 22:35 <shapor> will that make it take more or less time? ;)
2008-08-07 22:35 <flips> more
2008-08-07 22:35 <flips> you found the fleaf bug yet?
2008-08-07 22:36 <shapor> i haven't had a chance to look yet
2008-08-07 22:36 <shapor> busy with work all day
2008-08-07 22:36 <flips> that's sick
2008-08-07 22:36 <shapor> well some of us have day jobs!
2008-08-07 22:36 <shapor> :P
2008-08-07 22:40 <flips> new leaf blocksize = 4096
2008-08-07 22:40 <flips> root at 1
2008-08-07 22:40 <flips> leaf at 2
2008-08-07 22:40 <flips> inode at 123
2008-08-07 22:40 <flips> Thu Aug  7 22:39:52 2008: [10543] ileaf_lookup: Failed assertion "at < leaf->count"
2008-08-07 22:43 <tim_dimm> pardon my noobness- what does assert do?
2008-08-07 22:45 <shapor> similar to what it means in english
2008-08-07 22:45 <shapor> "abort if this is not true"
2008-08-07 22:46 <tim_dimm> just figured that out
2008-08-07 22:46 <tim_dimm> http://www.cplusplus.com/reference/clibrary/cassert/assert.html
2008-08-07 22:46 <tim_dimm> me learn to use the google
2008-08-07 22:46 <tim_dimm> teh google
2008-08-07 22:46 <shapor> yeah, the correct answer is http://www.justfuckinggoogleit.com/
2008-08-07 22:47 <shapor> but its not very nice to send that link
2008-08-07 22:47 <shapor> :P
2008-08-07 22:51 <flips> teh goggles
2008-08-07 22:53 <tim_dimm> beer goggles'
2008-08-07 22:55 <flips> lookup inode 123, 0 + 123
2008-08-07 22:55 <flips> 0 inodes, 4084 free:
2008-08-07 22:55 <flips> release buffer for 1
2008-08-07 22:55 <flips> release buffer for 2
2008-08-07 22:55 <flips> Thu Aug  7 22:54:52 2008: [10710] tuxopen: no inode 123
2008-08-07 22:55 <flips> good
2008-08-07 22:55 <flips> now to set the create flag and create the inode
2008-08-07 22:58 <tim_dimm> I know you guys are going to be at it until the wee hrs
2008-08-07 22:58 <flips> whowever wrote the ddsnap warn macro needs to be hurt
2008-08-07 22:58 <flips> not me
2008-08-07 22:58 <tim_dimm> wasn't I
2008-08-07 22:58 <flips> gotta recover
2008-08-07 22:58 <tim_dimm> I've been getting up at 6am for contractors all week
2008-08-07 23:00 <tim_dimm> <-crashing
2008-08-07 23:00 <tim_dimm> ttyl
2008-08-07 23:02 <flips> creating the inode requires abstracting add_extent_to_tree to also be able to add an inode
2008-08-07 23:03 <flips> starting with renaming as add_entity_to_tree
2008-08-07 23:33 <shapor> flips: not my bug :)
2008-08-07 23:34 <shapor> 
ACTION does a happy dance

2008-08-07 23:36 <flips> :-)
2008-08-07 23:36 <flips> mine?
2008-08-07 23:39 <shapor> actually, maybe mine
2008-08-07 23:39 <shapor> you've got mail :)
2008-08-07 23:40 <shapor> that was about 30 minutes of debugging
2008-08-07 23:41 <shapor> most of which was just isolating the issue
2008-08-07 23:42 <shapor> we should put a test in the main which tests all the boundary conditions
2008-08-07 23:42 <shapor> good call on grouplim = 7
2008-08-07 23:42 <shapor> would have taken us a lot longer to hit this otherwise
2008-08-07 23:43 <flips> yes
2008-08-07 23:43 <flips> use the same trick as I did in btree.c, redefine main as notmain, include in another file and go crazy
2008-08-07 23:44 <flips> there's a lot to be said for having a very simple single file sanity test as now, at this point
2008-08-07 23:44 <shapor> when i have more time
2008-08-07 23:44 <shapor> i can rest now knowing that one is fixed
2008-08-07 23:44 <flips> there's everntually going to be a generic fuzztester for all leaf methods
2008-08-07 23:44 <shapor> i dont want to discover 5 more which require 30 min of debugging each ;)
2008-08-07 23:45 <flips> but knowledge of the specific boundary condiditions means you can write a better single purpose test
2008-08-07 23:45 <shapor> not tonight anyway
2008-08-07 23:46 <flips> nice fix
2008-08-07 23:46 <shapor> did that used to read entry == entries
2008-08-07 23:47 <shapor> i think i might have changed it to !group->count when i added splitting
2008-08-07 23:47 <shapor> i remember poking it before, might have just been in testing
2008-08-07 23:48 <shapor> how do i go back to an old rev in hg
2008-08-07 23:48 <flips> test now succeeds indeed
2008-08-07 23:48 <flips> hg co <rev>
2008-08-07 23:49 <shapor> woo
2008-08-07 23:49 <flips> 		unsigned limit = !group->count || entry == entries ? 0 : (entry + 1)->limit;
2008-08-07 23:49 <flips> 		*entry = (struct entry){ .loglo = loglo, .limit = limit };
2008-08-07 23:58 <shapor> yeah, see anything wrong?
2008-08-07 23:58 <shapor> lgtm
2008-08-07 23:59 <shapor> if you're worried about limit=0, don't be, you go through and increment all the limits by 1 starting at entry*