2008-09-05 00:07 -!- stargazr5(~gaurav@59.95.6.25) has joined #tux3 2008-09-05 00:11 -!- cdk(~chinmay@59.95.14.95) has joined #tux3 2008-09-05 00:15 got to consider what to hack next 2008-09-05 00:15 I'll sleep on it 2008-09-05 00:17 wow, yummy fuse code from conrad 2008-09-05 00:17 unyum 2008-09-05 00:17 :D 2008-09-05 00:19 @konrad:: trying to compile ur fuse file....getting errors.. 2008-09-05 00:19 line 256 2008-09-05 00:19 hm? 2008-09-05 00:19 error? 2008-09-05 00:19 unknown filed 'key' 2008-09-05 00:19 specified in initializer 2008-09-05 00:20 is your checkout of tux3 up to date? 2008-09-05 00:20 ok....i thought so....one moment will get back 2008-09-05 00:21 installing fuse now 2008-09-05 00:22 enjoy the pain 2008-09-05 00:22 lots of segfaults 2008-09-05 00:22 and no proper readdir 2008-09-05 00:22 I'm not sure how it's done, so I just ignored it completely 2008-09-05 00:25 tux3fuse.c:113:47: error: macro "fuse_main" passed 4 arguments, but takes just 3 2008-09-05 00:25 eh, you must have a different version of fuse than me 2008-09-05 00:25 the example on the fuse site shows 3 args 2008-09-05 00:25 but my version takes 4 2008-09-05 00:25 I have 2.7.3 2008-09-05 00:27 just get rid of the NULL argument 2008-09-05 00:29 fuse: failed to exec fusermount: No such file or directory 2008-09-05 00:30 the mountpoint or the fake filesystem? 2008-09-05 00:31 I think there is no fusermount 2008-09-05 00:31 ah, weird 2008-09-05 00:31 what distro? 2008-09-05 00:32 or I don't have fuse compiled into my kernel 2008-09-05 00:32 debian etch 2008-09-05 00:32 I'll run it on something else 2008-09-05 00:32 hm 2008-09-05 00:32 I'm on Fedora 9 2008-09-05 00:32 need to install fuse-utils I think 2008-09-05 00:32 ah 2008-09-05 00:33 got the latest version .... still cant compile :: undefined reference to `btree_delete' 2008-09-05 00:33 are you compiling it correctly? 2008-09-05 00:33 cdk, change that to tree_chop 2008-09-05 00:33 oh, did it change? 2008-09-05 00:33 k 2008-09-05 00:33 it did 2008-09-05 00:33 gratuitous 2008-09-05 00:33 heh, I guess I'm out of date :D 2008-09-05 00:33 going to have to be more careful about that now 2008-09-05 00:34 you might want to post a rebased version 2008-09-05 00:34 just for now until its merged 2008-09-05 00:34 I'll do that tomorrow, got to sleep now 2008-09-05 00:34 fusermount: failed to open /dev/fuse: No such file or directory 2008-09-05 00:34 I suppose I have to make the devnode 2008-09-05 00:34 is the fuse module loaded into your kernel? 2008-09-05 00:35 or somit 2008-09-05 00:35 no 2008-09-05 00:35 I'm not totally familiar with fuse 2008-09-05 00:35 but I need to make the devnode anyway 2008-09-05 00:35 I think 2008-09-05 00:35 ok compiled 2008-09-05 00:35 cdk is going to beat me ;-) 2008-09-05 00:35 heh 2008-09-05 00:36 sudo mknod -m 666 /dev/fuse c 10 229 2008-09-05 00:37 fusermount: fuse device not found, try 'modprobe fuse' first <- maybe as far as I get tonight 2008-09-05 00:37 yes....mounted and visible.. 2008-09-05 00:37 :) 2008-09-05 00:37 :) 2008-09-05 00:38 and I'm sure, very breakable 2008-09-05 00:38 cdk: now count the seconds until segfault 2008-09-05 00:38 :D 2008-09-05 00:38 we'll fix that 2008-09-05 00:38 :D 2008-09-05 00:39 compiling a fuse module 2008-09-05 00:39 stupid make decided to recompile the whole kernel 2008-09-05 00:39 oh no 2008-09-05 00:40 actually it did the right thing 2008-09-05 00:40 but I compiled the wrong thing 2008-09-05 00:40 configfs :p 2008-09-05 00:41 eh 2008-09-05 00:41 heh 2008-09-05 00:41 ok....loop while deleting file.. 2008-09-05 00:41 found it 2008-09-05 00:41 "filesystem in userspace support" 2008-09-05 00:42 cdk, loop? 2008-09-05 00:42 have more that one file on the fs....after fuse shows only one file hello??? 2008-09-05 00:42 cdk: yes. 2008-09-05 00:42 readdir doesn't work 2008-09-05 00:42 so remember the name you used before and cat the file 2008-09-05 00:42 that much should work 2008-09-05 00:42 mounted 2008-09-05 00:42 k 2008-09-05 00:43 readdir will work tomorrow morning ;-) 2008-09-05 00:43 need to sleep now 2008-09-05 00:43 there is the hello file 2008-09-05 00:43 how did it get there? 2008-09-05 00:43 where are u guys...its 1:30 afternoon here 2008-09-05 00:43 flips: look at tux3_readdir() 2008-09-05 00:43 it's static 2008-09-05 00:44 cdk: Pacific time, west coast USA 2008-09-05 00:44 in india 2008-09-05 00:44 konrad, nice 2008-09-05 00:44 I'll make it real pretty soon 2008-09-05 00:44 excellent 2008-09-05 00:45 konrad, this is a most pleasant development 2008-09-05 00:45 good 2008-09-05 00:45 cdk, maybe I'll drop by one of these days ;-) 2008-09-05 00:45 ofcourse 2008-09-05 00:45 always welcome 2008-09-05 00:46 and you're good enough to get fuse working faster than me ;-) 2008-09-05 00:46 course I'm kind of lame at stuff like that 2008-09-05 00:46 ;-) 2008-09-05 00:46 shapor is missing the fun 2008-09-05 00:46 yeah only good at making stuff work....no good at devel 2008-09-05 00:46 cdk, we can fix that 2008-09-05 00:46 hi flips 2008-09-05 00:47 everybody's running tux3-fuse while you're... 2008-09-05 00:47 doing something ;-) 2008-09-05 00:47 having a life maybe 2008-09-05 00:47 yeah, a bit of one anyway lol 2008-09-05 00:47 sh-3.1# echo hello world >foo/foo 2008-09-05 00:47 sh-3.1# cat foo/foo 2008-09-05 00:47 hello world 2008-09-05 00:47 sh-3.1# 2008-09-05 00:48 amazing 2008-09-05 00:48 :D 2008-09-05 00:48 it's mountable 2008-09-05 00:48 it is 2008-09-05 00:48 developers should come pouring in now 2008-09-05 00:48 tomorrow morning readdir will work 2008-09-05 00:48 yah 2008-09-05 00:48 I need to prepare for this by sleeping now ;-) 2008-09-05 00:49 oh 2008-09-05 00:49 I'll check it in first 2008-09-05 00:49 rebased and all? 2008-09-05 00:49 yes, that was easy 2008-09-05 00:49 k, good 2008-09-05 00:49 saves me 30 seconds tomorrow morning 2008-09-05 00:50 konrad: great work on getting tux3 up in fuse ;) 2008-09-05 00:50 ok....flips u beat me to cat... 2008-09-05 00:51 i still cant do that 2008-09-05 00:51 thanks shapor 2008-09-05 00:51 heh 2008-09-05 00:51 I caught up 2008-09-05 00:51 tux3fuse.c, ok konrad? 2008-09-05 00:51 and have it in the same directory for now 2008-09-05 00:51 sure 2008-09-05 00:51 its great to see it looking like a filesystem before the kernel port 2008-09-05 00:52 really 2008-09-05 00:52 will make testing quite scriptable 2008-09-05 00:52 when can we expect the kernel port? 2008-09-05 00:52 and tux3 university starts soon 2008-09-05 00:52 heh, tomorrow night, and if it doesn't happen, blame flips 2008-09-05 00:52 sounds good 2008-09-05 00:52 :) 2008-09-05 00:53 a few of my friends are also interested in tux3 might get them together and get some devel work done. 2008-09-05 00:55 ok...cat working now.. 2008-09-05 00:55 @konrad:: great works.... 2008-09-05 00:56 great work 2008-09-05 00:57 konrad, want your email in the commit message or not? 2008-09-05 00:57 hm? 2008-09-05 00:58 Port of tux3 to fuse, contributed by Conrad Meyer <- for example 2008-09-05 00:58 oh, sure 2008-09-05 00:58 it's in 2008-09-05 00:58 no more version skew ;-) 2008-09-05 00:59 yay 2008-09-05 00:59 added it to make? 2008-09-05 00:59 nope ;-) 2008-09-05 01:00 next 2008-09-05 01:02 tux3 is not in make either... 2008-09-05 01:03 some lamer didn't put it in I guess 2008-09-05 01:03 heh 2008-09-05 01:03 night night 2008-09-05 01:03 night 2008-09-05 01:04 i am off as well 2008-09-05 01:04 nice development 2008-09-05 01:04 thanks again konrad 2008-09-05 01:04 np 2008-09-05 01:06 -!- cdk(~chinmay@59.95.14.95) has left #tux3 2008-09-05 01:09 ok, makefile is there 2008-09-05 01:09 test isn't 2008-09-05 01:09 patches gratefully accepted 2008-09-05 01:09 there are some warnings to clear up 2008-09-05 01:21 sleepy time 2008-09-05 01:21 today was fun 2008-09-05 01:34 -!- Danjel(~chatzilla@c-a721e255.1143-1-64736c12.cust.bredbandsbolaget.se) has joined #tux3 2008-09-05 01:52 ok, rm doesn't segfault if the file doesn't exist now 2008-09-05 02:06 -!- MaZe(~MaZe@c-24-6-86-168.hsd1.ca.comcast.net) has joined #tux3 2008-09-05 02:07 flips: heh you make almost the identical change to the Makefile that I did in that commit 2008-09-05 02:07 was only a few characters off, heh 2008-09-05 02:08 hey flips 2008-09-05 02:10 do you think that the ZFS notion of having sha1 hashes all over the place is sufficient to guarantee the integrity of a volume given that you can fix the corruption from redundant copies ? 2008-09-05 02:10 in place of online checking or any kind of checking ? 2008-09-05 02:12 ooh too bad 2008-09-05 02:13 droppoed off 2008-09-05 02:13 -!- flips(~phillips@phunq.net) has joined #tux3 2008-09-05 02:13 dropped off 2008-09-05 02:13 flips: you there ? 2008-09-05 02:21 sha1 in expensive 2008-09-05 02:21 probably needlessly 2008-09-05 02:21 and unless they are real real careful about the order of operations 2008-09-05 02:21 they still can't guarantee memory/cpu/disk channel corruption won't go undetected 2008-09-05 02:22 s/in/is/ 2008-09-05 02:22 [not that you can ever be 100% sure, if your cpu/memory could be bad - unless you're making a net fs] 2008-09-05 02:23 yeah, they could be summing a bad chunk of memory 2008-09-05 02:24 what else ? 2008-09-05 02:24 what about phantom writes ? 2008-09-05 02:46 ok night 2008-09-05 03:48 -!- nobody(c8c32a07@67.207.141.120) has joined #tux3 2008-09-05 04:27 -!- nobody(c8c32a07@67.207.141.120) has left #tux3 2008-09-05 05:06 -!- cdk(~chinmay@59.95.49.53) has joined #tux3 2008-09-05 05:07 -!- cdk(~chinmay@59.95.49.53) has left #tux3 2008-09-05 05:17 -!- stargazr5(~gauravstt@59.95.17.124) has joined #tux3 2008-09-05 07:12 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3 2008-09-05 09:36 -!- data(~data@echo489.server4you.de) has joined #tux3 2008-09-05 10:18 -!- vcgomes(~vcgomes@li17-238.members.linode.com) has joined #tux3 2008-09-05 10:35 -!- stargazr5(~gauravstt@59.95.19.213) has joined #tux3 2008-09-05 10:41 where does fuse log to? 2008-09-05 10:41 funny it doesn't just write to console when running on foreground 2008-09-05 10:58 oh it does 2008-09-05 10:58 silly me 2008-09-05 11:16 whoever came up with the readdir interface needs to be badly hurt 2008-09-05 12:07 hey 2008-09-05 12:07 hi 2008-09-05 12:11 moonbase:/more/src/hg/tux3/user/test# ls foo 2008-09-05 12:11 /bar /bar2 /bar3 /bar4 /bar5 /bar6 /foo /foo2 /foo3 /hello 2008-09-05 12:11 not sure what all the slashes are about 2008-09-05 12:14 the kernel filldir interface is the most amazing stinking pile of poo I have ever seen 2008-09-05 12:14 http://lxr.linux.no/linux+v2.6.26.3/fs/readdir.c#L146 2008-09-05 12:14 next time I'm in public I'll wear a bag over my head that says "no I am not a linux kernel hacker" 2008-09-05 12:15 dang it intel 2008-09-05 12:16 i'm looking for a x48 PCIe chipset, and they named theirs x48 2008-09-05 12:16 but it only supports 38 lanes 2008-09-05 12:16 I know I've seen a x48 lane pcie chipset out there 2008-09-05 12:16 bounders] 2008-09-05 12:17 that's a lot of lanes 2008-09-05 12:35 fuse marches on 2008-09-05 12:35 directory listing works now 2008-09-05 12:35 seems to 2008-09-05 12:36 I don't think it should be printing / for every file though 2008-09-05 12:36 wonder why it does that 2008-09-05 12:36 exercise for konrad ;-) 2008-09-05 12:46 http://hardware.slashdot.org/comments.pl?sid=954803&cid=24889733 2008-09-05 12:47 someone passed on my call for helpers 2008-09-05 12:48 nice 2008-09-05 12:49 btw, I pinged wook for an assist on the visualization we discussed 2008-09-05 12:50 no response yet 2008-09-05 12:57 wooklag 2008-09-05 13:31 wook_lag 2008-09-05 13:31 :-) 2008-09-05 13:57 wow- tux3 got a slashdot lkml bounce 2008-09-05 13:58 ? 2008-09-05 13:58 its #2 on the lkml hottest message list now 2008-09-05 13:59 ah 2008-09-05 13:59 heh 2008-09-05 13:59 speculating that was from the slashdot reference earlier 2008-09-05 13:59 i bet its just flips clicking reload ;) 2008-09-05 14:00 actually probably due to the fact that its been on the front page of kerneltrap for the past couple days 2008-09-05 14:00 well past day 2008-09-05 14:00 could be 2008-09-05 14:38 hullo 2008-09-05 14:42 hi konrad 2008-09-05 14:42 see the extra slashes on the ls output? 2008-09-05 14:42 I did not have time to investigate 2008-09-05 14:42 nope, build error :) 2008-09-05 14:42 tux3fuse.c:251:41: error: macro "fuse_main" requires 4 arguments, but only 3 given 2008-09-05 14:42 fun fun 2008-09-05 14:43 wow 2008-09-05 14:43 back up cheek by jowel with Linus's post 2008-09-05 14:43 lkml? 2008-09-05 14:43 slashdot blowback 2008-09-05 14:43 http://lkml.org/ 2008-09-05 14:44 konrad, I guess you have a different version of fuse 2008-09-05 14:44 too bad they couldn't keep the interface stable 2008-09-05 14:44 just add the NULL back 2008-09-05 14:44 we'll figure out what to do later 2008-09-05 14:45 konrad, anyway it should be tux3fs.c by now 2008-09-05 14:45 right 2008-09-05 14:45 got to go out for a bit 2008-09-05 14:45 heh, typo 2008-09-05 14:45 it's fux3fs.c 2008-09-05 14:47 whoops 2008-09-05 14:47 hm, getattr seems to be failing here 2008-09-05 14:47 yes, getattr is totally not implemented 2008-09-05 14:47 yet 2008-09-05 17:27 -!- tim_dimm_(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3 2008-09-05 17:32 konrad, which version of libfuse2 do you have? 2008-09-05 17:32 2.7.3 2008-09-05 17:32 2.5.3 here 2008-09-05 17:33 it's a shame they couldn't keep the api stable across 2 point releases 2008-09-05 17:33 I guess we need a #if on version 2008-09-05 17:33 hard to see how that will work 2008-09-05 17:34 heh 2008-09-05 17:34 or I upgrade 2008-09-05 17:34 and we define to work only on recent fuse 2008-09-05 17:34 :p 2008-09-05 17:34 if you feel like it 2008-09-05 17:34 I don't 2008-09-05 17:34 it's a backport to etch 2008-09-05 17:34 could just pass a define to gcc or somit 2008-09-05 17:34 hate fiddling with that 2008-09-05 17:35 yes, but the define has to key on something 2008-09-05 17:36 konrad, can you do ls /usr/lib/libfuse* 2008-09-05 17:37 then have stuff like /usr/include/fuse/fuse_compat.h 2008-09-05 17:38 and /usr/include/fuse/fuse_lowlevel_compat.h 2008-09-05 17:38 flips: you going to try and get this running on fuse ? 2008-09-05 17:38 bh, it's already running on fuse 2008-09-05 17:38 ok 2008-09-05 17:38 konrad did it 2008-09-05 17:38 nice 2008-09-05 17:38 when did konrad come on board ? 2008-09-05 17:38 now debugging the bugs 2008-09-05 17:38 nice 2008-09-05 17:38 earlier this week 2008-09-05 17:38 nice 2008-09-05 17:38 I think 2008-09-05 17:38 maybe last week 2008-09-05 17:38 I have /usr/include/fuse/fuse_compat.h 2008-09-05 17:39 but no /usr/lib/libfuse 2008-09-05 17:39 ah, redhat puts them somewhere else I suppose 2008-09-05 17:39 ok, out 2008-09-05 17:39 for no good reason 2008-09-05 17:39 well, and 64-bit 2008-09-05 17:39 /usr/lib64 2008-09-05 17:39 ah 2008-09-05 17:39 and what's there? 2008-09-05 17:40 nothing 2008-09-05 17:40 libfuse2 it should be 2008-09-05 17:40 ah 2008-09-05 17:40 /lib64/libfuse.so 2008-09-05 17:40 should be a bunch more 2008-09-05 17:41 ls /usr/lib64/libfuse.so* 2008-09-05 17:41 ls /usr/lib64/libfuse*.so* 2008-09-05 17:42 nope 2008-09-05 17:42 rpm -ql fuse-devel 2008-09-05 17:42 ah, I dimly recall rpm -ql 2008-09-05 17:42 /lib64/libfuse.so 2008-09-05 17:42 /lib64/libulockmgr.so 2008-09-05 17:43 that's it 2008-09-05 17:43 for .so's 2008-09-05 17:43 packaged quite differently 2008-09-05 17:43 and incompatibly 2008-09-05 17:43 I suspect the problem isn't the fuse guys 2008-09-05 17:43 but "creative" redhatters 2008-09-05 17:45 #define FUSE_USE_VERSION 26 <- aha 2008-09-05 17:45 that must have something to do with it 2008-09-05 17:46 ah, does 2.5.3 do version 26? 2008-09-05 17:46 read your fuse/fuse.h 2008-09-05 17:49 #ifndef FUSE_USE_VERSION 2008-09-05 17:49 #define FUSE_USE_VERSION 21 2008-09-05 17:49 #endif <- then there are lots of function mismatches and no .readdir 2008-09-05 17:49 hm 2008-09-05 17:50 #define fuse_main(argc, argv, op) \ 2008-09-05 17:50 fuse_main_real(argc, argv, op, sizeof(*(op))) 2008-09-05 17:50 what have you got there? 2008-09-05 17:51 477 #define fuse_main(argc, argv, op, user_data) \ 2008-09-05 17:51 478 fuse_main_real(argc, argv, op, sizeof(*(op)), user_data) 2008-09-05 17:51 idiotic api breakage 2008-09-05 17:52 :D 2008-09-05 17:52 well we can grep fuse.h for user_data 2008-09-05 17:52 in the makefile 2008-09-05 17:52 and write nasty notes there 2008-09-05 17:52 or we can ask on their mailing list 2008-09-05 17:53 better style 2008-09-05 17:53 there's probably an irc channel on this server 2008-09-05 17:53 I think I tried actually 2008-09-05 17:53 here and freenode 2008-09-05 17:53 nope 2008-09-05 17:54 http://lists.sourceforge.net/lists/listinfo/fuse-devel 2008-09-05 17:55 konrad, how would you like to be our representative on the fuse-devel list? 2008-09-05 17:55 :-) 2008-09-05 17:55 joining 2008-09-05 17:56 got an idea how to frame the question? 2008-09-05 17:56 "Why did you break the %#& API?" 2008-09-05 17:56 :-) 2008-09-05 17:56 excellent way to get silence 2008-09-05 17:57 right 2008-09-05 17:57 "what is the expected workaround for the extra user_data parameter added to fuse_main? 2008-09-05 18:02 konrad, you know what I'm going to do? 2008-09-05 18:03 hack my fuse.h 2008-09-05 18:03 and add an extra parameter that is ignored 2008-09-05 18:03 k 2008-09-05 18:03 I'll rev tux3 pretty soon to user the 4 parameter flavor of fuse_main 2008-09-05 18:03 skate first 2008-09-05 18:03 enjoy :) 2008-09-05 18:56 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3 2008-09-05 20:37 konrad, fixed the rename damage and adapted to the 4 arg from or fuse_main 2008-09-05 22:34 -!- tim_dimm(~timothyhu@cpe-76-90-98-247.socal.res.rr.com) has joined #tux3