irc.oftc.net #zumastor log beginning Sat Sep 1 00:00:01 PDT 2007 irc.oftc.net #zumastor log beginning Sun Sep 2 00:00:01 PDT 2007 2007-09-02 00:58 -!- erwan_taf(~erwan@konilope.linuxeries.org) has joined #zumastor 2007-09-02 01:35 -!- erwan__taf(~erwan@konilope.linuxeries.org) has joined #zumastor irc.oftc.net #zumastor log beginning Mon Sep 3 00:00:02 PDT 2007 2007-09-03 00:48 -!- erwan_taf(~erwan@LAubervilliers-151-13-63-69.w217-128.abo.wanadoo.fr) has joined #zumastor 2007-09-03 00:55 -!- erwan__taf(~erwan@visage.seanodes.com) has joined #zumastor irc.oftc.net #zumastor log beginning Tue Sep 4 00:00:01 PDT 2007 2007-09-04 10:52 jiangz: ~phillips/bio.throttle-2.6.22 2007-09-04 10:53 murb, there? 2007-09-04 10:53 the iram is a consumer product, it falls a little short of server grade, otherwise yes 2007-09-04 16:45 flips? 2007-09-04 16:49 hi jiayingz 2007-09-04 16:49 I just looked at the bio patch again and got a question about bio_queue 2007-09-04 16:50 bi_queue? 2007-09-04 16:50 why do we need it? 2007-09-04 16:50 yes 2007-09-04 16:50 so we can release the throttle in bio->endio 2007-09-04 16:50 need some way of finding the counter and wait queue 2007-09-04 16:51 can we just call bdev_get_queue? 2007-09-04 16:51 the bdev may have changed by this time 2007-09-04 16:51 if the bio is resubmitted to a lower level device 2007-09-04 16:51 hmm, i c 2007-09-04 16:52 ok. I will get the new patch irc.oftc.net #zumastor log beginning Wed Sep 5 00:00:01 PDT 2007 2007-09-05 11:57 hi flipz 2007-09-05 11:58 the sysrq trace is sysrq-fsync-hang under my home dir 2007-09-05 12:19 -!- erwan_taf(~erwan@konilope.linuxeries.org) has joined #zumastor 2007-09-05 16:12 thanks much jiayingz irc.oftc.net #zumastor log beginning Thu Sep 6 00:00:02 PDT 2007 2007-09-06 01:07 -!- erwan_taf(~erwan@81.80.43.67) has joined #zumastor 2007-09-06 09:45 -!- cbsmith(~user@adsl-75-56-197-116.dsl.lsan03.sbcglobal.net) has joined #zumastor 2007-09-06 09:48 ACTION waves at everyone 2007-09-06 09:48 hi chris 2007-09-06 09:49 Looks like UUASC has a storage software vendor coming to speak with them next week. 2007-09-06 09:49 BakBone? You heard of them? 2007-09-06 09:49 UUASC? 2007-09-06 09:49 Unix Users Association of Southern California 2007-09-06 09:49 have not heard of bakbone 2007-09-06 09:49 Looks like they have a remote snapshotting product 2007-09-06 09:50 unless they have a satellite link to the cruise ship I'm out of luck next week 2007-09-06 09:50 http://www.bakbone.com/ 2007-09-06 09:50 Cruise ship? 2007-09-06 09:51 taking me dad on a cruise next week 2007-09-06 09:52 so far I have not learned whether bakbone is a fiber channel thing, iscsi or what 2007-09-06 09:54 yeah, that was not immediately clear to me either. 2007-09-06 09:54 It looks like they are software that sit above all that. 2007-09-06 09:54 since they refer to specific kernels, they may even be a block device 2007-09-06 09:55 anyway, they seem to know the territory 2007-09-06 09:56 yup 2007-09-06 09:56 are you going to drop in? 2007-09-06 09:56 Might be worth sending someone down to Irvine just to see what their solution looks like. 2007-09-06 09:56 flipz: I normally don't go to the OC meetings, but maybe I should. 2007-09-06 10:11 ACTION reading the Joel Report 2007-09-06 10:12 Joel report? 2007-09-06 10:12 oh 2007-09-06 10:12 just a sec 2007-09-06 10:15 wow, he is really hitting us hard 2007-09-06 10:15 yup 2007-09-06 10:15 ACTION heads into the office 2007-09-06 10:15 we mostly survived.... although there were those nasty kernel problems and the corruption of the root partition. 2007-09-06 10:15 ACTION salutes flipz 2007-09-06 17:26 Hey guys. Regarding code coverage. Right now I'm compiling -pg versions of the code in to a different target directory. This is a real pain with our current Makefile structure (doesn't use rules for starters, so transforming each target is a pain). Thoughts on a better approach? 2007-09-06 17:27 I could use an include file which enhances the normal $cflags variable, but that seems at least equally hackish. 2007-09-06 17:31 You are all at TGAIF aren't ya? ;-) 2007-09-06 17:38 Yep. :P 2007-09-06 17:43 nice 2007-09-06 17:47 Makes me wish for the BSD makefile structure. Just add CFLAGS+=-pg at the top level. 2007-09-06 18:27 fmayhar: Yes, that was kind of what I was hoping to do, but apparently Daniel P doesn't like using suffix rules and prefers having explicit rules. 2007-09-06 18:28 That is, explicit rules for each target. irc.oftc.net #zumastor log beginning Fri Sep 7 00:00:03 PDT 2007 2007-09-07 07:40 -!- erwan_taf(~erwan@81.80.43.67) has joined #zumastor 2007-09-07 11:00 -!- erwan_taf(~erwan@konilope.linuxeries.org) has joined #zumastor irc.oftc.net #zumastor log beginning Sat Sep 8 00:00:02 PDT 2007 2007-09-08 02:03 -!- Prens(~ss@85.100.47.145) has joined #zumastor irc.oftc.net #zumastor log beginning Sun Sep 9 00:00:02 PDT 2007 2007-09-09 11:54 -!- erwan_taf(~erwan@konilope.linuxeries.org) has joined #zumastor irc.oftc.net #zumastor log beginning Mon Sep 10 00:00:02 PDT 2007 2007-09-10 01:13 -!- erwan_taf(~erwan@81.80.43.67) has joined #zumastor 2007-09-10 01:13 lo 2007-09-10 11:24 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-10 12:07 -!- juuva_(juuva@peili.org) has joined #zumastor 2007-09-10 12:07 -!- murb(~murbix@soapstone.yuri.org.uk) has joined #zumastor 2007-09-10 12:07 -!- erwan_taf(~erwan@81.80.43.67) has joined #zumastor 2007-09-10 12:07 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-10 12:07 -!- flips(~phillips@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-10 12:07 -!- shapor(~shapor@yzf.shaptech.com) has joined #zumastor 2007-09-10 14:21 -!- zumalog(~zumalog@yzf.shaptech.com) has joined #zumastor 2007-09-10 14:23 -!- jiayingz(~jiayingz@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-10 14:51 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-10 16:23 -!- fmayhar(~fmayhar@207.47.98.129.static.nextweb.net) has joined #zumastor irc.oftc.net #zumastor log beginning Tue Sep 11 00:00:01 PDT 2007 2007-09-11 06:26 -!- gektop(~chatzilla@219.134.231.69) has joined #zumastor 2007-09-11 13:14 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-11 14:31 -!- flipz(~phillips@phunq.net) has joined #zumastor 2007-09-11 14:35 flipz: Review the ddsnapd monitoring patch! :) irc.oftc.net #zumastor log beginning Wed Sep 12 00:00:02 PDT 2007 2007-09-12 06:21 -!- jdries3(~jdries3@72.14.224.1) has joined #zumastor 2007-09-12 06:23 -!- jdries3(~jdries3@72.14.224.1) has joined #zumastor 2007-09-12 07:45 -!- zumalog(~zumalog@yzf.shaptech.com) has joined #zumastor 2007-09-12 10:20 -!- shapor_(~shapor@yzf.shaptech.com) has joined #zumastor 2007-09-12 12:57 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor irc.oftc.net #zumastor log beginning Thu Sep 13 00:00:03 PDT 2007 2007-09-13 10:18 -!- cbsmith(~user@adsl-75-56-197-116.dsl.lsan03.sbcglobal.net) has joined #zumastor 2007-09-13 11:32 -!- fmayhar(~fmayhar@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-13 11:33 -!- erwan_taf(~erwan@konilope.linuxeries.org) has joined #zumastor 2007-09-13 12:53 So, I'm seeing a bunch of compiler warnings on my system because we play fast and loose with chunk_t's and u_int64_t's. In particularly the "exception" parameter in snapshot_chunk_unique(). Anyone care to explain why we don't just declare exception as of type "chunk_t*"? 2007-09-13 15:12 -!- erwan_taf(~erwan@konilope.linuxeries.org) has joined #zumastor 2007-09-13 15:13 ACTION taps the screen 2007-09-13 16:53 -!- daniel__(~phlipz@216.7.145.250) has joined #zumastor 2007-09-13 16:56 greetings shapor_, fmayhar, cbsmith 2007-09-13 17:23 hi phlyps 2007-09-13 17:23 still on the high seas ? 2007-09-13 17:23 I am 2007-09-13 17:23 and I have 2 hours oif internet connection to use up by the end of Friday 2007-09-13 17:24 we've been having fun with some interested bugs the last couple days 2007-09-13 17:24 interesting* 2007-09-13 17:24 hence just hanging on irc instead of the quick sortees I've been doing so far 2007-09-13 17:24 I'm not cut out for floating around on the sea 2007-09-13 17:24 to boring ;) 2007-09-13 17:24 yeah i never saw the appeal of a cruise 2007-09-13 17:25 I was impressed by massive chunks of ice falling of glaciers into the sea and going KABO"OM 2007-09-13 17:25 somewhat 2007-09-13 17:25 I see this as a floating retirement home 2007-09-13 17:25 with a low bandwidth internet connection <- the worst part 2007-09-13 17:25 costs $.40/minute 2007-09-13 17:54 -!- jiayingz(~jiayingz@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-13 18:04 hi jiayingz 2007-09-13 18:10 hi phlyps 2007-09-13 18:11 looking forward to being back ;) 2007-09-13 18:11 welcome back :) 2007-09-13 18:11 are you going to be back next week 2007-09-13 18:11 monday 2007-09-13 18:11 how is ur trip? 2007-09-13 18:11 I'm about 1,000 miles north of vancouver right now 2007-09-13 18:12 floating on a boat isn't that exciting for me 2007-09-13 18:12 especially with a low bandwidth expensive net connection 2007-09-13 18:12 how long have you been on the boat? 2007-09-13 18:13 I did however resume working out, after 2 years of not 2007-09-13 18:13 so hopefully I will continue 2007-09-13 18:13 pilates 2007-09-13 18:13 good for u 2007-09-13 18:13 the ship is 11 decks high 2007-09-13 18:13 I don't take the elevator ;) 2007-09-13 18:14 so that is the workout you are talking about? 2007-09-13 18:14 that is besides the working out 2007-09-13 18:14 I take pilates classes too 2007-09-13 18:15 i c. otherwise it is hard to continue since we are on the first floor 2007-09-13 18:15 we should introduce pedal powered workstations 2007-09-13 18:15 and we can donate the excess energy to the california power grid, earning carbon credits 2007-09-13 18:16 an additional benefit is, when you stop pedalling, your screen saver starts 2007-09-13 18:16 hmm, I wonder how that conversion can work 2007-09-13 18:17 you mean under each desk? 2007-09-13 18:17 well, maybe we have to replace the desks with exercise bikes 2007-09-13 18:17 just joking :) 2007-09-13 18:17 not a bad idea 2007-09-13 18:17 there are also ballroom dance lessons, which I enjoy 2007-09-13 18:18 all on the boat? 2007-09-13 18:18 yes 2007-09-13 18:18 and there is an amazing dance troupe 2007-09-13 18:18 just know there are so many things you can do on a boat 2007-09-13 18:18 there are enough things to do all right, the problem is, they are not necessarily what I want to do 2007-09-13 18:19 maybe it is time to try them 2007-09-13 18:20 ok, I need to save some internet minutes for tomorrow 2007-09-13 18:20 so I will go away and read some stuff offline 2007-09-13 18:20 see you 2007-09-13 18:20 ok. talk to u tomorrow irc.oftc.net #zumastor log beginning Fri Sep 14 00:00:02 PDT 2007 2007-09-14 04:13 -!- erwan_taf(~erwan@81.80.43.67) has joined #zumastor 2007-09-14 07:47 -!- jdries3(~jdries3@72.14.224.1) has joined #zumastor 2007-09-14 13:32 -!- flipzz(~phlipz@216.7.145.251) has joined #zumastor 2007-09-14 13:33 yay, SCOX files for bankruptcy 2007-09-14 13:33 now lets see if they can hide their ill gotten loot from NOVL and IBM 2007-09-14 15:43 -!- dkegel(~chatzilla@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-14 15:44 testing 2007-09-14 15:44 1..2...3... 2007-09-14 15:50 So, this bit twiddler walks into a bar... 2007-09-14 16:08 Lot of dead air eh? ;-) 2007-09-14 17:00 -!- jdries3_(~jdries3@c-69-249-52-124.hsd1.nj.comcast.net) has joined #zumastor 2007-09-14 17:32 -!- jdries3(~jdries3@72.14.224.1) has joined #zumastor 2007-09-14 18:48 -!- jdries3_(~jdries3@c-69-249-52-124.hsd1.nj.comcast.net) has joined #zumastor 2007-09-14 20:14 -!- daniel__(~phlipz@216.7.145.250) has joined #zumastor 2007-09-14 21:42 -!- daniel__(~phlipz@216.7.145.250) has joined #zumastor irc.oftc.net #zumastor log beginning Sat Sep 15 00:00:02 PDT 2007 irc.oftc.net #zumastor log beginning Sun Sep 16 00:00:04 PDT 2007 2007-09-16 17:09 -!- juuva(juuva@peili.org) has joined #zumastor irc.oftc.net #zumastor log beginning Mon Sep 17 00:00:02 PDT 2007 2007-09-17 11:03 good morning! 2007-09-17 11:18 good evening 2007-09-17 11:33 -!- cbsmith`(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-17 13:30 ACTION reappears 2007-09-17 13:39 ACTION waves to flipz 2007-09-17 13:40 trying to grok the deep meaning of (*pending)->holdcount = 1; 2007-09-17 13:40 desperately needs comments 2007-09-17 13:41 flipz: This is something like the "parent's curse" ;-) 2007-09-17 13:42 I don't know that curse 2007-09-17 13:44 flipz: The parent's curse is essentially: "may you have a child like you someday". ;-) 2007-09-17 13:44 ooh, well my kid is nice so I guess nobody cursed me 2007-09-17 13:45 flipz: It's a curse that genetics ensures has a high probability of happening. 2007-09-17 13:45 flipz: Talk to me again when she's 15. ;-) 2007-09-17 13:45 ACTION marks it in his calendar 2007-09-17 13:47 I'd agree with you though: my son doesn't seem nearly as bad as I was, but there are some traits that are appearing that suggest that my future may closely resemble my parent's past. ;-) 2007-09-17 14:50 flipz: Did you see my follow to your point about snapshot_details? 2007-09-17 14:50 not yet 2007-09-17 14:50 just a moment 2007-09-17 15:01 cbsmith, reply on the way 2007-09-17 15:01 flipz: kk 2007-09-17 15:05 flipz: I think that code is being misused then. 2007-09-17 15:05 flipz: oh wait, it isn't. 2007-09-17 15:05 flipz: dude, that is soooo evil. 2007-09-17 15:06 I know, but the alternative, having a separate size-calculating function is equally evil 2007-09-17 15:06 so I chose the most robust of two evils 2007-09-17 15:06 flipz: Not so bad if the snapshot_details function uses the calculating function. 2007-09-17 15:07 to put this in perspective, I think I will show you the code before it got rewritten this way 2007-09-17 15:07 Also, given that you are using C99's bits for variable length arrays, I believe you could actually use C99's runtime sizeof() and address functions to make things work. 2007-09-17 15:07 hehe 2007-09-17 15:07 try your hand at it, good luck ;) 2007-09-17 15:08 this one isn't broken though 2007-09-17 15:08 flipz: agreed 2007-09-17 15:08 flipz: just evil 2007-09-17 15:08 note that there are two levels of variable sized vector 2007-09-17 15:08 that pretty much throws a spanner into the C mechanism 2007-09-17 15:08 C is just really bad at this 2007-09-17 15:09 C89 is bad at this. C99 is just.... kludgey. 2007-09-17 15:09 python, lisp, others beat the pants off C for this kind of thing 2007-09-17 15:09 flipz: Heck, even C++ is an improvement in this regard. ;-) 2007-09-17 15:09 yes, then we get to enjoy template hell 2007-09-17 15:10 Why is status_reply.details set as being of type "char"? 2007-09-17 15:10 ok, the "right" thing to so is write a wrapper function with "size" in the name, and use that instead of inlining it with the address calculation 2007-09-17 15:10 ok? 2007-09-17 15:10 It would seem that if it was defined as "struct sanpshot_details" all would then be fine. 2007-09-17 15:11 a rational person would think so 2007-09-17 15:11 flipz: yeah, that's what I was thinking, but I think we can get the compiler to do the address computation for us. 2007-09-17 15:11 I did try that ;) 2007-09-17 15:11 flipz: So what am I missing? 2007-09-17 15:11 it doesn't work 2007-09-17 15:11 the details of why are painful 2007-09-17 15:12 it's the two levels of variable sized vector that do in that idea 2007-09-17 15:12 "char" at least sends a clear message that we are going to do our own address message here 2007-09-17 15:13 err, address massaging I meant 2007-09-17 15:13 flipz: Yeah, but as long as the inner array's size is fixed, the C99 VLA support *should* work if we use struct snapshot_details in status_reply. 2007-09-17 15:14 the inner array's size is not fixed 2007-09-17 15:14 ACTION decides to write a test case to see what he's missing 2007-09-17 15:14 oh, and gcc will compile incorrect code for this without complaint 2007-09-17 15:14 flipz: It's not fixed at *compile* time, but surely it is fixed at runtime, else snapshot_details() is broken. 2007-09-17 15:14 flipz: Yeah, I'm not doing a "does it compile" test. I'm doing a "does it compute the right addresses" test. 2007-09-17 15:15 of course, fixed at run time, I think that maps onto the halting problem ;) 2007-09-17 15:15 it compiles and computes the wrong address 2007-09-17 15:15 we've been through this before ;) 2007-09-17 15:15 flipz: fixed at runtime is what C99 VLA's supports. 2007-09-17 15:15 flipz: I don't grok why it would compute the wrong address. Is it because the structs are packed? 2007-09-17 15:15 not in this case though 2007-09-17 15:16 it should <- there's my favorite word again 2007-09-17 15:16 hehe 2007-09-17 15:18 I think you should be able to write struct foo { int num; int vec[num]; int foo; } and the compiler should generate code that reads num in order to calculate the address of foo, but the C committee thought otherwise 2007-09-17 15:18 flipz: yeah, you definitely can't do that. 2007-09-17 15:19 but you can, sort of, in parameter lists 2007-09-17 15:19 so this is just inconsistent 2007-09-17 15:19 flipz: They basically encoded in to the compilers what people were doing withing the confines of C89 anyway. ;-) 2007-09-17 15:20 anyway, don't get too obsessed with this, that way lies madness 2007-09-17 15:21 it is after all, mainly the deficiencies of C that gaurantee lifetime full employment for anybody who can put up with them 2007-09-17 15:22 lol, that's why C++ programmers get paid even more. ;-) 2007-09-17 15:22 this way lies happiness: http://finance.yahoo.com/q?s=scox 2007-09-17 15:23 close at $.23, which is exactly, um, 23 cents more than it is worth 2007-09-17 15:23 still a good short 2007-09-17 15:23 ACTION speculates that their existing business is still probably worth on the order of a million bucks. 2007-09-17 15:23 flipz: but yeah 2007-09-17 15:24 not after subtracting their owables to NOVL 2007-09-17 15:24 flipz: Oh yeah, I keep forgetting about how f'd they are on that count. 2007-09-17 15:24 then there are the IBM counterclaims 2007-09-17 15:25 then there is the "follow the money" part 2007-09-17 15:25 flipz: Those aren't nearly as threatening in my book. They were just their to piss people off. 2007-09-17 15:25 short ratio is 13.3! 2007-09-17 15:27 SIXTH COUNTERCLAIM 2007-09-17 15:27 Breach of the GNU General Public License 2007-09-17 15:27 FIFTH COUNTERCLAIM 2007-09-17 15:27 Unfair and Deceptive Trade Practices 2007-09-17 15:27 FOURTH COUNTERCLAIM 2007-09-17 15:27 2007-09-17 15:27 Intentional Interference with Prospective Economic Relations 2007-09-17 15:27 THIRD COUNTERCLAIM 2007-09-17 15:27 Unfair Competition 2007-09-17 15:28 SECOND COUNTERCLAIM 2007-09-17 15:28 Lanham Act Violation 2007-09-17 15:28 FIRST COUNTERCLAIM 2007-09-17 15:28 Breach of Contract 2007-09-17 15:28 48. SCO has breached its express duties and obligations under the AT&T 2007-09-17 15:28 Agreements and Amendment X by, among other filings, purporting to terminate 2007-09-17 15:28 IBM's irrevocable and perpetual UNIX rights and/or refusing to provide IBM 2007-09-17 15:28 adequate notice and opportunity to cure its alleged misconduct. 2007-09-17 15:29 remember that "irrevocable" thing? 2007-09-17 15:29 if there is an ounce of meat left on the carcass after NOVL is done, the Nazgul will get it 2007-09-17 15:30 Hehe. Yeah, but NOVL is going to strip the bones bare, I suspect. 2007-09-17 15:31 I'd love to see Boise have to pay back some of that dirty money 2007-09-17 15:31 didn't exactly have an arm's length relationship in this 2007-09-17 15:32 I suspect his net karma score is positive 2007-09-17 15:32 Besides, he sucked a ton of money out of them. ;-) 2007-09-17 15:32 not after getting disbarred in florida 2007-09-17 15:32 All the while doing a horrid job representing them. ;-) 2007-09-17 15:32 He got disbarred? 2007-09-17 15:32 yes 2007-09-17 15:33 For what? 2007-09-17 15:33 http://www.law.com/jsp/article.jsp?id=1071351152026 2007-09-17 15:34 has some other shady goings on as well 2007-09-17 15:36 That sounds like kind of a flakey (admittedly incestuous) set of charges. 2007-09-17 15:36 sorry, he seems to have got off in florida 2007-09-17 15:38 yeah, I'm not too surprised after reading the article. 2007-09-17 15:40 ACTION is beginning to understand the address computation problem. 2007-09-17 15:43 "In testimony on March 7, a former supervisor for Habie's company said Habie talked about having Lewis' legs broken. " <- Boies appointed Habie's wife CFO of his law firm 2007-09-17 15:43 just to be clear about what sort of character we have here 2007-09-17 15:44 as far as having him on the side of right in the microsoft case goes, I feel dirty about that 2007-09-17 15:44 flipz: He was also on the "wrong" side of the IBM case. 2007-09-17 15:44 flipz: And Al Gore's 2000 fiasco. 2007-09-17 15:45 funny that 2007-09-17 15:45 ok, time to lkml 2007-09-17 15:45 So the chief problem I'm seing with the sizeof() problem here comes from the fact that you can't specify the size of the snapinfo array when trying to get the address of a row in the status_reply array. Does that sound right? 2007-09-17 15:46 it's stirring old memories, yes 2007-09-17 15:46 Okay, I think I get the gist of the problem. Much suckage. 2007-09-17 15:47 just have to grin and bear it imho 2007-09-17 15:47 yes, I just want to make sure I understand it completely. C's VLA's are "funky". 2007-09-17 15:47 if you feel strongly, then just take the ugly bit of code with a cast and give it a nice name 2007-09-17 15:48 flipz: The truth is that this is a common idiom... I just thought it died with C99, and I'm realizing it didn't. 2007-09-17 15:48 I should have done that in the first place, avoiding the need for this exchange ;) 2007-09-17 15:48 flipz: Nah, this exchange was needed, so as to fill out my incomplete knowledge. 2007-09-17 16:16 flipz: Hmm.. just checked the C99 standard. It turns out, what we are doing is illegal according to the standard, so all bets are off. 2007-09-17 16:16 which is that? 2007-09-17 16:16 address subtraction not in the same vector? 2007-09-17 16:16 "A structure or union shall not contain a member with incomplete or function type (hence, 2007-09-17 16:16 a structure shall not contain an instance of itself, but may contain a pointer to an instance 2007-09-17 16:16 of itself), except that the last member of a structure with more than one named member 2007-09-17 16:16 may have incomplete array type; such a structure (and any union containing, possibly 2007-09-17 16:16 recursively, a member that is such a structure) shall not be a member of a structure or an 2007-09-17 16:16 element of an array." 2007-09-17 16:17 Basically, if a structure has an incomplete array, it can't be a member of another structure. 2007-09-17 16:17 true 2007-09-17 16:17 we only do that at the end of structures as far as I know 2007-09-17 16:18 Well incomplete array types can *only* be at the end of structures, but the key thing is once you've done that, you can't nest that structure inside another structure, which we are doing. So we effectively end up with something the compiler lets us do for the heck of it. ;-) 2007-09-17 16:20 Where did we do it? 2007-09-17 16:21 snapshot_details has "sharing", which is an incomplete array. 2007-09-17 16:21 status_reply has "details" which is actually an array of snapshot_details. 2007-09-17 16:21 And *that* the standard says is a boo-boo, which is probably another reason you define details as being an array of char. ;-) 2007-09-17 16:22 status_reply->details is a char array 2007-09-17 16:22 your rules above are exactly why we don't declare it as a struct array 2007-09-17 16:22 Not mine, the languages. ;-) 2007-09-17 16:23 But yeah, now I finally grok all that. I didn't know about that rule. 2007-09-17 16:23 neither did I the first time I tried it 2007-09-17 16:24 basically, C sucks at info structures 2007-09-17 16:24 2D arrays are even too much for it 2007-09-17 16:25 now let me see, somebody on lkml wants to know all our anti-deadlock measures 2007-09-17 16:27 flipz: anti-deadlock measure: don't lock. ;-) 2007-09-17 16:27 don't start the computer is a simple, effective solution 2007-09-17 16:27 flipz: Oooh! yes! 2007-09-17 16:27 rather similar to some of the solutions that have been proposed in the past in fact 2007-09-17 16:28 "just don't do that" 2007-09-17 16:28 "but I need to do that" 2007-09-17 16:32 flipz: hehe, I always like the "just don't do that" critiques. 2007-09-17 16:33 surprisingly effective at stalling the progress of system improvements 2007-09-17 16:34 lkml is mostly people who have to maintain code rather than make progress, and they are rightly fearful of stuff that is painful to maintain. 2007-09-17 16:56 flipz. 2007-09-17 16:56 hi 2007-09-17 16:56 I just convinced myself that you don't need that (*pending)->holdcount = 1; 2007-09-17 16:57 that makes two of us 2007-09-17 16:57 byebye? 2007-09-17 16:57 Just lose it and the decrement, then. Yep. 2007-09-17 16:57 actually, byebye + unit testing 2007-09-17 16:57 Yeah. 2007-09-17 16:58 there are still some unit testing stubs hanging around in ddsnapd.c related to that 2007-09-17 16:58 We can kill them as we notice them. 2007-09-17 16:58 I think we should keep them and make them better 2007-09-17 16:58 *shrug* 2007-09-17 16:58 suggestion: add an ifdef'd main function to ddsnapd.c that calls unit tests 2007-09-17 16:59 otherwise, it is very difficult to make effective unit tests 2007-09-17 16:59 the only reason this code works as well as it does is those little unit tests 2007-09-17 16:59 I'm far too deep in figure-out-the-code-and-document-it mode to think about that right now. 2007-09-17 16:59 So, I just added this to calm my annoyances with C: 2007-09-17 16:59 static inline struct size_t snapshot_details_calc_size(unsigned row, unsigned snapshots) 2007-09-17 16:59 { 2007-09-17 16:59 return (size_t)snapshot_details(NULL, row, snapshots; 2007-09-17 16:59 } 2007-09-17 16:59 2007-09-17 16:59 exactly 2007-09-17 17:02 There's a paren missing there, but you get the idea. 2007-09-17 17:03 Oh damn... we don't have size_t 2007-09-17 17:03 No? 2007-09-17 17:03 that's what the compiler is telling me. 2007-09-17 17:04 some header 2007-09-17 17:04 #include (or ) 2007-09-17 17:05 Why does dm-ddsnap.h not have any #includes? 2007-09-17 17:05 (or ) 2007-09-17 17:05 fmayhar: lol 2007-09-17 17:05 cbsmith, because it is included in both kernel and userspace 2007-09-17 17:05 this needs some cleanup 2007-09-17 17:05 anyway, you don't need size_t there 2007-09-17 17:06 flipz: but it is the right thing to do 2007-09-17 17:06 because you know the compiler will do the right thing even if you cast to int, long, unsigned or whatever 2007-09-17 17:06 it isn't, really 2007-09-17 17:06 it's a C standardism, not a Right Thing 2007-09-17 17:07 ok, one more read through this longish lkml post 2007-09-17 17:07 I just know this is one of the ones that will end up heavily indexed by googlebot 2007-09-17 17:08 so having relatively few typos would behoove me 2007-09-17 17:11 cook, kmail's spell checker knows that userspace is a word 2007-09-17 17:11 but not "boundedness" 2007-09-17 17:11 userspace isn't a world, it is a universe! 2007-09-17 17:28 ok, there we go, that was _one_ email answered 2007-09-17 17:28 only about 60 to go 2007-09-17 17:58 -!- fmayhar(~fmayhar@207.47.98.129.static.nextweb.net) has left #zumastor irc.oftc.net #zumastor log beginning Tue Sep 18 00:00:02 PDT 2007 2007-09-18 09:39 -!- fmayhar(~fmayhar@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-18 11:53 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-18 11:54 Hey, any suggestions out of the ridiculous number of places to eat around here? 2007-09-18 12:13 scox worth 18 cents now :) 2007-09-18 12:14 here means santa monica? 2007-09-18 12:22 flipz: yes 2007-09-18 12:22 flipz: yeah, its sinking like a stone 2007-09-18 15:15 shapor_: yt? irc.oftc.net #zumastor log beginning Wed Sep 19 00:00:02 PDT 2007 2007-09-19 02:34 -!- juuva(juuva@peili.org) has joined #zumastor 2007-09-19 03:31 -!- erwan_taf(~erwan@72-255-38-191.client.stsn.net) has joined #zumastor 2007-09-19 11:01 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-19 13:19 -!- cbsmith(~user@207.47.98.129.static.nextweb.net) has joined #zumastor 2007-09-19 14:58 ACTION waves 2007-09-19 15:23 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor irc.oftc.net #zumastor log beginning Thu Sep 20 00:00:02 PDT 2007 2007-09-20 08:01 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-20 08:10 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-20 09:22 -!- cbsmith(~user@adsl-76-240-81-238.dsl.lsan03.sbcglobal.net) has joined #zumastor 2007-09-20 10:32 rewriting the journal replay code 2007-09-20 10:33 doing a little head scratching on the most obvious way to compute the "highest" journal sequence number, where the sequence number may wrap 2007-09-20 10:55 flipz: loop and break when the number goes down? 2007-09-20 10:56 that qualifies as obvious 2007-09-20 10:56 more obvious than what I just wrote in fact 2007-09-20 10:56 flipz: agreed, and simple. Am I missing something. 2007-09-20 10:57 ? 2007-09-20 10:57 no you didn't miss anything as far as I can see 2007-09-20 10:57 well, "down" isn't exactly an obvious concept when the number can wrap 2007-09-20 10:58 so I retract my above remark ;) 2007-09-20 10:59 "doesn't go up by one" is the winning concept 2007-09-20 11:00 and we also need to verify that that happens exactly once, otherwise the journal is corrupt 2007-09-20 11:00 flipz: oooh! I like that idea. 2007-09-20 11:00 ACTION is a big fan of integrity checking when it comes to persistent data 2007-09-20 11:34 -!- erwan_taf(~erwan@mailgate.filetek.com) has joined #zumastor 2007-09-20 13:35 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-20 14:22 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-20 15:01 -!- murb(~murbix@soapstone.yuri.org.uk) has joined #zumastor 2007-09-20 20:47 -!- jdries3(~jdries3@14.sub-75-195-153.myvzw.com) has joined #zumastor 2007-09-20 21:01 -!- jdries3_(~jdries3@231.sub-70-198-168.myvzw.com) has joined #zumastor 2007-09-20 21:14 -!- jdries3(~jdries3@72.14.228.1) has joined #zumastor 2007-09-20 21:44 -!- jdries3(~jdries3@c-69-249-52-124.hsd1.nj.comcast.net) has joined #zumastor irc.oftc.net #zumastor log beginning Fri Sep 21 00:00:02 PDT 2007 2007-09-21 06:41 -!- jdries3(~jdries3@249.sub-75-194-125.myvzw.com) has joined #zumastor 2007-09-21 07:49 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-21 10:41 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-21 12:18 -!- cbsmith(~user@adsl-76-240-81-238.dsl.lsan03.sbcglobal.net) has joined #zumastor 2007-09-21 12:38 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-21 14:44 -!- jdries3(~jdries3@72.14.228.89) has joined #zumastor 2007-09-21 15:19 So, for those of you who hate "typeof", does this make it any more digestable?: tyepof (buf2block()->sequence) seq[jblocks] 2007-09-21 15:20 s/tyepof/typeof/ 2007-09-21 15:42 Not so much, no. 2007-09-21 15:46 I do have to say that "fieldtype(commit_block, sequence) seq[jblocks];" is a lot less ugly. It doesn't make my eyes bleed _nearly_ as much. 2007-09-21 15:47 fmayhar: I figured using the function call would be better, as that's where the sequence is actually coming from. 2007-09-21 15:47 fmayhar: But hey, if a macro helps, I have no objection. 2007-09-21 15:48 hi cbsmith 2007-09-21 15:48 hi 2007-09-21 15:49 c is just plain ugly 2007-09-21 15:49 agreed 2007-09-21 15:49 you have to start with that point, and then just strive for degress of slightly less ugly 2007-09-21 15:49 Which I think we're all striving for, each improving on the original mess. 2007-09-21 15:49 ACTION observes that Boost.Typeof would be soooo handy here. ;-) 2007-09-21 15:50 c really ought to have a set of field size attributes by now 2007-09-21 15:50 it's rediculous how much of the lrest of the language has bloated up, and fundamental issues like this are still left as lore 2007-09-21 15:50 flips: Actually, C++ is going to overload the "auto" keyword to allow the compiler to do type inferencing. If the C committee could get on board with that this could would look pretty nice. 2007-09-21 15:51 Should be a slash there between could and would... must have slipped. 2007-09-21 15:51 coulda, woulda, shoulda... 2007-09-21 15:52 Anyway, I think frank's suggestion of using the macro would help with readability... except for those that tried to decipher the macro. 2007-09-21 15:52 That was flips, actually. 2007-09-21 15:52 now that the ugliness issues are for the moment in abeyance, how about the correctness questions? :-) 2007-09-21 15:52 ah 2007-09-21 15:52 Working on that. 2007-09-21 15:52 that NULL->field thing is a meme 2007-09-21 15:52 Likewise. Trying to clean up my own mess before pointing at yours. ;-) 2007-09-21 15:52 it's just ugly when used inline 2007-09-21 15:53 if you want to really hurt your eyes, go feast them on the libc headers where such macros are declared 2007-09-21 15:53 flips: Yeah. I just hate the meme. I'm a big fan of "if you are going to use typeof, get the typeof the thing that is the actual source of the data", in this case, the function. 2007-09-21 15:53 go read offsetof for some real pain 2007-09-21 15:54 flips: Did that when I was trying to avoid that ugliness with the VLA's the other day. :-( 2007-09-21 15:54 ACTION cried like a baby 2007-09-21 15:54 Did that in the past. It's perverted and weird. 2007-09-21 15:54 yes, and yet it's not easy to see how to clean it up 2007-09-21 15:55 the compiler really ought to be doing this 2007-09-21 15:55 Exactly. That was my thought until I actually ended up looking at the code. 2007-09-21 15:56 flips: Well, there are lots of languages with good type inferencing.... just not C. 2007-09-21 15:56