**** BEGIN LOGGING AT Mon Jul 25 02:59:57 2011 Jul 25 03:24:10 sorta :) Jul 25 05:13:49 dtzWill, http://ompldr.org/vOW1hNA Jul 25 08:26:23 finally here Jul 25 08:38:12 rwhitby, wow...thats like way further along than I thought Jul 25 13:02:18 ka6sox: that's what 12 hours on a plane with a power socket can do :) Jul 25 13:50:50 whats the patch portal url again? Jul 25 13:53:00 http://patches.webos-internals.org/ looks about right Jul 25 13:53:33 so how come the patch on my linux box creates empty dirs Jul 25 13:53:38 but the one on my touchpad does not Jul 25 13:55:12 dofferent versions of patch Jul 25 13:55:20 inside those empty tirs just do a touch empty Jul 25 13:55:25 and then it should work no matter what Jul 25 13:56:12 well that would defeat the point of using patch Jul 25 13:56:32 something i could submit as a patch now have to make an ipk Jul 25 13:56:36 that has just a postinst Jul 25 13:56:37 :/ Jul 25 13:56:50 no, the patch would create that file, in that dir Jul 25 13:57:07 do the touch empty in the empty dirs so there is a file in there Jul 25 13:57:10 then make the patch file Jul 25 13:57:24 ehh i dont really follow Jul 25 13:57:27 then when the patch is applied, it should create that dir with that file in it Jul 25 13:57:50 http://dpaste.com/577302/ Jul 25 13:57:53 that was my patch Jul 25 13:59:05 I have no clue then Jul 25 14:39:36 is what I did what you were suggesting or were you suggesting something else? Jul 25 15:01:53 rwhitby in hte section up above on the top (with current sizes), need to add free space in the lvm Jul 25 15:01:59 parsed from ; Jul 25 15:02:03 pvscan Jul 25 15:14:05 PuffTheMagic, I have no clue what you are doing with that patch... my brain is now working correctly today Jul 25 15:14:20 lol Jul 25 15:14:25 usually if I have a folder that is empty that I want to make sure is created by a patch Jul 25 15:14:33 cryptk: just wanted to add a file to /etc/profile.d Jul 25 15:14:37 I just throw an empty file in it before I generate the patch file Jul 25 15:14:51 ka6sox: this lvm tool, i assume has a plugin or service Jul 25 15:15:00 ka6sox: couldnt it be using liblvm? Jul 25 15:15:05 instead of having to parse stuff manaully Jul 25 15:15:12 i could prob help out with that if needed Jul 25 15:16:34 * cryptk bought orange sherbet and sprite earlier Jul 25 15:16:38 to make floats Jul 25 15:16:40 cryptk: oh i think i get what you are saying i will try it out later Jul 25 15:17:21 because then when patch does it's work, it will create that file, and that folder... some versions of patch won't create empty folders, but if there is a placeholder file in it, it will Jul 25 15:17:27 I had a guinness float this week :D Jul 25 15:17:45 err last week Jul 25 15:17:49 I make mine with sprite, orange sherbet and vodka Jul 25 15:18:01 that sounds great though Jul 25 15:18:07 orange sherbet is awesome Jul 25 15:18:28 it is great... you should totally make one sometime Jul 25 15:18:36 I plan on making them at SCaLE this year Jul 25 15:18:39 PuffTheMagic, yes it has a C-service... Jul 25 15:19:08 http://fedoraproject.org/wiki/LVM/liblvm Jul 25 15:19:15 might be useful Jul 25 15:20:09 ya, we have partitioning, formatting, reducing, increasing, etc... Jul 25 15:20:15 quite a bit of different stuffs. Jul 25 15:20:52 liblvm should be able to do all that stuff from what the page says Jul 25 15:21:34 http://sourceware.org/git/?p=lvm2.git;a=tree;f=liblvm;h=43385553d9c3adbb975fb9a70aee8f40736ae386;hb=HEAD Jul 25 15:22:42 ya, its a good thing(tm) Jul 25 15:24:49 * cryptk is getting ready to write a regression test script for WIDK to find packages that don't have their dependencies correct Jul 25 15:25:13 basically step through every package... build it... then clobber all of them and repeat Jul 25 15:25:25 and log any of the builds that exit with a bad errorlevel Jul 25 15:26:14 why arent the things in widk part of build.git Jul 25 15:26:28 PuffTheMagic, that might be good stuff. Jul 25 15:26:39 ka6sox: ? Jul 25 15:26:51 the lib Jul 25 15:27:09 oh, i thought you had already dismissed it Jul 25 15:27:45 widk is where the initial hacking is done... so sometimes it breaks... some times things don't make the cut, et Jul 25 15:27:47 etc* Jul 25 15:27:53 so they never get put in build.git Jul 25 15:32:29 I wonder how long the regression test will take... hours... if not over a day Jul 25 15:36:07 PuffTheMagic, nope...just looking Jul 25 15:56:05 dtzWill: for Xsdl - bringing up the keyboard - how about using multiple finger swipe? Jul 25 16:22:20 rwhitby, thanks...that looks great! Jul 25 16:47:25 I am almost done writing the regression test script Jul 25 16:47:40 rwhitby, after I am done writing it, I am going to test it out... if it works, do you want me to check it in? Jul 25 16:47:59 it basically go's through every package... does a make stage, then clobbers all packages... then make stage on the next Jul 25 16:48:24 stores a log file for every packages build, and if a package build exits with a non-zero runlevel then it stops Jul 25 16:48:40 it should be able to detect things like dependency problems, build problems, etc Jul 25 16:49:01 by default it only does it for armv7, but that is configurable with command line options Jul 25 16:50:07 and it has a handy feature where it won't run the test, but it will jsut clobber all packages Jul 25 16:50:18 which could likely be done just as well through a make target if it isn't already Jul 25 16:51:30 rwhitby: sounds good. will have to sort out the right way to detect that Jul 25 16:51:47 but should be straightforward and if implemented right is an excellent idea Jul 25 16:51:59 small problem is i'm unclear if Xsdl gets SDL mouse events for the keyboard area Jul 25 16:52:16 what sounds good? Jul 25 16:52:22 :-) Jul 25 16:53:35 bash magic http://pastebin.com/srnvAPQs Jul 25 16:55:15 please don't run that yet though... it still isn't quite right Jul 25 17:00:57 linuxjacques: oh, using multi-finger swipe for keyboard Jul 25 17:01:05 actually kinda surprised webos doesn't already just have that Jul 25 17:01:24 * dtzWill remembers-ish apple having patents on some of that? Jul 25 17:01:59 cryptk: is that for finding erros in the widk? Jul 25 17:02:08 dep-wise, etc, from clobbered build? Jul 25 17:02:12 dtzWill, yes Jul 25 17:02:19 cryptk: have you seen the 'debug' target? :) Jul 25 17:02:28 although yours looks a lot more usable. Jul 25 17:02:47 if you end up finishing that and like it/etc, should rip out the debug stuff then. Jul 25 17:02:51 dtzWill: multi-finger to show/hide keyboard? or for modifiers ? Jul 25 17:02:58 linuxjacques: show/hide. Jul 25 17:03:01 linuxjacques: modifiers is interesting. Jul 25 17:03:10 also, we need right-click guys :) Jul 25 17:03:18 indeed. Jul 25 17:03:22 gah all this usability stuff is hard :( Jul 25 17:03:29 i just wanna play with xservers and such :P :( Jul 25 17:03:54 yeah, it's tedious Jul 25 17:03:59 dtzWill, mine does fanciness like creating a log file for every package build, etc etc Jul 25 17:04:11 cryptk: yeah yours is cleaner and probably just plain better :) Jul 25 17:04:56 dtzWill: have you seen issues with screenshots of 16-bit SDL apps? Jul 25 17:05:30 also, do pixel and depth buffers need to fit in 32 bits? Jul 25 17:05:44 can I have 24-bit color and 16-bit depth? Jul 25 17:08:04 dammit did they reboot the http proxy again? Jul 25 17:08:33 hello? Jul 25 17:09:08 hello! Jul 25 17:09:51 thanks for responding halfhalo, this webchat is crap Jul 25 17:10:04 it stops working and I don;t know til much later Jul 25 17:10:21 I was trying to get a bot to respond to me to see if it was working, but it wasn't Jul 25 17:10:40 yeah. I do hate webchats. Jul 25 17:11:31 about 30 minutes later I get a "connection closed because it didn't respond to several requests" or somesuch Jul 25 17:11:33 linuxjacques: sorry for delay :) Jul 25 17:11:41 linuxjacques: i haven't seen issues, no. Jul 25 17:11:49 are you talking about writing an SDL app for webos itself? Jul 25 17:11:52 (nothing related to x?) Jul 25 17:12:12 dtzWill: haven't seen issues with which? screenshots of 16-bit SDL apps? Jul 25 17:12:17 b/c if you force SDL to give you a 16bit surface in webos, it'll just use a 32bit anyway, allocate you a 16bit one and blit it internally Jul 25 17:12:28 dtzWill: ouch Jul 25 17:12:29 or at least it did in..... 1.4.5 or whenever i last looked closely Jul 25 17:12:31 that sounds slow Jul 25 17:12:43 linuxjacques: right, so it was always better to use 32bit itself iirc Jul 25 17:12:54 I ported a game over, but it uses 16-bit pixel and 16-bit depth buffers Jul 25 17:13:16 when I was taking screenshots, I was getting what looked like pixel and depth info Jul 25 17:13:21 i'm curious about benchmarking raw SDL 2d blitting performance vs the GL texture stuff i'm doing these days.... Jul 25 17:13:31 so the screenshots looked terrible and nothing like the screen Jul 25 17:13:34 (on newer webos versions, and newer hardware) Jul 25 17:13:45 should just write that up and have the source handy for testing... Jul 25 17:14:18 I'm wondering if I can tell SDL to use 24-bit color, 8-bit alpha, and 16-bit depth Jul 25 17:14:39 and then it would be both faster and screenshots would work? Jul 25 17:14:47 (i don't know) Jul 25 17:15:07 maybe it's b/c i'm tired or hungry--but regardless i honestly have no idea what color is versus depth Jul 25 17:15:25 tkae that back, i just simply don't know. no state qualifiers needed. Jul 25 17:15:26 z-buffer Jul 25 17:15:27 ok fixed a couple of bugs in the regression-test http://pastebin.com/8U15dHtw Jul 25 17:15:31 giving it it's first run now Jul 25 17:15:37 this is gonna take FOREVER Jul 25 17:15:38 depth buffer == z-buffer Jul 25 17:15:38 linuxjacques: o_____O sdl has z-buffer? Jul 25 17:15:39 lol Jul 25 17:16:07 clobber all packages... build first one... clobber all packages... build second one... and so on... Jul 25 17:16:24 linuxjacques: oh, are you using SDL+GL(ES)? Jul 25 17:16:26 dtzWill: you going to be around for a bit? I need to setup the laptop and find the code so I can pastebin it Jul 25 17:16:31 dtzWill: yes Jul 25 17:16:34 except it didn't pick up that the first package didn't build correctly... Jul 25 17:16:51 GLES Jul 25 17:18:16 linuxjacques: ahhhhhhh Jul 25 17:18:19 okay then i'm completely on board. Jul 25 17:18:57 and my cmment 16 vs 32, i don't know how that'll work. don't know that the sdl_surface returned by sdl_setvideo mattters at all then Jul 25 17:19:08 i mean i don't think with GL you'll have the copying i was talking about at all Jul 25 17:19:48 ok, it does run pretty fast Jul 25 17:19:52 even in 1.1_CM mode Jul 25 17:20:12 which on the tp is not in hw but implemented with 2.0 shaders Jul 25 17:20:21 so they recommend 2.0 for performance Jul 25 17:20:40 but I can't get good screenshots :-( Jul 25 17:20:44 rwhitby, did you see the comment..re: freespace in the LVM? Jul 25 17:27:56 hrm... it seems that piping things through tee hides their errorlevel... Jul 25 17:28:25 so even though the make ended with an errorlevel of 2... the tee finished successfully with an errorlevel of 0... making checking the errorlevel impossible Jul 25 17:33:53 ka6sox: yes, that is shown in the text box at the bottom where you type in amount, and it updates automatically Jul 25 17:34:26 cryptk: "(foo || exit) | tee" might work Jul 25 17:35:57 rwhitby, gonna test that Jul 25 17:37:02 rwhitby, in about an hour I'll push up the removal script for ext3fs (which I will genericize to include f00 Jul 25 17:37:04 ) Jul 25 17:37:33 also should we create something that manages the swap or do already have htat? Jul 25 17:38:28 rwhitby, you are magic! Jul 25 17:40:57 rwhitby, and that allows me to add another option to the script... continue processing on a fail Jul 25 17:41:51 cryptk: yep, use ( foo || ${EXIT} | tee Jul 25 17:42:00 and then define EXIT appropriately Jul 25 17:42:26 the 'foo' in this case is make stage arch=SOMETHING Jul 25 17:43:17 rwhitby, I got tired of going through the WIDK and finding problems... so I am writing a script to do that for me... for every package... Jul 25 17:46:04 ok, giving it another test run Jul 25 18:04:43 rwhitby, because it is in a subshell, that isn't working... it is exiting that subshell, but it isn't exiting the script.. Jul 25 18:04:57 but I think I know a way to do it... Jul 25 18:13:23 anythign that happens in the child shell can't really be detected by the parent shell... except for one thing... Jul 25 18:13:30 filesystem changes Jul 25 18:13:59 cryptk: exit 1 or exit 0 can be detected Jul 25 18:14:18 that isn't what I was seeing Jul 25 18:14:28 I had it doing an exit, the script continued Jul 25 18:14:41 because it was exiting the subshell but not the parent Jul 25 18:14:56 unless I want to have a chain of exits and checking errorlevels Jul 25 18:15:19 but having it output to a log is ok... because it is a regression test anyway, and I was already having it make a log file that I can test Jul 25 18:25:14 ok, now I have something that works! Jul 25 18:25:32 going to do a from-scratch setup of WIDK and let her rip! Jul 25 18:25:50 how much you guys wanna bet that navit fails? Jul 25 18:37:03 4 bitcoins Jul 25 18:40:37 thats how many cpucycles? Jul 25 18:40:45 .0032 Jul 25 19:17:23 btw, the test found a problem with glibc... it fails complaining about not having the libs for the thread implimentation Jul 25 19:18:11 cryptk: haha, yeah.... Jul 25 19:18:16 it also provides those headers? :( Jul 25 19:19:15 so... how do we fix that? Jul 25 19:23:52 I hit that at one point. IIRC it wasn't true. I worked around it. Jul 25 19:24:11 BTW, we talking glibc or glib ? Jul 25 19:24:55 if glib, I worked around it on 11.04 by downgrading my qemu-arm to 0.13.0 Jul 25 19:25:38 sorry glib... not glibc Jul 25 19:25:58 cryptk: and you are having the problem on a 11.04 build host? Jul 25 19:26:18 yeppers Jul 25 19:26:50 I hacked on it for days before arriving on the qemu downgrade workaround Jul 25 19:26:57 I dunno if there is a better way Jul 25 19:27:31 the pthread test app which configure is building is segfaulting when run under qemu-arm Jul 25 19:27:47 (the stock 11.04 version of qemu-arm) Jul 25 19:30:23 @$&*)(*&(*(%^*( Jul 25 19:30:30 (*&%^&% Jul 25 19:30:35 what did I miss? Jul 25 19:30:45 what's the last thing you guys saw me say? Jul 25 19:31:05 [14:20] I hacked on it for days before arriving on the qemu downgrade workaround [14:20] I dunno if there is a better way [14:21] the pthread test app which configure is building is segfaulting when run under qemu-arm [14:21] (the stock 11.04 version of qemu-arm) Jul 25 19:32:18 yep, I saw that Jul 25 19:32:45 ok, webchat died on me with the "connection closed after several requests failed" Jul 25 19:32:51 man... does that mean I am gonna need to do a manual downgrade? 0.14.50 is the only one in the repos... Jul 25 19:32:58 I never know which requests it's talking about Jul 25 19:33:22 cryptk: yeah I built it and moved the qemu-arm binary into place :-\ Jul 25 19:33:38 sonofa Jul 25 19:34:17 I tried a few other versions (like git HEAD) too, but 0.13.0 was the first one that worked Jul 25 19:37:44 I may try latest git HEAD to see if they've fixed it Jul 25 19:39:29 linuxjacques_, did you just get it from qemu.org? Jul 25 19:39:57 cryptk: yeah Jul 25 19:40:16 http://wiki.qemu.org/download/qemu-0.13.0.tar.gz Jul 25 19:40:41 weird thing is that my qemu-arm is reporting as being version 0.14.50.... the qemu.org site says that 0.14.1 is the latest, lol Jul 25 19:41:34 same here Jul 25 19:41:42 but in git they have 0.15.0 Jul 25 19:42:58 where the hell are the build instructions... the README tells me nothing but to read the contents of a file that doesn't exist Jul 25 19:43:09 is it just s standard configure and make? Jul 25 19:44:18 haha, the file the readme points to doesn't exist until after you build the damn thing Jul 25 19:44:35 that's a great way to do it... don't give them the docs telling them how to compile it until after they compile it Jul 25 19:45:28 yeah, it's not friendly - gimme a sec Jul 25 19:45:37 I am already building it Jul 25 19:45:48 it was a ./configure and a make Jul 25 19:47:13 ok, I add " --target-list=arm-linux-user" Jul 25 19:47:20 because that's the only one we need right now Jul 25 19:48:36 add that to the configure or the make? Jul 25 19:48:49 configure right? Jul 25 19:51:09 ok, qemu-arm built, original copied out of the way, new one in place Jul 25 19:53:46 cryptk: sorry, yesm the configure Jul 25 19:53:59 s/yesm/yes/ Jul 25 19:54:08 yep, it is already built Jul 25 20:10:54 ok, thanks linuxjacques_ that did fix it Jul 25 20:11:03 now to start the regression test up again Jul 25 20:13:56 cryptk: cool :-) Jul 25 20:14:37 now if only I could find a workaround for building gtk :-( Jul 25 20:17:32 there we go... set it up to where if a build fails it will also play a siren sound Jul 25 20:17:46 now I can go watch a movie and it will yell at me when something breaks Jul 25 20:19:23 heh Jul 25 20:20:56 I should run this thing around time so that I see how long it takes to run... Jul 25 21:21:05 plus the software I use splits it into chunks... then has multiple processes transcode it... then puts it back together Jul 25 21:21:13 multithreaded transcoding... yumm... Jul 25 21:22:06 but the software isn't done yet Jul 25 21:22:19 my buddy is writing it, it's basically a re-work of dvdrip Jul 25 21:38:36 I used to use dvdrip, these days I use handbrake Jul 25 22:44:33 jhoweb, this one is smaller than slim Jul 25 22:44:43 2.7M vs 2.9M Jul 25 22:44:49 there has got to be something wrong there Jul 25 22:49:56 jhoweb, so would you welcome some patches once the wifi stuff is fixed? Jul 25 22:54:00 jhoweb, that one booted, 1:21 Jul 25 22:54:50 jhoweb, this should fix some power consumption > http://dpaste.com/577781/ Jul 25 22:56:15 jhoweb, http://dpaste.com/577783/ <- here is another one to fix power consumption Jul 25 22:56:24 jhoweb, it booted, still no wifi though Jul 25 22:56:25 PuffTheMagic: is that from https://lkml.org/lkml/2010/7/17/34 ? Jul 25 22:56:29 yup Jul 25 22:56:43 err Jul 25 22:56:47 it was on lkml Jul 25 22:56:50 let me double check Jul 25 22:57:23 rwhitby, ya its almost that patch ;) Jul 25 22:57:26 that part that matters Jul 25 22:57:30 rwhitby, ka6sox, i vote tailor be re-named resizuh Jul 25 22:57:35 I concur, lol Jul 25 22:58:21 that comment is only about 3 days late. it was originally called resizah, then partitionah, then tailor. Jul 25 22:58:31 aww damn... Jul 25 22:58:40 lolz Jul 25 22:58:42 resiz'U'h doesn't make sense Jul 25 22:59:00 resizah Jul 25 22:59:16 why did you go away from the *ah name? Jul 25 22:59:30 threats from oil Jul 25 22:59:39 the author of the app chose the name Jul 25 22:59:42 ahh Jul 25 23:01:50 jhoweb, it booted, no wifi Jul 25 23:01:53 same puke in dmesg Jul 25 23:02:05 jhoweb, http://dpaste.com/577786/ <- vmscan inconsistency Jul 25 23:02:49 jhoweb, what HZ setting do you use on warthog? Jul 25 23:03:01 whatever the stock kernel uses Jul 25 23:03:08 i havent screwed with it Jul 25 23:04:22 i think its at like 250 or 100 then, unless palm upped it to 1000 where it should be if NOHZ is enabled Jul 25 23:04:39 tailah ? Jul 25 23:05:46 CONFIG_NO_HZ=y CONFIG_HZ=100 Jul 25 23:05:51 is what palm has it at Jul 25 23:07:29 cryptk: can you check /proc/config.gz and see if spinlock debug was indeed re-enabled? Jul 25 23:07:49 jhoweb, http://dpaste.com/577793/ > make 1000hz default Jul 25 23:08:08 linuxjacques: its not in db1 Jul 25 23:08:25 linuxjacques, CONFIG_DEBUG_SPINLOCK=y Jul 25 23:08:41 wait... that is stock kernel Jul 25 23:08:43 so that wasn't it? :-\ Jul 25 23:08:56 ah Jul 25 23:09:09 db1 did boot, but no wifi, db2 = boot>/dev/null Jul 25 23:09:24 but jhoweb just said that db1 didn't have spinlock enabled anyway Jul 25 23:10:20 100hz should be used for servers not for interactive destktop, 1000hz will be better Jul 25 23:11:59 PuffTheMagic: do you have that 200 liner patch in a version that will apply to 2.6.35? Jul 25 23:12:24 http://www.phoronix.com/scan.php?page=article&item=linux_2637_video&num=1 Jul 25 23:12:48 jhoweb, the just deletes the unused function Jul 25 23:13:30 and its not 200 lines Jul 25 23:13:33 which one? Jul 25 23:13:41 this http://marc.info/?l=linux-kernel&m=128978361700898&w=2 ? Jul 25 23:14:09 no not that one Jul 25 23:14:18 that adds code Jul 25 23:14:31 the patch about the power usage removes stuff Jul 25 23:14:59 that's fine Jul 25 23:15:15 now if that was in a version avail for the veer Jul 25 23:15:27 auto groups? Jul 25 23:15:33 or the power issue patch Jul 25 23:15:46 i dont think the power issue patch is needed on .29 Jul 25 23:15:56 it was a fairly recent regression Jul 25 23:16:14 ok Jul 25 23:16:33 autogroups should apply to .35 but I havent tried Jul 25 23:17:26 is the wifi issue fixed yet? Jul 25 23:17:31 no Jul 25 23:17:43 still trying to triangulate down to the offending CONFIG Jul 25 23:18:31 the autogroup patch prob wont help much since nothing has a controlling tty Jul 25 23:18:40 and the autogroups are based on tty Jul 25 23:20:19 even with CONFIG_DEBUG_SPINLOCK=y the nodebug kernel gives the undefined symbol errors on load attempt? Jul 25 23:31:37 cryptk: try this kernel https://www.box.net/shared/c3j0rn8zk46cell02s8i Jul 25 23:33:12 can't, gotta run Jul 25 23:33:41 should I test this? Jul 25 23:34:44 if you have 3.0.0 running Jul 25 23:34:49 you can memboot it Jul 25 23:36:42 ya i have 3.0.0 Jul 25 23:36:47 it this supposed to have working wifi or not? Jul 25 23:37:39 well Jul 25 23:37:49 that's why it's called trying things out Jul 25 23:37:58 it should Jul 25 23:38:21 but its not tested? i will give it a shot Jul 25 23:38:44 of course not tested, i just typed make Jul 25 23:38:57 that what you guys are for Jul 25 23:39:06 guinea pigs. mwhahahahahaha Jul 25 23:39:06 lol Jul 25 23:39:37 novacom boot mem:// < ~/Downloads/uImage-tp-db3 Jul 25 23:39:40 is that not right Jul 25 23:39:45 i havent done this in months? Jul 25 23:40:35 http://www.webos-internals.org/wiki/Memboot Jul 25 23:40:55 ya thats what i have Jul 25 23:40:57 wtf Jul 25 23:41:21 is it in recovery mode? Jul 25 23:41:25 and on that note... Jul 25 23:41:28 ;) Jul 25 23:41:28 * cryptk is out the door Jul 25 23:41:33 forgot about that part Jul 25 23:42:34 i wonder if u can hold down the vol key to get a fb console on boot like u could on the phone Jul 25 23:43:32 jhoweb, ar6000 is loaded Jul 25 23:43:51 ok Jul 25 23:43:58 so what was it? Jul 25 23:43:58 so wifi works Jul 25 23:44:02 and pdk too? Jul 25 23:44:37 from looking at kernel/Kconfig.locks it looks like CONFIG_DEBUG_SPINLOCK=y is the key Jul 25 23:45:32 jhoweb, preware sync worked and angry birds plays Jul 25 23:45:41 linuxjacques: not just that Jul 25 23:46:41 jhoweb: what else? Jul 25 23:46:53 probably DEBUG MUTEX Jul 25 23:48:12 I see lots like this: Jul 25 23:48:16 config INLINE_SPIN_LOCK_IRQSAVE def_bool !DEBUG_SPINLOCK && !GENERIC_LOCKBREAK && \ ARCH_INLINE_SPIN_LOCK_IRQSAVE Jul 25 23:48:54 and we want !INLINE_SPIN_LOCK_IRQSAVE Jul 25 23:50:50 I wonder how much performance changes by not having the spinlocks inlined Jul 25 23:52:17 linuxjacques: patches welcome Jul 25 23:53:19 the stock TP has CONFIG_ARCH_INLINE_SPIN_LOCK_IRQSAVE=y Jul 25 23:53:43 yeah I saw that Jul 25 23:54:13 not sure how that's different from CONFIG_INLINE_SPIN_LOCK_IRQSAVE Jul 25 23:54:26 which is not set stock Jul 25 23:56:09 bb Jul 25 23:56:14 bbl Jul 26 00:08:15 PuffTheMagic: try this one https://www.box.net/shared/v2rfzvmh7zpdvtm92nxt Jul 26 00:08:24 it should be the final one Jul 26 00:21:17 jhoweb, whats new in this once Jul 26 00:21:18 one Jul 26 00:21:43 just toggling debug options Jul 26 00:21:48 i think i have it narrowed down Jul 26 00:22:27 jhoweb, should merge in those power saving patches and switch to 1000HZ for db5 ;) Jul 26 00:25:25 im booted Jul 26 00:27:25 wifi? pdk? Jul 26 00:27:35 both work Jul 26 00:27:40 awesome Jul 26 00:31:22 ok Jul 26 00:31:24 all pushed Jul 26 00:31:41 that should be in the builder shortly Jul 26 00:35:13 https://lkml.org/lkml/2010/7/17/34 has already been applied as part of palm's patches Jul 26 00:36:07 jhoweb, what about http://kerneltrap.org/mailarchive/linux-kernel/2010/8/2/4601475 Jul 26 00:38:44 that one i just applied Jul 26 00:42:02 jhoweb, here is a fun one :D http://ompldr.org/vOW1wdg/bfq.patch Jul 26 00:42:07 BFQ IO scheduler Jul 26 00:47:15 lkml / lwn link? Jul 26 00:47:37 http://lwn.net/Articles/275978/ Jul 26 00:47:57 Stochastic Fair Blue Network scheduler -> http://dpaste.com/577874/ Jul 26 00:48:06 http://www.pps.jussieu.fr/~jch/software/sfb/ Jul 26 00:48:15 i also need to know what I need to turn on in defconfig for BFQ Jul 26 00:48:37 oh Jul 26 00:48:45 i disabled that part of the patch Jul 26 00:48:55 i didnt think you wanted it set to default Jul 26 00:49:26 no, give me the defaults Jul 26 00:49:30 I dont like CFQ Jul 26 00:49:37 CONFIG_IOSCHED_BFQ Jul 26 00:50:05 http://dpaste.com/577877/ Jul 26 00:50:08 that makes it default Jul 26 00:50:12 but does not change defconfig Jul 26 00:50:39 I want to know what to add/modify in defconfig Jul 26 00:50:57 -CONFIG_DEFAULT_IOSCHED="cfq" Jul 26 00:50:58 +CONFIG_DEFAULT_IOSCHED="bfq" Jul 26 00:51:10 CONFIG_DEFAULT_BFQ=y Jul 26 00:51:14 CONFIG_IOSCHED_BFQ=y Jul 26 00:51:38 my friend's advisor is the developer of this sched Jul 26 00:51:54 so Jul 26 00:51:56 CONFIG_IOSCHED_BFQ=y Jul 26 00:52:02 CONFIG_DEFAULT_BFQ=y Jul 26 00:52:09 CONFIG_DEFAULT_IOSCHED="bfq" Jul 26 00:52:22 ya Jul 26 00:52:29 that should do it Jul 26 00:52:29 you will need the CFQ patch I have already Jul 26 00:52:34 applied Jul 26 00:52:46 webos is silly enuff that it tries to force cfq always Jul 26 00:53:03 thats the one in preware that unsets it right? Jul 26 00:53:14 i believe i have that applied already Jul 26 00:53:36 yeah Jul 26 00:53:52 that i have it Jul 26 00:56:03 def set HZ_1000 to default while you are in there Jul 26 00:56:53 jhoweb, here is a better link for bfq http://algo.ing.unimo.it/people/paolo/disk_sched/ Jul 26 00:59:51 dude, dont make me do that much extra work Jul 26 01:00:03 give it to me in an easily applied patch Jul 26 01:00:31 its a long multistep process to get things in already Jul 26 01:00:46 same thing with the sfbn Jul 26 01:02:21 sfbn? Jul 26 01:02:26 for sfbn, backport this commit https://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=e13e02a3c68d899169c78d9a18689bd73491d59a Jul 26 01:02:34 Stochastic Fair Blue Network scheduler Jul 26 01:02:48 i already gave you the patch Jul 26 01:02:53 for .35 Jul 26 01:03:22 its an older version Jul 26 01:03:50 scroll to bottom of http://www.pps.jussieu.fr/~jch/software/sfb/ Jul 26 01:04:49 its slightly modifies because things changed > .35 Jul 26 01:10:56 im not sure if i want to apply this one Jul 26 01:12:06 not convinced of the benefits Jul 26 01:22:22 where's the corresponding info for the vmscan patch? Jul 26 01:22:23 http://dpaste.com/577786/ Jul 26 01:39:24 PuffTheMagic: try this kernel https://www.box.net/shared/hspaijijfxm62jtv8lzv Jul 26 01:39:40 this has bfq, 1000hz and tick skew Jul 26 01:43:02 :D Jul 26 01:43:25 bfq is gonna be fun Jul 26 01:54:16 dtzWill, ping Jul 26 02:27:42 PuffTheMagic: pong Jul 26 02:27:43 :) Jul 26 02:31:21 dtzWill, trying to think about how to handle the library and group and folder data Jul 26 02:31:43 like, should I clear the db and then push new shit each time sync happens Jul 26 02:31:50 or try updating/merging Jul 26 02:32:08 cant make up my mind Jul 26 02:37:23 dtzWill, have you set up to actually build the plugin? Jul 26 02:39:25 # cat /sys/block/mmcblk0/queue/scheduler Jul 26 02:39:26 noop deadline [cfq] Jul 26 02:45:39 PuffTheMagic: yeah, i can build it **** ENDING LOGGING AT Tue Jul 26 02:59:56 2011