**** BEGIN LOGGING AT Thu Dec 28 02:59:57 2006 Dec 28 07:55:14 Merry Xmas to all ! Dec 28 08:43:02 ~seen mr_mice Dec 28 08:43:15 niv_one_three: i haven't seen 'mr_mice' Dec 28 08:44:04 oh well Dec 28 09:17:34 morning Dec 28 09:32:32 good morning all! Dec 28 09:41:21 hey koen Dec 28 09:45:12 * koen celebrates another year of added wisdom Dec 28 09:45:28 koen: you an older bastard? Dec 28 09:46:36 yep Dec 28 09:46:58 24 years old Dec 28 09:47:19 same age as me wife Dec 28 09:47:31 happy b'day koen Dec 28 09:47:38 * XorA sings Happy Birthday Dec 28 09:47:55 :) Dec 28 09:49:04 so revoke koens keys for the day, so we dont get any drunken checkins :-) Dec 28 09:51:24 XorA: i guess I shouldn't try to update the infrastructure on oe.org to mtn 0.32 :) Dec 28 09:52:03 :-D Dec 28 09:53:54 koen: is that a network protocol change? Dec 28 09:54:23 rwhitby: no Dec 28 09:54:39 rwhitby: see my mails to the list from a few minutes ago :) Dec 28 09:55:02 it adds a extra cache table to the db Dec 28 09:57:20 XorA: how's the work on the remote coming along? Dec 28 09:58:38 koen: I managed to get it to read the buttons last night Dec 28 09:58:44 cool Dec 28 09:58:53 koen: now I need to clean/fix RP's test driver Dec 28 09:59:55 will it be exposed as an extra keypad? Dec 28 10:01:03 it would be neat to map the to those newfangled multimedia keys and have mediaplayers 'just work' Dec 28 10:01:06 (TM) Dec 28 10:02:08 koen: I beleive that is the plan Dec 28 10:03:21 * XorA wonders who chose birmingham over all the decent places in the UK Dec 28 10:05:15 * koen looks at http://gitweb.freedesktop.org/?p=xorg%2Fxserver.git&a=search&h=05f915050cad72d4fb39cbb886be57beeac18749&s=kdrive Dec 28 10:08:35 anything in particular? Dec 28 10:10:12 a bunch or renames Dec 28 10:10:31 I bet a lot of our patches need to get rediffed Dec 28 10:11:35 (as usual) Dec 28 10:12:35 :-D Dec 28 10:14:23 * XorA needs to get round to setting up a spam filter Dec 28 10:15:43 * koen wonders how well Xati will run on his radeons Dec 28 10:15:49 kdrive Xati, that is Dec 28 10:16:25 * XorA gave up on radeons Dec 28 10:35:44 * koen wonders how the 23c3 OE crowd is doing Dec 28 10:36:42 koen: Sleeping? Dec 28 10:37:48 koen: 12:00 here, means congress started only 30 minutes ago. :) Dec 28 10:40:29 koen: http://events.ccc.de/congress/2006/Fahrplan/events/1554.en.html should be the next talk I'll see zecke and mickey|23c3 Dec 28 10:43:12 Is Angstrom planning to use an initramfs? Dec 28 10:48:23 blaster8: use it for what? Dec 28 10:48:46 for booting assistance... Dec 28 10:49:03 i.e. to assist boot from USB/Network etc Dec 28 10:49:43 blaster8: that would be a choice specific to a machine Dec 28 10:50:06 blaster8: I think we should start with de-brainwashing you from working on slugos Dec 28 10:50:10 but there is no neat way of building an initramfs currently in OE... Dec 28 10:50:10 hehe Dec 28 10:50:39 I'm thinking about what work should be done for the next release Dec 28 10:51:01 reduce ipx4xx.conf a 5 lines Dec 28 10:51:13 it contains way too much 'common' stuff Dec 28 10:52:39 do all of Angstrom's current targets purely boot from flash? Dec 28 10:52:41 I seperated out the thumb stuff, so that can be removed now Dec 28 10:52:44 blaster8: nope Dec 28 10:52:58 or other persistent drives... Dec 28 10:53:16 I boot my ep93xx from nfs, my h2200 from mmc, etc Dec 28 10:54:30 eek http://www.openembedded.org/repo/org.openembedded.dev/conf/machine/include/ixp4xx.conf Dec 28 10:55:20 right Dec 28 10:57:16 'I seperated out the thumb stuff, so that can be removed now' - what does this mean exactly? Dec 28 10:57:51 last time I checked in here uClibc building was pretty broken - has that improved? Dec 28 10:58:10 (not suggesting that's your fault, lack of upstream release is very annoying) Dec 28 10:58:39 http://www.openembedded.org/repo/org.openembedded.dev/conf/machine/include/tune-thumb.conf Dec 28 10:59:04 I'm waiting for uclibc .29 before doing any work on uclibc builds Dec 28 10:59:08 same here Dec 28 10:59:32 (actually .30 for the nptl stuff, but .29 is a good start) Dec 28 10:59:51 I'm thinking we may want to pull thumb support... Dec 28 11:00:24 thumb support is policy, the choice to enable/disable it belongs in the distro Dec 28 11:01:31 got it Dec 28 11:01:59 http://ewi546.ewi.utwente.nl/tmp/thumb/comparison.txt <- size comparision between thumb and arm packages Dec 28 11:02:43 pretty damn minimal Dec 28 11:02:43 I wasn't really impressed Dec 28 11:02:51 only samba stands out Dec 28 11:02:58 samba_3.0.14a-r15_armv5te.ipk: 6920 7488 (kB) Dec 28 11:03:06 but for what speed decrease... Dec 28 11:03:43 I'll suggest pulling support for it from SlugOS for now, to enable at least a bit of cleanup Dec 28 11:05:43 slugos and all nslu2 packages need to get looked at if they are really slugos/nslu2 specific Dec 28 11:06:01 the sysconf stuff is usefull for more machines Dec 28 11:06:07 and that's just one example Dec 28 11:06:20 niv_one_three: ping Dec 28 11:11:11 briefly going back to the initramfs thing - it really annoys me Dec 28 11:11:34 there was going to be a minimal libc in the kernel, but development has completely stalled Dec 28 11:12:03 klibc? Dec 28 11:12:04 so to create a minimal initramfs you have to produce a second uclibc toolchain Dec 28 11:12:42 which is non-trivial, especially given that uClibc is old/broken atm Dec 28 11:13:29 and uclibc svn has problems with linux-libc-headers Dec 28 11:13:42 not a great situation... Dec 28 11:13:52 blaster8: dietlibc is also nice Dec 28 11:14:00 heard it was very very buggy Dec 28 11:14:15 blaster8: see line 11 thru 15 of http://www.openembedded.org/repo/org.openembedded.dev/conf/distro/include/angstrom.inc Dec 28 11:14:39 the DEPLOY_DIR trick will save you a lot of headaches Dec 28 11:14:51 ooh Dec 28 11:15:16 otherwise you end up with a mixed glibc/uclibc deploy/ipk dir Dec 28 11:15:56 ok... so can one distro compile two toolchains? in theory and neatly? Dec 28 11:16:02 yes Dec 28 11:16:24 you need to 'inherit multimachine' Dec 28 11:17:17 koen@bitbake:/data/build/koen/OE/build/tmp/angstrom/work$ ls Dec 28 11:17:18 all-angstrom-linux-gnueabi armv5te-angstrom-linux-gnueabi h5000-angstrom-linux-gnueabi i686-linux Dec 28 11:17:51 great Dec 28 11:17:52 uclibc use a 'uclibcgnueabi' as a suffix, so that gets separated out by default Dec 28 11:19:14 so eventually (theoretically) we can come under the angstrom umbrella AND build our two userlands (uclibc/initramfs & glibc/rootfs) Dec 28 11:20:21 yes Dec 28 11:20:35 apart from there being no 'angstrom umbrella' :) Dec 28 11:20:51 well, nearly right Dec 28 11:21:48 then we can remove the two stage flash piggy-back boot thing from SlugOS :) Dec 28 11:22:13 you discover lots of new and exciting stuff when trying to build weird distro/machine combos Dec 28 11:22:21 initramfs makes the decision whether to boot from USB disk (by UUID) or from flash Dec 28 11:23:00 one pivot_root less Dec 28 11:23:27 well, it's switch_root with initramfs, but yeah Dec 28 11:24:21 and initramfs can handle netconsole/nfs booting too, hopefully Dec 28 11:24:45 mickey|23c3 and I were thinking about a initramfs to do a netinstall on machines that could support that Dec 28 11:25:17 initramfs is a very good idea, but it needed to be followed through Dec 28 11:25:44 personally, I reckon that all of which root disk to mount logic should have been moved to userspace Dec 28 11:26:04 and the kernel guys taken over maintainership of this very minimal early userspace Dec 28 11:26:36 that wouldn't work Dec 28 11:26:41 why? Dec 28 11:26:44 just look at the debacle udev is Dec 28 11:26:58 every new release needs a newer kernel Dec 28 11:27:11 that would be ok if udev came with the kernel... Dec 28 11:27:22 and kernel people jsut don't understand that people don't run 2.6.released-yesterday Dec 28 11:28:19 btw, nslu2 bugs are leading the charge on lkml :) Dec 28 11:29:04 http://lkml.org/lkml/2006/12/28/40 Dec 28 11:31:02 mm is apparently a bit of a nightmare Dec 28 11:34:43 blaster8: does the nlsu2 suffer from the cache coherency issues described on LAK? Dec 28 11:37:22 I definitely haven't seen it Dec 28 11:37:29 but I don't see why we should not be affected Dec 28 11:40:25 huhu Dec 28 11:41:14 koen: Nico Pitre's latest post makes for scary reading Dec 28 11:41:48 'I thought that XScale most probably inherited a couple SA odd behaviors … and despite not being documented anywhere doing the same alternating flush regions on XScale made the flush bug go away. So this issue wasn't investigated any further at that point.' Dec 28 11:50:32 koen: how does angstrom handle /dev? Dec 28 11:51:27 I'm not entirely happy with the half-static half-dynamic setup we use in SlugOS atm Dec 28 12:08:52 blaster8: udev, but you need some static mappings Dec 28 12:09:14 blaster8: I lost some hours to find out you need a /dev/console before udev can run Dec 28 12:15:23 thanks Dec 28 12:15:25 gtg Dec 28 12:29:00 hi zecke|23c3 Dec 28 13:09:20 hi. does anyone have experiences with embedded browsers that come along with a late version of the flash plugin? Dec 28 13:42:33 ~hail RP Dec 28 13:42:43 * ibot bows down to RP and chants, "I'M NOT WORTHY!!" Dec 28 13:59:10 koen: :) Dec 28 14:15:39 a stupid question: how do you make a bb file? Dec 28 14:16:07 i want to cross compile a custom version of sdl 1.2.9, and i can't work out how to do it Dec 28 14:16:42 Orkie: a stupid answer would be - with vi :) Well, I started by looking at existing .bb files Dec 28 14:17:00 I think there is a .bb file for sdl already, if it does not match your version you could try adapting it Dec 28 14:17:21 i tried that, but it seemed to build an x86 version instead of arm Dec 28 14:17:42 .bb file just defines how a package should be build in OE, it does not define the architecture Dec 28 14:18:00 you have to select a distro and a target machine Dec 28 14:18:13 oh, i've got other packages compiled for arm fine Dec 28 14:18:40 what has thrown me is that it doesn't seem to use the configure scripts Dec 28 14:18:50 sdl? Dec 28 14:18:52 or what? Dec 28 14:19:05 bitbake Dec 28 14:19:20 it does Dec 28 14:19:33 i know exactly how to cross compile this sdl manually, i just can't work out how to pass bitbake the options it needs Dec 28 14:19:39 it's the inherit autoconf stuff Dec 28 14:20:02 have you looked at the libsdl-x11_1.2.9.bb? Dec 28 14:20:24 you see the EXTRA_OECONF = stuff? those are the parameters for configure Dec 28 14:20:33 morning Dec 28 14:20:39 ah Dec 28 14:20:46 but target architecture and so on is set elsewhere, so I wonder why you are having this problem Dec 28 14:20:49 hi chouimat Dec 28 14:21:02 so i shouldn't need to pass --target etc? Dec 28 14:21:11 no Dec 28 14:21:35 i'll give it another go then Dec 28 14:21:46 you have to select target + distro in your local conf Dec 28 14:21:58 I am trying to do a build using a MasterMakefile, but end up with the error: ERROR: Exception:exceptions.UnboundLocalError Message:local variable 'md5data' referenced before assignment (quilt complains) Dec 28 14:22:06 any tips on that? Dec 28 14:23:35 (first time I try to bake a nslu2 image on my mac) Dec 28 14:26:40 Jelle: running OSX? Dec 28 14:28:01 Jelle: building on OSX is largely untested, have a look at http://www.openembedded.org/buildonosx Dec 28 14:30:00 Jin^eLD: no idea what i've done, but it works now - thanks for your help :D Dec 28 14:30:23 Orkie: no problem :) Dec 28 14:31:45 koen: well, it is time to test it.. :) Dec 28 14:52:43 hi. does anyone have experiences with embedded browsers that come along with a late version of the flash plugin? Dec 28 15:04:19 easy: well seen the closed state of flash, no really available arm-based plugin and current gnash system requirements (same for flash 9) you will soon be out of luck with that. Dec 28 15:05:56 easy: the only linux machine with a browser and flash I know about is the Nokia 770 and it only has flash 6 with opera. And the machine is too slow for some flash animations already (more the fault of flash than anything else) Dec 28 15:06:22 i heard of flash lite. unfortunately, it is not available for linux Dec 28 15:06:56 * philippe thinks flash is evil and should be banned of the net. Dec 28 15:07:14 it's a customer decision :/ Dec 28 15:07:36 thanks for your ideas Dec 28 15:08:11 easy: well there is nothing much you can do except talk to macromedia.... Dec 28 15:08:35 easy: which will cost you lots and lots of money. Hope your customer has that. Dec 28 15:09:25 he'll definitely priorize a free solution :) Dec 28 15:09:44 macromedia even charges per unit Dec 28 15:10:27 03rpurdie 07org.oe.dev * r34a5a68b... 10/ (5 files in 2 dirs): linux-rp: Add xscale_cache_workaround patch to fix segfaults until we find a better solution Dec 28 15:49:50 03koen 07org.oe.dev * re06c1912... 10/ (1 packages/libschedule/libschedule_0.16.bb): libschedle: clean up and add dependency on libgpewidget Dec 28 15:50:00 lalala Dec 28 15:53:50 koen: Angstrom should be good to go with the removal of the Heisenburg now? :) Dec 28 15:54:30 RP: fit for general testing Dec 28 15:55:20 koen: Are there any other known EABI problems? Dec 28 15:55:43 gdb and strace Dec 28 15:56:25 koen: gdb 6.6 compiled for me yesterday. I don't know if its any better than 6.4 for EABI... Dec 28 15:56:27 and shit like the compulab kernel modules Dec 28 15:57:28 RP: http://bugs.openembedded.org/showdependencytree.cgi?id=1573 is still a pretty big list Dec 28 15:57:41 koen: strace has EABI issues even with the patches we have? Dec 28 15:57:56 issues is a big word Dec 28 15:58:05 scary warning about umoven and stuff Dec 28 15:58:52 I'd go through some of these bugs after the cache fix as I think some could be solved by it Dec 28 15:59:00 gpe-info segfault for example Dec 28 16:00:25 The fix solved both Xfbdev segfaults and the infamous loadkeys segfault for me... Dec 28 16:01:16 gpe-info might have weird expectations about pcmcia-cs being present Dec 28 16:01:30 or all sorts of arcane touchscreen interfaces Dec 28 16:01:59 If that was the case, I'd remove it from angstrom images ;-) Dec 28 16:05:10 RP: you rock! Dec 28 16:12:48 zecke|23c3: :) Dec 28 16:12:55 I can think about bitbake hacking now... Dec 28 16:19:21 morning all Dec 28 16:20:04 hi Dirk Dec 28 16:20:30 hey Richard Dec 28 16:20:41 RP: do you have a sec? Dec 28 16:21:11 do13: yes Dec 28 16:21:27 RP: does symlinking eventx -> touchscreen0 works for you? Dec 28 16:21:37 do13: yes Dec 28 16:22:11 do13: Whats the modalias of the device you're having problems with? Dec 28 16:22:18 do13: Also, is it a recent udev? Dec 28 16:22:57 RP: udev is 092 Dec 28 16:23:08 RP: modalias is "input:b0000v0000p0000e0000-e0,3,kra0,1,18,mlsfw" Dec 28 16:24:22 do13: udev 092 is fine. I think its the lack of e1 that might cause it Dec 28 16:24:31 See udev's local.rules file Dec 28 16:24:33 RP: yep Dec 28 16:24:46 RP: if i change this it's working Dec 28 16:25:10 I think that is a button event? Dec 28 16:25:49 EV_KEY Dec 28 16:26:03 The udev rule needs to be relaxed... Dec 28 16:26:18 RP: ok , the corgi driver sends key events Dec 28 16:27:11 do13: Yes, but touchscreens don't have to. It was just based on corgi and I thought they did at the time :) Dec 28 16:27:45 that might be the trouble on htcuniversal as well Dec 28 16:29:41 I'll push a fix Dec 28 16:30:28 RP: thx Dec 28 16:30:41 koen: which probs on thcuniversal? Dec 28 16:31:37 03rpurdie 07org.oe.dev * r17fc8ade... 10/ (7 files in 5 dirs): udev: Update touchscreen rule so it doesn't require the touchscreen to have EV_KEY support Dec 28 16:32:18 do13: no symlink for eventX Dec 28 16:33:25 * RP also updates poky as that's caused me problems there before. I thought that had been fixed a while back tbh... Dec 28 16:33:49 wasn't that something else? Dec 28 16:34:00 I know I relaxed the rule for hx4700 Dec 28 16:34:21 koen: It could have been, yes. I thought we'd made a change once before... Dec 28 16:37:40 http://www.openembedded.org/bonsai/view/rev/17767/ Dec 28 16:38:18 +SUBSYSTEM=="input", KERNEL=="event[0-9]*", SYSFS{modalias}=="input:*-e0,1*,3,*a0,1,*18,*", SYMLINK+="input/touchscreen0" Dec 28 16:38:18 -SUBSYSTEM=="input", KERNEL=="event[0-9]*", SYSFS{modalias}=="input:*-e0,1*,3,*k*14A,*a0,1,*18,*", SYMLINK+="input/touchscreen0" Dec 28 16:39:48 Interesting that the hx4700 advertises EV_KEY but no k* actual key events :} Dec 28 16:40:27 Greetings Dec 28 16:40:33 hey Ifaistos Dec 28 16:40:41 hi Ifaistos Dec 28 16:40:54 Merry Xmass to all Dec 28 16:41:04 koen happy birthday :) Dec 28 16:41:15 Ifaistos: thanks :) Dec 28 16:41:23 koen: When you rebuilt kernels, did you apply the xscale patch to machines outside of linux-rp? Dec 28 16:41:51 RP: no, I'm only rebuilding linux-rp machines Dec 28 16:42:09 koen: Anything xscale is probably open to the bug Dec 28 16:42:13 yeah Dec 28 16:42:29 but I'll leave it to psokolovsky to get the fix into hh.org cvs Dec 28 16:42:47 agnstrom is using 2.6.16 for those machines anyway Dec 28 16:42:52 koen: ok, I just wanted to make sure you knew that :) Dec 28 16:42:53 koen: your birthday today? happy birthday Dec 28 16:42:58 Hi! Dec 28 16:43:00 koen, yep, will commit soon ;-) Dec 28 16:43:06 2.6.16 is actually probably open to the bug too... Dec 28 16:43:08 do13: yes, thanks! Dec 28 16:43:20 koen, Happy Birthday! Dec 28 16:43:20 koen: happy birthday :) Dec 28 16:43:27 koen: how old are you? Dec 28 16:43:27 :) Dec 28 16:43:29 24 Dec 28 16:43:54 Are you counting in hex? Dec 28 16:44:02 heh Dec 28 16:44:10 lol :)) Dec 28 16:46:12 oh, birthday? happy birthday koen :) Dec 28 16:46:22 * Jin^eLD hates getting older. Dec 28 16:46:28 in hex i probably had to shoot myself Dec 28 17:04:31 NAiL, was it you that was working on the QNAP? Dec 28 17:04:55 HopsNBarley: "working" and "working" Dec 28 17:05:04 I haven't touched it for some time. Dec 28 17:05:37 just got my 201 yesterday. nicely made hardware all around. Dec 28 17:05:46 I'm impressed by the 101 Dec 28 17:06:16 (except that two of the screws to fasten the hdd broke right off) Dec 28 17:06:25 we measured the 201 at 20W with two 7200 rpm sata drives spinning. Dec 28 17:07:03 got the console working right away... but u-boot doesn't have any ethernet compiled in. did you flash over the serial line (kermit) Dec 28 17:11:55 HopsNBarley: Haven't flashed. Waiting for a converter so I can debug ;9 Dec 28 17:12:05 Jin^eLD: http://www.angstrom-distribution.org/repo/?action=details&pnm=mediatomb Dec 28 17:13:33 cool! :) allthough I am not sure if you maybe should wait till the release before it becomes part of a distro; this is still dev code and allthough it works it needs quite some testing :> Dec 28 17:21:05 Koen, Happy birthday :-) Dec 28 17:24:36 thanks :) Dec 28 17:35:06 koen : happy bday :) Dec 28 17:35:28 thanks :) Dec 28 17:40:38 happy birthday Dec 28 17:41:03 btw , koen , u got a blog? Dec 28 17:41:29 http://dominion.kabel.utwente.nl/koen/cms/ Dec 28 17:41:34 hey koen, happy birthday. how old? Dec 28 17:41:46 24 years young :) Dec 28 17:46:23 koen, wow! lotta good stuff ahead. Happy Birthday! Dec 28 17:48:05 does ARCH=ppc == powerpc for kernel >= 2.6.15 ? Dec 28 17:48:26 HopsNBarley: sort off Dec 28 17:48:34 HopsNBarley: they are migrating Dec 28 17:48:48 yeah, flat device tree and such. not finished? Dec 28 17:48:57 my efika board has half the stuff in ppc and the other half in powerpc Dec 28 17:48:58 koen: thanks for posting the FIC spec. I know I sound to obsest about it. sorry about it. I guess I have a lousy love life ;-) Dec 28 18:06:15 03pfalcon 07org.oe.dev * r67bd2a01... 10/ (1 conf/distro/angstrom-2007.1.conf): Dec 28 18:06:15 angstrom-2007.1: Add h3900 to supported machines, requires fresh kernel. Dec 28 18:06:15 * Approved by Koen Kooi. Dec 28 18:07:01 is anybody using imagemagick? Dec 28 18:07:39 HopsNBarley: I do occasionally, I think I even tried using their library to code something Dec 28 18:08:15 there are some current build problems. i was looking to partner up to verify fixes. Dec 28 18:08:58 I have to admit I did not try it in OE yet Dec 28 18:09:07 okay. Dec 28 18:21:29 psokolovsky: don't forget FEED_ARCH_h3900 Dec 28 18:22:13 koen, ok! Dec 28 19:17:41 RP, the USB gadget patch on your site for the PXA270 is it pretty stable? I want to use it to make my board look like a mass storage device. It seems to be working for me, however it seems a little slow and also it does not seem to respond to windows when I try to disconnect, I always get a message (on the windows box side) saying I need to wait.. Dec 28 19:19:34 RP, interesting is it seems to work at a nice speed unless I'm mapping it to my mmc (uSD) device. I guess I need to look at that module :-/ Dec 28 19:32:06 does anybody know how to mount a yaffs2 image as loopback? Dec 28 19:32:13 i want to extract a directory tree/file listing from one Dec 28 19:32:32 so far i've managed to build a module for my main pc, but it won't mount Dec 28 19:32:54 jffs2 FSes are hard to loopback; IIRC they need to be mounted on block flash devices.. Dec 28 19:33:25 damn... is there any way of extracting a listing from it without mounting then? Dec 28 19:33:56 Couldn't tell you off the top of my head.. might be worth googling around for the jffs2 project homepage, or alternately, generate jffs2 and tar images Dec 28 19:34:51 it isn't actually my image - it is one i want to look inside to see how the official firmware works Dec 28 19:34:58 Ahh Dec 28 19:35:15 There are ways to do a loopback; I can't remember exactly how though Dec 28 19:35:17 It's a bit hacky Dec 28 19:36:09 i think i may have found something about mtdram. i'll give it a go quickly Dec 28 19:38:04 yeah, you have to emulate a mtd backend Dec 28 19:38:31 * koen mounted a jffs2 file via loopback once Dec 28 19:38:46 3 or 4 years ago Dec 28 19:39:03 * koen eats some more pie Dec 28 19:48:12 * philippe hopes koen has some beer with that pie :-) Dec 28 19:49:04 Gerrath: Also look at the DMA - I've had reports of problems but it might be fixed now Dec 28 19:49:19 03dirk 07org.oe.dev * r3961d9fa... 10/ (1 conf/machine/include/tosa-2.6.conf): conf/machine/include/tosa-2.6.conf: remove obsolete file Dec 28 19:49:48 RP, the DMA for the gadget driver, or the mmc (SD)? Dec 28 19:50:06 Gerrath: gadget driver. I can't speak for your mmc driver Dec 28 19:50:25 RP, I had to disable the DMA for it to work at all. Dec 28 19:52:29 Gerrath: How recent was that code? I made a DMA tweak recently Dec 28 19:55:54 RP, I was using your patch: usb_pxa27x_udc-r1.patch Dec 28 19:56:19 RP, I think I grabbed it about 1 month ago. Dec 28 19:56:52 Gerrath: There's a -r3 there now which some fixes (and updates to make it work properly with 2.6.19) Dec 28 19:57:18 RP, I'm using 2.6.17-rc5 kernel.. I guess I could update my kernel. Dec 28 19:58:09 Gerrath: It should still run with 2.6.17 Dec 28 19:59:00 hi mickey|hotel Dec 28 19:59:27 hey RP Dec 28 19:59:28 RP, ok thanks.. There are so many patches on your site (which is a good thing) I'm never sure which ones to use.. I see there are several for the PXA270 MMC, which I'm using as well. Is there a file (or site) discribing the patches? Dec 28 19:59:48 RP: got any further with that mind boggling VM bug? Dec 28 20:00:01 Gerrath: The linux-rp kernel .bb files are usually a good guide ;-) Dec 28 20:00:43 mickey|hotel: We have a "fix". The bug is at least cornered, pending some thoughts from people who actually understand the Xscale minicache (rmk/nico) ;-) Dec 28 20:01:03 RP, so it is pretty safe to apply all the patches in that file. Dec 28 20:01:31 Gerrath: Apart from the ones marked as hacks, yes Dec 28 20:01:34 RP: ok, that's a very good result Dec 28 20:01:42 RP, thanks again. Dec 28 20:01:51 mickey|hotel: It fixes Angstrom nicely :) Dec 28 20:02:21 mickey|hotel: http://www.rpsys.net/openzaurus/patches/archive/xscale_cache_workaround-r0.patch Dec 28 20:02:34 hey mickey|hotel Dec 28 20:08:18 cheers do13 Dec 28 20:09:35 RP: nasty. why didn't we stumble over this w/ earlier kernels? Dec 28 20:11:07 do13: did you manage to book the same flight? Dec 28 20:11:22 mickey|hotel: Luck? :-/ Dec 28 20:11:42 mickey|hotel: yep. on friday we have the same flight Dec 28 20:11:57 mickey|hotel: I think this access pattern only happens in glibc >= 2.4 Dec 28 20:12:52 do13: great! Dec 28 20:12:55 RP: i see Dec 28 20:13:02 ~seen koen Dec 28 20:13:15 koen was last seen on IRC in channel #oe, 34m 29s ago, saying: '3 or 4 years ago'. Dec 28 20:14:16 btw., anyone know whether we still have the dropbear SIGSEGV in 2.5/OABI? Dec 28 20:14:24 ~hail zecke|23c3 Dec 28 20:14:28 * ibot bows down to zecke|23c3 and chants, "I'M NOT WORTHY!!" Dec 28 20:14:35 mickey|hotel: hey! Dec 28 20:14:40 mickey|hotel: Don't run?! Dec 28 20:14:55 mickey|hotel: what do the thousand lines look like to define a GType? Dec 28 20:15:13 mickey|hotel: Any segfault on xscale with glibc >=2.4 needs retesting with that patch applied :-/ Dec 28 20:16:20 zecke|23c3: i have a newgobject.py, but I need to reboot into Linux to access it. I'll mail it to you later tonite Dec 28 20:16:49 mickey|hotel: with the cache workaround dropbear works Dec 28 20:17:16 mickey|hotel: you are in linux now anyway? you do mails on windows? Dec 28 20:17:30 BTW: mickeyl or his wife made my mobile crash Dec 28 20:17:34 do13: cool, thanks for confirming Dec 28 20:17:50 zecke|23c3: i'm on win atm. *cough* let me reboot into Linux... Dec 28 20:18:02 hehe, no wait :) Dec 28 20:18:03 * mickey|hotel blames his Wifey for bad Windows Mobile charma Dec 28 20:18:06 I can't access mail anyway Dec 28 20:18:57 If anyone finds that workaround patch fixes previous problems, can they please mention it to me. I'm going to need evidence to get mainline to take it seriously Dec 28 20:19:47 If I can say it fixed x different mysertious userspace crashes, that adds weight to my arguments... Dec 28 20:20:03 Particuarly accross two ABIs :) Dec 28 20:20:43 right Dec 28 20:20:56 Crofton_ was the first who identified that dropbear problem on OABI Dec 28 20:26:03 RP: what is the performance impact? Dec 28 20:26:40 mickey|hotel: do you have a offline copy of gtk doc? or can you search me the URL? Dec 28 20:26:54 (glib/gobject doc might make sense as well). The net is crappy ATM Dec 28 20:27:05 zecke|23c3: Shouldn't be too bad. Its certainly lighter than a complete dcache flush Dec 28 20:27:56 zecke|23c3: http://developer.gnome.org/doc/API/2.2/gobject/howto-gobject-code.html might help Dec 28 20:28:18 like i said, i have a newgobject.py that asks you a couple of questions and then generates .c/.h for you Dec 28 20:28:34 mickey|hotel: got a all in one version? I could retry a wget a lot of times Dec 28 20:28:42 mickey|hotel: downloading mail is next to impossible :) Dec 28 20:30:30 can you download http://le-hacker.org/papers/gobject/gobject-book-a4.pdf ? Dec 28 20:30:30 it helped me a lot Dec 28 20:33:13 anjuta contains a gtype wizard as well Dec 28 20:33:17 mickey|hotel: no net, no download :) Dec 28 20:33:50 bummer, how do you IrC then? :) Dec 28 20:35:51 ok, when you have net again, do that: Dec 28 20:35:55 http://burtonini.com/cgi/gobject.py Dec 28 20:36:05 that does roughly the same as my stuff Dec 28 20:36:10 it's a cgi app :) Dec 28 20:41:30 hey mickey|hotel Dec 28 20:43:02 wb ich, did you get my last link= Dec 28 20:43:03 ? Dec 28 20:43:04 yo koen Dec 28 20:43:30 ~praise koen for new fontconfig version Dec 28 20:43:35 All hail koen for new fontconfig version! Dec 28 20:43:41 most of our aliasing problems are gone Dec 28 20:43:59 mickey|hotel: xora added it :) Dec 28 20:45:03 anyway, performance is much better than 4 weeks ago Dec 28 20:45:38 cool Dec 28 20:45:45 * koen ate too much pie Dec 28 20:46:28 RP, the usb_pxa27x_udc-r3.patch fails with the 2.6.17 kernel.. I'll check and see if it is an easy fix. Dec 28 20:47:20 koen: isn't your birthday very near? Dec 28 20:47:36 mickey|hotel: near as in today :) Dec 28 20:47:38 koen: CONGRATS! Dec 28 20:47:49 26 ? Dec 28 20:47:52 24 Dec 28 20:48:07