**** BEGIN LOGGING AT Thu Oct 06 02:59:56 2005 Oct 06 03:07:07 rwhitby: debian as per DiskImageInstall/Jacmet's page Oct 06 03:07:50 how are you building your kernel? just with crosstool? or using make-kpkg ? Oct 06 03:08:01 crostool Oct 06 03:08:15 make ARCH=arm CROSS.... Oct 06 03:09:28 so would you like to have a CVS repository to keep your patchset in? Oct 06 03:11:20 if you think it could be useful, yes. Oct 06 03:12:12 I'm thinking of a repository on nslu.sf.net, where kernel patches are kept, and are used both by people building kernels by hand, and also by the OE build stuff in monotone. Oct 06 03:12:38 cause the people building debian by hand have no interest in running monotone or OpenEmbedded. Oct 06 03:12:59 ok Oct 06 03:16:24 ok, find your sf.net id, and I'll get you write access Oct 06 03:16:48 found: azummo . i' just need to work out the password :) Oct 06 03:20:41 dwery: you now have write access to nslu project on sf.net Oct 06 03:20:57 thanks! Oct 06 03:21:16 i have to go now, we'll talk later about how to handle the patchset Oct 06 03:21:25 I'll set up a "kernel" module. Any ideas on directory structure under there? we'll want to support both BE and LE kernels, and potentially quite a few kernel versions in parallel Oct 06 03:22:00 maybe we'll mirror what jbowler set up in the nslu2-kernel directory in monotone ... Oct 06 03:31:40 it should be ok.. see ya later! Oct 06 03:41:06 03koen 07org.openembedded.dev * re456a231... 10/packages/meta/ (gpe-image.bb meta-gpe.bb): Oct 06 03:41:06 packages/meta/: image tweaks Oct 06 03:41:06 * make gpe-image use ROOT_FLASH_SIZE to determine what to install Oct 06 03:41:06 * introduce gpe-task-connectivity Oct 06 03:41:06 * ship gpe version of industrial Oct 06 04:16:32 03rwhitby * 10unslung/ (make/ufsd.mk sources/ufsd/postinst): Updated ufsd to match new firmware file from Linksys Oct 06 05:01:05 03koen 07org.openembedded.dev * r6d6af9eb... 10/packages/matchbox-panel/matchbox-panel_0.9.2.bb: packages/matchbox-panel/matchbox-panel_0.9.2.bb: use small icons for h2200 Oct 06 05:16:04 03koen 07org.openembedded.dev * r7e26dca9... 10/packages/meta/meta-gpe.bb: packages/meta/meta-gpe.bb: add RDEPENDS for gpe-task-connectivity Oct 06 05:40:12 dwery-away: pong? Oct 06 05:41:22 Jacmet: ping Oct 06 05:42:28 Jacmet: I'm gonna set up a "kernel" cvs repository on nslu.sf.net so that we can put all the nslu2-linux kernel patches there, and then use that same patchset for openslug, ucslugc, le debian and be debian. Oct 06 05:56:12 03koen 07org.openembedded.dev * red3a73f5... 10/packages/gpe-bootsplash/ (files/cairofix.patch gpe-bootsplash_1.14.bb): packages/gpe-bootsplash/gpe-bootsplash_1.14.bb: add patch to make it compile against the cairo we are using in gpe-2.7 Oct 06 06:29:01 dwery-away: cvs repository has been created - see the post in nslu2-developers Oct 06 06:34:24 rwhitby-asleep: ok, sounds good Oct 06 06:36:34 Jacmet: feel free to put your 2.6.11 patches in there, as long as they work for both BE and LE (like the 2.6.14 patches do) Oct 06 06:38:25 rwhitby-asleep: about the directory structure Oct 06 06:38:49 i haven't see it yet Oct 06 06:38:57 bu i would have a directory for each kernel release Oct 06 06:39:37 that's what I have done Oct 06 06:39:42 great. Oct 06 06:39:56 how to handle successive patchsets for the same kernel? Oct 06 06:40:25 cvs tags Oct 06 06:40:46 ok Oct 06 06:40:49 * NAiL tries building a kernel with the latest patchset Oct 06 06:40:59 NAiL: go! go! go! :-D Oct 06 06:41:15 hehe Oct 06 06:41:37 the #4 patch set has all the patches needed, right (except the anonymizer patch)? Oct 06 06:41:42 yes Oct 06 06:41:50 I also saw the latest copy-from from John Oct 06 06:41:55 you may want to test it Oct 06 06:42:19 where can I find that? Oct 06 06:42:23 dwery: #4 worked quite fine for me in le + dot eth working :-) Oct 06 06:42:47 I'm not trying LE though Oct 06 06:42:50 yvasilev: great. fo you have a cleaner build environment/patchset for the eth/ial? Oct 06 06:44:54 dwery: no, for now just tried gulsnepo's one with some additional tiny modifications to make them compile Oct 06 06:45:18 yvasilev: ok. Oct 06 06:45:28 yvasilev, Jacmet: what about the latest copy from? Oct 06 06:46:13 03koen 07org.openembedded.dev * r43606e02... 10/packages/gpe-bootsplash/files/ (splash-l.svg splash-p.svg): packages/gpe-bootsplash/files/splash-*.svg: update svgs to gpe 2.7 Oct 06 06:46:19 03koen 07org.openembedded.dev * r88df4b08... 10/packages/meta/meta-gpe.bb: packages/meta/meta-gpe.bb: add gpe-bootsplash back to gpe-base Oct 06 06:46:27 dwery: but this drivers still have a problem of not working well for addresses outside my lan Oct 06 06:46:48 i was unable to use dns Oct 06 06:46:54 i think there's a problem with udp Oct 06 06:48:16 no, for me sometimes it worked, others (by far the most common case) did not, the same with http Oct 06 06:48:37 dns mostly uses udp, but sometimes tcp Oct 06 06:48:51 did you changed that "stamp" line in the eth code? Oct 06 06:49:15 the one form your patchset? yes Oct 06 06:49:57 it wasn't mine ;) Oct 06 06:50:54 actually udp in lan is ok, nfs works w/o problems Oct 06 06:51:34 it's more like it discards routed packets Oct 06 06:51:56 quite strange.. Oct 06 06:52:10 yup Oct 06 06:54:13 flashing kernel Oct 06 06:55:54 NAiL: do you flash the kernel at each test? Oct 06 06:56:21 Linux slugave 2.6.14-rc3 #1 Thu Oct 6 15:45:31 CEST 2005 armv5teb unknown unknown GNU/Linux Oct 06 06:57:13 now I do, since I don't have serial Oct 06 06:57:46 you can download it with telnet Oct 06 06:57:47 dwery: #4 gives quite allot of warning because if (unfortunately I didn't logged it, but can reproduce if needed) some function disregarding qualifiers in ixp4xx.h (or something like that) Oct 06 06:58:19 yep.. they have been there since a while, never looked at them Oct 06 06:58:22 i'll do later Oct 06 06:58:36 hmm Oct 06 06:58:44 Can't see that I notice any difference Oct 06 06:59:06 did you compiled the speaker code as a module? Oct 06 06:59:47 probably not Oct 06 06:59:59 actually I didn't modify the config at all Oct 06 07:00:41 ok, can you compile it now? it is in drivers/input/misc Oct 06 07:00:45 NAiL: I have made a nice little script that uses netcat to telnet and sends all the needed commands to load and boot the kernel/initrd, if you want it I can send it to you Oct 06 07:00:45 in menuconfig Oct 06 07:01:21 yvasilev: cool, yes please Oct 06 07:01:22 :) Oct 06 07:01:33 to what address? Oct 06 07:01:59 nail at nslu2-linux.org Oct 06 07:02:35 ok, will clean it a little and send it to you in a moment Oct 06 07:03:18 thx Oct 06 07:16:28 NAiL: sent Oct 06 07:16:54 thx Oct 06 07:18:42 you can pass it parametrs form command line like `TFTP=aa KERNEL=bb INITRD=cc ./slugboot.sh` or modify the file (and I sincerely hope that I didn't brake it now with this cleanup) Oct 06 07:26:44 NAiL: modules ok? Oct 06 07:28:43 haven't gotten around to it, making food :P Oct 06 07:29:17 :) Oct 06 07:31:07 03koen 07org.openembedded.dev * re49417c7... 10/packages/meta/gpe-image.bb: packages/meta/gpe-image.bb: sort out DEPENDS vs. INSTALL business with tasks Oct 06 08:14:47 jbot_, test Oct 06 08:21:05 03koen 07org.openembedded.dev * r982dea7f... 10/packages/meta/gpe-image.bb: packages/meta/gpe-image.bb: it now builds all deps Oct 06 08:24:16 jbot_, test Oct 06 08:31:04 03rw 07org.openembedded.dev * rc1171d13... 10/classes/kernel.bbclass: kernel.bbclass: don't run 'make dep' for 2.6 kernels Oct 06 08:42:19 OT: Anybody has a digitizing tablet to suggest that would work under linux? Oct 06 08:46:08 VoodooZ_Log: Wacom usually works fine Oct 06 08:46:35 really? cool. I'm currently looking at their page. Oct 06 08:46:42 Any specific model to recommend? Oct 06 08:47:40 I've a graphire 3 Oct 06 08:47:57 or sapphire.. can't remember Oct 06 08:48:18 the code is CTE-430/B Oct 06 08:48:19 Is this USB or ?? Which linux distro are you using it with? Oct 06 08:48:23 usb Oct 06 08:48:24 debian Oct 06 08:48:41 Is it a hack job or there's binary actual drivers? Oct 06 08:48:43 I do no use it very much but it used to work quite fine Oct 06 08:49:06 it has been reve3rse engineered iirc, driver comes with X. Oct 06 08:49:09 Cool. Thanks man. You saved a lot work. Oct 06 08:49:36 np Oct 06 08:54:43 dwery, In fact they even mention the open-source driver on their support page. Oct 06 08:55:06 wow Oct 06 08:55:41 they say that Intuos2, graphire2 & 3 and Cintiq models are supported so it's not too shabby. Oct 06 08:55:47 Thanks for the lead! Oct 06 09:09:23 rwhitby-asleep: the 2.6.14 patches basically contains everything the 2.6.11 ones did (except for flash) - I'llcheck in the flash stuff as well Oct 06 09:09:38 rwhitby-asleep: Oct 06 09:09:42 rwhitby-asleep: ping Oct 06 09:18:27 anybody with hotplug experience ? Oct 06 09:19:11 I have a script that should be started when an usb camera/flash is connected to the slug Oct 06 09:19:43 this works like a charm as /etc/hotplug/usb/usb-storage on my regular debian mahcine Oct 06 09:20:06 thug slug however does not react Oct 06 09:20:18 any ideas ? Oct 06 09:21:04 03koen 07org.openembedded.dev * r42c5e248... 10/packages/gpe-bootsplash/ (files/splash-l.svg files/splash-p.svg gpe-bootsplash_1.14.bb): packages/gpe-bootsplash/gpe-bootsplash_1.14.bb: ship working svg files and bump PR Oct 06 09:21:24 (the openslug kernel does not have usb-storage compiled as a module, could that be the problem ?) Oct 06 09:47:43 hi all Oct 06 09:51:56 I've hit some problems compiling postgresql on the slug and wanted to try it again with another gcc version. I've set a PREFERED_VERSION_gcc ?= 3.3.4 in an attempt to bitbake a new native gcc package. But the compile stpes always fall on the libliberty header file: openslug/tmp/staging/armeb-linux/include/libiberty.h:96: error: syntax error before Oct 06 09:51:56 "ATTRIBUTE_SENTINEL" Oct 06 09:52:08 any hints? Oct 06 10:31:11 03frederic 07org.openembedded.dev * re5ed779b... 10/packages/libsdl/ (3 files in 2 dirs): Oct 06 10:31:11 libsdl: disable building static lib, don't install libtool .la file Oct 06 10:31:11 libsdl-qpe: fix sdl-config flags for compatibility with libsdl-x11 Oct 06 12:25:25 eno-away, ping? Oct 06 12:28:05 03jp30 * 10unslung/make/ (py-celementtree.mk py-elementtree.mk py-kid.mk): optware: correct missing underscores in several python module ipk names Oct 06 12:29:36 03jp30 * 10unslung/make/py-sqlobject.mk: optware: correct missing underscores in several python module ipk names Oct 06 12:49:22 Jacmet: ping Oct 06 13:46:12 03jbowler 07org.openembedded.dev * r1b00e951... 10/ (2 files in 2 dirs): Oct 06 13:46:13 ucslugc: le/glibc version: fix ucslugc-native for gcc and remove patcher Oct 06 13:46:13 Patcher is no longer required, so is not in ucslugc-packages.conf Oct 06 13:48:25 why I am getting this error when building nativly for os?: Oct 06 13:48:26 /usr/lib/gcc/armeb-linux/3.4.4/../../../../armeb-linux/bin/ld: ERROR: .libs/ao_aixs.o uses hardware FP, whereas .libs/libao.so.2.1.3 uses software FP Oct 06 13:48:26 /usr/lib/gcc/armeb-linux/3.4.4/../../../../armeb-linux/bin/ld: failed to merge target specific data of file .libs/ao_aixs.o Oct 06 13:50:49 cheef_daniel: you got this while building natively ? Oct 06 13:52:22 yes Oct 06 13:52:40 -->=directly on nslu2 Oct 06 13:53:05 and on many different sources/packages Oct 06 13:53:46 it is not only ao_aixes.o Oct 06 13:54:00 did you compile gcc yourself (on a desktop) ? Oct 06 13:54:05 no Oct 06 13:54:23 its the one from openslug 2.7 Oct 06 13:55:49 another example: Oct 06 13:55:51 gcc -g -O2 -W -Wall -O3 -o irmp3d irmp3d.o irmp3mod.o ../commommon.a mod_lcdproc.o mod_playlist.o mod_playlistlist.o mod_cd.o mod_mixer.o mod_netctl.o mod_env.o rm.o mod_idle.o mod_timer.o p.o mod_autoshutdown.o mod_debug.o wser.o mod_fifo.o Oct 06 13:55:51 /usr/lib/gcc/armeb-linux/3.4.4/../../../../armeb-linux/bin/ld: ERproc.o uses hardware FP, whereas irmp3d uses software FP Oct 06 13:55:51 /usr/lib/gcc/armeb-linux/3.4.4/../../../../armeb-linux/bin/ld: fae target specific data of file mod_lcdproc.o Oct 06 13:55:52 /usr/lib/gcc/armeb-linux/3.4.4/../../../../armeb-linux/bin/ld: ERylist.o uses hardware FP, whereas irmp3d uses software FP Oct 06 13:55:55 /usr/lib/gcc/armeb-linux/3.4.4/../../../../armeb-linux/bin/ld: fae target specific data of file mod_playlist.o Oct 06 13:56:29 it is from a totaly different source Oct 06 13:57:11 cheef_daniel: try gcc -msoft-float ... Oct 06 13:58:40 while we are at it... Oct 06 13:59:08 anybody successful with -mthumb ? Oct 06 13:59:39 opendebianslug, gcc 3.4.4 Oct 06 14:01:45 a77ila: how can I specify -msoft-float global? Oct 06 14:05:00 ah with CFLAGS Oct 06 14:06:57 cheef_daniel, really, you should read more and ask less. Oct 06 14:07:07 I know Oct 06 14:19:31 how can I find out the .config of a certain openslug kernel ? It doesn't seem to be in the kernel-image ipkg Oct 06 14:30:34 a77ila: it's in the monotone repo Oct 06 14:47:04 rwhitby: yes, yes, I thought there was a simpler way of getting it Oct 06 14:47:41 and it's still no guarantee that a slug has THAT config Oct 06 14:49:44 there is a linux kernel option that causes .config to embedded in the kernel binary - is that the sort of thing you want? Oct 06 14:51:02 either that or have a /boot/config sort of thing like the 'regular' Debian has Oct 06 14:55:18 a77ila: if you can work out a way to get OpenEmbedded to do that, then please feel free. Oct 06 15:06:04 rwhitby: not much to do actually Oct 06 15:06:15 rwhitby: CONFIG_IKCONFIG=y Oct 06 15:06:39 rwhitby: that's it - the downside being that we wasted 6.5K of memory and flash space Oct 06 15:07:25 well, since our official config doesn't change that often, then perhaps anyone who changes their own config can just enable that too ? Oct 06 15:08:56 well yes, but this is not an option that pokes you in the eye to make sure you notice it Oct 06 15:09:35 and you might end up with a kernel on the slug for which you're not sure what the config was Oct 06 15:10:07 or an older version that would need you to dig deeper into the repo Oct 06 15:10:39 it most certainly is not a must, but usually when you need it you need it REAL BAD :) Oct 06 15:13:20 a77ila: that's true. Oct 06 15:13:53 a77ila: make a proposal on nslu2-developers. If no-one objects to the extra 6.5K, then we'll put it in. Oct 06 15:14:08 ok Oct 06 15:14:34 thanks for looking into it. Oct 06 15:47:30 this hotplug this is however going to be the end of me... Oct 06 15:47:52 what hotplug thing? Oct 06 15:47:52 s/\sthis/thing/ Oct 06 15:48:01 heh Oct 06 15:48:12 I didn't even notice the spelling error :P Oct 06 15:48:23 won't start user scripts Oct 06 15:48:46 where are the scripts? Oct 06 15:49:14 /etc/hotplug/usb Oct 06 15:49:48 place 'em in /etc/hotplug.d/usb ;) Oct 06 15:49:50 I reckon this might be because usb-storage is not a module Oct 06 15:50:03 aaah... Oct 06 15:50:16 ;) Oct 06 15:50:21 this openslug/debianslug dichotomy is terrible :) Oct 06 15:50:28 hehe Oct 06 15:51:45 however this means that any deb package that relies on hotplug WILL be broken ? Oct 06 15:54:42 hmm Oct 06 15:55:37 the easy fix is probably ln -s /etc/hotplug.d/ /etc/hotplug Oct 06 15:55:46 ;) Oct 06 15:55:55 or replacing hotplug Oct 06 15:56:04 been there done that no joy Oct 06 15:56:40 replace hotplug with what ? that will also break a bunch of dependancies :( Oct 06 15:57:13 are you talking debian now? Oct 06 15:57:29 (or debonaras) Oct 06 15:57:32 yes Oct 06 15:57:40 debonaras Oct 06 15:57:42 sorry Oct 06 15:57:56 well, the hotplug supplied by debonaras should look in the correct dir... Oct 06 15:58:09 yes, I just checked it Oct 06 15:58:18 it DOES look in /etc/hotplug Oct 06 15:58:23 like it should Oct 06 15:58:26 Oct 6 23:27:53 slug scsi.agent[1103]: sd_mod: can't be loaded (for disk) Oct 06 15:58:29 hmm... Oct 06 15:58:31 aha Oct 06 15:58:58 however, then why arent the usb modules called ? Oct 06 15:59:14 s/modules/user scripts/ Oct 06 16:00:27 dunno Oct 06 16:00:35 maybe I'm just attaching my script to the wrong hook (usb-storage), but I'm not sure what else is really loaded on storage connect Oct 06 16:00:56 scsi stuff Oct 06 16:02:17 hm, worth a try Oct 06 16:19:59 rwhitby: ping Oct 06 16:21:39 dwery: gu Oct 06 16:21:59 rwhitby: hi.. I would like to talk about the mtd partitions Oct 06 16:22:04 sure Oct 06 16:22:24 others are invited to speak of course.. Oct 06 16:22:33 if I understood the problem correctly, Oct 06 16:22:47 the FIS and partitions are stored in the flash in BE way. Oct 06 16:22:49 right? Oct 06 16:24:09 if you mean the entries in the last flash block, then yes, that's how our current image creation tools do it Oct 06 16:24:45 and that also the way the stock nslu2 is, right? Oct 06 16:25:01 stock firmware slugs do not have a partition table in the last block of flash. Oct 06 16:25:23 the only have partitions? Oct 06 16:25:31 withtou the FIS directory? Oct 06 16:25:40 linksys's firmware developers (sercomm) didn't implement redboot partition table parsing in the kernel - they have hard-coded partitions. Oct 06 16:26:10 I actually have a partition tablem but can;t remeber if it has been created when i trien openslug.. is that possible?> Oct 06 16:26:14 the existence of the partition table was introduced by nslu2-linux firmware Oct 06 16:26:39 it would have been created whenever you install any nslu2-linux firmware, at the time of flashing. Oct 06 16:27:13 and when you flash stock firmware back on the device, it disappears again Oct 06 16:27:29 ok. should we create always in BE and let the LE patch handle the conversion or Oct 06 16:27:42 create based on the endiannes? Oct 06 16:29:13 if redboot actually ready it (which it doesn't) then I would say the former. Since neither redboot or stock firmware reads it, then I would say the latter. Oct 06 16:29:23 s/ready/read/ Oct 06 16:30:02 what if an LE user want to retrieve the data that is on his stock slug? Oct 06 16:30:32 there is no data in that block (apart from the SerComm signature) on a stock slug. Oct 06 16:31:26 ok. regarding the data on the partitions? Oct 06 16:31:47 that will be LE anyway, won't it? Oct 06 16:32:09 but there's existing data... Oct 06 16:32:09 ? Oct 06 16:32:41 I'm missing something - can you explain the steps the user took to get to that point, and what they want to read? Oct 06 16:33:06 let's say the user launched an LE kernel Oct 06 16:33:10 via tftp Oct 06 16:33:17 in memory Oct 06 16:33:21 without writing it to flash Oct 06 16:33:33 which should be the common test scenario Oct 06 16:33:36 for a debian system Oct 06 16:33:55 FIS would be empty Oct 06 16:34:44 then they can't read the data anyway, unless the fall-back built-in FIS table hard-coded in the kernel matches the layout in flash. We don't support that. Oct 06 16:35:25 ok.. too much work that isn' probably strictly needed. Oct 06 16:36:03 so the FIS directory will be created during installation by either debian installer or open slug or whatever? Oct 06 16:36:43 FIS directory is created as part of the flashing of the full image. Oct 06 16:37:00 and it should match the endianness of the other things in that full image Oct 06 16:37:50 *rofl* Oct 06 16:38:01 * NAiL just hit a snag Oct 06 16:38:06 USB Mass Storage support registered. Oct 06 16:38:06 i2c /dev entries driver Oct 06 16:38:27 right after that, I apparently trigger the PowerOff GPIO Oct 06 16:38:27 rwhitby: can FIS be created afterwards? Oct 06 16:39:24 rwhitby: given that our users has not yet writed the full image.. he's just trying things Oct 06 16:39:30 s/users/user/ Oct 06 16:39:41 dwery: no, you can't try things without writing a full image to flash once Oct 06 16:40:08 (at least from a stock slug) Oct 06 16:40:43 ok, that should be anough. Oct 06 16:41:01 for a non-stock slug, then you might get lucky and have a partition table that matches what your init scripts and such are expecting (i.e. openslug turnup requires mtd4 to be the jffs rootfs), but you can't depend on that. Oct 06 16:41:02 regarding the cvs Oct 06 16:41:21 wouldn't be better to have a 2.6.14-rc3 directory? Oct 06 16:42:01 dwery: do we really expect our patches to change much between release candidates on a kernel version? Oct 06 16:42:17 sometimes yes, sometimes not Oct 06 16:42:38 and we wouldn't be supporting two release candidates of the same kernel at the same time. we would have a stable previous version, and an unstable latest rc version. Oct 06 16:43:15 debian, for instance, doesn't even write 2.6.12.2 - they just go to 2.6.12 Oct 06 16:43:16 mmm ok Oct 06 16:43:30 I found a little bug in patchset #4 Oct 06 16:43:40 and use version numbers after that (and we would use CVS tags for that) Oct 06 16:43:57 so i'm going to fix it Oct 06 16:44:06 remove the files that john suggested to remove Oct 06 16:44:40 and retag it. Oct 06 16:44:42 correct? Oct 06 16:45:00 yes Oct 06 16:45:19 great :) Oct 06 16:45:38 I think I used the tag "kernel_20051006" and we can always add a,b,c, on the end for multiple tags in the one day Oct 06 16:46:01 but we could also start a version-based tag convention if you like Oct 06 16:46:24 like 2.6.14-rc3-5 you mean? Oct 06 16:46:44 yeah, except you can't have dots or dashes in cvs tags - only underscores Oct 06 16:46:50 (and it can't start with a number) Oct 06 16:47:24 actually, tag it with NSLU2_2_6_14_rc3_5 Oct 06 16:47:31 then we can have DS101 tags too Oct 06 16:47:50 ok Oct 06 16:47:54 for the huge amount of patches that I'll commit.. or something... Oct 06 16:48:03 I assume you're a seasoned CVS user ... Oct 06 16:48:56 NAiL: you may just want to tag existing files for the set that you're trying at one point in time Oct 06 16:49:10 yeah Oct 06 16:49:11 rwhitby: I used cvs but need to restudy it a little bit Oct 06 16:49:13 to give yourself a stable baseline to work from Oct 06 16:49:24 i'm quite sure I used dots in some tags a while ago :0 Oct 06 16:49:35 I was quite sure.. Oct 06 16:49:35 hmm - maybe I'm mistaken then Oct 06 16:49:50 no probably it's me.. can't remember Oct 06 16:49:55 maybe it was just dashes Oct 06 16:50:22 the one that's checked in now is patcjset #4? Oct 06 16:51:30 dwery: yeah, the one which is checked in is from your email labelled patchset #4, but the file was named -rc2 Oct 06 16:51:48 mmm... i must have made some mistakes then :) Oct 06 16:52:16 the file was named nslu2-2.6.14-rc2.tar.bz2 Oct 06 16:52:39 email subject was "2.6.14-rc3 patchset #4 " Oct 06 16:52:47 doh! i sent the wrong one! Oct 06 16:53:15 that's way john talked about files i thought weren't there anymore! Oct 06 16:53:18 i'm very sorry Oct 06 16:53:21 dwery: ok, I'll leave it to you to fix it up in the CVS then :-) Oct 06 16:53:26 ok :) Oct 06 16:53:33 hmm... how could I figure out the IRQ assignments on the ds101 again? Oct 06 16:53:59 NAiL: it's on the gpio wiki page isn't it? Oct 06 16:54:27 no, not which IRQ Oct 06 16:54:37 dwery: can you also generate a diff from rc2 (which John last merged from) to the latest in CVS, and send it to John? Oct 06 16:54:51 INT[ABCD] are there Oct 06 16:54:54 (latest in CVS being after you check in the correct one) Oct 06 16:54:58 ok Oct 06 16:55:07 but what irq PB/RB generates? Oct 06 16:55:37 NAiL: is that in the bootlog? or in the ds101 source code? Oct 06 16:59:48 jbot, test Oct 06 16:59:51 Oh, no! There's a test and I haven't studied! Oct 06 17:02:23 we have said we will not have defconfigs , right? Oct 06 17:03:40 right - we should have a nslu2_defconfig in the normal place, put there by a patch. Oct 06 17:04:33 I remember John was against that.. iirc Oct 06 17:04:47 against what? Oct 06 17:05:25 having nslu2_defconfig Oct 06 17:06:00 I thought every arch *had* to have an _defconfig in arch/arm/configs/... Oct 06 17:06:23 why should we be different? Oct 06 17:06:50 dunno :) Oct 06 17:07:04 maybe I have misinterpreted some mails.. Oct 06 17:07:40 of course, every person using this patchset will probably have their own defconfigs in their own build system, but we do need to have a default one which is as close as possible to the existing ixp4xx_defconfig, but which makes sense for the nslu2. Oct 06 17:09:16 ok Oct 06 17:09:42 03azummo * 10kernel/2.6.14/ (9 files): 2.6.14-rc3 patchset #4, the correct one. Oct 06 17:11:29 oh.. nice feature :0 Oct 06 17:11:40 03azummo * 10kernel/2.6.14/ (25-nslu2-arch-reset.patch 60-nslu2-beeper.patch): Those twos were still missing from 2.6.14-rc3 patchset 4 Oct 06 17:12:56 dwery: if we want to, we can have distro directories (like debian, ucslugc, openslug, etc) for the different defconfig files. Not sure whether they would be alongside the kernel version directories, or underneath each kernel version directory ... Oct 06 17:13:19 let's see how different will they be.. Oct 06 17:13:59 I know that ucslugc will be different (due to the thumb stuff), so that might have been the root of john's concern about it. Oct 06 17:14:28 but we must have a standard nslu2_defconfig for the upstream patch Oct 06 17:14:35 ack Oct 06 17:16:13 dwery: are you happy to be the one which sends the patches upstream? Oct 06 17:16:39 uhmm.. ehr.. wait, let me think about that :-D Oct 06 17:16:56 well, you seem to have had good success so far ... Oct 06 17:17:10 just one got approved... Oct 06 17:17:24 and it was a one-liner :-D Oct 06 17:18:10 anyway yes, I can do that.. :) Oct 06 17:18:15 that's a 100% success rate then ... Oct 06 17:18:18 :-D Oct 06 17:18:46 and the people doing the work of preparing the patches deserve the credit Oct 06 17:18:55 of course! Oct 06 17:19:00 the next one would be 90-ixp4xx-pci-le.patch Oct 06 17:19:06 which Jacmet already proposed Oct 06 17:19:19 ok, make sure there is a good description in the patch Oct 06 17:19:25 that's the problem. Oct 06 17:19:34 basically, this patch is against intel specs Oct 06 17:19:39 if we have that in the CVS, we can have other people contribute to the description Oct 06 17:19:45 if i've understood correctly Oct 06 17:20:11 but Jacmet told me things did not worked without it Oct 06 17:20:35 empirical evidence is sufficient :-) Post the description to the list and ask for input. Oct 06 17:21:07 wasn't there some slashdot story about Linus and specs recently ;-) Oct 06 17:21:28 ok Oct 06 17:21:42 yes, I saw it ,,, basically specs are useles.. Oct 06 17:22:00 ok, I gotta go soon. anything else I can help with? Oct 06 17:22:41 At least now we know that upstream are interested and are waiting for these patches, so it should be easier to get them approved ... Oct 06 17:23:03 "A "spec" is close to useless. I have _never_ seen a spec that was both big enough to be useful _and_ accurate. Oct 06 17:23:07 And I have seen _lots_ of total crap work that was based on specs. It's _the_ single worst way to write software, because it by definition means that the software was written to match theory, not reality. " -- linus Oct 06 17:23:08 and we have agreement onthe leds now too :-) Oct 06 17:23:18 :) i think you can go then :) Oct 06 17:23:22 thansk Oct 06 17:23:24 thanks Oct 06 17:23:52 no, thank *you*. Without your efforts, the upstream patch effort would still be stagnant Oct 06 17:24:18 it's your energy and ability which is driving it forward now Oct 06 17:24:33 (and the efforts of others as well, of course) Oct 06 17:25:13 back later Oct 06 17:34:03 03azummo * 10kernel/2.6.14/50-nslu2-general.patch: Oct 06 17:34:03 2.6.14-rc3 patchset #5 Oct 06 17:34:03 - wrong irq setup in nslu2-power.c Oct 06 17:38:01 Anyone volunteering to help me a bit with the ds101 kernel? :D Oct 06 17:38:25 once I get one, yes. Oct 06 17:38:41 Anyone volunteering to buy me a ds101 to help me help NAiL? :D Oct 06 17:38:54 hehe Oct 06 17:40:13 dwery: It's *almost* the same as a slug ;) Oct 06 17:40:27 ok, let's try then :) Oct 06 17:41:32 The GPIO stuff is different, as are (possibly?) the IRQs Oct 06 17:41:43 ds101: http://www.nslu2-linux.org/wiki/DS101/GPIOConnections Oct 06 17:41:45 nslu2: http://www.nslu2-linux.org/wiki/Info/GPIOConnections Oct 06 17:42:07 ok Oct 06 17:43:04 There's one other difference, and that's the IDE-interface on irq 28 Oct 06 17:43:29 I have it *kindaworking*, in that it identified the chip, and sees the hdd Oct 06 17:43:44 but it looses a bunch of IRQs, so something isn't right Oct 06 17:45:44 you mean the software doesn't see ints from the chip? Oct 06 17:46:17 ide2 at 0x1050-0x1057,0x1062 on irq 28 Oct 06 17:46:17 hde: max request size: 1024KiB Oct 06 17:46:17 hde: lost interrupt Oct 06 17:46:28 a few more lost interrupts, then : Oct 06 17:46:28 hde: 312581808 sectors (160041 MB) w/8192KiB Cache, CHS=19457/255/63 Oct 06 17:46:29 hde: lost interrupt Oct 06 17:46:31 etc.. Oct 06 17:47:23 maybe it th ekind of interrupt.. i.e. level triggered, edge triggeredn, active low, active high.. Oct 06 17:47:36 ugh Oct 06 17:47:43 that's kind beyond me yet Oct 06 17:47:45 you have to check the docs of the chip Oct 06 17:47:54 to see which kind if int they use Oct 06 17:49:01 dwery: we have the source code of the stock kernel - it should all be in there if you know where to look ... (which I don't easily) Oct 06 17:49:55 NAiL: are you just recompiling the stock kernel or forward-porting the patches? Oct 06 17:50:05 dwery: none of the above Oct 06 17:50:31 I was hoping to adapt the nslu2 patches, more or less Oct 06 17:51:31 which patches? Oct 06 17:52:17 the ones that would be necessary to get it to work :P Oct 06 17:53:16 ok, I think i've understood :) Oct 06 17:53:27 can you point me at the stock kernel? Oct 06 17:53:53 it's a rather huge download Oct 06 17:54:04 ftp://gpl.synology.com/gpl/ds101-200.tgz Oct 06 17:54:10 some 100+mb Oct 06 17:54:40 mm.. it would be better if you can produce a diff between the stock linux kernel and the ds101 one, same version Oct 06 17:54:48 uh Oct 06 17:54:55 that's gonna be a *sizeable* diff Oct 06 17:54:58 rwhitby-treo: are the developer cvs and the anonymous one separate? Oct 06 17:55:12 the kernel has been heavily modified Oct 06 17:55:25 NAiL: well, yes.. but it;'s required Oct 06 17:56:16 ok, I'll download a stock kernel Oct 06 17:57:02 dwery: there is an up to 5 hour delay Oct 06 17:57:10 rwhitby-uh, ok :) Oct 06 17:57:37 all the nslu2-linux developers have developer access Oct 06 18:00:51 diffing Oct 06 18:01:39 diff is 3.6mb... not half as much as I expected Oct 06 18:02:53 still a lot Oct 06 18:03:13 you may wanto to navigate it with midnight commander Oct 06 18:06:26 haha Oct 06 18:06:37 ? Oct 06 18:06:38 * NAiL googled for cm-102 usb audio linux... first hit is nslu2-linux :P Oct 06 18:06:47 :) Oct 06 18:08:34 NAiL: I have to go now... I would check the diff in the areas of the ide driver and the mach initialization code Oct 06 18:10:47 I'm cleaning out the cruft first Oct 06 18:11:03 There's a *lot* of it ;) Oct 06 18:11:26 you may want to diff adding -X kernel/Documentation/dontdiff Oct 06 18:11:47 dontdiff is a file with all the extensions that should be avoided Oct 06 18:11:50 while diffing Oct 06 18:11:52 aha Oct 06 18:11:57 well, too late :P Oct 06 18:12:28 there's heaps and heaps of rcs crap Oct 06 18:12:35 but that's easy to remove Oct 06 18:12:38 MMRRRGH Oct 06 18:12:48 ~lart lousy wifi conections Oct 06 18:12:48 * jbot beats lousy wifi conections to within 2.54cm of his life Oct 06 18:13:02 NAiL: it's worth starting again with the dontdiff Oct 06 18:13:44 see ya tomorrow.. Oct 06 18:13:54 hmm Oct 06 18:13:58 no dontdiff file Oct 06 18:14:29 in either kernels Oct 06 18:15:12 they must be old kernels then.. Oct 06 18:15:45 2.4.22 Oct 06 18:15:52 get a dontdiff here: http://www.xenotime.net/linux/doc/dontdiff Oct 06 18:17:21 thx Oct 06 18:25:43 gotta love midnight commander when managing a huge diff ;) Oct 06 18:39:46 ok, this is getting annoying **** BEGIN LOGGING AT Thu Oct 06 18:45:03 2005 Oct 06 18:48:38 could it be that the ipkg alsa-utils package for openslug is damaged? Oct 06 18:49:01 hmm? Oct 06 18:49:03 why? Oct 06 18:49:22 when I do a ipkg install... I got this error: Oct 06 18:49:23 An error ocurred, return value: 1. Oct 06 18:49:24 Collected errors: Oct 06 18:49:24 ERROR: Cannot satisfy the following dependencies for alsa-utils: Oct 06 18:49:24 ${ALSA_PACKAGES} Oct 06 18:49:41 2.7-beta? Oct 06 18:49:47 yes Oct 06 18:52:29 03repvik * r215 10/releases/OpenSlug-2.7-beta/openembedded/packages/alsa/alsa-utils_1.0.8.bb: Fix broken RDEPENDS Oct 06 18:52:52 cool, willt test it Oct 06 18:54:09 I'll have to rebuild the fieed first Oct 06 18:54:10 feed Oct 06 18:54:21 ok Oct 06 18:54:53 I tryed to build alsa-1.0.9, but have no luck Oct 06 18:55:18 03repvik * r216 10/releases/OpenSlug-2.7-beta/openembedded/packages/nfs-utils/nfs-utils_1.0.6.bb: Add update-rc.d Oct 06 18:56:12 03repvik * r217 10/releases/OpenSlug-2.7-beta/openembedded/packages/ipkg-utils/ipkg-utils_1.6cvs.bb: Fix python RDEPENDS Oct 06 19:34:54 03repvik * r218 10/releases/OpenSlug-2.7-beta/ (2 files in 2 dirs): Add madplay Oct 06 19:38:02 03repvik * r219 10/releases/OpenSlug-2.7-beta/openembedded/packages/meta/openslug-packages.bb: Add madplay to feed Oct 06 22:05:39 03jbowler * 10kernel/2.6.14/50-nslu2-include.patch: This patch is now in nslu2-general. Oct 06 22:20:08 03jbowler * 10kernel/2.6.14/ (20-timer.patch 50-nslu2-arch.patch): Oct 06 22:20:08 This is the set of changes currently being tested on the monotone head. Oct 06 22:20:08 Added: 20-timer.patch, this has the FREQ hack in it. Oct 06 22:20:08 Modified: 50-nslu2-arch.patch: this change removes the hardwired machine Oct 06 22:20:08 type with NSLU2_CONFIG (if the boot loader is broken - i.e. LinkSys Oct 06 22:20:08 RedBoot - this must be fixed by a prefix on the kernel image to set r1 Oct 06 22:20:10 to the required value.) Oct 06 23:01:17 03jbowler 07org.openembedded.dev * r8cb9fe03... 10/packages/linux/ (5 files in 2 dirs): nslu2-kernel: integrate 2.6.14-rc3 patchset 4 Oct 06 23:55:17 jbot, test Oct 06 23:55:19 Test Failed! Oct 07 01:01:26 03koen 07org.openembedded.dev * r063f5fcd... 10/packages/blueprobe/ (blueprobe-0.15/h6300.patch blueprobe_0.15.bb): packages/blueprobe/blueprobe_0.15.bb: add patch for h6300 support from #373 Oct 07 02:36:17 03koen 07org.openembedded.dev * r3c33d888... 10/packages/libgpewidget/libgpewidget_cvs.bb: packages/libgpewidget/libgpewidget_cvs.bb: add (disabled) cvs version of libgpewidget **** ENDING LOGGING AT Fri Oct 07 02:59:56 2005