**** BEGIN LOGGING AT Wed May 23 02:59:57 2007 May 23 09:42:44 kaloz * r7306 /trunk/tools/mkimage/src/ (crc32.c image.h mkimage.c): upgrade mkimage the current git version May 23 12:28:46 kaloz * r7307 /trunk/tools/mkimage/src/crc32.c: fix mkimage compile May 23 13:10:42 nbd: re optware build system and openwrt May 23 13:11:42 the main thing about optware is that it specifically installs everything in /opt so that it's guaranteed not to collide with vendor firmware (optware is designed to work with vendor firmware, and only recently did we start building it for custom firmware like slugos and openwrt too) May 23 13:12:21 i think /opt should not be hardcoded May 23 13:12:33 packages for optware should have a configurable prefix May 23 13:12:37 that would make them more flexible May 23 13:12:43 i mean configurable at build time May 23 13:12:48 agreed. if that was configurable there would be no reason not to have a common build system May 23 13:12:56 exactly May 23 13:13:13 is it configurable in buildroot-ng? May 23 13:13:17 no May 23 13:13:21 <[mbm]> that's one of thos fsstnd things that I have trouble with May 23 13:14:05 [mbm]: oh, optware doesn't even follow the fsstnd - we completely bastardized it up May 23 13:14:06 <[mbm]> technically third party packages belong in somewhere like /opt or /usr/local but then you get into a mess with the package management system May 23 13:14:23 <[mbm]> rwhitby: shame May 23 13:14:55 <[mbm]> but the nice thing about standards is that there's so many of them - did you follow any? May 23 13:14:57 yeah, some decisions you look back on and know now how you could have done it better ... May 23 13:15:19 we just put /opt in front of the normal / layout (i.e. /opt/bin, /opt/lib, etc) May 23 13:15:35 <[mbm]> hmm May 23 13:15:42 we used /opt like a /prefix May 23 13:16:12 (or like an alternate /usr, except that we have /opt/var and /opt/etc as well :-) May 23 13:16:15 actually the change from /opt to something else can be done at a later point in time May 23 13:16:29 first convert the build system, then use a sed hack to s/\/opt/$(PREFIX)/ May 23 13:16:33 <[mbm]> there's that tricky subject of "I just added external storage, how do I install more packages" May 23 13:16:35 in the build system makefile May 23 13:16:37 s May 23 13:17:09 <[mbm]> I have to say that I absolutely hate the ipkg link script May 23 13:17:20 yeah May 23 13:18:31 <[mbm]> the ipkg system has the ability to target different root devices but it doesn't quite work in practice May 23 13:18:58 <[mbm]> hmm .. May 23 13:19:01 and different prefix would be much more useful to change than different root filesystem May 23 13:19:03 <[mbm]> just had a thought May 23 13:19:25 <[mbm]> suppose we take the normal squashfs+jffs2 May 23 13:19:35 <[mbm]> and then add usb as an overlay ontop of that May 23 13:19:35 is there a way to have squashfs+jffs2+usb? May 23 13:19:49 yeah, what he said :-) May 23 13:19:58 [mbm]: the problem is due to an ugly hack, mini_fo will probably have trouble with stacking May 23 13:19:58 <[mbm]> somewhat confusingly you'd have /rom, /jffs, /usb May 23 13:20:23 <[mbm]> what I'm wondering though is how smart ipkg is May 23 13:20:32 <[mbm]> if I can set my root target May 23 13:20:34 don't depend on ipkg being smart May 23 13:20:36 because it isn't May 23 13:20:37 :) May 23 13:20:45 someone with enough time should rewrite ipkg as a real busybox applet :p May 23 13:20:46 <[mbm]> then I can ssay to install to /jffs or /usb May 23 13:21:00 [mbm]: it cna do it with -t or what May 23 13:21:07 [mbm]: if you have multiple targets May 23 13:21:22 I'd be happy enough with "just unmount usb first if you want to install to jffs2" May 23 13:21:37 <[mbm]> rwhitby: don't think you need to May 23 13:22:08 but then I guess you need to worry about usb files overlaying the stuff you just installed in jffs2 May 23 13:22:08 <[mbm]> you'd probably stack squashfs -> jffs -> usb May 23 13:22:42 <[mbm]> yeah, although that could also be a feature if used properly May 23 13:23:06 <[mbm]> ie when I plug in my usb device, it has a different config file which overrides whatever config is on jffs2 May 23 13:23:11 the way we use that in nslu2-linux is to have enough stuff in the jffs2 to recover from a failed disk May 23 13:23:27 <[mbm]> we do basically the same with squashfs May 23 13:24:28 <[mbm]> oh, while I'm thinking of it, we need some mini_fo utils May 23 13:24:40 <[mbm]> namely a 'revert' util May 23 13:25:21 <[mbm]> if I edit /etc/banner then the copu from squashfs is copied to jffs2, and overrides the copy on squashfs May 23 13:26:02 <[mbm]> and while you can revert your changes, there's no easy way to remove the jffs2 copy May 23 13:26:47 <[mbm]> you have to manually delete the file or metadata off jffs2 and reboot May 23 13:26:57 btw. i think instead of squashfs+jffs2+usb we should allow people to switch from squashfs+jffs2 to squashfs+usb May 23 13:27:32 either would work for me May 23 13:27:34 with a script that prepares the usb root with a copy of jffs2 May 23 13:27:40 <[mbm]> I like the idea of reducing the number of binary images May 23 13:27:53 * rwhitby heads off to bed - will read the backlog in the morning. May 23 13:28:12 <[mbm]> no need to create a special usb image if you can just install a package onto jffs2 that does it May 23 13:28:40 i didn't mean a special image May 23 13:28:49 just a script that runs at boot time and switches to the new root May 23 13:29:34 <[mbm]> what we need are packages that take over /sbin/init on jffs2 and chain boot into whatever root filesystem May 23 13:30:06 <[mbm]> although I still haven't figured out all the implications of what happens when someone says <*> to those packages in menuconfig May 23 13:31:46 it could work with special preinst/postinst scripts May 23 13:32:31 <[mbm]> ? May 23 13:32:37 brb May 23 13:41:19 re May 23 13:42:11 <[mbm]> . May 23 13:42:20 the package that overrides /sbin/init could do so in postinst May 23 13:42:39 and busybox could move away /sbin/init in preinst and move it back in postinst May 23 13:42:50 so it wouldn't matter in which order busybox and the other package get installed May 23 13:43:41 <[mbm]> interesting but I was thinking of the recursion problem of /sbin/init calling itself May 23 13:44:04 it could call /bin/busybox init May 23 13:44:48 <[mbm]> well, you still want to allow chaining, you just want to control it May 23 13:51:41 who'd actually use that chaining? May 23 13:52:28 <[mbm]> I just don't like the assumption that it will always be busybox init that's called May 23 13:52:58 <[mbm]> it's a nice fallback when you detect a loop though May 23 14:16:46 <[mbm]> ... May 23 14:30:09 nbd * r7308 /tags/kamikaze_pre1/package/madwifi/patches/113-security_fixes_r2366.patch: add madwifi security fixes from r2366 May 23 14:30:44 *sigh* rebuilding pre1 again :) May 23 14:31:01 nbd: this new madwifi thing works stable, but throughput isn't as good as it was before May 23 14:31:26 I can't watch a movie over WLAN May 23 14:31:36 how's the load average on the device? May 23 14:31:46 and how much throughput are you getting? May 23 14:33:03 I'm currently back to r7075 but I will test tommorrow with a recent build again (last test was with r7211, you fixed some madwifi things after that) May 23 14:52:41 loswillios: use sample instead of micrel May 23 14:53:31 loswillios: i'm pretty sure you would get better throughput May 23 14:53:48 (and it's just rmmod/insmod) May 23 14:54:17 nbd * r7309 /trunk/package/madwifi/ (4 files in 2 dirs): update madwifi to latest trunk (refcount and hal-0.9.30.13 got merged) and include a security fix that was merged in 0.9.3.1 May 23 14:54:36 nbd: PCCard (PCMCIA/CardBus) support (PCCARD) [N/m/y/?] (NEW) May 23 14:55:20 nbd: Is there anything more I can do to debug the PCMCIA-issue? May 23 14:57:01 nbd * r7310 /trunk/target/linux/generic-2.6/config-template: make CONFIG_PCCARD disabled by default May 23 14:57:01 Kaloz: I'll try, thanks May 23 14:57:12 loswillios: there you go. May 23 14:57:23 heh, ok :) May 23 15:30:32 loswillios: when you have time, please try the new madwifi version that i checked in May 23 17:15:18 nbd * r7311 /trunk/package/busybox/patches/920-awx.patch: fix awx segfaults May 23 18:15:10 nbd * r7312 / (6 files in 6 dirs): fix x86 boot filesystem allocation (#1736) May 23 18:21:33 nbd * r7313 /tags/kamikaze_pre1/package/busybox/patches/920-awx.patch: sync awx with trunk May 23 18:33:59 nbd * r7314 / (2 files in 2 dirs): awx: fix a typo May 23 19:48:36 kaloz * r7315 /trunk/ (34 files in 7 dirs): break trunk temporary - upgrade to 2.6.21.1 and iptables 1.3.7 May 23 19:49:01 Kaloz: s/_NOW_/r7314/ May 23 19:49:32 be precise May 23 19:49:35 :) May 23 20:19:00 nbd: ui. its only 250kb/s with the new code while 1,9mb/s with the old one May 23 20:19:58 nbd: no cpu-usage though. but a lot errors: http://pastebin.ca/506016 May 23 20:22:39 Kaloz: which module do you mean with 'sample', I don't seem to have one by that name May 23 20:22:46 loswillios: please try recompiling with RATE_CONTROL:=sample in madwifi May 23 20:22:59 loswillios: in the build system makefile May 23 20:23:02 ok May 23 20:23:11 I see May 23 20:23:48 nbd: oh, we only compile one algorythm? May 23 20:24:54 i think it compiles all of them, but only packages one May 23 20:26:50 maybe we should package them separately.. or dunno, honestly i think 802.1x and wep is rarely used by people for exmaple May 23 20:33:32 insmod: cannot insert '/lib/modules/2.6.19.2/ath_rate_sample.ko': Invalid parameters (8): Invalid argument May 23 20:34:17 (I copied ath_rate_sample.ko from build_mips/linux-2.6-atheros/madwifi-ng-r2362-20070522/ath_rate/sample/) May 23 20:36:00 * loswillios recompiles madwifi May 23 20:40:06 Kaloz: 802.1x is quite freqnetly used... May 23 20:40:42 it is for instance an ARINC standard for gate link wireless APs May 23 20:42:19 murb: i mean less then 5% of the users are using it May 23 20:42:24 murb: so it cna be a separate package May 23 20:44:01 is it planned for the madwifi upstream to adopt the new wirless infrustrcutre in the upstream kernel? May 23 20:44:42 murb: dadwifi using dscape already May 23 20:46:38 Bartman007: is that branch being developed? May 23 20:47:22 yep May 23 20:47:51 dadwifi worked ok with my wgt634u (2.6.19.2) May 23 20:48:26 btw it is now called mac80211 rather than dscape May 23 20:49:00 so they've dumped their 802.11 stack? May 23 20:49:54 trunk is still madwifi so no May 23 20:51:18 we are upgrading to .21, then asap to 22-rc May 23 20:51:26 that has mac80211 in tree already May 23 20:51:32 we don't need to upgrade to 22-cc May 23 20:51:33 rc May 23 20:51:41 backporting mac80211 will be easy, imho May 23 20:51:41 but i want and need to May 23 20:52:13 and i want to follow mainline more closely.. way easier to fix a few stuff 5 times then fix all them at once May 23 20:52:29 and 22-rc have a lot of arm and avr32 fix May 23 20:52:35 nbd * r7316 / (2 files in 2 dirs): awx: fix translation May 23 20:53:11 for the arm and avr32 fixes, ok May 23 20:53:20 would have been unnecessary just for mac80211 May 23 20:53:21 :) May 23 20:53:48 Kaloz: which platforms were forward ported to 2.6.21 properly? May 23 20:55:51 ixp4xx, currently working on iop, atheros and aruba May 23 20:56:10 on atheros there is an error that's not there on ixp4xx, so has to be arch specific May 23 20:57:01 * nbd starts working on x86-2.6 May 23 20:57:11 ip_set_malloc_atomic want PAGE_KERNEL May 23 20:57:17 wants* May 23 20:57:29 what's that? May 23 20:58:01 Applying /home/nbd/wrap/target/linux/generic-2.6/patches/002-lzma_decompress.patch using plaintext: May 23 20:58:09 [...] FAILED May 23 20:58:19 forgot to commit or remove something? May 23 20:58:41 oh, wait May 23 20:58:42 no May 23 20:58:52 never mind May 23 20:58:57 it's the build system forgetting to rebuild something May 23 20:59:02 yeah May 23 21:11:47 nbd * r7317 /trunk/ (Makefile include/scan.mk): fix target rescan (version kernel changes) May 23 21:40:19 nbd * r7318 /trunk/target/linux/generic-2.6/patches/013-ipset_compile_fix.patch: fix compile error in the netfilter ipset module May 23 22:01:38 arch/i386/boot/compressed/lzma_misc.o: In function `decompress_kernel': May 23 22:01:38 lzma_misc.c:(.text+0x12a5): undefined reference to `end' May 23 22:02:03 :o May 23 22:02:07 endless compression? May 23 22:04:05 has anyone experience with kgdb for debugging the kernel? May 23 22:09:44 nbd * r7319 /trunk/target/linux/generic-2.6/patches/600-x86_lzma.patch: fix x86-2.6 compile May 23 23:10:09 kaloz * r7320 /trunk/target/linux/ (11 files in 6 dirs): update netfilter patches, fix atheros May 23 23:12:18 * nbd is fighting with the x86 lzma loader May 23 23:24:31 kaloz * r7321 /trunk/target/linux/iop32x-2.6/ (Makefile config/default): fixup iop32x, needs testing, but should work now May 23 23:29:12 nbd * r7322 /trunk/target/linux/generic-2.6/patches/600-x86_lzma.patch: fix x86 lzma loader May 23 23:49:12 nbd * r7323 /trunk/target/linux/ (generic-2.6/config-template x86-2.6/config/default): update x86 config and move some config defaults to the config template May 24 00:04:20 kaloz * r7324 /trunk/target/linux/ (4 files in 4 dirs): aruba fixes, work in progress May 24 00:32:06 <[mbm]> Kaloz: clean up the patch, you have a Makefile.orig in there May 24 00:52:36 nbd * r7325 /trunk/target/linux/generic-2.6/patches/600-x86_lzma.patch: nuke vga console support from the x86 lzma loader, it seems to cause problems at least in qemu **** ENDING LOGGING AT Thu May 24 02:59:57 2007