**** BEGIN LOGGING AT Mon Aug 29 02:59:57 2011 Aug 29 03:01:53 nbd: what's the correct way to reflect that a kmod builds with 2.6.38 and earlier only? Aug 29 03:53:35 [florian]: updated the alix patch for 3.0. Aug 29 04:44:22 round and round and round it goes... Aug 29 06:29:42 wtf? is sysupgrade (on wgt) not respecting the -n flag? Aug 29 06:33:05 gah, no it isn't Aug 29 06:33:33 as of r28087 Aug 29 06:33:51 makes remote flashing a little more challenging :-( Aug 29 06:44:28 ls Aug 29 06:44:32 oops Aug 29 06:44:52 looks like its finding the old jffs and going with that Aug 29 06:45:01 on reboot Aug 29 06:45:28 even a simple mtd -r write foo.trx linux is resulting in the same behavior Aug 29 06:46:06 jow_laptop: ping Aug 29 06:46:46 violates least astonishment Aug 29 06:50:41 jow_laptop: i will be AFK .. sorry .. will ping later Aug 29 08:04:01 nbd * r28109 /trunk/package/mac80211/patches/ (3 files): ath9k: minor fixes for the antenna control patch Aug 29 08:04:08 nbd * r28110 /trunk/package/mac80211/patches/590-ath9k_fix_5ghz_adc_cal.patch: ath9k: fix calibration on 5 ghz Aug 29 08:13:06 russell--: sounds like a bug in image padding to me Aug 29 08:27:14 nbd * r28111 /trunk/package/hostapd/patches/ (570-eapol_key_type.patch 770-group_key_timeout.patch): hostapd: add a patch for ignoring the WPA/WPA2 key type to work around some broken clients (#9561) Aug 29 08:28:35 nbd * r28112 /branches/backfire/package/hostapd/patches/ (570-eapol_key_type.patch 770-group_key_timeout.patch): hostapd: add a patch for ignoring the WPA/WPA2 key type to work around some broken clients (#9561, backport of r28111) Aug 29 09:16:51 russell--: interesting Aug 29 09:18:59 wow. size of packages went up 20% with the new 4.6-linaro compiler Aug 29 09:35:58 bah Aug 29 09:36:00 sucks Aug 29 09:36:20 what target? Aug 29 09:41:56 nbd: x86 Aug 29 09:59:12 nbd * r28113 /trunk/package/mac80211/patches/ (3 files): ath9k: fix a tx aggregation padding issue Aug 29 09:59:17 nbd * r28114 /trunk/package/mac80211/patches/ (2 files): ath9k: fix processing of rx frames on channel change Aug 29 10:27:23 loswillios: do you have grub working then ? Aug 29 10:27:26 i doubt it Aug 29 10:28:25 WillieNL: I didn't try to boot the image Aug 29 10:30:59 loswillios: during build the old grub will wine about stage1 stage2 mismatch if not i wonder how you fixed that since i switched back to gcc 4.4.6 Aug 29 10:31:13 next to the 20% more buildtime Aug 29 10:31:42 also wonder why the gcc compiler gets updated faster then everything else inside the sdk :( Aug 29 10:32:20 how is it getting updated faster than everything else? Aug 29 10:32:31 we've been at 4.5-linaro for a quite a while now Aug 29 10:32:36 before that we had 4.3.3+cs very long Aug 29 10:32:55 nbd: people request to drop gcc 4.4 for newer gcc and get a "ok" just like that and then people complain about build errors . Aug 29 10:33:05 gcc 4.4 wasn't used very much at all in openwrt Aug 29 10:33:07 nbd: hmm ok so 4.4.6 will not be removed ? Aug 29 10:33:39 i've never used 4.4.6 myself Aug 29 10:33:42 or must i ask via a ticket to keep it in the system as stable compiler ? Aug 29 10:33:45 when it's not being used, it will be removed Aug 29 10:33:58 why keep it as a stable compiler, what doesn't work with 4.5-linaro? Aug 29 10:34:43 nbd: without the linaro patches everything is like how i know/remember and then i have a working gcc in my wrt target image in case cross compiling a package fails. Aug 29 10:35:33 the -linaro version flag is not managed by the target gcc package so that won't build and even then with the patches it is a pain.. Aug 29 10:35:40 without it is going flawless Aug 29 10:36:04 thats why i avoid -linaro and gcc 4.6.x and 4.5.x Aug 29 10:36:35 also all i do is for x86 Aug 29 10:37:13 since i feel like everything i use/need will get removed soon :( Aug 29 10:37:21 ah, so the only issue is with cross-compiling gcc? Aug 29 10:37:44 nbd: no main reason still is basicly rrdtool. Aug 29 10:38:13 huh? Aug 29 10:38:24 why did you just say something completely different then? Aug 29 10:38:28 also not done anything the last 2 weeks so when i start tonight it will be a new tour true changes Aug 29 10:38:53 nbd: for that i do need gcc 4.4.X and do want gcc in the target image so i can install stuff without the sdk Aug 29 10:38:59 why do you need gcc 4.4 for that? Aug 29 10:39:20 avoid issues that get triggered in new gcc and new gcc won't build inside the target images anyway Aug 29 10:39:27 specialy not with -linaro Aug 29 10:39:27 what issues? Aug 29 10:39:43 dependency version mismatches etc etc Aug 29 10:40:05 the way to go is not to desperately cling to old crap just for some specific packages Aug 29 10:40:16 fixing the new stuff is much better Aug 29 10:40:25 we talk about this in 5 minutes but reproducing and finishing this tour take a lot longer.. Aug 29 10:40:53 nbd: yes i know i asked for upgrading rrdtool but that will be tobig for the embedded targets. Aug 29 10:41:06 how about fixing the existing package? Aug 29 10:41:07 so i'm on my own there Aug 29 10:41:15 no bugs are fixed upstream Aug 29 10:41:27 yes, but openwrt allows you to patch stuff Aug 29 10:41:45 how much bigger is the new version anyway? Aug 29 10:41:51 and what's causing it to be bigger? Aug 29 10:42:40 nbd: gcc 1.3.8 is working for me as long using gcc 4.4.6 or 4.6 but had a lot autoconf/m4 issues around the perl inclusions during cross compiling. Aug 29 10:42:54 thats basicly the reason why i wanna build it native in the wrt image Aug 29 10:43:02 same about rrdtool 1.4.5 Aug 29 10:43:11 s/gcc 1.3.8/rrdtool Aug 29 10:43:29 i think building rrdtool natively is a waste of time Aug 29 10:43:51 sadly there is no replacement for rrdtool anyway so it is the hardway or no graphs Aug 29 10:44:29 and rrdtool before 1.3.8 sucks tomutch memory anyway compared to 1.3.8 and newer Aug 29 10:44:47 nbd * r28115 /trunk/package/mac80211/patches/ (2 files): ath9k: fix disabling ani on reset Aug 29 10:44:53 it was growing by day :( Aug 29 10:46:53 anyway busy with other stuff then openwrt atm.. (catalog/order cnc parts) Aug 29 10:47:12 are you writing to rrd files on the owrt device itself? Aug 29 10:48:44 thats the idea on a x86box with 512M ramdisk yes Aug 29 10:49:16 else i will endup with 10 machines for just graphing and monitoring next to the router :S Aug 29 10:50:03 nbd * r28116 /trunk/package/mac80211/patches/ (2 files): ath9k: fix a compile error Aug 29 10:50:22 karlp: the idea is using that box as rrd host Aug 29 10:50:35 +rrdweb Aug 29 11:00:49 10 machines? Aug 29 11:01:15 12:42:54] anyway busy with other stuff then openwrt atm.. (catalog/order cnc parts) Aug 29 12:05:04 nbd * r28117 /trunk/package/mac80211/patches/ (3 files): ath9k: add some cleanup patches Aug 29 12:05:07 nbd * r28118 /trunk/package/mac80211/patches/593-ath9k_ar9280_cold_reset.patch: ath9k: use cold reset instead of warm reset on ar9280 to make recovery from stuck states more reliable Aug 29 12:25:57 nbd: i don't quite grasp the details, but i understand that the squashfs image pads something magical that identifies the beginning of space to use for jffs2. can you point me at the moving parts? Aug 29 12:28:47 this looks suspiciously like it: /aux/src/openwrt/staging_dir/host/bin/trx -o /aux/src/openwrt/bin/brcm47xx/openwrt-brcm47xx-squashfs.trx -f /aux/src/openwrt/build_dir/linux-brcm47xx/loader.gz -f /aux/src/openwrt/build_dir/linux-brcm47xx/vmlinux.lzma -a 1024 -f /aux/src/openwrt/build_dir/linux-brcm47xx/root.squashfs -a 0x10000 -A /aux/src/openwrt/build_dir/linux-brcm47xx/fs_mark Aug 29 12:44:47 nbd: that looks wrong https://dev.openwrt.org/ticket/9999#comment:3 Aug 29 12:49:22 what looks wrong? Aug 29 12:49:53 you closed 9999 as duplicate of 9999 Aug 29 12:50:06 oh, copy&paste bug Aug 29 12:50:07 indeed Aug 29 12:50:18 added a comment Aug 29 12:50:33 thanks Aug 29 12:50:37 np Aug 29 12:57:24 kerneis * r28119 /packages/net/babel/ (Makefile files/babeld.config files/babeld.init): Aug 29 12:57:24 [packages] babel: upstream release 1.1.99.1 Aug 29 12:57:24 * INCOMPATIBLE CHANGE: default port changed to 6696 Aug 29 12:57:24 * New option: 'diversity' Aug 29 15:35:35 loswillios * r28120 /packages/net/openvpn/ (Config.in Makefile): [packages] openvpn: fix lzo dependency and rename config options Aug 29 16:49:01 nbd: ping? Aug 29 16:49:29 full: ok, here goes v1/v2 integration attempt Aug 29 16:49:31 wish me luck Aug 29 16:55:01 fullstop:hm, the options in the kernel menuconfig allow you to select support for both. wouldn't such a kernel "just work" on both? I still don't have a v2 here Aug 29 16:55:11 fullstop: pong Aug 29 16:59:06 Chocky: I think that means that both images will be built. Aug 29 16:59:39 Chocky: the image needs to know what board it is for. I assume that this happens down in the bootstrapping setup Aug 29 16:59:47 hm, ok Aug 29 16:59:50 * Chocky looks Aug 29 17:00:01 nbd: I was wondering if you had a chance to look at the ath9k gpio stuff yet? Aug 29 17:00:15 I think you might be wrong tho; the image handling is openwrt, not kernel build Aug 29 17:00:40 and I had all the machine support turned on in kernel, for ar71xx, just turned that all off Aug 29 17:02:17 fullstop: not yet Aug 29 17:05:57 nbd: thanks for the update Aug 29 17:06:42 Chocky: in target/linux/ar71xx/image, the makefile there specifies the machine name / arguments for each image built. Aug 29 17:06:58 I know; but are those different for v1/v2? Aug 29 17:07:29 Chocky: That's how the image "knows" what hardware it has. Aug 29 17:07:50 that doesn't really answer the question Aug 29 17:07:52 It's explicitly set for each image built. Aug 29 17:08:10 We may be speaking on different wavelengths here. Aug 29 17:08:33 is the same vmlinux used, with support turned on for all machines, as it is by default, to generate the ROM image? Aug 29 17:09:03 It needs to know which "MIPS_MACHINE" to use. Aug 29 17:09:10 There's no magic here. Aug 29 17:09:36 ok, like ARM, there's a value in the header passed by u-boot to the kernel Aug 29 17:09:38 +? Aug 29 17:10:43 I'm pretty sure that most information passed to the kernel by u-boot is ignored. Aug 29 17:10:54 yes, but you aren't answering my question Aug 29 17:11:02 What is your question then? Aug 29 17:11:20 is MIPS_MACHINE a value set in the generated image? Aug 29 17:11:32 and would therefore, the v1 and v2 images differ by only that byte? Aug 29 17:11:46 the kernel command line is different, yes. Aug 29 17:11:52 for nh2 it is wzrhpg300nh2_cmdline=board=WZR-HP-G300NH2 console=ttyS0,115200 Aug 29 17:12:05 and for G300NH it is wzrhpg300nh_cmdline=board=WZR-HP-G300NH console=ttyS0,115200 Aug 29 17:12:09 ok, it's part of the embedded kernel line Aug 29 17:12:22 and, obviously, the board part of that is what matches the appropriate setup code. Aug 29 17:12:46 so the code is identical otherwise, and it's therefore only a matter of detection if boards were to merge. Aug 29 17:12:53 Yes. Aug 29 17:12:55 I had thought the problem more complex Aug 29 17:13:06 no, there is less magic than you thought. ;-) Aug 29 17:13:17 actually, slightly more :p Aug 29 17:13:27 haha.. we have different definitions of magic! Aug 29 17:13:28 I had thought the machine support was separate for each Aug 29 17:13:33 but that's ok Aug 29 17:14:02 it's been a little while since I hacked sub-arch support in the kernel Aug 29 17:14:15 separate = ifdef for each, etc Aug 29 17:14:18 anyway Aug 29 17:14:52 There is SoC detection code, but that is common for all ar71xx. Aug 29 17:15:01 right Aug 29 17:15:07 time for a break Aug 29 17:15:20 kit-kat Aug 29 17:50:07 loswillios * r28121 /packages/sound/pulseaudio/Makefile: [packages] pulseaudio: update to 1.0-rc3 Aug 29 18:34:25 * Chocky pokes at kernel Aug 29 18:45:23 nbd * r28122 /trunk/package/mac80211/patches/ (5 files): ath9k: fix a few crash issues on hardware reset Aug 29 18:53:12 004d0000 de ad c0 de 00 00 00 00 00 00 00 00 00 00 00 00 |................| Aug 29 18:53:34 the fs_mark is there Aug 29 19:06:32 fullstop: boo Aug 29 19:06:54 ok, my nominally v2 kernel boots on v1; all did was have it call different setup code based upon SoC Aug 29 19:07:00 let me see if I can tidy this all up now Aug 29 19:12:20 * russell-- works on narrowing down when it started Aug 29 19:18:58 Chocky: that will work for now. No guarantees for the future, though. Aug 29 19:19:09 fine with me Aug 29 19:41:44 hmm. when i go from r28010 to r28087 initially, it notices that "jffs2 not ready yet; using ramdisk", but when I flash a different image of r28087 it likes the jffs2 just fine and uses it: "JFFS2 notice: (287) jffs2_build_xattr_subsystem: complete building xattr subsystem, 1 of xdatum (0 unchecked, 0 orphan) and 1 of xref (0 dead, 0 orphan) found." Aug 29 19:42:37 so, r28087+files -> r28087+differentfiles, doesn't rebuild jffs2 Aug 29 19:43:02 that can happen if you mange to disable mini_fo Aug 29 19:43:08 i didn't Aug 29 19:43:19 no, a bit hard. but it's one explaination Aug 29 19:43:24 -i Aug 29 19:47:48 mhei * r28123 /packages/lang/php5/ (3 files in 2 dirs): [packages] php5: PECL: include support for http Aug 29 19:48:35 trying r28087+differentfiles -> r28087+files Aug 29 19:48:47 different-files is a little bigger Aug 29 19:49:44 but 64k Aug 29 19:49:46 hmm. Aug 29 19:50:50 i vaguely recall that flash block sizes on wgt are different than other brcm47xx devices, because of the 8M flash parts or something Aug 29 19:51:26 going back to r28087+files works Aug 29 19:52:25 that makes me think that things are looking in the wrong spot for de:ad:c0:de Aug 29 19:54:30 * Chocky slaps pidgin for smilying that Aug 29 19:54:48 works: 004c0000 de ad c0 de 00 00 00 00 00 00 00 00 00 00 00 00 Aug 29 19:55:13 works-not: 004d0000 de ad c0 de 00 00 00 00 00 00 00 00 00 00 00 00 Aug 29 19:57:36 fullstop: oh crap. the openwrt network config is gonna be different, right? Aug 29 19:57:55 I think we can fix that up at boot time Aug 29 19:57:57 first boot Aug 29 20:10:19 Does anyone here know if "enable-3pcc" (http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#enable-3pcc) option is global or can be set on each trunk? Aug 29 20:12:13 Oops, just ignore the above. Posted on the wrong channel. :( Aug 29 20:20:14 Chocky: yes, they are different. Aug 29 20:20:45 I think there's some logic in there already for 300 vs 301. or perhaps that's entirely handled in the kernel Aug 29 20:23:01 Chocky: it pulls it from /etc/defconfig based on the board name passed to the kernel Aug 29 20:23:09 aye Aug 29 20:23:17 so, mm. Aug 29 20:23:19 but, you've wiped that slate.. ;-) Aug 29 20:23:50 actually, I think the board name may be modified by one of the scripts later, in our case, based upon SoC Aug 29 20:23:56 which is used for that Aug 29 20:24:35 ok, so I essentially combined v1 and v2 machine files. Pretty easy. Aug 29 20:36:49 yes, so I don't see why we can't put extra logic in ar71xx.sh Aug 29 20:50:57 russell--: try this: http://nbd.name/brcm-align.patch Aug 29 20:54:19 nbd: from where did "nbd" come from? Aug 29 20:54:26 out of nowhere Aug 29 20:54:37 smashing keys on a keyboard? Aug 29 20:59:27 my new nick will be: gjkls Aug 29 21:11:46 ah yeah, /me finally just did the math, on the bad image, rootfs_data is at 4E0000 Aug 29 21:15:38 nbd: it looks to me that the alignment should be 0x20000 rather than 0x10000, at least that's what the mtd partitioning stuff seems to be doing Aug 29 21:15:59 alignment of 0x10000 should be enough Aug 29 21:16:06 as long as it's repeated 64k later Aug 29 21:17:23 the wgt has a 128k erase block, i seem to remember Aug 29 21:17:42 yes, i know Aug 29 21:17:45 okay Aug 29 21:17:58 * russell-- shuts up and tries the patch Aug 29 21:18:24 fullstop: see the list Aug 29 21:25:58 oh duh, yeah, i see you just added another fs_mark Aug 29 21:26:42 so, yeah, that works Aug 29 21:26:56 i have a better fix which i'm going to commit now Aug 29 21:27:12 since i figured out the root cause of the issue Aug 29 21:27:22 padjffs2 does not add multiple tags on empty files Aug 29 21:27:34 so if i create the first one in the file, padjffs2 will create the rest Aug 29 21:27:51 nbd * r28124 /trunk/target/linux/ar71xx/files/arch/mips/ar71xx/devices.c: ar71xx: fix ethernet PLL setting on ar7242 Aug 29 21:27:55 nbd * r28125 /trunk/target/linux/brcm47xx/image/Makefile: brcm47xx: fix padding of trx images, padjffs2 does not work properly on empty files by default Aug 29 21:28:00 there, that should take care of it Aug 29 21:35:02 * russell-- will test to confirm Aug 29 21:50:57 nbd, i can say that r28015 works here, thanks! Aug 29 21:51:08 sorry, r28125 Aug 29 21:51:15 fingers-- Aug 29 23:38:16 nbd * r28126 /trunk/package/base-files/files/etc/sysctl.conf: base-files: remove an old network tunable tweak which is messing up network stack performance on modern systems Aug 29 23:39:33 nbd * r28127 /branches/backfire/package/base-files/files/etc/sysctl.conf: base-files: remove an old network tunable tweak which is messing up network stack performance on modern systems (backport of r28126) Aug 30 00:27:01 mazilo * r28128 /packages/net/freeswitch/ (28 files in 19 dirs): updated to latest git version, added HU lang support, and minor fixes to the Makefile **** ENDING LOGGING AT Tue Aug 30 02:15:47 2011 **** BEGIN LOGGING AT Tue Aug 30 02:16:14 2011 **** ENDING LOGGING AT Tue Aug 30 02:18:15 2011 **** BEGIN LOGGING AT Tue Aug 30 02:20:16 2011 **** ENDING LOGGING AT Tue Aug 30 02:59:57 2011