**** BEGIN LOGGING AT Sat Sep 18 02:59:57 2010 Sep 18 06:25:57 build #82 of ps3 is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/ps3/builds/82 Sep 18 08:02:14 build #100 of gemini is complete: Success [build successful] Build details are at http://tksite.gotdns.org:8010/builders/gemini/builds/100 Sep 18 11:40:04 cshore: there? Sep 18 12:21:17 build #116 of brcm63xx is complete: Failure [failed compile_4] Build details are at http://tksite.gotdns.org:8010/builders/brcm63xx/builds/116 Sep 18 12:25:47 jow_laptop, do you know the version of 2.4 kernel used in x86 port? Sep 18 12:25:56 in backfire i guess Sep 18 12:30:21 well thats just non existent for backfire x86 Sep 18 12:30:48 but i guess most recent release was 2.4.37 Sep 18 12:35:27 we don't use a 2.4 kernel for x86 Sep 18 12:37:57 indeed Sep 18 13:39:05 hi. I have manually cloned the linux-mips tree and set the path to it in menuconfig, but the image still builds with the default kernel. Do I have to add anything else? Sep 18 13:55:06 no Sep 18 13:59:12 but it still uses the default kernel... Sep 18 13:59:44 quilt enabled? Sep 18 13:59:58 Try make target/linux-clean Sep 18 14:02:02 hmm, for some reason, quilt wasn't installed on my system Sep 18 14:02:18 although I used it a while ago... Sep 18 14:02:24 anyway, how do you enable quilt? Sep 18 14:04:47 You need to disable it Sep 18 14:05:03 If it wasn't installed, it's obviously disabled Sep 18 14:05:19 I would try a make clean anyway Sep 18 14:05:55 doing a clean now Sep 18 14:24:30 I think I got it: Under "Use external kernel tree" I should have used a path, not a random string Sep 18 14:24:40 I can see it in the ln -s command Sep 18 14:24:57 but then, what is the reference repository for? Sep 18 14:26:31 Ehm, yeah. Random string is probably bad... Sep 18 14:26:42 There are two facilities Sep 18 14:27:06 You can point to an external tree with "use external kernel tree", _OR_ you can use the external git tree facitity. Sep 18 14:27:36 And the reference repos is just part of the external git tree thingie. Sep 18 14:28:22 Read the helptext anyway. It's pretty good. Sep 18 14:28:34 I did, but I didn't get this info. Sep 18 14:29:07 Maybe I didn't read it carefully... Sep 18 14:29:16 One more question though: That path won't be patched, right? Sep 18 14:29:20 Also note that the generic and platform patches are not applied, if you set an external tree. So you need to manually make sure the external tree is compatible with openwrt. Sep 18 14:29:27 :) Sep 18 14:29:32 aah, ok Sep 18 14:29:42 that's even better :P Sep 18 14:33:32 mb__: seems to be ok now, thanks for the help Sep 18 16:44:45 on the package makefile, can I refer one package install section on another? Sep 18 16:45:00 but not being the same, being able to add more things Sep 18 16:45:52 $(call Package/x-var1/install)? Sep 18 16:46:35 Why would you want to do that? Sep 18 16:47:27 because de variant2 of the package just needs another file... Sep 18 16:50:17 ok, yeah you can do that Sep 18 16:51:14 but "$(call Package/x-var1/install)" doesn't work... Sep 18 16:59:38 You would usually define a generic part and call that from all install sections. Sep 18 17:05:49 umhh, any example package? Sep 18 17:09:49 define mypack/generic-install .... endef Sep 18 17:09:56 $(call mypack/generic-install) Sep 18 17:31:24 mb * r23081 /trunk/target/linux/omap24xx/ (config-2.6.36 patches-2.6.36/200-omap-platform.patch): omap24xx: Add support for GPIO switches on the n810 Sep 18 17:43:35 anyone around who has edit rights to the Wiki? Sep 18 17:45:12 here Sep 18 17:46:07 doc/howto/build says to use "make V=99 |tee build.log" ... that won't capture stderr. you're better off with: Sep 18 17:46:40 script -c "make V=99" build.log ; dos2unix build.log Sep 18 17:47:48 ok Sep 18 17:48:30 also... doc/howto/installing doesn't mention the case of a device that just takes a CF. I've done a build, I've got some images... how do I get them onto the CF? Sep 18 17:48:42 dd Sep 18 17:48:53 the big images Sep 18 17:49:02 not the .fs or kernel/uimage Sep 18 17:49:29 those contain bootloader, partition table, kernel, rootfs, boot partition Sep 18 17:50:28 sorry, which ones? the root.squashfs image? Sep 18 17:53:35 I did this all about 4 months ago when I brought up the net5501 package, and I can't remember the steps. sigh. Sep 18 17:54:14 no Sep 18 17:54:36 blah-combined-blah.img[.gz] Sep 18 17:56:40 got it. Sep 18 17:57:16 mb__: why whould that work and not the "$(call Package/x-var1/install)", aren't they both regular "define"s? Sep 18 17:57:37 philipp64|laptop: don't forget to gunzip ;) Sep 18 17:57:42 if applicable Sep 18 17:57:57 yup. Sep 18 17:58:46 fetching it now. if the box comes up, I'll have a patch to be code reviewed shortly for adding the Traverse Technologies "Geos". Sep 18 17:58:53 mb * r23082 /trunk/target/linux/omap24xx/patches-2.6.36/ (300-nokia-board.patch 900-n810-battery-management.patch): omap24xx: Add n8x0 specific GPIO switch code Sep 18 17:59:13 nunojpg: Well, it should both work. Sep 18 17:59:15 nice little box. Geode LX800, 2 ADSL2+ ports, miniPCI, 2 GE ports, USB, serial... Sep 18 18:00:17 <[florian]> philipp64|laptop: adsl2+ using traverse solos cards ? Sep 18 18:11:51 [florian]: SBC. Sep 18 18:12:29 xMff: http://pastebin.ca/1943663 -- got a couple of minor issues booting up. Sep 18 18:14:33 not sure if microcode support is requisite... or why I'm seeing the ioremap backtrace... or why the solos_pci driver is getting unmatched symbols. Sep 18 18:14:40 mb * r23083 /trunk/target/linux/omap24xx/patches-2.6.36/ (500-cbus.patch 510-retu-tahvo-user-debugging.patch): omap24xx cbus: fix ioctls Sep 18 18:17:03 xMff: so solos_pci depends on "atm"... which contains the missing symbols... why didn't a "on demand" load happen? Sep 18 18:19:38 hmmm.... no m-i-t.... so using insmod instead of modprobe... Sep 18 18:19:54 doing an "insmod atm" fails silently with exit code 2. Sep 18 18:20:05 what am I missing? Sep 18 18:28:05 philipp64|laptop: 2GE? i am only finding one with 2 FastEthernet-Ports... Sep 18 18:29:00 8139cp Sep 18 18:29:55 ok, so where does "lspci" and "modprobe" live? what packages do I need? Sep 18 18:30:35 lspci = pciutils Sep 18 18:34:01 instead of modprobe try insmod... Sep 18 18:40:04 I did, that's what I'm saying. I did an "insmod -v atm" followed by "echo $?" and got 2. Sep 18 18:41:23 ah okay, i see.... Sep 18 18:42:25 logread does not show anything? Sep 18 18:43:46 nothing useful. Sep 18 18:50:21 ok, dumb question... when I run "make menuconfig", I can select certain packages as "builtin" or "module"... for the kernel, this makes sense. but what does it mean for "tcpdump"? Sep 18 18:52:55 buildin = installed in the image, module = only as package, not installed Sep 18 19:12:14 AlexanderS: ok, so CONFIG_PACKAGE_kmod-usb-storage=m means... build it as a module? or build it but don't include it in the image? Sep 18 19:12:54 build the package, but don't include it in the image Sep 18 19:13:16 the kmod-* packages are allways modules... Sep 18 19:19:58 mb * r23084 /trunk/scripts/dl_cleanup.py: dl_cleanup: Add support for GIT-SHASUM Sep 18 19:23:55 anyone have a minute to throw an eyeball on ftp://ftp.redfish-solutions.com/pub/openwrt-geos.patch ? Sep 18 19:24:55 In package/kernel/modules/netdevices.mk, for the package KernelPackage/solos-pci, if I put: Sep 18 19:25:23 DEPENDS:=@TARGET_X86@ +kmod-atm Sep 18 19:25:50 does that then mean that selecting solos-pci also requires kmod-atm to be built? Sep 18 19:29:27 well, I'll build and find out. Sep 18 19:57:05 philipp64|laptop: yes Sep 18 20:01:02 hmmm... util-linux-ng seems to generate a broken fdisk/Makefile. sigh. Sep 18 20:01:47 [florian]: got a minute? Sep 18 20:13:13 guess not. Sep 18 20:14:50 and Kaloz isn't around. Sep 18 20:16:36 well, I turn off util-linux-ng for now and fix that later. Sep 18 20:52:17 mb__: in fact both don't work, there is some issue with the substitution of $(1) Sep 18 20:53:30 when using a defined section elsewhere, if gets replaced by a empty string Sep 18 20:54:37 I'm not sure what you're doing there. Sep 18 20:54:49 What does "elsewhere" mean? Another makefile? Sep 18 21:06:52 all: what's the syntax for cleaning package x? make what? Sep 18 21:07:26 make package/x/clean Sep 18 21:08:11 ok, thanks... that always wigs me out. :-) Sep 18 21:15:33 I'm seeing some really borked dependencies... Sep 18 21:16:00 philipp64|laptop, regarding fdisk? Sep 18 21:16:21 I selected "fdisk", for instance, but it's also trying to build cfdisk... even though the fdisk/Makefile.am has cfdisk_OBJECTS commented out. Sep 18 21:16:22 anybody ever tried to cross compile subversion's mod_dav_svn? Sep 18 21:16:51 yes... and also setterm... setterm isn't selected, but it's being built anyway... and it requires ncurses, but ncurses isn't being built first...etc. Sep 18 21:17:05 RealOpty: what am I missing? Sep 18 21:17:09 philipp64|laptop, yeah it depends on cfdisk i think Sep 18 21:17:10 one sec Sep 18 21:17:50 do I need to back rev certain directories to unborked versions? Sep 18 21:18:11 philipp64|laptop, using trunk? Sep 18 21:19:17 yup... since I'll be committing patches... Sep 18 21:19:33 what do I need to steer clear of? Sep 18 21:20:39 I deleted "cfdisk" from the "make" line in package/util-linux-ng/Makefile and that fixed one problem... since I believe it gets built as a target of "all" in the fdisk directory anyway. Sep 18 21:21:14 i think it depends on cfdisk Sep 18 21:21:21 do you have the package libncurses selected? Sep 18 21:24:47 yeah, but I shouldn't need to! right at the top of package/util-linux-ng/Makefile is: PKG_BUILD_DEPENDS=...libncurses Sep 18 21:25:07 http://paste.ubuntu.com/490197/ Sep 18 21:25:16 did you get errors like that? Sep 18 21:25:35 philipp64|laptop, yeah that was added few days ago https://dev.openwrt.org/changeset/22982/trunk/package/util-linux-ng/Makefile Sep 18 21:26:25 that would be it. Sep 18 21:26:30 mb__: elsewhere is a generic define in the file, to be called be several install sections Sep 18 21:26:31 define mypack/generic-install .... endef Sep 18 21:26:38 let me make a pastebin... Sep 18 21:27:03 RealOpty: svn -r22981 package/util-linux-ng ? Sep 18 21:27:36 nope Sep 18 21:27:41 then.... Sep 18 21:28:00 AlexanderS: and to build package/x? Sep 18 21:28:45 i think that fdisk depends on cfdisk Sep 18 21:29:13 philipp64|laptop, try reselecting fdisk Sep 18 21:29:26 unmark it from makemenu then remark it Sep 18 21:30:14 RealOpty: if I backrev util-linux-ng, then ncurses gets built. Sep 18 21:30:44 philipp64|laptop, show me the output of make V=99 Sep 18 21:31:39 mb__: http://pastebin.com/ZZFMe315 Sep 18 21:34:49 RealOpty: too late... I revved up again. Sep 18 21:36:03 ok, rebuilding. Sep 18 21:36:48 nunojpg: You are using $(1) in generic, yet you are not passing that parameter. Sep 18 21:38:00 ping jow_laptop Sep 18 21:38:25 RealOpty: http://pastebin.ca/1943769 Sep 18 21:39:04 xMff: did you have a minute to eyeball my geos platform patch? nathan (from Traverse) said it looked ok. Sep 18 21:39:32 if I can ever get a successful build of packages, I'll put the image through burn-in.. Sep 18 21:39:47 philipp64|laptop, try selecting that cfdisk to be installed. Sep 18 21:40:54 philipp64|laptop, xMff made a patch that i cant find now, but i think that it selected the cfdisk that fixed this Sep 18 21:41:46 philipp64|laptop: make package/x/compile Sep 18 21:41:58 yup... figured it out. thanks. Sep 18 21:42:14 wondering if building with V=99 should also force -j1 implicitly. Sep 18 21:43:16 RealOpty: ffs even selecting cfdisk=y (by hand in .config) doesn't fix it. Sep 18 21:44:12 RealOpty: but the root cause is the source file itself. Sep 18 21:44:42 mb__: so it's that, thanks Sep 18 21:47:08 looking at util-linux-ng-*/fdisk/Makefile I see: Sep 18 21:47:14 #am_cfdisk_OBJECTS = cfdisk.$(OBJEXT) \ Sep 18 21:47:16 # $(am__objects_1) Sep 18 21:48:00 and a little later on: cfdisk_OBJECTS = $(am_cfdisk_OBJECTS) Sep 18 21:48:15 so of course when it gets built, there are no object files... Sep 18 21:48:32 why is it being commented out? where is that coming from? Sep 18 21:48:38 I'm not an automake guru. Sep 18 21:50:05 idk im trying to compile x86 image now Sep 18 21:50:23 RealOpty: ok, I can't spend any more time on this. I need to get the geos platform up this weekend. Sep 18 21:50:43 anyone have an idea what the quickest workaround is to getting trunk to build? Sep 18 21:50:59 remove fdisk Sep 18 21:51:04 and use busybox fdisk? Sep 18 21:51:16 why u need fdisk lol? Sep 18 21:52:20 my build failed :| Sep 18 21:52:37 on the larger scheme of things... I'm surprised that there's not an automated build daemon that runs after every checkin and emails the committer breakage logs... Sep 18 21:53:15 same error i486-openwrt-linux-uclibc-gcc: no input files Sep 18 21:53:35 well, I need fdisk because I'm testing various other features as well, such as using a squashfs image partition, and overlaying writable files via a unionfs or aufs. Sep 18 21:53:46 yeah... that's the kicker. Sep 18 21:56:10 think i found the prob Sep 18 21:56:12 testing now Sep 18 22:03:37 Who is Jo-Philipp? Sep 18 22:03:47 err... what's his handle? Sep 18 22:04:12 philipp64|laptop: jow_laptop / xMff Sep 18 22:04:16 ^ Sep 18 22:04:22 ah... Sep 18 22:04:31 fix didnt work. Sep 18 22:04:34 :/ Sep 18 22:04:46 xMff: can you fast-track commit the patch I just sent in? Sep 18 22:05:11 it's a trivial missing dependency.... Sep 18 22:07:15 ya. its weird lol i had this working the other day for brcm47xx based off the changeset i showed you. Sep 18 22:07:35 hmmm.... I'm wondering if there isn't a trivial way to do cross-checking of the DEPENDS:= values (in netdevices.mk) versus the what the module itself reports via modinfo. Sep 18 22:07:44 any build system experts out there? Sep 18 22:11:06 philipp64|laptop, just to be clear, you need fdisk working not cfdisk? Sep 18 22:14:15 correct. Sep 18 22:14:41 though I've never used cfdisk, so I'm not sure what the difference is. Sep 18 22:15:01 if it's functionally similar enough, I could get by using cfdisk I suppose. Sep 18 22:15:04 philipp64|laptop, i think i have a fix, confirming now Sep 18 22:15:12 deja vu. :-) Sep 18 22:15:27 or deja entendu, anyway. Sep 18 22:16:31 ok so in the makefile for util-linux-ng Sep 18 22:16:39 in this section Sep 18 22:16:40 define Build/Compile Sep 18 22:17:00 * RealOpty waits for compile to finish Sep 18 22:18:30 on this line $(MAKE) -C $(PKG_BUILD_DIR)/fdisk fdisk cfdisk sfdisk Sep 18 22:18:34 remove cfdisk Sep 18 22:18:54 temp fix :\ Sep 18 22:18:57 then Sep 18 22:19:00 tried that already... it worked, but then setterm failed. Sep 18 22:19:17 you might want to add libncurses Sep 18 22:19:25 I think that is the correct fix, though. Sep 18 22:19:29 you useing setterm? Sep 18 22:19:42 because cfdisk gets built when fdisk/ gets built, right? Sep 18 22:19:49 yup Sep 18 22:19:53 so no explicit special build should be required. Sep 18 22:20:44 so idk this fix worked for me. i only have the package fdisk selected. Sep 18 22:21:05 though the real issue needs to be resolved, im sure it will be soon. Sep 18 22:22:32 if u want u can just remove fdisk and all that stuff and just install a package Sep 18 22:23:13 yow! those ultra-III CF's are *really* fast, even over USB. Sep 18 22:23:30 why did I every buy discount CF's???? :-( Sep 18 22:25:21 ok... have some initialization questions.... Sep 18 22:25:57 how do I specify that I want to build mppe.ko, but not necessarily have it be insmod'd at startup? Sep 18 22:26:41 also, where do I specify target specific module options (i.e. "options cs5535_gpio mask=0x....") goes where for a specific target? Sep 18 22:27:22 and lastly.... I need to do some target-specific mknod's just after cs5535_gpio gets loaded up... where do I do that? Sep 18 22:28:43 well... not just mknods... I also need to a do an: echo "1" > /dev/gpio_27 Sep 18 22:28:46 for instance... Sep 18 22:30:13 philipp64|laptop: modules.d has module files to load and order/options Sep 18 22:30:47 goodasGNU: I know... but where do targets specify private settings? Sep 18 22:31:09 in that file after the module name Sep 18 22:33:35 sorry, not explaining myself very well. Sep 18 22:33:45 not on the running system. Sep 18 22:34:10 where in the source tree should a platform contain these settings and scripting? Sep 18 22:35:22 make a files/etc/modules.d directory in the root of the build tree, add the cs5335 file with the desired options there. Sep 18 22:37:17 goodasGNU: I think you missed some early conversation. I'm bringing up a new platform, and I want to add *all* the settings-in-a-can to the source tree so that everyone can build this target, and get all the configuration state they need. Sep 18 22:37:59 sure, then put it in target/linux//files Sep 18 22:38:11 I will commit my changeset, and want target/linux/x86/geos to contain *all* the state necessary... Sep 18 22:40:28 sorry its "base-files" for the root tree Sep 18 22:40:54 build #102 of cobalt is complete: Success [build successful] Build details are at http://tksite.gotdns.org:8010/builders/cobalt/builds/102 Sep 18 22:40:56 so target/linux//base-files Sep 18 22:41:18 ok.... what about init.d scripts? Is there anything particular about those? Sep 18 22:41:47 same place, any files to the rootfs will be added per target from there Sep 18 22:42:30 they should overwrite same files frin the generic build, accept for files/ in build root Sep 18 22:48:53 anybody know of an intelligent way to detect link status on a port Sep 18 22:53:29 ethtool eth0 | grep Link detected: Sep 18 22:54:06 goodasGNU: ok... so that takes care of passing a "mask=" option to insmod... but if I want to customize /etc/init.d/led ? Sep 18 22:55:17 goodasGNU: nope ethtool, mii-tool ifplugd state no link but its definatley up Sep 18 22:56:32 ok, and backing up a bit... so I want to build certain kernel modules (protocol modules) but I don't want to explicitly have them loaded up at boot time... like pppoatm.ko for instance... Sep 18 22:57:35 philipp64|laptop:: make your target/linux//base-files/etc/init.d/led It will be used when your target is selected Sep 18 22:59:44 yeah...problem is that if the "standard" way to initialize LEDs evolves, then my platform will diverge, or I'll have to keep importing those changes. Sep 18 23:00:10 might be easier if I can figure out a way to layer my stuff atop the standard stuff. Sep 18 23:02:25 so make a led- and call it after the std led script with START=97 Sep 18 23:03:25 OutBackDingo: status should be correct in /proc/net/dev then Sep 18 23:04:13 carrier says 0 Sep 18 23:04:29 goodasGNU: cat /sys/class/net/eth0/carrier says 1 Sep 18 23:04:42 /proc/net/dev says 0 under carrier Sep 18 23:04:50 but has stats Sep 18 23:04:52 kinda odd Sep 18 23:04:58 atheros soc Sep 18 23:05:05 the stats are the clue Sep 18 23:05:30 its odd tho Sep 18 23:12:25 goodasGNU: having to do: MAJOR="`grep cs5535_gpio /proc/devices | cut -d' ' -f1`" Sep 18 23:12:35 inside a script... is that acceptable? Sep 18 23:12:53 can I count of grep and cut always being present? Sep 18 23:13:16 or I could do it using awk... Sep 18 23:14:12 cut and grep are defaults in busybox cfg as well as awk Sep 18 23:15:25 is there any way to declare that two packages can not be installed at the same time? Sep 18 23:15:48 ...a negate dependency? Sep 18 23:16:36 hmmm.... MAJOR=$(awk '$2 == "cs5535_gpio" { print $1; }' /proc/devices Sep 18 23:17:09 nunojpg: depends on !PACKAGE_X Sep 18 23:18:27 havent done that one yet.. Sep 18 23:20:25 goodasGNU: ? Sep 18 23:20:39 negative deps Sep 18 23:20:45 ah. Sep 18 23:21:21 nbd: thats definately his territory :) Sep 18 23:30:37 philipp64|laptop: DEPENDS:=!python? Sep 18 23:31:41 nunojpg: !@PACKAGE_python Sep 18 23:38:22 cshore: @!PACKAGE_python instead. Sep 18 23:38:38 nunojpg: ooops, sorry Sep 18 23:38:40 but the idea is to be able to select and package it, but not to install it together Sep 18 23:39:13 nunojpg: no such possibility exists Sep 18 23:39:19 nunojpg: at present Sep 18 23:57:50 cshore: saved me some time, thanks Sep 19 01:08:57 goodasGNU: hmmm..... my private copy of /etc/modules.d/50-gpio-cs5535 is getting stomped (if it ever gets installed... I'm not seeing it) by the generic one. what am I doing wrong? Sep 19 01:34:28 location? Sep 19 01:35:13 eh? Sep 19 01:35:29 where is your copy located? Sep 19 01:36:02 in target/linux/x86/geos/base-files/etc/modules.d/50-gpio-cs5535 Sep 19 01:38:35 see ftp://ftp.redfish-solutions.com/pub/openwrt-geos.patch Sep 19 01:41:53 check the olpc target.mk , this target has some base-files added. Sep 19 01:43:21 does it work if you put it in /x86/base-files? Sep 19 01:48:09 no... the mask value for gpio-cs5535 is very target sepcific. Sep 19 01:48:17 *specific too. Sep 19 01:51:26 ack, I mean if you place it there does it make it to the image. If not, you may have other issues with you buildroot Sep 19 02:04:57 goodasGNU: have a look at include/kernel.mk ... ModuleAutoLoad Sep 19 02:08:06 don't stare at it too long. my head is starting to explode. Sep 19 02:10:34 dont think you should need any post install unless your doing the module as its own ipk. Sep 19 02:12:49 this is the modules for the IDE controler right? Sep 19 02:13:15 line 70. Sep 19 02:13:56 no, GPIO. i.e. ttl-levels used to drive LED's. Sep 19 02:14:20 see it Sep 19 02:14:30 the /etc/modules.d/ file is created right about there. Sep 19 02:14:41 not sure what the getvar mod is about. Sep 19 02:14:54 who are thepeople and nbd? Sep 19 02:15:21 nbd: ping Sep 19 02:15:38 thepeople: ping Sep 19 02:16:10 nbd is the person to best answer you question on that. thepeople mainly did the x-wrt extensions Sep 19 02:16:18 was hoping to get an authoritative answer about how this stuff works. Sep 19 02:16:48 nbd can help you for sure, he was around about 40 min ago on #openwrt **** ENDING LOGGING AT Sun Sep 19 02:59:56 2010