**** BEGIN LOGGING AT Mon Jun 09 02:59:56 2008 Jun 09 07:57:40 bonjour Jun 09 08:07:50 03deller 07org.oe.stable * re33c0ba6... 10/ (4 files in 3 dirs): Jun 09 08:07:54 poodle: add pcmcia modules to MACHINE_EXTRA_RRECOMMENDS Jun 09 08:07:56 Acked-by: Koen Kooi Jun 09 08:07:58 Acked-by: xjqian Jun 09 08:08:00 Acked-By: Marcin Juszkiewicz Jun 09 08:08:05 03koen 07org.oe.stable * r7214a86e... 10/ (1 conf/distro/angstrom-2007.1.conf): angstrom: bump to r18 Jun 09 08:45:28 hi Jun 09 08:45:48 I'm trying to flash an akita, but the flash tool keeps saying it can't find the CF card? Jun 09 08:45:56 the card works fine with a collie Jun 09 08:46:44 any suggestions? Jun 09 10:43:05 good morning Jun 09 10:43:12 broonie: ping Jun 09 10:43:21 pong Jun 09 10:46:25 broonie: Do you have a minute... I ran into trouble with the wm9705 in my device here. Jun 09 10:46:42 Sure... Jun 09 11:04:34 03tmbinc 07org.oe.dreambox * r5cd11fd9... 10/ (1 conf/machine/dm7025.conf): dm7025.conf: some modules are compiled in kernel Jun 09 11:15:54 oops Jun 09 11:16:53 broonie: I guess you remember that I applied a patch from your tree to my one here some time ago. Jun 09 11:17:29 broonie: The touchscreen driver worked out of the box, but now I started to work on audio support. Jun 09 11:20:53 broonie: The audio codec driver for the wm9705 and the new soc alsa framework I found in the mobile-linux tree. Jun 09 11:21:59 But soc alsa does not seem to be compatible with the arm alsa drivers which are used by the touchscreen driver. Jun 09 11:22:21 ASoC and the touch drivers should be compatible. Jun 09 11:22:43 What problems are you seeing? Jun 09 11:23:19 hmmm, is it normal that on a C1000, write speeds to CF/SD are absurdly slow? Jun 09 11:25:02 broonie: I had to turn off the alsa arm driver to make soc alsa build, but doing this my touchscreen stopped to work. Jun 09 11:26:03 Could you go into more detail on what "stopped to work" involved? Jun 09 11:26:28 You do need to disable the vanilla AC97 driver, you can't have two drivers for the same thing. Jun 09 11:26:52 You also need to provide a machine driver for your system. Jun 09 11:28:22 broonie: ah, maybe that's what I'm missing. I do not have one yet and this worked without just using the vanilla ac97 and touchscreen driver. Jun 09 11:28:58 If you don't provide an ASoC machine driver then no devices will be discovered on the AC97 bus. Jun 09 11:29:49 This will cause the touch driver to fail to probe. Jun 09 11:32:21 broonie: okay, then I'll add one now... looks quite simple. Thanks! Jun 09 12:00:35 Tartarus: ping Jun 09 12:08:03 Do somebody tries to boot any arm distro in qemu-system-arm with -mtdblock option? :) Jun 09 12:08:16 I need some advice.. Jun 09 12:17:06 * * OE Bug 4368 has been created by jorik(AT)kippendief.biz Jun 09 12:17:08 * * Grub 0.97 fails to build wih automake 1.10 Jun 09 12:17:10 * * http://bugs.openembedded.net/show_bug.cgi?id=4368 Jun 09 12:24:56 I am missing lots of packages in deploy/glibc/ipk/armv5te/, what bitbake commands remove packages? Jun 09 13:39:09 does anyone know where libpthread_nonshared.a is coming from? Jun 09 14:15:39 second question: do somebody tries to cross-compile fvmw2? Jun 09 14:47:02 is there a package containing a non-native mkisofs in oe? Jun 09 14:47:24 cdrtools is only available in native-form, and I definitely want to avoid looking into the schily makefile system Jun 09 14:48:19 I don't see one :( Jun 09 14:48:32 send him a friendly mail Jun 09 14:48:55 isnt cdrtools forked to something sane by debian? Jun 09 14:49:13 wodim and others Jun 09 14:49:35 cdrkit i think Jun 09 14:50:14 hi Jun 09 14:54:28 hi hrw Jun 09 15:06:02 broonie: Looks good so far... the mainstone ts driver was easy to adapt (with only a very few changes). Now I only need to find the bits to make sure probing starts with correct device :) Jun 09 15:08:17 florian: Probing starts with correct device? Jun 09 15:10:02 florian: look at tosa? it use wm9712 but for audio and ts Jun 09 15:11:50 broonie: wm97xx-ts probes wm97xx-touch... but which one is responsible for probing wm97xx-ts? Jun 09 15:12:45 It's probed when the AC97 device comes up. Jun 09 15:15:45 broonie: That's what I expected... then I still have somethign wrong here. The driver is there but it is never probed. Jun 09 15:16:08 florian: Is your SoC audio subsystem coming up? Jun 09 15:16:20 hrw: Yep... this one told me to to add it manually :) Jun 09 15:17:16 broonie: yes... and the wm97xx-ts driver is there as well. (e.g. you can see the matching driver entry in sysfs) Jun 09 15:17:20 The SoC machine driver needs to be hand crafted for each machine. It needs to explicitly configure all the parts of teh DAI. Jun 09 15:17:40 Registering the driver != the driver starting up. Jun 09 15:18:00 broonie: Yes, this one I have already. Jun 09 15:18:06 broonie: sure Jun 09 15:18:27 I only see the driver, but not a matching device. Jun 09 15:18:35 03jeremy_laine 07org.oe.dev * r0f761bf9... 10/ (1 packages/linux/linux.inc): linux.inc: add a symlink from /boot/devicetree to /boot/devicetree-${KERNEL-VERSION} Jun 09 15:19:03 Your machine driver needs to register the device as well as the driver. Jun 09 15:20:23 s/as well as/instead of/ sorry - the driver is registered by the SoC core Jun 09 15:24:12 | checking if mkstemp is safe... configure: error: cannot run test program while cross compiling Jun 09 15:24:27 how to avoid such problem? Jun 09 15:24:35 what is true way? Jun 09 15:26:08 Jay7: You need to put in the variable into the matching site file. Jun 09 15:26:41 hmm.. Jun 09 15:27:21 what do you mean at "matching site file"? :) i'm not autotools guru yet :) Jun 09 15:27:43 right now i just comment out this check in configure.ac Jun 09 15:29:16 bye guys Jun 09 15:30:00 Jay7: Depending on the architecture you build for. There is a directory "site" in OE's root directory containing the site files ordered by target arch. You can get the variable name from config.log usually. Jun 09 15:30:23 zecke: We have a couple of issues with bitbake atm - one is the ongoing environment discussion. I put a proposal to the list after our last discussion, did you look at it? Jun 09 15:30:25 ok, I'll try :) Jun 09 15:30:48 zecke: Secondly, I added a fix which replaced dashes in function names with underscores Jun 09 15:31:35 broonie: right, I think I found the issue. I still register the "pxa2xx-ac97" device in my machine driver. Jun 09 15:31:39 zecke: That "fix" has broken all the postinst/preinst stuff :( Jun 09 15:33:44 RP: let me take a look at the bitbake patch again Jun 09 15:34:55 zecke: Not having dashes in shell function names is technically correct but we use the postinst stuff in strange ways... Jun 09 15:35:19 I'm torn between fixing all the function names, hacking package.bbclass or reverting the change to bitbake. Jun 09 15:36:38 RP: personnaly I have pinned bitbake to r1071 :) Jun 09 15:38:15 RP: i never did really understand the commit to bitbake, what does it do, rename all the functions so they are "shell compliant" ? Jun 09 15:38:18 jeremy_laine: When I looked through the functions I'm a bit shocked at the mess we have as Koen mentioned on the mailing list Jun 09 15:38:35 RP: pristine env: I'm happy to apply the patch. We do need some patches to bitbake.conf right? Jun 09 15:38:38 jeremy_laine: shell functions have their names tweaks so they can acutally be used in shell Jun 09 15:39:08 zecke: I think so, I can't remember if I merged the tweaks from poky or not Jun 09 15:39:09 RP: yeah but if the function is renamed, all the calls to the function need to change too, how do you do that? Jun 09 15:39:19 zecke: Obviously I would check that before merging Jun 09 15:39:42 jeremy_laine: You'd use the correct name at the callsites Jun 09 15:40:04 jeremy_laine: Having bitbake error if it sees it is looking more tempting tbh Jun 09 15:40:14 RP: and why should we rename the function then, why not barf if the function is not "shell correct"? Jun 09 15:40:15 RP: cool, so pristine env is resolved? let me take a look at the next issue Jun 09 15:40:49 zecke: I've been using that patch locally and it works Jun 09 15:40:57 RP: also, I think you have an issue when the function name needs expansion Jun 09 15:41:30 RP: I seem to remember working around the bitbake problem by using pkg_postinst_${PN} instead of pkg_postinst_my-package and it worked Jun 09 15:41:40 (not 100% sure) Jun 09 15:41:43 RP: sorry, I'm late in the game: http://svn.berlios.de/viewcvs/bitbake?view=rev&rev=1072 is causing issues? Jun 09 15:42:21 zecke: yes Jun 09 15:42:30 zecke: yup, see http://bugs.openembedded.org/show_bug.cgi?id=4316 Jun 09 15:42:43 zecke: #4316 is the report of the fallout Jun 09 15:43:15 jeremy_laine: you're faster than me :) Jun 09 15:47:18 ah okay, so the 2nd hunk in line 249 is manipulating the vars? Jun 09 15:47:57 florian: this is non standart variable defined in fvwm's acinclude.m4.. I think adding it to site config is wrong way.. Jun 09 15:48:31 zecke: yes Jun 09 15:49:54 RP: what do we need that for? with my current gut feeling I would revert that "fix" and look into other options to correct that. E.g. "inherit foo-bar" would trigger the error? Jun 09 15:52:19 zecke: The original bug was #2958 Jun 09 15:53:00 zecke: shell functions shouldn't have dashes in the name, that is a valid point. Perhaps the better fix is to have bitbake error if it finds any when parsing though Jun 09 15:55:11 Well... foo_BAR-FOO, we would have to error out when we attempt to call such a function? or emit it? it is valid to have that function in memory due the override system, when we want to call/emit it it might violate shell standards Jun 09 15:56:12 zecke: So we error if we try to emit such a function? Jun 09 15:56:58 RP: well, I'm certain we do emit these functions.. :( Jun 09 15:57:54 RP: so we would have to stop emitting them when they are illegal and see what is breaking... anyway I would revert the current patch and search for another way to have dashes in class names :} Jun 09 16:00:25 zecke: I don't think we can be using them but agreed, it needs a different approach. Thanks for the second opinion Jun 09 16:22:15 So I'm having a problem where we have an inhouse build system that uses make heavily. Unfortunately, when I attempt to build my package with bitbake, make echo's the command to be run rather than exec'ing it. I'm curious if anyone has experienced a similar issue with make under bitbake. Jun 09 16:39:45 broonie: OK, found it... a bug in the codec driver. It did not set the dai type and therefore the probe never happened Jun 09 16:40:01 Fidget: I've never seen or heard of that problem before... Jun 09 16:42:18 RP: Its a heavily templatized make system, and its internals break when executed from within OE. I've run strace and make -d, and the only thing that stands out is that it writes the command to stdout rather than performing an exec. I was kinda hoping someone else might use a make system that tries to build cross platform stuff and calling that from within OE. Jun 09 16:43:05 Whats 'env' look like? Jun 09 16:43:17 I've seen bitbake fail when odd things (like CDPATH) are used Jun 09 16:43:19 Tartarus: sane Jun 09 16:43:23 maybe make has a similar setting Jun 09 16:43:26 hmm Jun 09 16:43:28 I can recheck that Jun 09 16:43:43 Don't just go for sane go for minimal :) Jun 09 16:44:18 Well, I was running: /usr/bin/env -i CC=${CC} PATH=${PATH} ./make Jun 09 16:44:23 and it failed Jun 09 16:45:02 Can I post 5 lines of code in here, or use pastebin? Jun 09 16:46:25 http://pastebin.com/d4c395085 Jun 09 16:47:19 So that do_compile fails, whereas if I run the exact same from the working directory, it succeeds Jun 09 16:49:08 Fidget: One rather suspects that one would need to see what the makefile does in order to figure out what might be going on. Jun 09 16:49:49 mwester: make -d is the best I can really do. Its a build system that uses make for the mechanics Jun 09 16:51:04 :) I'm not sure how anyone can help you then. Or perhaps I'm misunderstanding your question. Jun 09 16:52:27 mwester: I'm asking if anyone has a similar problem where make invoked via bitbake write's output to stdout instead of exec'ing the line. That is, in a normal shell, I run make and it will exec lines. Via bitbake, make is run in do_compile, but instead of exec'ing lines, it writes them to stdout. Jun 09 16:53:21 so everywhere gcc -c foo.c -o foo.o, rather than exec'ing, it echo's it to stdout Jun 09 16:54:28 Ah, ok. To answer your question, No I've never heard of nor encountered such a thing with bitbake. Or any other case where Make is used, for that matter. Jun 09 16:54:39 mwester: neither had I Jun 09 16:55:11 mwester: which is why I find this perplexing and annoying, it seems that makes behavior changes and I cannot explain how or why. Jun 09 16:55:19 To offer my opinion: the makefile has special code in it that does something exceedingly strange, that can be understood only by examination of the Makefile itself. Jun 09 16:55:49 Make has little built-in behavior; almost all of it's behavior comes from the Makefile itself. Jun 09 16:55:59 mwester: It isn't a Makefile. Its a make build system. Jun 09 16:56:14 same difference :) Jun 09 16:56:23 Ok, same thing -- instead of a Makefile it's a collection of scripts, programs, tools, command, operating systems, etc. Jun 09 16:56:39 Tartarus: Then can I call bitbake just some python code? Jun 09 16:56:49 Yes. Jun 09 16:56:57 oh Jun 09 16:57:27 it happens to be "some python code" that performs a specific task. Jun 09 16:57:37 Just as a Makefile performs a specific task. Jun 09 16:58:04 But semantics aside, the issue is that you have a collection of "stuff" (to use a generic term :) that behaves oddly when run from bitbake. Jun 09 16:58:42 succinctly, that is my problem Jun 09 16:59:03 or at least one of my identifiable problems. Jun 09 16:59:44 My experience has been pretty consistent -- when make stuff misbehaves under bitbake (or any other tool), then one has to instrument and examine the code and config files (and Makefile) used by make to sort out the problem. Jun 09 17:00:11 There is little behavior that is specific to the make utility, it is a remarkably stupid build tool in that sense. Jun 09 17:04:51 mwester & Tartarus: thanks for the input, I'll delve into the debug and see if I can see a magic flag that changes makes behavior Jun 09 18:05:09 broonie: Your codecs are too complex ;) I can't find the right switch to turn the mono output on. (but hp out works) Jun 09 18:16:05 XorA|gone: can you commit http://linuxtogo.org/~lgorris/linux-kaiser_ltg.diff ? Jun 09 18:17:11 dcordes: I wont commit SRCREV=$AUTOREV Jun 09 18:20:50 hi ph5 Jun 09 18:21:13 evening woglinde Jun 09 18:33:38 XorA|gone: why not? Jun 09 18:33:58 dcordes: that doesnt belong in a recipe Jun 09 18:34:08 I'm not going to break the repository Jun 09 18:34:15 ok I will change this to the current git head Jun 09 18:34:37 dcordes: no, Ill commit the rest, put the AUTOREV bit in your local.conf Jun 09 18:34:53 ok good Jun 09 18:34:56 dcordes: AUTOREV doesnt belong in a recipe Jun 09 18:35:09 * XorA|gone is just puling git then will commit Jun 09 18:36:15 XorA|gone: are you pulling with SRCREV = "4134160a774aad737ac3efadd81366e3c9f27c08" ? Jun 09 18:36:23 dcordes: no, AUTOREV Jun 09 18:40:52 03xora 07org.oe.dev * ra42c91c4... 10/ (1 packages/linux/linux-kaiser_2.6.24+git.bb): linux-kaiser_2.6.24+git.bb : update from dcordes Jun 09 18:57:06 * * OE Bug 4369 has been created by gwossum(AT)acm.org Jun 09 18:57:08 * * binutils-2.17 won't build for AVR32 Jun 09 18:57:10 * * http://bugs.openembedded.net/show_bug.cgi?id=4369 Jun 09 19:01:47 lots of packages won't build. Jun 09 19:13:30 pplz, which package provides xft-config? Jun 09 19:15:17 hm, already found other way :) Jun 09 19:28:05 * * OE Bug 4370 has been created by gwossum(AT)acm.org Jun 09 19:28:07 * * Eliminate need for gettext Jun 09 19:28:09 * * http://bugs.openembedded.net/show_bug.cgi?id=4370 Jun 09 19:31:17 gpephone-image failed when compiling with angstrom-2008.1 and org.openembedded.stable and machine omap5912osk. Jun 09 19:31:28 shouldn't stable be stable or am i doing something wrong. Jun 09 19:31:54 linphone is failing. Jun 09 19:35:58 Stable is stable, but not everything is bugfixed Jun 09 19:36:03 Can you pastebin your error log? Jun 09 19:37:24 okay Jun 09 19:39:56 I blew away my tmp/rootfs, is there some command I can tell bitbake to rebuild it? bitbake -c rebuild myimage doesn't seem to get the dependent packages. Jun 09 19:41:30 Fidget: bitbake should do it Jun 09 19:42:24 http://pastebin.com/m1ee3397 <--- here is local.conf, and error logs Jun 09 19:42:31 brb Jun 09 19:45:33 Note that tmp/rootfs isn't bootable directly Jun 09 19:45:46 re Jun 09 19:46:12 Fidget, if you want to see foopkg rebuilt and put into fooimage, you need to do bitbake -c rebuild foopkg && bitbake fooimage Jun 09 19:46:21 (or bitbake -c clean foopkg && bitbake fooimage) Jun 09 19:46:38 Tartarus: the clean is what I was missing, thanks Jun 09 19:47:09 markus, for OE.stable, I _think_ you want angstrom-2007.1 not 2008.1 Jun 09 19:47:12 not 100% sure tho Jun 09 19:47:42 That kinda looks like a "newer GCC is stricter than older GCC" kind of problem Jun 09 19:48:43 markus, yeah, see http://www.angstrom-distribution.org/building-angstrom Jun 09 19:51:20 XorA|gone: cheers Jun 09 19:52:08 ok, do i want to build gpephone-image with angstrom Jun 09 19:52:09 ? Jun 09 19:52:21 Tartarus: but still, it should be able to build? Jun 09 19:53:01 markus, with DISTRO=angstrom-2007.1 you'll get a different GCC version Jun 09 19:53:12 Which in turn won't have the warning there which is treated as an error Jun 09 19:53:13 I suspect Jun 09 19:53:30 aha... but with dev it will still fail? Jun 09 19:53:37 i will try 2007.1 Jun 09 19:53:45 markus, with .dev perhaps linphone has been fixed for that gcc version Jun 09 19:54:02 markus, .stable is a snapshot of .dev from a while back, with certain bugs fixed Jun 09 19:54:24 markus, I _think_ DISTRO=angstrom-2008.1 isn't really "valid" for the stable branch, but I don't know for certain Jun 09 19:54:45 i chose 2008.1 by mistake Jun 09 19:56:06 pplz, can *.rootfs.jffs2 image be mounted with qemu-system-arm -mtdblock ? Jun 09 19:56:36 I have trying to boot angstrom in qemu Jun 09 19:56:42 without success :) Jun 09 19:56:58 kernel is booting but can't mount rootfs Jun 09 19:57:11 Jay7, can you pastebin the error? Jun 09 19:57:18 iirc you can have qemu spit out to terminal Jun 09 19:57:43 I cat write :) Jun 09 19:58:28 when using jffs2 image - too much crc errors Jun 09 19:58:34 then kernel mounts rootfs Jun 09 19:58:41 but can't find init Jun 09 19:58:52 when using ext2 image - cant mount rootfs Jun 09 19:59:54 cmdline root=/dev/mtdblock2 rootfstype=jffs2 init=/bin/sh Jun 09 20:00:00 -M akita Jun 09 20:00:31 question is - can qemu-system-arm properly emulate zaurus NAND via -mtdblock? Jun 09 20:00:32 Yeah, it's hard to say without the whole console bootup log :) Jun 09 20:00:39 Ah, hmm Jun 09 20:00:45 I wonder, might be NOR only Jun 09 20:00:56 haven't played with qemu system that much Jun 09 20:01:39 I can pastebin entry bootlog, but it's more useless.. Jun 09 20:03:37 $OE_ROOT/tmp/staging/i686-linux/bin/qemu-system-arm -M akita -mtdblock x11-image-akita.jffs2 -kernel zImage-akita.bin -append 'root=/dev/mtdblock2 rootfstype=jffs2 init=/bin/sh' Jun 09 20:04:01 when somebody want's to try - it was my cmdline :) Jun 09 20:07:13 grr... Jun 09 20:07:31 I trying to compile fvwm under OE Jun 09 20:07:57 fvwm's configure takes xft-config from host system.. Jun 09 20:08:15 can this be resolved? Jun 09 20:08:49 Perhaps... Jun 09 20:08:59 Is the cross version not built because of a missing dep? Jun 09 20:09:05 and, lunch time :) Jun 09 20:09:10 I have recipe written already, but.. Jun 09 20:09:33 don't know... I can't find xft-config in packages Jun 09 20:09:51 pkg-config xft says nothing too.. Jun 09 20:10:11 lunch is good :) Jun 09 20:18:26 * flo_lap just found an image of the LinuxTag booth in c't magazine Jun 09 20:19:37 How can I query what packages provide certain interfaces? Jun 09 20:19:52 I need libstdc++ and libgcc1 Jun 09 20:19:55 not sure what provides those Jun 09 20:20:24 easy way - grep -- 'libstdc++' */*.bb Jun 09 20:29:38 Hmm Jun 09 20:29:49 Has anyone ever seen bitbake take a while to start printing that it's looking at the cache? Jun 09 20:29:59 yes Jun 09 20:30:05 Fix it? :) Jun 09 20:30:07 but I have other issues. Jun 09 20:30:12 heh, k Jun 09 20:30:33 Tartarus: Which bitbake version? Jun 09 20:30:41 1.8.8 Jun 09 20:30:59 1.8.10 had some optimisations for that Jun 09 20:31:11 hi rp Jun 09 20:31:18 ok Jun 09 20:31:19 hi woglinde Jun 09 20:39:24 btw, bitbake fails on reparse command :) Jun 09 20:39:59 flo_lap: The datasheets generally have a picture which is helpful :) Jun 09 20:40:56 broonie: yeah, printed it alerady... i'll take a look at it tomorrow :) Jun 09 20:42:28 broonie: To make things even worse this device has external switches for audio lines. (no idea why... i guess someone did not understand what the wm9705 does) Jun 09 20:43:00 Could just be that the WM9705 was dropped into an existing design; there is a standard pinout for AC97 devices. Jun 09 20:45:11 broonie: maybe... but judging from the other parts of the design i'd say it is just a good example for an over-engineered piece of hardware Jun 09 20:45:46 :) Jun 09 20:45:52 What is it, OOI? Jun 09 20:58:06 broonie: OOI? Jun 09 20:59:07 ... i hate building the toolchain... grr :) Jun 09 20:59:33 OT, but anyone know what would cause an int3 trap in the kernel from an application, but without an app segfault? I've learned that it indicates app misbehavior, but not exactly what it's doing Jun 09 21:01:31 Out Of Interest. Jun 09 21:09:11 03likewise 07org.oe.dev * r2fce1b18... 10/ (4 files in 4 dirs): linux-rt-2.6.25: Bump to -rt6 and add efika. Note that -rt is unstable. Jun 09 21:38:57 kergoth: some kind of debug-related activity, presumably. single-stepping maybe? Jun 09 21:39:45 * kergoth doublechecks the bug Jun 09 21:41:17 iirc, int3 is the debug vector; it's special because there's a single byte instruction (0xcc) to trigger it, so you can always poke that on top of any other instruction to force a debug trap. Jun 09 21:43:10 I did it! fvwm2 compiled :) Jun 09 21:43:18 ahh, that's interesting Jun 09 21:43:55 it's an ltrace where a customer is seeing the messages, so i guess it's making use of it. apparently they can just echo 0 > /proc/sys/debug/exception-trace to shut it up Jun 09 21:45:35 kergoth, I think you need to focus Jun 09 21:45:42 hm? Jun 09 21:49:10 urg, I guess the ltrace remark was meant for this channel Jun 09 21:52:07 * * OE Bug 4371 has been created by deller(AT)gmx.de Jun 09 21:52:09 * * opie-mail is missing in the feeds Jun 09 21:52:11 * * http://bugs.openembedded.net/show_bug.cgi?id=4371 Jun 09 21:52:40 yep. OT, but there are often clueful folk in here, so figured i'd ask Jun 09 21:55:08 See, that was your first mistake.. :) Jun 09 21:57:03 kergoth, don't mind me, doing to many things at once Jun 09 21:57:16 know how that is.. kinda miss it, actually... Jun 09 22:11:40 Bah, OE.dev needs the RDEPENDS=binutils-symlinks that poky has Jun 09 22:11:56 Tartarus: hey Jun 09 22:12:01 hi ant Jun 09 22:12:10 samba stuff? ;) Jun 09 22:12:13 Tartarus: have you checked your udev commit? Jun 09 22:12:15 ANGSTROM_EXTRA_INSTALL = "mc gpe-bluetooth gpsd speechd flite navit gpsdrive" Jun 09 22:12:22 ?? sorry Jun 09 22:12:31 http://gitweb.openembedded.net/?p=org.openembedded.dev.git;a=commit;h=55df9cf2cd11aed3b1b9333f72d518d1763b180f Jun 09 22:12:35 ant_, you mean if it's still there, or if something went wrong with it? Jun 09 22:12:49 yea, is it as intended? Jun 09 22:13:13 Yes Jun 09 22:13:21 uh oh... Jun 09 22:13:25 Why? Jun 09 22:13:33 !oebug 4118 Jun 09 22:13:34 * * Bug 4118, Status: NEW, Created: 2008-03-25 08:47 Jun 09 22:13:35 * * xjqian(AT)gmail.com: udev-118 failed to create device nodes on Tosa Jun 09 22:13:36 * * http://bugs.openembedded.net/show_bug.cgi?id=4118 Jun 09 22:14:07 the init from 115 seems to solve BTW Jun 09 22:14:16 ok Jun 09 22:14:29 but it is definitely unfinished state Jun 09 22:14:36 (115) Jun 09 22:15:13 ah ha, looks like i need to swipe http://www.x86-64.org/pipermail/discuss/2005-August/006413.html for this customer Jun 09 22:15:19 hm Jun 09 22:15:29 ant_, actually have the hw? Jun 09 22:15:45 Running the udev script w/ sh -x should help show whats not happening Jun 09 22:16:17 ok, I tried udevmonitor but did not sense the card insertion Jun 09 22:17:33 hm Jun 09 22:18:55 but *alot* of packages could be broken at this point (util-linux-mount)...bitbake -> berzerk Jun 09 22:19:30 Well Jun 09 22:19:36 Can you see if udevd really is running? Jun 09 22:19:40 Let me rebuild with udev_118, I have 092 right now... Jun 09 22:19:44 ok Jun 09 22:22:01 I was suspecting - # if this directory is not present /dev will not be updated by udev Jun 09 22:22:01 - mkdir -p /dev/.udev/ /dev/.udev/db/ /dev/.udev/queue/ /dev/.udevdb/ Jun 09 22:22:35 Shouldn't be Jun 09 22:22:43 or a broken udevsettle Jun 09 22:23:25 but koen explicitely mentions mount / umount are broken Jun 09 22:23:42 in which? Jun 09 22:24:01 any postint with a - in the name wont run Jun 09 22:24:15 http://lists.linuxtogo.org/pipermail/openembedded-devel/2008-June/005540.html Jun 09 22:24:21 ant_, but keep in mind what modern udev initscripts do/don't do for udev Jun 09 22:25:04 sorry, I've yet to make my mind (used *nix 10 years before udev) Jun 09 22:25:15 heh Jun 09 22:25:19 udev is relatively new for me Jun 09 22:25:25 for everyone Jun 09 22:25:35 phhew Jun 09 22:25:37 it's still a silly damn thing in a lot of regards Jun 09 22:25:39 ;-) Jun 09 22:25:54 (Internally we added a hook to prepopulate /dev and just use udev for hotplug) Jun 09 22:26:41 (Or rather, stop after doing prepopulate) Jun 09 22:27:40 Tartarus: I see each "big" distro has own customizations Jun 09 22:27:55 and special ruleset for udev Jun 09 22:28:08 kinda, yeah Jun 09 22:28:10 Gentoo for one let the option of a static tarball Jun 09 22:28:30 which is just fine for a server btw Jun 09 22:28:42 heh Jun 09 22:28:53 And embedded devices too :) Jun 09 22:29:08 not sure Jun 09 22:29:13 It's just desktops that are fast enough, but small enough in number of devices, to dynamically make every node and not notice Jun 09 22:29:30 I have to hack for working CF bluetooth Jun 09 22:29:36 We've got "new" and speedy enough HW that takes a good 30sec to generate all dev nodes Jun 09 22:29:39 That's not OK Jun 09 22:29:44 probably an udev rule would do this ... Jun 09 22:29:57 So static start nodes, udev for hotplug Jun 09 22:30:07 I see Jun 09 22:30:27 image compiled, let flash Jun 09 22:31:20 ok Jun 09 22:31:22 Tartarus: hmm, are yout hacks public? Jun 09 22:31:45 XorA|gone, no, but it's not worth pushing either :) Jun 09 22:32:00 Tartarus: but you have cut a good 10s off boot times? Jun 09 22:32:08 (I do bug & post anything that's not application-specific) Jun 09 22:32:39 XorA|gone, yeah, but only with the knowledge of what's going to be there exactly at boot-time Jun 09 22:33:04 Tartarus: I am very interested in that then for certain devices Jun 09 22:34:30 XorA|gone, http://pastebin.com/m71d8ea5e Jun 09 22:35:07 brb Jun 09 22:35:31 * mwester shares that interest; nobody will be hacking to change the core devices on the NSLU2 for example. :) Thanks! Jun 09 22:35:42 Tartarus: so you store a static copy of /dev in /lib/udev/devices/ Jun 09 22:36:40 * XorA|gone thinks a lot of OE devices would benefit from that Jun 09 22:37:21 Simple. Obvious. Easy to change for those who wish to. Jun 09 22:37:22 XorA|gone, yeah Jun 09 22:37:40 * XorA|gone will have to try and integrate that for GTA01 tomorrow Jun 09 22:38:36 I have a sticky note to fix up device handling for SlugOS, so I'll be messing about with that code as well. I'll wait to see what XorA|gone comes up with and model it from that,though. Jun 09 22:39:06 mwester: yeah let me get paid to steal Tarturus code :-) Jun 09 22:39:24 heh Jun 09 22:39:30 hehe! In this industry we call it "re-use" and we get lauded for it! Jun 09 22:40:06 XorA|gone, I'd been thinking the "best" way to do this right might be to have a "here's out tmpfs /dev" and "here's start udev for hotplug" and "we have no dev nodes, fakeplug everything now" Jun 09 22:40:17 since we want tmpfs /dev, we want udev for hotplug Jun 09 22:40:24 we just don't want that horrible startup time Jun 09 22:40:34 hey ant_ Jun 09 22:40:43 soo Jun 09 22:40:53 it fails after "Synyhesizing the.." Jun 09 22:41:16 ant_, how exactly is it failing? Jun 09 22:41:16 mount: mounting proc on /proc failed Jun 09 22:41:28 anyway its late and I must sleep Jun 09 22:41:56 avahi is broken too :( Jun 09 22:41:57 ant_, well, here's the easy way to rule out 3886 changes, or not Jun 09 22:42:06 ant_, copy udev-100/init to udev-118/init, rebuild Jun 09 22:42:19 That'll bring us back to the previous initscript Jun 09 22:42:22 I reverted to 115 Jun 09 22:42:37 and it somehow works (touchscreen ok) Jun 09 22:42:52 but fails for thge dynamic rules Jun 09 22:43:03 ant_, huh? Where do you get 115 from? :) Jun 09 22:43:10 poky Jun 09 22:43:14 heh Jun 09 22:43:32 btw 119 compiles fine Jun 09 22:44:46 ant_, yeah, ok, poky has none of the changes atm Jun 09 22:44:55 ant_, try bringing the current OE.dev initscript into poky Jun 09 22:45:04 Or, er, wait Jun 09 22:45:09 but the busy /proc thing happens with udev_092 too, Jun 09 22:45:12 Are you saying that poky's 115, sans 3886 changes, is busted? Jun 09 22:45:31 Or just the /proc thing, but not the lack of /dev/shm, etc Jun 09 22:45:46 I need to recheck, but I did customize Jun 09 22:46:03 I'd say vanilla no-go Jun 09 22:46:16 but it was late...) Jun 09 22:46:32 heh Jun 09 22:46:59 now, why does it fails mounting /dev/shm ? Jun 09 22:47:52 I'll have to hack inittab to log all this mess ... Jun 09 22:48:16 ok, proc busy Jun 09 22:48:26 Well, I'm wondering if some magic config option is turned off Jun 09 22:48:46 I'd have sworn you couldn't turn off the stuff udev needs to run, tho Jun 09 22:48:46 some odd netlink option Jun 09 22:50:07 root: mount: mounting tmpfs on /dev/shm failed: no such file or directory Jun 09 22:50:14 ^^^^ Jun 09 22:52:14 Tartarus: the paths /dev.udev exists and is populated Jun 09 22:52:37 db names failed uevent_seqnum Jun 09 22:54:10 and /dev/udev_network_queue contains only "irda0" Jun 09 22:54:51 ant_, with poky udev? Jun 09 22:55:11 no, 118 Jun 09 22:55:16 ok Jun 09 22:55:24 Try sticking udev-100/init into udev-118/ Jun 09 22:55:45 ok Jun 09 22:56:59 it i'm simply replacing /etc/init.d/udev on zaurus Jun 09 23:00:16 urgh, the sd card refuses to mount Jun 09 23:00:25 reboot Jun 09 23:03:49 urgh..mount errors, modprobe errors... Jun 09 23:04:04 and /dev/shm not mounted Jun 09 23:04:38 that was with vanilla init from udev_100 Jun 09 23:06:36 OK Jun 09 23:06:43 Something is up in the kernel config, I think... Jun 09 23:06:46 Can you note on the bug? Jun 09 23:08:19 Tartarus: 092 works on same kernel Jun 09 23:08:31 ant_, right Jun 09 23:08:36 and 115 too, with minor quirks Jun 09 23:08:38 udevd[270]add_matching_files: unable to open '/dev/.udev/rules.d': No such file or directory Jun 09 23:08:39 [20080608 01:13:40] *** Jun 09 23:08:39 ant_, but since 092, netlink has become a requirement Jun 09 23:09:01 these are the dynamic rules, isn't? Jun 09 23:09:02 ant_, but if 118 fails with the 110 (and 115) initscripts Jun 09 23:09:30 Tartarus: fails only on c7x0 and tosa, atm Jun 09 23:09:42 no other report Jun 09 23:10:02 (no other testers?) Jun 09 23:10:43 Not sure, heh Jun 09 23:11:22 the bug is outstanding, though Jun 09 23:11:53 Lemme see if I can get an N810 image up sometime tomorrow Jun 09 23:12:00 the init seems more logical... Jun 09 23:12:03 in 118 Jun 09 23:12:25 now wuth udevsettle Jun 09 23:12:32 Yeap :) Jun 09 23:12:54 ok, very late here, need some sleep Jun 09 23:13:00 thank you very much Jun 09 23:13:25 g'night Jun 09 23:58:41 pulseaudio is failing for me Jun 09 23:59:50 http://tinderbox.openembedded.net/packages/31034/ Jun 10 00:29:50 03mickeyl 07org.oe.dev * rd3f40a7a... 10/ (1 packages/bash/bash.inc packages/bash/bash_3.2.bb): bash 3.2 change the way how mkbuiltin is compiled for host Jun 10 00:29:54 03mickeyl 07org.oe.dev * r9baecd01... 10/ (6 files in 3 dirs): bash 3.2 fix compiling, this time for real Jun 10 00:30:00 03mickeyl 07org.oe.dev * r52c7363a... 10/ (1 packages/gstreamer/gst-plugin-pulse_0.9.4.bb): gst-plugin-pulse 0.9.4 remove buggy version, leaving 0.9.7 as the only working one Jun 10 00:30:04 03mickeyl 07org.oe.dev * r6c0c4205... 10/ (1 conf/distro/include/preferred-om-2008-versions.inc): preferred-om-versions: yank removed versions Jun 10 00:30:09 03mickeyl 07org.oe.dev * re57c2aac... 10/ (1 packages/bash/bash.inc): bash don't say a word... please **** ENDING LOGGING AT Tue Jun 10 02:59:56 2008