**** BEGIN LOGGING AT Mon Aug 13 02:59:56 2007 Aug 13 05:30:59 polyonymous, hi, are you around? Aug 13 05:58:34 hi all Aug 13 05:59:04 is there a usable open source os for wiza200 ? HTC - wizrad Aug 13 06:03:15 oops sorry just treda the pae oe is for creating custom packages Aug 13 06:30:23 * Bernardo is back. Aug 13 06:31:07 good morning Aug 13 06:34:19 good morning Bernardo Aug 13 07:05:11 * Bernardo is away: Ausente por agora. Aug 13 07:23:52 good morning all Aug 13 07:24:00 koen : fic arrived ! Many thanks Aug 13 07:46:52 good morning all Aug 13 07:50:25 steliosk: how do you like the screen? Aug 13 07:54:27 hey XorA Aug 13 07:54:40 hey koen Aug 13 07:54:43 morninal all Aug 13 07:55:57 koen : haven't used it yet... need to finish a few things before i get to start playing with it Aug 13 08:05:28 To get a kernel to build for the Xilinx fpgas we need to copy a file from a dir in the Xilinx install path to the kernel. Currently i am handling this in do_configure in the kernel but that means people need to change the kernel .bb file to put the dir where the files are Aug 13 08:05:38 morning all Aug 13 08:05:47 hey RP Aug 13 08:06:39 koen: I think I have the fix for 2601 but I need zecke to confirm it Aug 13 08:06:40 would that make a good candidate for a bbclass and an additional variable to be used in local.conf ? i.e XILINX_EDK_PATH or something similar Aug 13 08:06:49 RP : hi Aug 13 08:07:26 i am kind of reluctant to create more variable etc Aug 13 08:08:11 steliosk: Is Xilinx unpacked by OE? Aug 13 08:08:34 RP : No. Aug 13 08:09:31 When you make a design, the Xilinx environment generates a BSP (board support package) which is saved at the project dir Aug 13 08:10:08 steliosk: Then your only option is to add some kind of path variable for it Aug 13 08:12:10 RP : kernel config looks like that currently _>http://rafb.net/p/sVeBZc40.html Aug 13 08:13:49 RP : basicaly you need to copy xparameters_ml40x.h from the xilinx dir to kernel and rename it to xparameters_ml403.h or whatever board you are using Aug 13 08:16:34 steliosk: I think your only option is something in local.conf... Aug 13 08:22:40 RP: I also figured out with bitbake waits so long between executing two tasks: tinderbox Aug 13 08:27:06 koen: Right, I didn't know it did ;-) Aug 13 08:27:35 it contacts the server at the start and end of each task and sends a log Aug 13 08:53:31 koen: how far along the autoconf 2.61 timetable are we? Aug 13 08:53:48 XorA: www.openembedded.org/ical/oe-events.ics Aug 13 09:01:22 hi koe Aug 13 09:01:24 hi koen Aug 13 09:02:57 koen, i folow your instruction when build x86 image w/ OE http://pastebin.com/m6847c532 Aug 13 09:03:31 koen, but i have error when do_package linux-libc-headers_2.6.18.bb. Here is my error output http://pastebin.com/m67cdd35a Aug 13 09:03:48 koen, could you give me some sugesstion ? Aug 13 09:03:55 ~lart gnome for no working calendar software Aug 13 09:03:56 * ibot shoots gnome in the head for no working calendar software Aug 13 09:04:58 * * OE Bug 2747 has been created by autobuild(AT)openembedded.org Aug 13 09:05:00 * * failme-1.0-r0-do_configure Aug 13 09:05:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2747 Aug 13 09:05:10 * * OE Bug 2739 has been RESOLVED (FIXED) by Aug 13 09:05:12 * *  linux-omap1-2.6.20-omap1-r0-do_deploy Aug 13 09:05:15 * * http://bugs.openembedded.org/show_bug.cgi?id=2739 Aug 13 09:05:22 * * OE Bug 2747 has been RESOLVED (FIXED) by Aug 13 09:05:24 * *  failme-1.0-r0-do_configure Aug 13 09:05:26 * * http://bugs.openembedded.org/show_bug.cgi?id=2747 Aug 13 09:06:27 RP: ping Aug 13 09:09:44 zecke: pong Aug 13 09:10:50 Your patch is interesting but I'm scared to death changing the data.py stuff without checking the result :) Aug 13 09:11:09 and I think your method might be suited for delVar? Aug 13 09:11:53 zecke: I ran my dictonary compare code and it said the change was safe... Aug 13 09:12:04 alterdata._special_values[i].remove(key) should probably happen in delVar (I'm still not sure how this stale data makes a difference) Aug 13 09:12:42 alterdata._special_values[i].add(ekey) should already happen Aug 13 09:12:48 zecke: The problem is if you run update_data after expand_keys, it no longer spots that the renamed variable has appends Aug 13 09:13:24 RP: I want something more formal, even for the future to see the difference the code has :) Aug 13 09:13:59 zecke: add(ekey) does not happen, that is the problem... Aug 13 09:14:02 what about a renameVar for data_smart.py Aug 13 09:14:22 zecke: Yes, thats basically what expandKeys does Aug 13 09:14:45 then let us move it to data_smart.py Aug 13 09:15:24 zecke: I'm not sure I want to make that much of a change in the 1.8 branch Aug 13 09:15:59 zecke: On the other hand, this is a regression compared to 1.6 :/ Aug 13 09:16:05 NOTE: preferred version 2.5 of glibc-intermediate not available Aug 13 09:16:10 what changed over the weekend then? Aug 13 09:16:37 RP: didn't I introduce _special_values for 1.6? Aug 13 09:17:07 zecke: ah, yes Aug 13 09:17:17 so the question, why is it hit now :) Aug 13 09:17:35 zecke: Regression over 1.4 then? :/ Aug 13 09:18:25 zecke: Perhaps RDEPENDS_gtk+_ was replaced with RDEPENDS_${PN}_ ? Aug 13 09:19:51 zecke: http://www.rpsys.net/openzaurus/temp/bitbake_compare_data.patch is my rather evil compare patch (and yes, I know its not perfect) Aug 13 09:20:13 zecke: I played with both expanding the data and not expanding it Aug 13 09:23:18 RP: I can see how _special_values get lost and say good catch :) Aug 13 09:23:35 koen: importing that calendar into evolution gives one appointment that is visible and two invisible ones :-) Aug 13 09:23:52 RP: I don't want data.py to manipulate _special_values though. So I would propose to create a renameVar method and move that to data_smart.py Aug 13 09:24:11 zecke: Mainly luck I found it ;-) Aug 13 09:24:18 zecke: I agree a renameVar would be better Aug 13 09:24:32 RP: and, I would like to have a formal testing/baseline for such stuff, I might feel like doing it tonight Aug 13 09:24:39 RP: I would call it 'force' Aug 13 09:25:08 zecke: I agree we need something to test things like this. Aug 13 09:25:45 zecke: Can we make the test suite functions easily triggerable from bitbake itself, not hidden in other utilities though? Aug 13 09:26:42 RP: like bzr selftest? Aug 13 09:26:45 Hi everybody! Aug 13 09:26:49 XorA: only pkgconfig left Aug 13 09:26:59 XorA: no automake yet Aug 13 09:27:14 zecke: Kind of, yes Aug 13 09:27:33 I have a problem with OE. Can anybody help me? Aug 13 09:29:29 zecke: I added a -P profiling option btw, I got sick of maintaining my out of tree hacks ;-) Aug 13 09:30:23 zecke: The latest bit of insanity removed was 60,000 stat calls resulting in a 14% parsing speed improvement for poky :) Aug 13 09:30:45 koen: its a sad day that the only way I can read a calendar is to get you to read it to me :-) Aug 13 09:30:54 * XorA shoulda loaded it in gpe Aug 13 09:31:24 RP: wow, I profile at the camp site and number one was IIRC setVarFlag called and then the feeder Aug 13 09:32:28 my oldest to do item is to dump SRCDATE's, PREFERRED_PROVIDER and PREFERRED_VERSION of a current run to a file Aug 13 09:32:47 zecke: posixpath:stat was about 20th on the list Aug 13 09:33:02 zecke: Mostly due to SRC_URI_OVERRIDES_PACKAGE_ARCH Aug 13 09:33:10 So? Anyone? Aug 13 09:33:30 v1p3r100: Mention the problem and maybe someone can help Aug 13 09:33:35 03koen 07org.oe.dev * re1b38fd4... 10/ (1 classes/seppuku.bbclass): seppuku: remove extra '?' that confuses bugzilla Aug 13 09:34:13 zecke: seppuku 'works' now in .dev, I'm trying to find out why it doesn't reopen bugs properly Aug 13 09:34:30 hey lrg Aug 13 09:34:39 koen: need to work on my day job now Aug 13 09:35:02 zecke: I was just mentioning the current state Aug 13 09:35:22 ~zap lrg Aug 13 09:35:29 * ibot takes out a cattle prod and gives lrg a good jolt. Aug 13 09:35:29 03koen 07org.oe.dev * rbe168d12... 10/ (1 packages/gsm/libgsmd_svn.bb): gsmd: remove SRCREV Aug 13 09:35:30 hey koen Aug 13 09:35:41 XorA: ouch Aug 13 09:35:49 zecke: setVarFlag would be the next place to look at opitmisation, its the most heavily used function now :/ Aug 13 09:36:24 OK! Well: I try a simple build for DISTRO=generic-uclibc.conf and MACHINE=x86.conf... But when it try build uclibc, it drop an error. Aug 13 09:37:03 This is the error: /stuff/tmp/cross/lib/gcc/i486-linux-uclibc/4.1.1/../../../../i486-linux-uclibc/bin/ld: warning: ld-uClibc.so.0, needed by /stuff/tmp/cross/lib/gcc/i486-linux-uclibc/4.1.1/libc.so, not found (try using -rpath or -rpath-link) Aug 13 09:37:04 | /stuff/tmp/cross/lib/gcc/i486-linux-uclibc/4.1.1/libc.so: undefined reference to `__libc_stack_end' Aug 13 09:37:29 I install all application from OE doc. Aug 13 09:37:46 What is the problem? Aug 13 09:38:22 v13r100 : distro generic most probably. It has not been maintained for sometime now Aug 13 09:38:35 v13r100 : try angstrom Aug 13 09:39:07 * steliosk adds one more thing in the todo list Aug 13 09:39:45 v1p3r100 : try angstrom Aug 13 09:39:48 And must I have install cross toolchain tools, or the OE does it? Aug 13 09:39:58 OK. I try it. Aug 13 09:40:02 OE builds the toolchain Aug 13 09:40:18 I think so... Aug 13 09:59:17 good morning Aug 13 10:06:58 * * OE Bug 2748 has been created by mike(AT)blueroot.co.uk Aug 13 10:07:00 * * librsvg fails to build Aug 13 10:07:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2748 Aug 13 10:10:08 hey florian Aug 13 10:13:36 RP : Is there any easy way to export the vars used in local.conf to bash script ? I am trying to avoid to write the bbclass in python :/ Aug 13 10:22:37 steliosk: export SOMEVAR = "xyz" Aug 13 10:23:02 steliosk: export SOMEVAR might also export a previously set variable Aug 13 10:24:18 RP : well it seems that i was not using the proper variable i.e if [ -z "$XILINX_BOARD" ]; then instaed of the if [ -z "${XILINX_BOARD}" ]; then Aug 13 10:24:45 if i use it like this it get access to all the vars in OE Aug 13 10:25:05 no need to export anything Aug 13 10:25:11 steliosk: ok Aug 13 10:25:28 steliosk: If you exported it, $XILINX_BOARD would also work Aug 13 10:26:20 RP : yes it would. i just need to stay close to the OE way of doing things... i.e add a variable to local.conf and we take care of the rest :) Aug 13 10:41:41 ~seen do13 Aug 13 10:42:12 do13 was last seen on IRC in channel #oe, 19d 1h 39m 38s ago, saying: 'hi Richard'. Aug 13 10:44:50 eeks! Aug 13 10:44:58 * koen curses neo startup sound Aug 13 10:48:19 koen: scary shit isnt it :-) Aug 13 10:48:25 indeed Aug 13 10:50:32 heh Aug 13 10:50:42 the vibrator doesn't stop after a call Aug 13 10:56:22 the kernel has a config option for the TRANCEVIBRATOR Aug 13 11:02:35 err... someone plz can help-me with OE compilation? Aug 13 11:03:22 ive tested http://www.palmlinux.cz/modules.php?name=Downloads&d_op=viewdownload&cid=7, but bluetooth locks when i try discover Aug 13 11:03:55 so ive tried to build from mtn, but i cannot create de initrd and the opie image Aug 13 11:04:24 ive stoped here: source source-me ; pwd ; cd org.openembedded.dev/ ; mtn pull ; mtn update ; bitbake opie-kdepim-image ; bitbake angstrom-minimal-image ; bitbake angstrom-console-image ; bitbake angstrom-x11-image Aug 13 11:22:02 bother, the guy I know with the evm is working at home today .. Aug 13 11:31:26 heh Aug 13 11:31:37 I just noticed the evm is like $2500 Aug 13 11:31:41 yeah Aug 13 11:31:45 they are not cheap Aug 13 11:32:10 Back when Rishi would asnwer email, he thought he could get a stripped one without camera and some other stuff Aug 13 11:32:32 he's probably on the road or something Aug 13 11:32:36 yeah Aug 13 11:32:48 He's pretty bad about replying to email Aug 13 11:44:56 03koen 07org.oe.dev * r83c27aa0... 10/ (1 packages/hal/ohm_git.bb): ohm: add from poky Aug 13 11:52:17 koen: you have a avr32 working in build system ? Aug 13 12:33:30 ~lart koen Aug 13 12:33:30 * ibot cuts off koen's head with a halberd that could have been a little bit sharper Aug 13 12:35:00 oh yeah, scox down 52% in pre-market trading Aug 13 12:36:04 * koen wonders what happened now Aug 13 12:36:23 koen: your ${NEATSTUFF} causes glibc 2.3.2 to be built Aug 13 12:36:34 koen: or not built as the case is Aug 13 12:37:00 aha Aug 13 12:37:15 I suspect and RPROVIDES/PACKAGES_DYNAMIC problem Aug 13 12:38:56 koen: was my suspicion Aug 13 12:39:16 koen: as all glibc have same PACKAGES_DYNAMIC set by glibc-package.bbclass Aug 13 12:40:54 COMPATIBLE_HOST = ".*linux" ? Aug 13 12:41:04 so no EABI stuff chooses it? Aug 13 12:42:52 that way leads to bad meta data, as when glibc bumpts to 2.8 but 2.5 is still a valid version, we will still select the wrong one Aug 13 12:43:39 need a REALLY_ONLY_EVER_USE_glibc="2.5" Aug 13 12:44:28 anyway I was talking to RP about the problem on another chan, give him chance to get back from lunch Aug 13 12:44:46 yeah Aug 13 12:54:08 Do'h Aug 13 12:54:18 lrg: I spot a small problem... Aug 13 12:57:19 koen: :( Aug 13 12:57:40 koen: I can mail Aug 13 12:58:38 that seems to be the best option for now Aug 13 12:59:20 koen: ok, will do, although Remco will probably be around looking you. Best to hide ;) Aug 13 12:59:33 XorA: I'm back now Aug 13 12:59:47 I'll be working at a music festival when he's back in town :) Aug 13 13:00:10 RP: cool, it was koen's ${NEATSTUFF} hack in gtk+-2.10.inc that triggers the problem Aug 13 13:00:22 I'll hide behind the nine inch nails and basement jaxx Aug 13 13:00:23 XorA: How does gtk end up influencing glibc versions? Aug 13 13:00:33 RP: its the gconv-* RRECOMMENDS that pulls in glibc 2.3.2 I think Aug 13 13:00:47 RP: as all glibc have PACKAGES_DYNAMIC = "gconv*" Aug 13 13:00:58 XorA: Not all glibc's provide that? Aug 13 13:01:29 RP: they all provide it, they are not all buildable in Angstrom Aug 13 13:01:47 XorA: but it should pick glibc-2.5... Aug 13 13:02:19 I did 'mtn mv glibc*2.3* *2.2* obsolete/foo' eons ago Aug 13 13:02:26 that's why I didn't see that Aug 13 13:03:04 Well, one problem is that glibc-intermediate doesn't provide those things so their PACKAGES_DYNAMIC is wrong Aug 13 13:03:25 but it appears bitbakes dependency logic isn't working quite as well as I'd like :/ Aug 13 13:04:10 RP: intermediates dont use glibc-package.bbclass Aug 13 13:04:54 RP: I meant all final on device glibc, I forgot about intermediates Aug 13 13:05:30 PACKAGES = "" Aug 13 13:05:30 PACKAGES_DYNAMIC = "" Aug 13 13:05:34 is in the intermediates Aug 13 13:12:16 XorA: So why was it building a glibc intermediate? Aug 13 13:14:48 RP: I see the problem, 2.3.2 is missing PACKAGES_DYNAMIC = "" Aug 13 13:14:54 RP: gimme a sec to test Aug 13 13:15:48 RP: I am getting confused with all these damn package with similar names and versions Aug 13 13:16:23 XorA: np, if that version was missing the = "", that would explain it Aug 13 13:17:07 It still suggests bitbake's dynamic package handling needs work but not as badly as it would have if that hadn't been set... Aug 13 13:17:39 I think I know why it might work the first time and not the second after its built things :/ Aug 13 13:18:33 RP: I think that change works, I shall check it in Aug 13 13:19:01 XorA: great. The underlying bitbake problem needs fixing too though :/ Aug 13 13:19:37 Ill have to do a from fresh build again to test it properly Aug 13 13:30:50 03xora 07org.oe.dev * r188bf8cf... 10/ (1 packages/glibc/glibc-intermediate_2.3.2+cvs20040726.bb): Aug 13 13:30:50 glibc-intermediate_2.3.2+cvs20040726.bb : PACKAGES_DYNAMIC = "" because Aug 13 13:30:50 intermediate doesnt supply things like gconv. Aug 13 13:42:53 RP: another ping :) Aug 13 13:48:56 zecke: pong :) Aug 13 13:49:18 Crofton|home : Just got an email from Grant Likely that maintains (among others) the efika git tree we use Aug 13 13:49:51 Crofton|home : Said that we should use Sylvain's tree as its more up to date Aug 13 13:50:09 RP: the braunschweig mafia, had some issues with the rm_work on bitbake1.8. And admitingly I have not really much clue about your work. But it seems like rm_work is only in the depchain of the 'main' task Aug 13 13:50:18 Crofton|home : Are you using the git kernel or the standard+patches ? Aug 13 13:50:25 this reminds me, I need to work out external toosl cahin Aug 13 13:50:30 standard + patches Aug 13 13:50:58 RP: so bitbake foo, will call rm_work on foo but not on the other items. We all know rm_work is a bastard child and it has issues to get invoked at the right spot Aug 13 13:51:25 RP: I would like to introduce finalizer (destructor) which is called when everything is done Aug 13 13:51:26 Crofton|home : I contacted him about the Xilinx tree and told me that he is moving the server so there are going to be problems Aug 13 13:51:35 ah Aug 13 13:51:49 RP: e.g. if you have two finalizer, the first wins :) Aug 13 13:51:55 there is a xinlinx tree at source.mvista.com/git Aug 13 13:52:10 Crofton|home : yep. i am making 2 kernels ;) Aug 13 13:52:17 heh Aug 13 13:52:33 xilinx-ml403_2.6.x.git Aug 13 13:52:39 for the mvista one Aug 13 13:52:44 I sure hope no one owned scox .... Aug 13 13:52:47 cool Aug 13 13:52:55 and xilinx-ml403_2.6.20+git for Grands Aug 13 13:53:01 zecke: Some people aren't going to always want rm_work to run, they might only want it to run when the task is built Aug 13 13:53:05 (successfully) Aug 13 13:53:16 cbrake, ping Aug 13 13:53:49 RP: well, introduce two different finalizer's :) Aug 13 13:54:10 RP: and if you have two finalizer's that do competing operations (e.g. copying vs. deleting) it is a user fault Aug 13 13:54:23 zecke: I know there are issues with rm_work but its more to do with bitbake's task scheduler. It see no urgency in runnning the rm_work tasks. Adding finalisers won't help that Aug 13 13:54:51 RP: they are currently not called at all, besides for the main package you build Aug 13 13:55:07 RP: bitbake foo, fill call foo:rm_work but not bar:rm_work while foo DEPENDS on bar Aug 13 13:55:25 RP: and the way rm_work hooks into the task tree, is just stupid and needs to be cnahed Aug 13 13:56:10 zecke: then the rm_work task dependencies need fixing... Aug 13 13:56:39 RP: right, fixing as in a proper way to be called at the END Aug 13 13:56:48 RP: instead of handwaving when something could be done :) Aug 13 13:58:05 actually, it is just an event handler :) Aug 13 13:58:43 Turning rm_work into an event handler might be a better way to do things Aug 13 13:59:14 will do that now Aug 13 14:00:19 Crofton|home: hello Aug 13 14:00:42 someone mentioned you had a write up on using external tool chains with OE Aug 13 14:01:06 zecke: How are you going to tell when a given task is ready to be cleaned? Aug 13 14:01:48 RP: using task suceeded/taks failed? Aug 13 14:02:01 Crofton|home: yes, I've done it once: http://bec-systems.com/web/content/view/56/9/ Aug 13 14:02:18 Crofton|home: I think there is something in OE users manual as well Aug 13 14:02:22 thanks Aug 13 14:04:54 zecke: I think I foresee problems :/ Aug 13 14:05:19 RP: well, then this is a bug of bitbake :) Aug 13 14:05:33 RP: it is supposed to tell when it is done, and this used to work Aug 13 14:06:02 zecke: It will run each task individually in bitbake 1.8 Aug 13 14:06:20 zecke: I made the multithreading granular down to the tasks Aug 13 14:06:25 I know Aug 13 14:07:16 You will get a succeeded/fail message but you won't know which is the last one Aug 13 14:07:37 then we are back at the finalizer :) Aug 13 14:08:31 A finalizer event would make more sense than marking tasks... Aug 13 14:08:55 RP,zecke : What about a list of finished tasks ? Aug 13 14:09:27 steliosk: why do you want that? I could understand a list of tasks :) Aug 13 14:10:12 zecke : To see which taks rm_work should work on Aug 13 14:10:55 steliosk: in my mind rm_work is just executing the clean task Aug 13 14:11:35 zecke: The ugly problem is when you consider that "bitbake abc" wants xyz's package_write function to run but "bitbake bdc" wants xyz's populate_staging function to run Aug 13 14:11:55 zecke: If you clean it in the meantime, you have to build most of xyz twice Aug 13 14:12:55 zecke: So rm_work needs to ensure all build functions run, then clean. This then becomes a dependency problem. Aug 13 14:12:59 RP : that's why i said for a list of finished task and then "let" rm_work do its work Aug 13 14:13:02 RP: aeh? why is that? (technically this is due task deps, but why?) Aug 13 14:13:17 zecke: Why is what specifically? Aug 13 14:13:52 zecke: When is a task considered ready to be cleaned? Aug 13 14:14:01 In which case do we depend in package_write and in which case in populate_staging and why can't both depend on the goodl old do_build? Aug 13 14:14:40 zecke: bitbake doesn't run do_build for every task now unless something sets that as a dependency Aug 13 14:14:49 zecke: It might be such a dependency is correct for rm_work Aug 13 14:16:08 I go as far as saying do_build makes sense, and is not an issue/obstacle to paralelisation Aug 13 14:16:20 As I said a long time ago, the best thing is to improve the syntax so we can "addtask rm_work before build after populate_staging deploy package_write rootfs Aug 13 14:16:42 We need to make addtask take multiple tasknames Aug 13 14:16:51 that is still error prone and keeps braking Aug 13 14:17:09 zecke: Why? Aug 13 14:17:24 it did break with bitbake1.8.6 or earlier Aug 13 14:17:37 and just adjusting the task deps is not the right(tm) solution Aug 13 14:17:40 zecke: I guess we need to set something to ensure bitbake always runs do_build for everything but there are dependency options that can do that Aug 13 14:18:34 so we don't use the BB_DEFAULT_TASK anymore? or where is that package_write and populate_staging dep coming from? Aug 13 14:18:57 zecke: We do, when you run bitbake foo, it runs the do_build task of foo Aug 13 14:19:07 oh, I get interrupted... I think i have to quit my job Aug 13 14:19:37 From there we have a dependency tree and BB_DEFAULT_TASK is ignored... Aug 13 14:19:49 ~lart failed microSD card Aug 13 14:19:50 * ibot says "boot to the head" and knocks failed microSD card over Aug 13 14:20:24 summary from my point of view: BitBake could know how many tasks for a given bb file are going to be executed and could send a bbfile-completed event Aug 13 14:20:37 once all of these tasks are executed Aug 13 14:20:41 zecke: Send it to where? Aug 13 14:20:48 default event handler Aug 13 14:20:55 we have that system in place... Aug 13 14:21:03 zecke: so you'd just wipe work in total? Aug 13 14:21:53 oh okay, evicting the stamps might not be the most sane idea :) Aug 13 14:21:57 Is there a "--pretend" switch for bitbake -- so that I can just see what it *would* do before I do it? Aug 13 14:22:10 mwester: bitbake --help Aug 13 14:22:29 zecke: You just need to understand the way the task handling works now :/ Aug 13 14:22:52 zecke: do_build isn't run for every task now, it only runs as much of each task as it needs Aug 13 14:23:11 RP: I don't, I know how evicting of the workdir should work Aug 13 14:23:36 RP: yes, see the above sentences, I do not mention a task name at all Aug 13 14:23:36 zecke: ok, you don't need me then ;-) Aug 13 14:24:06 yeah, I plan to do it myself :) Aug 13 14:24:49 zecke: Just wipe tmp/work on the BuildCompleted event Aug 13 14:25:05 lol Aug 13 14:25:13 heh Aug 13 14:25:20 I thought you mentioned buildcompleted is called too early? Aug 13 14:25:29 .... :) Aug 13 14:25:45 zecke: No, your turn to read back now, you were talking about tasksucceed/fail Aug 13 14:25:59 and yes calling 'clean' is bad as it will remove the stamps Aug 13 14:26:18 noone can help in my trouble with zire72? Aug 13 14:26:39 RP: succeed/fail is due the asymmetry of not having a build-failed signal, but only task-failed and I always confuse names :) Aug 13 14:26:59 zecke: bitbake foo -c rmall Aug 13 14:28:07 zecke: I'm am adament that a finialiser is not needed and will not work without fixing the dependency tree anyway as you can't determine an end point without doing that. An end point is something that rm_work will have to define, not bitbake Aug 13 14:28:51 koen : grand just told me that the vga patches for the xilinx ip are in his git tree, so i think we are going to have one more board for angstrom :) Aug 13 14:29:16 steliosk: what has vga to do with angstrom? Aug 13 14:29:31 RP: we chase each other around misunderstanding. Finalizer would be the time when all tasks are finished, which used to be "BuildCompleted" Aug 13 14:30:26 get x running also Aug 13 14:30:52 zecke: Right, we are misunderstanding. By all means use the BuildCompleted signal but I'd create a different class to rm_work for that as its a bit of a different way of working Aug 13 14:41:16 zecke: btw, I did add a ConfigParsed event for sanity.bbclass to hook into Aug 13 14:41:27 I have seen that :) Aug 13 14:41:28 awesome Aug 13 14:49:48 well, can someone tellme where can i find help for my trouble/ Aug 13 14:51:59 morning Aug 13 14:59:52 RP: argh, I want PkgSucceeded Aug 13 15:05:55 zecke: We're probably back to the problems I previously mentioned then :/ Aug 13 15:09:55 hi crofton Aug 13 15:11:18 RP: well, firing PkgSucceeded on task success is IMHO a bug :) Aug 13 15:12:58 * * OE Bug 2507 has been marked as DUPLICATE of bug 2671 by waldecir(AT)gmail.com Aug 13 15:13:00 * *  building for angstrom-2007.1 tries to build glibc 2.3.2 Aug 13 15:13:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2507 Aug 13 15:13:51 khem, we should be good to go on adding omap2 kernel Aug 13 15:14:06 I made images for the davinci EVM yesterday, but no EVM to test on Aug 13 15:15:28 zecke: It raises the question of when we should fire it though? Aug 13 15:16:11 zecke: do_build? do_populate_staging? This is what I meant before by something in OE having to declare an "end" point Aug 13 15:17:26 RP: like I think I said. We know how many tasks we are going to execute in this run of BitBake, when all of these tasks are done fire it Aug 13 15:17:38 RP: Now I don't know the implementation and issues with synchronisation Aug 13 15:19:08 zecke: When you say "tasks" do you mean all tasks or just tasks for a given package? Aug 13 15:19:18 for a given package Aug 13 15:20:20 Crofton: I have done that locally Aug 13 15:20:40 omap2? Aug 13 15:20:46 What board? Aug 13 15:20:47 Crofton: yes Aug 13 15:20:53 2430 Aug 13 15:21:12 is there a specific eval board? Aug 13 15:21:20 but 2420 is also an omap2 Aug 13 15:21:22 zecke: Should the task fire if a different task fails somewhere? Aug 13 15:21:39 yes omap2430sdp Aug 13 15:22:23 RP: Well, TaskFailed => BuildFailed (some how we don#t call PkgFailed) Aug 13 15:22:25 I m getting problems in bitbake Aug 13 15:23:01 http://nishanthmenon.blogspot.com/ Aug 13 15:23:36 zecke: The notion of "Pkg" is certainly not very good in the current bitbake. I'm still not sure its worth fixing either :/ Aug 13 15:23:37 http://pastebin.ca/655649 Aug 13 15:23:45 hi khem Aug 13 15:24:00 RP: where is Pkg different from task now? Aug 13 15:24:17 zecke: Its not Aug 13 15:24:17 khem, who sells the sdp? Aug 13 15:24:39 zecke: I agree there is a problem there Aug 13 15:24:40 RP: so it is either remove or fix. And as I need it :) Aug 13 15:25:05 Crofton: Must be TI I am not sure Aug 13 15:25:12 khem that error makes no sense to me Aug 13 15:25:22 RP: Simple approach: You build the rundeps. For each package/bb-file we remember how many tasks are going to be executed Aug 13 15:25:32 bitbake guys, can you make sense of khems pastebin? Aug 13 15:25:36 Crofton: It started to happen when I added a new bb file for uboot Aug 13 15:25:38 RP: for each task that got executed of a bbfile we decrement the number and fire an event? Aug 13 15:25:56 RP: now do you think that decrementing work, or do we have synchronisation issues now? Aug 13 15:26:01 but what amazes me now that the error is persistent even when I reverted the changes Aug 13 15:26:07 this has happened to me second time Aug 13 15:26:16 zecke: Its probably even simpler, just check is any tasks for a given file still exist in the queue, if not, fire the event Aug 13 15:26:34 Now I know I have to get a fresh checkout to get it fixed Aug 13 15:26:37 khem, can you use existing uboot-git file? Aug 13 15:26:47 that is what I use for davinci-dvevm Aug 13 15:26:51 Crofton: It does not have 2430 support :( Aug 13 15:26:57 hmmm Aug 13 15:27:04 zecke: I still maintain the approach is broken though as per my do_populate_staging vs do_package_write comments earlier Aug 13 15:27:05 git + patch? Aug 13 15:27:15 I mean u-boot upstream Aug 13 15:27:25 zecke: Well, adding the events isn't but using them the way you're going to is Aug 13 15:27:32 Ti has not been good about getting stuff in to uboot Aug 13 15:27:53 Some other guys finally got davinci support in Aug 13 15:27:56 khem: I'd guess a malformed PREFERRED_PROVIDER somewhere Aug 13 15:27:59 Crofton: I thought of that but the uboot I have to use it from TI website Aug 13 15:28:12 RP: hmmm lemme check Aug 13 15:28:28 khem look at the blog link I just posted Aug 13 15:29:29 from Nishant ? yes I did thanks it mentions the patch list Aug 13 15:29:57 so I will figure out what patches we need and if they can be applied cleanly Aug 13 15:30:09 RP: well, I would say there should only be one end Aug 13 15:30:42 zecke: but the end depends on the package you're building Aug 13 15:32:05 wait a second, work... Aug 13 15:33:19 Crofton: Now if I create a linux-omap2 kernel that means it should run on both omap2420 and omap2430 Aug 13 15:33:57 Crofton: It would need kind of hack in bb file to select a proper defconfig Aug 13 15:34:16 yes Aug 13 15:34:22 see linux-davinci Aug 13 15:34:23 Crofton: It will conflict with exisiting eabi and non eabi snafu Aug 13 15:34:29 dir Aug 13 15:34:46 RP: I see that we have different ends now. I wonder about the difference btw staging and package_write and why it scales better than do_build? Aug 13 15:34:46 I am thinking of just supporting eabi ... Aug 13 15:35:00 that makes sense Aug 13 15:35:23 need to talk to the local gurus, but it would make things easier Aug 13 15:35:28 Crofton: oabi is thing of past and atleast on dev tree it would be ok to remove it Aug 13 15:35:31 there are ways to edit the kconfig Aug 13 15:35:35 ok Aug 13 15:35:47 yeah, and I do not think there are any OABI users for OMAP Aug 13 15:35:55 zecke: Its was more a generic task based approach I was interested in rather than the specifics. If the dependency tree is good and working, we don't need do_build at the end of it Aug 13 15:35:56 not any more Aug 13 15:36:10 there were some years back when EABI was not there Aug 13 15:36:23 zecke: Perhaps the best approach is going to be BB_DEFAULT_TASK = "rm_work" combined with do_rm_work[recrdeptask] = "do_build" and then addtask rm_work after do_build ? Aug 13 15:37:04 I think for davinci I just made two copies of the same config so the fetch would work :) Aug 13 15:37:48 zecke: Some games with immediate expansion could remove the references to do_build too... Aug 13 15:38:22 03koen 07org.oe.dev * rff97b52d... 10/ (3 files in 2 dirs): bluez: update to 3.14 Aug 13 15:38:31 Crofton: Locally I did same for 2430 :) Aug 13 15:38:41 khem, I'm going for lunch, you should review how I did the davinci-dvevm addtions Aug 13 15:38:45 Crofton: But my oabi defconfig was empty Aug 13 15:38:49 heh Aug 13 15:39:07 ok I will have to drive to office now Aug 13 15:39:08 RP: what was 'rec' again? And the initial problem is that when building as dep vs. directly package_write/populate_staging might not be called and that another build would rebuild from scratch to 'just' execute this task and fail due an empty directory? Aug 13 15:39:10 * leleobhz ask if someone can help-me with OE compilation fro palm Aug 13 15:39:10 l8r Aug 13 15:39:16 bbiah Aug 13 15:39:35 zecke: Thats where the rec comes in - recursive. I think I even added that to the manual :) Aug 13 15:41:09 zecke: That will require every do_build task to run for every package in the chain. It should probably be do_rm_work[recrdeptask] = "do_rm_work" actually... Aug 13 15:42:53 RP: I still need some light on package_write and populate_staging. When building a package as dep we depend on the package_write task, but it will stage anyway so populate_staging is actually called Aug 13 15:43:14 for some -native packages we depend on populate_staging Aug 13 15:43:49 zecke: Yes, populate_staging currently depends on package_write Aug 13 15:44:31 not the other way around? Aug 13 15:44:52 zecke: No, koen changed it for packaged staging Aug 13 15:45:00 zecke: Some tasks just depend on do_package for example Aug 13 15:45:31 An RDEPENDS with debian.bbclass just adds a do_package depends, not a package_write or staging Aug 13 15:45:38 let me take a look at the graph, if staging -> write and DEPENDS on foo -> abc:write Aug 13 15:45:46 then staging would not be called, but it is :) Aug 13 15:46:17 03koen 07org.oe.dev * rd0268fbb... 10/ (1 packages/hal/ohm_git.bb): Aug 13 15:46:17 ohm: fix depends and packaging Aug 13 15:46:17 * x11 dependants are in ohm-plugin-x11, leaving a small and contained ohm and libohm package Aug 13 15:46:50 RP: I suspect ross would like to have those ohm fixes in poky as well :) Aug 13 15:47:10 koen: ok, thanks :) Aug 13 15:47:25 zecke: Feel free to take a look :) Aug 13 15:47:56 RP: looks sane :) Aug 13 16:00:56 back Aug 13 16:01:04 haha now bitbake is downloading the same file twice :} Aug 13 16:01:47 RP: I'm going to change the time when PkgSuccess is called but we had agreement here Aug 13 16:02:58 RP: for determining the time of cleaning. We might clean too early and subsequent runs of bitbake could fail due an empty working dir Aug 13 16:13:58 * * OE Bug 2749 has been created by autobuild(AT)openembedded.org Aug 13 16:14:00 * * patchutils-0.2.30-r0-do_configure Aug 13 16:14:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2749 Aug 13 16:15:30 hmmmm monotone 0.36 Aug 13 16:17:51 koen: what was the fix? Aug 13 16:18:13 * Bernardo is away: Ausente por agora. Aug 13 16:21:26 zecke: which fix? Aug 13 16:22:24 koen: seppuku Aug 13 16:23:17 zecke: reopning closed bugs still doesn't work, but creating bugs has worked since I fixed the regexp yesterday :) Aug 13 16:37:51 anyone know why --with-headers isn't specified for gcc-cross? Aug 13 16:43:16 Hello all Aug 13 16:43:44 hi likewise Aug 13 16:43:58 * * OE Bug 2750 has been created by autobuild(AT)openembedded.org Aug 13 16:44:00 * * qt4-x11-free-4.3.0-r5-do_populate_staging Aug 13 16:44:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2750 Aug 13 16:44:13 hey likewise Aug 13 16:44:28 hehe, qt4 uses cp -a for copying... Aug 13 16:45:29 hey Aug 13 16:46:01 perl-5.8.8 is fighting with own TOPDIR variable in the Cross/Makefile, and my TOPDIR seems to win (which breaks its build). Known issue to someone? Aug 13 16:48:08 hey likewise Aug 13 16:48:15 likewise: didn't you have a davinci evm? Aug 13 16:48:42 koen: yes, on the shelves (shame on me) Aug 13 16:48:58 likewise: Crofton has built an image for it Aug 13 16:50:41 koen: does Crofton have the h/w? Aug 13 16:50:50 he doesn't :) Aug 13 16:51:31 I'd be happy to test when I come bacl, but that will only be after the 25th of august :P Aug 13 16:51:40 s/bacl/back/ Aug 13 16:52:03 03koen 07org.oe.dev * r2c8d4988... 10/ (1 packages/patchutils/patchutils_0.2.31.bb): patchutils: update to 0.2.31 Aug 13 16:53:18 Crofton: ping Aug 13 16:53:22 pong Aug 13 16:54:18 Crofton: I pushed a small update for the evm that will prevent troubles in the future Aug 13 16:54:28 ok Aug 13 16:54:46 s:_:-:g Aug 13 16:55:06 How can I reproduce the davinci build here? I can get the dust of the board and test somewhere next week Aug 13 16:55:21 set machine to davinci-devevm Aug 13 16:55:22 MACHINE=davinci-dvevm Aug 13 16:55:33 bitbake angstrom-minimal-image :) Aug 13 16:58:28 im can install .pc file in stagin puting pkgconfig in inherit parameter, but how i can make bitbake install include headers ins staging ? Aug 13 16:58:50 s/ins /in Aug 13 16:59:57 likewise: just curious, any news on the wis board, or is this an abandoned? Aug 13 17:00:38 Jin^eLD: NDA had to be refreshed, I'm waiting for a new NDA... Aug 13 17:01:04 Jin^eLD: you done any work on it yet? Aug 13 17:01:35 likewise: yes, we did a demonstrator on it Aug 13 17:02:36 a upnp mediarenderer with own UI engine Aug 13 17:03:04 03koen 07org.oe.dev * r5dae7b1d... 10/ (3 files in 2 dirs): davinci-evm: rename it so overrides work properly Aug 13 17:03:08 03koen 07org.oe.dev * r5bc92477... 10/ (7 files in 4 dirs): linux-davince: catch up with machine rename Aug 13 17:04:37 koen, this will break u-boot build .... Aug 13 17:05:03 machine in u-boot is davinci-dvevm Aug 13 17:05:12 bbl Aug 13 17:06:48 Crofton: mtn pull :) Aug 13 17:08:07 03koen 07org.oe.dev * r397efbec... 10/ (1 conf/machine/davinci-dvevm.conf): davince-dvevm: set UBOOTMACHINE Aug 13 17:10:30 Jin^eLD: using OE or the Linux SDK or anything else? Aug 13 17:10:58 likewise: I converted the SDK shipped with the unit to OE Aug 13 17:11:43 likewise: but they asked me to wait till they finish hacking their drivers and untill they officially release etc. Aug 13 17:12:17 then I could follow up with my conversion Aug 13 17:12:32 brb Aug 13 17:12:36 k Aug 13 17:12:59 * * OE Bug 2751 has been created by autobuild(AT)openembedded.org Aug 13 17:13:00 * * patchutils-0.2.31-r0-do_configure Aug 13 17:13:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2751 Aug 13 17:13:10 * * OE Bug 2751 has been RESOLVED (FIXED) by Aug 13 17:13:12 * *  patchutils-0.2.31-r0-do_configure Aug 13 17:13:14 * * http://bugs.openembedded.org/show_bug.cgi?id=2751 Aug 13 17:13:22 * * OE Bug 2749 has been RESOLVED (FIXED) by Aug 13 17:13:24 * *  patchutils-0.2.30-r0-do_configure Aug 13 17:13:26 * * http://bugs.openembedded.org/show_bug.cgi?id=2749 Aug 13 17:14:19 anyone has an idea why this is happening ? -> http://rafb.net/p/o0s8mg87.html Aug 13 17:15:29 line 9, does the file exist? Aug 13 17:16:04 Crofton : yes it does Aug 13 17:16:27 you have a linux-xilinx dir, or you have it in linux/files Aug 13 17:16:55 http://rafb.net/p/XZr8KA28.html Aug 13 17:17:42 but if looks in files Aug 13 17:17:52 it tries to find it in "files" but not in the dir of the kernel Aug 13 17:17:52 maybe you need the data on the end of the dir name? Aug 13 17:18:28 hmm Aug 13 17:18:30 maybe not Aug 13 17:18:34 I copied the kernel config from omap1-2.6.18+git Aug 13 17:18:59 yeah, it works for me Aug 13 17:19:09 look at 2.6.x+git Aug 13 17:19:18 linux-davinci Aug 13 17:19:20 this is what i looked Aug 13 17:19:23 that built for me yesterday Aug 13 17:23:04 steliosk: in what dir should it look? Aug 13 17:23:05 re Aug 13 17:23:23 linux-xilinx-2.6.20+git Aug 13 17:23:47 OE will look in PN, PN-PV and files Aug 13 17:23:54 not in PV-smoethingelse Aug 13 17:24:04 or PN-somethingelse Aug 13 17:24:42 koen did you rename the EVM machine file to davinci-evm? Aug 13 17:25:07 * koen waits for someome to realize that PV=2.6.x+git can never be upgraded using packagemanagement Aug 13 17:25:09 Crofton: yes Aug 13 17:25:24 hmm, why don't I see that Aug 13 17:25:34 Crofton: underscores in machine names don't work as you might expect Aug 13 17:26:28 ok Aug 13 17:26:40 The comment said davinci-evm Aug 13 17:26:49 not davinci-dvevm :) Aug 13 17:26:55 that is what was confusing me Aug 13 17:27:35 I don't care if the git kernels are upgradable easily Aug 13 17:27:45 adding a version confused things Aug 13 17:28:10 confused what? Aug 13 17:28:11 the git recipes exist for emergency use Aug 13 17:28:20 well 2.6.18+git Aug 13 17:28:42 being used when kernel is now 2.6.22+git Aug 13 17:29:19 that why you keep updating such recipes, or lock them down to a specific revision Aug 13 17:29:22 now, if we need to "release" something, we will need to use a certain git rev Aug 13 17:29:36 I agree with locking Aug 13 17:29:45 Crofton: mtn log packages/cairo/cairo_git.bb Aug 13 17:29:53 I just hate worrying about deleting the old git recipe Aug 13 17:34:31 03florian 07org.oe.dev * r09957ec1... 10/ (1 conf/distro/include/preferred-gpe-versions-2.8.inc): preferred-gpe-versions-2.8.inc: Use latest libmimedir release. Aug 13 17:34:36 03florian 07org.oe.dev * r89f29ea4... 10/ (1 packages/tinymail/tinymail_svn.bb): tinymail: License is LGPL, not GPL Aug 13 17:34:40 03florian 07org.oe.dev * ra3cc50e8... 10/ (1 packages/gpephone/libsettings_0.3.bb): libsettings: Add 0.3 Aug 13 17:34:46 03florian 07org.oe.dev * r73007e84... 10/ (4 files in 3 dirs): sqlite3: Add 3.4.1 Aug 13 17:34:51 koen : The only way to update git kernels not pinned down is to increase the PR everytime there is a new build Aug 13 17:34:51 03koen 07org.oe.dev * r701525eb... 10/ (4 files in 2 dirs): linux-davinci: fix PV so packages can be upgraded at runtime Aug 13 17:35:03 one topic for oedem should be the kernel summit :) Aug 13 17:35:46 basically, we can never release a non-upgradable kernel into a feed, right? Aug 13 17:35:55 steliosk: no it isn't Aug 13 17:36:04 what's the latest consensus on dealing with time ? Aug 13 17:36:06 steliosk: that's what SRCDATE and SRCREV are for Aug 13 17:36:19 time waits for no man! Aug 13 17:36:26 mickeyl: GMT + pool.ntp.org Aug 13 17:36:31 re Aug 13 17:36:36 koen: ok, but what about initial time? Aug 13 17:36:37 woglinde: wb Aug 13 17:36:45 hi mickeyl Aug 13 17:36:46 1970 is soo last century Aug 13 17:36:51 hey woglinde Aug 13 17:36:57 * Crofton was born b4 1970 Aug 13 17:37:11 mickeyl: I'd say default to build-time Aug 13 17:37:32 koen: i say that as well. and IIRC I did that once, but apparantly it has been ripped out Aug 13 17:37:55 iirc there was a bug in your appoach Aug 13 17:38:33 on unclean shutdowns on a system with an RTC strange things happened Aug 13 17:39:08 hello how are kernel recipes named like linux-omap1_2.6.22-omap1.bb why does omap1 appears twice ? Aug 13 17:39:21 * khem is trying to name recipe for omap2430sdp Aug 13 17:39:30 khem: to have 'omap1' in the version (dunno why) Aug 13 17:40:04 * Crofton tried to remember who told him to do that .. Aug 13 17:40:07 zecke? Aug 13 17:40:08 ok I am going to build from git HEAD so I am calling it linux-omap2_2.6.x+git.bb Aug 13 17:40:15 or hrw Aug 13 17:40:16 is that ok ? Aug 13 17:40:31 khem: linux-omap2_git.bb Aug 13 17:40:45 mickeyl: I don't think it was removed, it just doesn't seem to work anymore :/ Aug 13 17:40:46 ok I like that Aug 13 17:40:48 and then PV = "2.6.+git${SRDATE}" inside the bb Aug 13 17:40:53 * Crofton agrees Aug 13 17:41:06 SRCDATE? Aug 13 17:41:13 koen : if you do that then you get the problem i had Aug 13 17:41:41 what is the directory for files in that case? Aug 13 17:41:56 RP: oh. right, grep timestamp still shows some files. I'm going to take a look at it, seeing if i can fix it and being nice wrt RTC as well Aug 13 17:42:08 bbiab, changing termiansl Aug 13 17:42:19 steliosk: your directory is named wrong Aug 13 17:42:19 * steliosk needs a clear head to clear this thing Aug 13 17:42:25 probably Aug 13 17:42:37 steliosk: FILESDIR has PN, PN-PV and files in it Aug 13 17:42:53 steliosk: not PN-somethingthatlookslikeaversionbutisn't Aug 13 17:42:58 guys what about dropping ARM OABI from dev tree Aug 13 17:43:10 khem: no Aug 13 17:43:14 khem nO! Aug 13 17:43:14 at least for omap? Aug 13 17:43:29 khem or I will hit you with my simpad Aug 13 17:43:35 ha ha Aug 13 17:43:38 khem: OE supports lots of configurations and people are free to choose that ugly OABI Aug 13 17:43:50 khem: and woglinde has a strongarm cpu :) Aug 13 17:43:59 not only me Aug 13 17:44:02 oh my god Aug 13 17:44:02 how about for omap machines? Aug 13 17:44:09 Crofton: why? Aug 13 17:44:16 well, we are lazy Aug 13 17:44:20 the others will throw their zauri at you Aug 13 17:44:22 what's wrong with OABI, out of curiousity? /me is outta the loop Aug 13 17:44:30 it is old :) Aug 13 17:44:33 kergoth nothing Aug 13 17:44:37 it works like charm Aug 13 17:44:40 kergoth: float word order != native word order Aug 13 17:44:46 kergoth: Aug 13 17:44:51 its obsolete Aug 13 17:45:06 kergoth: and you can't mix hardfloat libs with softfloat libs Aug 13 17:45:07 ahh right Aug 13 17:45:25 basically, we do not know of anyone using OABI on the OSK Aug 13 17:45:33 * Crofton needs to check with hrw though Aug 13 17:45:46 there are awful lot of issues that I would nt go tying here Aug 13 17:45:46 the celinux distro might be OABI Aug 13 17:45:49 CELF probably is Aug 13 17:46:01 oh hmmm ok Aug 13 17:46:10 using gcc3 as well, btw Aug 13 17:46:13 yeah Aug 13 17:46:18 I remember Aug 13 17:46:19 *g* Aug 13 17:46:27 koen: Arent they on a branch ? Aug 13 17:46:33 some sort of release branch Aug 13 17:46:42 not that I know off Aug 13 17:46:56 I though dev was for bleeding edge Aug 13 17:47:06 they may just not update Aug 13 17:47:17 I don't know much about that distro Aug 13 17:47:33 koen : how do we handle the case when we have diffrent git trees. i.e mvista and Aug 13 17:47:36 khem, another thought on the omap git build Aug 13 17:47:54 steliosk: different PN Aug 13 17:47:54 does it get mentioned on the name of the file Aug 13 17:47:55 we should be able to build omap1 and omap2 kernels from the same tree Aug 13 17:48:09 Crofton: Yes Aug 13 17:48:28 Crofton: Combining all together would create a very tasty recipe Aug 13 17:48:32 yes Aug 13 17:48:34 xilinx-ml403-mvista_2.6.x+git for example ? Aug 13 17:48:46 no one would be able to touch it without breaking other arches Aug 13 17:48:54 xilinx-ml403-mvista_git.bb Aug 13 17:49:09 and PV = "2.6.+git${SRDATE}" inside the bb Aug 13 17:49:28 Crofton: I think omap divide is ok IMHO Aug 13 17:49:44 yes, for specific builds Aug 13 17:49:55 I would hate to ahve two recipes using same repo Aug 13 17:50:01 but Aug 13 17:50:10 we need to be able to get work done also Aug 13 17:50:21 koen : and the dir name for the config ? Aug 13 17:50:32 * Crofton would like to overhaul the omap recipes when he has time Aug 13 17:50:46 steliosk: the defaults in FILESDIR, or override FILESDIR Aug 13 17:50:53 http://www.pastebin.ca/655803 Aug 13 17:51:10 PN, PN-PV (which changes everyday) or 'files' Aug 13 17:55:05 hey guys, somebody had this error when trying to configure a gcc 3.x.x? Aug 13 17:55:07 configure: error: installation or configuration problem: C compiler cannot create executables. Aug 13 17:55:40 i'm building for a generic i586 Aug 13 17:55:49 birunko: see conf/machine/include/tune-arm926ejs.conf Aug 13 17:56:00 ah, different problem then Aug 13 17:56:29 =\ Aug 13 17:57:48 time to go home. goodnight everyone Aug 13 17:57:56 r Aug 13 17:57:58 e Aug 13 17:58:22 hi florian Aug 13 17:58:38 nite steliosk Aug 13 17:58:47 gn Aug 13 18:13:03 re crofton Aug 13 18:20:22 nite steliosk Aug 13 18:20:27 koen: `/local/sandbox/openembedded/org.openembedded.dev/packages/linux/files/defconfig.eabi': No such file or directory Aug 13 18:20:32 when I added PV Aug 13 18:21:29 koen: How can I make it to look into `/local/sandbox/openembedded/org.openembedded.dev/packages/linux-omap2-svn/files/defconfig.eabi Aug 13 18:22:03 hi all, i have a newbie question: when I do MTN checkout, it tells me there are two heads. Which should I use? koen or florian? Aug 13 18:22:46 * khem should use FILESDIR Aug 13 18:23:19 http://www.pastebin.ca/655803 Aug 13 18:23:19 anville: Wait for a while till they get merged by automerger Aug 13 18:23:22 anville choose one and try 20 inutes later Aug 13 18:23:52 thanks! Aug 13 18:26:11 how can my .bb file install devel files in staging ? need some special config ? Aug 13 18:27:18 psycoman autotools? Aug 13 18:27:20 psycoman: you can call one of the autotools_stage_... functions from do_stage or copy stuff manually to ${STAGING_someDIR} Aug 13 18:28:16 psycoman there are a lot of packages which overrides staging Aug 13 18:29:12 Crofton|laptop: What conf did you use for davinci in uboot package ? Aug 13 18:29:41 davinci_dvevm Aug 13 18:34:54 Crofton|laptop: it is upstream git that has it ? Aug 13 18:36:16 yes Aug 13 18:36:21 commited last week Aug 13 18:38:20 Crofton|laptop: I think I have to use the uboot from http://linux.omap.com/pub/bootloader/2430sdp/source/ Aug 13 18:39:14 Crofton|laptop: There are no patches I found that I can apply on upstream git and I will let the TI guys do that meanwhile I will pull this version of uboot into a new uboot recipe Aug 13 18:39:25 How does that sound Aug 13 18:41:00 ok Aug 13 18:42:02 sounds like you need a special bb file Aug 13 18:46:19 woglinde: im using right this inherit autotools pkgconfig Aug 13 18:47:39 pH5: im making openbox package and the compilation install files to include dir,and install .pc files to, when i add pkgconfig to inherit .pc file go to stage dir Aug 13 18:48:18 i realy need install devel stuff manually to put it in include dir of staging ? Aug 13 18:51:08 pH5: im making openbox package and the compilation install files to include dir,and install .pc files to, when i add pkgconfig to inherit .pc file go to stage dir Aug 13 18:52:39 psycoman again, there are serval packages which do_staging() manually please look at these Aug 13 18:53:00 woglinde: k Aug 13 18:56:14 khem, I think you will need to make a new bb file Aug 13 18:57:03 woglinde: another question where i can read about inherit ? Aug 13 18:57:17 since you have a unique source tarball Aug 13 19:00:40 psycoman in the python doku Aug 13 19:00:43 I guess Aug 13 19:00:54 but inherit and require mostly do the same Aug 13 19:35:22 Crofton|laptop: yeah I did Aug 13 19:36:53 ok Aug 13 19:38:38 Crofton|laptop: PREFERRED_VERSION_u-boot ?= "omap2430sdp" Aug 13 19:38:44 would that be ok ? Aug 13 19:38:58 I am calling it u-boot-omap2430sdp Aug 13 19:39:40 03cbrake 07org.oe.dev * r8fbe7f4f... 10/ (10 files in 3 dirs): compulab-pxa270: drop 2.6.20 kernel now that we have 2.6.22 Aug 13 19:40:13 let me check Aug 13 19:41:32 I'm trying to think how you would specify the package name Aug 13 19:41:45 in the EVM machine file I have Aug 13 19:41:53 PREFERRED_VERSION_u-boot = "git" Aug 13 19:42:15 yeah so you should be using u-boot_git.bb then Aug 13 19:42:16 because I need the git u-boot version, which is not the default version Aug 13 19:42:54 I think you need to have a PREFERRED_PROVIDER_u-boot = u-boot-ompa2430sdp Aug 13 19:43:24 you will get the latest "version" because you have not set the DEFAULT_PREFERENCE in it Aug 13 19:48:54 Crofton|laptop: I have created a conf/machine/omap2430sdp.conf and this I added there Aug 13 19:49:27 btw. I add files to mtn and then later I decide that they should be renamed what is good way to do it in mtn Aug 13 19:49:53 some thing like 'mtn move' Aug 13 19:50:24 mtn mv Aug 13 19:50:50 mtn can be funny with moves, although I think later versions do the right thing Aug 13 19:53:59 * * OE Bug 2754 has been created by autobuild(AT)openembedded.org Aug 13 19:54:01 * * openmoko-devel-image-1.0-r0-do_rootfs Aug 13 19:54:03 * * http://bugs.openembedded.org/show_bug.cgi?id=2754 Aug 13 19:54:10 * * OE Bug 2755 has been created by likewise(AT)gmx.net Aug 13 19:54:12 * * Perl's Cross/Makefile re-uses TOPDIR from environment Aug 13 19:54:14 * * http://bugs.openembedded.org/show_bug.cgi?id=2755 Aug 13 19:54:19 I use 0.35 mtn Aug 13 19:55:26 eek I'm still at work Aug 13 19:58:46 is anyone building the lastest minimo in the dev branch? it seg faults for me in my target Aug 13 20:13:51 03cbrake 07org.oe.dev * r226cc08d... 10/ (1 conf/local.conf.sample): conf/local.conf.sample: update comments with all the image formats Aug 13 20:13:55 03cbrake 07org.oe.dev * r0d2a716d... 10/ (3 files in 2 dirs): compulab-pxa270: add correct recipe file for v2.6.22 Aug 13 20:19:37 How can I find which files are in my tree which are not part of mtn repo yet Aug 13 20:19:51 hm Aug 13 20:19:55 if you did add them Aug 13 20:19:57 mtn diff Aug 13 20:20:14 ah Aug 13 20:20:15 sorry Aug 13 20:21:00 * khem looks at http://monotone.ca/monotone.html Aug 13 20:21:26 khem: mtn list unknown Aug 13 20:21:50 and for those that you have modified since you checked them out, mtn list changed Aug 13 20:23:02 hmmm mtn list I almost reached there Aug 13 20:23:06 thanks mwester Aug 13 20:23:30 thanks mwester Aug 13 20:23:32 * khem was reading 5.4 Informative section Aug 13 20:24:09 oh wow so many mtn list Aug 13 20:26:44 it also has mtn mkdir Aug 13 20:27:10 I feel like if I prepend mtn to any shell command it seems it will work Aug 13 20:29:23 mwester: mtn list unkown does not show me contents of a newly created directory which is not added to mtn yet it shows the directory itself as unknown Aug 13 20:29:57 khem: Yes, mtn versions directories themselves. Aug 13 20:30:10 ok Aug 13 20:30:23 So mtn add is same for a file or a directory right ? Aug 13 20:30:31 yes Aug 13 20:30:32 Yes. Aug 13 20:30:39 mtn add -R will add dir + content Aug 13 20:30:42 ofcourse I have to individually add the new files Aug 13 20:30:55 ah koen read my mind Aug 13 20:31:09 'mtn help' has been greatly improved in 0.36 Aug 13 20:31:27 * khem is using 0.35 Aug 13 20:31:37 I use 0.35 as well :) Aug 13 20:31:48 directory versioning is also why it is usually desirable to use the "mtn mv" command rather than the shell "mv" command followed by "mtn add" -- the latter loses the history. Aug 13 20:32:12 'mv ; mtn add' will create missing files as well Aug 13 20:32:39 git knows you moved it, since it tracks content, not files Aug 13 20:32:45 its only redeeming feature Aug 13 20:32:55 upstream omap git tree is broken :( Aug 13 20:32:55 I cant test my recipes Aug 13 20:32:56 which also means you can't checkin empty dirs..... Aug 13 20:33:10 git doesn't support empty dirs? Aug 13 20:33:18 mwester: it doesn't Aug 13 20:33:35 Yuck! That could be a problem. Aug 13 20:33:48 you can work around it with dot files and other voodoo Aug 13 20:34:16 Seems kludgey though. A directory with no content is no less a directory. Aug 13 20:34:26 It seems logical that it would be versioned. Aug 13 20:35:27 FWIW, mtn didn't do directories till 0.26 or so Aug 13 20:35:48 it only tracked files and implicitly directories holding the files Aug 13 20:45:36 anybody useing 0.36? Aug 13 20:45:44 hm Aug 13 20:45:51 I am not Aug 13 20:46:44 I think I am Aug 13 20:46:52 yep Aug 13 20:47:25 Crofton|laptop: I posted the omap2 patch can you review plz Aug 13 20:48:11 will do Aug 13 20:48:22 ty :) Aug 13 20:48:37 I am not able to build kernel but that omap git's problem Aug 13 20:53:21 bummer Aug 13 20:53:47 nite Aug 13 20:53:58 * * OE Bug 2750 has been RESOLVED (WORKSFORME) by Aug 13 20:54:00 * *  qt4-x11-free-4.3.0-r5-do_populate_staging Aug 13 20:54:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2750 Aug 13 20:55:14 khem, you posted patch to mailing list? Aug 13 20:55:38 yes I think I did Aug 13 20:56:06 I ahven't seen it come in Aug 13 20:56:15 It may be tomorrow AM b4 I can look at it Aug 13 20:56:15 me either Aug 13 20:56:59 its here http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-August/002795.html Aug 13 20:57:07 I need to talk to my friend who is interning at TI and have him look up some people and grab some boards Aug 13 20:57:44 the list may strip the attachment, can you add it inline Aug 13 20:57:51 or send direct to me? Aug 13 21:00:45 03rpurdie * r881 10bitbake/lib/bb/ (cooker.py data.py): cooker.py: Start UI merge: Cleanup exit status handling and collect exit calls into one place. Move buildSetVars() function calls. Cleanup whitespace. Make some of the core functions more generic. Aug 13 21:02:16 03rpurdie * r882 10bitbake/lib/bb/data.py: data.py: Comment last accidental commit Aug 13 21:03:24 Crofton|laptop: the patch is there if you look at the link right at the bottom of that post Aug 13 21:03:26 http://lists.linuxtogo.org/pipermail/openembedded-devel/attachments/20070813/6e8b9ea7/attachment-0001.obj Aug 13 21:03:30 is the link to patch Aug 13 21:05:56 ok, I mailed the link to myself :) Aug 13 21:06:23 I need to work through the "right" way to select the uboot bb file Aug 13 21:06:35 what dsp does the 2430 use? Aug 13 21:12:18 Crofton|laptop: http://focus.ti.com/pdfs/wtbu/TI_sdp_omap2430.pdf Aug 13 21:14:01 Crofton|laptop: DSP is Embedded TMS320C55x Aug 13 21:14:08 this suggests C55 yeah Aug 13 21:14:24 The DSP guys I know prefer the C6x series Aug 13 21:14:33 but I think power is lower in c55 Aug 13 21:16:15 I know arm side of omap only :) Aug 13 21:18:55 I need to learn both sides Aug 13 21:27:31 03rpurdie * r883 10bitbake/lib/bb/cooker.py: cooker.py: Refactor pkgs_to_build into a function, further cleanups Aug 13 22:52:09 03rpurdie * r884 10bitbake/lib/bb/ (__init__.py cooker.py msg.py): msg.py: Add plain message type, remove legacy logging code Aug 13 22:56:58 * * OE Bug 2756 has been created by autobuild(AT)openembedded.org Aug 13 22:57:00 * * linux-2.6.21-r5-do_unpack Aug 13 22:57:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2756 Aug 13 23:02:31 Hi, I am trying to build openssh for Angstrom but it fails: http://paste.stgraber.org/2517 Aug 13 23:02:46 Does anybody else see this problem? Aug 13 23:04:43 And even more important, how do I get rid of it? Aug 13 23:13:11 03rpurdie * r885 10bitbake/lib/bb/cooker.py: cooker.py: Switch to using bb.msg.plain Aug 13 23:47:57 03rpurdie * r886 10bitbake/lib/bb/msg.py: msg.py: Add checks for unset domains Aug 13 23:49:09 03rpurdie * r887 10bitbake/lib/bb/runqueue.py: runqueue.py: Mark completed tasks as buildable to avoid sanity check errors Aug 13 23:55:28 lichtgestalt: I've added openssh 4.6p1 - it fixes that compile problem. Please test it. Aug 14 00:10:55 RP: sqlite3 and pysqlite2 means that OE can no longer build on RHEL4. Aug 14 00:12:32 rwhitby: :/ Aug 14 00:12:54 I'm looking at upgrading the machine to CentOS 5. Aug 14 00:13:58 rwhitby: Sorry to cause problems. We needed a multi process safe data store and sqlite seemed the easiest way to do it... Aug 14 00:14:28 RP: yeah, I've been following the rationale, and don't disagree with it :-) Aug 14 00:14:46 so what are supported hosts for OE is there a list somewhere ? Aug 14 00:15:29 khem: In theory anything that provides our prerequisits Aug 14 00:15:45 khem: there are *no* supported OE hosts as far as I can tell ;-) Cause each of the core team members uses a different distro, so it breaks in different ways each time a major change happens :-) Aug 14 00:16:04 * rwhitby is not complaining, just setting expectations Aug 14 00:16:09 rwhitby: Poky uses ubuntu as a base... Aug 14 00:16:12 I see I use Fedora 7 so far all seems smooth for me Aug 14 00:16:33 khem: Debian Stable has been the least trouble for me with OE Aug 14 00:16:39 rwhitby: Sadly the version of sqlite in debian stable is too old for bitbake too :/ Aug 14 00:16:49 really, in Etch? Aug 14 00:17:03 I have it running in Etch (which is now debian stable) Aug 14 00:17:12 Ubuntu is your best bet on debian-like systems I guess Aug 14 00:17:21 rwhitby: Perhaps it was something older then... Aug 14 00:17:34 RP: it was probably sarge. Aug 14 00:17:56 rwhitby: Probably. As you say, etch is sqlite 3.3.8 which is goo enough Aug 14 00:18:30 meh - I have to test upgrading to CentOS 5 anyway for work, so I might as well do it on my OE build box first :-) Aug 14 00:18:34 rwhitby: Anyhow, since each of the core team runs a different distro, we get good coverage ;-) Aug 14 00:18:55 RP: yeah, that's the glass half full way of looking at it :-) Aug 14 00:18:58 Centos 5 is like Fedora Core 6 I believe Aug 14 00:19:13 CentOS X == RHEL X. Aug 14 00:19:33 oh well Fedora is their community version Aug 14 00:19:38 yep Aug 14 00:19:49 but Fedora != RHEL in a lot of cases Aug 14 00:20:05 hmmm thats expected :) Aug 14 00:20:09 rwhitby: If sqlite becomes a major problem, we can write a drop in replacement for persist_data.py which uses files and locks if sqlite is unavailable Aug 14 00:20:21 but I wonder how come CentOS=RHEL Aug 14 00:20:25 rwhitby: I'd prefer not to do that if we don't have to though... Aug 14 00:20:36 khem: CentOS is RHEL sources recompiled with logos stripped out. Aug 14 00:20:41 rwhitby: Great. Thank you. I will test it out and let you know how it goes. Aug 14 00:21:06 rwhitby: then why do pay RHEL for subscriptions Aug 14 00:21:13 We don't actually need any SQL, it was its optimised locking I wanted... Aug 14 00:21:26 anyhow, 'night all Aug 14 00:21:28 khem: only people who have not heard of CentOS do that. Aug 14 00:21:32 * RP -> sleep Aug 14 00:22:11 (or people who have some corporate wish to spend money on "support" that will never give them the help they need anyway. Aug 14 00:22:12 rwhitby: hmmm I use CentOS 3 here on one of servers the machine is running for last year and half now Aug 14 00:55:10 khem, people pay Redhat because they want someone to answer their questions Aug 14 00:57:41 Crofton|home: yeah :) Aug 14 00:57:43 bbiab Aug 14 00:57:44 or Aug 14 00:57:54 why do people pay MV, when OE is "free" Aug 14 00:59:24 because MV will sign a contract, has insurance, and a price list. ;) Aug 14 00:59:44 Same for RedHat... Aug 14 00:59:58 * * OE Bug 2757 has been created by autobuild(AT)openembedded.org Aug 14 01:00:00 * * linux-nokia770-2.6.16-osso26-r2-do_deploy Aug 14 01:00:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2757 Aug 14 01:00:10 * * OE Bug 2758 has been created by autobuild(AT)openembedded.org Aug 14 01:00:12 * * linux-nokia770-2.6.16-osso26-r2-do_populate_staging Aug 14 01:00:14 * * http://bugs.openembedded.org/show_bug.cgi?id=2758 Aug 14 01:01:36 rwhitby: Thanks. openssh compiled fine after removing the default preference of -1 Aug 14 01:03:08 lichtgestalt: once a couple of other people test it out I will make it the default preference. Aug 14 02:26:58 * * OE Bug 2759 has been created by autobuild(AT)openembedded.org Aug 14 02:27:00 * * angstrom-minimal-image-1.0-r3-do_rootfs Aug 14 02:27:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2759 **** ENDING LOGGING AT Tue Aug 14 02:59:56 2007