**** BEGIN LOGGING AT Fri Aug 17 02:59:56 2007 Aug 17 03:48:58 * * OE Bug 2794 has been created by jjpfeifferjr(AT)comcast.net Aug 17 03:49:00 * * Unable to fetch source for intltool Aug 17 03:49:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2794 Aug 17 04:27:55 mwester: ping Aug 17 07:35:16 morning all Aug 17 07:37:57 RP: hello :) Aug 17 07:38:10 koen: I have a problem with fic-gta01-angstrom-linux-gnueabi/linux-2.6.21-r5, http://rafb.net/p/D07fht23.html Aug 17 07:38:23 Any hint what's wrong? Aug 17 07:42:50 RP: mwester has some patches to get kexec cmdline processing working on arm. I just managed to succesfully test them on poodle. Hopefully he'll post them to the list tomorrow. Aug 17 07:48:04 koen: Never mind, copying the defconfigs from the linux-gta01 folder solved it Aug 17 07:51:23 hvontres|home: Sounds good Aug 17 07:52:10 RP:yup... now if I could only figure out the video corruption on poodle, we could get some decent test images ready. Aug 17 07:52:24 RP: any thoughts? Aug 17 07:53:25 hvontres|home: Not really, I can only suggest going back to when that didn't happen and see what changed Aug 17 07:54:46 morning Aug 17 07:55:40 RP: hmmm....well, I took a look at pxafb.c and there was no obvious place in the changes between 2.6.17 and 2.6.20.... Aug 17 07:55:59 hvontres|home: So did 1.6.19 work? 1.6.18? Aug 17 07:56:13 morning XorA Aug 17 07:58:12 RP: well, the .bb's for those two revs are gone now.... Aug 17 07:58:33 hvontres|home: So find them and bring them back. We use an SCM... Aug 17 07:58:59 RP:ok, I'll take a look at that this weekend..... Aug 17 07:59:18 RP: time to go to sleep. Aug 17 08:02:05 * hvontres|home is away: Gone away for now. Aug 17 08:03:17 RP: moin Aug 17 08:03:53 RP: :) Aug 17 08:06:37 Is there a way to get monotone to list which files are in a workspace directory that neither exists in the repository or have been added locally yet? Aug 17 08:06:58 esben: status? Aug 17 08:07:09 or ls --help Aug 17 08:08:35 mtn ls --unknown Aug 17 08:09:24 XorA: thanks, "mtn ls unknown" does the job :-) Aug 17 08:21:30 hi zecke Aug 17 08:25:09 zecke: Is http://rafb.net/p/jKL3l869.html a good or bad idea? Aug 17 08:25:47 zecke: 12% speed up, removal of 400,000 function calls ;-) Aug 17 08:26:41 one question :) Aug 17 08:26:53 will you add key more than once to the todolist? Aug 17 08:27:34 zecke: Not unless we can have multiple variables of the same name? Aug 17 08:28:12 can you make todolist an array ([]) Aug 17 08:28:29 and add tuples (todolist.append( (key,ekey))) Aug 17 08:28:48 and iterate for (key,ekey) in todolist: renameVar(key, ekey, alterdata) Aug 17 08:29:11 zecke: I could I guess although we should never hit the multiple variable case Aug 17 08:33:32 Is rm_work known to be broken at the moment? Aug 17 08:34:03 It seems to be only working on an image package, and no other packages. Aug 17 08:34:17 (bitbake 1.8 tip) Aug 17 08:35:33 rwhitby: We were dicussing that yesterday. It needs its dependencies setting correctly for 1.8 Aug 17 08:35:49 rwhitby: The plan is http://rafb.net/p/b6rLId11.html Aug 17 08:36:48 rwhitby-n800: You can use a version of that with s/RMWORK_ORIG_TASK/build/ for now. Using the fancy variables works in 1.8 svn head only Aug 17 08:37:21 Or we check that in now and people wanting to use it use svn head? Aug 17 08:39:21 any reason why stuff is changing so that it doesn't work with 1.8 stable? Aug 17 08:39:55 rwhitby-n800: rm_work doesn't work with 1.8 atm and never has Aug 17 08:39:58 good morning Aug 17 08:40:11 ah, ok Aug 17 08:40:23 rwhitby-n800: We're trying to fix that, we need a tweak in bitbake to make it work as nicely as I'd like Aug 17 08:41:31 RP: do you have new numbers? Aug 17 08:41:34 rwhitby-n800: I will check in something that works with 1.8 stable Aug 17 08:41:46 zecke: new numbers for what? Aug 17 08:42:55 RP: the proposed change to not use a hash and do lookups? Aug 17 08:43:48 RP: many mokians will thank you Aug 17 08:44:53 rwhitby-n800: pushed Aug 17 08:45:14 zecke: I don't follow? Aug 17 08:46:10 RP: you say your proposed patch is a 12% speedup. And I wonder if we can further improve it by not using a hash but a list of tuples Aug 17 08:46:25 zecke: That isn't where the speedup comes from Aug 17 08:47:01 zecke: data.expand has the expand_cache I added. That cache is wiped on any write operation on data. renameVar is a write operation Aug 17 08:47:05 RP: right, but when you optimize for speed, why don't do it the best way? Aug 17 08:48:32 zecke: dicts are that much slower than the tuples? Aug 17 08:49:32 ciao a tutti Aug 17 08:49:58 RP: I want to find that out :) Aug 17 08:51:15 zecke: It gives no performance difference that my benchmarks can detect. My statistical error is higher Aug 17 08:52:54 try the same with map instead of for :) Aug 17 08:53:24 zecke: It won't make any difference. The for loops are not the slow part Aug 17 08:53:52 zecke: The speedup is the loss of 400,000 other function calls since we start hitting the cache Aug 17 08:53:52 RP: I know, it is more a general finding Aug 17 08:54:18 and I agree with moving the renameVar calls out of the primary loop Aug 17 08:54:37 zecke: You'll need to write some tests with higher numbers of loops to spot the differences, my benchmarks don't show that Aug 17 08:55:00 zecke: Did you see my change to add_task -> add_tasks ? Aug 17 08:55:18 zecke: That was another 6% and 600,000 function calls :) Aug 17 08:55:32 zecke: I added it to trunk, I'm not sure whether to add to the 1.8 branch too Aug 17 08:56:29 I'm once again reading http://wiki.python.org/moin/PythonSpeed/PerformanceTips Aug 17 08:56:35 zecke: I started at 8.8million function calls FWIW as an idea of scale Aug 17 08:56:37 and think we should try map + get Aug 17 08:56:52 zecke: Its pointless when we have much bigger problems Aug 17 08:57:38 I would not call it pointles but low hanging fruits Aug 17 08:58:04 at the camp i profiled and the most often called method was getVarFlag, is that still the case? Aug 17 08:58:26 zecke: It will gain 0.0001% speed differences. It is therefore not a low hanging fruit Aug 17 08:58:37 zecke: run bitbake -pdP, find a function which a suspicously large number of calls and look at them. They are the low hanging fruit Aug 17 08:59:01 zecke: getVarFlag is 6th on the list Aug 17 08:59:34 zecke: The profile output also shows where functions are called from so you can see who is generating the large numbers of calls Aug 17 08:59:48 RP: I don't agree with your 0.0001%, but changing most loops and variable assignments to prove you wrong is not doable for me atm :} Aug 17 09:00:26 zecke: ok, you might get 0.1% but I stand by my point, that there are bigger and easier gains to be made elsewhere Aug 17 09:01:41 03rpurdie 07org.oe.dev * rcdc64e6b... 10/ (1 classes/rm_work.bbclass): rm_work: Fix to work with bitbake 1.8. Changes to make this optimal need 1.8.8 so they're in comments for now Aug 17 09:02:21 need to start working now, I hope to not be too tired tonight Aug 17 09:03:53 RP: I get circular dependencies on do_rm_work: http://rafb.net/p/U3Ovsf76.html Aug 17 09:04:12 your todolist change looks good there is one potential difference. If you have the variable twice in the list (which shouldn't be the case) we used to rename the first occurence (whatever that was) and now we will rename the last one Aug 17 09:07:25 zecke: There is another subtle problem. After each rename we updated the data so one rename could have used data from some previous rename. Aug 17 09:07:52 zecke: I'm tempted to think of that as a bugfix though as the data it deterministic now Aug 17 09:08:58 do you want to do a release before or after this change? Aug 17 09:12:14 bitbake has a strange behaviour. A build starting from the command line did not work, but starting it from the interactive shell it starts the compilation. strange. http://pastebin.ca/660793 Aug 17 09:12:28 it is a desired behaviour? Aug 17 09:16:22 cyberdeck: Something certainly looks wrong Aug 17 09:16:26 cyberdeck: bitbaking 'build' is never going to work Aug 17 09:16:50 bitbake angstrom-x11-image , will work hopwever Aug 17 09:17:00 however even Aug 17 09:17:14 cyberdeck: right :) Aug 17 09:17:14 * cyberdeck should catch some sleep. sorry for bothering you! Aug 17 09:17:35 rp: thank you! Aug 17 09:17:42 ade|desk: Thanks for spotting that, I was thinking it was some weird side effect of the rm_work change :} Aug 17 09:17:51 not today Aug 17 09:17:53 ;) Aug 17 09:19:36 * XorA waves an electric eel in lrgs direction Aug 17 09:26:04 hi all Aug 17 09:33:47 RP: I think i have another "instance" of the dependency problem. Doing a clean build git-native fails because expat was not build Aug 17 09:34:08 s/expat/expat-native Aug 17 09:34:30 RP: do we need to update the manual regarding SRC_REV? Aug 17 09:34:43 zecke : hi Aug 17 09:47:39 03hrw 07org.oe.dev * r0495ced6... 10/ (1 packages/python/python-pyode_1.0.0.bb): python-pyode: depend on ode Aug 17 09:47:43 03hrw 07org.oe.dev * rf4c81cd8... 10/ (1 packages/python/python-pygtk-1.2_0.6.12.bb): python-pygtk-1.2: fix build on amd64 Aug 17 09:51:23 zecke: Probably Aug 17 09:51:44 steliosk: Which bitbake are you using? Aug 17 09:54:27 RP: 1.8.7 Aug 17 09:54:46 steliosk: which svn revision though? Aug 17 09:55:56 RP: hmmm not sure. I am using svn localy and had to remove bitbakes svn stuff Aug 17 09:56:38 RP: should i get the latest ? Aug 17 09:56:40 steliosk: I fixed somethings to do with dependency problems in bitbake svn recently which is why I ask Aug 17 09:57:00 RP: it should be over a month or so old the version i am using Aug 17 09:57:11 steliosk: I'd update then ;-) Aug 17 09:57:46 steliosk: The fix I have in mind went in during the last few days Aug 17 09:57:56 RP: you know what the funny thing is. The problem does not show up on amd64 machines Aug 17 10:55:40 RP: it happened again after the pull :( Aug 17 10:56:25 RP: the main difference between the 2 build machines is that the amd64 does not do paraller builds Aug 17 10:58:36 steliosk: :-( Aug 17 10:58:48 I'm going to need to be able to repoduce it... Aug 17 10:59:28 RP: i can give you access to the build cluster Aug 17 11:00:16 steliosk: Can you describe the problem exactly first please? Aug 17 11:02:25 RP: started clean build and then git-native fails -> http://rafb.net/p/N8KFOq36.html Aug 17 11:03:49 steliosk: ok, can you run bitbake xyz -g instead and share the task-depends.dot file? Aug 17 11:03:58 (with a clean build) Aug 17 11:05:14 RP: ok will do Aug 17 11:07:51 steliosk: The idea is to see whether the expat-native dependency appears on that graph Aug 17 11:12:59 * * OE Bug 2795 has been created by simpad(AT)secretlab.de Aug 17 11:13:01 * * busybox-1.2.1 failed to build Aug 17 11:13:03 * * http://bugs.openembedded.org/show_bug.cgi?id=2795 Aug 17 11:31:58 * * OE Bug 2794 has been RESOLVED (WORKSFORME) by philippedeswert(AT)scarlet.be Aug 17 11:32:00 * *  Unable to fetch source for intltool Aug 17 11:32:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2794 Aug 17 11:38:58 * * OE Bug 2793 has been RESOLVED (WORKSFORME) by philippedeswert(AT)scarlet.be Aug 17 11:39:00 * *  change SRC_URI Aug 17 11:39:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2793 Aug 17 12:02:58 * * OE Bug 2796 has been created by jstoefer(AT)schuett-elektronik.de Aug 17 12:03:00 * * Build of matchbox-keyboard fails Aug 17 12:03:03 * * http://bugs.openembedded.org/show_bug.cgi?id=2796 Aug 17 12:11:22 rwhitby: The other odd feature of this bug is it only happens ever other bitbake run :/ Aug 17 13:13:20 RP: it looks like 'rev' for svn is not honored anymore Aug 17 13:14:25 oh my bad Aug 17 13:21:56 zecke : found an issue with tinderbox and new bitbake. If parallel make is used and build fails the tinderbox still waits for data (build colour stays yellow) instead of signalling failure (i.e red) Aug 17 13:22:26 zecke : http://ewi546.ewi.utwente.nl/tinderbox/showbuilds.pl?tree=oplinux Aug 17 13:22:28 steliosk: don't use parallel make with tinderbox :} Aug 17 13:23:08 zecke : actually is parallel bitbake that's the issue Aug 17 13:23:30 zecke : or more correct the multi-thread Aug 17 13:23:32 steliosk: parallel bitbake :) Aug 17 13:23:55 steliosk: PkgStarted/Succeeded are fired more than once Aug 17 13:25:14 zecke : hmmm and tinderbox looses "count" of what's going on i assume... Aug 17 13:26:12 zecke : when you get a failure from one of the threads but the others keep sending PkgStarted/Succeded Aug 17 13:26:14 steliosk: if PkgStarted/PkgFailed would be correct, it should be 'okay' Aug 17 13:26:21 reading the results would still be no fun Aug 17 13:27:38 zecke : disabling multi-threads in bitbake slows things down unfortunately Aug 17 13:28:47 steliosk: if I just could quit my job :) Aug 17 13:29:18 zecke : would it make sense to write all tinderbox results locally and send them to the tinderbox server once the build finishes (either successfull or not) Aug 17 13:29:48 steliosk: it would make sense to finish my tinderbox server, and all problems just go away Aug 17 13:30:00 zecke : you have to work for a leaving ? :) Aug 17 13:30:21 zecke : its so windblow$ :) Aug 17 13:43:58 * * OE Bug 2797 has been created by sugia(AT)yandex.ru Aug 17 13:44:00 * * kernel-module-sharpsl-rc make kernel panic. Aug 17 13:44:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2797 Aug 17 14:09:14 03crofton 07org.oe.dev * r8f0990c1... 10/ (1 packages/uboot/uboot-utils_1.2.0.bb): uboot-utils : Update SRC_URI to ftp.denx.de. Aug 17 14:17:48 * lrg is looking for another kernel/driver engineer (UK - Edinburgh, Swindon, High Wycombe, or Germany - Stuttgart) with some OE experience. Aug 17 14:18:46 Blacksburg, VA? Aug 17 14:20:01 Crofton|home: I'd have to check, we may be able to do Boston or San Diego Aug 17 14:20:17 nope :) Aug 17 14:20:28 I am stubborn Aug 17 14:20:38 We have a lovely research park here Aug 17 14:20:41 Crofton|home: np Aug 17 14:21:32 I'm probably not the most qualified kernel/driver guy on the bloxk either Aug 17 14:22:21 Hopefully I got the last two SRC_URI updates in u-boot correct :) Aug 17 14:22:43 I guess everyone that builds those u-boots never cleans bb_sources Aug 17 14:27:29 RP: http://paste.lisp.org/display/46285, r=+? Aug 17 14:28:40 zecke: missing_lis) ? :) Aug 17 14:28:53 awesome :) Aug 17 14:29:25 I'm not sure I get all missing deps, but it worked here for the DEPENDS case Aug 17 14:29:43 and is a usability improvement (I'm not forced to use -v) Aug 17 14:29:57 Yes, we need to do something like that certainly Aug 17 14:30:31 okay to commit (with the additional 't')? Aug 17 14:32:36 RP: When using the git fetcher how does it handle the situation where we need to merge 2 different git reps ? Aug 17 14:33:07 does it clone both ? clones the first and pulls from the second ? Aug 17 14:34:00 steliosk: Its a clone/pull. It should never need to merge Aug 17 14:34:08 I checked both DEPENDS, RDEPENDS now and will commit Aug 17 14:34:27 zecke: Thank you for fixing that issue properly :) Aug 17 14:34:37 in some situation people recommend to clone Linus tree for example and pull from theirs due to limited bandwidth etc Aug 17 14:34:41 RP: irony? Aug 17 14:35:23 zecke: No, I came under a lot of critisim for not just changing the log levels on that a while back Aug 17 14:38:20 RP: they recommend something like "clone tree a, cd to it, pull tree b" Aug 17 14:38:46 RP: so that only differences are pulled from tree b Aug 17 14:38:53 RP: I'm not sure I got all calls of fail_fnid, but the output already helps me Aug 17 14:41:22 zecke: Yes, this is the way I wanted to fix it but never had time :) Aug 17 14:41:43 steliosk: You'll have to write a custom task to fetch that, its beyond the fetcher Aug 17 14:43:21 RP: so the fetcher actually clones both trees (or all those mentioned in the SRC_URI) ? Aug 17 14:43:39 steliosk: yes Aug 17 14:45:32 03crofton 07org.oe.dev * r56cb1415... 10/ (3 files in 2 dirs): u-boot : Update SRC_URI to the Denx ftp server. Aug 17 14:47:06 03zecke123 * r908 10/ (4 files in 4 dirs): Aug 17 14:47:06 Print the list of missing DEPENDS/RDEPENDS on the "No buildable providers available for required build target..." error message Aug 17 14:47:06 I'm not certain to get all occurences of fail_fnid but the printed message is a usability improvement. So this change might be not complete but good enough to be applied. Aug 17 14:49:12 good afternoon Aug 17 14:49:29 hmm I miss bitbake hacking Aug 17 14:49:34 I have a little problem that I am encountering from time to time, Aug 17 14:49:55 maybee someone here can give me a hint. Aug 17 14:50:39 When I am bitbaking an image, at the last stage, whenits creating the rootfs, it fails. Aug 17 14:51:08 jan_js: what is the error? use pastebin for posting the message Aug 17 14:51:10 ~pastebin Aug 17 14:51:11 somebody said pastebin was a place to paste your stuff without flooding the channel - try http://pastebin.ca, or http://channels.debian.net/paste, or http://rafb.net/paste/, or http://pastebin.com is usually painfully too slow and unresponsive to use, use one of the other pastebin sites, or dpaste.com is a very nice pastebin as well Aug 17 14:51:32 In the log is something from img_cmd saying that there is no space left on the device. Aug 17 14:51:48 and that is not true? Aug 17 14:51:52 However, du tells that I am only using 29% on that device Aug 17 14:52:01 do you have quota? Aug 17 14:52:27 zecke: sorry,what do you mean Aug 17 14:54:55 zecke: What do you mean with 'quota' ? Aug 17 14:57:43 morning Aug 17 14:57:54 jbs: is this disk controlled by you? or is it property of your university or such? Aug 17 15:01:08 jan_js: is this disk controlled by you? or is it property of your university or such? Aug 17 15:06:49 * Bernardo kicks his modem and hopes for a stable connection Aug 17 15:49:58 * * OE Bug 2796 has been RESOLVED (FIXED) by philippedeswert(AT)scarlet.be Aug 17 15:50:00 * *  Build of matchbox-keyboard fails Aug 17 15:50:02 * * http://bugs.openembedded.org/show_bug.cgi?id=2796 Aug 17 15:54:23 03philippe 07org.oe.dev * r6872bbed... 10/ (4 files in 2 dirs): Aug 17 15:54:23 matchbox-keyboard_svn: fix compilation issues due to oudated patches (fixes bug #2796) Aug 17 15:54:23 * updated 2-Add-new-modifier--layout--Used-to-cycle-thru-all-available-layouts.patch Aug 17 15:54:23 * updated 3-Changes-to-improve-layout-rendering--especially-after-adding-support-for.patch Aug 17 15:54:23 * updated 4-Add-rendering-debug-logging.patch Aug 17 16:09:14 hi all Aug 17 16:09:25 morninga Aug 17 16:09:33 -a Aug 17 16:11:35 hi koen|770 Aug 17 16:15:51 hi koen|770 Aug 17 16:19:17 stinking nokia crap Aug 17 16:21:41 where is definied task-base-extended ? Aug 17 16:22:26 it is in RDEPENDS of angstrom-console-image, but where i can set or get list of packages for this ? Aug 17 16:27:57 * * OE Bug 2723 has been RESOLVED (FIXED) by Aug 17 16:28:00 * *  makedepend do_compile problem: No rule to make target `makedepend.@APP_MAN_SUFFIX@', needed by `all-am'. Aug 17 16:28:01 * * http://bugs.openembedded.org/show_bug.cgi?id=2723 Aug 17 16:28:59 03hrw 07org.oe.dev * r4843e0d9... 10/ (3 files in 2 dirs): makedepend(-native): fix DEPENDS (taken from Poky) - close #2723 Aug 17 16:31:18 RP: I have a weird deadlock with bitbake-1.8 Aug 17 16:33:56 zecke: Hmm. What kind of deadlock? Aug 17 16:34:24 RP: do_fetch of a local file is not succeeding. I'm not certain how to debug this Aug 17 16:34:33 RP: using -b this builds okay Aug 17 16:35:25 zecke: What does stty on your terminal say? tostop? Aug 17 16:36:14 yupp Aug 17 16:36:23 zecke: stty -tostop Aug 17 16:36:49 We should test for that in sanity.bbclass... Aug 17 16:37:57 hmm can't we call stty -tostop ourselves? Aug 17 16:38:16 zecke: We could I guess Aug 17 16:38:29 zecke: I'd prefer to warn users their environment was broken and let them fix it Aug 17 16:42:15 good evening Aug 17 16:45:17 hi pH5 Aug 17 16:46:51 hi RP Aug 17 16:48:23 Hi guys, not sure if this is the correct channel, but, I am trying to build a lightweight X11 image for an Akita. Aug 17 16:48:44 Something like xterminal-image for angstrom. Aug 17 16:49:08 I need some help figuring it out. Aug 17 16:49:38 which packages are definied in task-base-extended ? Aug 17 16:50:13 There are quite a few, not sure which are required for X and what is needed for GPE. Aug 17 16:53:10 : Aug 17 16:54:12 I built a GPE image that worked, booted etc. but I only had about 5MB of RAM left. I need about 30MB free. Is that possible on an akita? Aug 17 16:54:38 GPE image has about 3000 packages! Aug 17 16:54:45 at the moment you can instal console image Aug 17 16:54:57 and before install via ipk the x server Aug 17 16:55:04 dothedog: X+Gtk+ needs ~40-60mb RAM to be usable Aug 17 16:55:15 ouch Aug 17 16:56:06 at least this was my findings Aug 17 16:56:18 What does the xterminal-image take? Aug 17 16:56:27 no idea :) Aug 17 16:57:11 zecke: i know how i can re-define task-base-extended ? Aug 17 16:57:26 I tried building the xterminal-image in .dev but the xserver didn't start when I tried to boot. Aug 17 16:57:45 dothedog: have tou tried startx ? Aug 17 16:58:11 Yeah, I think there is a problem with the .bb file for the xterminal-image. Aug 17 16:58:45 you can try install this ipkg (xserver-nodm-init) too Aug 17 16:58:54 Simple Xserver Init Script (no dm) Aug 17 16:59:43 psycoman: is this an image.bb file? or something you ipkg install to console image? Aug 17 17:01:04 ipkg install to console image Aug 17 17:01:38 psycoman: I'll give it a shot. Aug 17 17:02:02 psycoman: also, what do you need to change in task-base-extended? Aug 17 17:03:04 dothedog: i want make a personal image Aug 17 17:04:08 psycoman: its defined in task-base.bb line 109. I think you can just add/delete stuff in the RDEPENDS_task-base-extended section. Aug 17 17:05:35 psycoman: what I have been doing is copying out tasks that I need for my image to a "mypackages" dir, renaming them and changing to my heart's content. Aug 17 17:08:29 that way I don't lose the changes when I do a pull/update Aug 17 17:16:12 dothedog, BBCOLLECTIONS Aug 17 17:19:42 dothedog: http://bitbake.berlios.de/manual/ch04s02.html#id870544 link to BBCOLLECTIONS manual Aug 17 17:32:56 Yeah, That's what I did, I created a new directory in /stuff and defined it in BBFILES. Aug 17 18:00:31 huhu Aug 17 18:21:11 Hi, I have OE setup, I now want to create a distro and get it installed onto Soekris 4826, can anyone point me at the next steps Aug 17 18:22:47 I need something that I can load up over pxe/tftp so I can copy a full image to the onboard flash Aug 17 18:23:06 Is there something like bitbake initrd-image ? Aug 17 18:33:25 do we have a machine.conf for that? Aug 17 18:35:18 budgester, look for CosmicPenguin Aug 17 18:35:26 he should be able to help some Aug 17 18:36:01 1 Aug 17 18:36:10 2 Aug 17 18:36:36 The soekris uses a geode processor so thats what I was starting from Aug 17 18:37:01 CosmicPenguin, is familiar with geode stuff Aug 17 18:37:11 Cool, I'll keep a look out Aug 17 18:37:18 I am not sure what machines we have that would be good starting places Aug 17 18:37:41 for distro, try angstrom-2007.1 Aug 17 18:37:48 Initially to learn about some of this stuff, I figure create an image then run it on qemu Aug 17 18:38:05 Its been quite a while since my name has come up in this room.. :) Aug 17 18:38:09 BTW, extreme newbie to this stuff Aug 17 18:38:46 there are a couple of machine config's with "geode" in the name Aug 17 18:39:17 budgester, has a Soekris 4826 Aug 17 18:40:05 yup, the geodelx and geodegx, the cpu in these is a geode SC1100 Aug 17 18:41:11 What I would like to finally end up with is a simple distro, that runs php with ldap, ssh Aug 17 18:41:28 So I am now trying to work out the steps I need to take Aug 17 18:42:22 SC1100 should probably be just a generic 386 Aug 17 18:42:32 there definately isn't a kernel config for it Aug 17 18:50:37 ok, gonna play a bit, prolly be back asking questions Aug 17 18:53:16 is there a way to force a distro to use a certian version of the kernel? Aug 17 18:53:37 mickey|out: ping? Aug 17 18:54:20 i'm only using the rootfs, the kernel is someone else's Aug 17 18:57:38 hello all Aug 17 18:57:44 hello Aug 17 19:01:37 hey likewise Aug 17 19:02:34 local.conf... i feel like a fool Aug 17 19:03:33 Hey all Aug 17 19:04:08 hello again likewise Aug 17 19:04:11 likewise, I have got as far as installing the denx tool chain for ppc :) Aug 17 19:06:05 Crofton: which one, the OSELAS (sp?) one? Aug 17 19:06:54 not sure Aug 17 19:06:57 from denx.de Aug 17 19:07:23 Crofton Aug 17 19:07:40 Crofton: was mistaken, I though of this one: http://www.pengutronix.de/software/ptxdist/ Aug 17 19:07:58 ah Aug 17 19:08:06 I have the denx one Aug 17 19:08:07 Crofton: "reproducable" is something OE could work on :-) Aug 17 19:08:13 yeah Aug 17 19:08:22 Crofton: OK, I'll probably try this one. Aug 17 19:08:35 this morning I fixed some bad source url's in the older u-boot packages Aug 17 19:08:46 masked by bb_sources on my builds Aug 17 19:09:11 Crofton: We still did not switch to u-boot because of the IXP4xx network driver stuff... Aug 17 19:09:12 not only do we need autobuilds, we need something to test source availibilty Aug 17 19:09:44 Crofton: I remember a fetchall, but it probably only fetches the packages in the deps, not all. Aug 17 19:10:10 runned autobuilder without using cached sources Aug 17 19:10:17 isn't there bitbake -cfetch world or something like that? Aug 17 19:11:00 We need a way to test every bb file to look for bad sRC_URI entries Aug 17 19:13:50 03stefan 07org.oe.dev * r7250bdd6... 10/ (9 files in 3 dirs): libgsmd: Catch up with upstream changes. All patches are applied now. Aug 17 19:14:43 does anyone know of a way to configure OE to build to a particular kernel major and minor version? Aug 17 19:15:06 or would it just be better for me to create symlinks on the rootfs? Aug 17 19:15:19 which kernel? Aug 17 19:15:51 it's building 2.6.18-omap1, i need it to build 2.6.16.27-omap1 Aug 17 19:16:36 hmm Aug 17 19:16:43 which machine btw? Aug 17 19:16:47 nokia 700 Aug 17 19:16:49 770 Aug 17 19:17:02 ca0abin: with PREFERRED_VERSION one can set versions, but these not be available/buildable :) Aug 17 19:17:18 basically, you would need to make a bb fiel that fetches that version Aug 17 19:17:25 zecke - saw that, only lets you control major version on 3 platforms Aug 17 19:17:40 ??? Aug 17 19:17:43 Do you have a source URL for this version? Aug 17 19:18:07 not as such, but i'd imagine it would be available on the same mirror as the 2.6.18 Aug 17 19:19:07 http://www.muru.com/linux/omap/patches/old/ Aug 17 19:19:38 I think the .27 suggests patches against 2.6.16 Aug 17 19:21:41 unfortunately i don't know much about bitbake. can anyone recommend a good place to learn what i need to know? Aug 17 19:21:56 well, there is the manual Aug 17 19:22:04 but a lot of us learn by doing Aug 17 19:22:20 in your case, you need to figure out how you wouls manually build the kernel Aug 17 19:22:37 and adapt one of the linux-omap1 bb fiels to do what you need Aug 17 19:23:01 for this, I would start by identifying what source packages are needed for the build Aug 17 19:26:15 i should be able to copy the packages/linux/linux-nokia800-2.6.18-osso40 (what it's being built with now) to start Aug 17 20:15:33 RP: BTW: is bitbake creating .lock files? Aug 17 20:18:40 re Aug 17 20:18:58 wb Aug 17 20:19:04 any luck with the greenphone? Aug 17 20:22:01 zecke: getting a little bit closer: i have the rs232 ic now, but not yet time to connect it... Aug 17 20:22:34 :) Aug 17 20:23:22 if forking humans wouldn't take that long ;-) Aug 17 20:29:57 is that BB issue? http://pastebin.com/m4804f924 Aug 17 20:34:13 ynezz: looks like the runqueue fails if there are no elements in it Aug 17 20:34:25 but your real problem is the failed libtool-native compilation Aug 17 20:34:29 "ERROR: see log in /home/ynezz/data/devel/om2007.2/ep93xx/tmp/work/i686-linux/libtool-native-1.5.10-r6/temp/log.do_compile.18322" Aug 17 20:35:00 yes i know Aug 17 20:35:43 it's that strange error 'configure: error: source directory already configured; run "make distclean" there first' again Aug 17 20:36:47 i've deleted everything in tmp/work/i686-linux/libtool-native-1.5.10-r6 and rerun, but it says the same Aug 17 20:37:52 ynezz: also remove the build stamps from tmp/stamps/i686-linux or do bitbake -crebuild libtool-native? Aug 17 20:38:05 yes stamps also of course Aug 17 20:38:15 will try crebuild Aug 17 20:40:45 RP: that is something for you, I don't understand how in this case there are elements in build_targets that are not tuples: http://pastebin.com/m4804f924 Aug 17 20:41:45 same error, but now in do_compile Aug 17 20:44:37 http://pastebin.com/m4c093365 Aug 17 20:48:24 do_compile log http://pastebin.com/m3bf3c482 Aug 17 20:49:28 i've had same problem with other packages before, but deleting the tmp/work/ files and stamps solved it Aug 17 20:55:31 zecke: what do you mean? Aug 17 20:57:12 ljp: hey Aug 17 20:57:21 RP: http://paste.lisp.org/display/46308, r=? (without the print "Bla") Aug 17 21:28:34 * ynezz hides Aug 17 21:28:51 /home/ynezz/data is symlink to /media/data Aug 17 21:29:18 and that's why the weird test fails Aug 17 21:36:20 ynezz: oww :) good to see it's solved though Aug 17 21:44:29 when doing bitbake -c compile it's executing do_compile task right? Aug 17 21:45:00 where can i find defininition/source of that do_compile task? Aug 17 21:46:45 ynezz: base.bbclass, base_do_compile Aug 17 21:46:48 nm, found it, now looking for oe_runmake :p Aug 17 21:47:12 same file Aug 17 21:47:26 just want to take a look why it don't work with symlinks Aug 17 21:56:37 lol, that problem with libtool-native is in OE FAQ Aug 17 21:57:24 the process of creating the sparseimage for OS X, is this image expandable? Aug 17 21:57:40 that is, will it grow as more space is needed/ Aug 17 22:19:58 night Aug 17 22:20:35 zecke: gn Aug 17 22:22:29 hi, anyone around? Aug 17 22:22:49 Marex: sure Aug 17 22:22:58 hi flo_lap Aug 17 22:23:08 I got across a weird issue :( Aug 17 22:23:44 when I compile angstrom using OE and use thumb instead of arm instruction set, but use Opie as a gui, qpe segfaults Aug 17 22:24:25 (I have the following in my local.conf for thumb) Aug 17 22:24:32 ARM_INSTRUCTION_SET="thumb" and THUMB_INTERWORK="yes" Aug 17 22:24:52 but qpe just segfault both in qemu and on real hardware Aug 17 22:24:56 Marex: other applications work? Aug 17 22:25:09 flo_lap, yea, it boots, but opie doesnt start Aug 17 22:25:12 * flo_lap has no idea about opie Aug 17 22:25:47 I tried recompiling qpe with arm/xscale optimizations (I have pxa270 in palmld), but it dies on some opie package then Aug 17 22:25:55 flo_lap, hmm :( Aug 17 22:26:26 time to struggle on :) Aug 17 22:28:40 03freyther 07org.oe.dev * r5f324e98... 10/ (1 classes/sanity.bbclass): Aug 17 22:28:40 classes/sanity.bbclass: Check stty settings and complain when tostop is present Aug 17 22:28:40 tostop is known to make bitbake-1.8 processes hang. Make the user aware of the Aug 17 22:28:40 upcoming issue and propose a solution to this problem. Aug 17 22:58:13 zecke: Sanity patch looks good Aug 17 22:58:26 good night Aug 17 22:58:28 zecke: Are there any internal python functions we can use to trigger syscalls? Aug 17 22:58:50 zecke: and yes, bitbake creates the .lock files. It doesn't remove them since it could cause races :-( Aug 17 23:14:12 how can I tell what packages are either fetched as archives, fetches as a specific revision from SCM, or HEAD? The same build can fail in the latter case, and I need to make sure I have a SCM snapshot in the second case. Aug 17 23:14:27 s/fetches/fetched Aug 17 23:17:50 good nite Aug 17 23:25:10 03rpurdie * r909 10/ (2 files in 2 dirs): taskdata.py: Improve dump_data debugging code Aug 17 23:28:52 03rpurdie 07bitbake-1.8 * r910 10/lib/bb/ (build.py parse/parse_py/BBHandler.py): BBHandler.py/build.py: Rework add_task to be more efficient (6% speedup, 7% number of function calls reduction) Aug 17 23:31:33 03rpurdie * r911 10/ (4 files in 4 dirs): bb/__init.py: Sort digraph output to make builds more reproducible Aug 17 23:35:12 03rpurdie * r912 10/ (4 files in 4 dirs): data.py: Split expandKeys into two for loops to benefit from the expand_cache (12% speedup) Aug 17 23:41:49 03rpurdie * r913 10/ (4 files in 4 dirs): runqueue.py: Fix idepends handling to avoid dependency errors, improve debugging capability Aug 17 23:43:00 http://pastebin.com/m6a8b130e Aug 17 23:43:22 this BB patch allows me to build libtool-native Aug 17 23:44:27 nm Aug 17 23:45:38 ynezz: symlinks in paths are known to be bad Aug 17 23:45:54 ynezz: That might fix some issues but you'll hit others Aug 17 23:46:07 hm Aug 17 23:50:55 'night a;; Aug 17 23:50:57 all Aug 18 01:06:15 ynezz: should re-do that patch as part of sanity.class, and make it refuse to build unless the user gets rid of the symlink in the path Aug 18 01:06:44 that's a sane thing for sanity to check for. Aug 18 01:07:14 (I agree with RP that bitbake shouldn't try and correct it for you) Aug 18 01:30:56 halloo Aug 18 01:31:32 how can i make a recipe compile into a machine, not an arch specific package? Aug 18 01:32:36 its for a third party load of binary stuff Aug 18 01:33:14 yes i know that sucks ... :) Aug 18 01:34:38 actually, i worked it out... **** ENDING LOGGING AT Sat Aug 18 02:59:56 2007