**** BEGIN LOGGING AT Tue Sep 25 02:59:56 2007 Sep 25 06:44:01 morning. i see a python-psyco bitbake file python-psyco_1.5.bb but i cant find any provider with BB>> match providers python-p* Sep 25 06:44:21 how can i build the python psyco bb package? Sep 25 08:49:11 what can cause a defconfig option not to end up in the kernel .config after do_configure ? Sep 25 09:05:12 nud:copied the wrong file Sep 25 09:56:27 XorA: You should have something like this, but with a switch for higher than d6 dices: http://www.makezine.com/blog/archive/2007/09/electronic_batteryless_di.html Sep 25 10:09:35 hi Sep 25 10:15:54 hm.. had to remove unit tests from the build process because otherwise dpkg-buildpackage would (which needs to be run as root) fails .. and that's in current source release Sep 25 10:22:19 uhm.. i just compiled/installed monotone 0.36 and grabbed http://www.openembedded.org/snapshots/OE.mtn.bz2 .. when I try mtn --db=/stuff/OE.mtn pull monotone.openembedded.org org.openembedded.dev I get an error saying the database is laid out according to old schema Sep 25 10:22:39 can i somewhere get snapshot in the proper layout which will work with 0.36 ? Sep 25 10:24:38 ah.nevermind.. in the snapshot directory Sep 25 10:36:00 morning all Sep 25 10:37:42 morning Sep 25 10:39:39 morning gilligan_ , Richard Sep 25 10:40:12 * gilligan_ is taking his first steps with oe at the moment.. Sep 25 10:41:32 I wouldn't be lucky enough to have someone in this channel who is using OE for the compulab PXA270 ? :) Sep 25 10:43:13 I know at least two people using it Sep 25 10:44:47 gilligan_: I have done before but quite a while ago Sep 25 10:44:54 i see Sep 25 10:45:29 well the hardware specifics don't even matter that much at this point.. right now i've gotta say oe is still pretty much a black box to me :] Sep 25 10:47:05 and the documentation is a bit scattered and sometimes somewhat confusing.. but i'll get there eventually Sep 25 10:47:28 any of you guys using OE for projects at work? Sep 25 10:53:22 i suppose for a first test I could just stick to the "generic" distrib ? Sep 25 10:55:57 gilligan_: I would start with Angstrom Sep 25 10:58:11 cbrake_a1ay, out of any specific reason? Sep 25 10:59:04 cbrake_a1ay, because it seems to be much bigger which leads to longer builds and it includes stuff I don't need.. first goal is to just get something booted on the compulab board Sep 25 11:02:31 gilligan_: angstrom gets a lot more testing, and is what most people use, so there is a much higher chance it will work. Sep 25 11:02:43 gilligan_: start with something that works, and then optimize for size later Sep 25 11:03:51 gilligan_: angstrom-console-image currently takes about 5Gb of space on the host to build, which is reasonable. Sep 25 11:04:07 cbrake_a1ay, alright..thanks for the input Sep 25 11:04:52 so.. "dist=angstrom-2007" or what is it ? Sep 25 11:04:54 gilligan_: angstrom does not really set the image size, that is more a function of the image you build. There are many images you can build, but I would start with angstrom-console-image Sep 25 11:05:30 gilligan_: ls openembedded/conf/distro/angstrom-200* Sep 25 11:05:52 angstrom-2007.1.conf angstrom-2007.1-legacy.conf angstrom-2008.1.conf Sep 25 11:05:53 gilligan_: so, angstrom-2007.1 or angstrom-2008.1 Sep 25 11:06:04 alright Sep 25 11:06:13 thanks again Sep 25 11:06:23 gilligan_: I think either should work, but 2007.1 is considered somewhat stable, 2008.1 is somewhat testing Sep 25 11:06:33 alright Sep 25 11:13:42 we really need to remove the name angstrom from teh image names Sep 25 11:13:45 cbrake_a1ay, next dull question.. so now that I set up my local.conf (modified $PATH as desired etc..) what's the bitbake target that I first want to build now ? you mentioned `angstrom-console-image` which sounds like it could be a target :) don't see it in the angstrom conf file though Sep 25 11:13:50 it cause so much confusion Sep 25 11:14:07 packages/angstrom Sep 25 11:14:35 there is also angstrom-minimal-image Sep 25 11:14:44 and no good "in between" image :) Sep 25 11:14:53 ah.. ok Sep 25 11:14:54 ;] Sep 25 11:19:11 gilligan_, There are quite a few of us using it at work, I believe :-) Sep 25 11:19:11 $ bitbake angstrom-console-image <-- that should work shouldn't it? I get "ValueError: substring not found" in tmpHandler Sep 25 11:19:18 gilligan_: yes, angstrom-console-image is the target. Sep 25 11:19:34 maybe my packages path is wrong..hum Sep 25 11:19:53 gilligan_: does bitbake parse any packages when you run it? Sep 25 11:19:59 gilligan_: should take several minutes Sep 25 11:20:02 gilligan_: the first time Sep 25 11:20:14 gilligan@toontown:~/projects/ATM$ bitbake angstrom-console-image Sep 25 11:20:14 NOTE: Psyco JIT Compiler (http://psyco.sf.net) not available. Install it to increase performance. Sep 25 11:20:14 ERROR: no files to build. Sep 25 11:20:14 NOTE: Parsing finished. 0 cached, 0 parsed, 0 skipped, 0 masked. Sep 25 11:20:14 NOTE: build 200709251318: started Sep 25 11:20:16 nope Sep 25 11:20:37 echp $BBPATH Sep 25 11:20:41 er echo\ Sep 25 11:22:36 /home/gilligan/projects/ATM/stuff/build:/home/gilligan/projects/ATM/stuff/org.openembedded.dev Sep 25 11:22:40 that appears to be correct Sep 25 11:23:03 gilligan@toontown:~$ ls /home/gilligan/projects/ATM/stuff/build Sep 25 11:23:03 conf Sep 25 11:23:03 gilligan@toontown:~$ ls /home/gilligan/projects/ATM/stuff/org.openembedded.dev Sep 25 11:23:03 classes conf contrib COPYING.MIT delete.txt files MAINTAINERS _MTN packages README removal.txt site Sep 25 11:26:49 and BBFILES in local.conf is right? Sep 25 11:27:04 when this happens, I usually made a type in one fo those two settings Sep 25 11:27:38 no..because i am an idiot :] Sep 25 11:27:43 typo Sep 25 11:27:44 that too Sep 25 11:27:45 :) Sep 25 11:29:02 alright.. now that seems much better Sep 25 11:30:40 Crofton|home, what are your experiences with enabling parallel builds ? working or too many packages making problems ? Sep 25 11:31:19 hm..really wonder how long this build is going to take ;] Sep 25 11:31:20 how can i tell OE to create a ramdisk image instead of jffs2 image? Sep 25 11:31:32 PARALLEL_MAKE = "-j 4" Sep 25 11:31:34 is fine Sep 25 11:35:53 any German speakers handy? Sep 25 11:38:03 abschicken = send Sep 25 11:38:11 nik0n: IMAGE_FSTYPES -- see openembedded/conf/local.conf.sample Sep 25 11:38:30 this time I received a confirmation Sep 25 11:39:51 cbrake_alay so its possible to turn a ext2 into a ramdisk image with some modifications? Sep 25 11:49:11 hm..there we go.. build failed because qemu failed oe_runconf complaining that it does not want to be compiled with gcc 4.x Sep 25 11:50:07 sure as hell don't need qemu actually ;-] Sep 25 11:50:25 unless of course that is used for cross compilation Sep 25 11:53:04 gilligan_: add a gcc 3 toolchain to your system and it will use that Sep 25 11:54:32 will do Sep 25 11:58:12 gilligan_, look the ENABLE_BINARY_LOCALE in local.conf Sep 25 11:58:22 and disable, that stops qemu from building Sep 25 11:58:46 since I do not care about internationalization, this is not a problem for me Sep 25 11:58:57 ah yeah.. don't care either Sep 25 12:27:09 nik0n: it is probably easiest to start with the *.tar image if you want to hack it Sep 25 12:27:41 nik0n: but, it is best to make OE generate your image rather than rely on post ops Sep 25 12:35:04 The RH is fully booked Sep 25 12:35:05 help! Sep 25 12:45:13 zecke, ping Sep 25 12:46:35 Crofton|home: there was another hotel on same street I found using google maps Sep 25 12:46:48 Crofton|home: looked quite nice, but slightly more expensive Sep 25 12:46:49 ok Sep 25 12:46:55 Crofton|home: cant remember name though Sep 25 12:47:07 something russian I think Sep 25 12:47:14 I could also try hostel and see if it is possibel to join you for breakfast Sep 25 12:47:17 heh Sep 25 12:47:20 commies Sep 25 12:48:20 Crofton|home: pong Sep 25 12:48:35 zecke the RH and GH hotels are full Sep 25 12:48:40 know of any others to try? Sep 25 12:49:27 Crofton|home: Hotel chez zecke ;) Sep 25 12:49:36 Crofton|home: wait a second Sep 25 12:49:38 it may come to that Sep 25 12:49:41 :) Sep 25 12:49:44 I snore Sep 25 12:49:49 badly apparently Sep 25 12:50:14 Crofton|home: there is one more on the other side of the Invalidenstrasse and I have to search the name Sep 25 12:51:19 what number on invalidenstrasse Sep 25 12:51:48 Crofton|home: it is not there with the google maps search Sep 25 12:51:55 Crofton|home: I'm asking some one for the name Sep 25 12:54:18 Crofton|home: MERCURE HOTEL BERLIN AN DER CHARITE Sep 25 12:54:22 ok Sep 25 12:54:27 it is in google :) Sep 25 12:56:17 Crofton|home: if it is too expensive let me find another one Sep 25 12:56:29 let me run through google Sep 25 12:56:33 I have location now Sep 25 12:56:38 I will provide updates :) Sep 25 12:56:42 bbl Sep 25 12:57:14 ok Sep 25 12:57:16 urg Sep 25 12:57:29 rates would be grim in usd ... Sep 25 12:59:01 you guys are going to Berlin? Sep 25 12:59:05 kinderinsel is childrens hotel .... Sep 25 12:59:21 http://www.openembedded.org/wiki/OEDEM2007 Sep 25 12:59:43 http://www.hotel-zarenhof.de/ Sep 25 13:00:03 ah Sep 25 13:00:07 nice Sep 25 13:00:20 i won't go to Berlin for december.. for the CCC Congress :) Sep 25 13:00:44 Crofton|home: the google busines search is not too bad either Sep 25 13:00:54 Crofton|home: http://maps.google.com/maps?f=l&hl=en&geocode=&q=hotel&near=Novalisstra%C3%9Fe,+10115+Mitte,+Berlin,+Germany&ie=UTF8&ll=52.526927,13.395467&spn=0.006044,0.015213&z=16&om=1 Sep 25 13:01:02 zecke: do you already know if you are going to Linz? Sep 25 13:01:55 Jin^eLD: the talk was accepted Sep 25 13:02:08 cool, see you there then :> Sep 25 13:02:13 anyone else going? Sep 25 13:02:15 Jin^eLD: I better show up :) Sep 25 13:02:30 Jin^eLD: at least one guy of o-hand.com will present as well Sep 25 13:02:49 well anyone else from #oe I mean? Sep 25 13:02:55 zarenhof ok, but full some nights Sep 25 13:03:08 not necessarily presenting, just - if someone whom I could know from he is joining? :> Sep 25 13:04:01 Crofton|home: we find you a comfy cardboard box Sep 25 13:05:58 http://www.novalis-berlin.de/home.htm Sep 25 13:07:30 http://www.hotelclub.net/hotel.reservations/Best_Western_Kant_Hotel_Berlin.htm is a good hotel, not too expansive Sep 25 13:07:48 but don't know where that is w.r.t where you need to be Sep 25 13:07:55 going to see what Novalis qyotes Sep 25 13:08:45 why do they have a check box for booking or accomodation Sep 25 13:18:17 hm Sep 25 13:18:22 seems liek ftp.rpm.org is down Sep 25 13:18:42 from which oe tries to fetch ftp://ftp.rpm.org/pub/rpm/dist/rpm-4.1.x/popt-1.7.tar.gz Sep 25 13:21:05 any workaround to this? Sep 25 13:22:35 was easy enough to find a mirror using google Sep 25 13:22:35 hmmm Sep 25 13:22:41 try again until it succeed Sep 25 13:22:49 then do not remove your local sources .... Sep 25 13:23:10 well I am pretty sure it is down Sep 25 13:23:19 so i could retry till tomorrow Sep 25 13:23:59 or at least i am on the 4th retry or so already Sep 25 13:24:14 I had this issue the other day and I eventually got it Sep 25 13:24:25 I can email the stuff for your sources dir .... Sep 25 13:25:08 well i could just grab popt-1.7.tar.gz from elsewhere and put it somewhere, where OE finds it right ? Sep 25 13:26:10 give me your email and I can send it along with the md5 etc Sep 25 13:27:05 tobias.pflug@gmx.net Sep 25 13:28:03 going out now Sep 25 13:29:02 got it Sep 25 13:29:06 thanks Sep 25 13:29:47 where do I need to put those? also.. is it "safe" to cancel the build process now while it is endlessly retrying to fetch the file from the server? :] Sep 25 13:30:11 morning Sep 25 13:30:34 chouimat, morning Sep 25 13:31:02 gilligan_, put them in DL_DIR Sep 25 13:31:14 from local.conf Sep 25 13:31:18 right right,, Sep 25 13:31:39 so i can cancel the build now and restart when i copied the files ? Sep 25 13:31:46 yeah Sep 25 13:34:28 well.. thanks once again Sep 25 13:38:00 http://youtube.com/watch?v=IuwMOBM5xfU Sep 25 13:38:08 Ninja outbreak in Pittsburgh Sep 25 13:38:26 o_0 Sep 25 13:52:58 Is there a good way to override variables set in a mkefile using from a bb file? Sep 25 14:21:56 Crofton, can't reach http://ftp.info-zip.org/pub/infozip/src/unzip552.tar.gz either.. happen to have that floating around on your disk as well? :] Sep 25 14:23:09 or,well..actually i can just grab it myself from elsewhere..nevermind Sep 25 14:44:08 gm Sep 25 14:46:03 gm Sep 25 14:51:10 Where's the viewmtn site again? I need to find out what's broken the openmoko builds... Sep 25 14:51:38 Unless someone knows what work has been going on involving cairo Sep 25 14:59:00 cairo-directfb was added Sep 25 14:59:38 Yeah, that's probably the problem. Sep 25 14:59:48 Now the trick is to figure out how to resolve it. Sep 25 15:00:09 Looks like svolpe might be able to help? Sep 25 15:17:11 angstrom is quite big.. Sep 25 15:19:18 gilligan_: 892kB over here Sep 25 15:19:23 I wouldn't call that big Sep 25 15:20:36 well I don't know.. i'm building angstrom-console-image and that appears to contain a LOT of software per default Sep 25 15:22:13 gilligan_: you can create your own custom image based on the Angstrom distro -- Angstrom does not dictate what goes in the image Sep 25 15:22:45 gilligan_: but, I still recommend you get angstrom-console-image running first so you have a working point of reference Sep 25 15:23:27 cbrake, yeah i get the point.. what i basically meant was "the default selection is rather big" then ;] Sep 25 15:23:55 cbrake, well with the minimal space available on the board there is no way i can use the default setup Sep 25 15:24:42 gilligan_: gota run -- I'll be back in a couple hours ... Sep 25 15:24:45 cbrake, but i just wanted to successfully build a distrib before I start anything else anyway (even if I can't test it on the actual hardware) Sep 25 15:24:48 ok Sep 25 15:24:49 cu ;] Sep 25 15:25:14 gilligan please read on the oe-devel list Sep 25 15:25:23 there is a disscusion about it Sep 25 15:25:51 hi bernado Sep 25 15:29:07 woglinde, discussion on what exactly? (but you are right.. i should subscribe to the oe devel-list in either case..) Sep 25 15:29:32 discussion about to shrink your image Sep 25 15:29:54 and how oe can be made easier for normal user Sep 25 15:30:06 ah ok Sep 25 15:39:23 and what a normal user is :) Sep 25 15:39:35 * Crofton heads to school/lunch Sep 25 15:43:39 in local.conf I set IMAGE_FSTYPES to jffs2 and tar (or rather didn't touch those defaults) .. are the respective images also stored in the tmp dir ? Sep 25 15:44:38 ah Sep 25 15:44:39 there Sep 25 15:44:43 in deploy Sep 25 15:46:33 gotta say i'm rather impressed with openembedded Sep 25 15:50:26 hi woglinde Sep 25 15:50:38 * Bernardo was having some network problems Sep 25 15:52:30 /me wonders why u-boot git fetches and vuilds at home, but not at work .... Sep 25 15:52:52 a firewall at work? Sep 25 15:53:12 no Sep 25 15:53:14 :) Sep 25 15:53:33 i am working from home right now because of the idiotic firewall ;] Sep 25 15:53:54 talk about locking yourself in instead of keeping others out Sep 25 15:54:08 re crofton Sep 25 15:54:48 re Sep 25 15:56:27 crofton sure the it dont have a transparenrt fw? Sep 25 15:56:42 or maybe problem with transparent proxy Sep 25 15:56:45 git works Sep 25 15:56:58 I'll poke around some more Sep 25 16:47:15 I need confirmation on an OE-behavior thing from one of the wizards here! Sep 25 16:47:46 Assume that I have two recipes: pango-1.2.3.bb and pango-directfb-1.2.3.bb Sep 25 16:47:57 Something has a dependency on "pango" Sep 25 16:49:48 Both pango recipes build the same set of packages (but with different content) - is it true that the bitbake logic will select pango-directfb-1.2.3.bb because the string "directfb-1.2.3" compares greater than "1.2.3"? Sep 25 16:50:25 If so, why would "PREFERENCE = "-1" in pango-directfb-1.2.3.bb not change the results? Sep 25 16:51:10 (I could just delete pango-directfb-*.bb, but I think that would anger someone -- although it makes openmoko build just fine again) Sep 25 16:57:23 mwester hm ask koen now Sep 25 16:59:43 * koen hides? Sep 25 16:59:48 *g* Sep 25 17:00:23 it seems the intotruduce of gtk-directfb has some sideeffects Sep 25 17:00:34 pango-directfb_1.2.3.bb has "1.2.3" as PV Sep 25 17:01:07 it being selected was random before RP made some changes for PREFERRED_PROVIDERS Sep 25 17:02:56 * mwester contemplates duplicate PVs and PREFERENCE... Sep 25 17:03:14 mwester: the whole directfb thing needs more thought, please respond to the thread on the oe ml Sep 25 17:03:27 * koen will be away for work most of the time this week Sep 25 17:05:07 I shall locate said thread, and respond. :) Thanks - in the meantime, should we rename or do something with the problematic recipe? All new openmoko builds will fail. Sep 25 17:06:33 the two options I see: remove *irectfb or add PREFERRED_PROVIDERS for pango, cairo, etc to angstrom on top of the existing gtk ones Sep 25 17:06:54 the second option is the easiest, but basically rules out developing directfb with angstrom Sep 25 17:07:46 it's also the best option for the short term Sep 25 17:12:34 koen, off to do more fittings? Sep 25 17:13:34 service days Sep 25 17:13:45 so fixing the misfittings Sep 25 17:13:53 and the people that lost weight Sep 25 17:14:03 heh Sep 25 17:14:09 one person today lost 19kg the past year Sep 25 17:14:28 clean living Sep 25 17:14:35 I need to lose some weight Sep 25 17:14:44 * koen as well Sep 25 17:15:25 and the best way to do that is finish my thesis Sep 25 17:18:21 hi there. Sep 25 17:19:01 is there anyone how tried to build an image for wrt54? Sep 25 17:19:37 s/how/who Sep 25 17:27:55 bedboi: Not with OE AFAIK. Sep 25 17:28:34 btw the conf files are there Sep 25 17:28:48 now i don't know if these config files will work Sep 25 17:28:50 but they are there. Sep 25 17:28:58 For angstrom? or what distro? Sep 25 17:30:22 there is a wrt54 distro Sep 25 17:30:33 well actually wrt54oe Sep 25 17:38:27 hrw uses that to create packages for his wrt, but not images Sep 25 17:38:46 we need someone with a serial lead attached to a wrt to get OE wrt support into shape Sep 25 17:44:21 koen: btw at least we can create packages Sep 25 18:07:58 bedboi: please work on adding openwrt kernels to OE, that would be very nice Sep 25 18:28:38 anyone know anything about NIOS processors? Sep 25 18:40:00 03koen 07org.oe.dev * r57d7542e... 10/ (17 files in 8 dirs): linux-ezx: catch up with svn and unbreak booting by enabling ext2 Sep 25 19:07:54 koen: how do you get back to holland after OEDEM? Sep 25 19:15:36 03jluebbe 07org.oe.dev * r881b2437... 10/ (3 files in 3 dirs): openmoko-base: use startup_unintrusive.wav instread of startup_openmoko.wav Sep 25 19:21:26 03jluebbe 07org.oe.dev * r13a65151... 10/ (4 files in 4 dirs): Sep 25 19:21:26 base-files: mount mmc card async Sep 25 19:21:26 * it is not possible to remove the mmc card from the neo while it is on Sep 25 19:21:26 so we can safely mount the card async Sep 25 19:54:46 hello all Sep 25 19:58:21 hey likewise Sep 25 19:58:33 Jin^eLD: got my board :-) Sep 25 19:58:37 finally :) Sep 25 19:59:27 which SDK version came along with it? Sep 25 20:12:16 Jin^eLD: May 2006? Sep 25 20:12:37 uhm.. actually it should have a version number :) Sep 25 20:13:01 Jin^eLD: actually, I don't think I have the SDK, it's an eval board for lend.... Sep 25 20:14:32 likewise: uhm.. but how would you boot the board and try the decoders etc. then? Sep 25 20:14:49 Jin^eLD: I guess there are binaries on the HDD Sep 25 20:15:15 Any ALIX users here yet? Sep 25 20:16:15 likewise: uhm, that's weird... Sep 25 20:16:32 I'd at least expect some board support package to play with Sep 25 20:16:57 Jin^eLD: we didn't pay for anything yet, the SDK is 25k we heard. Sep 25 20:17:29 ooh.. I thought the board was 25k Sep 25 20:28:29 * CM learned some more about timers today. :D Sep 25 20:30:50 CM: POSIX timers? Sep 25 20:31:20 likewise: I read cbrake's blogpost :) Sep 25 20:31:34 ah, let me see Sep 25 20:32:08 likewise: http://bec-systems.com/web/content/view/78/9/ Sep 25 20:32:57 ah yes Sep 25 20:33:05 Basic, but that's what I needed. Sep 25 20:33:08 I can show you an example implementation as an extra Sep 25 20:33:19 Sure :) Sep 25 20:35:24 CM, cbrake: http://www.pastebin.ca/713962 Sep 25 20:35:42 contains timers.c source and Makefile in the end Sep 25 20:36:26 likewise: Ah, neat. Thanks Sep 25 20:36:32 CM: clock_id = CLOCK_REALTIME; can be set to CLOCK_MONOTONE if you need absolute timeouts (see cbrake's blog on why) Sep 25 20:37:24 Why would there be any advantage to use CLOCK_REALTIME if you need a timeout interval? Sep 25 20:37:43 If I use it for a wake-up alarm, I want realtime though. Sep 25 20:38:13 CM: none. This is some old source I used to verify what happens in all cases. Sep 25 20:38:56 kergoth: I'm so in love with git-rebase Sep 25 20:39:07 CM: yes, sure, you can add real-time thread scheduling etc. Sep 25 20:39:21 hehe Sep 25 20:39:25 rebase is very nice Sep 25 20:39:39 CM: the timer triggers the thread, which you can make real-time. Sep 25 20:39:40 i like stgit too, convenient rather than rearranging the history manually Sep 25 20:39:55 thanks to reflog, its like quilt but with patch modification history Sep 25 20:39:58 kergoth: I had more issues with it than pleasure :) Sep 25 20:40:00 likewise: I see that. Nice :) Sep 25 20:40:15 kergoth: git-rebase -i and git-mergetool just rock Sep 25 20:40:32 just need to make sure not to git gc on a repo using an stgit branch, or itll throw away your old reflog entries eventually, would lose patch mod history Sep 25 20:40:42 lately i'm loving git bisect Sep 25 20:40:52 kergoth: you do customer support right? Sep 25 20:41:09 not phone support, but it's support yeah, fixing the bugs a specific customer reports Sep 25 20:41:38 * cbrake is liking git-rebase a lot as well :-) Sep 25 20:41:51 kergoth: see, I decided two things in my life. Not to fix other ones bugs and not to introduce bugs myself :) Sep 25 20:41:54 kergoth: so stgit adds enough value to mess with? Sep 25 20:41:59 bisect helps a lot for what i'm doing, since i can quickly search upstream for where a fix happened that i need to backport to one of our kernels Sep 25 20:42:02 zecke: hehe Sep 25 20:42:44 cbrake: imo it does. it makes pushing the patch series into your main branch and then upstream quick and easy, and like i said lets you have history for the modifications of the "patches" Sep 25 20:42:49 OEDEM item: move to GIT? Sep 25 20:43:13 likewise: psst Sep 25 20:43:29 * zecke sits on git-fast-import to support bidirectional git<->mtn Sep 25 20:43:30 zecke: pssst or sssshhh? :-) Sep 25 20:43:52 hey all, can somebody point me where to look for power management issues for ARM based devices (like Sony PRS 500, Nokia 770)? Sep 25 20:44:19 likewise: I think we can just stay with mtn, and people wanting to use git, hg, darcs can do that Sep 25 20:44:32 likewise: thanks for the timer code -- grokking now Sep 25 20:44:35 * kergoth uses git-svn a lot for interacting with upstream svn's Sep 25 20:44:36 zecke: as long as the repo's scale along Sep 25 20:44:51 likewise: and with a working bridge it doesn't really matter Sep 25 20:44:55 cbrake: the timer code defaults to REALTIME, it is something I was playing with long ago Sep 25 20:45:54 cbrake: just define the variable at the the top to use MONOTONE. In fact, I used this code base to make the timer triggered thread real-time for a real-time application I was developing. Sep 25 20:47:01 likewise: e.g. I want to use git to try merging .dreambox to .dev Sep 25 21:00:52 likewise: I looked at signal handers early in my project, but decided to use a while loop in a separate thread with clock_nanosleep. Sep 25 21:01:39 likewise: I was curious if a signal handler can call a function before the previous call had exited Sep 25 21:02:46 likewise: but never tested. I ran into one other issue with the signal handler, but can't recall what it was ... Sep 25 21:03:27 cbrake: clock_nanosleep() is perfectly fine, simple and does the job. Sep 25 21:04:31 likewise: must have been I was just uncomfortable putting a huge amount of code in a signal handler ... Sep 25 21:05:34 cbrake: oh yes, I only perform mutex_unlock()s to awaken another thread. clock_nanosleep() is much cheaper I suspect. Sep 25 21:06:05 cbrake: I think at that time I needed to couple a real-time process to a real-time wallclock (which came across a broadcast video link). Sep 25 21:06:48 cbrake: so I was experimenting with all rare casing off the real-time clock (which was non-monotone in case of loss of video or leap seconds etc) Sep 25 21:06:53 s/casing/cases/ Sep 25 21:07:05 likewise: nod, that makes sense Sep 25 21:08:15 cbrake: one of the rare cases is where the video feed has a leap second, and my system has a leap second, but not quite at the same time. Sep 25 21:15:45 hmm, freescale uses ltib to build Linux images: http://www.freescale.com/webapp/sps/site/overview.jsp?nodeId=02VS0l320822D022DB Sep 25 21:20:18 I think stelios is talking to them about OE :) Sep 25 21:20:23 bahh, written in perl, and depends on OE Sep 25 21:20:26 * cbrake is not impressed Sep 25 21:20:28 he is giving a talk at the PPC dev conference Sep 25 21:21:32 Crofton: ltib source tree is only 10MB -- can't be much there ... Sep 25 21:21:38 heh Sep 25 21:22:08 gahh, make that "bahh, written in perl, and depends on _RPM_" Sep 25 21:22:42 I hope stelios shows up and lets us know how his talk went Sep 25 21:22:58 Crofton: nod Sep 25 21:25:13 Crofton: I sometimes send processor/board vendors the following like and ask them if they feel like they are missing out :-) http://www.openembedded.org/filebrowser/org.openembedded.dev/conf/machine Sep 25 21:26:02 what just need to make sure they all work :) Sep 25 21:26:17 Crofton: yeah, that's a challenge Sep 25 21:28:18 * RP sighs at #3066 Sep 25 21:29:17 SRCREV? Sep 25 21:30:31 Crofton: yes Sep 25 21:31:04 a) its not a bitbake bug b) it was a design decision discussed on the mailing list Sep 25 21:31:56 we need a FAQ Sep 25 21:32:04 this will come up again .... Sep 25 21:32:14 yes Sep 25 21:34:47 I suppose you should close the bug with an explanation of how to get the desired behavior and reference the list discussion .... Sep 25 21:34:57 then I can read up on the solution :) Sep 25 21:35:06 Crofton: LOL :-) Sep 25 21:35:12 Crofton: (and me with you) Sep 25 21:35:32 brain death is setting in Sep 25 21:35:45 nite is setting in. good nite all, cu Sep 25 21:35:46 I need to move thesis edits from paper to disk, but no energy for that Sep 25 21:35:49 gn Sep 25 21:35:58 good luck Sep 25 21:37:10 I'm trying to build my own distribution with a TARGET_OS of linux-uclibcgnueabi Sep 25 21:37:27 why not use Angstrom? Sep 25 21:37:44 when it gets to gcc-cross-initial, it fails to compile because sys/types.h is missing Sep 25 21:38:04 Angstrom should be able to use uclibc Sep 25 21:38:12 I'm not sure how to use angstrom correctly, and oe doesn't have good documentation Sep 25 21:38:17 I am not sure how to make i tdo that Sep 25 21:38:18 hehe Sep 25 21:38:20 I've built things with my own distro before Sep 25 21:38:36 well, koen worked out how to do it with uclobc and Angstrom Sep 25 21:38:54 it seems as though uclibc-initial may need to be installed prior to gcc-cross-initial being built Sep 25 21:39:11 but when staging, uclibc-initial needed gcc... Sep 25 21:39:17 kinda circular Sep 25 21:44:04 anyone know where sys/types.h comes from? Sep 25 21:44:25 Chase: libc's headers Sep 25 21:44:35 yeah... Sep 25 21:44:35 Chase: be that uclibc or glibc Sep 25 21:44:59 when building gcc-cross-initial, should it be using my system's sys/types? Sep 25 21:45:21 Chase: Since you're running gcc on that system, yes Sep 25 21:45:36 hmm, then why wouldn't it be looking there? Sep 25 21:45:40 because I have it Sep 25 21:46:39 I seem to remember glibc-initial installing headers for some cases... Sep 25 21:46:58 gah! 6-12 weeks for an MX27ADS board. anybody have a good substitution recommendation? Sep 25 21:47:29 ... suppose i could look at i.mx[23]1 boards. commercial vendors? Sep 25 21:49:19 it wouldn't be in linux-libc-headers would it? Sep 25 21:49:36 maybe I'm just getting all confused Sep 25 21:49:45 I didn't see anything from glibc Sep 25 21:50:54 one thing that bothers me is I don't know what this is doing: Sep 25 21:51:02 "${@['virtual/${TARGET_PREFIX}libc-initial',''][bb.data.getVar('TARGET_ARCH', d, 1) in ['arm', 'armeb', 'mips', 'mipsel']]}" Sep 25 21:52:09 HopsNBarley: the LogicPD i.MX31 board Sep 25 21:52:25 * HopsNBarley googles... Sep 25 21:53:41 Chase: that is Python code -- if that helps Sep 25 21:55:42 Chase: are you building for ARM? Sep 25 21:55:48 yes Sep 25 21:55:52 ep93xx Sep 25 21:56:04 linux-uclibcgnueabi Sep 25 21:56:10 Chase: I'll try a build tomorrow and see what happens Sep 25 21:56:38 alright Sep 25 21:56:39 thanks Sep 25 21:56:49 Chase: I typically use Angstrom, and then set a var in local.conf to specify uclibc ... Sep 25 21:56:52 Chase: looking up now ... Sep 25 21:57:07 how does one "use" angstrom Sep 25 21:57:19 Chase: ANGSTROM_MODE = "uclibc" Sep 25 21:57:25 I want to build a very small image Sep 25 21:57:32 one that hopefully will fit in 16 MB Sep 25 21:57:37 but doesn't HAVE to Sep 25 21:57:45 that will act as a router Sep 25 21:58:00 Chase: DISTRO="angstrom-2007.1" in local.conf Sep 25 21:58:05 and I was wary of angstrom cause I couldn't tell what packages it uses Sep 25 21:58:13 or how to define what packages it uses Sep 25 21:58:20 Chase: I would use Angstrom, and then bitbake angstrom-minimal-image Sep 25 21:58:30 ok Sep 25 21:58:35 I'll look into that Sep 25 21:58:43 Chase: Angstrom does not dictate image contents -- more just versions, and such Sep 25 21:59:03 that was my other concern Sep 25 21:59:03 Chase: so you will benefit from package versions being set to things that we know work Sep 25 21:59:28 I will be using rt2x00 drivers Sep 25 21:59:35 and I need a fairly up to date kernel Sep 25 21:59:53 they use stuff in the LATEST kernel Sep 25 21:59:54 Chase: should not be a problem -- you can define a kernel recipe easily to build whatever you want Sep 25 21:59:57 as far as I can tell Sep 25 22:00:03 ok Sep 25 22:00:09 Chase: I'm build 2.6.23-rc6 for compulab modules right now in OE Sep 25 22:00:19 Chase: gotta run, catch you later Sep 25 22:00:22 alright Sep 25 22:00:25 thanks for your help Sep 25 22:00:31 Chase: good luck :-) Sep 25 22:43:51 has anybody thought about a VMware target? Sep 25 23:53:12 mwester, I removed pango-directfb until I figure out how to handle the issue you posted.. I replied back to your post explaining the situation in more detail :-) Sep 25 23:56:36 mwester, what I don't get is if I just do bitbake pango it builds pango and NOT pango-directfb. Sep 26 00:10:11 svolpe: that's because bitbake's selection is more or less random Sep 26 00:10:46 psokolovsky, any suggestion on how I should fix the issue other than deleting the package :-) Sep 26 00:13:23 svolpe: I posted my thoughts to oe ML. fix should be in bitbake. For now, apart from deleting, commenting out RPROVIDES would do. Or stuffing bunch of PREFERRED_PROVIDERs somewhere deep in config, like in bitbake.conf Sep 26 00:14:39 the conflicting RPROVIDES should be removed IMHO Sep 26 00:14:51 (until the issue is solved) Sep 26 00:15:25 psokolovsky, could I set the pango-directfb RPROVIDES="pango-directfb" would that stop pango-directfb from being built in place of pang? Sep 26 00:15:48 rwhitby, I removed pango-directfb already until I get the issue resolved. Sep 26 00:16:31 I did test it on my system bitbake pango and bitbake pango-directfb before commiting and all worked as expected but I guess there is something in package selection I'm missing. Sep 26 00:16:32 svolpe: sure, it will Sep 26 00:17:02 psokolovsky, that seems like the best solution for now, whats your thoughts? Sep 26 00:17:20 of course, that will "break" package management boon it offers, but the aim is to stabilize build of course Sep 26 00:17:25 svolpe: yep Sep 26 00:18:14 psokolovsky, not really because anything that needs it calls pango-directfb as the dependancy, or is there something I'm missing? Sep 26 00:19:04 svolpe: well, I hope you understand the whole magic behind RPROVIDES: Sep 26 00:19:35 once you have it done right, you can have either gtk-x11 or gtk-directfb system, Sep 26 00:20:19 but application package can be installed on any of them. the same application package. Sep 26 00:20:48 psokolovsky, I obviously do NOT understand it or I would not have botched everything up already :-) I guess I will try to figure it out better. Sep 26 00:21:36 well, it's like interface and class in OOP Sep 26 00:21:53 RPROVIDES is interface, package name is class Sep 26 00:22:37 if you have two packages doing the same thing, but in different ways, you want to define the same RPROVIDES for them, to enjoy polymorphism Sep 26 00:23:00 the current catch - bitbake handles that lousily Sep 26 00:24:01 psokolovsky has just identified something very key to OE -- its behavior is probably more understandable to the set of people who thing in OOP terms! Sep 26 00:25:05 svolpe: Just to make sure, my email was expressing my concerns more about bitbake and OE rather than your work with pango-directfb -- it seems quite logical to me that your approach should have worked without destabilizing anything. Sep 26 00:25:31 mwester, that is how I took it. Sep 26 00:25:35 I don't think it's particular to OE ;-) Sep 26 00:26:12 Well, the one item -- resistance to branches -- is universal :) Sep 26 00:27:56 I have some experiance with OOP although I have spent more time just using C since I have done more low level programming and hardware design :-) Sep 26 00:28:30 I confess that I started learning assembler, and finished my formal education when Pascal and PL/1 were the rage -- perhaps that hinders my understanding and acceptance of the apparent non-determinism in OE :) Sep 26 00:29:26 mwester: we just don't have enough resources to maintain such branches. on the merge, you'll get all the breakage at once, that's the only difference ;-). imho. Sep 26 00:29:26 I'm a hardware design engineer but I got hooked on Linux and now embedded Linux so things have changed :-) Sep 26 00:30:13 psokolovsky: That's exactly the same situation that every commercial user of SCM tools faces, and most of them have found that the "all at once" is more manageable and tolerable. Sep 26 00:31:23 ewwwww Sep 26 00:31:34 I think the problem is that in a formal group, there is an individual who is paid on results. That person is able to dictate who will do what, and it often is not "fair" -- one developer (or group) will pay the price of doing the merge, while another group reaps the benefit. Sep 26 00:31:41 psokolovsky, so really RPROVIDES won't fix the problem becuase it just says this package is a replacement (equiv) to the the interface (package) it lists. So RPROVIDES="pango" means, even thought not named it" this package provides a pango equiv. Sep 26 00:31:46 That doesn't work in open-source. Sep 26 00:32:07 Hence we work in an environment that pretty much "limps along" in terms of stability. Sep 26 00:32:55 Perhaps I need to respond to that OE thread... Sep 26 00:33:26 I'll shut up now. I got it to build, and it's not too late to order more chicken entrails and incense for the next build ;-) Sep 26 00:33:32 mwester: well, other projects are different. I speak based on previous OE experience. of OE branches I know, dreambox one is not being merged back, OZ used to be merged somehow, Familiar was not merged and died... But if someone thinks he can break the trend, he can try ;-) Sep 26 00:35:07 svolpe: yes, exactly. so, bitbake takes pangs-dfb as possible candidate when building just pango, and sometimes randomly selects it. that randomness is bug. the way to work it around for now is to give up the same RPROVIDES as in pango for now. Sep 26 00:35:40 psokolovsky: I can't make others use branches. But I can promise that if I knowingly (this pango-directfb problem is not such a situation) embark on a project that has the potential to destabilize others, I will branch, and I will merge back. I feel it is common courtesy to the other developers. JMO. Sep 26 00:35:53 mwester: anyway, maybe angstrom stable branch will change it, if/when it will be created Sep 26 00:36:47 We'll see. I also admit to some skepticism about monotone and branching, but that's another topic. Sep 26 00:37:10 mwester, would you have time to repeat the issue if I fix the packages.. I would like to see if it DOES fix the issue. Sep 26 00:38:09 Of course! Sep 26 00:38:16 psokolovsky, mwester, thanks for the help. Sep 26 00:38:31 np Sep 26 00:39:08 It's actually easy to test -- if you move aside the live tmp directory, and issue a bitbake command with the "dry-run" option, and the option to generate the dependency list, you can just examine the *.dot files to see what bitbake and oe came up with. Sep 26 00:48:00 mwester, pushing the changes now so when you get a change please test them.. Sep 26 01:28:02 svolpe: If you read the log, the changes did not work - RPROVIDES seems to add additional information, but doesnt' change bitbake's conviction that pango-directfb is the best provider of pango. :( Sep 26 02:10:51 somebody seen CoreDump|afk around? Sep 26 02:11:14 Not for a long time. Sep 26 02:11:22 ~seen CoreDump|afk Sep 26 02:11:24 coredump|afk is currently on #oe (18h 26m 6s) #openmoko (18h 26m 6s) #openzaurus (18h 26m 6s), last said: 'SpeedEvil: I'll probably dig out the scope today. If you need anything that is not on the wiki page, post it in this channel, addressed to me. I'll know'. Sep 26 02:12:02 mwester: he's always idle because of his logs. I'm trying to reach him but he doesn't answer Sep 26 02:12:32 I need some help with oebug 3007 and I herad he was the person to contact about it since he made altboot Sep 26 02:13:14 s/herad/heard **** ENDING LOGGING AT Wed Sep 26 02:59:56 2007