**** BEGIN LOGGING AT Thu Mar 10 02:59:57 2011 Mar 10 03:01:22 tyco, the stable branch is getting a little old now Mar 10 03:01:41 I suggest you look at the .dev or master branch Mar 10 03:05:21 thanks grg Mar 10 04:03:02 Tartarus: another problem with it: echo dch $dist $urgency $version -r "dummy" Mar 10 04:03:05 grr Mar 10 04:03:17 Tartarus: http://paste.debian.net/110217/ Mar 10 04:11:19 Tartarus: the fix has introduced a dependency loop Mar 10 04:11:44 what? Mar 10 04:11:53 Tartarus: check the paste Mar 10 04:13:58 * Tartarus gives up Mar 10 04:14:14 03Tom Rini  07master * r6358b23b4c 10openembedded.git/recipes/glib-2.0/ (6 files): Mar 10 04:14:14 Back out recent glib-2.0 changes Mar 10 04:14:14 Signed-off-by: Tom Rini Mar 10 04:14:43 Tartarus: sorry by pestering you but it was broken Mar 10 04:15:33 yeah, no, you're right to do so Mar 10 04:15:40 I'm just not having a great day it seems Mar 10 04:16:01 * otavio tests the building again Mar 10 04:16:53 Tartarus: seems to work. Mar 10 04:17:36 Tartarus: I will remove tmp and do a full rebuild once again to be sure. Mar 10 04:25:55 Tartarus: don't blame yourself. You've done many nice cleanups and this is the first one that I spot something broken Mar 10 04:26:09 Tartarus: so it is not a bad day et all Mar 10 07:06:31 hi, what's sane way to avoid building kernel in oe? Mar 10 07:06:55 I have no kernel recipe for my machine, and don't want to mess with it right now Mar 10 07:18:04 anarsoul: you need to provide virtual/kernel I guess Mar 10 07:18:30 khem: added virtual/kernel to ASSUME_PROVIDED, it still pulls linux Mar 10 08:44:06 hi florian Mar 10 09:19:21 I'm getting a warning "Note consider defining a PREFERRED_PROVIDER entry to match java2-runtime" even when I specify PREFERRED_PROVIDER_java2-runtime = "openjdk-6-jre", Am I missing something Mar 10 09:20:52 pixtmux hm Mar 10 09:22:45 I get the same message for gdbserver "NOTE: consider defining a PREFERRED_PROVIDER entry to match gdbserver" Mar 10 09:23:42 hm but it should just build Mar 10 09:25:59 PREFERRED_PROVIDER_gdbserver = "gdb" && bitbake gdb doesn't build gdb but gdbserver though... Mar 10 09:26:47 It is working for a bunch of other recipes, but not for these ones. Mar 10 09:30:04 it should build both Mar 10 09:35:19 03Koen Kooi  07org.openembedded.dev * r4cafb38524 10openembedded.git/recipes/linux-firmware/linux-firmware_git.bb: Mar 10 09:35:19 linux-firmware: bump SRCREV Mar 10 09:35:19 This includes, among other things, update wl12xx firmware Mar 10 09:35:19 Signed-off-by: Koen Kooi Mar 10 09:41:11 moin Mar 10 10:07:22 good morning Mar 10 10:12:05 hi mckoan Mar 10 10:19:44 03Koen Kooi  07org.openembedded.dev * radfc8b4896 10openembedded.git/recipes/kinect/libfreenect_git.bb: Mar 10 10:19:44 libfreenect: add git version Mar 10 10:19:44 Signed-off-by: Koen Kooi Mar 10 10:19:55 03Koen Kooi  07org.openembedded.dev * r40a850649a 10openembedded.git/recipes/xmms/xmms_1.2.10.bb: Mar 10 10:19:55 xmms: bandaid some more host path leakage Mar 10 10:19:55 Signed-off-by: Koen Kooi Mar 10 11:47:06 Any recipes with scripts/utils to handle KEY_POWER events? Mar 10 11:52:33 cd oe; git grep KEY_POWER Mar 10 12:00:33 ynezz, that's a good start of course, but it could that the work KEY_POWER is not in the recipe but in repo/package it fetches? acpid is maybe what I want though. Mar 10 12:31:53 hi woglinde, all Mar 10 12:32:08 hi pb Mar 10 12:32:11 hi pb__ Mar 10 12:36:29 gm Mar 10 12:38:11 woglinde_, hi Mar 10 12:38:38 does someone knows very well shell scripting, here's my problem: Mar 10 12:39:14 if [ -e ${S}/bin/*.in ] doesn't work obviously Mar 10 12:39:33 so I wondered what was the *best way* to achieve that Mar 10 12:39:47 there are some way but they are hackish Mar 10 12:40:05 such as comparing ${S}/bin/*.in to '${S}/bin/*.in' expanding the S in the second part Mar 10 12:40:07 for instance Mar 10 12:40:29 if * = '*' it generally means that the dir is empty Mar 10 12:40:41 it's for autoconf.inc Mar 10 12:41:02 autoconf.inc doesn't check if Makefile.am and *.in exists before doing a sed Mar 10 12:41:12 so with autoconf213 it fails Mar 10 12:41:53 eFfeM, hi Mar 10 12:45:03 I quite don't get what do you want to achieve Mar 10 12:45:10 test if the file exist? Mar 10 12:45:24 thest if there is at least one file in *.in Mar 10 12:45:25 I think he wants to test if any file named '*.in' exists Mar 10 12:45:37 presumably the simple '-e' thing fails if *.in expands to multiple filenames Mar 10 12:45:39 else it fails to sed Mar 10 12:45:47 indeed Mar 10 12:45:56 (and, given lack of quotes, probably also fails if it expands to a filename with a space in) Mar 10 12:46:04 GNUtoo|laptop: Mar 10 12:46:27 for i in $S/*; do something ; done Mar 10 12:46:49 or if you are only interested in whether there is a .in file Mar 10 12:47:03 eFfeM, I wondered if there was something simplier, but if there is no simplier thing I'll use that Mar 10 12:47:05 ls $S/*.in | wc -l Mar 10 12:47:08 and compare that to 0 Mar 10 12:47:10 or 1 Mar 10 12:47:12 ahhh Mar 10 12:47:14 good idea Mar 10 12:47:15 better Mar 10 12:47:16 or find Mar 10 12:47:31 the wc -l is great Mar 10 12:47:51 thanks a lot Mar 10 12:48:06 yw Mar 10 12:48:07 or ls ${S} | grep -q '\.in$' Mar 10 12:48:50 which would avoid an error from ls if there aren't any files, and gives you the exit code you want (I think) Mar 10 12:48:53 wc `ls $S/*.in` Mar 10 12:49:00 wc -l ofc Mar 10 12:49:36 true didn't think of the case nu file exists Mar 10 12:50:05 "ls ${S}/*.in" might also be prone to do weird things if ${S}/foo.in is a directory :-} Mar 10 12:50:20 not that this is very likely though Mar 10 12:50:26 hmmm in ls $S/*.in | wc -l let's say *.in doesn't exist, will it cause some bitbake problem, as it will return 0 and also raise an error Mar 10 12:50:35 if you want to do ls and $S exists you can also do ls -d $S/*.in . | wc -l and subtract one for the . Mar 10 12:50:40 GNUtoo|laptop: tre Mar 10 12:50:42 true Mar 10 12:51:19 pb__ 's solution is more robust and so would a find based solution as suggested by ynezz Mar 10 12:51:26 ok Mar 10 12:51:29 I'll look Mar 10 12:58:17 hi Mar 10 12:58:36 hi Mar 10 13:18:25 Hi. I'm making a class file and I want to use a helper function. Is there any special trick to make this helper function available, so it can be used by other functions of the class? I keep getting "NameError: global name 'helper_function' is not defined". Mar 10 13:19:40 BTW, I'm declaring it as "python helper_function () { ... }" Mar 10 13:36:02 It seems that declaring the helper function like "def helper_function(d):" does the trick... Mar 10 13:44:22 maybe is an old problem, but anybody knows more about help2man_1.38.2.bb do_configure failed? Mar 10 13:46:48 Tartarus: hi, help2man_1.38.2.bb do_configure failed : any hint? http://pastebin.com/ecitq02k Mar 10 13:48:12 mario-goulart: use: def helper_function () { ... } Mar 10 13:48:43 otavio: read some lines above. :-) Mar 10 13:48:58 mario-goulart: ah heh Mar 10 14:19:24 * kergoth_ thinks the two syntaxes for defining python functions need to be consolidated, add support for arguments to the other syntax.. python foo(alpha, beta) {} or so.. Mar 10 14:21:49 kergoth_: out of curiosity: has using a single language ever been considered? (e.g., e2, which uses lua) Mar 10 14:22:25 I wouldn't be opposed to it, personally, i quite like e2's use of lua, and like lua in general, but I don't see us changing file formats anytime soon, we're heavily invested in our metadata Mar 10 14:22:27 IMO, having a mix of shell, python and a custom configuration laguage is too much confusion for a single brain. Mar 10 14:22:48 I see and fully understand your point. Mar 10 14:22:50 well, I don't think what we have is much different than make + python. make too is a mostly declarative format with shell for scripting Mar 10 14:23:40 I do think it'd be nice to consider giving bitbake the capability to use python plugins, find them in the bbpath, and us provide them as a way of altering the build process without having so much python in the classes and recipes themselves Mar 10 14:33:36 From a user perspective, I find the infrastructure confusing (I don't want to sound harsh nor demotivating -- I'm just giving some feedback as a user). There are too many tricks you have to be aware to make a proper recipe or class. A lot of things which are intended to provide convenience end up being too confusing. Mar 10 14:42:53 I agree with you. We have a shitload of metadata and it's scattered all over, and it's not clear to the user what's there, what they can set, where it call comes from, or how it fits together Mar 10 14:43:25 Yeah, that's exactly my impression. Mar 10 14:44:52 I think increasing discoverability would help a lot. if i could sit down and say okay, what variables can i set, and what values are available for them, what images can i build, etc, it'd go a long way. and of course making it easy to inspect the metadata in a way that shows what came from where and why.. but the file format should be improved regardless, i agree, but it'll likely be a slow move. we'll create something, and they'll both Mar 10 14:45:38 cooexist for a long, long time :) Mar 10 14:45:38 of course, the biggest reason our file format is fairly declarative and is directly parsed rather than something like shell, which is what portage used, was so we could programmatically translate it to another format in the future Mar 10 14:45:38 unfortunately, it's not so simple Mar 10 14:45:46 much of our metadata relies upon implementation details of the format and bitbake itself Mar 10 14:45:53 a direct translation might result in something that doesn't even function Mar 10 14:46:51 kergoth_: hey Mar 10 14:47:10 it seems to me bitbake master is slower than before since some days Mar 10 14:47:41 apparently at the end of the task...after rm_works takes some more time Mar 10 14:48:01 bisect :) Mar 10 14:48:19 is nmaybe som new checksum in the code? Mar 10 14:48:40 I'll translate in english... Mar 10 14:48:50 are there new checks Mar 10 14:48:53 ? Mar 10 14:50:46 the checksumming code only affects the up front parse time, as the information is cached and made available to the tasks. doesn't affect the task execution code. and of course, for oe it doesn't affect anything at all, since we aren't using a checksumming class that actually does anything yet Mar 10 14:51:32 ah, I see, I've just read some references to checksums : now I understand Mar 10 14:51:53 I'll take some timings with older git tags Mar 10 15:15:38 morning kergoth Mar 10 15:15:41 * pb__ stabs siemens Mar 10 15:17:09 pb__, customer or just in general? ;) Mar 10 15:19:23 Tartarus: apparently the glass door on our oven/microwave combi has "spontaneously" shattered Mar 10 15:19:34 my wife denies having anything to do with it :-) Mar 10 15:20:10 luckily though I have just found the spare part in their online shop and it isn't nearly as expensive as I feared it might be. Mar 10 15:20:36 Ah, owch Mar 10 15:20:45 Under warranty or anything still? Mar 10 15:21:16 I've had luck, but usually smaller than siemens type companies, just saying "___ just up and broke, can you send a new one?" Mar 10 15:21:47 I think it is still under warranty in theory, but I have no idea where the paperwork is. Mar 10 15:22:05 If the part had been expensive then I was going to phone them and make a fuss, but for £15 I think I might as well just suck it up. Mar 10 15:25:23 k Mar 10 15:25:31 Might just need to call up and have the serial # handy Mar 10 15:25:47 But yeah, 15 I think that was a EUR sign, isn't too bad :) Mar 10 15:27:57 looks like the problem is spreading, see ML. help2man-native: task configure fails with `configure: error: perl module Locale::gettext required` Mar 10 15:29:24 pb__, this is part of the remodeling/ Mar 10 15:30:05 Tartarus: that was meant to be a pound-sterling sign, but the exchange rate is probably near enough to 1 nowadays that it doesn't make a lot of difference :-} Mar 10 15:31:35 Crofton: not as such. I think the oven failure was unrelated. Mar 10 15:31:48 this is still in the "old" kitchen :-} Mar 10 15:31:50 so it wasn't brand new Mar 10 15:31:52 ah Mar 10 15:31:58 no, about a year old Mar 10 15:32:14 well, 9-10 months, I think I bought it last June or thereabouts Mar 10 15:32:53 it's not totally impossible that my daughter might have smashed the glass, she does rather like bashing objects together at the moment. Mar 10 15:33:06 :) Mar 10 15:33:26 ask florian about his child and the drink powder sometime :) Mar 10 15:40:32 Hey, I just tried to upgrade to bitbake 1.12.0, and I'm getting the "ImportError: No module named process" error. Is there a solution to that problem yet? Mar 10 15:45:48 xobs, python --version please? Mar 10 15:46:05 Tartarus: Python 2.6.6 Mar 10 15:46:13 xobs: if it was a known problem, it'd likely be fixed :) Mar 10 15:46:29 that's strange, many of us are on 2.6.6 and it works fine Mar 10 15:46:29 I see mention of it on the oe mailing list, but no solution. Mar 10 15:51:09 This is the error, though I'm getting it on 2.6.6: http://lists.linuxtogo.org/pipermail/openembedded-devel/2011-February/030472.html Mar 10 15:54:48 well, lib/bb/server/process.py exists, i don't see how that would happen :\ Mar 10 15:55:39 I'm not very familiar with python, but I can run "import bb" and "import bb.server", but "import bb.server.process" fails. Mar 10 15:56:59 doesn't happen with the python 2.6.6 i have installed. what distro is this? Mar 10 15:57:37 Ubuntu 10.10. Running in a VMWare instance that I SSH into. Mar 10 15:58:41 * kergoth_ tries creating an ubuntu 10.10 chroot Mar 10 15:58:57 hi, does anyone here know whether oprofile call graphs should work(tm) without the code being compiled with frame pointers? Mar 10 15:59:08 (aside: 10.10 has I/O performance issues atm, can result in doubling build times, Tartarus knows more) Mar 10 16:01:56 That's good to know about performance. Though I've been running bitbake 1.10 for a while. I was hoping the upgrade to 1.12 would help speed things up. Mar 10 16:02:51 1.12 would, but ubuntu 10.10 has general perf issues Mar 10 16:03:43 03Tom Rini  07master * r81abfc6b78 10openembedded.git/recipes/help2man/ (help2man_1.36.4.bb help2man_1.38.2.bb): Mar 10 16:03:43 Revert "help2man: Update to use 'inherit gettext'" Mar 10 16:03:43 This reverts commit bccf626b1428d7087d261bf4ad91744b5b5d5418. Mar 10 16:09:50 orjanf: I don't think so. You'll get immediate caller information but nothing beyond that. Mar 10 16:10:33 pb__: any idea if it works *with* frame pointers? Mar 10 16:10:46 I think it does on at least some targets. Mar 10 16:10:57 iirc, arm and x86 are meant to work. I don't know about others. Mar 10 16:12:33 pb__: ok, thanks Mar 10 16:14:16 tartarus: do you know what's wrong with u10.10, or is it just reported to suck? Mar 10 16:18:59 Lemme see if I can find it again Mar 10 16:20:25 http://www.frontsidebus.net/2010/10/ubuntu-10-10-benchmarks/ Mar 10 16:20:31 We see simply awful build times Mar 10 16:22:03 mm, that article seems to be talking about some kind of ext4 performance regression. I wonder if that's it. Mar 10 16:22:15 Nope Mar 10 16:22:25 funnily enough we have just switched our main build host from xfs to ext4 to improve the performance :-} Mar 10 16:22:26 We had someone try ext3 and ext4, same HW, same poor results Mar 10 16:23:09 weird. I wonder what's up with that. Mar 10 16:23:42 Well, perhaps I'd be better off just redoing the build system rather than trying to hunt down the problem with bitbake. Mar 10 16:24:02 Sounds like it'd save time in the long run. Mar 10 16:24:04 xobs, well, it should _run_ Mar 10 16:24:11 And it's kinda worrying that it doesn't Mar 10 16:24:21 But yeah, 10.04 is a better bet, imho Mar 10 16:26:18 I'm going to guess it's a botched install.. Mar 10 16:27:02 Yes, that's it. "sudo python setup.py install" didn't install any libraries. Perhaps it detected the 1.10 libraries were already installed? Mar 10 16:28:52 In any case, manually removing bitbake from /usr/local/lib/python2.6/dist-packages/ has fixed the problem. Mar 10 16:29:05 ...and then re-running "sudo python setup.py install". Mar 10 16:29:49 interesting. must be a distutils behavior, we just use it Mar 10 16:31:36 xobs, note that most folks don't install bitbake globally Mar 10 16:31:48 just unpack the archive and add /path/to/bitbake-VER/bin to PATH or similar Mar 10 16:32:38 Tartarus: Ah, that makes sense. I'll have to add that to my setup script. Still, the temptation to run the Python equivalent of "make install" is a strong one. Mar 10 16:33:07 I personally really like the idea of just installing it with standard mechanisms and running it that way, but historically very few people do it Mar 10 16:33:20 plus often in the past development was quite rapid, so you'd have to reinstall a lot :) Mar 10 16:35:05 * pb__ is using system-installed bitbake at the moment, seems to work well Mar 10 16:35:19 (now that debian sid contains a useful version again) Mar 10 16:35:55 one less source of aggro when starting to build. the fewer env vars I have to mess with, the better. Mar 10 16:36:03 agreed Mar 10 16:36:11 that + bblayers and you shouldn't have to set any vars at all Mar 10 16:36:14 which is pretty nice Mar 10 16:36:29 right Mar 10 16:37:46 at my new job we use perforce, and one of my pet hates there (along with a long list of other perforce dislikes) is that I now need to set this silly P4CLIENT var in the environment all the time. Mar 10 16:39:04 will tinderbox ever work again ? Mar 10 16:39:14 yeah, one day Mar 10 16:39:31 it has some performance issues that need to be resolved Mar 10 16:44:22 Hi, I look for a document about oe but all wikis manuals are just say "enter this command", "change this value" etc. Mar 10 16:45:08 I need extensive informations about oe build steps to add my application and making my own distrubition Mar 10 16:45:28 is there anyone know a document like this? Mar 10 16:47:16 I think just a few people know how oe works, probably just people who maintain oe project Mar 10 16:47:52 so why anybody try to write a nice extensive document for other people Mar 10 16:49:49 well, there's the openembedded user manual Mar 10 16:50:22 and bitbake has its own user manual Mar 10 16:50:52 perhaps neither of those are exactly what you want, but both of them are indeed intended to be a "nice extensive document for other people" Mar 10 16:52:09 dcordes, and by performance issues pb__ means that I killed it by doing things like http://dl.dropbox.com/u/3284539/release-2011.03-build-matrix.pdf, a lot Mar 10 16:52:22 (and I'll update the release-2011.03 table this weekend w/ that pdf Mar 10 16:52:37 and it's ~9MB if anyone is about to open it) Mar 10 16:52:46 yeah. not that what you were doing was unreasonable, but it turns out that the infrastructure couldn't quite cope with that. Mar 10 16:53:23 pb__, heh. Yeah, it's just a bigger load all at once than it was designed for Mar 10 16:53:31 Which is kinda a good problem to have Mar 10 16:53:39 it's possible that the right way to proceed is to host tinderbox somewhere else, so that at least it won't take out everything else using mysql if it goes wrong again. Mar 10 16:54:50 right now it seems to be on a db which is shared by not only other bits of oe infrastructure but other projects entirely, and it probably is not very fair to swamp them with stats traffic. Mar 10 16:55:53 pb__: Tartarus: I see. thansk for the input. looking forward for it to return! Mar 10 16:56:11 pb__, yeah, that too Mar 10 16:56:25 I look forward to, and wish I had more time for, working on the new and improved concept Mar 10 16:56:51 Morning folks Mar 10 16:56:59 hi ka6sox-away Mar 10 16:57:06 howdy... Mar 10 16:58:03 anyone have a clue on current size of OE repo from a full clone? Mar 10 16:58:16 I think we may overrun github... Mar 10 16:58:38 sec Mar 10 16:58:55 trini@easource:/opt/git/upstream/openembedded.git$ du -sk . Mar 10 16:58:55 273564. Mar 10 16:59:04 thats in k? Mar 10 16:59:05 semi recent pull in my internal mirror Mar 10 16:59:06 yes Mar 10 16:59:28 okay thanks. Mar 10 16:59:57 that isn't anything bitbake or -core right? Mar 10 17:00:17 right, just openembedded.git Mar 10 17:00:26 thanks. Mar 10 17:01:24 bitbake is trivial anyway, it's only a megabyte or two I think Mar 10 17:02:33 ya a MB or 2 among 270ish is noise. Mar 10 17:04:12 oe-core is like 65-70mb at the moment, it seems Mar 10 17:05:22 kays_: What pb__ said is right-- the Bitbake manual and the OE manual are the right place to look. They are a little sparse, but as far as I can tell, they're accurate. Mar 10 17:05:46 kays_: I've also found this Poky Linux appendix useful: http://www.yoctoproject.org/docs/poky-ref-manual/poky-ref-manual.html#ref-bitbake Mar 10 17:06:37 kergoth_, thanks that was going to be my next question. Mar 10 17:07:00 I don't think I need to worry about bb mirroring right? thats handled? Mar 10 17:07:22 thanks pingswept but I think there should be a simpler and step by step explained document, life will be easier on oe then Mar 10 17:07:47 kays_: Also look at this Gumstix tutorial: http://www.gumstix.net/Setup-and-Programming/view/Build-system-overview/Hello-world-tutorial/111.html Mar 10 17:08:32 Agreed that the documentation could be improved, but you won't hear me complaining about something I got for free. Mar 10 17:09:19 pingswept, kays_: we have a wiki and accounts are free... Mar 10 17:10:48 ka6sox-away: An excellent point. Any idea how to contribute to the bitbake manual? Mar 10 17:11:26 I'm sure if you ask that question the bb folks would love help with documentation. Mar 10 17:11:43 wait for it...they will speak up I'm sure :D Mar 10 17:11:45 pingswept: the same way you contribute to the code, it's docbook in bitbake git Mar 10 17:11:56 pingswept: it's a docbook file, the source is in the bitbake distribution (and the git tree) Mar 10 17:12:06 patches are welcome Mar 10 17:12:12 Ah, nice. Thanks. Mar 10 17:12:32 Send a patch to whatever the bitbake dev list is called? Mar 10 17:12:47 the bitbake manual even whenw e first wrote it was only really reference material -- it'd be nice to flesh that out for some things that are missing first, then perhaps move on to more introductory or troubleshooting material Mar 10 17:12:56 pingswept: yep Mar 10 17:13:25 Yeah, that seems like a good step. I actually think the Bitbake manual is pretty good as a reference. Mar 10 17:14:08 heh, not sure anybody has ever said that before. Mar 10 17:14:42 heh, i think it covered the file format pretty well originally, but that's about it, and even that is incomplete. it doesn't even mention EXPORT_FUNCTIONS for example Mar 10 17:15:13 * kergoth_ thinks he personally might contribute to such things more if they were sphinx/rst instead of docbook, has an aversion to xml :P Mar 10 17:15:47 Interesting. I'm pretty new to Bitbake. As a result, I didn't know that EXPORT_FUNCTIONS exist. I guess every manual seems complete unless you wrote the code. Mar 10 17:16:01 yeah, and I suspect it must be way out of date. The last time I can remember editing it was when I was sitting in a hotel room in Ottawa, must have been five or six years ago at least. Mar 10 17:16:25 yeah, that seems about right Mar 10 17:16:27 pb its always 5 years ago when you touched something Mar 10 17:16:31 *g* Mar 10 17:16:50 woglinde: yeah, that is probably true Mar 10 17:17:00 Mar 10 17:17:20 right Mar 10 17:17:55 * Tartarus really hates on DEPENDS_virtclass-native semantics Mar 10 17:18:31 Tartarus: heh, it's not very special, i think you just hate the way bitbake variables are processed ;) Mar 10 17:18:38 kergoth_, my wife thinks I have a mistress called OE. Mar 10 17:18:47 ka6sox-away: hehe Mar 10 17:19:01 ka6sox-away: at least it's not outlook express Mar 10 17:19:02 ka6sox lol Mar 10 17:19:16 kergoth_, thank ghod! Mar 10 17:19:32 kergoth_, yes, that's it too Mar 10 17:19:48 I just really want to treat it like it wasn't already one level of override in Mar 10 17:19:48 hm the kvm people made the update path to 0.14 really hard for me Mar 10 17:19:54 and my virtio setup Mar 10 17:20:00 yeah, I think my wife used to suspect I was having an affair with nagios when I used to do sysadmin duties. Mar 10 17:20:01 kergoth_: Is the OpenEmbedded manual also a docbook file somewhere in the OE repo? Mar 10 17:20:30 pingswept: exactly Mar 10 17:20:36 it's in docs/usermanual/ Mar 10 17:20:57 pb its incinga now Mar 10 17:22:03 pb__: Will it move to openembedded-core? Mar 10 17:22:23 Looks like there is some other documentation there already, but not the usermanual. Mar 10 17:22:51 pingswept: poky has a user manual too, might be good to start with theirs instead, not sure what the plan is at the moment Mar 10 17:23:25 kergoth_: because of the Yocto merge (or whatever is happening)? Mar 10 17:23:47 pingswept: good question. I have no idea I'm afraid. Mar 10 17:23:57 tartarus might be able to tell you what the plan is there. Mar 10 17:24:47 hm 10 million sold kinect Mar 10 17:26:50 ah, nearly time to go home Mar 10 17:26:57 another evening with fermacell beckons Mar 10 17:27:05 Tartarus: I think we'd have been better off with a different implementation of overrides. probably would've been better off just having the ability to make a conditional block.. if foo ... endif or so, which is evaluated right then and there, and if you need something postponed explicitly, you can make a damn event handler and do it there Mar 10 17:27:06 heh Mar 10 17:28:10 kergoth: yeah, probably true Mar 10 17:28:42 bye pb Mar 10 17:29:22 Tartarus: the special bits for DEPENDS_virtclass-native are tied up in both base.bbclass and native.bbclass. native.bbclass saves the current DEPENDS into DEPENDS_virtclass-native, and base.bbclass prepends to DEPENDS_virtclass-native. there's interactions with evaluation order, though. DEPENDS_prepend will be applied after the DEPENDS are saved into DEPENDS_virtclass-native, but before DEPENDS_virtclass-native becomes DEPENDS, so if you Mar 10 17:29:22 define DEPENDS_prepend, you also have to define DEPENDS_virtclass-native_prepend. Mar 10 17:29:29 Tartarus: which is just painful and headache inducing Mar 10 17:30:13 ga Mar 10 17:30:16 So Mar 10 17:30:26 I maybe have something right, will pastebin and then post to the ML Mar 10 17:30:43 of course, some of the design with our file format was due to how it evolved. event handlers weren't there originally, they were added after the initial design and implementation Mar 10 17:32:02 http://pastebin.com/L38EULcJ Mar 10 17:32:40 As the problem with broken recipes is that they also modify DEPENDS_virtclass-native and so while ORIG_DEPENDS is right, the mangled one ends up being wrong Mar 10 17:33:11 it also depends on more ordering issues. did they inherit native before or after gettext? if after, those prepends aren't needed. if before, they are Mar 10 17:33:18 * kergoth_ gets a tylenol Mar 10 17:34:01 kergoth_, too early for anything stronger :( Mar 10 17:34:40 kergoth_, ah, so it's all BBCLASSEXTEND recall, so it's always "before" i think Mar 10 17:36:16 http://www.newegg.com/Product/Product.aspx?Item=N82E16833316041 Mar 10 17:36:34 oops sorry...new backend infra. Mar 10 17:36:36 bbclassextend loads native.bbclass after the recipe and all of its inherits are parsed. need a damn map to keep track of when things happen :| Mar 10 17:36:46 * Tartarus idly wonders if we shouldn't have USE_NLS_virtclass-native = no Mar 10 17:38:38 non-english speakers might want the build tools we build to show messages in the log files in their language? Mar 10 17:38:39 * kergoth_ shrugs Mar 10 17:38:57 heh, ok, true Mar 10 17:39:20 okay now another try with kvm 0.14 Mar 10 17:39:30 course, bitbake doesn't emit its own messages in anything but english.. Mar 10 17:39:31 * Tartarus posts the patch Mar 10 17:39:31 we need to fix that Mar 10 17:44:22 I'm looking for a way to get ${WORKDIR}/.. returned to me normalized (i.e. w/o the ..) any suggestions on the easiest way to do this? Mar 10 17:45:13 dirname? Mar 10 17:45:19 Or not in shell? Mar 10 17:45:24 fray: os.path.normpath()? Mar 10 17:45:28 no, via bitbake in a bbclass.. I thought there was something like expandPath Mar 10 17:45:56 normpath is likely what i need.. Hmm Mar 10 17:47:10 fraxinas: bb.data.expand("${FOO}/bar", d) Mar 10 17:47:15 er, mistabcomplete Mar 10 17:47:25 or os.path.join(d.getVar('FOO', True), 'bar') Mar 10 17:47:41 what I have is bb.data.expand("${WORKDIR}/..", d) Mar 10 17:47:45 kergoth: that won't normalize .., will it? Mar 10 17:47:51 but what I need is the normalized path.. so normpath will do it Mar 10 17:47:56 oh, right, i see Mar 10 17:48:01 os.path has lots of useful functions, see pydoc :) Mar 10 17:48:03 phew i was like... omg what have i done wrong again this time :) Mar 10 17:48:07 fraxinas: :) Mar 10 18:25:31 It appears that user account creation has been disabled on the wiki since January. How can I get an account? Mar 10 18:30:28 03Eric Bénard  07org.openembedded.dev * r712d43b10d 10openembedded.git/recipes/busybox/busybox_1.18.3.bb: Mar 10 18:30:28 busybox: add latest fix Mar 10 18:30:28 klogd: fix the problem of delayed handling of ^C and SIGTERM Mar 10 18:30:28 Signed-off-by: Eric Bénard Mar 10 18:30:33 03Eric Bénard  07org.openembedded.dev * rb298c9e327 10openembedded.git/recipes/qt4/ (11 files in 2 dirs): Mar 10 18:30:33 qt4: add 4.7.2 Mar 10 18:30:33 * details on the 4.7.2 release are available here : Mar 10 18:30:33 http://qt.nokia.com/developer/changes/changes-4.7.2/ Mar 10 18:30:33 * please note that x11 versions are not yet tested Mar 10 18:30:33 Signed-off-by: Eric Bénard Mar 10 18:42:58 pingswept, sec, lets see who is in... Mar 10 18:47:24 hi risca Mar 10 18:48:18 hm mplayer task would be cool too Mar 10 18:49:06 uh ups Mar 10 18:54:20 evening Mar 10 19:17:04 03Koen Kooi  07org.openembedded.dev * r664005f4dd 10openembedded.git/recipes/powervr-drivers/ (libgles-omap3_4.03.00.01.bb omap3-sgx-modules_1.6.16.3977.bb): Mar 10 19:17:04 libgles-omap3: update 4.03 release to 02 Mar 10 19:17:04 Signed-off-by: Koen Kooi Mar 10 20:05:51 Any thoughts on resetting PACKAGES_DYNAMIC in native.bbclass the way we do for PACKAGES? Mar 10 20:06:03 currently, gtk+ and gtk+-native are both seen to be emitting gdk-pixbuf-loader-* Mar 10 20:06:10 can result in preference conflicts with bitbake Mar 10 20:06:11 yes Mar 10 20:06:13 saw it too Mar 10 20:06:36 k, will send a patch/rfc to the list and see if anyone objects, and will poke RP directly Mar 10 20:06:44 the alternative would be to mangle it Mar 10 20:06:58 but that's more complex, since some emitted packages will contain ${PN} and some won't Mar 10 20:07:31 so e.g. gtk+-native-dev shouldn't be mangled, but gdk-pixbuf-loader-* would need native in it somewhere.. Mar 10 20:07:36 think emptying it for now is the best bet Mar 10 21:06:21 kergoth_: yes I see that problem Mar 10 21:06:33 kergoth_: resetting PACKAGES_DYNAMIC in native.bbclass is good idea Mar 10 21:07:20 kergoth_: actually can we just take all PACKAGES and add -native to them ? Mar 10 21:07:29 in native.bbclass would be nice Mar 10 21:07:32 khem hm no Mar 10 21:07:48 woglinde: why not Mar 10 21:07:51 I mean all packages and put -native to them Mar 10 21:08:00 or only for dynamic ones? Mar 10 21:08:21 I meant for all Mar 10 21:09:04 if a recipe generates 10 different packages then BBCLASSEXTEND = "native" should get -native appended to all 10 Mar 10 21:10:52 khem: yeah i thought about the mangling thing too, but its non-trivial, since PN is often in package names, and PN often contains the -native already Mar 10 21:11:00 still, worth looking into in the future Mar 10 21:11:31 kergoth_: hmmm so we will have to lookout for that yes Mar 10 21:11:54 * khem has successfully set a internal git repo Mar 10 21:12:09 one possibility would be to createCopy the datastore, remove the postfix from PN, and re-expand and then mangle the PACKAGES Mar 10 21:12:15 * kergoth_ shrugs Mar 10 21:12:38 yeah something like that would work Mar 10 21:12:49 or may be look for -native Mar 10 21:12:51 and of course, ideally -native wouldn't always be on the end.. i'd rather see foo-native-dev than foo-dev-native Mar 10 21:13:05 ah right Mar 10 21:13:12 same for dbg and doc.. Mar 10 21:13:32 so may be looking for string -native and ignoring it for mangling would work Mar 10 21:14:06 that sounds somewhat iffy though. what if hte recipe is for some project that has -cross in it's name? ;) Mar 10 21:14:09 or whatever Mar 10 21:14:28 yeah hmm Mar 10 21:14:35 khem see Mar 10 21:14:38 can we annotate packages Mar 10 21:14:42 i'd think 1) expand with BPN not PN to remove the -native, 2) split off any postfix (-dbg, -dev, -doc, there's a variable for it), 3) add -native postifx Mar 10 21:14:49 4) re-assemble Mar 10 21:14:59 but there's multiple possibilities Mar 10 21:16:03 alternatively, we could make recipe maintainers add a variable to the package name of each package that doesn't include PN, to add the postfix, but that sounds problematic :) Mar 10 21:16:13 kergoth_: yeah and hope that there is no BPN containing -native Mar 10 21:16:53 ah I dont know what I typed Mar 10 21:16:57 need food and sleep Mar 10 21:16:59 hehe Mar 10 21:17:12 anyway, yeah, good idea in general, it'd be nice if packaging was consistent across everything Mar 10 21:17:31 at which point we could consider something along the lines of waht oe-lite does, really.. Mar 10 21:17:38 very long term, anyway Mar 10 21:23:29 03Koen Kooi  07org.openembedded.dev * r941a7800d1 10openembedded.git/recipes/ (4 files in 2 dirs): Mar 10 21:23:29 *beagleboard-validation: collection of tasks iand images to serve as basis for beagleboard.org validation effort Mar 10 21:23:29 Signed-off-by: Koen Kooi Mar 10 21:23:42 03Koen Kooi  07org.openembedded.dev * re0a7d4ea00 10openembedded.git/recipes/tasks/task-omap-drivers.bb: Mar 10 21:23:42 task-omap-drivers: task for commonly needed drivers on omap chips Mar 10 21:23:42 Signed-off-by: Koen Kooi Mar 10 22:55:48 Anyone want to review my gettext patch? Mar 10 23:07:46 Tartarus: looks like the cron script for automatically creating the testing branch worked :-) Mar 10 23:08:13 maybe it will actually get done on time now ... Mar 10 23:11:58 Tartarus: let me look Mar 10 23:13:13 hmmm git add will take a snapshot of the given files and stages that content in the index, ready for inclusion in the next commit. Mar 10 23:13:52 so does it store multiple copies Mar 10 23:14:28 how is a blob contructed ? is it some sort of compressed thing Mar 10 23:39:50 <[Rui]> spank spank spank perl-native is borked Mar 10 23:40:37 <[Rui]> damn it, tinderbox is down :( Mar 10 23:42:36 tinderbox is administratively down some time ago Mar 10 23:42:55 so.. you should forget about it until someone will announce something other Mar 10 23:43:14 <[Rui]> I only noticed it now :) Mar 10 23:43:22 yeah.. Mar 10 23:43:30 <[Rui]> it would make it easier to point to my error :) Mar 10 23:43:53 <[Rui]> usually who helps me on this things is JaMa but he's away until monday Mar 10 23:45:09 pastebin your logs Mar 10 23:46:18 <[Rui]> I had already cleared them, but I'm rebuilding. I basically updated my local setup a couple of hours ago, cleared the tmp/ and bb -k someproject Mar 11 00:51:50 ah to answer myself http://book.git-scm.com/7_how_git_stores_objects.html Mar 11 00:52:34 so packfile it is Mar 11 01:10:14 kergoth: is there a way to extract date/time info given a ref Mar 11 01:10:30 talking about git Mar 11 01:12:31 Look at git help log. You can specify a custom format string -- git log --pretty=format:, iirc. Mar 11 01:21:10 kergoth: I was thinking we could use that instead of git rev in SRC_URI Mar 11 01:21:29 that will always have increasing number YYYYMMDDHHMMSS Mar 11 01:23:16 khem, that wont work Mar 11 01:23:22 the time has to refer to a commit Mar 11 01:23:28 but a commit to which tree? Mar 11 01:23:40 grg: hmmm Mar 11 01:23:59 we already use repo in SRC_URI anyway Mar 11 01:24:09 or repo url Mar 11 01:24:31 no, i mean that the commit time will be the time that the patch is first committed *somewhere* Mar 11 01:24:42 its then merged, etc Mar 11 01:24:42 oh i see Mar 11 01:24:49 thats correct Mar 11 01:25:11 with a pull model in the future, i suspect that will not work Mar 11 01:26:07 I don't think that is a very good idea. Mar 11 01:27:13 kergoth: yeah easier is to just bump the PR along Mar 11 01:29:38 Poky puts the checksum in the stamp. Long term could avoid the manual bumping entirely in favor of srcrev-like incrementing from a local database for packaging, once it won't be needed for the rebuilding piece Mar 11 01:29:47 * kergoth shrugs Mar 11 01:30:31 kergoth: yes thats there sure I was looking for easy hack right now Mar 11 01:30:47 Ah Mar 11 01:32:18 * kergoth ponders Mar 11 01:53:43 * Crofton is getting behind on email ... **** ENDING LOGGING AT Fri Mar 11 02:59:57 2011