**** BEGIN LOGGING AT Thu Jan 24 02:59:58 2013 Jan 24 03:42:52 Hi All Jan 24 03:43:53 anybody here? Jan 24 05:58:59 mostly digesting... Jan 24 06:12:13 Are MACHINE vars ever used outside of their bps layers? Jan 24 06:12:21 bsp Jan 24 06:19:33 sr105: MACHINE is used quite a bit Jan 24 06:19:48 its used to define overrides Jan 24 06:20:09 and some machine specific tweaks in recipes which may not be in bsp layerbut elsewhere Jan 24 06:20:35 khem: Well, I'm trying to figure out the repercussions of defining my own machine based on another, but I don't want to lose any overrides of the original. Jan 24 06:22:30 or am I going about this all wrong? I'm using an off the shelf processor board with a custom main board. So I want 95% of the stuff written for the purchased board, but with my tweaks when I need them. Jan 24 08:17:38 shoragan: hi Jan 24 08:39:00 good morning Jan 24 09:03:27 currently ext2 images seem to have some 200MB of extra space, how does one increase that? Jan 24 09:07:41 IMAGE_ROOTFS_EXTRA_SPACE seems to be it Jan 24 09:39:44 morning all Jan 24 10:02:08 hi zecke :) Jan 24 10:35:29 ant_work: found my kexecbootproblem, missing evdev ..... Jan 24 10:40:24 Hi Jan 24 10:41:01 Is there any bitbake recipe that adds new user to filesystem? I have recipe that requires new user to be added to /etc Jan 24 10:41:31 But I have not found any example. What is the proper way of doing this? Jan 24 10:41:50 I have few ideas but these are kind "hacks" Jan 24 10:49:07 there is an example somewhere in one of the recipes in oe-core Jan 24 10:49:30 foobar321: I know that avahi adds a user, so try checking that one Jan 24 10:51:52 jackmitchell: oh, great. I'll take look into that Jan 24 10:52:07 pwgen: great Jan 24 10:53:26 now i have a kexecboot image, with a lot of flaws. not every kernel is able to kexecboot, and no chance to read the mmc after executing kexec, i have to boot to an external usb stick Jan 24 10:55:02 seems if the kernel size gets over 5 mb then kexec fails, but not shure if that is the reason . Jan 24 10:58:21 "no chance to read the mmc after executing kexec" might be a buggy driver, or a buggy hard Jan 24 10:58:30 (ie not kexec's fault) Jan 24 11:02:34 iirc there was indeed a size limit but for the initramfs I think Jan 24 11:08:01 pwgen: you are passing that long cmdline full of params, isn't? Jan 24 11:08:56 like sdhci-bcm2708.emmc_clock_freq=100000000 Jan 24 11:21:41 has anyone noticed bitbake deadlocking on parsing recipes recently? Jan 24 11:26:40 hello Jan 24 12:44:06 sdhci-bcm2708.emmc_clock_freq=100000000 helped ... thx Jan 24 12:54:41 pwgen: http://www.raspberrypi.org/phpBB3/viewtopic.php?f=29&t=6201&start=650 Jan 24 12:55:46 pls take conservative defaults for the -kexecboot kernel Jan 24 12:57:38 ah, and send the config/patches for meta-handheld Jan 24 14:32:17 Quick question, if I modify the kernel config options, are they saved to defconfig in the tmpdir or is there a config saved elsewhere, I just want to diff the changes. Jan 24 14:35:23 the former Jan 24 14:36:19 ok, I'll take a copy and clean the kernel then, thanks. Jan 24 14:36:39 *then build again Jan 24 15:55:00 i'd like to write a class to append extra stuff to ipk's pre/post inst/rm scripts ? where should i hook ? Jan 24 15:56:37 afournier: you'd just prepend or append to the functions in the class as with any other function Jan 24 15:57:16 afournier: you can either do it directly or by using d.appendVar() in a python function, depending on which suits the situation Jan 24 15:57:30 the thing is that the appended script would be somehow generated so i cannot do it inline Jan 24 15:57:47 ok, python function is probably going to be the place I would think Jan 24 15:57:51 so i guess python appendVar would do it yes Jan 24 15:57:59 thx Jan 24 15:58:26 there are a number of examples, e.g. update-alternatives.bbclass Jan 24 15:58:33 thanks Jan 24 15:58:43 although that does getVar/setVar, the idea is the same Jan 24 15:58:48 maybe it has already been done, but my idea is to append mysql scripts Jan 24 15:59:46 so it'd grab the script name from the recipe, read it, and append it inside a "cat << EOF | mysql" Jan 24 16:00:21 if -z "$D" :) Jan 24 16:01:05 this way i don't have to install the script on the host, it can stay in /tmp while installing Jan 24 16:03:22 afournier: that sounds like it should work Jan 24 16:03:50 neat! Jan 24 18:03:13 Is it possible to define a new machine B based on A and tell bit bake that if there are overrides present, prefer B to A, but use A if present and no B. Jan 24 18:03:15 does anyone knows why the qt.conf generated has this option Jan 24 18:03:24 Binaries = ${STAGING_BINDIR_NATIVE} Jan 24 18:03:29 instead of: Jan 24 18:03:38 Binaries = ${STAGING_BINDIR} Jan 24 18:03:40 ? Jan 24 18:10:54 otavio: i saw that you did that commit.. could you please explain me why? Jan 24 18:11:10 i saw on the commit comment but it doesn't explain this situation. Jan 24 18:25:21 ftonello: Ah; it is need because you need to use moc, and other binaries, from host ... Jan 24 18:35:49 otavio: how to change the $$[QT_INSTALL_LIBS] for example? I thought qt.conf changed that Jan 24 18:36:10 because $$[QT_INSTALL_LIBS] from qmake2 installs on the host machine as expected Jan 24 18:37:01 ftonello: it all depends on qt you're using Jan 24 18:37:05 qt4? qt5? Jan 24 18:37:12 otavio: qt4 Jan 24 18:37:31 otavio: qt5 in near future, btw Jan 24 18:38:16 ftonello: you need to use QT_CONFIG and provide the path for the qt.conf Jan 24 18:39:04 otavio: what you mean? Jan 24 19:06:49 hi zecke Jan 24 19:16:49 otavio: wouldn't be necessary to move qt.conf to ${S}? Jan 24 19:17:14 ftonello: no; it is written when need Jan 24 19:19:08 anyone know much about opkg and "holding" Jan 24 19:19:20 I just got asked a question that made me look dumb :) Jan 24 19:19:27 Crofton|work: its kinda broken Jan 24 19:19:45 rofl Jan 24 19:19:50 yeah, that is what I was asked Jan 24 19:19:56 baically, is it broken Jan 24 19:20:14 its supposed to work like deb package holding, but opkg seems happy to upgrade anyway Jan 24 19:20:48 XorA, thanks I owe you a few beers for that Jan 24 19:21:07 I think just to annoy it will work in some cases and not others from memory Jan 24 19:24:29 XorA, could it be made to work? Jan 24 19:24:52 Crofton|work: it should be relatively simple for a competent coder Jan 24 19:25:00 opkg code isnt insanely complex Jan 24 19:25:56 Crofton|work: does OE support apt/deb properly yet? Thats a much saner solution for EMMC based devices anyway :-D Jan 24 19:26:14 Crofton|work: I only ask as Ive been out the loop on the packaging changes Jan 24 19:27:20 duno Jan 24 19:27:28 I was getting asked by some guys Jan 24 19:27:34 now I know what holding is Jan 24 19:28:05 you stick a 6inch nail in the package and nothing is allowed to move it :-D Jan 24 19:28:17 rofl Jan 24 19:28:35 * XorA may have been watching too much moonshiners while in USA Jan 24 19:36:58 hm can I enable verbose output for postinst when doing rootfs? Jan 24 19:50:44 woglinde: heh, we ned to fix postinst handling. today a *real* bug in a postinst will just be silently ignored until you boot the thing Jan 24 19:50:48 s/ned/need/ Jan 24 19:51:53 kergoth I get exit code 2 for my postinst but I cannt see why Jan 24 19:52:13 and the oe-core minimal image does not ship opkg stuff Jan 24 20:22:23 sr105: yes its ok to define a new machine if you really need one. But you can not piggy back on features of other machine unless there is a machine class where both of these belong to Jan 24 20:22:44 khem: thanks Jan 24 20:22:46 sr105: some layers have defined such overrides like meta-fsl-ppc does Jan 24 20:23:19 so in effect when you define a new machine inheriting from other machines its all ok you get all features of that machine barring any machineoverrides Jan 24 20:23:28 I gripped all of my layers for overrides and there are no odd ones outside of bps that I care about so I should be ok Jan 24 20:23:49 you can play games in your machine conf to pretent exactly like the machine you inherited from but then whats the point Jan 24 20:24:20 yes usually any machine specific tweaks now a days are in form of bbappends and reside inside the bsp layer Jan 24 20:26:41 well, I could either clone the existing BSP, modify it, and merge as needed. Or I can create a new BSP layer based on the former that does the minimum possible. I wasn't sure what made the most sense. Creating a new layer seems to me to be more manageable. Jan 24 20:28:19 we have different peripheral chipsets, LCD, and pin MUX to the original, too Jan 24 20:28:50 new layer that extends the original is a better way to do it Jan 24 20:29:14 makes it much easier to leverage bug fixes from the original, and clearly documents your deviations.. Jan 24 20:29:38 am I right to set my priority to +1 over the existing BSP layer? Jan 24 20:30:05 fray: thanks. Jan 24 20:30:33 the priority thing always messes me up.. I can't remember is +1 or -1 is higher Jan 24 20:31:09 I wish you could say MACHINE=B, INHERITED_MACHINE="A" Jan 24 20:31:20 +1 is higher Jan 24 20:31:35 sr105: depends if you want your layer to override the bsp layer you inherit from then yes Jan 24 20:31:42 I figured setting +1 was clearer than bblayers ordering Jan 24 20:32:00 khem: yes, definitely override Jan 24 20:32:20 sr105: for any .inc or .conf changes bblayer ordering is still honored Jan 24 20:32:29 and for classes too Jan 24 20:32:39 priority is only for recipes Jan 24 20:32:54 oh… good to know Jan 24 20:32:59 so depending upon what all you will be overriding you have to decide Jan 24 20:33:18 so, it's reverse order? last takes precedence. Jan 24 20:33:28 its like PATH Jan 24 20:33:37 whichever is found first is taken Jan 24 20:33:51 oh, then not reverse Jan 24 20:34:00 so you have to make sure that your layer appears before the contentions in BBPATH Jan 24 20:34:17 there are .= and =. operators Jan 24 20:34:20 read about them Jan 24 20:34:32 that explains why meta-yocto appears second... Jan 24 20:34:36 and choose the layering structure that fits your layer in whole cosmos Jan 24 20:35:28 I have read about those, thanks. I see. Jan 24 20:36:08 SIGFOOD bbl Jan 24 20:39:16 One thing about OE-core that I found incredibly confusing at the start was the use of "meta-name" for directories that were not layers, but rather collections of layers. All of the OE docs I've seen don't seem to address that. Jan 24 20:39:40 meta simply stands for "metadata" Jan 24 20:39:49 its a pokyism :-D Jan 24 20:39:54 yup.. Jan 24 20:40:04 * fray notes none of our commercial layers start with "meta" Jan 24 20:41:03 every doc, how to, or getting started seems to imply that layers are all meta-something. So when you see meta-intel or meta-openembedded, it's confusing for a new user. They're named like layers, but they don't "look" like layers inside. Jan 24 20:41:15 I see. Jan 24 20:42:00 Once I figured it out, it's quite simple, but it through me at first. Jan 24 20:42:04 threw Jan 24 20:42:26 the convention though is any layer (or group of layers) starts w/ "meta-" as an identifier.. Jan 24 20:42:33 but there is nothing enforcing that Jan 24 20:45:09 if I want to dump my existing TMPDIR, all I need to keep is the sstate_cache directory, correct? Jan 24 20:46:22 damn for changed postinst I need to recompile Jan 24 20:46:30 thats suckz for openjdk Jan 24 21:12:41 anyone here have experience using mercurial URLS in SRC_URI? Jan 24 21:13:03 other then I know they're supposed to work.. and the format is "hg://...." no.. Jan 24 21:13:13 I'm trying to do a simple "hg serve" from a repo and then use the URI in a recipe (temporarily) Jan 24 21:13:45 I've got this, but I'm running into errors: hg://minty:8000;module=;rev=db653ac03876 Jan 24 21:13:51 minty is localhost Jan 24 21:14:32 oh, it has to have a module name.... Jan 24 21:15:00 cmd = "%s clone %s %s://%s/%s %s" % (basecmd, " ".join(options), proto, hgroot, ud.module, ud.module) Jan 24 21:17:08 evening all Jan 24 21:17:17 hm how does update-alternatvies class work my bin is under $foo/dir2 Jan 24 21:17:17 * pb_ returns to the UK Jan 24 21:17:22 he pb Jan 24 21:17:23 good Jan 24 21:17:39 hi woglinde Jan 24 21:18:45 or ho does the class find out the bin Jan 24 21:18:51 hm read the source luke Jan 24 21:19:00 hi pb_ Jan 24 21:19:08 hi florian Jan 24 21:20:05 damn cold Jan 24 21:20:13 let my lying half the day in bet Jan 24 21:29:41 hello Jan 24 21:34:03 hi shinda Jan 24 21:34:35 hi woglinde Jan 24 21:35:39 Looks like bitbake can't pull from a mercurial repo hosted at the root of a server. Jan 24 21:35:59 I'll have to modify hg.py to be more flexible in the options Jan 24 21:37:52 hm Jan 24 21:38:00 send patches Jan 24 21:40:56 hmmm mercurial is not one of most used fetchers I must admit Jan 24 21:41:12 hi khem Jan 24 21:41:18 hey woglinde Jan 24 21:41:27 I have serval meta-java patches in the pipeline Jan 24 21:44:11 woglinde: cool. Jan 24 21:44:30 woglinde: one thing thats biting me atm is rebuilds from sstate Jan 24 21:44:49 jamvm hardcodes paths into binary Jan 24 21:44:58 Do we really need jamvm ? Jan 24 21:48:12 runtime or native? Jan 24 22:00:14 native Jan 24 22:03:54 yes we need it actual Jan 24 22:04:00 to build openjdk native Jan 24 22:04:16 hrms I do not understand update-alternatives class Jan 24 22:05:46 my binary is under $foo/$bla and not bindir Jan 24 22:11:53 hm its ALTERNATIVE_TARGET than Jan 24 22:15:08 woglinde: yes Jan 24 22:15:17 that class got revamped Jan 24 22:15:24 docu is bit confusing Jan 24 22:15:25 it looks good now though Jan 24 22:15:31 hmm that could be :)\ Jan 24 22:16:21 *sigh* waiting 20 minutes again only to see if update-alternatvies works Jan 24 22:45:20 good nite Jan 24 22:45:27 Hi. Is there a command to "prune", or whatever the term is, stale objects in the sstate-cache ? Jan 24 22:46:22 there's an sstate management script in scxripts/ in oe-core and opky reopsitories Jan 24 22:46:38 never used it, personally i just store them on filesystems with atime support and periodically drop archives i haven't used in the past week Jan 24 22:49:14 kergoth, ok, thanks ill check out the scripts then Jan 25 01:30:13 any easy way to change the target MACHINE instead of changing it manually in local.conf ? Jan 25 01:30:26 I tried MACHINE bitbake recipe and it didn't work Jan 25 01:30:37 MACHINE=bla bitbake Jan 25 01:36:14 ftonello: you need to have it in BB_ENV_EXTRAWHITE Jan 25 01:37:19 JaMa: thanks **** ENDING LOGGING AT Fri Jan 25 02:59:58 2013