**** BEGIN LOGGING AT Mon Oct 29 03:00:01 2012 Oct 29 03:58:09 how cryptic Oct 29 08:05:46 tried going from denzel to danny, and all is silent after "Starting kernel ...". any common pitfalls in this upgrade? I have the exact same kernel I used before (2.6.39), and have tried both new and old MLO+u-boot.img. Oct 29 09:11:48 good morning Oct 29 09:17:41 I would like recipes that set PARALLEL_MAKE = "" to include a comment describing the issue. preferrably linking to a bug report. looking at gcc-cross-canadian_4.7.bb now. ping RP_ Oct 29 09:21:41 why does the cross-canadian build need serial compilation when other gcc builds handle parallel? Oct 29 09:23:14 good question Oct 29 09:23:29 I agree there should be a comment indicating why PARALLEL_MAKE = "" was needed Oct 29 09:24:03 some of them that are left are quite old I suspect Oct 29 09:24:30 although we did go through and try to remove a bunch of them not that long ago Oct 29 09:25:57 morning all Oct 29 09:26:17 'morn Oct 29 09:28:32 yeah the cross-canadian one is from 2009-09-17. commit afa02c54 Oct 29 10:38:53 hi all Oct 29 10:52:50 hi all Oct 29 11:17:13 hi pb_, silvio_ Oct 29 11:18:03 hi bluelightning Oct 29 11:18:43 hi bluelightning Oct 29 11:21:49 I need fortran enabled in gcc it is safe and sufficient set " FORTRAN = ",fortran" " in the file "gcc-4.7.inc" or there is same specific reason because it it disabled? Oct 29 11:21:58 hi pb_ Oct 29 11:22:11 yes, that should be fine Oct 29 11:22:29 it's only disabled because most folks don't want it and don't want to spend the time to compile it Oct 29 11:22:53 of course, the support in the recipes might be bit-rotten for the same reason, but there is no particular reason you shouldn't be able to turn it on. Oct 29 11:23:11 What would people suggest is the best way to use dbus in an embedded system, glib, directly through the dbus library, anyone know of any other wrappers/helpers? Oct 29 11:23:28 ok thanks, I also see that there is also " gcc-csl-arm-2008q1.inc:FORTRAN = "" " abnd I m not sure where is needed Oct 29 11:23:48 jackmitchell: if you're using glib for your main loop, then the glib dbus bindings are probably going to be easiest. Oct 29 11:24:18 silvio_: the easiest thing would be to set it in your distro configuration Oct 29 11:24:20 jackmitchell, I used dbus for my needed Oct 29 11:24:25 otherwise, if you're using c++ then there are specific c++ bindings. failing that, probably the dbus native api. Oct 29 11:26:05 I am using C, and I am not using glib for anything else in my application - I'm just a bit worried about the size of glib and all the self defined types etc... Oct 29 11:26:49 if you're not using glib otherwise then GDBus is probably not going to buy you much. Oct 29 11:27:00 I would just use the native dbus api in that situation. Oct 29 11:27:16 ok, I'll see if I can find some up to date tutorials on it Oct 29 11:27:19 I Oct 29 11:27:28 bluelightning, the only way that I found is set FORTRAN = ",fortran", for enable it in the distro there is a smarter way? Oct 29 11:27:36 I didn;t want to start looking into it if using the native api was deemed 'stupid' Oct 29 11:28:09 silvio_: you would use the same value but because gcc-4.7 uses = and not ?= you will probably need to use an override Oct 29 11:28:49 silvio_: i.e. FORTRAN_somedistro = ",fortran" Oct 29 11:43:14 jackmitchell no glib support, hm so you have easy tasks or you are c guru Oct 29 11:44:41 bluelightning, the OE manual is leaky on the override, but if I understand I can put FORTRAN_mydistro = ",fortran" in layer.conf of my distro, I right? Oct 29 11:45:45 silvio_: or rather, it should be added to conf/distro/mydistro.conf in your distro layer Oct 29 11:46:20 woglinde: indeed, I don't think I would write C for a Linux system without glib myself, it's just too handy Oct 29 11:46:29 woglinde: possibly both ;) Oct 29 11:46:55 (and I'm not a C guru ;) Oct 29 11:48:20 woglinde: I've never really been introduced the glib as I write specifically for emebedded systems, and I've never written for desktop linux except in higher level languages (python, perl, bash) Oct 29 11:48:44 hm embedded is diffrent today Oct 29 11:48:51 woglinde: it's possibly just the next step in my developement Oct 29 11:48:55 and glib do not take this much space Oct 29 11:49:10 it has some nicer and easier to use abstraction than pure c Oct 29 11:49:17 the more I read the more I am starting to liek it Oct 29 11:49:35 especially for dbus Oct 29 11:49:37 *g* Oct 29 11:50:11 however I worry, as I am introducing embedded linux to a more "traditional" company, and while coming in as the new young grad student I'm trying not to step on toes Oct 29 11:51:21 using a c abstraction library may be more trouble than it's worth in the long run when it comes to the usual awkward Linux questions ;) Oct 29 11:51:57 yes if you are an c guru Oct 29 11:52:03 args a Oct 29 11:52:29 heh, I thought you were trying to make some sort of subtle pun on "ansi" Oct 29 11:52:31 :-} Oct 29 11:59:40 bluelightning, if I put it in the layer.conf the syntax remain FORTRAN = ",fortran" ? Oct 29 12:03:26 silvio_: don't put it in the layer.conf, that's not the correct place Oct 29 12:05:37 bluelightning, meta-mydistro/conf/distro/mydistro.conf ? Oct 29 12:10:33 silvio_: yes Oct 29 13:35:59 is it expected that the 'The recipe %s is trying to install files into a shared area' message is triggered for every rebuild? afais, the 'if search_output != None' check from 56268f6e4ed1fc11143173bb1717a8be78c728a5 holds everytime because 'search_output' never becomes None (only '') Oct 29 14:03:46 hi Oct 29 14:06:48 hi afournier Oct 29 14:06:56 hi silvio_ Oct 29 14:31:31 is there a way to bypass the license check ? Oct 29 14:33:13 LICENSE = "closed" iirc Oct 29 14:33:34 LICENSE = "CLOSED" Oct 29 14:42:01 thanks Oct 29 15:15:50 I'm trying to use the glib library however I'm not sure how to set up the compiler Oct 29 15:16:05 I have sysroot=/blah/blah/armv7... Oct 29 15:16:21 ${CC} `pkg-config --cflags --libs glib-2.0` -o myprogram source.c source.c ... Oct 29 15:17:08 That is kind of what I am doing, but when I try to include glib.h it says it can't find it Oct 29 15:17:18 so I use glib-2.0/glib.h which it finds Oct 29 15:17:33 but then in glib.h it references relative files which it can't find Oct 29 15:17:34 That suggests that your pkg-config is somehow malfunctioning. Oct 29 15:17:47 Using is wrong and, as you say, won't work. Oct 29 15:18:24 ok, I'm building through eclipse, and I don't actually think I am using pkg-config... Oct 29 15:18:32 ah Oct 29 15:18:47 so, when you said you were "kind of" doing what I said, I guess you meant "kind of" in the sense of "not". Oct 29 15:18:59 I suppose so :( Oct 29 15:19:13 I have glib-2.0 in my -l flag Oct 29 15:20:36 that's not a very good idea either, though in practice it will probably be ok. really, you want to just use what pkg-config tells you. Oct 29 15:21:12 ok, I'll have a read about using pkgconfig with eclipse Oct 29 15:21:23 ok, good plan Oct 29 15:22:00 I am also in the middle of transitioning my app over to auto tools which I hope will eliviate this kind of issue Oct 29 15:29:37 hey. how do i implement a -= in a .conf file? i need to remove some stuff from DISTRO_FEATURES Oct 29 15:36:16 CMoH: you can use := in conjunction with ${@oe_filter_out(...)} Oct 29 15:39:44 pb_: ok, fixed - I used a manual config rather than installing an eclipse plugin, but I used pkg-config to get it done, thanks for the help! Oct 29 15:40:05 bluelightning, thanks Oct 29 15:42:38 bluelightning, does this work in .conf files as well? or just in recipes? Oct 29 15:43:22 CMoH: should work anywhere Oct 29 15:43:50 i see Oct 29 15:44:22 when was this introduced? Oct 29 15:44:34 NameError: name 'oe_filter_out' is not defined Oct 29 15:45:25 it's quite old... Oct 29 15:45:58 current code is from 2010-05-05 at least Oct 29 15:46:00 it was in OE-Classic for a number of years Oct 29 15:47:03 thought I only see it used in recipes Oct 29 15:47:22 k... let me grep for it... Oct 29 15:47:32 is it sensible to define EXTRA_IMAGE_FEATURES in an image recipe, this is in order to get a "development" image and a "production" image. Oct 29 15:48:06 jackmitchell: I thought the whole purpose of the EXTRA variable was to set it from local.conf? Oct 29 15:48:26 Zagor: ok, so I could just use IMAGE_FEATURES in an image file Oct 29 15:48:30 jackmitchell you can do that, but most of the people I know simply change the setting in local.conf -- if you want it to be influenced by the environment you can do that as well Oct 29 15:49:10 bluelightning, oe_filter_out is defined in utils.bbclass, which i cannot inherit in a .conf file Oct 29 15:49:13 in that case, why does local.conf not use IMAGE_FEATURES += or IMAGE_FEATURES_append ? Oct 29 15:50:31 CMoH: doesn't INHERIT += "utils" help? Oct 29 15:50:56 CMoH: I guess this is automatic in OE-Core Oct 29 15:51:24 speaking of which; why not have a -= in bitbake to cover such situations? Oct 29 15:52:05 ty - will try that, bluelightning :) Oct 29 15:53:23 CMoH: because implementing it is nontrivial Oct 29 15:53:39 well... it's already implemented in oe_filter_out, it seems Oct 29 15:53:57 no Oct 29 15:54:02 removing a word is easy, obvious Oct 29 15:54:22 the implications, interactions with the parse process, *when* it happens, how it behaves with overrides, etc Oct 29 15:55:23 removing a word would be enough Oct 29 15:55:34 more than nothing anyway Oct 29 15:55:35 ? Oct 29 15:55:42 i'm not talking about doing more than removing a word Oct 29 15:55:54 i'm talking about making sure when and how that happens meets user expectations, always Oct 29 15:56:24 well, you have += which seems to work out fine; if -= would simply remove a word be applied in the same conditions as += Oct 29 15:56:34 wrong Oct 29 15:56:54 consider: FOO = "one two" BAR = "${FOO} three" Oct 29 15:56:58 BAR -= "one" Oct 29 15:57:00 what happens? Oct 29 15:57:04 does it force early expansion of BAR? Oct 29 15:57:11 normally that wouldnt expand until BAR is used Oct 29 15:57:38 well, what BAR is used with the -= Oct 29 15:57:51 ? Oct 29 15:58:04 sorry; rephrase: since user typed "BAR -=" it means he used BAR, so it should be expanded Oct 29 15:58:18 no Oct 29 15:58:43 that'll have all sorts of problematic implecations Oct 29 15:58:44 well, BAR -= "one" would mean "remove word 'one' from ${BAR}" Oct 29 15:58:52 we have lots of variables that will resolve to something useless if they're expanded at the wrong time Oct 29 15:59:10 yes, but again, the point is, it's *nontrivial*. it's not just a matter of saying here, i'll just remove a word, and call it a day Oct 29 15:59:21 there are implications, and those implications and behavior have to be documented Oct 29 15:59:35 sec brb Oct 29 15:59:49 if you want to write a patch to do so, and make sure it always does what people expect, then go right ahead, but it's not as easy as you think Oct 29 16:00:33 peter seebach recently did a patch to add removal, but he did so only as a end-of-parse operation ala _append/_prepend, specifically due to the issues I just mentioned Oct 29 16:00:57 I think it would have to be done that way if it's done at all Oct 29 16:01:18 on the whole I think it's better to avoid the need for such constructions Oct 29 16:01:31 * kergoth nods Oct 29 16:01:59 e.g. if you're fiddling with DISTRO_FEATURES you're pretty much forking the distro config... if it's a permanent change you might as well just go ahead and fork it properly Oct 29 16:03:13 and that's a totally reasonable thing to do, one distro is not going to fit everyone Oct 29 16:03:23 or rather, fit all use cases Oct 29 16:03:44 core problem is, that string is the only datatype in bitbake Oct 29 16:04:11 ensc|w: stronger typing won't solve the issue above Oct 29 16:04:46 and if it takes away the ability to constructively build up lists then it's only going to make things less flexible Oct 29 16:05:14 for distro features specifically, for easier *temporary* changes, we're using http://git.yoctoproject.org/cgit/cgit.cgi/meta-mentor/tree/classes/user_features.bbclass Oct 29 16:06:05 kergoth: interesting Oct 29 16:06:36 the trouble is, with the way DISTRO_FEATURES is evaluated it forces rebuilds of a lot of things that don't need rebuilding Oct 29 16:06:51 that *would* be helped by some support for list types within bitbake Oct 29 16:08:27 https://gist.github.com/2788410 - was just a 'hey, i wonder if this would work..' prototype, but it makes you think.. Oct 29 16:08:30 heh Oct 29 16:08:36 * kergoth thinks.. and also goes to get caffeine Oct 29 16:44:19 well, if i try to use oe_filter_out Oct 29 16:44:55 i get SyntaxError: EOL while scanning string literal Oct 29 16:47:02 bluelightning, INHERIT += "utils" does not help :) Oct 29 16:47:19 ok, I have no idea then... I assume this is OE-Classic? Oct 29 16:47:25 no Oct 29 16:47:26 oe-core Oct 29 16:47:28 early... oe-core Oct 29 16:47:42 some working commit from May this year Oct 29 16:47:51 we do use it in OE-Core in conf files and have done so for some time I think Oct 29 16:48:20 well actually... hmm Oct 29 16:48:41 no, that's wrong, sorry Oct 29 16:49:01 I think you're just going to have to set DISTRO_FEATURES explicitly Oct 29 16:49:39 if I wanted to use connman rather than the network/interfaces script, what would I remove and where in the chain is it? Oct 29 16:50:06 well, i'm trying to use python string replace, but i get SyntaxError: EOL while scanning string literal Oct 29 16:50:34 which i don't get, since DISTRO_FEATURES is expanded to a one-line string in the error message Oct 29 16:50:36 'alsa argp bluetooth ext2 ipv4 ipv6 irda largefile nfs pam pci pcmcia ppp smbfs usbgadget usbhost vfat wifi xattr zeroconf tk x11 ${DISTRO_FEATURES_LIBC}' Oct 29 16:52:02 jackmitchell, check out the netbase recipe Oct 29 16:52:15 CMoH: ah yes, I remember! thanks Oct 29 16:54:30 bluelightning, either that or add a -= to bitbake, despite kergoth's counterarguments Oct 29 16:55:35 CMoH: why not just create your own distro configuration? Oct 29 16:56:25 bluelightning, what do you mean? to fork away from angstrom completely? Oct 29 16:57:09 more or less yes; you're effectively doing that already by changing DISTRO_FEATURES Oct 29 17:05:53 Calling oe.utils.str_filter_out() on DISTRO_FEATURES is nontrivial in a .conf file for a bunch of reasons. You could make a class that does the munging and arrange for it to be INHERIT'd, though. Oct 29 17:06:36 well, i wouldn't if there was a way not to... but it seems task-base doesn't really care about what i write in the image recipe and keeps installing nfs and such Oct 29 17:09:50 If that's your only problem then you could avoid that by just not using task-base, or using a customised version of it. Oct 29 17:10:42 But in general, bluelightning is right. If you are modifying the DISTRO_FEATURES then you are now a distro maintainer and have a distro all of your own, and you should face up to this new responsibility by naming it as such. Oct 29 17:14:33 right, and then modifying the configuration becomes trivial - you just edit it in the distro .conf file. Oct 29 17:17:04 CMoH: did you looked at DISTRO_FEATURES_BACKFILL already? Oct 29 17:17:48 ensc|w, i'll have to check this out Oct 29 17:18:58 morning kergoth Oct 29 17:19:05 CMoH: I use https://www.cvg.de/people/ensc/distro-feature.txt in my distro Oct 29 17:19:24 CMoH: EOL while scanning string literal tends to occur when using a ${@} block with a ${} variable reference when that variable is undefined. i don't think anyone has looked into why it acts that way, but.. Oct 29 17:19:28 hey pb_ Oct 29 17:20:31 ty ensc|w Oct 29 17:21:04 kergoth, thanks for the details - i suspected something like that Oct 29 17:22:01 np Oct 29 17:34:49 i've implemented the -= as a dummy string replace in bitbake and it does the job... though following this discussion i'm still wondering if i should go along that path... Oct 29 17:39:23 well, I'm not sure if the case against can be stated much more clearly than it has been already Oct 29 17:53:21 it would be nice to know what distro features people are using Oct 29 19:14:35 crofton time for a survey Oct 29 19:29:48 woglinde, ? Oct 29 19:32:55 crofton -> 18:53 < Crofton|work> it would be nice to know what distro features people are using Oct 29 19:34:39 ah Oct 29 19:34:40 yes Oct 29 19:40:56 urg, what is the incantion to istall a package manager i the image Oct 29 19:41:24 EXTRA_IMAGE_FEATURES += "package-management" Oct 29 19:41:34 we ended up bbappending to core-image-base to add it there always Oct 29 19:41:38 heh Oct 29 19:41:44 * kergoth still thinks its odd that core-image-sato has it, but not core-image-base Oct 29 19:41:52 and it add opkg by defualt Oct 29 19:42:42 opkg is there because of update-alternatives Oct 29 19:43:01 core-image-base is the minimal 'base' system.. on-target package install and upgrade isn't a basic feature Oct 29 19:43:35 just annoying the first time you fall over it Oct 29 19:43:45 I'd say most of the systems my customers build will never have a package manager on them.. Oct 29 19:44:01 and I -hate- the fact that opkg provides update-alternatives.. that should be broken out.. so we can avoid opkg Oct 29 19:44:09 depends if it is a final image for product or for testing Oct 29 19:45:28 fray: we should see if alternatives.c in chkconfig will do Oct 29 19:45:49 http://git.fedorahosted.org/cgit/chkconfig.git/tree/alternatives.c#n62 Oct 29 19:45:53 seems like it should Oct 29 19:46:03 * kergoth never seems to find the time to look at it Oct 29 19:46:14 someone looked at it a while back and said why they thought it would.. they were not sure.. and we didn't have time to test it.. Oct 29 19:46:19 * kergoth nods Oct 29 21:32:10 i was just looking through angstrom's distro configuration Oct 29 21:32:36 if one takes all the required .inc files you get quite a big configuration for one distro Oct 29 22:26:29 CMoH: it's a very particular configuration for sure Oct 29 22:31:24 is there a guide on how to do this ? i mean the amount of variables that are managed there is quite high, and well... i have no clue what many of them actually do Oct 29 22:32:21 what exactly are you trying to accomplish? Oct 29 22:32:23 stupid question... of course there's no guide... Oct 29 22:33:11 well... the exact problem is the following: upon "some" upgrades involving changes to systemd services, systemd no longer starts all the units Oct 29 22:33:19 since you know, systemd is very... easy... Oct 29 22:33:25 i'm trying to remove portmap ... Oct 29 22:33:37 which is installed by the nfs distro feature Oct 29 22:33:48 distro features don't install things Oct 29 22:33:53 distro features affect how things are built Oct 29 22:34:03 it's a task or image feature that you need to find Oct 29 22:34:13 grep should do it quickly enough Oct 29 22:34:14 so far it seams that removing portmap relieves the pressure on systemd in actually doing what it's expected Oct 29 22:34:36 well... removing nfs from DISTRO_FEATURES in the image recipe didn't do the job as expected Oct 29 22:34:50 the image recipe doesn't define DISTRO_FETURES Oct 29 22:34:57 and doing so will do nothing Oct 29 22:35:08 distro features affects how things are built Oct 29 22:35:11 it doesn't affect images at all Oct 29 22:35:20 again, look for the task / image feature Oct 29 22:35:44 it'd be trivial if you'd look at bitbake -g or -g -u depexp we talked about earlier, to see what's depending on it Oct 29 22:35:51 so... i'm not supposed to derive my image from console-image, which installs task-base-extended, which brings in portmap and the like Oct 29 22:36:08 i think some angstrom guide told me to do so, so i did :) Oct 29 22:37:24 oh, i see what you meant about distro features and nfs, task-base pulls in the nfs bits based on the distro feature, that's likely true, yes Oct 29 22:37:41 so i stand corrected Oct 29 22:38:18 np - i am in so much need of help for these things :) Oct 29 22:38:26 but again, it doesn't affect images at all, so doing it there won't do it for you. you're est off manipulating distro features from the distro, from the config metadata Oct 29 22:38:30 not the recipe Oct 29 22:39:39 yeah... but the angstrom distro conf does some hocus-pocus when defining distro_features (keeps appending stuff, testing some other stuff) Oct 29 22:41:17 i just saw the effects of overwriting all these blindly in angstrom-master's gst-plugins-bad Oct 29 22:41:38 put http://git.yoctoproject.org/cgit/cgit.cgi/meta-mentor/plain/classes/user_features.bbclass in build/classes/; echo 'INHERIT += "user_features"' >>conf/local.conf; echo 'USER_FEATURES += "~nfs"' >>conf/local.conf Oct 29 22:41:49 might be easier Oct 29 22:41:54 depends on your goals Oct 29 22:42:09 well... thanks Oct 29 22:42:17 that would be cool Oct 29 22:42:21 but i'm off for today Oct 29 22:42:32 removing nfs did unlock systemd on one sd card Oct 29 22:42:56 however, the beloved system daemon decided to crash at the second test Oct 29 22:43:42 you can't imagine the hate i am accumulating for this systemd Oct 29 22:46:58 but i guess i'll fight the daemon (again) tomorrow... Oct 29 22:47:58 * kergoth`away wonders why he doesn't just switch to a distro that isn't using systemd **** ENDING LOGGING AT Tue Oct 30 02:59:57 2012