**** BEGIN LOGGING AT Tue Jul 22 03:00:01 2014 Jul 22 08:11:03 good morning Jul 22 08:29:10 gm Jul 22 08:35:39 morning all Jul 22 08:41:18 morning bluelightning Jul 22 08:41:24 hi anarsoul Jul 22 08:52:48 hi bluelightning, woglinde, all **** BEGIN LOGGING AT Tue Jul 22 09:55:14 2014 Jul 22 10:09:31 howdy all Jul 22 10:10:00 where in a recipy can I hook into the success cleanup state? Jul 22 10:10:10 do_work_rm() { Jul 22 10:10:11 ? Jul 22 10:10:21 donno how to search that in the manualk Jul 22 10:11:09 rogue2 what is your problem you want to solve? Jul 22 10:11:50 well Jul 22 10:12:14 I have a very manual recipy that needs to do cleanup after it has succeded... or after the do_install step Jul 22 10:12:32 you can add a custom task after rm_work, bitbake should only trigger it if rm_work succeeded Jul 22 10:12:40 (keep in mind rm_work is a bit special) Jul 22 10:12:41 perfect Jul 22 10:12:45 oh noes Jul 22 10:12:51 I understand Jul 22 10:13:02 so dont mess with rm_work, just add a step after Jul 22 10:13:04 got it Jul 22 10:13:06 developing... Jul 22 10:13:30 a custom class is probably easiest Jul 22 10:13:37 have a look at the rm_work classes Jul 22 10:13:45 not familier with how a class would do that Jul 22 10:13:58 foo() { stuff} Jul 22 10:14:06 addtask foo after rm_work before do_build Jul 22 10:14:11 or something like that Jul 22 10:14:14 ok Jul 22 10:14:23 it's been a while since did that stuff :) Jul 22 10:14:39 forgiven Jul 22 10:16:40 hi I am porting rdesktop from oe-classic (1.8.2) but I have an error during do_configure unrecognized command line option '-V' during configure do you have any hint? http://pastebin.com/isidqeNh Jul 22 10:17:06 mckoan why not look at configure.in/ac Jul 22 10:17:21 maybee rdesktop does something wrong Jul 22 10:17:46 woglinde: thx, it definitely does something wrong Jul 22 10:51:42 wow, what is that patchset? Jul 22 10:51:43 fix rdepends on bash perl and python Jul 22 10:53:26 xz: add bash to RDEPENDS_xz ?? Jul 22 11:47:18 bluelightning: hi, ^ Jul 22 11:47:47 same as http://lists.openembedded.org/pipermail/openembedded-core/2013-December/087160.html Jul 22 11:47:51 http://lists.openembedded.org/pipermail/openembedded-core/2013-December/087161.html Jul 22 11:49:20 debian doesn't list such runtime deps, nor does Gentoo Jul 22 11:51:25 please have a look before we end up with bash, perl and python on target Jul 22 11:52:07 ant who made the patch? Jul 22 11:53:00 http://comments.gmane.org/gmane.comp.handhelds.openembedded.core/53509 Jul 22 11:53:49 seems it is Jul 22 11:53:49 [YOCTO #1662] Jul 22 11:56:04 ant_work: there have been a lot of patches that added perl to RDEPENDS Jul 22 11:56:11 I think that if some sources do contain specific perl or python script these should be packaged separately Jul 22 11:56:18 ant_work: queries to move the perl bits to their own packages were ignored Jul 22 11:56:22 e.g. with i2c-tools Jul 22 11:56:24 :/ Jul 22 11:57:09 is 1662 one of those "someone looks at RPM output and sends a dozen patches without understanding the issue" ? Jul 22 11:57:13 I would check the dependencies matching with packages.debian.org Jul 22 11:57:53 or even being more granular Jul 22 11:58:33 well Jul 22 11:58:36 ant answer the patch Jul 22 11:58:38 koen: it has to be said, on typical x86 there are bash, perl and python Jul 22 11:58:50 the first step is to look at the source and figure out if the dependency is Jul 22 11:58:52 a) real Jul 22 11:58:53 ant on desktop maybee Jul 22 11:58:53 2) needed Jul 22 11:59:06 woglinde: yes, the POV of many devs Jul 22 11:59:15 I have bash and perl installed, mostly due to git Jul 22 11:59:30 perl is essential on debian Jul 22 12:00:13 koen uhm which git scripts have bashims? Jul 22 12:00:17 well, we just don't have space for them in some cases Jul 22 12:00:50 and I doubt git is normally installed on target images Jul 22 12:00:51 ant do not tell me I know I send patches to perf to rip off python and perl from core Jul 22 12:01:03 ;) Jul 22 12:01:11 sent even Jul 22 12:01:34 woglinde: busybox sh drives me insane, I don't need extra anguish when I'm dealing with git :) Jul 22 12:02:47 .oO( git applet into busybox ) Jul 22 12:04:35 koen okay Jul 22 12:04:48 I thought there were some bashims Jul 22 12:38:58 sent Jul 22 12:50:59 woglinde, bluelightning: rdesktop_1.8.2.bb built, thx Jul 22 12:51:21 mckoan: awesome... hopefully you can submit it somewhere as well? ;) Jul 22 12:52:08 bluelightning: I needed to add INHIBIT_PACKAGE_DEBUG_SPLIT = "1" though Jul 22 12:53:32 bluelightning: I have to find the time to read the submission rules Jul 22 12:55:42 rdesktop, that's blast from the past Jul 22 13:00:51 koen: someone still love fine aged packages ;-) Jul 22 13:02:48 * koen stuff OE-classic in an oak barrel Jul 22 13:03:36 koen: LOL Jul 22 13:22:55 about bitbake, is there any gui configure like menuconfig in buildroot? Jul 22 13:24:09 you do not need one Jul 22 13:24:57 ayaka: if you're feeling brave, Toaster Jul 22 13:25:06 it's under development so you'd be an early user Jul 22 13:25:21 there is Hob but it's unmaintained now that the focus is on Toaster Jul 22 13:25:22 rburton: it doesn't actually let you do configuration of the build though, only examine the results Jul 22 13:25:42 thats the caveat i was about to say, yeah Jul 22 13:26:05 because the doc of bitbake is too long Jul 22 13:26:28 no it isnt Jul 22 13:26:32 I have a read that doc but I can't make a target for my board Jul 22 13:26:33 that is good Jul 22 13:26:53 what is your board? Jul 22 13:27:05 my board is samsung exynos 4412 Jul 22 13:27:09 and there is no graphical tool for a board support package Jul 22 13:27:14 to create Jul 22 13:27:41 I don't event know how to select packages now Jul 22 13:28:25 maybe I need some time to convert from buildroot(openwrt) to oe Jul 22 13:28:26 packages are selected by image Jul 22 13:28:37 and you create a feed Jul 22 13:28:46 hm I would start with http://layers.openembedded.org/layerindex/branch/master/layer/meta-exynos/ Jul 22 13:29:08 and look into how other exynos bsp were made Jul 22 13:29:18 and yes oe works diffrent Jul 22 13:29:44 ayaka: the yocto manual has a section on creating and customising images, including how to pick what goes into an image. Jul 22 13:29:58 and don't ask me about exynos4 never seen one Jul 22 13:30:45 yes I use exynos4 Jul 22 13:31:55 how doc. I only know I need to read doc about bitbake but nothing else, both oe and yocto doesn't have a index of docs Jul 22 13:32:20 XorA: no Chromebook there ? Jul 22 13:32:33 ayaka: did you try the Documentation link on the yocto site? https://www.yoctoproject.org/documentation/current Jul 22 13:33:14 mckoan: chromebook is exynos5 Jul 22 13:33:23 XorA: yep, my fault Jul 22 13:33:54 rburton, thank I was finding user manual Jul 22 13:34:48 I enter the yocto by the link in wiki of yocto, I don't know it has a different view of that site Jul 22 14:17:14 this db6 business is annoy Jul 22 14:17:17 annoying Jul 22 14:17:28 is the C++ compile failing? Jul 22 14:17:40 no it blows up ice though Jul 22 14:18:01 and I am off on vacatoin Jul 22 14:18:11 I am wondering how many oher packages have issues with it Jul 22 14:18:14 well change your preferred to 5.3 is all I can suggest.. :( Jul 22 14:18:43 people had been asking for the db6 for a while now, and RPM was holding it back.. that was resolved and now we've got other issues.. (including the license change to AGPL-3.0) Jul 22 14:19:46 heh Jul 22 14:19:48 lovely Jul 22 14:20:22 tryin gto skim the ice phorums Jul 22 14:20:31 cat is in the way Jul 22 14:23:19 no matches for searching for db6 Jul 22 16:26:23 So: I would like to extract a deploy/ipk tarball on top of an existing build with its deploy/ipks. But after I do that, the next rebuild, it appears that all those packages get regenerated via do_package_ipk_setscene. How do I inhibit that, for just those packages? (And how on earth is bitbake figuring out when to setscene? Couldn't figure that out last night) Jul 22 16:28:23 bitbake tracks what tasks its run and what it hasn't Jul 22 16:28:34 it maintains stamps, and pulls prebuilt binaries from the sstate cache Jul 22 16:28:40 none of that pulls *from* deploy/ipk Jul 22 16:28:45 deploy/ipk is output, not input, to a bitbake build Jul 22 16:28:56 unless you configure an image to populate directly from an ipk feed and point it to the feed Jul 22 16:29:26 no, I'm reproducing this solely from `bitbake packagegroup-foo` Jul 22 16:30:12 although, now that you mention it, I *do* notice that if I run bitbake multiple times in a row, the 2nd and 3rd iterations often run setscene tasks -- they don't do nothing Jul 22 16:30:16 then bitbake will either build everythng from scratch, or if available, pull sstate archives from your sstate cache, both of which will blow away deploy/ipk, as as i said, deploy/ipk is output, not input to the build Jul 22 16:30:34 setscene tasks are pulling cached output of the associated task Jul 22 16:30:38 from sstate-cache Jul 22 16:31:40 so the overwriting of deploy/ipk likely has nothing to do with me overlaying files on top of it -- for whatever reason, bitbake is doing it on its own Jul 22 16:33:01 as i've explained 3 or 4 times now, bitbake keeps track of what tasks need building and what don't, and maintains its own binary cache. deploy/ipk isn't definitive, and it doesn't contain everything bitbake needs to complete a build Jul 22 16:33:20 putting your own files there will basically do nothing unless you first complete your bitbake build, *then* overwrite what bitbake has done Jul 22 16:34:30 you mean, as opposed to overwriting the files during a bitbake build? Jul 22 16:36:18 what exactly are you trying to accomplish here? Jul 22 16:36:56 if you want to build an image from an ipk feed, we provide a mechanism to do that independent of the usual build process. if you want to build from a cache rather than from scratch, then you should save SSTATE_DIR and/or set up an sstate mirror, not mess with deploy/ipk Jul 22 16:37:16 in either case, there's a better option than what you're doing now :) Jul 22 16:37:28 I'm building a feed server by rsyinc'ing the contents of deploy/ipk. That's a valid way of doing it, right? Jul 22 16:37:56 yes, but bitbake doesnt' build from a feed server unless you configure it to build an image from a feed Jul 22 16:37:58 which you haven't done Jul 22 16:38:11 so yes, its' rebuilding as it usually does, ignoring what you've done to deploy/ipk Jul 22 16:38:21 I'm not building from a feed server and I'm not building an image. I'm essentially building world. Jul 22 16:38:30 that really doesn't make any sense Jul 22 16:38:40 why would you write to deploy/ipk then? Jul 22 16:38:44 it's accomplishing nothing Jul 22 16:39:13 because our released/shipped hardware *is* built from images, from a totally differnet build machine, with *no* shared sstate Jul 22 16:40:04 so, if I naïvely build my repo, then e.g. opkg update;opkg install libc6-dbg on the target, the CRCs between libc6-dbg and glibc don't match. Jul 22 16:40:26 this is exactly what sstate is for. if you want to be able to reproduce bit-for-bit what's come out of another build, then use the sstate cache Jul 22 16:40:27 and in particular, call stacks in glibc, and thread debugging in general don't really work Jul 22 16:40:31 otherwise build your image from the feed Jul 22 16:42:43 The sstate cache on the build machine is deleted on every image build. :F It was a special request of mine just to have deploy/ipk archived off. Looks like I incorrectly chose what I needed to save off. Jul 22 16:43:41 yep, sounds that way. the purpose of sstate is to let you reproduce builds, and avoid having to rebuild everything from scratch. the ipk feed is a distribution mechanism to the device, or a way to build images from it, that's about it Jul 22 16:44:04 at least I can still overwrite deploy/ipk + opkg-make-index like I have been -- but it sounds like I need to keep it entirely outside bitbake's clutches Jul 22 16:45:15 I think I have enough to go on, thanks :) Jul 22 16:45:18 pretty much. like i said, deploy/ipk is output from bitbake's perspective, not input, with the one exception of image/rootfs creation. Jul 22 16:45:21 np Jul 22 18:14:36 how do I install makeinfo on ubuntu Jul 22 18:14:41 I am in a psecial place .... Jul 22 18:14:51 texinfo package Jul 22 18:15:03 for future reference: sudo apt-get install apt-file; apt-file update; apt-file search makeinfo Jul 22 18:18:38 kergoth does ubuntu no have this feature where you type a command and it can be found ubuntu says which package you could install Jul 22 18:19:51 that's a good point, i think so, but it probably depends on the install Jul 22 18:24:58 yeah Jul 22 18:25:02 this was a panic Jul 22 18:25:21 power supply fail in the machine I already had setup at a customers Jul 22 18:27:04 we should probably change the sanity test message to mention texinfo. makeinfo is the tool, but texinfo is the project/package on pretty much all distros Jul 22 18:28:48 yes **** ENDING LOGGING AT Wed Jul 23 03:00:01 2014