**** BEGIN LOGGING AT Fri May 23 03:00:01 2014 May 23 07:32:07 morning all May 23 07:37:48 gm bluelightning May 23 07:49:19 morning mckoan May 23 07:49:32 hey mckoan & bluelightning May 23 07:49:39 hi koen May 23 08:05:16 aargh bluelightning help me... May 23 08:05:27 pompomJuice: what's up? May 23 08:05:29 apk_base_postinst_chrome() { May 23 08:05:41 I dont see the script in the ipk May 23 08:05:43 it wont go May 23 08:05:56 recipy is chrome_38.7.bb May 23 08:06:35 is something like this legal? May 23 08:06:36 pk_base_postinst_${PN}() { May 23 08:07:10 shouldn't it be pkg_postinst_${PN} () { ? May 23 08:07:15 amagad May 23 08:07:18 ok lets see May 23 08:08:34 still not there May 23 08:08:56 inside the IPK there is this debian folder that is supposed to get the script May 23 08:09:01 well, that is the correct syntax May 23 08:09:02 it works for other packages May 23 08:09:22 in my system so it's just this one.. no other level I must pull that you can think of? May 23 08:13:30 pompomJuice: I just tried it here - I added pkg_postinst_${PN} to a recipe that I'd already built, built it again, and unpacked the ipk - the postinst file is there inside control.tar.gz within the ipk May 23 08:13:52 I did not notice the pkg_ May 23 08:13:54 let me try again May 23 08:14:01 the start differs May 23 08:14:11 right, that would stop it from working... May 23 08:15:26 it works! May 23 08:15:27 amagad May 23 08:15:30 thanks May 23 08:16:09 np May 23 08:16:20 i was cloning some other recipies, there was an anonymous python hack that does this -> d.setVar("pkg_postinst_%s" % pkg, d.getVar("apk_base_postinst", True) % apk) May 23 08:16:23 and it works! May 23 08:16:36 so no idea by apk_base_postinst works there... May 23 08:16:37 bizarre May 23 08:17:09 I don't think it does, that code is reading apk_base_postinst and setting pkg_postinst_ from it May 23 08:17:10 I wrote a system that installs APKs through opkg :) May 23 08:17:25 amagad May 23 08:17:35 oops May 23 08:17:40 take it slow May 23 08:18:37 did not see that the first part was not the same May 23 08:18:43 not easy May 23 08:19:44 I'm happy to take suggestions on how we can document this better May 23 08:20:35 maybe start by explaining in some document how bitbake->oe->angstrom->layers relate May 23 08:20:47 im so lost sometimes May 23 08:21:10 but the documentation is good, just you dont get time to read it May 23 08:21:15 there is a "Post-Installation Scripts" section in the development manual (under the new "Writing a New Recipe" section) May 23 08:21:15 shit must be finished today May 23 08:21:30 let me check May 23 08:21:35 my main issue is the following May 23 08:21:39 all the documentation is split May 23 08:21:44 I want one document May 23 08:21:48 the macdaddy of documents May 23 08:21:51 aha, then you are in luck :) May 23 08:21:55 where I can do a keyword search May 23 08:21:56 check out the "mega manual" May 23 08:22:02 awesome May 23 08:22:19 equally though I don't mind answering questions here, even if they are in the manual May 23 08:22:42 thats just pure lazyness then May 23 08:22:49 please give me a link to this manual May 23 08:22:54 I'm not just going to say "RTFM" although I may point to sections of the manual if they explain some aspect in greater detail than I feel like typing ;) May 23 08:23:00 one sec May 23 08:23:25 http://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html May 23 08:23:34 awesome thanks May 23 08:23:45 that's the 1.6 manual (the current release) May 23 08:23:50 the 1.5 manual is here: http://www.yoctoproject.org/docs/1.5/mega-manual/mega-manual.html May 23 08:24:03 omg its huge May 23 08:24:05 lol May 23 08:24:12 I don't feel bad anymore May 23 08:24:12 it does say "mega" ;) May 23 08:24:29 no wonder im so lost May 23 08:25:36 I don't think we expect people to read the entire thing May 23 08:25:48 yea it's not always neccesary May 23 08:26:09 if the sources are able to be built by autotools it's a no brainer May 23 08:26:15 the recipies are delf explanitory May 23 08:26:18 self* May 23 08:27:57 this is going to be awesome May 23 08:28:29 I hope that the new "Writing a New Recipe" section will be helpful... it's not in the 1.5 manual, it's new in the 1.6 version (but equally applicable to older versions) May 23 08:28:46 awesome May 23 08:29:14 im gonna check it out def, since the only oother writing a new recipy in the interwebs is so old May 23 08:29:32 maybe include some templates May 23 08:29:33 like May 23 08:29:37 autotool template May 23 08:29:41 cmake May 23 08:29:45 etc May 23 08:30:11 and one that shows different maneuvers like the ipk post install... May 23 08:35:32 maybe I have a legit question now May 23 08:35:42 sure, fire away May 23 08:35:52 I am busy adapting a recipy... May 23 08:36:02 in this recipy we have these post insttall scripts May 23 08:36:27 and one of the scripts contained: May 23 08:36:27 apk_base_postinst() { May 23 08:36:28 #!/bin/sh May 23 08:36:28 APK="%s" May 23 08:36:34 that %s May 23 08:36:36 what is that? May 23 08:36:39 how does that work? May 23 08:36:44 it makes no sense to me May 23 08:36:57 aah wait May 23 08:37:22 its a placeholder for the python line that inserts it May 23 08:37:26 sorry man May 23 08:37:28 plastic programming May 23 08:37:30 right May 23 08:37:32 no worries :) May 23 08:37:34 I solved it now May 23 08:37:58 so my question becomes May 23 08:39:33 can I insert bitbake variables into those scripts... May 23 08:39:41 testing it now... May 23 08:39:47 the manual does not say May 23 08:40:03 you can yes, just ${VARIABLENAME} May 23 08:40:25 since it seems that the reason why the previous guy went with the anonymous python is that he could not inject bitbake variables into his post install script May 23 08:40:30 aweome May 23 08:42:06 whoever invented bitbake is genius... this thing has no end May 23 08:44:03 it's very flexible for sure May 23 08:44:28 very, the things we have automated in our image is insane May 23 08:44:30 of course that does lead to a bit of a learning curve, but it's powerful once you get to grips with it May 23 08:44:35 we have a very unconventional image May 23 08:44:42 and we are using bitbake on all levels May 23 08:45:23 bitbake just sorts out all our problems May 23 08:47:52 are you actually packaging android stuff using bitbake ? May 23 08:47:58 (just being curious) May 23 08:50:58 yes May 23 08:51:07 the dudes before me did something strange May 23 08:51:18 they installed android ontop of linux May 23 08:51:36 instead of just adroid with gimped linux May 23 08:51:43 so we have a device May 23 08:51:55 with full angstom capability... but runs android May 23 08:52:10 things like systemd May 23 08:52:11 etc May 23 08:52:29 makes it then a very powerfull device that can easily be augmented will all kinds of trickery May 23 08:53:12 our android is in a jail May 23 08:53:29 pretty cool May 23 08:53:45 the android we use is rowboat, since it's a ti chip May 23 08:54:56 so you have an androidish kernel ? May 23 08:54:58 therefor the recipies I am writing now enables you to opkg install android apps May 23 08:55:05 yes the kernel is the bugger May 23 08:55:10 when they started this May 23 08:55:16 t4 years ago May 23 08:55:31 ti did not have a android kernel, so they took one and patched it May 23 08:55:37 locking the kernel in place May 23 08:55:43 that number is 2.6.37 May 23 08:55:45 pretty annoying May 23 08:55:57 so that means I can only take systemd to 204.. shit like that May 23 08:56:43 I could try find a new kernel that does support linux, and re-apply the patches... but how do you motivate that to the upper management. They don't see the value in keeping versions up to date May 23 08:57:36 the other tricky part was getting the to SGX 530 gl driver module into the kernel May 23 08:57:43 and getting android to use it May 23 08:59:44 you get werid issues May 23 08:59:45 like May 23 08:59:52 the other day I went from linaro 4.7 4.8 May 23 08:59:58 bam kernel brick May 23 09:00:04 then solved by patching memset May 23 09:00:09 bam sgx driver brick May 23 09:00:40 memcpy from syslink no good anymore May 23 09:01:09 things like that May 23 09:04:27 uugh I have another issue May 23 09:04:50 I wronte this mini recipy android-utils that contain 2 !#/bin/sh scripts May 23 09:04:53 wrote* May 23 09:05:05 the ipk contains the scripts at /usr/bin May 23 09:05:19 and the device says it is installed May 23 09:05:24 but the files are not there May 23 09:05:36 should I take special precaution in recipy for script targets? May 23 09:06:07 pompomJuice: no, no special handling should be needed May 23 09:06:16 mmm ok thanks. May 23 09:06:29 what does the package manager list when you ask it to list files within that package? May 23 09:06:40 checking May 23 09:08:16 strrange May 23 09:08:36 ./usr/bin/adb May 23 09:08:36 ./usr/bin/aapt May 23 09:08:43 now I am confused May 23 09:09:18 omg May 23 09:09:42 another package adds PACKAGES += "android-utils" in another package May 23 09:09:47 its a duplicatge May 23 09:09:55 bizarre May 23 09:10:18 my package name exists... thats why... May 23 09:10:21 ok let me fix it May 23 09:10:37 ah right May 23 09:11:00 though that should have triggered a warning, unless it happened that only one of the recipes actually got built May 23 09:12:01 yea May 23 09:12:06 was only building the one May 23 09:12:20 interesting though May 23 09:12:49 I have that recipy as DEPENDS="android-utils" in my chrome recipy... and it built my new one before it built the chrome recipy May 23 09:12:55 so I thoguht it worked May 23 09:48:38 RDEPENDS_${PN} = "android-install-utils" May 23 09:48:44 does not seem to work May 23 09:48:48 :/ May 23 09:49:18 is android-install-utils a valid package name? May 23 09:49:37 its recipy is android-install-utils_0.1.bb May 23 09:49:56 that version number is purely fictitious May 23 09:50:06 ok, and that recipe produces a normal main package that is non-empty? May 23 09:50:11 yes May 23 09:50:17 what fails then? May 23 09:50:20 does android-install-utils_0.1.bb generate a 'android-install-utils'? May 23 09:50:20 my other rdepends works May 23 09:50:21 well May 23 09:50:30 it's like the dependency is not there May 23 09:50:38 ? May 23 09:50:50 even though the chrome ipk contains in it's info a line: May 23 09:51:03 can you be more specific? May 23 09:51:17 Depends: android-install-utils May 23 09:51:29 that is inside the chrome.ipk's info file May 23 09:51:43 what's wrong then ? May 23 09:51:59 but when I do opkg install chrome it does not pull the android-install-utils first... May 23 09:51:59 and you don't have anything RPROVIDE'ing or RREPLACES'ing android-install-utils ? May 23 09:52:11 let me check that one too May 23 09:52:26 grep android-install-utils /var/lib/opkg -rn May 23 09:53:24 does not seem like it May 23 09:53:35 let me double check this May 23 09:53:40 since other rdepends in our system works May 23 09:54:18 yea May 23 09:54:28 opkg uinstall chrome May 23 09:54:35 opkg uninstall android-install-utils May 23 09:54:42 opkg install chrome May 23 09:54:50 worked May 23 09:54:59 sorry May 23 09:55:36 sounds like the chrome package didn't get upgraded on the target to include the new dependency May 23 09:55:45 yea May 23 09:55:52 I made a weird discovery May 23 09:56:03 you can install something for the first time by saying opkg upgrade chrome May 23 09:56:08 it was habit May 23 09:56:33 when it started downloading the chrome ipk I ctrl-C it thinking the deps dont work May 23 09:56:38 but it downloads the dep second May 23 09:56:39 amagd May 23 09:56:43 upgrade will work after that, but the version number has to change or it'll think nothing has tobe done May 23 09:56:49 I see May 23 09:56:59 install --reinstall May 23 09:57:06 hm, --force-reinstall May 23 09:57:18 yea that one May 23 09:57:52 and it works May 23 09:57:55 everything works May 23 09:57:58 :) May 23 09:58:29 sorry for going for the cheap route by asking questions here I was under time pressure it's easier to come plastic program here May 23 09:58:30 thanks all May 23 09:58:38 it works May 23 09:58:41 boss man is happy May 23 09:58:46 I can relax again May 23 09:58:50 now you can play around with OE ;) May 23 09:58:57 yea May 23 09:59:28 one interesting I notice though May 23 09:59:45 my pkg_postinst script contains echo "some status" to give some feedback May 23 09:59:56 the feedback is printed twice, but I am sure the script only runs once May 23 10:00:05 unless my dupcheck wored May 23 10:00:10 worked* May 23 10:00:12 anyway May 23 10:02:22 yea the script executed only once otherwise my dupcheck echoes would have printed.. May 23 10:13:29 koen legend btwe May 23 10:13:41 everyone in this office knows the name May 23 10:13:46 fyi May 23 10:13:49 ;) May 23 10:53:42 ah nice the google-pdf plugin May 23 10:53:50 now we can make it run under arm too May 23 10:54:00 for chrome/ium of course May 23 11:45:36 I was wondering, these pkg_postinst that you can add... how are those handled when you create the image? Those scripts of mine need to be deferred, they cant be switched since some of the post install commands are adb shell commands to the android system. Those won't work on image generation... is there something built in for this? May 23 11:46:41 pompomJuice: if it can't run safely, there usually is a -z $D check present to abort if opkg -o /path/to/root is used May 23 11:46:58 for some postinsts OE will try to run them under qemu or another wrapper May 23 11:47:11 e.g. gtk-pixbuf and icon cache postinsts May 23 11:47:24 yea I see my script is switching on $D like you said May 23 11:47:53 ok so am I to assume that those scripts are run and they fail? Or what? May 23 11:47:57 postinsts are just shellscripts opkg will run with some extra vars set (like $D) May 23 11:48:14 are they run when the image is generated May 23 11:48:15 ? May 23 11:48:17 surely May 23 11:48:30 what happens if they fail or "hang"? May 23 11:48:36 ok wait May 23 11:48:39 they are switching on 4D May 23 11:48:43 $D May 23 11:48:46 so they wont hang May 23 11:48:48 nm May 23 11:48:51 if they fail they are postponed until first boot May 23 11:48:58 aah May 23 11:49:10 that could come in handy May 23 11:49:17 exactly what I need May 23 11:49:21 top top May 23 11:49:23 will test it now May 23 11:49:43 iirc 1.5 introduced the qemu and wrapper stuff May 23 11:49:56 So this one IPK installs a android keyboard APK, that needs a post install command adb ime enable command that can only run on the meter May 23 11:50:14 interesting... that qemu move May 23 11:50:36 must you have a qemu configuration ready? May 23 11:50:37 if you want the script to be postponed when $D is set you must "exit 1" in that case so that it appears to fail May 23 11:50:51 thanks bluelightning May 23 11:50:57 testing that now May 23 12:19:23 hello, anyone knows if binary patch (GIT binary patch) is applied on bitbakes? I have a patch witch contains images and text, but only text is applied (using SRC_URI += ...) May 23 12:22:31 athoik: it won't be normally, no May 23 12:22:47 athoik: you may be able to get it to apply if you set PATCHTOOL = "git" within the recipe May 23 12:23:27 and if that fails, add the binaries to SRC_URI and overwrite the originals in do_configure_prepend May 23 12:23:37 do_unpack is python, right? May 23 12:24:11 yep May 23 12:24:52 bluelightning: thanks i will test first PATCHTOOL = "git" if that doesn't succed i was thinking to include photos in files and copy then manualy :) May 23 12:30:14 bluelightning: PATCHTOOL = "git" did the trick! really\ thanks a lot! May 23 12:32:45 athoik: np May 23 16:01:29 How can I track a patch as it goes through the oe-core sausage making machine? May 23 16:21:30 Crofton|work: sorry, what's the context? May 23 16:21:51 I sent a patch in, trying to figure out how I can follow its path into oe-core master May 23 16:22:05 so I know when to ask it go into diane May 23 16:22:16 daisy? dora? dylan? May 23 16:22:24 daisy May 23 16:22:33 to many d words May 23 16:22:51 I guess that's what patchwork is supposed to help with, unfortunately it's not very good at marking merged patches for OE-Core May 23 16:27:11 I need to make sure it doesn't fall on the floor, since I have a meta-oe patch that depends on it May 23 16:34:31 beyond patchwork and the mailing list we don't really have anything May 23 16:36:38 set up a cron job to update a clone and git log —grep= it for your commit periodically? ;) May 23 16:37:07 well, we do have the oe-commits list as well, that's probably easier than that ;) May 23 16:40:45 * kergoth sets up git url aliases to redirect oe-core -> openembedded-core and meta-oe -> meta-openembedded for the github repos May 23 20:01:56 hmm so it seems the layer index wasn't updating properly for the master branch since the server was moved May 23 20:02:12 mainly because I forgot to switch it to use the proper branch of bitbake May 23 20:02:25 I've now fixed it and reloaded all of the data May 24 02:44:21 is anybody working on a sqlite fix? if not i should have a patch after dinner sometime... **** ENDING LOGGING AT Sat May 24 02:59:58 2014