**** BEGIN LOGGING AT Thu Feb 04 02:59:59 2016 **** BEGIN LOGGING AT Thu Feb 04 04:10:39 2016 Feb 04 05:49:46 * nerdboy finding more stuff but still not sure what "it" is exactly... Feb 04 08:34:19 good morning Feb 04 08:50:51 :q Feb 04 09:49:51 Anyone had systems hanging at: systemd-journald[125]: Received request to flush runtime journal from PID 1 ? Feb 04 10:14:27 koen: this is angstrom, btw. seen anything like it? Feb 04 10:15:10 no Feb 04 10:15:13 but looks harmless Feb 04 10:15:55 koen: except that it never gets past that line. the system hangs forever. Feb 04 10:16:21 internet says other people have had this issue in the past, but I haven't found any solutions. Feb 04 10:39:15 overall, how stable/ready is systemd in OE? Feb 04 10:39:40 can it be safely considered ready for production? Feb 04 10:46:24 uh? Feb 04 10:46:41 nerdboy: ? Feb 04 11:13:58 Hi Feb 04 11:14:16 When my recipe needs bash or sh, how to define DEPENDS variable? Feb 04 11:14:35 Especially when i would like to use busybox instead of building bash. Feb 04 11:16:43 That's an interesting question. In theory it should be virtual/sh or something, but in practice I think everybody just assumes that /bin/sh is "always available" and there is no such virtual. Feb 04 11:17:32 If you need bash specifically then it's easy, you just need to DEPEND on bash itself. Feb 04 11:27:27 In fact i can use /bin/sh but bitbake will warn me that my recipe have inproper DEPENDS declaration. Feb 04 11:27:36 I wanted bitbake to be silent. Feb 04 11:27:38 :) Feb 04 11:28:14 Bitbake always warn that my recipe need bash. Feb 04 12:01:16 It's also stupid with systemd, because systemd has in DEPENDS util-linux. For example systemd needs mount. Feb 04 12:01:26 But busybox could be also mount provider. Feb 04 12:10:58 spaszPL: I've got a feeling that systemd requires the util-linux version of mount Feb 04 12:31:07 yes, it does Feb 04 14:04:25 rburton, Is systemd a virtual package? What should i do when i would like to have multiple systemd providers? Feb 04 14:05:37 no, it's not virtual Feb 04 14:05:41 use PREFERRED_PROVIDER Feb 04 14:08:07 when i make additional recipe called systemd-minimal with PROVIDES += "systemd" and i use PREFERRED_PROVIDER in my machine conf i have bitbake error Feb 04 14:08:15 ERROR: Multiple .bb files are due to be built which each provide systemd Feb 04 14:08:25 it build, but with error Feb 04 14:08:29 builds Feb 04 14:13:18 rburton, is it ok? Feb 04 14:14:00 your preferred provider is probably wrong Feb 04 14:57:19 more likely than that is you either missed another provider, or something rdepends on systemd and you don't rprovide it Feb 04 14:57:22 morning Feb 04 15:46:24 why does these 2 lines not expand to a RDEPENDS on both python-can and pytcep: RDEPENDS_${PN} += "python-can" Feb 04 15:46:27 RDEPENDS_${PN}_class-target_append += "pytcep" Feb 04 15:46:44 it just becomes "pytcep". so clearly, i'm screwing up with assignment operators OR overrides Feb 04 15:46:54 you are, yes Feb 04 15:47:09 instead of doing a conditional append, you're appending to a conditional value Feb 04 15:47:23 i.e. that appends to RDEPENDS_${PN}_class-target, which then replaces RDEPENDS_${PN} after htat Feb 04 15:47:44 you want RDEPENDS_${PN}_append_class-target, which is append to RDEPENDS_${PN} when class-target is in overrides Feb 04 15:48:18 both are occasionally of use, but i regret both existing now, too often a source of confusion, and anonymous python could fill in the gaps Feb 04 15:48:38 iirc this is covered by the bitbake manual, fyi, might be worth looking over Feb 04 15:54:44 but shouldn't RDEPENDS_${PN}_class-target_append += pytcep mean that we append to RDEPENDS_${PN}_class-target after parsing has occured? and then if class-target is in OVERRIDES, it becomes assigned to RDEPENDS Feb 04 15:55:22 no Feb 04 15:55:29 the append happens, then the override is applied Feb 04 15:56:03 but yes, your description is accurate. it appends to RDEPENDS_${PN}_class-target, which then *replaces* RDEPENDS_${PN} Feb 04 15:56:15 that's how overrides work. FOO_ replaces FOO if is in OVERRIDES Feb 04 16:00:20 https://www.yoctoproject.org/docs/1.6/bitbake-user-manual/bitbake-user-manual.html#variable-interaction-worked-examples Feb 04 16:00:26 oddly i can't find it mentioned in the yocto mega manual Feb 04 17:17:19 otavio: can you point me at the tests for this => "For boards from meta-fsl-arm-extra, the test cycle is performed by each mantainer." Feb 04 17:28:53 nerdboy: We used to have a test cycle Feb 04 17:29:10 nerdboy: people though it were not important and it has been discountinued Feb 04 17:29:23 nerdboy: please ask on mailing list so we can comment on this Feb 04 17:37:21 you want me to ask about something i've never seen? Feb 04 17:37:37 can you give me a clue about what it used to be? Feb 04 17:38:08 we can probably add the imx28 stuff to mfgtool Feb 04 17:38:18 since i have the evk... Feb 04 18:19:50 nerdboy: it was a test plan we used to have Feb 04 18:20:51 goes along any of the test images/tools? Feb 04 18:20:55 *with Feb 04 18:22:25 nerdboy: http://thread.gmane.org/gmane.linux.embedded.yocto.meta-freescale/8024/focus=8149 Feb 04 18:27:57 thanks Feb 04 18:28:16 a test plan on google docs and everything Feb 04 18:29:53 nerdboy: Daiane stopped to do this as it had weak commitment from people Feb 04 18:30:51 doesn't the linux test package thing cover most of that stuff? Feb 04 18:31:42 maybe we just need a test script/image to run it Feb 04 18:32:08 * nerdboy did that for ti camera project motherboard Feb 04 18:32:36 cheesy set of bash scripts to run for the manufacturing peeps Feb 04 18:33:18 nerdboy: I would love to see something like that Feb 04 18:33:32 nerdboy: we can trim machine-test image for thus purpose Feb 04 18:33:45 couple of adhoc tests and loopback cables Feb 04 18:36:19 so i see ltp_20150420.bb on jethro Feb 04 18:36:32 you weren't using that before? Feb 04 18:41:16 no Feb 04 21:02:19 Hi, I am relatively new to OE. I have a python module which successfully builds and installs in an image. I would also like to run the python on the build host it in a function in another module to build a custom data file. I can't seem to figure out how to access the python module on the build_host - it is in the DEPENDS of the second module. Feb 04 21:07:08 pfitz: you'll most likely need a native version of the module you need to import, so it ends up in the native python's search path, and then make sure you depend on it, and inherit pythonnative in the recipe that needs to import it Feb 04 21:07:22 pythonnative class adds the bits to the PATH so python-native can be run Feb 04 21:08:22 When I tried adding the -native, it complained about all the dependencies not having -native Feb 04 21:09:29 add BBCLASSEXTEND += "native" to the recipe with the module you need to import Feb 04 21:09:35 then its native variant will be available to build Feb 04 21:09:39 and that's what the other recip eneeds to depend on Feb 04 21:09:43 e.g. ERROR: Nothing RPROVIDES 'python-misc-native' (but Feb 04 21:10:07 that's.. odd. worst case you can override your deps for native to just python Feb 04 21:10:19 ie. DEPENDS_class-native = "python-native", etc Feb 04 21:10:25 empty the rdepends Feb 04 21:10:39 Or do I need to use ${BPN} in my RDEPENDS_ lines Feb 04 21:11:01 not a good idea in general, that'd break multilib, nativesdk, etc Feb 04 21:11:08 better to explicitly deal with native Feb 04 21:11:43 even if you have no intention to use those, it'd good policy and a good habit to develop, and you may well want to support nativesdk eventually if you want to add it to an sdk Feb 04 21:13:27 I am not sure I am following your "good habit" comment Feb 04 21:14:46 (I am somewhat fumbling in the dark, trying to learn OE from google and spelunking through OE .bb and .bbclass files) Feb 04 21:14:48 the correct pattern to follow is always use ${PN}, not ${BPN}, as the package name in RDEPENDS_ lines Feb 04 21:14:55 and override them in the native case where necessary Feb 04 21:21:55 OK, I have the files installed now in sysroot-destdir, but no PYTHONPATH var is set. Do I need to do that manually? Feb 04 21:25:49 kergoth, thanks for the help so far - unfortunately I have to get my daughter from daycare - I'll keep plugging away Feb 04 21:26:05 you don't need PYTHONPATH if you inherit pythonnative Feb 04 22:24:39 kergoth, you have gotten me so close. All I am missing are the python dependencies, like python-xml. How do I get those provided? Feb 04 22:26:06 if you need to import those at build time, add the appropriate reicpes to DEPENDS Feb 04 22:28:54 I need them available at runtime Feb 04 22:32:42 runtime doesn't mean much Feb 04 22:32:46 build time, or on target? Feb 04 22:32:57 if the former, DEPENDS, if the latter, RDEPENDS_${PN} Feb 04 22:34:02 On the host that is building. I have a function like do_something() { ${PYTHON} -m custom.module }, where custom.module is part of the python I am installing, and it uses python-xml Feb 04 22:34:12 as i said, add to DEPENDS Feb 04 22:34:31 Whenever I add to depends, it fails with the -native not provided by anyone Feb 04 22:34:56 ERROR: Nothing RPROVIDES 'python-xml-native' (but virtual:native:/home/pfitz ... Feb 04 22:35:07 does the recipe exist? does the recipe add native to BBCLASSEXTEND? Feb 04 22:35:12 if not, append it and fix it Feb 04 22:35:44 It would be the standard OE python recipes. I have not dug into them very far. Feb 04 22:36:02 (Which is where I was going next ) Feb 04 22:36:24 the xml module is in python itself, there's no python-xml recipe to depend on, last i checked Feb 04 22:36:38 just depending on python-native should work, unless python-native wasn't built with xml support Feb 04 22:38:12 I can RDEPEND on it for the standard non-native recipe RDEPENDS_${PN} += "python-math python-datetime python-logging python-xml" works fine Feb 04 22:38:59 yep, that's fine Feb 04 22:39:16 rdepends is for binary packages, native recipes don't emit binary packages Feb 04 22:43:28 Thanks, this is helping clear things up, and I think I have it working Feb 04 22:43:54 np Feb 04 22:44:06 On another topic, is it possible to call a function in another recipe? Feb 04 22:44:15 nope, recipes are isolated Feb 04 22:44:22 can't touch one naother's metadata directly Feb 04 22:44:58 I have another recipe which is responsible for installing public keys on the image. In the recipe I am working on I need to access an associated private key to sign the data. Feb 04 22:45:27 If I can't call into the other recipe, do I need to have the other recipe install the private key into the deploy hierarchy (or something similar?) Feb 04 22:45:32 sounds like you want to put the keys in the sysroot and DEPEND on it Feb 04 22:45:53 by default only certain paths go from ${D} into the sysroot, so you'd have to append to sysroot_stage_all Feb 04 22:45:55 (see staging.bbclass) Feb 04 22:46:33 That would make much more sense. I can have the other recipe add a PROVIDES for a private key package and install it Feb 04 22:47:04 (Since the private key would not go into most builds) Feb 04 22:55:58 Thanks again, kergoth, I am going to work that issue tomorrow Feb 04 22:58:31 np **** ENDING LOGGING AT Fri Feb 05 02:59:59 2016