**** BEGIN LOGGING AT Mon May 13 02:59:56 2019 May 13 07:23:00 Is there a quick way to delete the downloaded sources of a single recipe without manually removing them from the download directory? May 13 07:41:24 armpit: yes May 13 07:52:48 frsc: You can use the cleanall task if you're happy to also throw away the sstate and work directory for that recipe May 13 07:54:07 paulbarker: thanks, that's what I was looking for May 13 09:30:15 when working with kernel config fragments, do i need to add my fragment to KERNEL_CONFIG_FRAGMENTS in addition to an entry in SRC_URI? May 13 10:01:32 milloni, I can't find any reference to KERNEL_CONFIG_FRAGMENTS in oe-core.. is that a meta-poky thing ? I've only ever needed to put them in SRC_URI May 13 10:09:58 kroon: see https://gerrit.automotivelinux.org/gerrit/gitweb?p=AGL/meta-agl.git;a=blob;f=meta-agl-bsp/meta-intel/recipes-kernel/linux/linux-intel_4.19%25.bbappend;h=9827b3e7dfd0fe40ae9326ead69ace98e347f5aa;hb=HEAD May 13 10:20:15 milloni, I can't make sense of it. I guess you could use "bitbake -e " to check if it is ever referenced, at least for the kernel build May 13 11:01:07 seems like toaster is broken, or I guess I have to change to ubuntu 18 May 13 11:01:15 Failure while trying to setup toaster: Problem installing fixtures: no such table: orm_build__old May 13 11:28:41 RP: reading your status email; build artefact reuse is when something was built and was determined to be in fact equivalent to a previous cache item, we can then reuse the 'tree of cache items' at that point, instead of build the items anew? May 13 11:28:55 that'd be awwwwsome! May 13 11:31:53 kanavin_: yes, that would be the idea May 13 11:37:15 kanavin_: remembered to reply about webkit May 13 11:41:09 RP: right, there's just one point I am not able to understand. Let's say we changed recipe A, that forced it be rebuilt. Then recipe B, which depends on A was rebuilt and determined to be equivalent to a previous build of B. Then we have a recipe C, that depends on B. I don't think we can then just take it from the cache, as it has A in its sysroot, so where is the win? What is the scenario where we can short cut the build? May 13 11:42:52 kanavin_: we can just take C as it was as B was unchanged? May 13 11:43:24 yeah, but A was changed, and A is present in C's sysroot, so we can't assume changing A has no effect on C? May 13 11:45:14 kanavin_: the comparisons happen at an sstate task level May 13 11:45:32 kanavin_: so this would only apply if B hadn't changed and A hadn't changed in output May 13 11:45:46 assuming C uses A in its sysroot May 13 11:46:29 yes it was ubuntu 19 . May 13 12:00:11 Hi, I need to setup an init.d service but started manually only. I tried `INITSCRIPT_PARAMS = "start 99 . stop 99 0 6 ."` but I'm getting an error `expr: non-numeric argument`. Is there a better way or did I make a mistake ? May 13 12:01:26 RP: right, I guess I'm just unsure of specific example scenarios where this would happen? One kind of shortcut is if a recipe build output is the same, we can skip the packaging steps. Or if a native tool has changed, but its output has not, then only the direct dependencies of the tool need to be rebuilt to confirm that. Are there other obvious wins? May 13 12:04:21 kanavin_: a lot of the time there are syntax changes to classes/variables which doesn't change the output. This would allow large chunks of the build to be skipped in those cases May 13 12:04:46 kanavin_: or imagine a patch which say only changes the docs, the sysroot sstate would be unchanged and the rest of the build reused May 13 12:09:35 RP: I see - this would require a major change to the task queue though. Right now a tree of setscene tasks is made and processed, then a tree of real tasks is made and processed. May 13 12:10:01 with the new approach, the tree would be 'adjusted' after every task has completed. May 13 12:10:28 or a graph rather May 13 12:14:26 kanavin_: it needs a total rewrite of the way runqueue works May 13 12:15:57 never said it was easy :) May 13 12:26:22 RP: I was poking around and decided to give https://bugzilla.yoctoproject.org/show_bug.cgi?id=13323 a go on Friday... I don't think it will be too bad to get *something* working May 13 12:26:23 Bug 13323: enhancement, High, 2.8 M1, JPEWhacker, NEW , Add reproducibility automated testing May 13 12:32:42 RP: kanavin_: IIRC one of our developers did work on a similar approach a while ago. Randomly highlighting mario-goulart with keyword "CDEPENDS", maybe he can offer some insights from read-world use. May 13 12:32:59 s/read-world/real-world/ May 13 12:35:40 https://lists.yoctoproject.org/pipermail/yocto/2017-June/036919.html May 13 12:36:24 Yeah, the message pointed out by mcfrisk has more info on CDEPENDS. May 13 12:45:42 bitbake page is broken May 13 12:45:56 http://developer.berlios.de/projects/bitbake/ May 13 12:47:38 how do I update bitbake in poky (thud) ? May 13 12:52:40 " Bitbake version 1.39.1 is required and version 1.38.0 was found May 13 12:52:41 " May 13 12:52:50 Hmm, CDEPENDS is interesting.... May 13 12:56:32 neverpanic: I remember it, it just always seemed a bit too special case and fragile :/ May 13 12:57:39 CDEPENDS is basically: "this dependency is need to build, but doesn't cause a rebuild if it changes"? May 13 13:05:15 IIRC it never worked very well for us either, and we've since stopped using it, but maybe there are some lessons learned we could share. May 13 13:06:08 New news from stackoverflow: How to set path in bitbake append file? May 13 13:11:48 how do I fix my bitbake version problem? seems weird May 13 13:12:59 Ad0: if you're using poky the version of bitbake and oe-core shipped together should be compatible May 13 13:13:06 Ad0: are you mixing up several branches, or how did you end up in this situation? i'm relatively sure that latest thud builds without problems. May 13 13:13:19 yeah, I think I'll start over May 13 13:14:00 neverpanic: I'd be interested to understand the lessons. I remember talking about those patches with their author at the time.. May 13 13:14:41 Yeah, unfortunately the author doesn't seem to be here atm, but maybe you can drop him an email and ask whether he has a couple of minutes to chat with you about that. May 13 13:50:30 RP, LetoThe2nd - https://pastebin.com/tH2V1V3Z May 13 13:50:55 same happened heh May 13 13:53:10 Ad0: and without meta-raspberrypi? May 13 13:53:28 suggestion for recipe to build? :) May 13 13:53:39 core-image-minimal May 13 13:54:06 i don't use toaster, so can only guess. May 13 13:54:10 same error May 13 13:54:33 core-image-minimal qemux86 May 13 13:55:16 how do I check the actual bitbake version - when I do bitbake --version it says BitBake Build Tool Core version 1.40.0 May 13 13:55:36 so why the heck does it find 1.38 ? May 13 13:57:08 Ad0: I'd suggest you send a mail to the toaster mailing list, they can probably help better than we can May 13 13:57:20 thanks :) May 13 14:36:27 New news from stackoverflow: Add same files to multiple PACKAGES May 13 14:47:42 seems like toaster think it's on sumo, but says it's on thud May 13 14:50:46 a problem in lib/toaster/orm/fixtures/oe-core.xml May 13 14:59:05 http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/bitbake/lib/toaster/orm/fixtures/oe-core.xml?h=thud#n56 - says bitbake version 1 ... ? May 13 14:59:10 should be 4 I assume May 13 15:00:35 hm, no it should be 4 or 1. May 13 15:02:05 Ad0: I think you're right and it should be 4 May 13 15:02:12 yes May 13 15:03:12 Ad0: the toaster list would know for sure May 13 15:03:29 I can't find the toaster list, I only see the bitbake-devel May 13 15:03:50 Ad0: lists.yoctoproject.org May 13 15:03:53 is it that one or is there really an own toaster? :) May 13 15:04:01 thanks May 13 15:04:35 LetoThe2nd: is there a twitch stream today? or is my calendar confused ? May 13 15:05:29 RP: tomorrow, it should be. second tuesday. May 13 15:05:42 RP: already got stuff prepared :) May 13 15:05:54 right.. mine as well.. May 13 15:06:18 ndec: ah, andreea sent out 13th... strange May 13 15:06:21 "When: May 13, 2019, 5:00 PM CET" as announced by Andreea.. May 13 15:07:19 indeed. i mistyped may 13th, and in the text upon writing down the concept i wrote "every second tuesday". May 13 15:07:25 so shes right, my apologies May 13 15:07:26 RP: webkit upstream has reworked the SSE2 pieces again, so let's defer the webkit update until they make another release May 13 15:07:55 I will however make a patch to modernize qemux86 May 13 15:08:04 kanavin_: ok, that does sound helpful to us May 13 15:08:06 LetoThe2nd: looking at my inbox you are the one who proposed May 13 th ;) May 13 15:08:08 ndec: but i can't do a run right now, sorry :( May 13 15:08:10 kanavin_: agreed, we should May 13 15:09:43 ndec: you're absolutely right, i mistyped in the initial date list and didn't double-check the announcement May 13 15:10:21 but it looks like nobody is there.. not even andreea.. so maybe she also thinks it's tomorrow.. idk.. May 13 15:11:05 ndec: can you maybe trigger a short notice on the announcement that i messed up and will be streaming tomorrow? May 13 15:11:19 ok! May 13 15:12:07 ndec: sorry... i always knew there will be a date glitch one day, but didn't expect it so early :( May 13 15:13:13 LetoThe2nd: do we agree that 5PM CET is now, at least? May 13 15:14:32 ndec: 17:13 it is now. and one of my very next notices towards andreea woud've been that we change the announcement timezone to Europe/Berlin. it correlates better with the ics, and is less confusing, methinks May 13 15:14:45 so i will be live in 23:45hrs :) May 13 15:14:50 right! May 13 15:16:32 * LetoThe2nd is currently beating his head against a layer that shall change its kernel from non-xenomai to xenomai :/ May 13 15:18:43 RP, I'll mail them May 13 15:18:56 I changed the xml but it had no effect, I had to edit the sqlite file May 13 15:30:38 how do i stop yocto form building a package? May 13 15:30:38 hah it worked May 13 15:30:43 actually building not just installing May 13 15:32:00 milloni: Don't build anything that depends on the recipe that builds the package. May 13 18:09:58 not a yocto question, but i'm thinking someone here has run into this.. May 13 18:10:31 i create an /etc/X11/xorg.conf file. on reboot the file contents have been removed but the file still exists May 13 18:10:40 who's clearing the file?!? May 13 18:11:44 yates: if you're stuck for an explanation you could try enabling the audit subsystem and get it to log writes to the file May 13 18:12:12 ok, sure. how do i "enable the audit subsystem"? May 13 18:12:15 I did once use that to find out what was wiping out files in ~/tmp May 13 18:12:44 yates: https://wiki.archlinux.org/index.php/Audit_framework May 13 18:12:56 (I was using it on Fedora) May 13 18:13:36 shoot. i'm 4.1.15 May 13 18:13:40 not available there. May 13 18:14:13 ok, I'd missed it was that new May 13 18:14:51 you could take away write access to the file and see what breaks :) May 13 18:15:08 true. May 13 20:28:35 RP, there is not meta-yocto contrib so I am using poky-contrib May 13 20:42:45 armpit: that is fine May 13 23:38:29 the manual says the recommendation for settting FILESPATH in a bbappend is: May 13 23:38:47 FILESEXTRAPATHS := "${THISDIR}/${PN}:" May 13 23:38:51 as opposed to May 13 23:39:05 FILESEXTRAPATHS := "${THISDIR}/${BPN}:" May 13 23:39:31 even though base.bbclass only sticks BP/BPN into FILESPATH May 13 23:39:59 is that just an artifact of the fact that PN used to be included in FILESPATH by default? May 13 23:40:33 radsquirrel: no, it's not related to that May 13 23:41:04 bluelightning: ok - so it really is recommended to use PN there instead of BPN? May 13 23:41:24 radsquirrel: ${BPN} is probably better as it will work if the recipe is also BBCLASSEXTENDed to native May 13 23:41:55 the critical thing is that the expanded value must match up with whatever subdirectory you put your extra files in May 13 23:42:09 so you could in fact use "files" if you want May 13 23:43:05 bluelightning: would it make sense to update the docs to recommend BPN? May 13 23:43:17 I am in-fact dealing with native recipes May 13 23:43:43 probably yes May 13 23:43:48 and because in the base on BPN is used, but in the appended layers we used PN (per the docs) May 13 23:44:00 we wound up with an unfortunate directory hierarchy May 13 23:44:16 cool, I'll submit a patch May 13 23:44:18 thanks! May 13 23:45:01 well, so there is a caveat May 13 23:45:31 if your recipe is named "something-native" - ${BPN} will evaluate to "something" and not "something-native" May 13 23:45:36 isn't there always? :-) May 13 23:45:42 just something to be aware of May 13 23:45:46 right. May 13 23:46:19 I mean logically that's what you'd expect, but it's possible you could have the directory name not match up and then wonder why things aren't working May 13 23:47:42 https://pastebin.com/0N5JjCML May 13 23:47:59 that probably illustrates the "problem" I'm poking at. May 13 23:48:53 if I use FILESEXTRAPATHS := "${THISDIR}/${BPN}:" instead, I can get a more sensible directory hierarchy. May 13 23:49:25 right, fair enough May 13 23:50:14 I just wanted to make sure that $PN in FILESEXTRAPATH wasn't explicitly recommended over BPN for some reason I didn't know about. May 13 23:50:30 thx again. **** ENDING LOGGING AT Tue May 14 02:59:57 2019