**** BEGIN LOGGING AT Mon Jul 22 03:02:02 2019 Jul 22 07:19:41 Hi Yocto experts, I'm getting this error "Postinstall scriptlets of ['systemd'] have failed. If the intention is to defer them to first boot, then please place them into pkg_postinst_ontarget_${PN} (). Deferring to first boot via 'exit 1' is no longer supported" on do_rootfs() stage of Thud build. Any pointers? Jul 22 10:40:13 Hi I would like to activate kernel tracing with lttng, but for some reason the .cfg file in my additional layer is not applied to the kernel config Jul 22 10:41:11 bitbake-layers show-appends lists my recipe (which includes the SRC_URI modification) as .bbappend Jul 22 12:36:34 dear all, i new to yocto and i try to "generate" a wifi kernel module via bitbake for a yocto system here. the wifi chip is from laird connect and i found some how tos, but they were not working. now i started to investigate it deeper with devshell etc and i found out for example there is one of the first calls which will fail for example, the call Jul 22 12:36:34 of a makefile to request the kernel version in /lib/modules/4.XX.X-XX-generic/build, on an ubuntu there is really a make file placed there, but in my equivalent path in yocto for an arm, i hope i am right, is .../tmp/work-shared/XXX/kernel-build-artefacts there was a makefile but without target for a kernel version. and after a bitbake virtual/ker Jul 22 12:36:35 nel even this disappear. now my question: Could somebody explain me what is wrong with my yocto environment? best regards, Jul 22 12:36:39 Michael Jul 22 13:02:02 hello is this working? Jul 22 13:06:58 hi! very stupid question: i want to build a single recipe that i had to split over two layers (into a .bb and a .bbappend file) . know i can use bitbake -b to build a bb file, but how can i add the bbappend? Jul 22 13:08:02 ah i can at least read other questions :-) Jul 22 13:09:09 timblechmann: the bbappend is taken into account automatically, but you have to name like the recipe bb Jul 22 13:09:41 timblechmann: bitbake -b can't evaluate a bbappend AFAIK Jul 22 13:10:04 Michael78: your issue is related to kernel integration not Yocto Jul 22 13:10:29 mckoan: oh, i see ... thanks for the quick reply Jul 22 13:10:35 ok, i thought this is done by yocto Jul 22 13:10:58 is this than the poky channel? Jul 22 13:12:00 Michael78: probably I don't understand your question, sorry Jul 22 13:12:28 it was really a little bit to much Jul 22 13:12:52 basically i want to know why the yocto kernel build folder doesnt look like the "normal" one Jul 22 13:13:18 content and structure Jul 22 13:15:17 in /lib/modules/4.XX.X-XX-generic/build on an ubunut i have for example an makefile which gives me the kernelversion (as a target) Jul 22 13:19:41 Hi, Jul 22 13:19:42 is not knowing how to automatically build the encrypted rootfs and package it into the .wic file during the Yocto build process. Today our Yocto build produces a .wic with a boot partition and a plain rootfs which is then flashed to eMMC. We would like to replace the plain rootfs in the .wic with an encrypted one. Jul 22 13:19:42 taion or pointers on how to change my recipe so that it will build with an encrypted rootfs. Jul 22 13:19:54 timblechmann: if you want to build a single recipe just do bitbake recipename. -b doesn't include the dependencies, which means you don't get a compiler... Jul 22 13:22:42 rburton: i see ... i guess i have a toolchain in place from trying to build an image before ... i'm mainly trying to use -b to reduce the turnaround times (but mayber there are better ways?) Jul 22 13:23:03 timblechmann: it won't rebuild anything that doesn't need rebuilding Jul 22 13:23:40 if youre doing stuff like changing the core classes there are tricks to avoid rebuilding everything, but they're not needed for most 'i changed my recipe' builds Jul 22 13:24:39 rburton: yes, i was more trying to avoid the 'reparsing all recipes step' ... that's half a minute every time Jul 22 13:24:55 your machine sounds very slow, or you're deleting the cache every time Jul 22 13:25:17 persistant bitbake will help there Jul 22 13:25:50 RP: is memory resident bitbake the default in master? Jul 22 13:28:25 rburton: the machine is a 2-core skylake i7 (sub)notebook ... not extremely fast, but also not super slow Jul 22 13:34:05 RAM is normally the issue with notebooks, you just can't keep enough in the cache for speed Jul 22 13:50:13 timblechmann: for yocto builds, a 32 core system is a good start (of course such systems are not cheap but neither is time). Jul 22 13:57:06 rburton: no Jul 22 14:04:28 RP: are there known issues still? Jul 22 14:04:55 rburton: yes, I think there are open bugs Jul 22 14:05:07 rburton: I don't think there were serious issues but some Jul 22 16:58:06 rburton:gparted-0.33.0: gparted rdepends on hicolor-icon-theme, but it isn't a build dependency, missing hicolor-icon-theme in DEPENDS or PACKAGECONFIG? [build-deps] Jul 22 16:58:48 rburton: http://jenkins.nas-admin.org/job/oe_world_qemuarm64/370/console Jul 22 16:59:19 do we need to add hicolor-icon-theme explicitly to deps/rdeps Jul 22 16:59:46 I thought gtk-icon-cache.bbclass should take care of it Jul 22 17:00:44 https://errors.yoctoproject.org/Errors/Details/253713/ is the error log Jul 22 17:36:03 hey all, I'm trying to change SRCREV in an anon python function and it's generating a ton of basehash check errors, I've a bunch of lines to try and ignore the check, but it never works Jul 22 17:36:04 do_clean_remove[vardeps] = "PV SRCREV" Jul 22 17:36:50 any ideas? Jul 22 17:43:48 ebolton: changing state like this is going to bypass data integrity checks and will land you into more troubles I am guessing Jul 22 17:44:30 instead think of external way to induce right SRCREV to metadata before invoking bitbake, it will simplify it and make the system work well Jul 22 17:49:45 @khem I want to write a manifest, a single file that collects BRANCH and SRCREV info for several recipes in one place, I've tried it with a single include file containing a lot of override lines "BRANCH_pn-", it's problematic, so I'm trying to this Jul 22 17:50:48 @khem the only other external option I can think of is have some script manually scrape and replace variables in the recipes before bitbake is invoked Jul 22 17:54:15 yes you need a branch tool Jul 22 17:54:36 dont beat bitbake into doing it Jul 22 17:58:12 thanks Jul 22 18:06:35 Hi, all. My colleagues gave me a prebuilt DEB file they want included in our BSP image, like, *yesterday*. I'm not at all sure what the best approach for this is. Jul 22 18:06:44 I tried something along these lines: https://gist.github.com/evadeflow/fda5e7afe87f4098c457b38ea911c828 Jul 22 18:07:28 But `THISDIR` has the 'wrong' value in my `ROOTFS_POSTPROCESS_COMMAND`, so it's not finding the archive. Jul 22 18:08:20 How can I make the line `local archive=${THISDIR}/xyzinstaller_0.1-1.deb` actually find this DEB file (which is stored right next to my `fsl-image-qt5.bbappend` file)? Jul 22 18:09:58 set FILESEXTRAPATHS properly and add file://xyzinstaller_0.1-1.deb;unpack=no to SRC_URI Jul 22 18:10:03 then reference it in ${WORKDIR} Jul 22 18:10:24 i.e. handle it like any other source file/patch/etc Jul 22 18:11:05 Oh. Nice! Thank you, I'll try that right now. Jul 22 18:27:03 @kergoth, I modified my patch based on your suggestion: https://gist.github.com/evadeflow/fda5e7afe87f4098c457b38ea911c828 Jul 22 18:28:04 It's not finding the archive: `ar: /jenkins/workspace/ivs_bsp/build/tmp/work/imx8m_var_dart-poky-linux/fsl-image-qt5/1.0-r0/ivsinstaller_0.1-1.deb: No such file or directory` Jul 22 18:29:03 And it really seems not to be there, since `find tmp/work/imx8m_var_dart-poky-linux/fsl-image-qt5/ -name "*.deb"` turns up no hits. Jul 22 18:31:02 khem: is that not the gnome.bbclass changes? Jul 22 18:33:18 Is it possible to determine when the transmit buffer of a serial uart is empty so as to add a delay between transmission of serial messages? Jul 22 18:34:54 not really a yocto question Jul 22 18:40:26 I suppose its also a general purpose Linux question, but we are using a yocto distribution and I don't see anything supporting this in the posix termios documentaion so I was hoping there might be a solution aimed more at embedded linux users. Jul 22 18:46:57 RP: Reminder. Respond to the reproducible build thread on the ML Jul 22 18:47:17 rburton:yeh I think Adrian has been trying to use gnomebase directly and I wonder if thats causing this issue too ? Jul 22 18:53:59 khem: right, gnome inherits a load of stuff, which probably pulled in the depedns Jul 22 20:16:16 JPEW: thanks, I need to do that! :) Jul 22 20:17:41 JPEW: bit distracted/excited with another patch ;-) Jul 22 20:19:17 RP: lol, fair enough. That one is probably better to focus on; I'm not sure I'll have enough bandwidth to get both done for zeus Jul 22 20:20:34 Also, I've come to terms with reproduciblity being an "ongoing process", more than a single "task"... but good QA checks will make that easier to distribute that work Jul 22 20:32:52 JPEW: The QA checks are absolutely key Jul 22 20:33:38 JPEW: I also think the other work will raise the priority on reprodcibility as people join the dots Jul 22 20:33:51 RP: Ya Jul 22 20:41:24 JPEW: oddly in bigger builds I am seeing connection timed out to the hashserver :/ Jul 22 20:46:53 Hmm. Do you know if its on a GET or a POST? Jul 22 20:47:15 JPEW: I'd say POST Jul 22 20:49:18 JPEW: I've realised this is using HTTPServer and I think I remember that is shockingly bad at concurrency Jul 22 20:49:36 Ya, that an FWIW, I'm not sure if my sqlite querys are quite optimal in that case. It does end up taking two SELECTs and an INSERT for a new equivalent hash Jul 22 20:49:39 JPEW: it can't cope with server sstate Jul 22 20:49:59 JPEW: My bet is on HTTPServer rather than that Jul 22 20:54:17 JPEW: I'll throw a ThreadingMixIn in there Jul 22 20:56:35 Ok. I don't remember if sqlite cursors are atomic; it might need a semaphore around the SELECT/INSERT/SELECT in POST Jul 22 20:58:15 JPEW: hmm, this could be interesting as I didn't do that :) Jul 22 20:58:34 JPEW: Its now showing pages of warnings now though Jul 22 20:59:36 Ya, my sqlite expert just told me to use a semaphore :) Jul 22 21:00:38 JPEW: does it actually matter since only one task would be updating any given entry? Jul 22 21:01:44 Hmm.... maybe not? I guess I was thinking of the generic server case where you might have mutiple simultaneous builds.... Jul 22 21:03:00 JPEW: right, the potential problem is two builds running simultaneously if both made the first select and reach the insert, does it break anything even then? Jul 22 21:04:42 I think the INSERT would fail with duplicate primary key (in which case you could probably just ignore it) Jul 22 21:05:12 Hmm, maybe not.... it's been a while :) Jul 22 21:05:56 JPEW: right, just ignoring a duplicate insert would be much nicer than the locking. Also wondering why you need the third selet Jul 22 21:07:01 * JPEW shakes fist at past self for lack of comments Jul 22 21:07:29 JPEW: I think you're being clever with the data structures but at the expense of performance Jul 22 21:09:02 JPEW: Just tried this in a big build. Its basically locked the machine up at 100% cpu so definitely a performance issue somewhere Jul 22 21:09:19 (big == minimal image) Jul 22 21:10:12 I think I used to have index on the columns in the table to speed up the lookup (since they are so common). Jul 22 21:10:38 I original wrote the whole server in flask :) Jul 22 21:11:20 JPEW: I remember. Just a bit of a dependency minefield for us :/ Jul 22 21:11:35 JPEW: sqlite definitely needs indexes to work with any kind of performance Jul 22 21:12:00 Ya, I'm pretty sure I had the indexes in the flask version, let me dig back to see if I can find them Jul 22 21:12:15 The cooker sat for 2 minutes pinning a CPU, then crashed with a bug in my code Jul 22 21:12:29 * RP tried again with the typo fixed Jul 22 21:13:53 Ok, I had a uniqueness constraint on the ('taskhash', 'method', 'outhash') triplet, and an index on ('taskhash', 'method') Jul 22 21:19:49 did the long standing glibc-locale host-user-contaminated issue ever get resolved? Jul 22 21:20:18 ah, [OE-core] [PATCH] pseudo: Update to gain key bugfixes looks promising Jul 22 21:22:03 kergoth: sadly, no :/ Jul 22 21:22:22 JPEW: not sure that is the right index but I'll play Jul 22 21:25:51 ah, that's unfortunate Jul 22 21:28:04 I'm getting the following error with a recipe where I'm simply trying to install items from a tarball: Jul 22 21:28:33 `error: No file attributes configured` Jul 22 21:28:52 The recipe is here: https://gist.github.com/evadeflow/13a35f9fd52003515c85032267baf091 Jul 22 21:28:53 JPEW: good news is that it did figure out how to recover from "bitbake m4-native -c install -f" and not rebuild much. Bad news is it takes 3 minutes of 100% cpu to figure out what its doing :) Jul 22 21:29:23 The error comes from running `rpmdeps`. Any idea how to fix this? Jul 22 21:30:18 (And... should I be concerned that it's running `rpmdeps` when I'm only attempting to build DEB packages?) Jul 22 21:30:30 evadeflow: I've never seen that error before. and no, don't be concerned Jul 22 21:30:43 evadeflow: we use rpm tools to figure out dependencies Jul 22 21:30:45 RP: Ya, that's a bit unfortunate. It that with a clean tmpdir? It does take a while the first time to populate the local (persistDB) cache, but once thats populated it only talks to the server when something changes Jul 22 21:30:55 JPEW: no, existing dir Jul 22 21:31:17 Hmm, Ya I was pretty sure I had it to an unnoticable level. Jul 22 21:31:21 JPEW: I'll figure out where its spending the time, hopefully something silly Jul 22 21:31:45 JPEW: I'm happy it appears to do the right thing :) Jul 22 21:31:57 Ya, that is good Jul 22 21:59:32 JPEW: part of the problem is the code is passing around an sqlite cursor connection instead of a string for the unihash Jul 22 22:07:18 JPEW: The threadingmixin really did't get on well with sqlite! Jul 22 22:11:40 JPEW: 1.7 million sql execute calls are the bottleneck :) **** ENDING LOGGING AT Tue Jul 23 02:59:57 2019