**** BEGIN LOGGING AT Wed Jan 29 02:59:57 2020 Jan 29 03:29:11 New news from stackoverflow: How to store pyc files with setuptools3 in bitbake recipe in the same way setuptools does? Jan 29 06:17:59 Which meta-java branch is supposed to compile openjdk-8 with zeus and which host gcc version should I use? gcc9 seems to be too restrictive to compile without hacks. Currently I try to build icedtea7-native-2.1.3 with host gcc version 7.4.0. meta-java on master Jan 29 06:29:42 New news from stackoverflow: Kernel hang during boot-up Jan 29 07:26:50 toaster + bitbake gives me the "timeout waiting for the bitbake server". where to find the timeout option? Jan 29 08:41:08 Is Toaster ready for production? Jan 29 08:41:33 dv|2: depends on what you expect it to do Jan 29 08:42:52 dv|2: in a real life production environment i actually see little use for it besides build inspection. Jan 29 08:45:58 LetoThe2nd, I mean I constantly get different errors, timeouts and exceptions trying to run toaster in Build Appliance image. Jan 29 08:47:07 dv|2: i guess thats more related to the build appliance environment. sure, file a bug for it Jan 29 08:47:25 but i usually just ignore toaster as i have no use case for it. Jan 29 08:49:58 LetoThe2nd did you ever see any good manual/instruction how to setup Toaster? I already read Yocto manual for Toaster and simple (but outdated) instructions in yocto Wiki Jan 29 08:51:35 dv|2: https://www.yoctoproject.org/docs/latest/toaster-manual/toaster-manual.html Jan 29 08:51:53 this is the current state of affairs. if it is wrong, please file bugs :) Jan 29 08:53:25 but actually, i'd advise to check it works *outside* of the generated build appliance first. Jan 29 09:01:09 LetoThe2nd, thank you Jan 29 09:02:15 dv|2: but seriously. to avoid wasting time, i would take a minute to think about what you actually expect it to do, and how it will be helpful in your work. because it usually is not what people think it is. Jan 29 09:02:45 dv|2: first and foremost, it is *NOT* a GUI to magically point-click-create-build on yocto technology. Jan 29 09:04:45 LetoThe2nd, I'm building the development image for my customers. We need to let them buil the image for our device even if they just have VirtualBox. It looks very convinient if customer company may just download the image and start building the image for our device with or without some recipes included Jan 29 09:05:40 dv|2: the key point to get your head around is: what will they do once they are satisfied with the build? Jan 29 09:06:06 shall they keep that appliance image with the magic toaster settings around forever? Jan 29 09:06:39 or, doesn't it make more sense to pour everything into a image reipce then, which can be put into an SCM, CI workflow and everything? Jan 29 09:07:24 because in all cases i've seen so far, this is what people actually want. and then toaster is a bad fit. rather go for just devtool. Jan 29 09:10:32 LetoThe2nd, some of them are not experienced Linux users so it looks more convinient to give the image that has web interface to build the image for target device Jan 29 09:10:53 just sayin Jan 29 09:12:24 web interface all things? Jan 29 09:14:08 yes, everything is configured and they just need to run a couple of commands once then go to web interface and add some software to final image, click build and get the result Jan 29 09:14:41 dv|2: and then they put the VM into their git repo. :) Jan 29 09:15:17 Hi all. I found a strange behaviour using PREFERRED_VERSION using Zeus (but perhaps could be also in other versions) Jan 29 09:15:35 I have two recipes vesions of a custom kernel, my-kernel_4.9.11.bb and my-kernel_4.9.67.bb Jan 29 09:15:50 Into my machine .conf I set PREFERRED_PROVIDER_virtual/kernel = "my-kernel" Jan 29 09:16:23 LetoThe2nd, they are windows users. everything is possible :) Jan 29 09:16:34 dv|2: go ahead then. Jan 29 09:16:51 If I set PREFERRED_VERSION_my-kernel = "4.9%", the selected version during build is 4.9.11 Jan 29 09:17:21 If I don't set PREFERRED_VERSION the selected version is 4.9.67, which is what I expect also using PREFERRED_VERSION Jan 29 09:18:42 I know that in this situation the PREFERRED_VERSION is useless, but...it's strange Jan 29 09:22:33 ms_k: only thing that comes to my mind, have you compared bitbake -e virtual/kernel in both cases? Jan 29 09:25:04 exit Jan 29 09:52:21 Hello all, stupid question, I have probably miss something but in all devshell I got this followinf issue : gcc-version.sh: line 26: arm-poky-linux-gnueabi-gcc: command not found... Someone knows how to fix that ? Thx. Jan 29 09:55:22 LetoThe2nd: thank you. What I should compare in two environments? Jan 29 09:56:18 LetoThe2nd: I tried command bitbake -s | grep my-kernel Jan 29 09:57:52 LetoThe2nd: output with PREFERRED_VERSION is Latest Version:4.9.67-r0, Preferred Version 4.9.11-r0 Jan 29 09:58:26 LetoThe2nd: output without PREFERRED_VERSION is Latest Version:4.9.67-r0, Preferred Version Jan 29 09:58:53 PinkSnake: have you fixed your dts issue from yesterday (you wanted to cp it to the sources) Jan 29 10:01:50 qschulz Yes all right :-) Jan 29 10:03:46 qschulz FYI we can use something like that (file://zynq-bd51x2-2.0.dts;subdir=git/arch/${ARCH}/boot/dts) in order to add cp command :) Jan 29 10:05:33 hey there, can anyone tell me how licenses in staticly built/linked linraries are handeled in bitbake/yocto? Jan 29 10:06:21 say i write an application (MIT) which uses LibA (GPLv2) that is staically linked into the application. will both be listed in the license.manifest? Jan 29 10:09:15 PinkSnake: i'm wondering if you can't replace git by ${S} Jan 29 10:11:48 qschulz you're right with s = "${WORKDIR}/git". will do that :) Jan 29 10:12:33 PinkSnake: I'm not so sure, maybe be subdir is relative and not absolute, in which case ${S} wouldn't work Jan 29 10:13:29 PinkSnake: worth trying (and telling me :D) Jan 29 10:13:37 qschulz let's make a try to be sure :) Jan 29 10:15:02 qschulz Answer is no : Failure expanding variable WORKDIR, expression was ${BASE_WORKDIR}/${MULTIMACH_TARGET_SYS}/${PN}/${EXTENDPE}${PV}-${PR} which triggered exception RecursionError: maximum recursion depth exceeded while calling a Python object, now we are sure ^^ Jan 29 11:17:25 hi, i have submitted a patch, but i think it might have gotten under the radar with other ones i submitted, i was wondering if someone could let me know if its something that is wanted or not so i can forget about it? thanks http://lists.openembedded.org/pipermail/bitbake-devel/2020-January/020759.html Jan 29 11:37:31 FrazerClews, maybe reply with a "ping" Jan 29 12:15:11 I have exceptions running bitbake in the Build Appliance image: https://ibb.co/1Z3FZpg Jan 29 12:15:11 https://ibb.co/TY36mp6 Jan 29 12:22:34 I stuck at the first step of Toaster setup : [Errno 2] No such file or directory: ${BSPDIR}/externals/core/meta, it's not possible to have variable inside bblayer path ? o.0 Jan 29 12:28:11 PinkSnake: It is possible but not very easy. Where is BSPDIR defined? Jan 29 12:29:13 paulbarker Thx for repllying, BSPDIR is difine inside bblayers.conf (BSPDIR := "${@os.path.abspath(os.path.dirname(d.getVar('FILE', True)) + '/../..')}" Jan 29 12:29:58 And what's FILE? Jan 29 12:30:00 It's just to let developers use our Yocto conf where they want Jan 29 12:30:48 paulbarker it's a dummy var set by a script before call setup-environment Jan 29 12:31:33 I have make a try with full path, no issue all seems work, do you think relative path is enought ? Jan 29 12:31:54 PinkSnake: bitbake-layers add-layer? Jan 29 12:31:55 Have you added that variable to BB_ENV_EXTRAWHITE? Jan 29 12:33:02 paulbarker Pretty sure not :S Jan 29 12:33:08 i will double check the doc Jan 29 12:33:14 Ok, let me send you some links Jan 29 12:33:19 https://www.yoctoproject.org/docs/3.0/bitbake-user-manual/bitbake-user-manual.html#var-bb-BB_ENV_EXTRAWHITE Jan 29 12:33:49 Here's an example: https://github.com/SanCloudLtd/sancloud-arago/blob/thud/build/conf/setenv defines YOCTO_SOURCES and adds it to BB_ENV_EXTRAWHITE before bitbake is called Jan 29 12:34:01 https://github.com/SanCloudLtd/sancloud-arago/blob/thud/build/conf/bblayers.conf uses YOCTO_SOURCES Jan 29 12:34:13 paulbarker Thank you for the support :) Jan 29 12:34:40 You'll probably have to customise the setup-environment script you mentioned Jan 29 12:34:46 you save my day Jan 29 12:56:11 paulbarker Fixed! Thx you also for your apply.py script ;) Jan 29 12:57:30 what is the easiest way of getting a list of available packageversions within my current meta-layers? Jan 29 12:58:02 creich: maybe oe-pkgdata-util has something for that. Jan 29 12:58:19 thx, i'll check Jan 29 12:59:00 LetoThe2nd: btw, i have watched most of your videos. latest i've seen was the one about distro/machine/image configs. Jan 29 12:59:17 i'd say i'd liked that one most :) Jan 29 12:59:43 creich: glad to hear. yes, thats really one of the most important ones. Jan 29 12:59:55 not sure why, just had the feeling that it explained things very good and i found myself beeing more enlightened than before compared to the other videos ^^ Jan 29 13:00:18 bitbake-layers show-recipes? Jan 29 13:00:25 qschulz: ah! Jan 29 13:01:26 well it's all layers but it says which version of which recipe is where Jan 29 13:01:44 creich: well some topics are tough, others are not. the key about the live coding sessions is that they're *NOT* meant as fully prepared webinars, but as live tinkering. so sometimes things work out and play nicely, some other time they don't. Jan 29 13:03:21 LetoThe2nd: yeah i got that. Jan 29 13:03:24 creich: bitbake-layers will dump what it found Jan 29 13:03:26 didn't want to bother Jan 29 13:03:29 ah, LetoThe2nd beat me to it Jan 29 13:03:45 qschulz: rburton thx guys :) Jan 29 13:04:55 looks like usually there is only one version available. i guess that's supposed to be that way?! Jan 29 13:05:10 rburton: i beat you= Jan 29 13:05:11 ? Jan 29 13:05:29 i am wondering if there is something compared to the gentoo style, where i can (more easy?) choose between versions Jan 29 13:05:51 as far as i understood that should easily be possible with the preferre_version settings Jan 29 13:09:00 LetoThe2nd: I think he swapped your nick and mine ;) Jan 29 13:09:29 creich: indeed. Don't forget that two recipes can provide the same functionality as well. So you shouldn't look for one recipe only :) Jan 29 13:09:37 creich: c.f. PROVIDES in recipes Jan 29 13:09:59 creich: otherwise, PREFERRED_VERSION is the way to go (and if a different recipe name, PREFERRED_PROVIDER as well) Jan 29 13:10:54 thx! Jan 29 13:13:23 halstead: Ok, I will only need to grab one file from it and then it should be good for reuse Jan 29 13:15:16 tgamblin: there was another log failure in that build btw :/ Jan 29 13:17:27 RP: I see three instances of the parselogs test failing? What's the other? Jan 29 13:19:54 RP: You mean this one, right? https://autobuilder.yoctoproject.org/typhoon/#/builders/109/builds/402 Jan 29 13:23:39 On a related note, it looks like I am in fact able to recreate it now, although still in a semi-inconsistent manner Jan 29 13:24:48 tgamblin: https://autobuilder.yoctoproject.org/typhoon/#/builders/72/builds/1535 was the one I didn't share yesterday Jan 29 13:25:00 tgamblin, I'll grab that file for you now. Which is it? Jan 29 13:25:00 tgamblin: reproducing at all is good Jan 29 13:25:12 halstead: I would like to have /var/log/boot Jan 29 13:31:50 Trying to correlate commits to oe-core with the first report of the issue Jan 29 13:34:09 I caught these errors in the boot log for a local QEMU instance last night, while manually building and cleaning core-image-sato-sdk: Jan 29 13:34:10 hwclock: settimeofday() failed: Invalid argument Jan 29 13:34:10 touch: cannot touch '/var/log/wtmp': No such file or directory Jan 29 13:34:10 chown: cannot access '/var/log/wtmp': No such file or directory Jan 29 13:34:10 Failed to set owner -root- for -/var/log/wtmp-. Jan 29 13:34:10 chmod: cannot access '/var/log/wtmp': No such file or directory Jan 29 13:34:13 Failed to set mode -0664- for -/var/log/wtmp-. Jan 29 13:36:49 Hi, Is there a way for using xserver-nodm-init with .Xauthority ? Jan 29 13:44:03 what "ERROR: Layer version information not found; Check if the bitbake server was configured to inherit toaster.bbclass" means? Jan 29 14:17:14 I'm seeing weird errors when using a binary tool chain. I'm seeing multiple things similar to: Jan 29 14:17:14 WARNING: m4-1.4.18-r0 do_package_qa: QA Issue: m4-ptest rdepends on locale-base-fr-fr.iso-8859-1, but it isn't a build dependency? [build-deps] Jan 29 14:17:14 Is this a legit problem or some weird side effect of binary toolchains? Jan 29 14:25:08 side affect of a binary toolchain Jan 29 14:25:39 (glibc builds those usually) Jan 29 14:30:55 thanks rburton Jan 29 14:31:12 so your toolchain should provide it, or something? Jan 29 14:53:14 tgamblin: those sound related Jan 29 14:56:34 tgamblin: in a new build it looks to be the same things failing on that test so it may be more deterministic... Jan 29 14:57:11 jonmason: it is a real issue, almost certainly from the toolchain Jan 29 14:57:54 RP: Yeah. I'm currently trying to confirm that it's repeatable if you do the following: 1. bitbake core-image-minimal; 2. bitbake core-image-sato-sdk; 3. runqemu kvm slirp nographic Jan 29 14:58:04 If that ends up true then I should be able to make the logrotate test fail Jan 29 14:58:28 ok, I'll open internal bugs to track this and hopefully someone can be tasked with working on it Jan 29 15:00:49 jonmason: i saw similar problems with the linaro toolchain before Jan 29 15:01:13 rburton: now the linaro toolchain is the arm toolchain and lives in meta-arm :) Jan 29 15:01:25 i've a solution Jan 29 15:01:29 burn it in fire Jan 29 15:01:44 and if it's in meta-arm, then I'm going to test it constantly! Jan 29 15:01:53 external toolchains are a world of pain and should be destroyed Jan 29 15:02:12 * tgamblin approves of this solution Jan 29 15:03:19 if arm makes it, then i need to support it Jan 29 15:03:43 even if that means just pointing to a bug and saying "we'll get around to it" Jan 29 15:05:02 can the bug be "remove external toolchain" Jan 29 15:05:46 * rburton steps off his high horse Jan 29 15:10:02 https://twitter.com/TheYoctoJester/status/1222537173494308864 Jan 29 15:13:15 LetoThe2nd: hahahaha Jan 29 15:13:24 zeddii: khem https://www.irccloud.com/pastebin/2sNM4rTx/ Jan 29 15:13:28 thoughts? Jan 29 15:13:37 I wonder if I like and retweet if that would offend my employer Jan 29 15:14:32 Crofton|road: is your asm symlink busted? Jan 29 15:14:37 jonmason: i took it out of context so it should be SFW: :) Jan 29 15:14:45 hah Jan 29 15:15:57 Crofton|road. hmm, that should be there after scripts prepare or modules_prepare. Jan 29 15:16:38 without it, very little builds. I can see if something is missing in devsrc, I also have a custom patch you can try if you want ALL the kernel source on the target. Jan 29 15:16:46 this is with sdk made with populate_sdk Jan 29 15:17:21 on target I havesame issue, are you suggesting I run a command before building a module? Jan 29 15:17:26 aha. so not on target ? you'd still need to recreate the symlinks through prepare. Jan 29 15:17:45 yes, the devsrc can't package those things Jan 29 15:17:58 you always need to do a 'make scripts prepare' or 'make modules_prepare' Jan 29 15:18:28 ahah. sweet. early checkin was just confirmed at my hotel. Jan 29 15:18:39 * zeddii will be in Brussels @ 10am tomorrow. Jan 29 15:21:04 ah Jan 29 15:21:16 * Crofton|road is in Leiden Jan 29 15:23:08 so modules_prepare shoul dbe in the makefile in the -C directory? Jan 29 15:24:05 yup. where the kernel source is. Jan 29 15:25:19 make: *** No rule to make target 'modules_prepare'. Stop. Jan 29 15:25:26 hmmm Jan 29 15:25:54 got it Jan 29 15:26:26 Crofton|road: make help Jan 29 15:27:27 lol Jan 29 15:27:28 yeah Jan 29 15:27:37 this deosn;t work lik eit does on Fedora! Jan 29 15:27:38 headers_check - Sanity check on exported headers Jan 29 15:28:07 probably what he meant to say Jan 29 15:28:17 as there is no modules_check Jan 29 15:28:28 typos Jan 29 15:29:26 victory thanks Jan 29 15:29:35 woot. Jan 29 15:30:08 even using Xilinx borad Jan 29 15:31:57 * zeddii shudders Jan 29 15:32:00 ;) Jan 29 15:33:36 zedboard Jan 29 15:33:50 *even* better Jan 29 15:34:02 Also, I need to introduce you to a friend who wants to know why his ultrascale is painfully slow Jan 29 15:36:22 LetoThe2nd: you've seen the bugzilla quips right Jan 29 15:37:15 I'll make sure to have a few beers first ! Jan 29 15:38:46 rburton: i ain't seen nothing. Jan 29 15:38:53 https://bugzilla.yoctoproject.org/quips.cgi?action=show Jan 29 15:39:02 (yet) Jan 29 15:39:32 my favourite still being 'ross: RP and myself were looking at that last night and started pulling on the little thread of that bug. currently unravelling into a rat infested spool of razorwire surrounded by bees' Jan 29 15:39:35 rburton: what is it supposed to tell me? Jan 29 15:39:46 LetoThe2nd: feel free to donate quotes Jan 29 15:40:07 you get a random quote at the top of every query Jan 29 15:40:16 at the moment it donates "The username or password you entered is not valid" Jan 29 15:40:33 apparently you need to be logged in Jan 29 15:41:33 "apperently" Jan 29 15:41:55 rburton: nice. Jan 29 15:42:00 * LetoThe2nd likes. Jan 29 15:42:20 the first batch are from the old openedhand bugzilla, so they're proper vintage Jan 29 15:42:35 I think rm_work.bbclass is removing some stamp that causes my openjre-8 recipe to rebuild all the time.. any tips for how to debug which stamp it accidentally being removed ? Jan 29 15:43:14 rburton: I like that one :) Jan 29 15:43:53 "Our next release should be called Unicorny, who's with me!!???" Jan 29 15:44:15 too late, naming scheme decided and has HUNDREDS of candidates Jan 29 15:44:54 hehe Jan 29 15:44:58 I think I need to brush up on my geography... Jan 29 15:45:10 i like rhinos. they're like unicorns, just fatter. Jan 29 15:45:48 Hello! I have multiple partial defconfig for my kernel recipe, but when I do `bitbake linux-sap -c menuconfig` it doesn't seems to take in consideration the partial configuration, to add to this, I have an externalsrc setup for my kernel. Is it not supported? I'm probably doing something wrong... Jan 29 15:48:20 I added one :) Jan 29 16:10:55 "Sstate summary: Wanted 39 Found 37 Missed 2 Current 611 (94% match, 99% complete)". Where do I find which 2 tasks were missing ? Jan 29 16:11:49 Hi. I previously had a kernel with some patches applied along with device tree changes, but all of the patches (besides the device tree) are now in mainline. Jan 29 16:12:59 Is there a mechanism to build an out-of-tree dt which also lets me use the pre-processor like device tree does in the in-tree build? Jan 29 16:29:53 fullstop: Have a look at devicetree.bbclass in oe-core Jan 29 16:30:17 thanks! Jan 29 16:37:42 jonmason: does arm care about builds using thumb Jan 29 16:37:54 fullstop: almost forgot, here is an example for how you can use it (https://github.com/nathanrossi/meta-random/blob/master/recipes-toolbox/raspberrypi/rpi-device-tree.bb) Jan 29 16:39:01 nrossi: I'll check it out. My DT stuff is based off of a file in the kernel and requires the pre-processor to be built. Hopefully this works like that. Jan 29 16:40:45 fullstop: handles preprocessor, overlays and enables including kernel sources for headers/etc. Jan 29 16:40:53 Great!! Jan 29 16:41:58 rburton: I believe so. Is there an issue? Jan 29 16:42:08 jonmason: just gave you one of my bugs :) Jan 29 16:42:11 JPEW: looking at the ca-cert failure I have a hypothesis Jan 29 16:42:33 JPEW: could hashequiv break reproducibility? as it doesn't account for timestamps? Jan 29 16:42:52 rburton: thanks a lot ;-) Jan 29 16:43:00 nrossi: that was astoundingly easy. thanks! Jan 29 16:43:29 i.e. if I build without repro and have random timestamps on some files could that then leak into a repro build via hashequiv Jan 29 16:44:05 kroon_: bitbake -DDv and save the logs to a file, then grep Jan 29 16:44:31 where does the kernel recipe install sanitized kernel headers for user space use? Jan 29 16:45:12 milloni: into the sysroot in usr/include ? Jan 29 16:45:29 that's the linux-libc-headers isn't it? Jan 29 16:45:45 milloni: yes, guess it depends which headers you mean Jan 29 16:46:06 i'm asking in the context of the warning in linux-libc-headers.inc Jan 29 16:46:58 https://paste.debian.net/hidden/784fbbdb/ Jan 29 16:47:24 milloni: there are a few ways to get the other headers but its likely a dependency on virtual/kernel and having it stage the headers you need Jan 29 16:48:18 i was trying to use STAGING_KERNEL_DIR/include/uapi as my include path Jan 29 16:48:32 but it complains that these aren't sanitized headers Jan 29 16:49:04 milloni: then you probably need to have that kernel put sanitized headers somewhere Jan 29 16:49:04 RP: "having it stage the headers you need" by it do you mean virtual/kernel? Jan 29 16:49:24 ack, thanks Jan 29 16:49:30 milloni: to be fair sanitzed headers were created after that warning was written :/ Jan 29 16:49:53 but its still fair, you need to depend on virtual/kernel to get them Jan 29 16:50:05 er, s/fair/right/ Jan 29 16:50:19 https://kernelnewbies.org/KernelHeaders suggests keeping a copy of the headers I need with the user-space program that needs them Jan 29 16:50:36 people shouldn't ever be hacking linux-libc-headers regardless of how tempting that might look Jan 29 16:50:41 i've seen some open-source projects do that, for example drm keeps a copy of uapi Jan 29 16:51:01 milloni: right, there are a few different ways to achieve it Jan 29 16:51:01 but that doesn't seem like a good solution, having two sources of information and having to keep them in sync is an antipattern Jan 29 16:52:12 RP: why is hacking linux-libc-headers so bad? i can see it's bad, but why is it *that* bad? Jan 29 16:52:34 RP, find build/sstate-cache/|grep openjre-8, that should give me some results if there was any sstate-cache available, right ? Jan 29 16:53:25 RP, cause I get matches for every other recipe Jan 29 16:53:44 RP, it is as if saving sstate cache was disabled for this particular recipe :-/ Jan 29 16:55:30 milloni: you make glibc, the compiler, *everything* kernel specific to your device Jan 29 16:56:00 milloni: you really shouldn't have a device specific toolchain Jan 29 16:56:28 kroon_: I'd have expected so Jan 29 16:59:18 Crofton|road: it seems either recipe or this kernel module you are building doesnt have inc paths set right Jan 29 17:00:38 The prepare stuff solved the toolchain problem, working on the on target, but need to rewrite rootfs and taking the chance to learn as mux Jan 29 17:04:51 RP: Ya, possibly... perhaps the hash should include timestamps? I don't think we are expecting hash equiv to really work well without out reprodcible builds anyway Jan 29 17:05:03 RP: given that the standard kernel recipe doesn't do this, but it seems useful for cases like mine, would you accept a patch that makes the kernel recipe put sanitized kernel headers somewhere in ${STAGING_KERNEL_BUILDDIR}? Jan 29 17:05:30 JPEW: I think I'm starting to understand. We disabled reproducibile builds on the buildtools tarball. I suspect that is "infecting" the rest of the builds Jan 29 17:05:38 Ah Jan 29 17:05:52 We disabled it because Python was slow Jan 29 17:05:53 JPEW: adding in timestamp checks if repro builds are enabled would perhaps be the best way forward Jan 29 17:06:00 does anyone use package feeds and often think "i wish bitbake would run package-index for me"? Jan 29 17:06:26 rburton: Not anymore but when we did yes :) Jan 29 17:07:48 JPEW: darn Jan 29 17:07:54 was hoping to find a sucker to finish my patch Jan 29 17:08:02 rburton: how would you do it if your feeds are synced out remotely? Jan 29 17:08:50 denix: https://bugzilla.yoctoproject.org/show_bug.cgi?id=10749 is the background to my question, for people who just point a httpd at tmp/deploy. prototype patch enclosed! Jan 29 17:08:50 Bug 10749: enhancement, Medium, 3.99, ross.burton, NEW , Add a class to update package index automatically Jan 29 17:09:08 hoping to find someone other than unassigned to pass it to Jan 29 17:15:49 milloni. there's already a bugzilla for having a secondary package specific headers around. Jan 29 17:15:59 and no, I wouldn't recommend using the staging dir for that. Jan 29 17:16:07 arrange for something package specific period, or use libc-headers. Jan 29 17:17:51 RP: Ensure no timestamp is newer than $SOURCE_DATE_EPOCH Jan 29 17:17:53 ? Jan 29 17:17:55 zeddii: what do you mean by package-specific headers? Jan 29 17:18:03 meaning they can't go into a common location Jan 29 17:18:11 they are for whatever messed up userspace package needds them Jan 29 17:18:23 you can't put it int he common location, then you get some thing using libc-headers, some not. Jan 29 17:18:25 boom. Jan 29 17:18:40 there's a reason we don't do that already. Jan 29 17:23:49 zeddii: but there's no way to have package-specific headers yet? so if I dont want to use linux-libc-headers for reasons given by RP, copying to STAGING_KERNEL_BUILDDIR seems to be the best option for now? Jan 29 17:24:51 you can deploy anything you want to the recipe specific sysroot. Jan 29 17:25:10 JPEW: to ensure non-repro and repro builds don't collide hashes Jan 29 17:26:10 milloni: you can install anything to the sysroot in ${includedir}/my-special-headers or whatever Jan 29 17:27:00 ack Jan 29 17:28:01 milloni: I think we could support putting the sanitzed headers somewhere but needs to be separate from the kernel build, most people survive without this. Simlar to kernel-devsrc I guess Jan 29 17:28:29 ack, i'll try to write something up Jan 29 17:28:34 put it here: Jan 29 17:28:37 https://bugzilla.yoctoproject.org/show_bug.cgi?id=5305 Jan 29 17:28:38 Bug 5305: enhancement, Medium, Future, bruce.ashfield, IN PROGRESS IMPLEMENTATION , Make sanitized kernel headers available Jan 29 17:28:47 since we've already been talking about it for years :D Jan 29 17:29:12 I have patches for it, so we could finalize details and come up with something appropriate. Jan 29 17:29:34 JPEW: is this as simple as http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t222&id=5801b8262377346930d3e32364ca1267279691c6 I wonder? Jan 29 17:31:25 zeddii: https://autobuilder.yoctoproject.org/typhoon/#/builders/102/builds/238 - any ideas? Its your series added to -next :/ Jan 29 17:31:41 checking. Jan 29 17:32:03 zeddii: I didn't think those changes were active so something odd is going on Jan 29 17:32:44 yah. that is odd, since I build a module on target for that very config. but my new series did have a devsrc tweak. Jan 29 17:33:19 zeddii: where can i see your patches? Jan 29 17:33:39 I have to find them first. Jan 29 17:33:51 but I'm leaving for the airport in an hour, so it won't be for a while. Jan 29 17:34:02 that's okay, thanks Jan 29 17:34:12 since the feature request has been around since 2013, I'm sure that won't matter ;) Jan 29 17:34:25 RP: I'll spin up a mips build and see if I see the same thing. Jan 29 17:34:29 wont matter as in the patches will be out of date? Jan 29 17:34:40 no, meaning a few more days for me to find them :P Jan 29 17:34:53 ah, okay Jan 29 17:35:11 RP: so that build has my series, but not the kernel default changes, right ? Jan 29 17:35:20 zeddii: correct Jan 29 17:35:21 which means, it must be my devsrc tweak. I can try 5.2 + the new devsrc patch to see if that triggers it Jan 29 17:35:29 qemumips - alt is just systemd, right ? Jan 29 17:35:38 so I can cobble that together Jan 29 17:35:57 I mainly tested with 5.4 + mips for that, so yes, it could have slipped through. I'll debug and follow up. Jan 29 17:36:14 zeddii: DISTRO="pokyaltcfg" Jan 29 17:36:29 gotcha. will make that switch and build. Jan 29 17:36:31 er, poky-altcfg Jan 29 17:38:06 hello all. i have a qustion. what is best way to add opencv version 4.1.0 package. i have tried to add through my build/local.conf file and bitbake image again. and after that populate my sdk for cross compiling through that image. i am able to compile but unable to run it. Jan 29 17:38:12 tgamblin: I'm going to drop the logrotate patch for now, its confusing the builds more than it helps atm :/ Jan 29 17:38:37 tgamblin: I think it is exposing a pattern though Jan 29 17:38:37 RP: sure thing. I'm also done with that worker if you need to use it for another build Jan 29 17:39:39 RP: I've currently got builds for core-image-sato-sdk set up on two of our internal build servers, one of them is experiencing those error messages for wtmp at boot and one isn't. Makes me suspicious about what in the environment could be causing it Jan 29 17:39:53 Won't say too much more until I've got more data, though Jan 29 17:40:17 tgamblin: its at least good it replicates somewhere else and seems to be a bigger issue Jan 29 17:46:14 hello all. has anyone developed opencv applications with yocto ? Jan 29 17:50:03 RP: build started with that exact series and altcfg for mips. as I mentioned, I'm heading to the airport, so it'll be sometime tomorrow CET at the earliest when I can follow up. Jan 29 17:51:55 zeddii: np, thanks. I'll see if it repeats in this new build and report back Jan 29 17:52:10 zeddii: if it does I'll probably have to drop for now Jan 29 17:53:08 rizi_aquarian: what do you mean you're unable to run it? Jan 29 17:54:39 RP: no worries. it really will be an easy fix, so drop away, and I'll submit a v2 for the devsrc patch. or would you want the whole series as a v2 ? Jan 29 17:55:19 milloni: i added opencv in local.conf file and i am able to cross compile. but when i run this built executable in my qemux86 it says illegal instruction Jan 29 17:56:46 rizi_aquarian: sorry, i've not used qemu with yocto, but it could be just that qemu doesn't emulate whatever opencv needs? Jan 29 17:56:55 maybe try on real hardware? Jan 29 17:58:09 zeddii: whichever :) Jan 29 17:58:25 ack'd Jan 29 18:00:09 milloni: i have to actually scale the function in cloud testing so have to try to work it out for qemu. a question would be how to find out build time and runtime dependencies of the package? Jan 29 18:00:47 rizi_aquarian: i dont think that's the problem, i think qemu doesnt emulate something that you need Jan 29 18:01:05 you definitely dont want to run your computations under an emulated cpu Jan 29 18:01:21 if it's for the cloud you probably dont want to use yocto Jan 29 18:01:58 oh i think i misunderstood, you want to test your image in the cloud, but the target is real hardware? Jan 29 18:02:26 yeah yeah now you understood Jan 29 18:03:22 i would try to find out which instruction it crashes on Jan 29 18:03:42 perhaps by running the program under gdb? Jan 29 18:03:48 RP, should yocto-doc patches be sent to the new doc mailing list? Jan 29 18:04:23 so right now it only says "illegal instruction" Jan 29 18:04:39 armpit: yes Jan 29 18:04:48 armpit: we've not updated for that yet Jan 29 18:05:16 * RP is banging his head against the autobuilder failures :( Jan 29 18:05:25 milloni: sorry what do you mean by gbd? Jan 29 18:05:51 rizi_aquarian: gdb is a very useful unix debugger Jan 29 18:05:54 IMAGE_INSTALL += "gdb" Jan 29 18:06:33 you run your program with it, it should tell you why it crashed (more than just "illegal instruction" Jan 29 18:06:46 ok great thankyou. i will try that. also what IDE you use for crosscompiling? and debugging Jan 29 18:07:07 rizi_aquarian: unix is my ide :) Jan 29 18:07:23 i use vim as the editor Jan 29 18:07:36 gdb as the debugger, gcc as the compiler, and so on Jan 29 18:08:10 milloni: thankyou i will try that. i thought may be you use eclipse or any other :] Jan 29 18:08:11 * armpit sends RP a pillow to soften the blows Jan 29 18:16:22 RP, sent patch to update docs README Jan 29 18:16:40 I will update stable as needed Jan 29 18:19:04 armpit: thanks! Jan 29 18:31:49 zeddii, do we need to maintain https://wiki.yoctoproject.org/wiki/Linux_Yocto#Release_Cadence ? Jan 29 18:33:57 are entries in FILES recursive? Jan 29 18:34:16 i.e for example "/asdf" will include everything under "/asdf"? Jan 29 18:35:36 RP: Ya, probably Jan 29 18:36:21 Could someone please fix https://autobuilder.yoctoproject.org/typhoon/#/builders/75/builds/1506/steps/8/logs/warnings in meta-gplv2? :) Jan 29 18:36:29 Hi, Is there a way I can use xserver-nodm-init with .Xauthority ? Jan 29 18:36:38 * RP pushed a fix for the bukd breakage Jan 29 18:37:21 * JPEW really needs to drop meta-gplv2 Jan 29 18:37:28 RP: I'll give it a look Jan 29 18:38:01 did parsing/handling of DISTRO_FEATURES recently changed? I use @oe.utils.conditional() in distro.conf to alter it and now it throws exception "name 'oe' is not defined"... Jan 29 18:40:43 JPEW: an inherit might fix it Jan 29 18:40:49 RP: Ok Jan 29 18:41:05 JPEW: that also might not be the right fix, not sure Jan 29 18:41:47 denix: You've probably just been lucky in the past. It hasn't changed but variables do change which could influence when its expanded Jan 29 18:45:24 RP: hmm, is it valid to use @oe.utils.conditional() in distro.conf to alter DISTRO_FEATURES? is it too early? I guess I can re-write that condition with pure python, if oe lib is not yet available at that time... Jan 29 18:45:51 Why is shared-mime-info even in meta-gplv2? Jan 29 18:46:20 Ah, itstool Jan 29 18:46:46 RP: or use bb lib instead of oe lib... Jan 29 19:04:47 RP: Hmm, in my failing build environment I can consistently see that boot error, but "bitbake -c core-image-sato-sdk -c testimage" still passes :/ Jan 29 19:10:29 I'm tempted to try a v4 that at least adds another step in the test to make sure the /var/log/wtmp file has the right permissions Jan 29 19:12:12 JPEW: built-time only dependency i *think* Jan 29 19:12:56 rburton: Ya, that seems a little odd; I thought GPLv3 build time dependencies were allowed Jan 29 19:13:11 JPEW: it was marked as a RDEPENDS in master Jan 29 19:13:20 rburton: Ah, OK Jan 29 19:13:43 double-checking now because i''m paranoid i broke it Jan 29 19:14:34 Is a recipe not allowed to have a digit in its name, like "openjre-8" ? Jan 29 19:15:34 kroon: lz4 is a recipe in oe-core Jan 29 19:15:44 and m4 Jan 29 19:15:53 hmm, is the "-8" not allowed then ? Jan 29 19:16:04 more likely the hyphen breaking some bad parsing Jan 29 19:16:08 what actually happens? Jan 29 19:16:27 I dont get any sstate cache generated when building openjre-8 and openjdk-8-native Jan 29 19:16:41 JPEW: right, its build-time only. Jan 29 19:16:54 kroon: none at all? sounds like the recipe is breaking it very badly Jan 29 19:17:05 denix: I'd use lib/bb if you can Jan 29 19:17:56 rburton, "find sstate-cache/ | grep openjre" shows nothing Jan 29 19:18:04 rburton, the same if i grep for openjdk Jan 29 19:19:34 RP: yeah, done pure python condition for now and it parsed, but it's strange that is used to work for years - something now is trying to access DISTRO_FEATURES very early during initial parsing... Jan 29 19:21:21 py2 ; ) Jan 29 19:21:39 rburton: finally looking at the 3 year old bug, thanks...for nothing Jan 29 19:22:37 jonmason: you're welcome Jan 29 19:22:43 jonmason: ignore it for another three years, that's what i did Jan 29 19:24:06 kroon: you meant openjdk not openjre right Jan 29 19:24:22 i have openjdk-8 files in my sstate from when i poked at openjdk last month Jan 29 19:24:28 so its not broken by design at least Jan 29 19:24:59 rburton, I meant both recipes "openjre-8" and "openjdk-8-native" Jan 29 19:25:04 armpit: is your comment about py2 for me? I kind of suspect the same... :) Jan 29 19:25:14 rburton, not openjdk-8 Jan 29 19:26:06 rburton: Huh, weird. Seems like shared-mime-info might be in meta-gplv2 unnecessarly then... I'll still try to fix it anyway. Thanks! Jan 29 19:27:14 JPEW: when I added it, it seemed like the best solution. It may be it can be removed now Jan 29 19:28:06 rburton, but yeah the "-8" doesnt seem to make a difference Jan 29 19:28:19 so how in the heck can a recipe break sstate cache Jan 29 19:35:04 @kroon: PN may contain hyphens, that should be fine Jan 29 19:35:52 kroon: is the openjdk recipe in a public layer? Jan 29 19:35:54 denix, yeah.. maybe now with py3 things are different Jan 29 19:36:11 RP, its in meta-java, master-next Jan 29 19:36:20 py2 or py3 likely makes no difference Jan 29 19:36:40 @RP: meta-java Jan 29 19:36:42 RP, buts its a big thing.. Jan 29 19:37:38 RP, I was giving denix a way out Jan 29 19:39:49 @kroon: BTW I managed in the meantime to build meta-java, master-next, openjdk-8 and openjre-8 and it seems to run as well Jan 29 19:40:04 RobertBerger, do you have sstate-cache for now ? Jan 29 19:40:13 @kroon oh yes Jan 29 19:40:25 RobertBerger, find sstate-cache/ | grep openjre Jan 29 19:40:31 RobertBerger, shows up files ? Jan 29 19:40:34 wait Jan 29 19:41:17 kroon: I did just have a quick glance and didn't see anything obvious Jan 29 19:41:33 @kroon: both openjdk and openjre are in there Jan 29 19:41:54 denix, armpit: There was some change which did have an effect like this. Can't quite remember what it was though Jan 29 19:42:36 RobertBerger, are you on master oe-core ? Jan 29 19:43:16 @kroon - wait. I am on zeus everywhere except for meta-java Jan 29 19:43:35 RP: PARSE_ALL_RECIPES? Jan 29 19:43:37 @kroon - only meta-java is master-next with a few hacks Jan 29 19:43:55 denix: no, wasn't that recent Jan 29 19:44:06 denix: did this just happen? Jan 29 19:44:35 @kroon: master-next: 7d30e6ff7b8a54d67256ce528e8c4fd7e63f6b14 Jan 29 19:45:06 RobertBerger, ok I haven't tried zeus, im on master Jan 29 19:45:38 RP, thanks for checking Jan 29 19:45:46 @kroon: I am with all my other layers on zeus, but could not get the meta-java zeus branch to work, so I moved to meta-java master and master-next Jan 29 19:46:25 @kroon: I had to use host gcc 9 and adjust a few things to make gcc 9 happy Jan 29 19:47:16 @kroon: Now I get: root@6fca7a354060:/# java -version Jan 29 19:47:16 openjdk version "1.8.0_242-internal" Jan 29 19:47:16 OpenJDK Runtime Environment (build 1.8.0_242-internal-ga) Jan 29 19:47:16 OpenJDK 64-Bit Server VM (build 25.242-bga, mixed mode) Jan 29 19:48:17 RobertBerger, the problem I have is that building openjre-8 doesnt seem to generate any sstate cache Jan 29 19:49:33 @kroon: hmmm spooky Jan 29 19:49:48 @kroon: is icedtea in sstate? Jan 29 19:50:08 RobertBerger, yup, icedtea7-native Jan 29 19:50:34 so im suspecting the "-8" is fooling the sstate cache Jan 29 19:50:48 @kroon I have both jdk-8 and jdk-8-native there Jan 29 19:51:00 "openjre-8-242" is name and version Jan 29 19:51:14 RobertBerger, yes but you are on zeus right ? Jan 29 19:51:15 RP: may not be very recent, haven't tried bisecting yet... Jan 29 19:51:28 @kroon: yes poky/zeus Jan 29 19:54:13 @kroon: wait a bit Jan 29 19:54:39 pokyuser@196d78e83b84:/workdir/build/container-x86-64-java$ bitbake openjdk-8 -e | grep ^PN= Jan 29 19:54:39 PN="openjdk-8" Jan 29 19:54:39 pokyuser@196d78e83b84:/workdir/build/container-x86-64-java$ bitbake openjdk-8 -e | grep ^PV= Jan 29 19:54:39 PV="242" Jan 29 19:54:39 pokyuser@196d78e83b84:/workdir/build/container-x86-64-java$ bitbake openjdk-8 -e | grep ^PR= Jan 29 19:54:39 PR="r0" Jan 29 19:55:49 RobertBerger, sstate cache generation has changed inbetween zeus and master. can you try master ? Jan 29 19:56:23 @kroon: openjre-8-242 would be wrong, it is openjdk-8_242.bb Jan 29 19:57:26 @kroon: I will try. So you just want me to switch poky to master-next, yes? Hope not too many things will break. Jan 29 19:57:56 RobertBerger, master would be enough. but rebuilding would take some time I guess Jan 29 19:59:20 @kroon: The rebuilding time is not an issue. I can check tomorrow morning - I am in Germany and will build in Austria on my servers somewhere. Just fixing build issues will be more tricky while sleeping ;) Jan 29 20:00:06 RobertBerger, thanks, i'd appreciate it! Jan 29 20:27:41 @kroon: I need to turn all other layers also to master since quite a few bbclasses changed and are not avail anymore, so I am not sure i will be able to have results any time soon Jan 29 20:27:56 RobertBerger, yeah, its ok Jan 29 20:28:30 RobertBerger, no need to bother, its a little too much to ask for Jan 29 20:28:47 RobertBerger, ill see if I can bisect this Jan 29 20:29:56 @kroon: let me put my current hacks under version control an I'll see what I can do maybe on some other server so I don't f*ck up my current setup Jan 29 20:41:42 denix: I'd suspect 6-9 moths ago there was something Jan 29 20:55:03 RP: no, not that long :) few days, maybe a week or two ago it was working for me Jan 29 20:56:33 RP, I see the sstate files being generated under work/. But they arent copied to sstate-cache/ dir. could you point me to which function is supposed to take care of that ? Jan 29 21:00:32 can anyone point me to an example of PACKAGECONFIGs "building on top of each other" conditionally? Jan 29 21:00:49 for example, package mpv can be built for x11 or wayland Jan 29 21:01:22 so you can configure mpv with --enable-x11 or --enable-wayland, depending on a DISTRO_FEATURE, no problem Jan 29 21:02:24 but if you're also using opengl, then you can also configure with --enable-gl-x11 or --enable-gl-wayland Jan 29 21:03:14 the same goes for vaapi and vdpau, even leading to things like --enable-vdpau-gl-x11 Jan 29 21:04:53 kroon: sstate.bbclass creates the file in SSTATE_DIR, sstate_create_package() in that class Jan 29 21:05:02 so is there a way to say: "if DISTRO_FEATURES contains 'opengl' and 'x11' and the user has asked for PACKAGECONFIG[vdpau] then configure with --enable-vdpau-gl-x11"? Jan 29 21:05:30 @kroon: This is what I am trying at the moment: meta Jan 29 21:05:30 meta-poky Jan 29 21:05:30 meta-yocto-bsp = "master:1cebd3469e295b4b1ae46effe5cf521b5264ee05" Jan 29 21:05:30 meta-resy = "master:9b1976b0c750441a597ab579fdbdaa4b9dac1054" Jan 29 21:05:30 meta-java-master = "master-next:7d30e6ff7b8a54d67256ce528e8c4fd7e63f6b14" Jan 29 21:05:31 meta-oe = "master:4043bcc714d4aa7abaee7e8e5911f76701207497" Jan 29 21:05:47 denix: right, I don't know of a change that recently Jan 29 21:05:50 @kroon and I have a fresh SSTATE dir, so we shall see (hopefully soon) Jan 29 21:07:57 RP, its created there, but then it is copied/ln:ed to build/sstate-cache/ somewhere else ? Jan 29 21:08:49 RP, cause sstate_create_package() is exiting early since the file already exists Jan 29 21:09:14 kroon: where is SSTATE_DIR pointing? Jan 29 21:09:32 kroon: it should be created into the sstate cache there Jan 29 21:10:00 kroon: it sounds like SSTATE_DIR is incorrect Jan 29 21:11:09 SSTATE_DIR="/tmp/ramdisk/xxx/work/i686-oe-linux/openjre-8/242-r0/jdk8u-jdk8u242-ga/sstate-cache" Jan 29 21:11:10 @kroon: I see in my build: find | grep openjdk Jan 29 21:11:10 ./02/31/sstate:openjdk-8-native::242:r0::3:02310a5ba4da6551097a81232b3c780966a72f840299f571af9d57f6e9b39c72_fetch.tgz.siginfo Jan 29 21:11:10 ./23/78/sstate:openjdk-8::242:r0::3:23780bf55e6d385cfad7cae2180dfd458fcb1cd89d12a72840e57565984ae976_fetch.tgz.siginfo Jan 29 21:11:42 @kroon so I think it will populate my SSTATE Jan 29 21:12:07 RP, yes, it looks incorrect compared to other recipes Jan 29 21:13:28 RP, aha... Jan 29 21:14:02 RP, openjdk-8-common.inc does export TOPDIR="${S}" Jan 29 21:14:12 aha Jan 29 21:14:54 not in master, fwiw Jan 29 21:15:06 @kroon: SSTATE is being populated also with master as it looks like Jan 29 21:15:34 RP, thank you, this was driving me nuts... Jan 29 21:17:59 kroon: my openjdk-8-common in meta-java master doesn't set that Jan 29 21:18:14 rburton, no, this is with master-next Jan 29 21:18:17 ah Jan 29 21:19:23 best tell richard before it merges to master! Jan 29 21:19:39 rburton, yeah I just sent an email Jan 29 21:19:40 there is a JDK_TOPDIR used for presumably this exact reason :) Jan 29 21:19:45 right dog walk Jan 29 21:19:53 rburton, me too :-D Jan 29 21:20:39 @kroon, are you still searching why SSTATE is not populated with openjdk-8 stuff? Jan 29 21:21:18 @kroon: I have meta-java/master-next and poky/master and SSTATE is being populated Jan 29 21:28:59 JPEW: thanks! Jan 29 21:29:09 RP: np Jan 29 21:29:24 JPEW: sadly that timestamp change didn;t work so either its wrong or my theory is wrong Jan 29 21:30:38 RP: Hmm, OK Jan 29 21:33:54 JPEW: repro-fail/build-st-27398 is the difference, will investigate Jan 29 21:36:39 JPEW: 1970 vs 2019 timestamp differences Jan 29 21:36:59 RP: That looks the same as the one for which I logged the bug Jan 29 21:37:07 JPEW: yes :/ Jan 29 21:37:18 * JPEW Checks if it's the exact same date... Jan 29 21:37:45 JPEW: one is a timestamp of "0", the other is the repro build timestamp Jan 29 21:38:13 RP: Ya, I wonder where the 0 came from Jan 29 21:38:43 RobertBerger, ah sorry, no the TOPDIR thing was the reason Jan 29 21:38:59 JPEW: fixed_source_date_epoch ? Jan 29 21:39:05 @kroon I am still confused Jan 29 21:39:33 @kroon how does my poky/master, meta-java/master-next then populate SSTATE? Jan 29 21:39:53 JPEW: I'll pause a worker and run this manually I think Jan 29 21:40:43 RobertBerger, yeah thats a fair question Jan 29 21:41:14 RobertBerger, what do you get when you run "bitbake -e openjre-8|grep ^SSTATE_DIR=" Jan 29 21:41:40 @kroon - wait I switch the environment and let you know Jan 29 21:42:17 RP: OK. Perhaps grab the __source_date_epoch.txt file if you can Jan 29 21:43:04 JPEW: the previous build was cleaned up but hopefully reproduces and I can narrow it down Jan 29 21:43:16 JPEW: I suspect something to do with allarch Jan 29 21:44:13 Oh my, allarch.bbclass is busier that I would have expected Jan 29 21:45:26 JPEW: it doesn't just set packagearch=any, but it does its best to stop compilation entirely Jan 29 21:46:48 rburton: its signature problems which are the hard part Jan 29 21:47:29 I can't decide whats stranger, that the timestamp from sstate is 0, or that the clean build timestamp is the current date Jan 29 21:47:35 I think the latter Jan 29 21:47:53 At least the 0 date has an obvious path through the code Jan 29 21:50:09 @kroon: it looks good: Jan 29 21:50:10 bitbake openjdk-8 -e | grep ^SSTATE_DIR= Jan 29 21:50:11 SSTATE_DIR="/workdir/sstate_zeus_master/" Jan 29 21:50:11 pokyuser@2cc132871621:/workdir/build/container-x86-64-java-master$ bitbake openjre-8 -e | grep ^SSTATE_DIR= Jan 29 21:50:11 SSTATE_DIR="/workdir/sstate_zeus_master/" Jan 29 21:50:11 pokyuser@2cc132871621:/workdir/build/container-x86-64-java-master$ Jan 29 21:51:17 @kroon: /workdir/sources/meta-java-master/recipes-core/openjdk$ grep -r TOPDIR . Jan 29 21:51:17 ./openjdk-8-common.inc:export TOPDIR="${S}" Jan 29 21:52:02 @kroon: I use meta-java on master/next which contains: export TOPDIR="${S}" Jan 29 21:52:05 i'm guessing one of you has SSTATE_DIR set absolutely and the other has it set using TOPDIR Jan 29 21:52:21 ah yeah Jan 29 21:52:29 @rburton: I have it absolutely Jan 29 21:52:30 im using default SSTATE_DIR Jan 29 21:52:47 SSTATE_DIR ?= "${TOPDIR}/sstate-cache" <-- default Jan 29 21:52:49 JPEW: thanks to you, mpv-0.32.0 and chromium-79.0.3945.117 running on rock-pi-4 accelerated with panfrost! :-D Jan 29 21:53:01 tlwoerner: Awesome! Jan 29 21:53:21 i just need to figure out the configure maze for mpv before i can push the update Jan 29 21:53:29 @rburton: in site.conf: SSTATE_DIR = "/workdir/sstate_zeus_master/" Jan 29 21:53:42 RobertBerger, thats why you dont have the problem Jan 29 21:54:06 @kroon - that's why I don't have THAT problem ;) I have others with meta-java ;) Jan 29 21:54:19 RobertBerger, :-D Jan 29 21:54:47 what's the trick that allows a partition to auto-resize on bootup? Jan 29 21:57:18 RP: Hmm, I wonder if __source_date_epoch.txt is empty... Jan 29 21:59:42 rburton: For trying to prevent compiling, ca-certificates sure looks like it's compiling with allarch.bbclass :) Jan 29 22:00:09 rburton: Ah, probably not actually compiling, just has a Makefile Jan 29 22:50:42 JPEW: well, it totally blew up :/ Jan 29 22:54:16 JPEW: build restored from sstate has no epoch file and 1970 timestamps Jan 29 22:54:55 hm, logs say the file should be there Jan 29 22:55:17 ah, it is Jan 29 23:03:58 JPEW: I think I know the cause. its from when we started stopping sstate overwriting files with the same hash in sstate Jan 29 23:05:47 hmm, that doesn;t quite make sense though Jan 29 23:28:23 JPEW: the idea of tracking timestamps won't work as it doesn't apply for all tasks :/ Jan 29 23:29:03 JPEW: we don't track the timestamps of the created ipks/debs for example but do for do_package output apart from maybe the top level dir Jan 29 23:29:30 JPEW: but that also shows my theory is totally wrong and I cant see how it can restore files from 11Jan like this Jan 29 23:29:45 * RP has more questions than answers and will have to sleep on it Jan 29 23:33:13 * RP is sad to have to pass on that Scott Rifenbark has died. The docs won't be quite the same without him :(. Jan 29 23:35:01 Very sad to hear that RP. Thank you for passing on the news. Jan 29 23:38:11 oh, sad news, indeed. we'll miss Scott Jan 29 23:47:50 yeah.. very sad. he will be missed Jan 29 23:52:21 RP: :( Jan 29 23:53:37 RP: sad to hear about Scott, He would first try the procedure or instructions before documenting it, I was really impressed Jan 30 00:04:40 oh no, sorry to hear that :( Jan 30 00:12:01 RP: when you're up tomorrow https://github.com/kergoth/vim-bitbake/issues/21 could use a quick opinion Jan 30 00:30:05 thank you, RP, for sharing that news, he will be missed indeed Jan 30 00:44:54 JPEW: interesting data point, i bought one of these: https://www.canadacomputers.com/product_info.php?cPath=5_1336_96&item_id=137879 and now my rock-pi-4 boots just fine with the rpi serial cable plugged in (i.e. on the RX pin) **** ENDING LOGGING AT Thu Jan 30 02:59:57 2020