**** BEGIN LOGGING AT Mon Apr 12 02:59:56 2021 Apr 12 07:25:30 yo dudX Apr 12 08:52:40 good morning Apr 12 09:10:38 Morning! Apr 12 09:50:01 * LetoThe2nd mourns. Apr 12 11:49:13 https://www.instagram.com/p/CNkCiiJpC8y Apr 12 12:17:38 LetoThe2nd: :-) Apr 12 12:18:32 * LetoThe2nd wants to feel like one of the cool kids at least once! :) Apr 12 12:19:07 cool kids use tiktok Apr 12 12:25:59 derRichard: https://www.tiktok.com/@theyoctojester/video/6838971178602745094 Apr 12 12:26:57 LetoThe2nd: lol :-) Apr 12 12:27:01 * derRichard has the same t-shirt Apr 12 12:27:58 derRichard: and who says that i'm not already preparing the YPS announcement there? Apr 12 12:28:14 you're my hero! Apr 12 12:28:19 * LetoThe2nd bows. Apr 12 12:30:10 but on a more serious note: we're really trying to improve on "marketing" and working on understanding the audience and how to reach out. but it's "complicated" Apr 12 12:32:00 marketing to find new customers or to promote yocto itself? Apr 12 12:34:01 the latter, as speaking for those who are involved in yocto "advocacy". i'm not a consultant nor working for a company that sells such services, so finding customers is certainly none of my interests. unless you include babbling into a webcam while having a beer as paid consulting, then i'm onviously totally interested! Apr 12 12:35:24 IMHO promoting events on linkedin/xing works well Apr 12 12:37:44 LI is certainly one of the major platforms for our kind of project/business, yes. Apr 12 12:55:22 Hello, I am working on the node recipe. It need to build the host tools in 32bit. Apr 12 12:56:03 How can I add a dependency on a 32 bit native recipe (here zlib for example) Apr 12 13:04:41 kmaincent: i'd prefix the recipe name with "lib32-" to depend on the 32bits variants Apr 12 13:07:13 derRichard: i don't think lib32 would apply to native recipes Apr 12 13:07:21 oh. :-( Apr 12 13:12:41 nodejs already has a recipe. why do you need to build host tools in 32-bit? Apr 12 13:14:16 rburton: because currently it can not built snapshot, and we won't be able to move to nodejs v14 without fixing this. Apr 12 13:14:55 i'm struggling to believe that nodejs *needs* 32-bit binaries to build on a 64-bit host Apr 12 13:16:28 yes, the tools built need to be in the same 32/64 bit as the target to use them properly. Apr 12 13:16:40 the tools are stupid and should be fixed then Apr 12 13:18:05 mmh not enough expert to fix v8 engine. Apr 12 13:18:53 and I think there is a reason that v8 needs to have it working like that. Apr 12 13:19:08 normally, because people don't think about cross Apr 12 13:24:35 rburton: s/about cross//g Apr 12 13:24:40 lol Apr 12 13:26:14 :) Apr 12 15:03:56 Hi all! I have a tool that make use of the Tinfoil API, is there any way to use it from a recipe? When I try I get "No reply from server in 30s" Apr 12 15:21:31 RP: Mind another go at master-next of meta-mingw? I updated the SHA1s so the fetch should succeed now. Apr 12 15:21:52 It's also possible that the 7.0.0 -> 8.0.0 upgrade will fix GCC 11 Apr 12 15:22:56 JPEW: https://autobuilder.yoctoproject.org/typhoon/#/builders/89/builds/3303 Apr 12 15:55:12 how can I add a newline in a variable? Apr 12 16:03:11 vdl: what exactly do you need a newline for? Apr 12 16:05:08 qschulz: a layer I'm using requires a variable to have newlines in it Apr 12 16:06:51 qschulz: because it does file.write(d.getVar('FOO').replace(r'\n', '\n')) Apr 12 16:10:31 JPEW: failed but much later than last time: https://autobuilder.yoctoproject.org/typhoon/#/builders/89/builds/3303 Apr 12 16:16:32 Hi. is it possible to invoke strip with the --strip-debug option in the strip part of binary handling? I want to do this to get stack traces in the systemd journal when a program crashes Apr 12 16:16:41 vdl: shouldn't \n just work then? or you have some typo in your replace()? Apr 12 16:17:15 without this, only the current call name of each thread is shown in the log, and no complete call traces Apr 12 16:18:35 Scorpi: IIRC, you can install the debug package version (-dbg), which is not stripped, would that work for you? Apr 12 16:19:52 qschulz: I know that. My aim is to get call traces of devices in production enviroments Apr 12 16:20:22 qschulz: installing a debug package is fine for cases where I know that there is a problem and how to reproduce it Apr 12 16:21:25 sounds like you want to always install that one then. Apr 12 16:22:07 kergoth: that would be an option if it wouldn't be that large Apr 12 16:24:05 kergoth: that's why I want to use strip --strip-debug, to get binaries that are only slightly larger, but get full call traces Apr 12 16:25:44 qschulz: no typo. Do you need a literal \n in the string? is it what r\n means? Apr 12 16:26:18 Scorpi: https://sourceware.org/gdb/onlinedocs/gdb/MiniDebugInfo.html may be of interest, set ting PACKAGE_MINIDEBUGINFO = "1" will cause our package.bbclass split_and_strip to inject that compressed mini info into the binary without the -dbg package installed Apr 12 16:26:32 "The intent of this section is to provide extra minimal debugging information for use in simple backtraces" Apr 12 16:27:26 kergoth: oh, thanks, I'll try that Apr 12 16:27:45 i'd never heard of it before now, was reading package.bbclass to see how to change the strip arguments :) Apr 12 16:27:48 but seems promising Apr 12 16:28:53 damn, not supported by my poky version as it seems Apr 12 16:28:57 (aside: looks like the arguments are hardcoded in meta/lib/oe/package.bbclass in runstrip(), and the value of the STRIP variable is assumed to be a single executable name, not arguments, so you can't currently influence it) Apr 12 16:28:59 Hi all, I am trying to copy a folder containing files and symbolic links using "cp -r", but I still see files missing in RFS. any pointer what I could be missing ? Apr 12 16:29:05 https://github.com/openembedded/openembedded-core/blob/master/meta/lib/oe/package.py#L42 Apr 12 16:29:23 but you could presumably do INHIBIT_PACKAGE_STRIP and then add your own packagefunc to strip it the way you want Apr 12 16:30:21 kergoth: yes, I assumed that as a possible option, thanks for clarifying Apr 12 16:35:22 any clue about the best way to submit patches to meta-linaro ? their README suggests their gerrit, which has no visible "create account", and openembedded@lists.linaro.org, whose archives at https://lists.linaro.org/mailman/listinfo/openembedded stop abruptly on last Dec 1st. Would have asked Koen but he's not been very visible here lately, right ? Apr 12 16:37:58 vdl: r'\n' is a python raw string (basically, replace looks for the character \ followed by the character n and replaces it with \n (which is a newline) Apr 12 16:38:03 vdl: c.f. https://docs.python.org/3/reference/lexical_analysis.html#string-and-bytes-literals Apr 12 16:38:09 (learned something new today :) ) Apr 12 16:41:16 qschulz: got it, thanks. Apr 12 17:21:04 how do I need to format my SRC_URI to clone a local repository (just for testing)? Apr 12 17:21:42 bitbake isn't happy about git://${TOPDIR}/../project" Apr 12 17:24:06 vdl: just give it the absolute path on your system without using ${TOPDIR} or .. Apr 12 17:24:22 git:///home/vdl/yocto/some-git for example Apr 12 17:24:45 or maybe it works already, you need to add ;protocol=file after git://${TOPDIR}/../project Apr 12 17:28:43 qschulz: it complains about invalid SRC_REV even though the commit exists Apr 12 17:31:59 vdl: does it exist in the master branch? Apr 12 17:32:12 obviously SRCREV instead of SRC_REV works... Apr 12 17:32:18 * vdl face palm Apr 12 17:32:51 The variable naming can be confusing with OE/bitbake :-) Apr 12 17:34:31 qschulz: git:///home/vdl/yocto/some-git;protocol=file is what I needed btw (since I don't run a git server). Thank! Apr 12 17:34:33 +s Apr 12 17:42:15 vdl: unfortunately, I'm not sure we can detect which variables are not used Apr 12 17:42:55 I mean we probably can, but bitbake.conf (for example) which is used as base for all recipes probably contains lots and lots of variables that aren't used by the recipe Apr 12 17:49:37 qschulz: I meant that some variables use underscore, others don't, there are no strong naming convention for variables, making it hard to catch typos Apr 12 17:51:08 vdl: yes I know, the point was that I was not sure we would be able to catch typos Apr 12 17:51:18 but that'd be super neat Apr 12 17:52:30 hm, Koen's email as listed in meta-linaro/README is bouncing, any needs for concern ? Apr 12 18:03:07 yann: I think that README should be updated Apr 12 18:06:06 does yocto requires the setup.py file of a python application to be located at the root of the project, or can it be e.g. in an src/ subdirectory? Apr 12 18:24:58 Guest75177: we'll see if my emails at least reach someone... Apr 12 18:27:06 I was also wondering if it would not make any sense to have meta-optee as a standalone repo, rather than living in meta-linaro where it's not obvious it's even maintained Apr 12 18:28:46 Hello, i'm trying to install yocto on Garuda linux(Arch linux), following the instructions on Arch wiki, but when installing "bitbake" i'm getting dependency issue, that python-django18 package was not satisfied, when searched for that package on arch website, is there any other way for resolving this issue Apr 12 18:30:05 Guest52974: you shouldn't be "installing" bitbake at all Apr 12 18:35:14 yann: meta-arm has optee recipes too Apr 12 18:36:22 Guest52974: and Arch Linux is not supported by Yocto, we suggest to use containers instead. E.g. pyrex or CROPS (github projects, easy to find) Apr 12 18:38:18 qschulz: cloning a local copy ended up being stupid. How do I point to a local source tree? bitbake doesn't like SRC_URI = "file://${TOPDIR}/../app/subdir". Do I need EXTERNAL_SRC or something? Apr 12 18:38:20 qschulz: oh, sadly the layer index pointed me only to meta-optee when I searched for "optee" in layers, and I missed that :/ Apr 12 18:38:48 there are sooo many optee recipes in tons of repos! :fearful: Apr 12 18:39:21 ok, next step rebase my work onto meta-arm :) Apr 12 18:40:03 vdl: you said it worked, why has it stopped now? Apr 12 18:40:22 you can modify S if you want to have your source tree in a subdir of your git repo Apr 12 18:41:33 qschulz: the python app I'm writing the recipe for has its setup.py into a app_root/src/app subdirectory Apr 12 18:42:04 and I realized for quick testing git:// wasn't ideal (I'll have to commit any tweaks in the sources first) Apr 12 18:43:58 vdl: use devtool for debugging Apr 12 18:44:10 vdl: or externalsrc (which is the mechanism used by devtool) Apr 12 20:31:42 Hi. On my platform opkg is very slow to install packages, about a minute to install a 5kb package. Is there any way to speed this up? Apr 12 20:32:03 It seems like a second run is much faster than the first one, so I assume that it's io bound Apr 12 20:36:08 parsing stuff in /var/lib/opkg/info ? Apr 12 20:38:14 can a raw file on the SD card be mounted as a file with WIC? e.g. part /mnt/foo --source rawcopy --sourceparams "file=foo"? Apr 12 20:39:23 fullstop: seems related to io yes, perhaps you can benchmark your disk Apr 12 20:40:13 khem: That sounds like an exercise in sadness but I guess it needs to be done. Apr 12 20:41:01 It's overlayfs over encrypted ubiblock of squashfs, if that makes sense. Apr 12 20:41:16 Many layers. With that being said, it was slow before I added them. Apr 12 21:08:49 strace shows me that it's spending most of the time reading .list files from /var/lib/opkg/info Apr 12 21:09:05 There's close to 1000 of those. Apr 12 21:33:12 fullstop: I dont think opkg can help here, perhaps its your file system setup which is to blame here Apr 12 21:33:43 fullstop: maybe mount /vaar/lib/opkg on tmpfs and see if it speeds it up, if it does then you know for sure its not opkg Apr 12 21:34:09 right, I can just copy the files to a directory in tmp and set the config option for opkg to test that. Apr 12 21:42:03 khem: that significantly cut down the time required. Some of it was systemd dbus stuff after reinstalling the service, but it is about 50% of the time required before. Apr 12 21:42:33 this was measured after "echo 2 > /proc/sys/vm/drop_caches" Apr 12 21:43:19 hmm dropping the caches will slow it down. Are you measuring raw performance Apr 12 21:44:02 No, "time opkg install --force-reinstall PKGNAME.ipk" Apr 12 21:44:24 caches are dropped in both cases, though, so clearly a significant bottleneck is reading the files from NAND. Apr 12 21:44:40 I did not understand what reduced to 50% Apr 12 21:44:56 and how is systemd dbus coming into picturee Apr 12 21:45:21 Sorry, it was about 60s with the info directory on flash and about 30s with the info directory on tmpfs Apr 12 21:45:36 systemd dbus is coming into the picture because of postinst commands, presumably. Apr 12 21:46:44 yes, the postinst runs some systemctl stuff. This happens in both cases, though, and I understand that it will take some additional time on this particular package. This will not be the case with all packages, so I'm not overly concerned with the additional time that it takes.. Apr 12 21:47:19 right yes services will be restarted if they are updated etc Apr 12 21:48:10 so just changing the file system reduced it to half so you know where the problem is Apr 12 21:48:43 Yes, but unfortunately there's not much that I can do about that one. Apr 12 21:49:15 other than cheat.. and cat all of the files in the info directory at startup. :-) Apr 12 21:50:17 I remember doing that with an oracle database decades ago in order to pre-load certain tables in the cache after a cold boot. Apr 12 21:50:24 perhaps look into your file system as well. IMO opkg is perhaps symptom Apr 12 22:59:13 Can bitbake function without inotify capability? Apr 13 00:09:16 do you guys define kernel .cfg snippet per feature, or per kernel driver/subsystem? (e.g. a feature foo.cfg requiring both squashfs and dm-verity, or squashfs.cfg and dm.cfg) Apr 13 00:17:05 vdl: there is no hard and fast rule, but it cuts across features which work together and are dependent Apr 13 00:17:36 in your examples they look more like independent Apr 13 00:18:19 ecdhe: bitbake needs inotify **** ENDING LOGGING AT Tue Apr 13 02:59:56 2021