**** BEGIN LOGGING AT Wed Apr 29 02:59:57 2020 Apr 29 06:56:14 build #351 of kirkwood/generic is complete: Failure [failed images] Build details are at http://buildbot.openwrt.org/master/images/builders/kirkwood%2Fgeneric/builds/351 blamelist: Adrian Schmutzler , Pawel Dembicki Apr 29 09:24:45 * ldir (very late) mornings for another bright & breezy day in lockdown Apr 29 09:24:45 hrm mocp is the only victim of the ffmpeg 4 update. wonder if anyone uses it. Apr 29 09:25:58 actually it's overcast, dull & breezy. Not raining yay. Apr 29 09:26:30 warm and sunny here, since weeks already Apr 29 09:26:54 almost a drought again Apr 29 09:28:19 ha ha - success - ldir has turned jow English... he's talking about the weather. ;-) Apr 29 09:29:53 english would be talking about _wet_ weather Apr 29 09:30:47 lol Apr 29 09:36:02 once this build has finished and I can switch git branches, I'll submit the missing luci pr for sqm-stats. Apr 29 10:02:18 * ldir submits a luci pr - waits for screams in German equivalent to "what is this shit code?" - he he he :-) Apr 29 11:14:58 anyone ever heard of lgtm? https://lgtm.com/ Apr 29 11:15:11 ldir: how far is the collectd side? Apr 29 11:21:11 jow: I think it's there. Certainly there's nothing that screams at me. It would be better if it were a native collectd plugin of course but I'm not going there. https://github.com/openwrt/packages/pull/11941 Apr 29 11:24:20 The rrdtool logarithmic labelling patch makes things readable but isn't essential. Apr 29 11:24:38 luci side looks okay Apr 29 11:24:48 Seen your comment Apr 29 11:24:56 will move the script Apr 29 11:25:42 The permissions thing. Apr 29 11:26:10 https://unix.stackexchange.com/questions/312146/what-is-the-purpose-of-usr-libexec Apr 29 11:26:33 Am I correct in thinking that you can't have the permissions correct in the image file? Apr 29 11:26:48 So a 'postinstall' script doesn't help you? Apr 29 11:26:49 yep, thats a known deficiency in the build system Apr 29 11:27:11 we'd need some kind of fakeroot support to be able to create arbitrary permissions and ownerships at build time Apr 29 11:27:41 or somehow stage all files in tar archives or loop mounts Apr 29 11:27:43 ok. So my init script find thing is the best workaround Apr 29 11:27:59 if you do not want to pollute the init script then a postinstall secito nfor your plugin script might make sense too Apr 29 11:28:54 but if I build an image containing the plugin I'd still have the ownership/perms problem no? Apr 29 11:29:18 no since the postinstall is deferred to firstboot Apr 29 11:30:01 ah, ok. and firstboot still runs after a sysupgrade ? Apr 29 11:30:26 or so I thought, could be that I'm just confused Apr 29 11:30:34 lol Apr 29 11:31:27 I might have confused it with the luci package postinstall mechanism Apr 29 11:31:56 that one stages uci default scripts and emits postinstall code to source the uci-defualts script right away when isntalling on target Apr 29 11:32:24 I'll move the script. I looked but didn't see anyone changing perms in a post-install script, and when I tried it it didn't work. Apr 29 11:32:24 so it runs deferred on firstboot for builtin packages or immediately for opkg installed ones Apr 29 11:36:27 I was very pleased with my "find /usr/libexec/collectd ! \( -user nobody -a -group nogroup \) -exec chown nobody:nogroup '{}' '+'" Apr 29 15:20:59 https://openwrt.org/toh/gl.inet/gl.inet_gl-usb150 Apr 29 15:21:04 pretty cool Apr 29 15:39:56 that looks neat! does that work if you just plug it into a wall socket with usb adapter? Apr 29 15:43:40 I'm just dealing with a kernel size increase due to switch to DSA on kirkwood Apr 29 15:44:28 Interestingly, the buildbot config results in uimages being about 87k bigger than with default setup from "make menuconfig" Apr 29 15:44:49 its possible that feed packages enable kernel bloat Apr 29 15:44:54 or kmods Apr 29 15:45:09 so that would have been my question Apr 29 15:45:53 you should try a build with all pacakges from all feeds installed, then nuke .config and start with CONFIG_TARGET_kirkwood=y CONFIG_ALL_KMODS=y CONFIG_ALL_NONSHARED=y + make defconfig Apr 29 15:48:05 but after all, that's nothing extraordinary then, but just to be expected and taken into consideration regularly? Apr 29 15:48:22 you mean the size increase? Apr 29 15:48:53 yes, i did not expect just building the packages to increase the kernel Apr 29 15:49:18 it should not. I do not really suspect packages, I think the difference is building all kmods vs. building some kmods Apr 29 15:49:43 since some kmods will enable entire kernel subsystems when built Apr 29 15:53:10 so, being heretic, building with all kmods will unnecessarily increase the image for those not using those kmods Apr 29 15:53:44 but otherwise, some of the kmods would be unusable? Apr 29 15:53:56 thats the nature of modular systems. There'll always be overhead to accomodate for eventually needed functionality Apr 29 15:54:37 yes, otherwise those extra kmods will no be loadable Apr 29 15:54:38 that's no criticism, but just a question whether I understood it correctly Apr 29 15:54:53 they'll lack exported symbols from the kernel Apr 29 15:55:57 okay, thanks for explaining. Apr 29 15:56:45 still wouldn't have expected it to be that much, as well as i'm surprised about the 47k for DSA support Apr 29 15:56:59 maybe I dealt with tiny devices too much :-) Apr 29 15:57:12 thats progress Apr 29 15:57:45 more footprint for a fraction of the old functionality Apr 29 15:59:53 fortunately, those device have enough flash (64M) after all, and it's just the kernel partition which restricts Apr 29 21:47:16 OpenWrt builds on ubuntu 20 i assume? Apr 29 22:32:12 Slimey: no, definitely not. Apr 29 22:33:20 build #354 of ramips/rt3883 is complete: Failure [failed tools] Build details are at http://buildbot.openwrt.org/master/images/builders/ramips%2Frt3883/builds/354 blamelist: Daniel Golle , Adrian Schmutzler , David Bauer , Kevin Darbyshire-Bryant Apr 29 22:33:21 build #340 of layerscape/armv8_64b is complete: Failure [failed images] Build details are at http://buildbot.openwrt.org/master/images/builders/layerscape%2Farmv8_64b/builds/340 blamelist: Daniel Golle , Adrian Schmutzler , David Bauer , Kevin Darbyshire-Bryant bryant.me.uk> Apr 29 22:58:07 rotanid: it doesn't? where does it fail? Apr 29 22:58:27 i guess i could just not be lazy and try it ;) Apr 29 22:59:48 m4t: Slimey: sorry, i misunderstood the question. "builds on" versus "is based on" Apr 29 23:03:07 ah Apr 29 23:48:16 aparcar[m]: Hauke: how do I strip a signed image back down to the raw image? Apr 29 23:57:50 philipp64: Does the signing cause any trouble? It is automatically removed via the fwtool, does it break if you dd it on a drive? Apr 29 23:58:30 which option removes it? ‘-t’ or ‘-T’? Apr 29 23:59:28 I’m trying to mount the raw image with “losetup” (as a raw disk, so I can run “fdisk” or “sgdisk” on it) but “losetup” doesn’t like the non-disk data (the meta data)… so I have to strip it to do so. Apr 30 00:25:07 what is in openwrt the correct way to force some kernel config options for some specific board to be built into the kernel image, instead as a module? Apr 30 00:26:03 if i add these options into target/linux/mvebu/cortexa53/config-5.4, they will take for all mvebu cortex-a53 boards... Apr 30 00:39:39 philipp64: fwtool -t -s signature.sig image.bin Apr 30 00:43:10 and I guess the same with -i to extract the metadata? Apr 30 00:43:19 philipp64: yes Apr 30 00:44:20 okay… sgdisk is complaining about the backup GPT being corrupted… is that normal? Apr 30 00:46:18 honestly, I don't know Apr 30 01:05:14 remind me how to force 5.x kernel in a 4.19 build Apr 30 01:13:46 Slimey: you mean activating testing kernels? Apr 30 01:35:50 yes Apr 30 02:10:29 build #352 of kirkwood/generic is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/kirkwood%2Fgeneric/builds/352 **** ENDING LOGGING AT Thu Apr 30 02:59:57 2020