**** BEGIN LOGGING AT Wed Nov 13 02:59:58 2019 Nov 13 03:24:12 rmilecki: /me recalls you were working on sysupgrade recently, can you tell me why stage2 needs /usr/share/libubox/jshn.sh? see: http://patchwork.ozlabs.org/patch/1193891/ Nov 13 05:02:58 ath79 uses dts files instead of the glue files ar71xx used, so why does mpc85xx have both a dts file and a glue file for each device? Nov 13 05:47:43 russell--: i don't know... shouldn't you know the answer? ;) Nov 13 05:47:49 russell--: what error did you see? Nov 13 06:06:23 rmilecki: i didn't see any message at first, just noticed that the update didn't apply Nov 13 06:06:43 oh, that's even worse Nov 13 06:06:59 eventually, by short-circuiting the update (exec'ing a shell without the -c $COMMAND) Nov 13 06:07:24 i tried running the update manually and noticed the message about no jshn.sh Nov 13 06:09:23 i find a lot of the sysupgrade mechanics kind of byzantine and hard to follow Nov 13 06:10:14 not the say anything bad about the byzantinians Nov 13 06:21:40 rmilecki: here's what I'd see when it was failing: http://paste.debian.net/1116007/ Nov 13 06:21:58 the "Working" part is after my patch Nov 13 06:22:42 or, with an earlier version (circa october 17) Nov 13 06:22:42 russell--: i see, thanks Nov 13 06:22:58 i'll look at that later today Nov 13 06:23:23 i think it was choking on /usr/share/libubox/jshn.sh being missing the root-pivoted environment Nov 13 08:50:42 Good morning. Anyone have any ideas on this luci rpc problem https://pastebin.com/aYS4qR7A What have I done wrong this time? Nov 13 09:00:27 ldir: you are dir-EDB0 on github right? Nov 13 09:01:36 guilty as charged :-) (ldir was taken - ED B0 are the hex machine code values for the z80 ldir op-code) Nov 13 09:26:16 ldir: all right good to get the links. Sorry for the uuid clash Nov 13 09:29:18 aparcar[m]: ha - it was an interesting day yesterday, a sysupgrade outage of what should be 60 seconds turned into 2+ hours :-) But I learned things about the linux boot process Nov 13 09:30:45 bah, my upload from Intel 8265 on ath10k-ct AP is busted again. 30Mbit. download 300Mbit Nov 13 09:54:30 ldir: I guess the underlying ddns package is too old Nov 13 10:04:15 jow: can you clarify that - bear in mind that I rebuild & flash entire images based on doing git pulls and feed update -a Nov 13 10:33:28 xback: you can add my tested-by or whatever for 4.19.83 on x86 if you like. I also have a 4.19.84 bump in my tree & running x86 too - really straightforward refresh with a minor wrinkle in 902-debloat_proc.patch Nov 13 10:36:03 much easier than .83 - the bcmgetnet-Workaround stuff made no sense to me. I got the impression that upstream had basically fixed it, so dropped both patches Nov 13 10:36:35 but then I don't have hardware or time to test so.... who knows! :-) Nov 13 11:01:10 ldir: ah, then maybe feckert did some mistakes, or prematurely merged the luci side before the packages repo Nov 13 11:01:54 Oh, is there supposed to be a packages bump as well? Nov 13 11:04:31 the file timestamps here looke realllllly weird: https://github.com/openwrt/packages/issues/10536#issuecomment-553304824 Nov 13 11:04:57 board.json and urandom.seed with 2033 timestamps Nov 13 11:07:15 ewww, ./usr/share/rpcd/acl.d/luci-app-ddns.json - DOS end of lines - how did that happen? Nov 13 11:17:01 ldir: thanks. just testing more here currently .. (https://pastebin.com/raw/Sb2Ch5yP) Nov 13 11:17:53 ldir: I share the idea to drop both patches, but will do in another commit Nov 13 11:18:09 first one added the workaround, while the 2nd one just deletes the exact same lines Nov 13 11:18:32 yeah - all a bit strange :-) Nov 13 11:19:39 ewww, kernel panic :-) Nov 13 11:29:24 ldir: yeah, just found out it's 4.14.152 or 4.14.153 causing it Nov 13 11:29:44 ldir: now making a build with 4.14.152 only included Nov 13 11:30:15 I'm seeing this on ar71xx (RB922) using ath10k wave1 radio's Nov 13 11:30:21 19.07 branch Nov 13 11:30:41 seems to trigger when using 4+ boards in IBSS on wave 1 Nov 13 11:41:15 ldir: found the culprit: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v4.14.154&id=a9de6f42e945cdb24b59c7ab7ebad1eba6cb5875 Nov 13 11:43:21 I think pkgadd waved at this one a few days ago and said 'one to watch' Nov 13 11:52:54 hence the extensive testing at my side :-) Nov 13 13:34:50 jow: can you please take a look if it would be possible to somehow share the git.oo setup so we can contribute changes somehow? Nov 13 13:36:01 1. possibility to somehow extend the git pre commit(push?) hooks, 2. s/LEDE/OpenWrt/ in the sent emails and links Nov 13 13:40:23 1.1 prevent DMARC text in commit description 1.2 prevent not portable stuff like `head --bytes/-c` in the commits to *.mk files Nov 13 13:46:12 ynezz: out of curiosity, how to verify 1.2? Nov 13 13:47:45 gch981213: hi, do you have access to mt7629 ref board? Nov 13 13:47:49 hmmmz, don't really recall how the hook works on our setup Nov 13 13:49:28 pepe2k: https://imgflip.com/i/3g7bwt Nov 13 13:49:41 multi-hook-driver Nov 13 13:49:47 whitelist or black list from hell :) Nov 13 13:50:20 honestly, just building it on multi platform is probably easier as long as it breaks fast enough. Nov 13 13:50:23 files are all owned by git so I suspect this is all done via gitolite Nov 13 13:50:30 trying to prevalidate in a commit hook sounds like madness IMO Nov 13 13:51:21 karlp: :D Nov 13 13:54:37 -ENOCI Nov 13 13:55:58 grep changed *.mk files for `head --bytes` or `head -c` additions ^+ Nov 13 14:01:08 yeah, like I said, a whitelist/blakclist that will become ever bigger Nov 13 14:02:06 you'll spend a lot of time on it, when just having builds on a different platform will catch all the unknown things that also fail, just "a bit later" instead of precommit. but hey, whatever you want to spend your time on Nov 13 14:03:32 we're talking about builds on different platforms for some time already Nov 13 14:03:43 great! Nov 13 14:04:04 I've tried FreeBSD on that osusl farm Nov 13 14:04:27 it won't work Nov 13 14:05:15 so the remaining options are Travis and macOS VM or jenkins slave on macOS Nov 13 14:05:45 but this is going to produce unusable SDK on Linux x86/64 and so on. Nov 13 14:06:01 we don't consider inability to build on freebsd a bug that should be fixed? Nov 13 14:06:13 be my guest :) Nov 13 14:06:42 be my guest writing whitelists and grep patterns to "verify" that they will build on other platforms. Nov 13 14:06:50 if we don't care about other platforms, we don't need to bother doing it anyway. Nov 13 14:07:04 if we care about them, making them build seems more productive than grepping for a few things we know we've broken in the past Nov 13 14:08:06 the problem is that it didn't fail during build, but at runtime Nov 13 14:08:16 I guess we should add set -e somewhere :) Nov 13 14:08:34 because producing non-bootable images is not acceptable Nov 13 14:08:42 I much rather have build failure Nov 13 14:09:10 ynezz: git clone git@git.openwrt.org:gitolite-admin Nov 13 14:09:15 ynezz: I made you gitolite amdin Nov 13 14:09:28 thenyou can't ever do VAR=$(shell something) in makefiles. Nov 13 14:09:45 its a more or less standard Gitolite setup so the (quite comprehensive) online manual for it applies Nov 13 14:10:21 I'm using gitolite as well, but it's setup&forget experience Nov 13 14:11:20 stintel: https://zerobin.net/?8e0702905f93838a#EGh8ccHQGXZ46ttZWveCSyrys8L3Wb/bZictGO+PQDU= Nov 13 14:12:01 I think, that you can check for $(shell something) failures in the Make Nov 13 14:12:18 each tabbed line in a Makefile is executed as new shell Nov 13 14:12:34 nvm, wrong context Nov 13 14:14:08 IIRC there is something like .ONESHELL Nov 13 14:14:51 still applies only to the recipes, not the $(shell blah) function call Nov 13 14:15:32 .SHELLSTATUS Nov 13 14:15:34 .SHELLFLAGS = -e Nov 13 14:15:46 seems towork in that little test I pasted though, so that's better than I thought possible Nov 13 14:17:44 eh, no, that breaks differely Nov 13 14:17:55 just passes it in front of everything else. Nov 13 14:44:24 jow: is 34fa5122f which converted luci-app-ddns a decent example of converting cbis to "new fancy" js? or is there a better one? Nov 13 14:45:03 the new docs "Client side JavaScript API documentation" linked from https://github.com/openwrt/luci/wiki/Documentation have lots of text, but not a lot of examples I can see. Nov 13 14:48:29 karlp: https://github.com/openwrt/luci/commit/9c16090780b7430ffe88c770732604f6eb6cfda6 Nov 13 14:51:16 karlp: or maybe this as a less complex example: https://github.com/openwrt/luci/commit/f878b21c8c4ef9444a9de6bc02b58ed4f91b0b2d Nov 13 15:03:43 thanks. have a new project that will only be from 1907+ so thought I'd try and do the new cool thing :) Nov 13 15:04:47 so view() will now find js files as well as htm files automatically then. Nov 13 15:05:57 we're basically just moving lots of the old controller lua stuff into rpcd plugins now right? Nov 13 15:06:12 which at least means it's not blockign the webserver. Nov 13 15:09:30 jow: I may have stumbled across a 'feature' in the new js ssh management. Nov 13 15:10:19 the 'ssh-keys' view is a representation of the content of /etc/dropbear/authorized_keys file yes? Nov 13 15:16:54 ldir: yes Nov 13 15:17:09 karlp: yes Nov 13 15:18:16 ecdsa-sha2-nistp256 AAAAE2VjZHNhLXNoYTItbmlzdHAyNTYAAAAIbmlzdHAyNTYAAABBBG7aanu3ysko1wId9tJXIbjIVLZP1rVaP+kBeNOjkRxDVtn5/IfU56ELvc/xVYQhDRDhyfyqgO314PpDVAvkXzU= ldir@darbyshire-bryant.me.uk Nov 13 15:19:01 is accepted by the interface, and indeed is in authorized_keys - but the ui says 'no public keys present' Nov 13 15:19:41 ldir: interesting. I cannot reproduce it unfortunately Nov 13 15:19:52 which browser is this happening with? Nov 13 15:20:40 safari Nov 13 15:22:23 is it happening with chrome too? Nov 13 15:22:41 I've a OS X catalina VM at home, but not right here with me Nov 13 15:22:47 let me try !safari :-) Nov 13 15:22:54 so can't test safari Nov 13 15:23:54 doesn't work with firefox. Maybe my key is wonky in some way Nov 13 15:24:12 I just pasted it on my router Nov 13 15:24:23 and it worked as expected, thats why I am wonderting Nov 13 15:24:25 jow: have you done any testing of 802.11r with that catalina? Nov 13 15:24:38 DonkeyHotei: since it is a VM, no Nov 13 15:27:13 so basically the only mac user here is ldir? Nov 13 15:28:47 I followed in the great footsteps of nbd who uses macs AFAIK Nov 13 15:29:50 I was shocked the other day to discover that ALL of my wireless devices as manufactured by Apple - I've gone from everything windows to virtually everything mac Nov 13 15:31:16 ldir: the problem at hand as it pertains to you is that without openwrt serving up wifi, you don't get to see interoperability issues Nov 13 15:32:01 jow: is there specific reason to not send commit emails from @project-repos? Nov 13 15:32:11 I mean to lede-commits Nov 13 15:32:24 ynezz: no Nov 13 15:32:53 ohter than the lede name at least :) Nov 13 15:33:20 I don't know what it would mean to rename that list Nov 13 15:33:30 DonkeyHotei: No. But I do still get to play with 'doesn't build correctly on mac os' issues..and that scared the sh*t out of me for 2 hours yesterday 'cos it built but wouldn't boot completely. Nov 13 15:36:42 jow: something specific to that key/key type 'cos my rsa key displays no problem. Nov 13 15:37:21 it's definitely suboptimal if mac interoperability issues have to be investigated without the direct involvement of any mac users Nov 13 15:37:22 ldir: do you see any errors in the javascript console? Also do you copy-paste it our do you drag & drop it? Nov 13 15:39:29 I've tried both copy paste & drag'n'drop - both methods show it when initially adding the key... and it appears in authorized_keys, but if I navigate away to another page and come back, the ecdsa key is gone (but still in auth keys file thank goodness) Nov 13 15:46:26 no obvious error in javascript console - but then I'm pretty hopeless at driving a web browser at best of times Nov 13 15:47:29 ldir: the navigate away part was missing, I can reproduce it Nov 13 15:50:23 yeah, sorry about that, poor description of fault. At least you can repro. Nov 13 15:57:02 ldir: fixed now Nov 13 15:57:59 genius - ldir pulls latest luci Nov 13 17:27:31 * karlp gets a basic cbi.js working, but discovers all his theme is b0rked :) Nov 13 18:10:31 In a target's image/blah.mk, is it correct that the "|" character is simply a separator (looks like it from `split_args`, `build_cmd`, `concat_cmd`, ...) Nov 13 19:45:49 jeffsf: yes Nov 13 19:46:34 well it micks shell pipes somewhat as the macro magic will turn it into a transform command pipeline behind the scenes Nov 13 19:47:33 Thanks @jow I was looking for a way for one part of the command-sequence to "abort" creation of the image. Nov 13 19:48:14 However, it looks like letting it complete, then removing it at the end if the condition fails is going to have to do for now Nov 13 19:48:44 (Use case is that the kernel, for -factory.img, is too large for U-Boot, but is OK for sysupgrade) Nov 13 19:49:25 Ref: https://github.com/openwrt/openwrt/pull/2184#issuecomment-553232401 Nov 13 19:59:54 jow: do you want to send out the Security Advisory or should I do it now? Nov 13 20:00:04 I will just add the last changes which wrere requested Nov 13 20:38:34 ynezz: ping Nov 13 20:42:28 after some discussion with the gitlab team they granted us access to https://gitlab.com/openwrt Nov 13 20:46:46 :+1: Nov 13 20:51:32 aparcar[m]: nice Nov 13 20:58:27 aparcar[m]: ooh, victory! nice! Nov 13 21:07:27 ynezz: btw some builds only take some 20 minutes https://buildmaster.aparcar.org/#/ Nov 13 21:07:27 Ill try to make the buildbots smart to stop rebuilds for targets if only files in other target folders changed and add multi branch builds Nov 13 21:08:01 this way a single master to... suggest build steps to them all Nov 14 02:18:46 https://github.com/vshymanskyy/wasm3 - An (experimental) high performance WebAssembly interpreter written in C. Nov 14 02:58:28 18.06.4 -> .5 was a seamless upgrade for me. thanks devs **** ENDING LOGGING AT Thu Nov 14 02:59:58 2019