**** BEGIN LOGGING AT Wed Dec 11 02:59:58 2019 Dec 11 03:49:42 mangix: grep -E -i -o 'CVE-[[:digit:]]{4}-[[:digit:]]{1,}' package/feeds/packages/libvpx/patches/400-CVE-2019-9232_9325_9371_9433.patch shows all 4, not that 18.06 has a PKG_CPE_ID for libvpx or a local CPE_ID, or that cvechecker isn't busted atm Dec 11 07:14:04 build #202 of omap/generic is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/omap%2Fgeneric/builds/202 Dec 11 07:28:14 nbd: ping Dec 11 07:44:44 should my wzr600dhp still have wifi radios? ... fresh build seems to be missing them :-( Dec 11 07:53:19 /etc/config/wireless is empty, ath9k kernel module is loaded, nothing radio related in dmesg afaict Dec 11 08:01:12 CONFIG_PACKAGE_kmod-owl-loader=y in a 2 week old .config, not in my current .config Dec 11 08:19:07 russell--: check https://patchwork.ozlabs.org/patch/1196298/ Dec 11 08:24:16 pkgadd: yeah, i'd found that commit with vigorous git grep'ing Dec 11 09:02:24 ynezz, https://paste.centos.org/view/b105d944 Dec 11 09:02:28 ynezz, thanks Dec 11 09:10:06 r11614-82a741c774 is busted, trying the one before ... Dec 11 09:18:43 and ... r11613-f1410902e6 is not busted. Dec 11 09:20:41 Hauke: you ack'd 82a741c774dc1aa1e2c2e4b2e445bf79158de1a1 ... any clues what went wrong? Dec 11 09:22:51 maybe just a dependency problem Dec 11 10:26:02 russell--: are you able to reproduce it with official images? Dec 11 10:26:27 to me it seems like it's a config problem 'Depends on: TARGET_PER_DEVICE_ROOTFS [=n] && m && MODULES [=y]' Dec 11 10:26:38 so you perhaps need TARGET_PER_DEVICE_ROOTFS=y ? Dec 11 10:28:24 and I'm wondering if it makes sense to report ar71xx bugs in master :) Dec 11 10:28:47 as long as it exists there, yes Dec 11 10:28:55 ah, it's ath79 anyway Dec 11 10:31:07 it's ath79 Dec 11 10:31:21 reverting the commit fixes it Dec 11 10:31:30 on current HEAD Dec 11 10:32:20 ynezz: i tried stuffing TARGET_PER_DEVICE_ROOTFS=y into my .config, and make defconfig helpfully erased it Dec 11 10:32:44 then you're probably missing other dependency, arent you? Dec 11 10:33:05 i looked, i haven't figured out what it is yet Dec 11 10:33:17 maybe if it's easy, you can point me at it? ;-) Dec 11 10:34:41 you're holding it wrong Dec 11 10:34:54 use menuconfig Dec 11 10:35:18 doesn't perdevice rootfs only work if you are uysing config stubs? selecting the devices manually after doesn't work last I'd hearad Dec 11 10:35:55 karlp: fwiw, i'm starting with a stub Dec 11 10:37:14 per device rootfs also needs the devices to be selected or something, Dec 11 10:37:25 because it's a "different" device selection vs the single profile selection Dec 11 10:37:59 curl http://downloads.openwrt.org/snapshots/targets/ath79/generic/config.buildinfo > .config; make defconfig; grep -i owl .config Dec 11 10:38:09 CONFIG_MODULE_DEFAULT_kmod-owl-loader=y Dec 11 10:38:22 yeah, that's in there but not the PACKAGE_ Dec 11 10:40:20 in make menuconfig: Symbol: PACKAGE_kmod-owl-loader [=PACKAGE_kmod-owl-loader] Dec 11 10:41:41 if i revert the commit, and make defconfig on my stub i get this: Dec 11 10:41:42 CONFIG_DEFAULT_kmod-owl-loader=y Dec 11 10:41:43 CONFIG_PACKAGE_kmod-owl-loader=y Dec 11 10:42:14 without the revert, i get just the CONFIG_DEFAULT Dec 11 10:42:41 (from grep owl-loader .config) Dec 11 10:46:55 maybe it worked differently previously Dec 11 10:47:37 http://downloads.openwrt.org/snapshots/targets/ath79/generic/openwrt-ath79-generic-buffalo_wzr-hp-ag300h-squashfs-sysupgrade.bin contains ath9k_pci_owl_loader.ko and Dec 11 11:00:48 looking at the commit, i wonder what the KCONFIG:=CONFIG_OWL_LOADER does Dec 11 11:07:38 probably selects kernel config option ? Dec 11 11:29:52 jow: pong Dec 11 11:54:39 nbd: sorry, forgot to unping. I've already answered my question Dec 11 12:17:09 jow: thanks for the comment, updated version https://patchwork.ozlabs.org/patch/1207614/ Dec 11 12:23:57 jow: can you explain why the buildbots use a shorter feed commithash than debian 10? did git change the minimal length in more recent versions? https://aparcar.gitlab.io/-/rebuild/-/jobs/375951940/artifacts/output/19.07-SNAPSHOT/ath79/generic/feeds.buildinfo.html Dec 11 12:39:03 either its a git change or it has something to do with shallow history cloning Dec 11 12:45:53 dejavu? I remember looking it up Dec 11 13:10:33 jow: I think the buildbots use full aka no shallow copy Dec 11 13:10:41 ynezz: you also remember your previous answer? Dec 11 13:12:54 I think jow found it, somewhere in the perl script, you need to force the same hash format or such Dec 11 13:40:09 in case there was any suspicion about the state of my build tree, i did a dirclean, problem of missing owl-loader persists. Dec 11 13:51:51 ath9k_pci_owl_loader.ko ? Dec 11 13:52:18 I don't follow this stuff, nor I'm willing to look it up, so you need to be explicit Dec 11 13:53:00 and as I hinted before, try to reproduce the problem with prebuilt images Dec 11 13:55:55 ynezz: for giggles, i'm installing http://downloads.openwrt.org/snapshots/targets/ath79/generic/openwrt-ath79-generic-buffalo_wzr-hp-ag300h-squashfs-sysupgrade.bin Dec 11 13:56:39 well, that image has ath9k_pci_owl_loader.ko Dec 11 13:56:51 (if thats the missing module you need for working wifi) Dec 11 13:59:31 indeed the snapshot does. that doesn't change the fact that the commit broke the build for a perfectly reasonable .config stub Dec 11 14:01:00 maybe someone with mips16/32 insights can check this plz: https://github.com/openwrt/packages/pull/10751 Dec 11 14:04:20 russell--: you're still arguing about config changes, have you tried to build image with that .config stub? Dec 11 14:06:10 ok, now I see difference, it's CONFIG_DEFAULT_kmod-owl-loader=y in your config and CONFIG_MODULE_DEFAULT_kmod-owl-loader=y in the buildbot config Dec 11 16:30:01 ynezz: We do not happen to have actual versioning beyond git revions for libubox, do we? Dec 11 16:31:22 nope Dec 11 16:33:17 Ok, can you finalize the !WIP commit then? I'll have to include a reference to the fixed version Dec 11 16:34:13 Or are you still hunting for more bugs? Dec 11 16:36:22 Also we should probably add some kind of stable versioning in the future. That will make the tracking bugs a lot easier Dec 11 16:37:50 you're too fast for me :) Dec 11 16:39:32 Just an idea. We are not really in a hurry though Dec 11 16:41:33 * tsys will be afk for a while Dec 11 17:35:58 russell--: https://gitlab.com/ynezz/openwrt/commit/6dd22e5f6d745f205bffce016fa213fd62e52334 Dec 11 17:50:51 ynezz: if you touch libubox anyway, can you give it an SOVERSION so that we have a proper aby version? Dec 11 17:51:11 *abi Dec 11 18:35:07 Is it possible to for me to separate SSID's on two different networks (br-lan, br-lan2) with two different subnets (192.168.1.0/24, 192.168.2.0/24) but still have stuff like chromecast, sonos, etc show up no matter which network you join? Dec 11 18:37:03 Maybe with a proxy arp service? Dec 11 18:38:37 rubberduck: Maybe, sounds complicated. Ill look into it. Dec 11 18:39:45 jow, stintel: remember a few weeks ago with my weird iptables rules to have different lookup tables depending if clients were connected to wlan0/wlan1 ? This reason with chromecast identification is a good reason why I preferred not to split the networks up Dec 11 20:08:04 ynezz: i'll test that here in a bit, thanks. i must be misunderstanding what @PCI_SUPPORT because my config has CONFIG_PCI_SUPPORT=y Dec 11 20:08:43 well the fix is 'rm -fr tmp' Dec 11 20:10:58 its probably because the package wasn't renamed and just moved between the parts of the buildroot Dec 11 20:12:09 package/kernel/linux/modules/wireless.mk -> package/kernel/mac80211/ath.mk Dec 11 20:28:54 I missed those toolchain-mips_24kc_gcc-8.3.0_musl/gcc-8.3.0/gcc/dwarf2out.c:15406:1: internal compiler error: Segmentation fault Dec 11 20:30:50 hope it was just some cosmic ray Dec 11 20:45:17 barhom: what's the point of separate networks if you want it all on the same network? Dec 11 20:46:13 karlp: I do want it in the same network if I can manage to do my objectives. We had a lively conversation a while ago where people told me that it was easier to split the networks to achieve what I wanted to do. Dec 11 20:46:43 I couldn't say back then why I specifically wanted it to be the same network, now I can, for discovery of chromecasts and whatnot Dec 11 20:46:57 iirc, it seemed like complicated twisted goals too, and _those_ goals were best served by different networks, but whether those goals are reasonable has never been clarified :) Dec 11 20:46:59 barhom: what was the reason again why iptables physdev didn't work? Dec 11 20:47:22 jow: It did work actually. Dec 11 20:47:57 So wlan0 clients uses the "main" routing table while "wlan1" clients uses another routing table. It works perfectly. Dec 11 20:48:38 right, did I ask you to try enabling sysctl net.bridge.bridge-nf-call-iptables=1 ? Dec 11 20:48:53 jow: thats what you forgot I think or I missed it if you told me to Dec 11 20:48:59 ah okay, so why did you split nets when it works with physdev Dec 11 20:49:41 I did not yet, its still same network. My next issue is that I want to disable ipv6 on wlan0 but not on wlan1. Dec 11 20:49:57 The reason is because my ipv6-pd subnet comes from VPN (not from my ISP) Dec 11 20:50:16 And I want "wlan0" to be "VPN" (with v4, and v6) while wlan1 be without VPN. Dec 11 20:51:00 the physdev did solve that problem for me for v4 connections. But since I'm getting v6 assignments from the router on both wlan0 and wlan1 then both wlan's are "VPN:d" Dec 11 20:51:07 Hi all, I'm trying to figure out how to compile an oWRT sysupgrade-image for the O2 6431 router that allows the kernel to get it's bootargs/CDMLINE from the uBoot environment (if it exists). Reason is the Router has two CPU cores but the normal images reserve one for the telephony-stuff and ignores my modified bootargs. Dec 11 20:51:44 barhom: did you try a simple ip6tables -I FORWARD -m physdev --physdev-in wlan0 -j REJECT ? Dec 11 20:52:08 jow: That would work but the clients would still get the ipv6 assignments. Dec 11 20:52:14 oh and my ~three posts/topics in the forum didn't yield any usefull results so far (only some more users with similar wishes) Dec 11 20:52:29 stintel recommended to maybe block RA packets from the router, haven't tried that yet. Dec 11 20:52:37 barhom: ip6tables -I OUTPUT -m physdev --physdev-out wlan0 -j DROP Dec 11 20:53:00 jow: would that also drop any RA announcements to wlan0? Dec 11 20:53:29 not entirely sure, try it out Dec 11 20:53:33 I will, thanks Dec 11 20:53:49 it should unless odhcpd uses raw sockets Dec 11 21:18:12 I've already tried building an image with "Built-in command line overrides firmware arguments" unset but the uboot bootargs are still ignored - https://forum.openwrt.org/t/proposal-do-not-ignore-uboot-env-args-if-e-g-ubenvar-respect-ubenv-yes-is-set/28821/5?u=limer Dec 11 21:32:25 ynezz: doesn't make dirclean blow away tmp ? Dec 11 21:33:36 it should, yes Dec 11 21:37:30 your patch indeed fixed it for me, i applied it, built, and flashed, and getting ath9k_pci_owl_loader messages on boot and wifi works again. Dec 11 21:42:10 ok, happy it works for you, but now you need to convince Christian, that it's probably really still some issue Dec 11 21:42:14 :) Dec 11 21:44:20 what does the @ in @PCI_SUPPORT mean? Dec 11 21:47:25 my understanding is that @ depends on selection Dec 11 21:48:35 "If you say @SYMBOL that means that CONFIG_SYMBOL must be defined by OpenWrt in order for the package to be available for selection." Dec 11 21:50:38 * russell-- scratches head in confusion ... since CONFIG_PCI_SUPPORT was defined before, so the ||'s seem like they should be redundant. Dec 11 21:53:13 ldir: what happened with that mips FPU emulation patch? Dec 11 22:01:55 anyone know how i can add a package depends for a config symbol and 64bit targets? Like +(MY_CONFIG&&TARGET_x86_64):deps ?? Dec 11 22:02:36 you can't use the + auto selection with boolean deps I think, only the "depends" Dec 11 22:03:02 ok Dec 11 22:03:27 (I think) Dec 11 22:04:11 erm something like +OPENSSL_WITH_EC works Dec 11 22:04:23 sorry, +@OPENSSL_WITH_EC Dec 11 22:04:42 it messes up the buildbot defaults but it works Dec 11 22:05:14 mhh i'm looking for a way to only add a package if a optional config symbol is selected and its a 64bit target Dec 11 22:06:23 does +@(MY_CONFIG&&TARGET_x86_64) work? Dec 11 22:07:25 need to try Dec 11 22:09:56 nope Dec 11 22:13:34 andy2244: try without the +; @(MY_CONFIG&&TARGET_x86_64) should work Dec 11 22:15:43 andy2244: if you want a conditional select, you can use +TARGET_x86_64:MY_CONFIG Dec 11 22:16:22 @(MY_CONFIG&&TARGET_x86_64) this hides my config option until i select the depends = not what i want Dec 11 22:17:41 the concrete example is this: Samba4 ad-dc can use the new lmdb, but only on 64bit targets Dec 11 22:19:21 if thats not possible, i will just disable the lmdb not a biggy Dec 11 22:20:01 don't want to create a variant just to support this one option Dec 11 22:20:30 andy2244: you may have to give up the idea of having altenraive deps _and_ being visible Dec 11 22:20:33 kconfig just doesn't do that. Dec 11 22:20:54 (it should, sure.... but it doesn't have any "greyed out" equivalent) Dec 11 22:21:31 thats fine, was just asking in case i missed a syntax Dec 11 22:33:40 mhh was/is (perl5/5.28.1) bugged? my build keeps failing at this package for target Dec 11 22:34:29 also seeing some weird long output: Dec 11 22:34:32 * andy2244 ...//../lib/perl5/5.28.1/perl5/5.28.1/perl5/5.28.1/perl5/5.28.1/perl5/5.28.1/perl5/5.28.1/perl5/5.28.1/perl5/5.28.1/perl5/5.28.1/TAP/Object.pm Dec 11 22:36:14 andy2244: you can put the depends also just on whereever you defined the MY_CONFIG (i.e. "depends on TARGET_x86_64") Dec 11 22:37:12 but yeah, there is no visible but not selectable option Dec 11 22:37:44 yeah ok, not sure if i want to make this a extra config option just for x64 Dec 12 01:03:07 whats the secret to taking good pictures of boards Dec 12 01:17:30 Slimey: tripod and good, slightly diffuse lighting - in the past I've made very good experiences with (non-CCD) flatbed scanners (if you do it regularly, you just might want to improve lighting, to reduce drop shadows and paralaxes) Dec 12 02:36:54 pkgadd thanks ill give it a shot Dec 12 02:38:03 an older/ cheap compact camera might lead to better results than a smartphone camera as well (not even talking about better equipment) Dec 12 02:52:33 yeah i realized a dedicated camera is better **** ENDING LOGGING AT Thu Dec 12 02:59:58 2019