**** BEGIN LOGGING AT Thu Apr 02 02:59:57 2020 **** BEGIN LOGGING AT Thu Apr 02 09:00:55 2020 Apr 02 09:11:00 FATAL: parse error in symbol dump file Apr 02 09:11:02 (...) Apr 02 09:11:03 make[3]: *** [Makefile:49: /home/rmilecki/openwrt/openwrt-master-bcm53xx/build_dir/target-arm_cortex-a9_musl_eabi/linux-bcm53xx_generic/gpio-button-hotplug/.built] Error 2 Apr 02 09:11:07 what what the quick fix for that? Apr 02 09:11:18 (not dirclean) Apr 02 09:12:50 rmilecki: i had to do a make distclean to get rid of that recently Apr 02 09:13:00 nevermind :P Apr 02 09:13:29 uh, I don't like compiling toolchain :/ Apr 02 09:14:03 neither do I and i have 8C/16T :P Apr 02 09:16:34 distclean sounds a bit extreme for that Apr 02 09:24:56 (it might have been dirclean) Apr 02 09:25:16 i actually did a rm -rf staging_dir/ build_dir/ i think Apr 02 09:34:30 rmilecki: yeah i get that when i switch between the normal and testing kernels Apr 02 09:36:56 this was enough: rm -fR staging_dir/target-* build_dir/target-* Apr 02 09:37:22 rmilecki: :) Apr 02 09:37:43 russell--: good point. i think i accidentally switched between 4.19 and a 5.4 PR as well. Apr 02 09:39:14 yes, I get that after every 4.19 <-> 5.4 switch Apr 02 11:49:58 May I ask someone to test https://git.openwrt.org/?p=openwrt/staging/ldir.git;a=commit;h=0ff5be10e2c103744a89a72983ce3fc8bbf51d83 please Apr 02 11:50:33 prove that it still works on linux Apr 02 12:14:41 jow: the packages for tegra and mvebu cortexa9 changed path from arm_cortex-a9_vfpv3 to arm_cortex-a9_vfpv3-d16, and I don't see that on download server (when browsing), do You need to manually adjust server for that path to appear? Apr 02 12:26:15 ldir: it works here, please add proper credit for his work Apr 02 12:26:34 ldir: Suggested-by: 李国 Apr 02 12:26:53 ldir: feel free to add my ack/reviewed-by Apr 02 12:27:23 thanks for taking care of that! Apr 02 12:38:28 tmn505: yes Apr 02 12:38:46 tmn505: does this also mean that all existing installations won't be able to fetch packages anymore? Apr 02 12:40:13 * karlp laughs Apr 02 12:40:27 ynezz: sure, working on it :-) Indeed the filtering of no-plt was the missing bit because I'd tried various combinations of pie/nopie etc etc and got as far as changing the error to the one that could be solved by no-plt :-) Apr 02 12:41:43 jow: I think yes. Apr 02 12:42:38 phew... crazy arm stuff Apr 02 12:42:49 yep Apr 02 12:43:16 I wonder when we reach the point of one arm variant per router model Apr 02 12:43:20 to hell with build resources Apr 02 12:43:24 whats a few GB more Apr 02 12:43:43 * jow will kick the builders Apr 02 12:43:46 its just replace Apr 02 12:43:58 no new cpu subtype Apr 02 12:44:08 ynezz: and months of bug reports Apr 02 12:44:35 plus an incompatibility of <= 19.07 to later Apr 02 12:45:36 jow: on sysupgrade one replaces the image anyway with new userspace. Apr 02 12:46:32 jow: http://lists.infradead.org/pipermail/openwrt-devel/2020-March/022510.html Apr 02 12:46:33 the old path could be kept for now for older images Apr 02 12:47:22 I am just playing the grumpy guy on the sidelines. I don't actually care about that arm madness Apr 02 12:47:38 give me a minute to restart the builders Apr 02 12:50:26 packages should appear within the next three days Apr 02 12:55:30 thanks Apr 02 12:59:21 is there any libubus api documentation? Apr 02 13:01:00 https://openwrt.org/docs/techref/ubus is mostly cli Apr 02 13:01:44 I get this issue with libgd-2.3.0 error: openwrt/build_dir/target-mipsel_24kc_musl/libgd-default/libgd-2.3.0/config/getlib.sh: No such file or directory Apr 02 13:02:57 tested on last git commit "af5ccfbac7 openssl: bump to 1.1.1f" Apr 02 13:04:28 last part buidl output: https://paste.debian.net/1137991/ Apr 02 13:08:11 russell--: unfortunately not. Do you have any specific use case in mind? Apr 02 13:21:03 someone suggested i try using umdns instead of avahi, i'm investigating what it would take to announce a service via ubus from within the application or something Apr 02 13:22:11 right now the application is writing a service file Apr 02 13:22:23 which avahi-daemon consumes Apr 02 13:24:25 well, I meant umdns via the procd init files, Apr 02 13:24:54 mhei was suggesting you could keep avahi and just use it's library, which is _may_ be better documented that ubus<->umdns Apr 02 13:25:49 i don't know if i know the port number before the application runs Apr 02 13:26:25 when it restarts, the port might be stuck, so we are taking the port number offered Apr 02 13:29:39 karlp: btw, i appreciated your suggestion of umdns, since i think i was unaware of its existance previously, whether i end up using it or not. Apr 02 13:33:54 * russell-- is no great fan of avahi, fwiw ;-) Apr 02 13:38:03 ynezz: lol Apr 02 13:38:12 russell--: umdnsd Apr 02 13:38:50 russell--: basically umdnsd will pick up mdns announcements from service dexcriptions Apr 02 13:39:16 russell--: if you check e.g. `ubus call service list` you should see one such entry for the dropbear service Apr 02 13:39:48 the only way to update that at runtime is to issue an `ubus call service set '{ ... }'` call which updates the service state Apr 02 13:40:25 you need to send the entire service description including cmdline, respawn settings etc. plus a data: { ... } member containing the mdns announcmenet Apr 02 13:40:40 iirc there is no simple api to just update the `data: { ... }` node at runtime Apr 02 13:40:52 umdnsd should then pick it up and start announcing it Apr 02 13:41:03 jow: we could add that easily Apr 02 13:41:08 usually that announcement is setup once at service start by the init script Apr 02 13:41:16 correct Apr 02 13:41:28 or via a raw txt file in /etc/umdns.d/ Apr 02 13:41:33 *raw json Apr 02 13:41:55 blogic: hello!! Apr 02 13:42:05 mrkiko: bonjour Apr 02 13:42:30 blogic: I know I am asking for something that's little bit of a stretch but... do you have still some WL-330n3G around? This device is still supported but can't boot with master, it boots with 19.07 Apr 02 13:42:39 blogic: :) :) Apr 02 13:42:50 what is wl-330n3g Apr 02 13:42:52 rt305x ? Apr 02 13:50:41 yes Apr 02 13:50:56 single port Apr 02 13:51:34 19.07 boots fine when in failsafe mode, something breaks when I disable the wan interface, that BTW doesn't exist on the device Apr 02 13:51:51 well, the switch presents itself to the kernel like having 4 ports or something, but has one physically exposed Apr 02 13:58:17 mornin' Apr 02 14:06:57 i'm getting make world build failures in libssh2 which are cured with a make package/libssh2/{clean,compile} Apr 02 14:07:06 on ath79 in this case Apr 02 14:16:01 Hi are builds for mevbu broken? Apr 02 14:16:40 mvebu* Apr 02 14:19:29 * russell-- looking for differences between 4.19 and 5.4 builds, found this unexpected difference in the way net-snmp-config.h is constructed: http://sprunge.us/lv2jz2 Apr 02 14:24:45 russell--: is that just changes in defaults in net-snmp itself, or is that breaking things? Apr 02 14:25:41 i'm looking for why we see the gpio-button-hotplug build failure when switching kernels Apr 02 14:26:11 ... since the known fix is probably overly aggressive Apr 02 14:26:55 i can't see how the net-snmp difference would be relevant Apr 02 14:29:08 Tapper: are they? http://buildbot.openwrt.org/master/images/builders/mvebu%2Fcortexa9 says it's building. Apr 02 14:39:56 tmn505 I am building at home mate Apr 02 14:40:18 tmn505 I did a make dirclean I will let you know how i get on. Apr 02 14:40:55 Yeah build just messed up again. Apr 02 14:41:33 mrkiko: I will have quick look whether it's a config issue do to the device rename in ramips. Since ramips is still on kernel 4.14 as in 19.07 release, other reasons for your observation should be limited Apr 02 14:45:18 https://www.theregister.co.uk/2020/04/02/cisco_rations_staff_vpn/ Apr 02 14:48:02 mrkiko: looks correct to me Apr 02 14:58:17 backlog, reading now Apr 02 14:59:50 adrianschmutzler: oh, unfortunately I have no easy way to tell due to the fact I have no serial access, but after all the device behaves corrctly, because it threats lan as lan... Apr 02 14:59:57 with master device never comes up Apr 02 15:00:43 adrianschmutzler: thank you very much for looking. Well, all i can if you want is to give you ssh access to a device Apr 02 15:04:02 mrkiko: I cannot help you much with that; I just wanted to sort out the easy and obvious Apr 02 15:19:47 adrianschmutzler: broadcom-wl is the proprietary broadcom wifi driver Apr 02 15:19:56 an alternative to b43 and b43legacy Apr 02 15:20:35 jow: that was quick Apr 02 15:22:46 unfortunately b43 performance has always been atricious and essential features like multi ssid support are missing Apr 02 15:22:57 while wl is stable but stuck in the dark ages feature wise Apr 02 15:23:13 I see Apr 02 15:23:16 it is supporting some limited pre-11n features on very few selected devices Apr 02 15:23:47 not sure how many people actually use wl in production though Apr 02 15:24:01 last time I tested in on a e3000 it hardly worked properly Apr 02 15:24:26 maybe its time to nuke it from orbit Apr 02 15:24:31 but it seems to be selected by default, since I just did a stupid build test from scratch for that Apr 02 15:25:04 actually, I wanted to send a patch for bumping that target to 4.19 after five months, but this doesn't look like it Apr 02 15:25:18 hm, some brcm63xx stakeholder should weigh in Apr 02 15:25:40 maybe there is some response on my mail Apr 02 15:26:04 I wonder how noltari could runtest it, since I can't even build Apr 02 15:26:06 jow: only way to be sure Apr 02 15:27:07 iirc there's multiple profiles for 63xx Apr 02 15:27:13 tmn505 It's umdns that is causing my builds to break Apr 02 15:27:13 with and without wl Apr 02 15:27:32 I just started grepping :-) Apr 02 15:27:40 I tryed to build with the new package luci-app-dawn to test it out. Apr 02 15:28:05 adrianschmutzler as jow said I tested it on a Comtrend AR-5387un, which relies on brcmsmac, not on wl Apr 02 15:28:15 after removeing luci-app-dawn, dawn and umdns build builds again Apr 02 15:28:33 Noltari: can you do something about the #lede channel? you still own it Apr 02 15:28:58 DonkeyHotei how did that happen? xD Apr 02 15:29:08 last man standing? Apr 02 15:29:19 jow: Ah, I see it now. It's just selected for three individual devices in the target Apr 02 15:29:22 what do you want me to do? Apr 02 15:30:00 Noltari: maybe https://epir.at/2016/06/26/redirecting-freenode-channels/ if you have sufficient permissions? Apr 02 15:30:23 OpenWrt has not ever worked on the e3000 proper! Apr 02 15:30:53 Noltari: also reassign ownership to people who are actually around Apr 02 15:31:03 I tryed it out about 4 years ago and it was aweful Apr 02 15:31:36 Tapper: we have an adjective for that. Its spelled Broadcom :P Apr 02 15:31:52 * Tapper grins Apr 02 15:33:28 As in o no that router is a bit Broadcom! Apr 02 15:33:34 I like it. Apr 02 15:34:05 the hardware may be nice but the lack of suitable drivers (at least for those old units) makes it completely unsuitable for OpenWrt Apr 02 15:34:52 Broadcom Profane word to refer to feces Broadcom is a word generally considered to be vulgar and profane in Modern English... Apr 02 15:35:56 Should I make a report about umdns? Apr 02 15:39:36 jow I'm trying but I think I need OP in both channels, right? Apr 02 15:50:03 but, regarding Broadcom DSL ... is someone trying to recerse engineer that? It's so much diffuse hardware Apr 02 15:50:27 yesterday, or the day before, I was having a look at the bcm dsl drivers source I found around Apr 02 15:51:03 I mean, if we could work out that hw, a lot of hardware could open up to interesting uses Apr 02 15:51:12 Noltari: hm, don't know - but might be the case, yes Apr 02 15:52:49 jow: there's been a network hiccup on slashdirt-02, which has been resolved 4h ago but it seems doesn't seem to refresh its network config (the machine is alive I get a prompt at virsh console): should I reboot it? Apr 02 15:58:22 Noltari: maybe just make wigyori an owner of #lede then? Apr 02 16:00:08 http://sprunge.us/YtWaE1 Apr 02 16:05:39 i only see it going from 4.19 to 5.4. Apr 02 16:11:26 adrianschmutzler: fyi it still seems there's some breakage with partial erasesize/4K_SECTORS_LIMIT in 4.19. Switching ramips will break the mikrotik devices Apr 02 16:16:13 f00b4r0: yeah, just reboot it I'd say Apr 02 16:16:47 jow: ok. Could you maybe check afterwards if whatever dhcp client never gives up? :) Apr 02 16:19:11 jow: clean rebooted (reacted correctly to send-key) Apr 02 16:22:11 the DHCP client is wahtever debian uses by default Apr 02 16:22:25 I never had to change anything there to make it persistently do ... things Apr 02 16:22:25 dhclient i guess Apr 02 16:22:33 maybe it died due to oom / kernel foo? Apr 02 16:22:36 i don't remember how it behaves tbh Apr 02 16:22:54 no the network died overnight and i guess it just stopped trying to renew the dhcp lease Apr 02 16:23:06 hm interesting Apr 02 16:24:50 and interestingly it doesn't seem to try to pick up after boot either Apr 02 16:24:57 * f00b4r0 scratches head Apr 02 16:27:36 jow: just to be sure, you didn't upgrade the vm recently? ;) Apr 02 16:27:42 f00b4r0: no Apr 02 16:27:45 ok Apr 02 16:38:14 jow: I rebooted the metal, now everything is back ;) Apr 02 16:39:04 when I get the time and the tooling to get a cable through a 90cm-thick wall there should no longer be these problems ;P Apr 02 17:06:24 I have a house with 70cm walls. You need a 100cm drill for that and much time as the long drill eats much of the power. Apr 02 17:09:44 rubberduck: indeed :) Apr 02 17:11:33 No normal machine. You should use at least sds plus or better sds max Apr 02 17:13:15 *nod* Apr 02 17:13:22 >5J ideally Apr 02 17:14:33 I have a machine to kill biggest stones with 60J Apr 02 17:14:47 wow. Handheld? Apr 02 17:15:28 2 Hands and you should look like Arnold Schwarzenegger in his best Years... Apr 02 17:15:40 lol Apr 02 17:15:54 But not drilling. Only a Big Big Hammer with a Motor Apr 02 17:16:00 i see Apr 02 17:16:27 How ist this called the hammer for digging roads with pressed air? Apr 02 17:16:42 Presslufthammer in German Apr 02 17:16:53 pneumatic hammer iirc Apr 02 17:16:57 Such a thing but no air but electric power Apr 02 18:32:20 * ldir is satisfied - it looks like a successful day re commits Apr 02 19:26:03 does the build system create a single rootfs which is then used for all created profiles of the target or does it create individual files? in other words, can I add a profile specific file in the rootfs during build time? Apr 02 19:47:57 ire Apr 02 20:44:52 trying to port a new device that only has eMMC, and /proc/mtd is empty although /proc/partitions show mmcblk*. How can I define/recognize the partition layout with MTD? (needed to fetch caldata from ART partition) Apr 02 20:49:16 did github just die? Apr 02 20:50:37 yes Apr 02 20:51:08 for almost half an hour already Apr 02 20:51:20 they shut it down to free up slots in azure ? :P Apr 02 20:51:23 it was working for me just moments ago Apr 02 20:51:47 maybe they're just having a bad patch Tuesday. On a Thursday ;) Apr 02 21:02:14 time to switch to our own inrastructure... Apr 02 21:06:02 while that's always a good idea, uptime and reliability is not the part you can beat github with (and it's up again) Apr 02 21:06:30 :) Apr 02 21:10:51 is block2mtd obsolete now? **** ENDING LOGGING AT Fri Apr 03 02:59:56 2020