**** BEGIN LOGGING AT Tue Mar 10 02:59:57 2020 Mar 10 03:36:20 all glibc mirrors seem to be unreachable (currently) Mar 10 03:37:00 actually they all return 404 Mar 10 03:39:01 https://sourceware.org/git/?p=glibc.git itself however lists http://sourceware.org/git/glibc.git - which doesn't exist.. hm Mar 10 05:25:57 buffalo_wzr-hp-ag300h on 5.4 from a few days ago, no ethernet Mar 10 05:29:18 [ 0.830709] ag71xx: probe of 19000000.eth failed with error -2 Mar 10 05:29:18 [ 0.836602] ag71xx: probe of 1a000000.eth failed with error -2 Mar 10 05:30:06 r12503-e5ffb94dc5 Mar 10 05:39:25 also, ubnt,airrouter, but tplink,tl-wdr3600-v1 is okay on 5.4 Mar 10 06:18:11 /* driver matched but the probe failed */ from drivers/base/dd.c Mar 10 07:41:10 https://bugs.openwrt.org/index.php?do=details&task_id=2889 Mar 10 08:04:20 russell--: it's upstream driver, poke upstream :p Mar 10 08:06:58 aparcar[m]: no, matrix is broken Mar 10 08:12:55 ynezz: okay. Just wanted to ping you a final time to get a command of the last profiles.json patch. It's all done except the merging part where I can't figure out the correct naming of target Mar 10 08:24:29 no its not done Mar 10 08:25:30 * russell-- wonders if Hauke has seen this or has ideas? Mar 10 08:26:33 a lot of people is aware about those issues since it landed in upstream Mar 10 09:08:52 btw, when switching between 4.19 and 5.4, i get build errors in gpio-button-hotplug which resist attempts to correct, including package/gpio-button-hotplug/clean, target/linux/clean. so far the only thing that's been reliable is dirclean. Mar 10 09:09:24 maybe there is something more targetted Mar 10 10:17:32 blogic, morning, you around or still yawning? Mar 10 11:04:22 around Mar 10 11:22:35 https://github.com/openwrt/openwrt/pull/2804 Mar 10 11:22:45 regarding 5.4 on mvebu Mar 10 11:28:52 i think i'll order a ac320 Mar 10 11:30:48 russell--: I've seen that too, on my UAP-LR and UAP-AC-PRO. I too have been trying to get the attention of various people without success so far. Mar 10 11:32:06 russell--: `make clean` solved that for me Mar 10 11:32:43 blogic, leaving wifi aside (which is used only for internet browsing without issues), the 3200acm is a beast of a machine Mar 10 11:33:28 add a usb port and a usb/e-sata port to that and you get a mini-server Mar 10 11:50:18 nitroshift: i have real servers Mar 10 11:50:29 nitroshift: I am just considering to get it to test mvebu Mar 10 13:02:31 Hi. Is there anything special required for netifd to handle an interface that is renamed? Mar 10 13:03:27 I would like to rename an interface after it is connected (an LTE modem) and before I start configuring the device. Renaming works fine, but it seems netifd does not properly detect the interface after rename Mar 10 13:03:53 My steps for reproducing are 1) Restart network, 2) Connect modem, 3) Rename Mar 10 13:05:31 I have put my log here: https://pastebin.com/RuPWYzGB. The intrface is configured with a static address and if I manually do ifup, then I see that the address is configured on the interface but L2 state is not updated Mar 10 13:05:58 Restarting network resolves the issue Mar 10 13:44:30 I see the problem. The hotplug handler is run before the rename is done, which causes the ioctl() call to fail. Adding a call to device_set_present() in cb_rtnl_event makes the interface work reliably Mar 10 13:44:36 I will see about submitting a patch Mar 10 14:05:37 https://github.com/blogic/u80211d Mar 10 14:05:40 voila Mar 10 14:06:14 basically iwinfo on ubus Mar 10 14:06:18 ewwww github :D Mar 10 14:06:23 yes yes Mar 10 14:06:28 i will move it g.o.o Mar 10 14:06:44 i just used it as my staging area while i was doing lots of push -f Mar 10 14:07:08 nice work Mar 10 14:07:23 does background cac and normal scans if desired Mar 10 14:07:36 incase there is a dedicated scanning radio Mar 10 14:08:00 What's "quality": 33 ? Mar 10 14:08:00 considering to add load balancing and band steering support next Mar 10 14:08:11 signal strength Mar 10 14:08:13 33/70 Mar 10 14:08:30 same as in iwinfo Mar 10 14:09:02 its the ap from across the street hence the bad signal ;) Mar 10 14:09:23 nl80211 gives out just dBm, isn't this ratio just confusing? What's 70, where is it from and why it might be needed? Mar 10 14:09:40 as i said, its on par with iwinfo Mar 10 14:09:49 we could add raw dbm aswelli guess Mar 10 14:09:57 there are still lots of options that can be added Mar 10 14:10:04 nl80211 spews out tons of attr Mar 10 14:10:08 *attrs Mar 10 14:10:16 Latest commit Mar 10 14:10:16 33b5da3 Mar 10 14:10:18 on Aug 14, 2017 Mar 10 14:10:26 But not "quality" :) Mar 10 14:10:28 this has been a long time in the making apparently :) Mar 10 14:10:34 nice to finally have it then Mar 10 14:10:35 stintel: yup Mar 10 14:10:44 stintel: derived of some code from 2017 apparently Mar 10 14:10:54 stintel: git rebase -i --root ;) Mar 10 14:10:59 yeah Mar 10 14:11:01 clean history ftw Mar 10 14:11:13 better than 120 "foo" commits Mar 10 14:11:21 +10000000 Mar 10 14:11:36 next up is utopologyd Mar 10 14:11:47 already got the core running Mar 10 14:11:56 Hm, and "signal": 161, <- I was told by johill the signal should be casted to (int8_t) after getting it from netlink u8. Mar 10 14:12:04 what's that supposed to do? Mar 10 14:12:15 stintel: draw a map of the network Mar 10 14:12:22 attached devices, where they are attached Mar 10 14:12:41 PaulFertser: really ? Mar 10 14:12:45 let me fix that just now Mar 10 14:13:05 stintel: basically neigh table + bridge port, dhcp name, ... Mar 10 14:13:10 fancy Mar 10 14:13:22 and then also lldp/mdns/ieee1904.2 .... Mar 10 14:13:24 blogic: that's what "iw" does. Not sure if we have the same attr in mind, but where nl80211 gives signal strength in dBm in u8 it should be casted to i8 on receiving side. Mar 10 14:13:41 ah Mar 10 14:13:47 let me check the code Mar 10 14:19:04 I just had a recent deep dive into the signal strength reporting: https://github.com/jaor/xmobar/pull/429 . ath10k has a fancy way to provide it: "status->signal = snr + ATH10K_DEFAULT_NOISE_FLOOR;" Mar 10 14:19:20 yep Mar 10 14:19:26 ath11k has the same Mar 10 14:28:15 PaulFertser: "signal": -95, Mar 10 14:28:22 now lets look at the scan rssi Mar 10 14:28:42 PaulFertser: quality is the legacy wext quality formula Mar 10 14:30:31 quality = max(min(signal, -40), -110) + 110 Mar 10 14:42:17 "rssi": 65426, Mar 10 14:42:20 strange Mar 10 14:44:31 jow: if it's indeed that formula (used in cfg80211 wext compatibility layer) then I'd say it shouldn't be used anywhere for anything because it's causing silent data loss (when reported signal level is above -40). Mar 10 14:44:53 PaulFertser: its working well enough for my use case Mar 10 14:46:51 I'd prefer to be warned when I get clamped data. Mar 10 14:47:13 for that you can read the raw rssi value Mar 10 14:47:58 And I can't see how it can be ever more useful than the reported signal level itself. This transformation seems arbitrary, I can't see when it might be advantageous. Mar 10 14:48:19 thats fine. I can see it, thats enough. I have no interest in discussing this further Mar 10 14:50:29 If somebody else can see it too and is willing to explain me the usecase, I'd be grateful to learn that. Mar 10 14:50:29 PaulFertser: for a nerd not Mar 10 14:50:29 for my granny 30/70 has a meaning Mar 10 14:50:29 regardless of it being correct Mar 10 14:50:46 remember this is dreamed up by marketing drones Mar 10 14:50:46 didn't it just say "30" though? paul had to ask about the /70 part? Mar 10 14:51:02 karlp: the 70 is in the next line Mar 10 14:54:11 When you're using a set of integers you have total order defined, so comparing 30 with 40 is as easy as -30 with -40 (assuming one looks at the numbers to tell when the signal is stronger). Mar 10 14:55:37 I admit my granny might make a mistake comparing negative integers though. But she's not OpenWrt target audience either, she's never had any clue wrt technology. Mar 10 14:56:56 PaulFertser: luckily you and her are the onl;y people on the planet, lets stop writing software for the rest and go shopping Mar 10 14:56:59 ;) Mar 10 14:57:34 blogic: my point is that OpenWrt target audience are able to compare negative integers just fine. Mar 10 14:57:58 #aspi Mar 10 14:58:02 it is also able to skip data not interesting for them just fine Mar 10 15:04:53 blogic: winning comment there, nice one. Mar 10 15:05:42 * ynezz didn't get it Mar 10 16:47:08 ok, apparently glibc builds are quite neglected - glibc itself can't be built due to 404 and no archives on the mirrors and several packages fail to build on 19.07 (e.g. tcpdump). is it worth fixing/backporting glibc related issues or was there any discussion/decision about it? Mar 10 17:41:17 build #287 of armvirt/32 is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/master/images/builders/armvirt%2F32/builds/287 blamelist: Leon M. George , Koen Vandeputte , Yousong Zhou , Rafa? Mi?ecki , Christoph Krapp Mar 10 17:41:17 , Kevin Darbyshire-Bryant , Petr ?tetiar Mar 10 17:55:07 Hi can any one tell me why I cant find kirkwood in 19.07.2? Mar 10 17:57:25 The target kirkwood. Mar 10 18:04:33 Tapper: build was triggered yesterday, AFAIK it was a mistake it wasn't there in the first place Mar 10 18:04:47 mirko thanks mate Mar 10 18:09:05 Tapper: http://downloads.openwrt.org/releases/19.07.2/targets/kirkwood/ (?) Mar 10 18:09:56 mirko thanks some one was asking me on the openwrt-help twitter feed Mar 10 18:13:35 Tapper: yesterday somebody asked here as well and ynezz triggered the build Mar 10 18:14:31 mirko OK I am glad we could get a quick fix for people. Mar 10 18:21:24 the problem with overwritten art partition is fixed in master now? (and was only 4.14?) Mar 10 18:39:55 mirko: glibc should be cloned from git Mar 10 18:40:08 that worked for me some weeks ago Mar 10 18:54:12 Hauke: well, http://sourceware.org/git/glibc.git -> 404 Mar 10 18:56:38 mirko: ok that's not good Mar 10 18:57:35 ftp://ftp.gnu.org/gnu/glibc/ seems to be present - and git clone git://sourceware.org/git/glibc.git seems to work Mar 10 19:32:43 i'm just surprised why there's no tarballs on the openwrt mirrors Mar 10 19:33:04 probably only tarbarlls for musl builds are uploaded? Mar 10 19:42:13 mirko: yes Mar 10 19:42:19 we don't have any glibc auto builds Mar 10 19:42:30 hence no source uploading either Mar 10 20:10:45 jow: that explains why there's quite a few broken packages as well Mar 10 20:11:45 jow: could we maybe also make runs for glibc builds at least before (major) releases? 19.07 is full of glibc-related issues :/ Mar 10 20:11:49 or we state it's unsupported Mar 10 20:16:24 I think it is de-facto unsupported Mar 10 20:19:05 I'd vote for marking it broken Mar 10 20:23:18 any objection to https://git.openwrt.org/?p=openwrt/staging/ldir.git;a=commit;h=72ebdfcc0b69f92f5fe59e9a880d850bde69da4a ? Mar 10 20:26:53 ldir: +1 Mar 10 20:27:33 So that's one object - LOL ;-) Mar 10 20:28:50 objection - oh dear my brain and fingers don't agree - buffer overruns I think. Mar 10 23:48:43 ldir: could you test my x86 rework patch? Mar 11 02:13:38 mangix: thanks, both fixed until luasec switches back to tagging their releases the old way **** ENDING LOGGING AT Wed Mar 11 03:02:27 2020