**** BEGIN LOGGING AT Tue Nov 19 02:59:58 2019 Nov 19 09:15:35 any idea why kernel-debug.tar.bz2 is missing in the release builds? Nov 19 09:18:17 I'm wondering why ipq806x has such weird kernel stack traces https://bugs.openwrt.org/index.php?do=details&task_id=2616 Nov 19 09:35:59 ynezz: does you experienced eye ever seen this very unverbose error before? https://builder.prpl.dev/#/builders/65/builds/1/steps/42/logs/stdio Nov 19 09:37:09 which error? pls point me at the err line Nov 19 09:37:49 ynezz: thats mostly the problem. I don't see why make tools would fail Nov 19 09:38:09 but it does on the bots Nov 19 09:38:24 is it reproducible? Nov 19 09:38:47 i suspect -j24 is the culprit Nov 19 09:38:49 Not locally on my device Nov 19 09:39:00 mhh interesting Nov 19 09:39:09 how could that be a problem? Nov 19 09:39:21 some unhandled race/dependency condition Nov 19 09:39:45 I think it used to work on master Nov 19 09:39:55 but I'll slow it down to test it thank you! Nov 19 09:40:07 so this is 19.07 branch? Nov 19 09:40:35 this is the prpl branch which is basically openwrt-19.07 Nov 19 09:40:51 I've had frequent failures with -j12 that went away after running make -j12 again in many cases Nov 19 09:41:36 sometimes is enough to build with odd number of cores :p Nov 19 09:41:52 wait what :S Nov 19 09:42:55 make[1]: *** [tools/Makefile:159: tools/qsdksquashfskit4/compile] Error 2 Nov 19 09:43:05 I hate to lookup errors in -j24 output Nov 19 09:43:25 gzip_wrapper.c:27:10: fatal error: zlib.h: No such file or directory Nov 19 09:43:27 but turn on build logs and grep for *** in the log dir is my solution.... Nov 19 09:43:36 BUILD_LOG=1 ;) Nov 19 09:43:36 though build logs actually make the build noticeably slower :( Nov 19 09:44:09 karlp: well if you take into account the time lost searching for the actual error ... ;) Nov 19 09:47:50 ynezz: thank you! Nov 19 09:48:17 so there is no guarantee that zlib is compiled before other build tools? Nov 19 09:48:34 or actually a bad setup of the qsdk thing? Nov 19 09:57:28 or a slow disk write? Nov 19 09:57:50 stintel: oh, for sure, I always have build logs turned on :) Nov 19 09:58:03 How would I go about not distributing IPv6 addresses to my clients connected on wlan0 only? wlan1 should still get ipv6 addresses Nov 19 09:58:20 if the docs are correct, set ignore=1 on the interface.... Nov 19 09:58:22 ynezz: it's this line https://git.prpl.dev/prplwrt/prplwrt/blob/c22db67018adcf51cf4f772177f3f5d026197924/tools/Makefile#L48 missing Nov 19 09:58:23 but.... Nov 19 09:58:57 aparcar[m]: ah, qsdk :) Nov 19 09:58:58 thank you very much for checking that out, I didn't think of the 24 cores Nov 19 09:59:21 https://github.com/openwrt/odhcpd/issues/60 Nov 19 09:59:23 barhom: with that and your other routing problem, I would really advise you to make them into two different networks Nov 19 09:59:35 zorun: My other problem is now fixed Nov 19 10:00:04 zorun: news on the survey? Nov 19 10:00:32 aparcar[m]: it's taking longer than expected :( Nov 19 10:00:44 I hope that it will be ready next week Nov 19 10:01:55 why would you discriminate clients by not giving them IPv6 Nov 19 10:02:23 it's bad enough that ISPs still do it when it's almost 2020 Nov 19 10:02:25 stintel: Because I have split wlan0 to be WAN routing while wlan1 is VPN routing which works Nov 19 10:02:39 So basically my 2g is normal ISP internet while 5g is my vpn Nov 19 10:02:49 and my vpn provides ipv6 Nov 19 10:02:53 that sounds ... random Nov 19 10:03:01 I want it that way Nov 19 10:03:16 anyway, the issue now is when Im on 2g I still get v6 which is through the vpn Nov 19 10:04:24 stintel: Why is it random wanting to use different routing tables for wlan0 clients vs wlan1? It works Nov 19 10:04:54 It might not be possivle to *not* hand out ipv6 from the pd-subnet to wlan0 clients specifically, but I am here to find out if it is or not Nov 19 10:05:20 zorun: everyone is hyped, don't let it lose the paste ;) Nov 19 10:05:40 *pace Nov 19 10:06:12 barhom: the clean way would be to use different SSID, and put each of them on a different L2 network Nov 19 10:06:33 aparcar[m]: I know :) but if the students mess up the survey, we won't be able to use the results and that would be a shame Nov 19 10:06:37 e.g. if you now have a device that only supports 2.4G it can not use your VPN Nov 19 10:07:14 stintel: I am aiming at splitting them to different L2 networks but I want to avoid that as much as possible until I hit something I really cant do Nov 19 10:07:31 why do you want to avoid that? Nov 19 10:07:33 stintel: next step is try to blast out 2x 2G SSID and 2x 5G SSID and see if I can route diferently Nov 19 10:08:05 Adds complexity with different subnets in my scripts Nov 19 10:08:37 well, what you are trying to achieve now is different complexity on its own Nov 19 10:09:02 and while it is most likely perfectly doable... once something goes wrong, I imagine debugging it might be a pain Nov 19 10:09:30 can you direct me at what I should even be looking at if I wanted my wlan0 clients not to get assigned ipv6 from my pd-subnet? Nov 19 10:10:21 I dont even know what controls the ipv6 assigments in openwrt. All I do is assign ip6prefix. The rest happens automagically Nov 19 10:10:57 the problem is that it's enabled per network zone, I suspect lan which uses br-lan Nov 19 10:11:14 I did solve it one way, mangle on prerouting for ipv6 matching wlan0 using physdev and dropping. It works. But the bad thing is that the clients OS would think it does have v6 but cant use it Nov 19 10:11:15 and it is not possible to configure it per bridge member Nov 19 10:11:41 barhom: you should probably filter outgoing RA from the router on wlan0 Nov 19 10:12:20 Thats a good tip, is RA what is telling my clients what ipv6 subnet they should use? Nov 19 10:13:22 this is the ra, dhcpv6 and, at least in theory, ignore options are for in /e/c/dhcp Nov 19 10:13:42 karlp: he can't solve his problem in /etc/config/dhcp Nov 19 10:14:07 because that controls (br-)lan, which contains both wlan0 and wlan1. he just wants to disable IPv6 on wlan0 :) Nov 19 10:14:32 so with the physdev rules I would filter outgoing RA, incoming RS, incoming DHCPv6 requests Nov 19 10:14:38 it controls whatever you have. you have dhcp sections for all the interfaces. Nov 19 10:15:12 or "networks" rather than interfaces, but that's what they've got. Nov 19 10:15:14 stintel: I will try that. You know out of your head how to match ra in ip6tables? Nov 19 10:15:26 I mean does it use some specific port/ip etc I can match on Nov 19 10:15:38 karlp: not interface. network segment as defined in /etc/config/network. in this case lan. no way to configure br-lan's members individually? Nov 19 10:15:56 given what else they're doign, why would they have everyone in one br-lan anyway? Nov 19 10:16:10 beats me really :) Nov 19 10:16:15 ok then :) Nov 19 10:17:00 I might end up splitting to two different L2 nets eventually. I just want to see if I can avoid it Nov 19 10:17:13 sounds like you avoid it by doign a massive pile of iptables, Nov 19 10:17:22 instead of just declaring two nets and having things just work? Nov 19 10:17:36 still don't understand (if situation still like yesteday) why barhom doesn't just make 2 separate networks for catering both wlans and then just allow zone forwarding between them Nov 19 10:18:04 olmari: The main reason is that normal non-tech-savvy people always expect all clients to be in the same subnet. Nov 19 10:18:24 why does that matter? Nov 19 10:18:33 allowing forwarding means clients reachable by clients on other side too Nov 19 10:18:36 you're very explicitly not treating them the same way, why should they be int he same v4 subnet? Nov 19 10:18:56 I'm about to make a 3rd SSID here :) Nov 19 10:19:50 You are all making valid arguments. I agree with you all. I still however know that a "normal user" would be WTF?! if IPs are different. Nov 19 10:20:06 I'll try the RA ip6tables, and if I fail I'll probably go 2x L2's Nov 19 10:20:40 I'd argue a "normal user" nowadays doesn't even know what an IP address is, let alone how to look it up :) Nov 19 10:20:57 I thought about that as I was typing ;) Nov 19 10:21:35 "normal" user turns on playstation 4 and expects it to get online Nov 19 10:21:44 How is ath10k support for multiple SSID? 2x 2G and 2x 5g? Nov 19 10:21:44 and UPnP to work!!!! Nov 19 10:22:04 alos, again, on our exact situation there, why do you care Nov 19 10:22:13 your* Nov 19 10:22:13 barhom: I have 2 APs here with each 2x2.4 and 2x5GHz SSIDs Nov 19 10:22:23 since using ath10k-ct it is stable Nov 19 10:22:41 great, Ill give it a try. That way I can get my 2g only clients to use vpn Nov 19 10:22:47 I'm about to add a 3rd SSID on 2.4GHz. probably skip 5GHz as most IoT devices only do 2.4 anyway Nov 19 10:23:20 what was the ath10k issues you guys were talking about for the past 2 days? Nov 19 10:23:31 I didn't quite understand what was wrong Nov 19 10:23:37 some people reported "issues" with ath10k in 19.07-rc1 Nov 19 10:24:09 and it was suggested to revert 19.07 back to mainline ath10k instead of ath10k-ct Nov 19 10:24:21 but that would reintroduce other bugs solved by switching to ath10k Nov 19 10:24:27 so I would so far it's inconclusive :) Nov 19 10:24:35 + say Nov 19 10:24:48 ..and on the same notion, there wasn't apparently real figuring out what was actually failed, or that I gathered from history Nov 19 10:25:09 indeed. although mesh seems to be problematic with -ct Nov 19 10:25:47 but I would argue AP mode is more important and people building large mesh networks probably build their own images so they can easily s/ath10k-ct/ath10k/ Nov 19 10:26:37 +1 Nov 19 10:26:47 that's already the case with gluon if I remember correctly Nov 19 10:27:19 well, one can still switch to ath10k even in the official images, right? Nov 19 10:27:37 some opkg-fu, but still doable Nov 19 10:27:47 if you ahve enouhg flash free for it. Nov 19 10:27:52 I'm not sure how that will work out. it will waste space for sure Nov 19 10:28:12 is that issue with devices having ath10k radios? Nov 19 10:28:15 and given I have trouble getting my AP images to fit in <8MB .. :) Nov 19 10:28:17 I say for default openwrt we want usually exactly the most commonly working thing per default (inrelated to what subsystem talking about), beauty of openwrt is that the on ewho really wants can, and even easily even is compiling from 0, to make whatever suits oneself Nov 19 10:28:29 ynezz: it's always an issue :) Nov 19 10:28:43 for testing? Nov 19 10:28:57 just to find out if it's ath10k-ct causing the issue Nov 19 10:29:34 and with ath10k we have the apparent shitty OEM one, and then one less shitty that greearb__ is kind enough to keep "hacking" Nov 19 10:31:58 (or the one guy if I am mistaken from identity) Nov 19 11:38:56 i just pushed the pcre(1) jit support PR to github (https://github.com/openwrt/packages/pull/10588) and I am open for criticism. Nov 19 11:58:15 looks straightforward, mhei is pretty responsive normally too. Nov 19 11:59:47 greearb__: ping Nov 19 12:00:15 I think a recent change broke mesh Nov 19 12:00:42 I tested various ct firmware/kmods but I suspect it might be hostapd bump that broke it Nov 19 12:01:04 the mesh gets established, but no traffic is passed (mesh_fwd specified on both ends( Nov 19 12:02:23 trunk 19.07 of 2 months back works fine on same hardware and exact same config Nov 19 12:02:50 I was wondering if this is a known issue or I should submit a bug? Nov 19 12:04:11 I see this in log: mac80211: Failed command: iw dev wlan0 set mesh_param mesh_fwding 1 Nov 19 12:05:03 abenz: some more or less relating discussions happened lately, but proper issue bould be cool... with good analyzing and testing preferably done (like you seem to have something at least).. often times earlier reports has been crappy mostly, kind of "this is broken" without any real analysis what actually is broken =) Nov 19 12:07:40 olmari: thing is I have a wave1 chip (qca9980 in archer c2600), mesh doesnt work on this with official qca firmware, CT firmware enables mesh though and recently after a few upgrades (I upgrade every few months), mesh almost always breaks or needs some tweaking to make it work again.. hence I had to ask to check if this is a general issue or something specific to my abandonded/EOL chip like it has been in the past Nov 19 12:08:24 I just fail to see how failing iw dev command is related to hostapd bump Nov 19 12:08:24 or is that a wave2, first gen? Just not the 9984 which seems to get all the patches/fixes Nov 19 12:09:38 abenz: all the other related programs stayued same? I mean for examplke IW full vs IW "mini" and whatnot? Nov 19 12:09:48 ynezz: my bad, perhaps mac80211, changes that require updates to openwrt scripts perhaps? I'm not sure Nov 19 12:09:53 what was last working revision? Nov 19 12:10:04 revision=git hash Nov 19 12:10:23 olmari: yes exactly the same, same .config, same configs. I used the same diffconfig and ./files for the previous build and this new one Nov 19 12:11:31 all good info, as oftentimes that is the actual "error" =) Nov 19 12:14:04 unfortunately I can't know for sure which revision it was Nov 19 12:14:23 abenz: you should put all this information in a bug report Nov 19 12:14:24 but it was 19.07 branch, running kernel 4.14.141/2 Nov 19 12:14:58 git reflog? Nov 19 12:15:12 git reflog is great Nov 19 12:16:16 would've been if I don't git pull every couple of days to see what changed Nov 19 12:16:49 I'm almost sure about the kernel version though Nov 19 12:16:55 4.14.141 is Aug. 30 and 4.14.142 is Sep. 9 Nov 19 12:17:40 do any of you have a mesh setup? is it working fine on current 19.07 ? Nov 19 12:18:08 btw I flashed my routers to 18.06 and mesh came up (again same .files and diffconfig) Nov 19 12:22:34 try to 1st revert d7ea38036323582c594d00890719132b95404484 ath10k-firmware: update Candela Tech firmware images and then 757fd85402f931327c91c62d854430cc3c29bbb7 ath10k-firmware: update Candela Tech firmware images Nov 19 12:22:42 then next suspect might be cabd12d2e36a1db08f841e5430e620a9a242e4e2 mac80211: Update to version 4.19.79 Nov 19 12:23:54 ynezz: given I jumped between many different CT firmwares I think I will try your second sugguestion first Nov 19 12:24:04 thanks for taking the time Nov 19 12:24:13 or just git bisect Nov 19 12:24:18 I will compile a new build and report back Nov 19 12:24:32 but since you don't know good revision, it wont work Nov 19 14:02:39 is it possible for a package to select a KERNEL_ option ? Nov 19 14:02:56 i.e, to make sure that when a given user-space package is enabled, some kernel config option gets enabled Nov 19 14:03:27 using DEPENDS:=@KERNEL_, I'm able to get a "depends on" dependency, but that makes the package invisible unless KERNEL_ is enabled Nov 19 14:03:29 kos_tom: possible yes, endorsed now Nov 19 14:03:33 *not Nov 19 14:04:11 it will cause the kernel for all platforms to grow when the package happens to be compiled in the same tree, which generally is a nogo Nov 19 14:04:26 it might depend on the specific feature in question though Nov 19 14:04:44 if its just for a private repository or feed then its fine too, obviously Nov 19 14:04:45 related to selinux support I'm working on, which will I'm sure trigger lots of discussion on the mailing list once posted Nov 19 14:05:06 the syntax would be DEPENDS:=+@KERNEL_... Nov 19 14:05:20 however, things like selinux should be a toplevel kernel build option Nov 19 14:05:28 similar to cgroups Nov 19 14:05:53 jow: I'm sure people will have plenty of comments to my patch series, so at this point I want to send some first implementation, and get some feedback Nov 19 14:05:54 and I could imagine it being enabled by defualt for x86 and fat ARMs Nov 19 14:06:29 but on ath79, ath25 and similar constrained targets it'll likely be a non starter Nov 19 14:06:48 try the +@KERNEL_xxx syntax, it should work afair Nov 19 14:07:10 if not, you'll need to add a menuconfig entry (look at config/) similar to cgroups Nov 19 14:09:56 I wonder if there'll come a day where the solution to "my comptuer isn't behaving the way I expected" is somethign other than "turn of selinux" Nov 19 14:15:08 or just turn it off Nov 19 14:17:21 fix ur policies! Nov 19 14:17:52 it's not *that* hard tbh Nov 19 14:18:33 I tried man, I tried for years. Nov 19 14:19:04 logread | grep Failed Nov 19 14:19:17 mac80211: Failed command: iw dev wlan0 set channel 149 HT40 Nov 19 14:19:21 jow: thanks a lot, it seems to work Nov 19 14:19:26 mac80211: Failed command: iw dev wlan0 set mesh_param mesh_fwding 1 Nov 19 14:19:50 ynezz: ^ after reverting the mac80211 update Nov 19 14:19:57 we develop our product on CentOS 7 with SELinux fully enabled. we need to support it as sometimes customers require SELinux by internal policy Nov 19 14:20:10 I still get these failed notices, do they appear on your boxes? Nov 19 14:20:12 http://false.ekta.is/2014/12/selinux-is-teh-suck-ive-given-up/ Nov 19 14:21:56 I was actually working on enabling it on my public servers, but haven't gotten very far yet Nov 19 14:22:13 selinux scares me Nov 19 14:25:11 https://github.com/opencontainers/runtime-spec/blob/master/config.md Nov 19 14:25:18 https://github.com/opencontainers/runtime-spec/blob/master/config.md#configuration-schema-example Nov 19 14:25:26 i would like to see procd/ujail support this Nov 19 14:26:12 then have Nov 19 14:26:30 1) dynamic containers bsed on the json which is what ujail does now + some extras Nov 19 14:27:02 2) static containers as folders or mountable blobs and a tool called uxc to create/manage them Nov 19 14:27:52 --> https://hackmd.io/LJzcQLt4QyqcBj2RU6nW8Q Nov 19 14:31:17 I don't follow much this container universe, using mostly docker, but lxc seems quite light Nov 19 14:31:51 ynezz: its all a wrapper around "man clone" Nov 19 14:32:13 with bloaty app layers Nov 19 14:32:53 *buzzwords* orchestration and provisionon in the cloud with a machine learning/AI thingy Nov 19 14:34:10 you're talking about docker now? Nov 19 14:34:27 yeah, the app layer Nov 19 14:34:39 the container/isolation part is real simple Nov 19 14:34:56 the magic of all these variants is the application layer/user tooling around it Nov 19 14:39:38 well, lxc is marketed as low level container runtime, so it makes me wonder why the need for uxc Nov 19 14:40:29 lxc containerizes a whole rfs incl systemd Nov 19 14:40:42 we want to containerize services dynamically Nov 19 14:40:49 also have you looked at the lxc code ? Nov 19 14:40:57 it aint pretty Nov 19 14:41:12 no, just at the amount of dir/files Nov 19 14:41:26 integrating lxc is more work than simply extending ujail with the bits of code missing Nov 19 14:41:54 ok, then it makes sense, looking forward to try it one day :) Nov 19 14:42:34 if my plan works, mid Q1 Nov 19 14:42:37 ;) Nov 19 14:54:42 it was the mac80211 commit Nov 19 14:54:47 reverting that only, mesh came up Nov 19 14:55:00 rebooted nodes randomly and it comes back reliably Nov 19 15:29:01 I'm also seeing very low speeds on my RPi0W clients Nov 19 15:29:23 right after connect they are OK (>1MB/s), but after some time the speed drops to few 100KB/s Nov 19 15:29:41 sometimes even below 100 Nov 19 15:29:52 does wifi transfer require entropy? Nov 19 15:30:30 because that keeps being a problem on the pi0 Nov 19 15:30:56 it usually takes ~2m uptime before crng init done and wifi connects Nov 19 15:31:16 you don't have urngd? :p Nov 19 15:34:43 I do Nov 19 15:34:53 oh, I don't Nov 19 15:34:53 wtf Nov 19 15:34:58 it's default enabled, no ? Nov 19 15:35:10 urngd - 2019-10-21-40f939d5-1 Nov 19 15:35:12 it's installed Nov 19 15:35:36 [ 13.170030] urngd: jent-rng init failed, err: 2 Nov 19 15:35:50 there you go Nov 19 15:36:40 so urngd doesn't work on pi0/1, same as kernel: [ 2791.483606] jitterentropy: Initialization failed with host not compliant with requirements: 2 Nov 19 15:36:53 low resolution timer Nov 19 15:43:12 abenz: so with 4.19.66 backports mesh works, but with 4.19.79 doesnt? Nov 19 15:55:24 jow: luci-app-cifsd (for example) uses fs.trimmed and fs.write to put a file template into a text area, but when I try it, I get nothing, and putting in console logs shows that the fs.trimmed and fs.write are just returning Promise objects. Is there a trick to make them resolve automatically? Nov 19 15:55:55 (is the cifs app actually jsut broken?) Nov 19 15:56:14 ooh there is luci-app-cifsd now? we can say goodbye to samba3x now? :) Nov 19 15:57:23 luci-app-firewall approaches it compeltely differnetly, and provides a top level handleSave function, but it uses fs.write().then? /me tries at least the .then syntax Nov 19 16:09:28 karlp: did you declare acls to allow write access to these files? Nov 19 16:09:36 erm read and write access Nov 19 16:09:56 promises returned by .load() or .render() functions are automatically awaited and resolved Nov 19 16:10:04 yeah, in the debugger I can see the promises have the contents. Nov 19 16:10:09 this is on 1907? Nov 19 16:10:42 19.07 should be feature identical to master Nov 19 16:11:16 let me take a look at the application Nov 19 16:11:30 https://zerobin.net/?ca9a0e3586eeb9b8#pMLVL4h0DleXPs5cIW9VQiMsWEBiEUhJ2KQCkdiK/QE= Nov 19 16:12:01 I can see my "default" file contents are loaded when I have the promises version uncommented. Nov 19 16:12:29 you need to return the promise from the function Nov 19 16:12:48 return fs.trimmed(...) Nov 19 16:12:56 how can I try a different file then? Nov 19 16:13:30 return fs.trimmed("/file/a").then(function(s) { return s || fs.trimmed("/file/b") }) Nov 19 16:13:47 yay. I love js. Nov 19 16:14:18 stintel: I'm using cifsd now. The only issue I found is that Solid Explorer (A file manager on Android) is trying to send some discovery packet (I remembered is nmbd protocol?) before connecting the server. Nov 19 16:15:01 karlp: alternatively this: return Promise.all([fs.trimmed("/file/a"), fs.trimmed("/file/b")]).then(function(results) { return results[0] || results[1] }) Nov 19 16:15:12 which might fell more natural (it should also be slightly faster) Nov 19 16:15:18 But cifsd does not have such feature, so the file manager stuck at "Connecting to server" for about 30 seconds. Nov 19 16:15:54 yeah, the PromiseAll might be nicer, I was trying to do that, but obviously trying to set the return value wasns't doing what I wanted that way :) Nov 19 16:18:01 ysc3839: But if you are using Windows, it seems works well. Nov 19 16:21:30 ysc3839: recent windows? might have the same problem when you older/newer ... Nov 19 16:21:49 you -> you use Nov 19 16:23:02 I'm using Windows 10. Nov 19 16:24:38 Recently I found Windows 10 can't connect to cifsd. After report it's fixed now. Nov 19 16:25:13 So I think it's ok to replace samba. Nov 19 16:26:28 this might explain my RPi0W wifi trouble? https://pastebin.com/BPrWB03a Nov 19 16:27:31 ysc3839: windows 10 dropped support for a lot of legacy stuff, so maybe what you Android app does is some older mechanism not present in windows 10 anymore... Nov 19 16:28:24 ok, that's working well now jow, thanks. Any idea why textareas aren't resizeable? I can set o.rows=blah, and the handle is there, and I can resize sideways, btu not vertically. Nov 19 16:28:48 karlp: allegedly this is due to some "height: auto" line in the default CSS, there's an open bug about it Nov 19 16:29:11 I find it ... unexpected that height auto prevents resizing but well Nov 19 16:29:31 didn't find the time to investigate / refactor the CSS yet Nov 19 16:29:48 no problem, if it's a bug, not somethign I need to configure, that's fine with me :) Nov 19 17:03:50 abenz, here. wave-1 ath10k-ct firmware does NOT support mesh, it at least somewhat works with IBSS. wave-2 ath10k-ct might support mesh, I have not tested. Nov 19 17:08:55 jow: ok, loading/saving/removing all works, files are written, defualted, removed, all good. Nov 19 17:09:13 except that the modal "applying cahnges.... " dialog flashses "no changes to apply" Nov 19 17:09:42 it... _works_ just isn't an ideal indicator. https://zerobin.net/?0babdae2bc643dfc#LZWBFBAMav+hMOI3MOufYj2EEapHXTEp7vkGIesuHA8= Nov 19 17:18:45 karlp: that happens if no uci values were changed Nov 19 17:18:57 is this all non-uci values? Nov 19 17:19:19 in this case yeah, only hcanged the contents of the file backed text area. Nov 19 17:19:27 I mena the entire view Nov 19 17:19:36 no, all uci vars except a couple of text areas Nov 19 17:19:55 ok Nov 19 17:20:03 the you need to live with that for now Nov 19 17:20:19 fair enough :) Nov 19 17:21:55 I guess I can fiddle with the procd files to make sure that "reload_config" will still do the right thing I guess. Nov 19 17:23:37 Maybe I'll simply change the text to something else Nov 19 17:26:43 hrm. this should still be calling reload_config() somewhere though right? or has it decided not to do that because "no changes" ? Nov 19 17:26:52 I've added the files to procd_set_param file... Nov 19 17:27:41 need to check Nov 19 17:28:45 hrm, file params aren't in ubus callservice list Nov 19 17:28:59 did you pass the verbose param? Nov 19 17:29:10 * karlp learns to spell verbose properly :) Nov 19 17:29:33 * ldir laughs Nov 19 17:30:03 ok, files aren't in the verbose list either, even htough it's more verbose.... Nov 19 17:32:16 well, reload_config on the cli works, so they're obviously listed :) Nov 19 17:33:16 yeah, the cbi form save doesn't cause relaod_config. Nov 19 17:34:03 jow: can I do soemthing with formvalue/cfgvalue/textvalue to have the action of reading/writing the file be "extra" and have a uci value get saved anyway? like, savea timestamp there that will always change? Nov 19 17:34:40 yes, ultimately rpcd's uci apply is invoked which iterates the list of changed uci configs, it'll commit each config and emit a changed event for each which are then picked up by procd triggers Nov 19 17:35:01 without any uci changes, no events will be emitted Nov 19 17:35:16 writing some uci option would be one way to enforce such a trigger Nov 19 17:38:42 hrm, I'd probably almost prefer a DummyValue hidden that _always_ wrote a timestamp or something so that you got sometimes unnecessary saves, instead of sometimes no reload_config at all. Nov 19 17:38:55 will have to play with that tomorrow. thanks for your time today Nov 19 17:39:31 how does one get reload_config to be called if it _wasns't_ a UCI view? Nov 19 17:39:47 call ubus service reload yourself manually? Nov 19 18:24:36 ynezz: yes correct. Nov 19 18:25:59 greearb__: the card I'm referring to is the qca9980 (wave2 first gen), I'm using CT firmware with it (stock qca is over 4 years old, and doesn't support mesh) Nov 19 18:27:26 oh, that should support mesh I think, same general FW as the 9984 Nov 19 18:28:26 greearb__: yes, running mesh on it now Nov 19 18:45:40 karlp: ubus call service event '{ "type": "config.change", "data": { "package": "foo" }' Nov 19 18:45:47 karlp: where foo is the uci config of the service Nov 19 18:56:17 jow: can you please share the buildbot vpn setup? Nov 19 18:57:34 blogic: I read you wanted to document the security handling in OpenWrt, then please extend this: https://openwrt.org/docs/guide-developer/security Nov 19 19:00:44 aparcar[m]: its just a bunch of manual autossh settings Nov 19 19:01:07 aparcar[m]: for the new setup you should use something more robust, wireguard or openvpn Nov 19 19:02:36 okay, I'll replace the current https approach with wireguard Nov 19 19:07:58 blogic: please let me know why the kmods fails, it's the same error message as for master: https://builder.prpl.dev/#/builders/4/builds/2 Nov 19 19:11:51 aparcar[m]: it's right there in the output; "Lantiq GRX500 Platforms (USB_DWC3_GRX500) [M/n/?] (NEW)" - the symbol is appearently neither set in any kernel config, nor by any kmod Nov 19 19:12:47 KanjiMonster: thank you. is there a simple fix for that? Nov 19 19:13:31 aparcar[m]: whatever was responsible for seting USB_DWC3 should also set it Nov 19 19:18:00 mhh it works locally, than I might be a special buildbot thing which enables all kmods or something... Nov 19 19:18:09 stintel: CI will fail no matter what. boost-python takes too much memory for CircleCI. Locally I get no linking failures. Nov 19 19:18:30 mangix: it failed locally for me when I was adding the telldus dep myself Nov 19 19:18:42 weird Nov 19 19:18:47 linking error Nov 19 19:19:22 ah Nov 19 19:19:27 hmmm Nov 19 19:19:37 that might have been the python thing Nov 19 19:19:53 ah OK Nov 19 19:20:56 aparcar[m]: yes, buildbot enables all kmods by default; if you set ALL_KMODS or ALL_NONSHARED (IIRC) you'll get the same behaviour locally (with a fresh config of course) Nov 19 19:22:52 hmmm someone reported jool not compiling on 17.01 Nov 19 19:22:57 wonder if I should fix it. Nov 19 19:23:35 KanjiMonster: Thank you again! I'm afraid I'm to much of a kmod noob to figure out things even it I get it running (crashing) locally. I'll wait for blogic Nov 19 19:23:41 but thanks for the insights Nov 19 19:24:37 jow: in the buildbot config.ini.example is a line like `cc_version = lt 4.9`. I changed it to gt but what is your recommended value? Nov 19 19:25:10 .....this guy is confused Nov 19 19:27:18 musl 1.1.16... I think that's in 18.06, not 17.01 Nov 19 19:27:30 aparcar[m]: I think the idea is to test build with the lowest supported version (as we claim) Nov 19 19:28:06 isn't it not 7.n? Nov 19 19:29:14 wow Nov 19 19:29:20 lede is still getting updates Nov 19 19:30:40 aparcar[m]: https://github.com/openwrt/openwrt/blob/master/include/prereq-build.mk#L29 Nov 19 19:31:10 aparcar[m]: USB_DWC3_GRX500 is added by a patch, it is not in upstream Nov 19 19:31:52 Hauke: yes I found it in the intel_mips stuff, so how to deal with it? Nov 19 19:31:59 it is probably not handeld in the generic kernel config or in package/kernel/linux/modules/usb.mk Nov 19 19:32:36 you could add it to intel_mips/config-4.9 as "# CONFIG_USB_DWC3_GRX500 is not set" Nov 19 19:32:47 but an make kernel_oldconfig will probably remove it again Nov 19 19:33:10 hahahaha tools/scons does not build on 17.01 Nov 19 19:33:19 missing python2 dependency Nov 19 19:33:33 Hauke: wouldn't you want to set it to =y? Nov 19 19:34:45 Hauke: yea what is the clean solution for that? Nov 19 19:41:05 Hauke: next error at ATM_BR2684_MINI_JUMBO_FRAME_SUPPORT Nov 19 19:41:32 is there like an automatic fix? Nov 19 19:48:29 bah Nov 19 19:50:09 brcm2708/patches-4.19 contains a lot of unneeded cruft Nov 19 19:50:48 xback: did you notice you had to edit a patch to apply on 4.19.84 only to have to edit a patch that reverts that patch later ? Nov 19 19:51:03 lol Nov 19 19:52:30 or target/linux/brcm2708/patches-4.19/950-0091-net-Add-non-mainline-source-for-rtl8192cu-wlan.patch Nov 19 19:52:37 5+MB that we don't even use :/ Nov 19 19:52:53 jow: nice, I can throw that ubus call into the write handlers of the model. Nov 19 19:53:41 I can't wait to nuke that brcm2708 target someday Nov 19 19:56:04 aparcar[m]: whatever happens to ship as default with debian stable's build-essential Nov 19 19:57:11 iirc the intend of picking such an old cc was to ensure that resulting IBs and SDKs work among a wide range of distros Nov 19 19:57:15 centos in particular Nov 19 19:58:19 if I recall correctly, the newer CCs produced host utils not executable on the older distro Nov 19 20:10:58 jow: C++ ABI issue? Nov 19 20:59:11 stintel: yes he did notice - xback's plan is to drop the 2 patches later Nov 19 21:00:08 ok Nov 19 21:00:13 Noltari should have probably dropped those when importing the patches from rpi foundation Nov 19 21:00:27 but it's a pita and takes time and effort Nov 19 21:21:34 how do I build an older version of the master branch? Is resetting git head at that particular commit enough? Nov 19 21:23:48 t4h4: for the base buildroot yes that's enough.. I don't know how to have specific version(s) of feed(s) Nov 19 21:27:27 jow: would you mind moving the buildbot config.ini to config.yml? I would do the PR but do you have specific reasons to favor ini? Nov 19 22:01:19 I guess only way would be git reset feed packages individually? Checking out a branch does this automatically but resetting to a commit does not. Nov 19 22:04:19 Propably.. what for you want? Nov 19 22:04:21 aparcar[m]: there is probably a kmod which sets it to =y or =m Nov 19 22:05:28 trying to find which commit introduced a particular bug Nov 19 22:07:56 Mm.. well if you know where the affected program is then suffice to tinker with that repo only I'd day Nov 19 22:08:04 Say* Nov 19 22:09:09 https://forum.openwrt.org/t/sta-speed-slows-down-after-eap-reauthentication/48506/ Nov 19 22:10:07 "git bisect" Nov 19 22:25:48 thats a TIL, thanks Nov 19 22:35:24 aparcar[m]: yeah, ini is simple to parse with sed Nov 19 22:35:31 yml not so much Nov 19 22:35:55 I'd prefer to keep ini files Nov 19 22:52:13 jow: for me QCA9880 with ath10k-ct on 5GHz works Nov 19 22:52:22 in openwrt 19.07 Nov 19 22:53:06 the AP can do 180 MBit/sb Tx and only 40 Mbit/s RX over 8m with a wall in between Nov 19 22:56:17 with sae-mixed Nov 19 22:58:13 I see similar result on 9880 but master, haven't yet able to test would 11w 0 correct it (relating to chat history) Nov 19 23:01:47 I set ieee80211w=0 and see the same results Nov 19 23:01:58 how do I check if I am using 11w? Nov 19 23:02:43 Mmh.. Nov 19 23:03:09 Iw output? Nov 19 23:03:46 MFP is another name for that (to look for) Nov 19 23:04:17 olmari: from which subcomand of iw? Nov 19 23:04:54 Or mpf, i never remember Nov 19 23:05:15 Dunno on top of head.. already at bed with phone =D Nov 19 23:05:47 it is also very close to 40MBit/s not much chaning the lowest I saw in about 30 outputs from iperf3 was 438.8 MBit/s and the higest 41.2 MBit/s Nov 19 23:06:38 olmari: that's a good idea, I should also go to bed Nov 19 23:22:54 olmari: iw wlanX station dump Nov 20 00:06:36 here's an off-topic query: Any idea what the 'CC' column means in apple's 'airport' scan results: Nov 20 00:06:47 scan line: SSID BSSID RSSI CHANNEL HT CC SECURITY (auth/unicast/group) Nov 20 00:06:47 /scan line: MimoSeries1 04:f0:21:63:68:4a -50 100 Y SG NONE Nov 20 00:06:47 /scan line: ubnt-hd-ent 02:9f:c2:69:55:12 -55 60 Y -- WPA2(802.1x,FT-802.1x/AES/AES) Nov 20 00:08:43 greearb__: country code? Nov 20 00:09:27 good guess, let me inspect that a bit more Nov 20 00:12:45 yes, that seems to match nicely Nov 20 00:22:13 xback: I have problems building the kernel for ipq40xx (zyxel nbg6617) on 19.07. Errors out in drivers/dma/qcom/bam_dma.c, so possibly related to 040-dmaengine-qcom-bam-Process-multiple-pending-descript.patch Nov 20 00:22:19 xback: https://bpaste.net/show/VUTHU Nov 20 00:22:54 not sure when it broke but I can bisect it if need be Nov 20 01:55:42 xback: it is in fact the bump to 4.14.154 **** ENDING LOGGING AT Wed Nov 20 02:59:58 2019