**** BEGIN LOGGING AT Fri Oct 18 02:59:58 2019 Oct 18 07:03:12 aparcar[m]: is there any way to remap the paths used by Weblate? e.g. zh_Hans to zh-cn Oct 18 07:04:38 if not I would workaround it using symlinks for now... the problem is that the codes used in the directory path must correspond to locales used by browsers in their Accept-Language header Oct 18 07:05:38 there's zh-{cn,hk,sg,tw} for PRC, Hong Kong, Singapore and Taiwan respectively while Weblate uses zh_Hanc and the like Oct 18 07:08:00 jow: thanks for pointing that out. I'm sure there is a solution within weblate. I'll update you once I know more Oct 18 07:13:58 aparcar[m]: btw my weblate invitation token expired, can you reissue one? Oct 18 07:15:23 jow: done Oct 18 07:15:34 be warned, the website is pretty laggy Oct 18 07:16:22 jow: check these settings https://hosted.weblate.org/addons/openwrt/luci/367/ Oct 18 07:18:53 still waitingh for my invite mail Oct 18 07:19:42 hm, why it's not public? Oct 18 07:20:02 I mean, getting permission denied while trying access that URL Oct 18 07:20:18 ynezz: because its the admin settings for the luci org Oct 18 07:20:33 strange :) Oct 18 07:21:09 here's the public pages: https://hosted.weblate.org/projects/openwrt/ Oct 18 07:21:21 nice, thanks Oct 18 07:21:33 wow, long list Oct 18 07:22:12 * ldir is stunned that a kernel bump hasn't affected any generic kernel patches at all! Oct 18 07:22:19 its a bit unfortunate that luci-base (the most important part) is at the end of the last page Oct 18 07:22:46 but life is hard, so nvm Oct 18 07:23:02 jow: yes I see that too, I'll open an issue Oct 18 07:24:04 I invited jo at mein io Oct 18 07:24:20 yeah, its probably teh greylisting Oct 18 07:28:26 jow: Re: [buildbot] add helper script to (re)build docker images, we could start mirroring the repo to gitlab, add .gitlab-ci.yml so the images would be pushed to docker hub after push to master branch automagically, saving some time Oct 18 07:28:45 ynezz: yep, that probably makes sense Oct 18 07:29:07 speaking of gitlab, I've started with gitweb2gitlab migration script, just finished the projects/ but would like to do them all soon Oct 18 07:29:38 should we retain the same structure with feeds, project subgroups as it's now done on gitweb ? Oct 18 07:29:58 yes, please Oct 18 07:30:29 at some point in the past we discussed that external feeds should be mirrored through our repository server before referenced by feeds.conf Oct 18 07:30:56 indeed, makes sense Oct 18 07:31:23 jow: Error in parameter email: User with this E-mail already exists. Oct 18 07:31:23 Error in parameter email: A user with that username already exists. Oct 18 07:32:48 jow: any ambitions to update the buildbots to like 2.5 Oct 18 07:32:58 oh wow actually released 20 hours ago Oct 18 07:33:11 what's wrong with current version? :) Oct 18 07:33:24 it's like 0.x? Oct 18 07:33:25 aparcar[m]: if you feel like it, go ahead Oct 18 07:33:50 I'm happy to have it running more or less stable now so I'd like to avoid more potential regressions atm Oct 18 07:34:17 to make it future proof one could start replacing deprecated code though Oct 18 07:35:01 jow: what exactly is deprecated? Oct 18 07:35:05 aparcar[m]: regarding weblate; then I'll register normally and you need to appoint me to admin or something Oct 18 07:35:22 jow: fine for m Oct 18 07:35:23 e Oct 18 07:35:23 aparcar[m]: I think it will get printed on startup Oct 18 07:35:37 just start the compose setup of buildbot.git Oct 18 07:36:03 its something about WithProperties() vs. renderers or decorates or so Oct 18 07:36:26 parts of the master.cfg already are in newer format Oct 18 07:36:58 jow: yey sounds like a nice project Oct 18 07:38:42 aparcar[m]: well apparently I am unable to register and unable to login Oct 18 07:38:43 ynezz: should we make a PoC with gitlab runners? they manage most of the stuff on their own and are easier scaleable Oct 18 07:39:05 jow: can you provide another mail address? Oct 18 07:39:21 aparcar[m]: I don't follow with runners Oct 18 07:39:21 I could but then I'd be another acount and I'd get spammed twice Oct 18 07:39:38 the system already knows me and even reported "suspicious activity" because I tried to login Oct 18 07:39:47 maybe I can try a forgot password route Oct 18 07:41:01 aparcar[m]: before building yet another build setup from scratch I'd prefer to stabilize the current one Oct 18 07:41:21 it literally took years to find and squash most quirks in buildbot Oct 18 07:41:28 *buildroot Oct 18 07:41:44 password reset worked Oct 18 07:41:53 jow: all right but than I'll start fixing it with 2.5 if you don't mind Oct 18 07:43:11 aparcar[m]: buildbot has gitlab integration support nowadays Oct 18 07:43:15 https://github.com/buildbot/buildbot/blob/master/master/docs/examples/gitlab.cfg Oct 18 07:43:20 that might be a viable route Oct 18 07:43:41 sweet Oct 18 07:43:48 keep me upated on your mail Oct 18 07:46:08 jow: I don't know buildbot that much to understand what is going on with that config file Oct 18 07:46:26 does it mean, it would be possible to use buildslaves also as job runners for gitlab? Oct 18 07:48:46 ynezz: no, that config only integrates with gitlab repos as change sources Oct 18 07:48:56 ah, ok Oct 18 07:49:59 jow: why do you use an rsyn script and not the buildbot file uploadone? Oct 18 07:50:01 I'm undecided yet if gitlab runners are the way to go for really longrunning build processes Oct 18 07:50:10 aparcar[m]: ? Oct 18 07:51:21 aparcar[m]: well mainly because we don't want to upload the artifacts to the build master put to a dedicated 3rd storage server Oct 18 07:51:27 *but Oct 18 07:52:03 jow: thanks for your mac_generator function, I didn't realize that hexdump can open a file... I always use stdin, and in this case it was a great improvement Oct 18 07:52:16 and we don't want the build master, which is handling secrets, to reside on the storage server Oct 18 07:52:33 jow: thanks that makes sense. I though it's uploaded to the master so it can acutally sign stuff Oct 18 07:52:35 by the way, did you commit the rpcd bug fix to master? Oct 18 07:52:49 aparcar[m]: no, only a few small files are uploaded to the master for signing Oct 18 07:53:01 and this happens via the buildbot fileupload facility iirc Oct 18 07:53:32 https://git.openwrt.org/?p=buildbot.git;a=blob;f=phase1/master.cfg;h=6675f2570a7e64c405bf3cf6a6b452b6ef4b8c5c;hb=48df35bfe9f15dfe751bad9aa8a9aa9d2279a866#l971 Oct 18 07:53:42 see that and the following steps Oct 18 07:57:03 msgito: yw. you should convert the other `dd if=/dev/urandom ... | hexdump ...` too. E.g. in line 109/110 Oct 18 08:01:04 jow: ack I'll check that out too Oct 18 08:01:34 what about using some kind of cdn? the peakdownload of sdk etc is like... 10mbit Oct 18 08:02:24 can't we ask like... fastly? they support projects like alpine, tor etc too https://www.fastly.com/open-source Oct 18 08:06:07 sure Oct 18 08:06:32 switching to the sfc was part of the plan to be able to get proper sponsorship agreements with such cdns and other providers Oct 18 08:07:06 so I'd write them a mail on how to set things up? Oct 18 08:07:59 that'd be great. However I do not want these CDNs to be the canonical source, but rather a caching proxy network which we can use as default download and opkg repo source Oct 18 08:08:05 jow: btw could you figure out how to login to weblate? it is super slow, maybe I'm just to far off on this island Oct 18 08:08:15 yeah I am logged in and it is fast here Oct 18 08:09:36 jow: okay I guess you are the new translation leader then ;) Oct 18 08:10:08 jow: yea I'd suggest to have cdn.openwrt.org which could be a mirror of downloads. however wants speed by the cost of privacy can shose so Oct 18 08:10:19 maybe even offer it within luci Oct 18 08:10:25 yes jow, I already converted the other function hahaha Oct 18 08:10:27 means the build infra uploads to your own servers and either we push the stuff from there to the cdn or the cdn simply proxies these servers Oct 18 08:10:58 I don't really like our assets to be hosted in some U.S. juristication Oct 18 08:10:58 I just didn't push the changes yet, awaiting for more comments Oct 18 08:12:07 jow: I guess the clue of cdn is that it is hosted everywhere? Oct 18 08:12:28 aparcar[m]: caching / mirroring etc. is fine Oct 18 08:12:33 I mean we could also setup our own CDN I guess, just asking some universities if they have some bandwidth to share Oct 18 08:13:14 no, we need something with geo distribution, anycast routing etc. Oct 18 08:13:32 just having a bunch of mirrors does not work, they'll be unused and consume more bandwidth syncing than serving users Oct 18 08:14:08 so lets fastly or $insert_corp handle it Oct 18 08:14:13 yea I looked a bit into that and it seems like just another big project. I'll write fastly a mail asking how they work Oct 18 08:14:13 https://hosted.weblate.org/widgets/openwrt/de/luci/svg-badge.svg Oct 18 08:14:18 as long as it draws the contents from servers under our control Oct 18 08:14:22 btw luci-base is now at 100% Oct 18 08:14:59 (at least the german translation) Oct 18 08:19:08 jow: I think the weblate webhook does not work. it does not automatically pulls new changes. Maybe we should install the app and set luci to "pull request only", so the app does not direclty commit anything Oct 18 08:32:32 aparcar[m]: fastly is pretty straight-forward. they act as a reverse proxy to your backend server(s). you setup a CNAME record to point at fastly. Oct 18 08:33:03 agb: sweet that sounds doable Oct 18 08:37:01 jow: I've seen multiple people asking how to donate to openwrt. What do you think of me setting up opencollevtive to gather some money which is than paid to finance a "release manager" (likely you) over the next two weeks to eventually get 19.07 out? So you could work fulltime on it for a little while Oct 18 08:39:36 Caching is controlled primarily via HTTP caching headers emitted from your backend server. If you need something more fancy I think Fastly uses VCL (the Varnish Config Language) for custom configuration. Oct 18 08:41:10 agb: I guess the basics should be fine for starters Oct 18 08:41:24 any not openwrt servers beeing used for sdk/ib download would be nice Oct 18 08:41:50 yeah. hosting of static assets you don't need anything fancy. Oct 18 08:44:00 aparcar[m]: donations, yeah - but lets finalize the sfc topic Oct 18 08:44:29 is that glued together? Oct 18 08:44:41 aka would the sfc create such accoutns? Oct 18 08:46:50 aparcar[m]: not directly but it helps to have the organizational structure clarified before we start with other endeavours Oct 18 08:47:18 imho it makes sense to pool the donations in one account, then distribute as needed (hosting, funding, travel, events) Oct 18 08:48:00 of course everyone is still free to donate whereever he likes and however he likes Oct 18 08:49:09 but it would probably be better if the official donation channel is ready before alternative ones are advertised Oct 18 08:49:47 so create the account now, wait for sfc takeover, start advertising? Oct 18 08:58:36 wait Oct 18 09:01:28 yea I'm working on a project proposal for my subjects, I'm not gonna create any account soon Oct 18 09:03:23 Is this valid? "%s (%s), ":format(wiface.ssid, wiface.device) I am getting " unexpected symbol near ':' " Another way to reprocude it executing lua -e'a="TEST:"; b="%s (%s), ":format("za", "zb")' Oct 18 09:05:09 if this is Lua it should be ("text"):method() Oct 18 09:05:25 thanks ynezz, testing :D Oct 18 09:05:26 yes literal string expression need to wrapped in parens to invoke meta methods on them Oct 18 09:05:58 msgito: fyi, there's #lua :p Oct 18 09:06:03 jow: btw did you see my PM from yesterday regarding the bugs database dumps? Oct 18 09:06:10 that is working! :D Oct 18 09:06:54 aparcar[m]: yeah, later Oct 18 09:07:05 right now that rpcd leak is on my todo Oct 18 09:07:19 sure whatever Oct 18 09:08:11 jow: yea it is not time critical at all, just at some point nice in case the move becomes more concrete Oct 18 10:49:08 oooh - ash supports exit traps Oct 18 10:52:41 .oO(all that years...) Oct 18 10:59:21 I'm learning a lot about shell scripting from trying to tidy up someone else's script Oct 18 10:59:50 until 30 minutes ago I didn't know what an exit trap was :-) Oct 18 11:04:06 turns out shell scripts _can_ be robust, not just layers of hacks :) Oct 18 13:00:16 hello guys ! Oct 18 13:00:32 and girls maybe ? Oct 18 13:00:39 anyway... Oct 18 13:02:43 I'm trying to compile libwebsockets for openwrt 19.07 Oct 18 13:02:57 target is mipsel Oct 18 13:03:27 I just can't figure out where is the openwrt compiled openssl lib Oct 18 13:03:52 anyone has a hint ? Oct 18 13:04:54 what doesn't work? libwebsocket is packaged in packages feeds Oct 18 13:08:08 I can't compile on the device, it will be a custom server written in c. Oct 18 13:08:23 uh, on device? Oct 18 13:08:31 mt7688 Oct 18 13:08:50 128M ram, 32M flash Oct 18 13:08:57 it's tight Oct 18 13:09:01 and slow Oct 18 13:09:25 why can't you just opkg install it? Oct 18 13:09:43 it's not showing Oct 18 13:10:09 root@OpenWrt:~# opkg list | grep libwebreturn none Oct 18 13:11:03 *** root@OpenWrt:~# opkg list | grep libweb* returns nothing Oct 18 13:11:52 "Opkg update" first? Oct 18 13:12:25 nope Oct 18 13:12:52 I mean, already tried Oct 18 13:15:22 https://downloads.openwrt.org/releases/19.07-SNAPSHOT/packages/mipsel_24kc/packages/libwebsockets-full_3.1.0-1_mipsel_24kc.ipk Oct 18 13:15:25 its there Oct 18 13:15:31 maybe your opkg setup is incorrect Oct 18 13:15:39 maybe Oct 18 13:16:52 so, iic, I just need to cross compile my apps with the libwebsockets headers and it will use the lib as a shared lib ? Oct 18 13:17:03 yes Oct 18 13:17:23 I wonder why libwebsockets is ~100K large Oct 18 13:17:48 does not speak for the simplicity of the protocol Oct 18 13:18:16 How do you check the compilation flags of a packaged lib ? Oct 18 13:18:25 haha indeed Oct 18 13:18:48 its a boilerplate also, websockets, http server, tls etc... Oct 18 13:18:52 you can't really check that. The packages libraries are shipped without header and debug symbols Oct 18 13:19:11 you could take a look at the makefile Oct 18 13:19:20 I tried to use lighttpd but the documentation is so scarse and old that i just ditched it. Oct 18 13:19:45 okay will try that Oct 18 13:19:50 thanks Oct 18 13:20:49 jow: says a lot about the lws project too actually. Oct 18 13:20:53 it's gotten out of hand. Oct 18 13:21:27 but eyah, complex, lots of versions, client and server support, bleh Oct 18 13:22:01 I'd love to have some lightweight (sub 10K) websockets library I could tuck onto uhttpd Oct 18 13:22:24 but I'm not going to p*ss away 100K for fancy XHR polls Oct 18 13:22:42 I've seena couple of small ones for client side only, but not for server side. Oct 18 13:23:05 probably because they all ship own mainloops and server skelettons Oct 18 13:23:41 that's why i tought about linking the lib as static, so i don't have the whole boilerplate on the device libs Oct 18 13:24:02 okay, I'm missing libuv Oct 18 13:24:24 How did you find the 19.07 ipk ? Oct 18 13:24:50 linking the lib as static is dumb, I'm not even including all the parts of it already, it's just a big library :) Oct 18 13:25:20 tionebrr: well I know it must exist, so I checked the repo manually Oct 18 13:25:27 and indeed it was there Oct 18 13:25:44 so my opkg config is bjorked : Oct 18 13:27:08 meh, -full is only an extra ~10k. why bother :) Oct 18 13:27:55 jow: talk to https://github.com/zhaojh329/libuwsc about adding server side? it's even built on libubox :) Oct 18 13:28:40 it's only 13k so far Oct 18 13:55:41 aparcar[m]: your json files are there now: https://downloads.openwrt.org/snapshots/targets/mpc85xx/generic/ Oct 18 13:55:48 however I'm not sure I like it Oct 18 13:59:27 okay guys, thanks. Oct 18 14:00:15 My opkg distfeeds was missing the mipsel_24kc packages repo Oct 18 14:00:36 only the mt76x8 as there Oct 18 14:00:41 *was Oct 18 14:04:04 the mpc85xx is power pc right? Oct 18 14:07:39 build #25 of ramips/rt305x is complete: Failure [failed buildinfo] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/ramips%2Frt305x/builds/25 blamelist: Michael Heimpold , Koen Vandeputte , Dan Haab , Jo-Philipp Wich , Christian Franke Oct 18 14:07:39 Oct 18 14:13:09 gah Oct 18 14:14:26 damn buildinfo crap Oct 18 14:14:47 aparcar[m]: the builders share the same config for 18.06, 19.07 and master Oct 18 14:43:23 * karlp laughs Oct 18 15:42:48 i just read master as magic Oct 18 15:42:53 #equestria Oct 18 16:25:08 so that rpcd leak was definitely due to libiwinfo Oct 18 16:25:48 looked at the code for like 20 minutes before it made "click" Oct 18 16:26:36 * ldir waits for the fix commit before hitting 'build' :-) Oct 18 16:27:07 I already accidentally fixed while refactoring the code, I just retested an older version to confirm that it was actually the issue Oct 18 16:27:19 so ts fine in master Oct 18 16:30:25 ha :-) Oct 18 17:35:13 jow: when was the problem introduced? Oct 18 17:50:34 Hauke: I think it was always there but it recently surfaced when newer luci began to do certain iwinfo ubus calls involving radio names instead of interface names Oct 18 17:51:39 Hauke: https://git.openwrt.org/?p=project/iwinfo.git;a=blob;f=iwinfo_nl80211.c;h=4d20489ded3bf50686fe93ebcc01774f2d2f89f9;hb=2a95086db2b730346d87c320e2c753126c4ff795#l235 this function will loose the memory returned by the first glob() and overwrite &gl in the second glob() call Oct 18 17:51:49 * Slimey burp Oct 18 17:52:23 Hauke: the leak in rpcd can be easily provoked by running something like while true; do ubus call iwinfo info '{ "device": "radio0" }'; done Oct 18 17:53:15 Hauke: it has been (inadvertently) fixed by https://git.openwrt.org/?p=project/iwinfo.git;a=commitdiff;h=a29b7d4e26a8d97e48c2a899dda3665d385b95db Oct 18 17:53:24 Hauke: didn't yet check if 18.06 is affected Oct 18 18:00:52 ok thanks Oct 18 18:01:13 I think I also saw some memory leak in 19.07 when I last rebooted my production gateway Oct 18 18:01:20 bit was probably this one Oct 18 18:01:23 *it Oct 18 18:07:18 Hauke: why are you talking to yourself on the ML? :) Oct 18 18:23:42 only way he gets sensible answers? Oct 18 18:27:05 https://fccid.io/WBV-HIVEAP330/Internal-Photos/Internal-Photos-1498814 Oct 18 18:30:34 ldir: hehe Oct 18 18:31:18 jow: sorry for the trouble and that k you for the fixup. I'll now setup my own buildbots to reduce the number of such error introduced by me :/ Oct 18 18:40:59 aparcar[m]: hey Oct 18 18:41:07 did you see my PM Oct 18 18:41:26 sudo should work Oct 18 18:41:34 lemme know if it does not Oct 18 18:42:01 blogic: Yea my first slave! Oct 18 18:42:06 :-) Oct 18 18:42:17 actually worker which is more correct Oct 18 18:42:39 pfff Oct 18 18:43:15 I have black, gay, lesbian and disbaked friend they dont hate me for using IT terms Oct 18 18:44:00 god will be laughing her lesbian black arse of looking at us Oct 18 18:47:26 blogic haha true talk! Oct 18 18:47:53 Have you seen the film dogma? Oct 18 18:47:55 sorry i missed chruch was practising lesbian witchcraft Oct 18 18:48:08 Tapper: for sure mate Oct 18 18:48:17 i hae a special relation to god ;) Oct 18 18:48:19 I love that film. Oct 18 18:48:49 its amazing Oct 18 18:48:51 The humor is just the gratest Oct 18 18:48:57 yup Oct 18 18:53:11 YaY Oct 18 18:53:20 uh oh Oct 18 18:53:26 * blogic gets a dump of broadcoms flow HW aceel code under GPL Oct 18 18:53:47 its do ugly a mother could not even love it Oct 18 18:54:05 but its only 3k lines of code Oct 18 18:54:26 looks like we will see upstream flow table accel for northstar soon Oct 18 18:58:01 awesome Oct 18 19:01:18 CVE in openwrt it is clamed. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2019-17367 Oct 18 19:06:02 O it's fixt in master Oct 18 19:06:10 blogic: do you mind I run the buildbot on your server with -j12 but use ionice and nice so your builds always hae prio? Oct 18 19:08:52 jow: do you mind I move the json files in it's own box like "json info files" just below the "supplementary files"? this way it does not annoy anyone but is still available Oct 18 19:14:18 Tapper: jow fixed it already Oct 18 19:14:43 aparcar[m]: mi casa es su casa Oct 18 19:14:58 aparcar[m]: worste case i'll mail you that i killed soemthing Oct 18 19:15:10 arigato Oct 18 19:15:15 ;) Oct 18 19:15:41 everything on that server from me is inside a git Oct 18 19:16:10 just make sure it stays on ubuntu 16.* otherwise qsdk wont build on it Oct 18 19:16:37 because qualcomm likes pre-historic distros Oct 18 19:17:23 aparcar[m]: and if it is not powerull enough let me know we can upgarde the machien to something more powerfull Oct 18 19:27:53 I guess it is never powerfull enough... Oct 18 19:28:01 I keep yo uupdated. thanks! Oct 18 19:32:04 jow: zh_Hans seems to be the new zh-cn, therefore I'd suggest to change that in luci.mk instead of making weblate use legacy stuff Oct 18 19:32:18 At least thats what I understand following https://github.com/encode/django-rest-framework/issues/2688 Oct 18 19:33:54 aparcar[m]: well then you also need to write Lua code to remap incoming requests Oct 18 19:34:11 as I already explained, its not about the paths, its about mapping Accept-Language Oct 18 19:34:50 so browsers did not adapt this new style? Oct 18 19:36:07 https://addons.mozilla.org/en-US/firefox/addon/chinese-simplified-zh-cn-la/ Oct 18 19:37:13 https://stackoverflow.com/questions/14257598/what-are-language-codes-in-chromes-implementation-of-the-html5-speech-recogniti Oct 18 19:39:46 https://github.com/chromium/chromium/tree/2ca8c5037021c9d2ecc00b787d58a31ed8fc8bcb/components/strings Oct 18 19:40:06 so no, it seems its still zh-{cn,tw,sg,hk} Oct 18 19:40:41 somewhere in luci some path remapping logic is needed if weblate cannot be convinced to use other codes Oct 18 19:41:59 I guess it can but simply does not show zh-cn and zh_Hans at the same time. I'll try to find other project with zh-cn working and if so, move all zh_Hans file over to zh-cn in a PR Oct 18 19:42:19 thank you! Oct 18 20:09:48 "fixed", I'll create an issue upstream that newly created files don't follow the applied schema Oct 18 20:53:09 jow: https://github.com/openwrt/luci/pull/3209#issuecomment-543940484 Oct 18 23:13:34 anybody working on 11ax APs? Oct 18 23:14:33 I played with an engenius EW357AP, and interestingly, in the interface, you could see 'luci' Oct 18 23:14:56 so I would assume it might be somewhat "easy" to get openwrt on it Oct 18 23:15:19 EW377AP should be the same, it's just a 4x4 (357 is 2x2) Oct 18 23:19:20 it works fine as standalone AP if anybody's wondering, no need for engenius wireless managed switch Oct 19 00:37:15 Mister_X: ipq807x support in mainline is rather minimal Oct 19 00:49:52 Mister_X: blogıc was working on ipq807x/ ath11k a while ago, but I have no idea about the current state **** ENDING LOGGING AT Sat Oct 19 02:59:57 2019