**** BEGIN LOGGING AT Thu Nov 21 02:59:57 2019 Nov 21 07:36:09 ynezz: if you have a moment for me it would be appreciated Nov 21 07:36:28 would like a quick hand hold to get teh gitlab-ci setup for procd, fstools, ... Nov 21 09:46:07 xback: just a heads up - really wasn't sure what to do about cns3xxx/patches-4.19/080-sata_support.patch in 4.19.85 Nov 21 09:46:36 ldir: thanks. currently doing the bump here Nov 21 09:47:50 I hope it's an easier bump this time .. last one took a lot of work Nov 21 09:48:53 other than that it appears to be straightforward, with 3 'reverse-applied' deletions and just a refresh - the cns thing though is a bit WTH? to me Nov 21 09:49:16 xback: was that due to Eric's hash changes ? Nov 21 09:52:18 xback: I'm talking to you via an x86 4.19.85 router :-) Nov 21 10:08:22 I'm trying to build cifsd-tools on macos which has various dependencies, glib2 is failing https://pastebin.com/XbEzvR2N Assistance appreciated. Nov 21 10:10:01 it would be good if at least one of the SMB providing packages built under macos, but sadly neither samba4 nor cifs do (obviously the kernel bit is ok but user space tools are wonky) Nov 21 10:13:03 ldir: yeah also. the hash change from Eric, combined with a lot of patch editing, new symbols, reverse applieds, and a lot of testing. and even then I slipped and missed a brace .. Nov 21 10:15:12 multiplied by 3 branches .. Nov 21 10:19:30 see, this is why I dabble at it but trust the heavy lifting to an expert Nov 21 10:20:51 ldir: can't comment if that's the most wise decision ;-) Nov 21 10:21:58 morning :) Nov 21 10:23:10 stintel: o/ Nov 21 10:24:41 trying some mac80211 reverts here to see if that would improve my RPi0W wifi problems Nov 21 11:04:47 wtf Nov 21 11:04:51 https://fsfe.org/news/2019/news-20191120-01.html Nov 21 11:06:08 how is anything other than NTP at location A even up for discussion Nov 21 11:09:17 stintel: it's not but reality now is that at least in PL, ISPs are using B approach Nov 21 11:10:33 you get a black box you can't touch, access, modify its settings, exchange etc., that applies mostly for (but not only) DOCSIS based services Nov 21 11:11:17 pepe2k: in Belgium, major ISPs are giving a router/modem combination with very little control (you might be able to change the internal subnet and maybe configure some firewall rules) Nov 21 11:11:32 mostly configured via a public website of the ISP Nov 21 11:12:21 you could request Telenet to install a modem instead of a router and then your router would get the public IP, but I've heard it's getting increasingly hard to convince them to do so Nov 21 11:14:12 oh well I have a GPON "router" here as well, by default you get RFC1918 IP when you connect to it, but my both OpenWrt routers have their MAC registered at ISP and a static public IP linked to them Nov 21 11:14:30 not great, not terrible :P Nov 21 11:18:54 interesting build failure. suddenly started getting `make: getcwd: No such file or directory` and stuff like that Nov 21 11:32:32 stintel: disk full? Nov 21 11:32:41 rm -rf gone crazy? Nov 21 11:37:11 /dev/sda2 466G 141G 324G 31% / Nov 21 11:37:35 probably something went wrong in my build pipeline Nov 21 11:57:03 ha haaa! Have 'solved' it. Or worked around it at least. Nov 21 11:59:10 glib2 build is looking for 'msgfmt' and possibly other commands. Although I have brew's gettext installed, this is 'keg only', in other words it's not linked to /usr/local/bin. Nov 21 12:01:45 actually the needed commands are in /usr/local/opt/gettext/bin - so I've added that dir to my path. I wonder if there's a better way. Nov 21 13:10:47 quick luci (JavaScript) question: for the form.DummyValue whats the difference between cfgvalue vs textvalue ? Nov 21 13:11:48 ah and whats this modalonly for? Nov 21 13:12:22 ldir: bumps pushed to staging. I'm pretty sure these will be final Nov 21 13:15:50 andy2244: "it is used by TableSections and GridSections" Nov 21 13:20:25 I don't know much about mvebu, but can anyone explain to me why we need arm-trusted-firmware-mvebu in the master tree? Nov 21 13:22:39 It seems like nothing depends on it, it looks like kitchen sink, downloads/uses another ARM compiler, which probably works only on Linux Nov 21 13:25:14 xback: looks much more sensible than what I did :-) (re)-Building now Nov 21 13:27:45 andy2244: GridSections render a text representation of each section option in a table row. The textvalue function is responsible for formatting a cfgvalue as textual represnetation (e.g. transformin 0/1 into "yes", "no") Nov 21 13:28:34 andy2244: .modealonly controls whether an option appears in the row of a gridsection only (false), only in the Edit... popup (true) or both (undefined / null) Nov 21 13:29:15 if you don't use GridSections, nether .textvalue() nor .modalonly apply Nov 21 13:35:21 jow: if I want to add buttons and things to a cbi form js, are tehre any existings ones you can think of? do I have to do the E() stuff to create everything I might need, or are there shortcuts? Nov 21 13:35:53 if `ifstatus lan` returns errors: subsystem: interface, code: NO_DEVICE ... what could be the problem? Nov 21 13:37:08 (like ddns/overview uses a DummyValue with a cfgvalue that returns a full dom fragment. Nov 21 13:38:01 this is on a RPi0W. could it be because it takes too long for wpa_supplicant to associate due to low entropy? that udhcpc gives up? Nov 21 13:39:26 kind of looks like it. so wifi is fine but some netifd change might be the culprit instead? Nov 21 13:40:46 stintel: using urngd? Nov 21 13:41:15 karlp: there is form.Button Nov 21 13:41:49 mangix: urngd does not work on pi0 Nov 21 13:42:28 and it doesn't even matter, if for any reason associate takes long, it should not abort dhcp requests Nov 21 13:43:58 stintel: which protocol? Static? Nov 21 13:44:08 erm, nvm - dhcp Nov 21 13:44:47 guess I'll be bisecting again Nov 21 13:45:03 I don't recall having this problem on my previous images Nov 21 13:49:12 jow: ok, that's simpler :) any ideas on how to make a click handler pop a file download? I should probably just link it into the docroot somewhere maybe instead... Nov 21 13:51:24 karlp: you can post to /cgi-bin/cgi-download Nov 21 13:51:25 * karlp looks at datauris Nov 21 13:51:27 example https://github.com/openwrt/luci/blob/86f492173d1daab8b75e4d4bd70ba06872ea70fa/modules/luci-mod-system/htdocs/luci-static/resources/view/system/flash.js#L168 Nov 21 13:51:43 * stintel randomly builds a03219ba09a55ad49926e5c2d60ddff095fe5096 Nov 21 13:51:59 karlp: a less clunky method is L.post() w/ new FormData() as content Nov 21 13:52:51 nvm, that won't work as it does not initiate a file save dialog but provides the download date inline to js code Nov 21 13:52:58 should probably enable ccache on my pipeline :P Nov 21 13:53:16 stintel: for some reason ccache has little impact on the build system Nov 21 13:53:18 jow: yeah, I'm on "modern browsers only" so I think data uri should work. Nov 21 13:53:30 file is only small anyway Nov 21 13:53:38 jow: really? Nov 21 13:54:03 oh well, I can just wait, I can start a workflow with custom hash, it will auto-clone, build, scp to test device, sysupgrade Nov 21 13:54:48 stintel: yeah, it has an absymal default hit rate (~20%-30%) - if you tune it to use a less restrictive compiler detection (gcc -dumpmachine) it goes up to 70% but the overall build time seems not really affected Nov 21 13:55:09 never got around to debug it, in the end I just ripped it out everywhere as it was causing headaches anyway Nov 21 13:56:22 hmmm looks like it does not take my git hash parameter Nov 21 14:02:24 hmz. looks like wpa_supplicant in 19.07 could have a memleak. running for a week on a vessel suddenly shows 5x increase in ram usage Nov 21 14:02:37 valgrind to the rescue .. Nov 21 14:03:10 hauke: ^^ Nov 21 14:11:16 jow: ok thanks Nov 21 14:21:12 ok, so this makes a nice download, but a) the content promise hasn't been resolved, and b) is there a way of getting the value of another option in the js? https://zerobin.net/?b3c82fd0762b2122#4GODpTvbBwHJjrIo7D4XE68W+e6OeEHyf2zbE34FFpY= Nov 21 14:23:28 otheroption.cfgvalue(sectionid) works well enough for b). Nov 21 14:26:33 nvm, figured it out. justmore wrapping: https://zerobin.net/?a1b5a757c790b236#BkzSexLDaUU2FRj3mlLO87ucxEkTo2j92WlYQL4ZoJU= Nov 21 14:29:47 this is really nice now. super js client speedy :) Nov 21 14:47:13 Is there any documention for setting up IPv6 SLAAC? Nov 21 14:48:27 I mean I'm in a environment that upstream is using SLAAC, how can I make devices lan access to IPv6? Nov 21 14:50:29 * How can I make my lan devices access to IPv6? Nov 21 14:52:16 ..you need generally to have a method of IPv6 from upstream (aka ISP), or tunnel of some sort... SLAAC is then a mehtod to distribute teh address space to your clients Nov 21 14:56:40 But it's using SLAAC... Nov 21 14:57:05 jow: fyi, that random commit a03219ba09a55ad49926e5c2d60ddff095fe5096 does not have the DHCP problem Nov 21 14:58:08 bisecting now, doing the actual `git bisect` locally and build + deploytest via my CDS pipeline by passing the hash git bisect suggests as next Nov 21 14:58:42 maybe I can one day fully automate bisecting in the pipeline simply by giving it good and bad hash Nov 21 14:58:52 that would be an amazing feature Nov 21 15:01:06 ysc3839: ask your upstream, it will be much easier if they use DHCPv6-PD like everybody else Nov 21 15:03:26 ysc3839: but according to https://openwrt.org/docs/guide-user/network/ipv6/start#native_ipv6_connection it should still work out-of-the-box Nov 21 15:04:45 otherwise try 'option ra relay' and so on, further down in the page Nov 21 15:09:33 zorun: It doesn't work out-of-box. LAN devices get private IPv6 address. Nov 21 15:15:02 I tested that relaying configuration, and it works now. Thank you! Nov 21 15:24:53 karlp: btw fs.read() may fail Nov 21 15:26:35 karlp: either use something like return fs.read(...).then(download...).catch(function(err) { }) Nov 21 15:27:13 karlp: or fs.trimmed() or L.resolveDefault(fs.read(...), 'default-value').then(function(content) { ... }) Nov 21 15:34:50 that resolveDefaults looks interesting... is that promise result or fallback, also returning a promise? Nov 21 15:46:01 is there a hook for when someone deletes an instance of typed section? Nov 21 15:50:38 and yar, that snippet failed whenI tried to make a new section, and things haven't been selected yet :) Nov 21 16:00:07 catch can swallow it, but it needs to be notified of the change when the dependent field is finally selected. js is awesome. Nov 21 16:00:19 catch and swallow works if you just save the new section first. that will do Nov 21 18:14:34 xback: are you using mesh? Nov 21 18:14:41 and see your memleak? Nov 21 18:19:30 stintel: I plan to create a backports 5.4-rc8 release by today and then will also send a patch to the openwrt mailing list, if there is still enought time I will also create a release based on the most recent 4.19 branch for openwrt 19.07 Nov 21 18:25:00 Hauke: nice stuff Nov 21 19:09:53 Hauke: do you plan to push 4.19 based backports before rc2? Nov 21 19:15:10 pepe2k: depends if rx2 will really be tomorrow ;-) Nov 21 19:16:36 got it :) Nov 21 21:20:31 Hauke: pepe2k: are we talking about mac80211 backports? cause i'm not finding any in hauke's tree and 4.19.79 backports are already in 19.07 no? Nov 21 21:42:43 there is a bunch of fixes since v4.19.79 Nov 21 21:51:43 ok. i'll grab them when i'll see them then, so far i just saw a few mac80211 commits (the one xback pushed to the 19.07 tree is the latest one i'm seeing) Nov 21 22:21:40 whats the difference between ath10 and ath10-ct Nov 21 22:23:15 plain is the "OEM" one, -ct is improved by greaarb (or something like that), I take there is his workplace involved and all sorts of jazz with contracts etc Nov 21 22:24:17 well.. this has also short explanation https://github.com/greearb/ath10k-ct Nov 21 22:24:37 and person was greearb, so almost correct 😁 often times here too Nov 21 22:46:25 pepe2k: https://patchwork.ozlabs.org/patch/1199146/ Nov 21 22:54:24 what firmware would i load for a QCA9890 Nov 22 01:10:34 ynezz: I'm not very familiar with c static code checks but is this of value or boring? https://paste.sr.ht/%7Eaparcar/35c050d34730b3df297a8fbe7f378fff21ff521e it uses flawfinder Nov 22 01:36:57 looks like the dynamic wifi reconfig stuff is related to my problem on rpi0w Nov 22 01:37:04 Bisecting: 1 revision left to test after this (roughly 1 step) Nov 22 01:37:05 [155ede4f1fdb192b11f8ae2dbffda5f7ef4903bd] netifd: add dynamic wireless reconfiguration Nov 22 01:50:41 blogic: on RPi0W, it takes ~2m uptime for the CRNG to have enough entropy. due to that, the wifi is only able to connect after ~2m. the above commit causes `ifstatus lan` to return errors: subsystem: interface, code: NO_DEVICE in this case Nov 22 01:51:17 blogic: do you feel like looking into it, as these netifd changes are based on your work, or do you prefer I post this to the ML and Daniel Nov 22 01:52:44 or is Daniel Golle hanging around on IRC by chance :) Nov 22 01:55:08 `/names Nov 22 01:55:14 [Tue 2018-10-02 05:18:20 PM PDT] * dangole has quit (Ping timeout: 256 seconds) Nov 22 01:56:06 [Thu 2019-11-21 05:55:56 PM PST] -NickServ- Last seen : (about 7 weeks ago) Nov 22 02:27:07 hmmm wait it actually looks like it's the next commit 60fb4c92b6b0d1582d31e02167b90b424185f3a2 **** ENDING LOGGING AT Fri Nov 22 02:59:57 2019