**** BEGIN LOGGING AT Tue Nov 06 03:00:00 2018 **** BEGIN LOGGING AT Tue Nov 06 03:25:32 2018 Nov 06 05:39:48 https://bugs.openwrt.org/index.php?do=details&task_id=1618 Nov 06 07:05:03 stintel: i thought you already were Nov 06 08:00:14 does anyone has worked with juci? Nov 06 08:00:40 https://github.com/mkschreder/juci Nov 06 08:30:28 is that safe to use the same "struct ubus_context" for registering objects & also invoking other objects? Nov 06 08:57:23 blogic: ;) Nov 06 08:57:27 fair enough Nov 06 08:59:58 rmilecki: yes Nov 06 09:00:15 rmilecki: you should only ever have 1 single ctx Nov 06 09:00:24 rmilecki: look inside ustatusd Nov 06 09:00:32 it covers all types of ubus api Nov 06 09:00:44 incl send/recv a/sync notifications Nov 06 09:01:47 thx Nov 06 09:14:26 our office network completely died, and it wasn't the openwrt firewall \o/ Nov 06 09:16:47 build #416 of ar71xx/tiny is complete: Success [build successful] Build details are at http://phase1.builds.lede-project.org/builders/ar71xx%2Ftiny/builds/416 Nov 06 09:18:03 aaah ar71xx/tiny finally fixed \o/ Nov 06 09:22:53 stintel: do you mean to imply that it is that unreliable? ;) Nov 06 09:23:46 feckert: I don't think so. I looked a few years back and the main problem was that it requires various userland changes Nov 06 09:24:45 xback: fpu emulation is intentionally disabled on ar71xx due to size reasons Nov 06 09:25:02 jow: thanks. was just looking at the issue Nov 06 09:25:16 i admit I also enable it in my personal builds to avoid this issue Nov 06 09:26:33 jow: is the size issue still valid now that nearly all targets use dynamic partitioning? can you roughly estimate the size change? Nov 06 09:26:56 it is a matter of principle Nov 06 09:27:23 ok. noted Nov 06 09:27:27 just a little feature enabled here, another one enabled there and suddenly another flash block becomes unavailable Nov 06 09:27:59 jow: there have been some problems with it before, but since I pushed https://git.openwrt.org/ef025e64 Nov 06 09:28:17 I don't really recall any figures but I *think* the emulator added about 100k to the uncompressed kernel image Nov 06 09:28:23 might be totally wrong though Nov 06 09:28:43 stintel: ah firewall as in appliance, not as in iptables ruleset Nov 06 09:28:48 jow: I can run a build to verify it Nov 06 09:28:58 jow: yes Nov 06 09:33:44 jow: purely intended as a question ensuring i'm exchanging the OpenWrt vision on this :) : does the *principle* argument still hold when it leads to Illegal Instruction? Nov 06 09:48:13 xback: in this case its a packaging bug, packages should be built with software flow emulation Nov 06 09:48:29 yes I know, node.js is special since it uses machine specific asm Nov 06 09:48:38 in this case its simply unsupported Nov 06 09:49:06 yate can be probably fixed to do proper softfloat Nov 06 09:49:29 right, the only people who should ever really run into this are people who are explicitly tryign to link with hardfloat or hardwritten asm right? Nov 06 09:49:46 thats my take on it, yes Nov 06 09:50:27 my general reasoning is: I do not want to sacrafice precious flash space for all supported boards/targets just to accomodate for packages which aren't suitable for this platform to begin with Nov 06 09:51:05 thats like enabling the v4l subsystem in the ar71xx kernel just because someone might decide to attach an usb dvb dongle Nov 06 09:51:26 * karlp nods Nov 06 09:51:32 this is my opinion though, if the other devs think differently I won't stand in the way Nov 06 09:51:34 leave that to customizers if they like, Nov 06 09:52:40 I believe someone tried nodejs with enabled fpu emulation and it was unusably slow Nov 06 09:52:59 jow: i am playing around to get this working Nov 06 09:53:27 feckert: what... juci? Nov 06 09:54:14 jow: yes Nov 06 09:54:31 ok Nov 06 09:54:36 good luck :) Nov 06 09:55:06 thanks Nov 06 09:56:18 you are not so convinced? did you made bad experience Nov 06 09:56:35 its a mixture of different things Nov 06 09:57:04 its past maintainer insulted me Nov 06 09:57:11 the he took apache code and put it under gpl 3 Nov 06 09:57:27 the ui itself also had a larger footprint in the end Nov 06 09:57:43 the only remaining upside was the flashiness factor Nov 06 09:58:23 the fact that there's been another not-upstreamable ui in the end annoyed me, so I didn't look further into it Nov 06 10:00:08 The first appearance was good if you read the git page but there are no more commits since a year Nov 06 10:02:08 jow: thanks for elaborating on the FP EMU issue. appreciated! Nov 06 10:03:10 i'm building 2 identicals having 1 with FP EMU enabled to get some figures Nov 06 10:03:53 xback: what other package is it? node.js and? Nov 06 10:04:25 karlp: https://bugs.openwrt.org/index.php?do=details&task_id=1928&order=id&sort=desc Nov 06 10:04:59 and a copy for mt7621: https://bugs.openwrt.org/index.php?do=details&task_id=1929&order=id&sort=desc Nov 06 10:09:01 so node and yate Nov 06 10:09:32 node's known, and just is "unsupported in standard builds" Nov 06 10:19:57 jow: purely for info Nov 06 10:20:34 Uncompressed: +70,4KiB lzma Compressed: +21kiB Nov 06 10:21:03 thats indeed a lot for a niche public Nov 06 10:39:06 nbd: do you ack on this one? https://git.openwrt.org/?p=openwrt/staging/xback.git;a=commit;h=4acb406f72592a6eac1fc35221bf7fedc1f42153 Nov 06 10:40:11 yes Nov 06 10:40:29 I would also backport it to 18.06 Nov 06 10:40:35 thank you :) Nov 06 10:40:37 fine with me Nov 06 11:07:04 xback: we should probably reach out to jiri to see if the yate issue can be fixed. I guess it uses some kind of special asm which trips up here Nov 06 11:07:19 maybe its not even related to yate but a problem with one of the sound codec libraries Nov 06 11:35:01 jow: ok i'll ping him as soon as I have some time for it Nov 06 14:44:40 jow: what is the procedure for adding some buildslaves? Nov 06 14:44:54 I notice that there are only 3 slaves left which struggle to cope with the workload Nov 06 14:45:59 i'm willing to let my buildmachine help a bit when it's idling and if it's not too complicated (or require a VM) to install it as a slave Nov 06 14:46:25 there was an offer of powerful build servers from https://cfarm.tetaneutral.net/ Nov 06 14:48:05 oh actually the machines have already been installed, gcc125 & gcc126 there: https://cfarm.tetaneutral.net/machines/list/ Nov 06 14:48:22 =) Nov 06 14:48:34 I think they just need to be setup and added to buildbot Nov 06 14:51:23 I LOVE OpenWRT Nov 06 14:51:39 Hauke: do you think you can pull this patch in openwrt-18.06? https://patchwork.ozlabs.org/patch/985756/ Nov 06 14:52:09 zorun: thanks for the info :) didn't know that Nov 06 14:52:16 Hauke: otherwise, the next point release of 18.06 will not support most recent mikrotik boards Nov 06 14:52:28 Hauke: mikrotik seems to be using the new flash chips for many different models Nov 06 14:55:02 xback: there was a thread on openwrt-adm some weeks ago (also and unrelated, I'm involved with running the compile farm ;) ) Nov 06 14:55:45 nice nice :) exciting times here Nov 06 14:56:13 It's a very generous offer also Nov 06 15:01:34 Hello! Im sorry. I need answer about firmware version for TP-Link Archer C20 v4.1 device https://image.ibb.co/cofwyV/20181106-153107.jpg Nov 06 15:02:06 i find firmware v4 for this device Nov 06 15:02:21 its correct firmvare or not? Nov 06 15:06:01 xback: yes, these new machines are quite awesome Nov 06 15:06:35 OSUOSL received quite a lot of them Nov 06 18:55:01 pffft. procd is being a major pita again Nov 06 18:55:26 it refuses to start my new daemon Nov 06 18:58:39 https://pastebin.com/LH3XzTrS Nov 06 18:59:06 the last command doesn't crash and is effectively writing debug messages to syslog Nov 06 18:59:13 so it definitely works Nov 06 18:59:58 apparently this happened during boot: Tue Nov 6 18:53:03 2018 daemon.info procd: Instance gpio2mqttd::instance1 s in a crash loop 6 crashes, 1 seconds since last crash Nov 06 19:00:05 how does one reset that ? Nov 06 19:02:58 stintel: manula stop/start of the init.d Nov 06 19:03:03 aka a restart Nov 06 19:03:26 stintel: basically procd noticed, that the daemon stopped less then 30s after starting it Nov 06 19:03:32 it then waits 5s and respawns Nov 06 19:03:42 if that happens 5 times in a row it wont retrt Nov 06 19:05:12 hmmm, when I add strace in the init script and write to a file, it ends after: poll([{fd=4, events=POLLIN|POLLPRI}, {fd=0, events=POLLIN|POLLPRI}], 2, 1000) = 2 ([{fd=4, revents=POLLIN}, {fd=0, revents=POLLIN}]) Nov 06 19:09:00 stintel: the daemon *is* staying in foreground right? ie. it doesn't fork into background itself? Nov 06 19:09:45 when I run it manually it just keeps running, not returned back to the prompt Nov 06 19:10:13 ok, ruled out an obvious silly :-) Nov 06 19:10:17 feckert: a difefernt version of Juci can be found here: https://github.com/prplfoundation/prpl-feed/blob/demo/juci/Makefile Nov 06 19:10:35 Hauke: go wash your hands Nov 06 19:10:36 ldir: I don't even know how to write a real daemon that actually forks :P Nov 06 19:10:51 stintel: daemon() ? Nov 06 19:11:02 stintel: maybe i tries to poll the stdin/out Nov 06 19:11:11 there was a way to fiddle with those in procd Nov 06 19:11:19 but dont ask me how .... Nov 06 19:12:08 do you mean "procd_set_param stdout 1" or something different ? Nov 06 19:12:21 i wonder what that does Nov 06 19:12:25 try it out :-) Nov 06 19:12:46 blogic: Hauke will never get the stain out lol :-) Nov 06 19:12:57 ldir: :-) Nov 06 19:13:10 blogic: already did, doesn't help Nov 06 19:13:45 afaik that is just to write stdout (or stderr) to logd Nov 06 19:13:51 ah ok Nov 06 19:14:38 * ldir wishes he'd remembered about that a few days ago when he was fighting with znc startup problems Nov 06 19:14:42 oh that looks pretty, sadface Nov 06 19:14:56 zorun: xback: The patch adding the new flash chip used by mikrotik boards looks good for 18.06, xback will you take it or should I integarte it? Nov 06 19:16:07 blogic: Juci is the future of the coperate UIs ;-) Nov 06 19:16:13 haha Nov 06 19:16:22 blogic: here prplWrt for you: https://github.com/prplfoundation/prplwrt Nov 06 19:16:24 it does look pretty fancy tbh Nov 06 19:35:42 This prplwrt is what people were talking about then! :-D Nov 06 20:15:32 blogic: do we still accept patches for ar71xx? or do we require new board to be ath79? Nov 06 20:20:19 so the code was closing a file descriptor which made procd terminate the process Nov 06 20:20:59 lynxis: ar71xx is still accepted, but as far as I can see, no body is going to touch ar71xx any more Nov 06 20:21:07 lynxis: riding dead hourse and so on Nov 06 20:21:33 you say that, but there are so many boards still being sold and used that are supported by ar71xx :( Nov 06 20:21:52 jwh: but they should be supported by ath79 too Nov 06 20:21:58 sure Nov 06 20:22:01 *horse Nov 06 20:22:05 teach me how to port them :D Nov 06 20:23:08 jwh: most boards doesn't differ so much. usually you need to know: which soc has been soldered? how many ports does it use? which gpios are connected to the leds? .. Nov 06 20:23:10 ath79 would be nice for sure, but it looks like a pretty difficult task to port unless you're already well versed in doing it Nov 06 20:23:22 jwh: everything is basicly read out of the mach-$board.c file. Nov 06 20:23:45 jwh: i think a good start is to look on a supported ath79 board. copy the files and modify according to the mach file Nov 06 20:24:38 yeah I tried that, don't understand DTS either so it's basically a no-go Nov 06 20:25:53 it's one of those I guess, once it clicks and makes sense it's not so bad Nov 06 20:31:49 it looks rather easy if you know the DTS 'lingo' i assume Nov 06 20:33:39 yeah Nov 06 20:34:01 also requires at least rudimentary knowledge about hardware, memory addresses etc Nov 06 20:34:05 none of which I have :D Nov 06 20:34:06 hey Borromini, everything still running fine in your test? Nov 06 20:36:52 nbd: hey. all looks fine so far :) Nov 06 20:37:09 i will keep you posted :-) Nov 06 20:37:26 so far 25h+ uptime Nov 06 20:37:36 didn't get there before Nov 06 21:12:55 anyone know offhand what I have to enable to get an arduino device to load w/usb? Nov 06 21:18:26 ah, USB FTDI driver it seems Nov 06 21:38:58 Hauke: ok, thanks! Nov 06 21:41:04 oohhhhhhhh Nov 06 21:41:06 https://up-board.org/upsquared/specifications/ Nov 06 21:41:17 gonna get me one of those Nov 06 21:42:09 stintel: yeah, procd debug is "here's some json, it doesn't mean much, but it migh thelp debug this particuarl facet of your script" Nov 06 21:54:27 jwh, in case you put a 9984 NIC in it and it works, I'm interested to know :) Nov 06 21:54:39 oh, no wifi for me :D Nov 06 21:54:49 but theres no spare pcie lanes anyway Nov 06 21:54:50 oh, so you want stability...oh well :P Nov 06 21:54:58 and thus no mini pcie Nov 06 21:55:04 usb or bust Nov 06 21:55:07 it had a mini-pcie connector Nov 06 21:55:10 oh? Nov 06 21:55:11 hm Nov 06 22:01:18 yay, wrote a small tool that polls a GPIO pin on my RPi0W, connected to a motion sensor, events are published to mosquitto, and Domoticz reads that event to toggle the state of a dummy motion sensor Nov 06 22:01:54 greearb_: thats interesting, I thought apollo lake, like cherry trail, was severely restricted in lanes Nov 06 22:02:43 could possibly be USB only I guess, but that board doesn't have a lot of ethernet, so probably there are a few lanes available? Nov 06 22:03:05 2 ports, so in theory 2 lanes Nov 06 22:03:16 unless its sharing a single lane, depending on the nic Nov 06 22:04:11 apollo lake seems to be all round a more desktop/laptop orientated soc though, the older cherry trail stuff was almost exclusively tablets and had very little i/o Nov 06 22:09:17 pretty decent board actually, sata, m.2, mini pcie Nov 06 22:12:12 ok, I'm in a patch management jam again: I used the trick of {clean,prepare} in the kernel, then I applied it to a git tree, then I made my changes, and created a patch, and put that patch back in openwrt hack-4.14 dir... Nov 06 22:12:24 but now it won't build because it tries to re-apply the patch Nov 06 22:19:35 If you're going that route, you have to revert the patch from that git tree Nov 06 22:20:06 A git reset --hard ; might help also Nov 06 22:20:08 I think something already purged the tree Nov 06 22:20:22 like no .git in the kernel dir Nov 06 22:21:02 If you did the clean target, probably the buildsystem removed the whole directory Nov 06 22:27:51 ok, got it back to functional Nov 06 22:28:27 and yes, the build purges the git rep, at least when you add a new patch Nov 06 22:48:08 . Nov 06 22:48:33 lynxis: for avg fols ar71xx, for you ath79 :-) Nov 06 22:49:13 Hauke: now go also go wash your mouth Nov 06 22:50:02 Hauke: cant't believe you are "he who shall not be named"'ing it ... Nov 06 22:50:14 Hauke: srsly ? Nov 06 23:00:54 ohhhh Nov 06 23:00:55 https://forum.odroid.com/viewtopic.php?f=29&t=32536 Nov 06 23:01:03 forget the up board, odroid ftw Nov 06 23:02:09 stintel: you still have any quirks/issues with mosq/similar? anything else I can make more full of awesome? Nov 06 23:03:42 (have you seen lua-mosquitto? we put it together entirely to make it easier to do openwrt-ish things with mqtt) Nov 06 23:03:47 jwh: are there adaptors to splitthe 4 NVMe PCIe lanes into two normal mini-PCIe slots for wireless cards? if not (and I didn't really find a conclusive answer to that), it won't be a functional /wireless/ router Nov 06 23:03:58 karlp: I should test if that missing dependency issue with libmosquittopp is still there Nov 06 23:04:03 karlp: haven't seen lua-mosquitto Nov 06 23:04:05 pkgadd: on the up board? Nov 06 23:04:15 it looks like it has seperately wired lanes for mini pcie and the m.2 slot Nov 06 23:04:28 jwh: that's the big problem I've found with pretty much any x86 solution smaller than µATX (and within reasonable price ranges) Nov 06 23:04:44 but the diagram isn't that detailed Nov 06 23:05:04 if you mean the odroid, I doubt it will work, they don't sell routers Nov 06 23:05:07 heh Nov 06 23:05:12 don't know why, but I still expect full functionality from any 'router' (in a wide definition of the word) Nov 06 23:05:36 I'm still of the adage that routers shuold route and APs should radio Nov 06 23:05:38 yeah, I'm referring to the odroid h3/ x86 board you linked to Nov 06 23:05:40 should* Nov 06 23:05:52 I don't like this combined thing Nov 06 23:07:05 for home uses I very much like the combination - mostly to have the basics going with one device (plus external modem) Nov 06 23:07:09 depends how they've designed it I guess, might be doable Nov 06 23:07:21 yeah I guess Nov 06 23:07:33 but I wire my places with unifi aps anyway, so perhaps not a typical home user Nov 06 23:07:51 after experiencing how terrible typical home user internet access is again recently, do not want Nov 06 23:07:51 I do use WDS/ 4addr repeaters and additional gear, but I like to keep that (semi-) optional Nov 06 23:09:59 stintel: it still has potentially odd areas if you use coroutines a lot or something, but I'v ereplaced a lot of c code with lua and lua-mosquitto. https://openwrt.org/packages/pkgdata/lua-mosquitto Nov 06 23:10:12 pkgadd: ah Nov 06 23:10:24 (just mentioning it as something that might be easy/ier to use in some environments) Nov 06 23:10:49 (why domoticz has to be this great bit blob of c++mud I'll never truly understand, but ok..) :) Nov 06 23:11:02 karlp: but what's the alternative :) Nov 06 23:11:10 have you seen homeassistant.io? :P Nov 06 23:11:16 "hass" :) Nov 06 23:11:19 yeah, in passing Nov 06 23:11:21 nothing more Nov 06 23:11:27 that's multiple gigabytes of all kinds of stuff Nov 06 23:11:44 I don't have an alternative, I'm just often surprised by how shit most of this shit is. Nov 06 23:11:54 can't argue with that Nov 06 23:12:12 I've been running hass for some time, as a docker container on my workstation Nov 06 23:12:15 it's huge Nov 06 23:12:15 I'm quite happy with what we've done as a producer, but man, the consumer/actor side seems to be just complete garbage Nov 06 23:16:25 karlp: you suggest doing that gpio2mqtt bridge thing might be easier in Lua ? Nov 06 23:16:49 Hello. So I've noticed that OpenWRT is populating my /etc/modules.d with pretty much every module I've built. I know that uci replaces udev, but does it have support (or is it planned) for automatic module loading upon need? Nov 06 23:17:21 oh, I should improve my VPN redundancy :> Nov 06 23:17:41 reboot main router and Kodi (streaming from my VDR box in Belgium) hangs Nov 06 23:18:45 stintel: I've not actually looked at your gpio2mqtt more than the name, just mentioning it as something thatmight be nicert to use than C, an dsuitable for openwrt. Nov 06 23:23:46 anyone see wlanX called wlanphyX instead on boot sometimes? Nov 06 23:25:42 could that possibly be, (on a limb maybe, but shit's weird) be related to https://github.com/openwrt/packages/issues/7309 ? Nov 06 23:27:56 er, probably not, I'd guess. At start of my init.d startup script, wlan0 exists, but not wlan1. I then sleep for up to 30 seconds or so waiting for wlan1 to exist. When that timeout happens, something has renamed wlan0 to wlanwiphy0 (I call my phy devices wiphyX) Nov 06 23:28:37 * karlp walks away, out of my depth sorry :) Nov 06 23:28:51 I also created /etc/udev.d/* files like what would work on Fedora, but I have a manual script that does the rename. The names are messed up before I run the rename script as best as I can tell Nov 06 23:36:37 oh dnscrypt-proxy was rewritten in golang. guess I should look at unbound then Nov 06 23:36:52 * karlp laughs Nov 06 23:37:08 stintel: do yo have any ideas on that snmpd ifindex failure? Nov 06 23:37:36 karlp: hmmm? Nov 06 23:37:45 https://github.com/openwrt/packages/issues/7309 Nov 06 23:38:07 looks reallllllly weird, but I don't use that part of snmp. Nov 06 23:40:33 I would have to investigate, but it's getting late for that :P Nov 06 23:41:08 compltely understand. Nov 06 23:41:35 i'm only involved as a nominal user of net-snmp, Nov 06 23:42:33 down with snmp, json posts for all info! Nov 06 23:42:38 I can't wait Nov 06 23:42:59 so much more useful info from ubus Nov 06 23:43:02 sure, but what does that actually change? Nov 06 23:43:17 json-schema sure, Nov 06 23:43:19 well, from netifd I guess Nov 06 23:43:25 that's ~equvialent to mibs Nov 06 23:43:33 but actual functionally _differnet_ ? Nov 06 23:43:36 more info than the standard mibs describe Nov 06 23:43:50 hah, like anything's in "standard" mibs Nov 06 23:43:58 it's all in enterprise oids Nov 06 23:44:06 you think json makes that better Nov 06 23:44:11 well as far as the almost universally standard interface mib goes Nov 06 23:44:16 no Nov 06 23:44:17 "just add a key, it' sbackward compatible" Nov 06 23:44:20 it could be xml Nov 06 23:44:24 but there is that Nov 06 23:44:45 my point was more than the info you get from ubus is much better than ootb net-snmp Nov 06 23:45:01 you could probably write a whole bunch of scripts to implement the missing bits as private oids, but why bother Nov 06 23:45:03 don't get me wrong, asn1 is a pig, but I'm really not convinced that changing the data representation to json will actyually make anything better Nov 06 23:45:15 jwh: except that ubus is entirely openwrt specific Nov 06 23:45:27 sure, but its an example of how to do it sensibly :D Nov 06 23:45:44 and... if I may say so, has absolutely _zero_ guarantees of stability Nov 06 23:45:50 theres no reason system couldn't provide similar Nov 06 23:45:53 systemd Nov 06 23:45:54 as an example Nov 06 23:46:11 it's a competely internal format that happens to be pleasant, as long as youre on the current top of things Nov 06 23:46:20 hang on, why is systemd invovled now?! Nov 06 23:46:22 oh yeah for sure Nov 06 23:46:40 or anything, I was pointing out its a sensible way of doing it as an example to follow Nov 06 23:46:43 heh Nov 06 23:47:06 also means you don't actually need to be able to reach devices if they can just send posts out Nov 06 23:47:48 sure, I feel you're perhps downplaying how useful it is that MIBs are actually, for all their limititations, fairly cross platform. Nov 06 23:47:58 try calling ubus blah on some other system... Nov 06 23:48:14 yeah, dealing intimately with snmp for many years does give you a jaded view tbf Nov 06 23:48:21 *anything* is nicer to deal with Nov 06 23:49:20 although thats largely the implementation/vendors fault rather than snmp itself Nov 06 23:49:41 yeah, I hear you Nov 06 23:50:00 but I still kinda feel like, "it's a world of shit, of course shit is going to be involved" Nov 06 23:50:09 haha Nov 06 23:50:22 admit defeat, pretty much? Nov 06 23:50:32 I mean, netconf/yang stuff all looks enormous and even _more_ unwieldy than snmp Nov 06 23:50:45 it is horrific Nov 06 23:50:47 genuinely Nov 06 23:50:55 but perhaps this si the difference bewtween snmp form the point of view of "I just want to get some metrics on the side" Nov 06 23:51:08 vs "I need a tool for managing my infrastructure" Nov 06 23:51:08 it's like enterprise didn't learn from any of heir mistakes and made them all again Nov 06 23:51:25 well yeah, thats what netconf/yang is, management more than metrics Nov 06 23:51:47 describe an interface in the most dramatic and descriptively abstract way possible Nov 06 23:51:50 snmp _could_ do the management, but yeah, I'd never want to do that... Nov 06 23:51:51 then turn it into a 2 line config Nov 06 23:51:59 in yaml, no less Nov 06 23:52:51 don't forget to indent properly! Nov 06 23:53:01 don't even start with fucking yaml Nov 06 23:53:09 do I need 2 spaces here or none? Nov 06 23:53:14 maybe, maybe now Nov 06 23:53:15 not* Nov 06 23:53:25 seemed like a good idea, but man, what a train wreck Nov 06 23:53:28 hmm, is there any documentation for kconfig.pl? Nov 06 23:53:31 yes Nov 06 23:53:45 ooh! Nov 06 23:53:47 if your parser isn't even part of python's standard library, you failed, move on. Nov 06 23:53:53 again, enterprise (mostly, anyway), making the same mistakes over and over Nov 06 23:54:12 jwh: nah, that's unfair, yaml predates heaps of enterprise shit Nov 06 23:54:25 not yaml specifically, but the whole thing Nov 06 23:54:40 it's hobbyists throwing shit together and making something inadvaertantly useful, and yaml is a sideaffect Nov 06 23:54:41 you could always use toml Nov 06 23:54:44 no-one planned yaml Nov 06 23:54:49 like yaml, but also json, but not really either Nov 06 23:54:55 vagrant used raw ruby files Nov 06 23:54:56 Does anybody believe in documentation? :) Nov 06 23:55:05 dansan: it's voodoo Nov 06 23:55:09 HAH! Nov 06 23:55:11 dansan: if you want kconfig docs, speak to lkml Nov 06 23:55:19 Have you guys seen the .dts validation patch set yet? Nov 06 23:55:24 it uses json and yaml Nov 06 23:55:24 and ask more nicely than you are wont to do. Nov 06 23:55:26 I think he means the perl script that generates some config Nov 06 23:55:41 my eyes, my eyes, etc Nov 06 23:55:42 if something uses json _and_ yaml, it's already on the wrong track. Nov 06 23:56:00 multi format input, does it also take xml? Nov 06 23:56:13 how about ini-like? Nov 06 23:56:13 we replaced yaml with json in some other projects becase json was standard in python, yaml was extra package download Nov 06 23:56:28 karlp: nope. This is an OpenWRT invention Nov 06 23:56:53 I know which script you're talking about Nov 06 23:57:00 I couldn't find any docs either Nov 06 23:57:17 karlp: Well I haven't dug through it deeply yet, but it may become the new standard for .dts Nov 06 23:57:24 or rather, this validation scheme Nov 06 23:57:58 DTDs all over again? Nov 06 23:58:11 or XSD? Nov 06 23:58:31 jwh: Yeah, a lot of programmers claim that their code is the documentation, which is about like claming that your sh** is a crop of food -- sure it can fertilize one, but there's a lot of work in between the two points Nov 06 23:58:36 dansan, again, you seem to be having a trouble with writing clear sentences. Nov 06 23:58:51 karlp: I'm having a luck day apparently! Nov 06 23:59:02 * karlp cheers and blows up balloons Nov 06 23:59:07 lmao!!!! Nov 06 23:59:12 better be helium Nov 06 23:59:21 are you trying to talk about a schema of some sort of .dts files? Nov 06 23:59:38 karlp: just a side note as I saw you guys discussing yaml Nov 07 00:00:22 Here's the thread: https://lkml.org/lkml/2018/10/5/883 Nov 07 00:00:43 How would be the time to make objections known :) Nov 07 00:02:30 heh, again, I feel like your sentence structure is letting you down. Nov 07 00:02:46 validation is completely desirable. Nov 07 00:02:47 lol! It wouldn't be the only thing Nov 07 00:02:53 Sure it is Nov 07 00:03:15 I would like to personally add an in-kernel validation that's enabled via some kernel hacking option Nov 07 00:03:25 those aren't the sort of things that are edited by humans Nov 07 00:03:35 those are very carefully curateed metadata. Nov 07 00:03:43 lol Nov 07 00:03:56 they fail now because it's not enforced by antyhing, it' sjust text Nov 07 00:03:57 And I have some land I want to sell you too Nov 07 00:04:04 exactly Nov 07 00:04:10 but they're not, and have never been, anything that users had to write Nov 07 00:04:16 unliked dts files themselves Nov 07 00:04:25 how do you define "user"? Nov 07 00:04:40 adding syntax and validation to th ebindings files? fuck yeah, why would anyone object to that Nov 07 00:04:51 Oh wait, what are you talking about? I thought we were talking about .dts and .dtsi files Nov 07 00:05:03 * karlp boggles Nov 07 00:05:12 Yeah, I'm excited about it, but it will be a lot of work Nov 07 00:05:14 you know that thread is only about the bindings docs right? Nov 07 00:05:21 it doesn't change the dts files _one bit_ Nov 07 00:05:31 No, it doesn't Nov 07 00:05:49 it adds a validation layer, but each driver's bindings docs will become a specification Nov 07 00:05:56 instead of free text as they are now Nov 07 00:06:10 which they were, but without any sort of automatic validation Nov 07 00:06:42 right. But what part were you saying is not edited by humans then? Nov 07 00:09:11 the dts files are not yaml. Nov 07 00:09:39 the bindings are proposed as ebing converted from arbitrary text to yaml. Nov 07 00:09:44 No, I think the bindings will become yaml Nov 07 00:09:48 yeah Nov 07 00:10:30 so, again, your formation of sentences is confusing. Nov 07 00:10:40 stintel: where is your gpio2mqtt code? Nov 07 00:11:02 another week another total raid failure, I really hate HP Nov 07 00:11:48 Oh, I see. Sorry I meant for dts validation :( Nov 07 00:12:04 karlp: yeah, I gotcha now Nov 07 00:12:32 jwh: which reminds me, I need to submit a patch marking raid5-cache as (EXPERIMENTAL) because *damn* it's screwed Nov 07 00:13:12 probably more reliable than this piece of crap sas controller Nov 07 00:13:28 disk having trouble? ah screw it, lets just block all i/o Nov 07 00:13:40 lol!! Nov 07 00:13:41 karlp: I only started writing it today, it contains hardcoded stuffs etc, not ready for public at all Nov 07 00:14:25 No, it has some serious flaws that only arise once you have an abnormal shutdown -- and then you can't assemble the array EVER again Nov 07 00:14:35 well, only read-only Nov 07 00:14:36 nice Nov 07 00:14:51 I'd settle for read only at this point, at least then I could tell the controller to drop the disk Nov 07 00:15:09 instead of scrambling to find my windows xp vm because the ilo sucks Nov 07 00:15:13 I'm sure they didn't intend it that way, but some Facebook engineers wrote it. I'm glad that they are sharing with the FOSS world, but uhh... Nov 07 00:15:36 geeze Nov 07 00:15:45 design systems to fail, only way to win Nov 07 00:16:02 Well if they do it gracefully... Nov 07 00:16:24 That's my goal each day. I know I'm going to fail, but I try to do it gracefully now :) Nov 07 00:16:33 stintel: np, just curious what you're putting in it :) Nov 07 00:17:01 karlp: ugly stuff :D Nov 07 00:17:30 I wanna port mqttwarn to lua, but it's way to big.... Nov 07 00:17:38 can't run it all on ar71xx.... Nov 07 00:21:26 dansan: it also has this cute trick where if it loses a disk, when it power cycles (because all i/o is wedged), it'll bring the broken disk online and destroy the array Nov 07 00:21:41 oh dear god Nov 07 00:21:58 have to hope you interrupt boot in time and stop it Nov 07 00:22:00 wth is this hardware!? Nov 07 00:22:09 HP "enterprise" sas controller Nov 07 00:22:14 omg Nov 07 00:22:21 I'd rather take my chances with windows raid Nov 07 00:22:43 eh, it's fine for enterprise, they have backups Nov 07 00:22:56 I've had good luck with Linux software raid, at least until I tried the raid5-cache thing Nov 07 00:22:56 backups are fine but the service is still down :D Nov 07 00:22:59 lol!! Nov 07 00:23:11 yeah, there is really no excuse for hardware raid anymore, *and* ahci is usually faster Nov 07 00:23:24 plus you don't have to hope the controller can do trim Nov 07 00:23:45 seen plenty of "SSD" sas controllers that don't do trim and just destroy SSDs Nov 07 00:24:02 jwh: https://www.youtube.com/watch?v=JwZwkk7q25I Nov 07 00:24:42 lol Nov 07 00:24:59 I dunno if you're a homestarrunner fan, but it always reminds me of that damn song Nov 07 00:25:20 was when it was relevant :P Nov 07 00:25:35 yeah, relevance is relative Nov 07 00:26:12 They got real jobs, mostly in childrens television Nov 07 00:28:05 heh Nov 07 00:28:25 gonna be a long night I think, bbl Nov 07 00:28:35 ok, gl Nov 07 00:35:07 know its bad when grub erors on checksum when loading the kernel :D Nov 07 01:00:29 jwh: That Odroid had me all excited until I saw the NICs. Nov 07 01:00:46 Monkeh: what are they? Nov 07 01:00:52 Crabs. Nov 07 01:00:55 realtek? Nov 07 01:01:01 Yeah. Nov 07 01:01:04 Crabnics. Nov 07 01:01:06 you can actually do a lot worse Nov 07 01:01:11 modern realtek is just fine Nov 07 01:01:14 You can also do a lot better. Nov 07 01:01:22 they're cheap and low powered Nov 07 01:01:56 the odroid-c1 has a realtek and a crappy soc, will still do 700meg single stream tcp throughput though Nov 07 01:02:03 so it should be ok Nov 07 01:02:17 Also lack of wifi opportunity Nov 07 01:02:29 meh, wifi :D Nov 07 01:04:50 its more useful for storage stuff I guess, or maybe htpc Nov 07 01:04:55 dual gig nic though, so thats nice Nov 07 01:10:04 * Monkeh amuses himself exploding mt76 again Nov 07 01:10:46 moving this data at a whopping 800mbps, I *really* hope it doesn't die again, or need to pull data off a broken disk to finish Nov 07 01:14:08 Well MFP seems improved, anyway. Nov 07 01:14:28 i/ Nov 07 01:14:30 o/ Nov 07 01:14:43 4addr still explodes. Nov 07 01:15:12 I need to figure out why this kmod causes a kernel panic :( Nov 07 01:15:24 kinda hard to reproduce though Nov 07 01:15:31 any volunteers? Nov 07 01:15:56 I have enough exploded stuff :P Nov 07 01:16:16 lol Nov 07 01:17:00 think I'm gonna get one of those odroids when they come out, if they're reasonably priced Nov 07 01:17:08 which given hardkernel, usually is Nov 07 01:17:25 since they haven't announced any update to the xu4 Nov 07 01:17:49 I'm.. also magically missing a LED on this board, that's really weird. Nov 07 01:17:56 "missing" ? Nov 07 01:17:58 it fell off? Nov 07 01:18:19 No, it seems to be there, it just doesn't work Nov 07 01:18:22 oh Nov 07 01:18:40 well thats no good Nov 07 01:20:28 Now it works. I guess that's what I get for using cheap sockets. Nov 07 01:20:44 Cheap SoCs, too.. Nov 07 01:22:45 heh Nov 07 01:23:49 Ah, the DED (Dark-Emitting Diode). Nov 07 01:36:09 if you have problems with 4addr, I'd suggest testing to revert mac80211 back to its 4.14 based state - 4addr is also broken for me on ath10k and ath10-ct since the bump to 4.18/ 4.19 (but ath9k seems to be o.k.) Nov 07 01:37:58 pkgadd: This is not related Nov 07 01:39:00 o.k., just mentioning it as a potential candidate Nov 07 01:40:57 4addr works fine, this chip hangs during TX and will not reset. Actually, it doesn't seem to survive any reset attempt at all. Nov 07 01:42:11 * Monkeh experiments more Nov 07 02:11:21 mamarley: DED? :) Is this a Bell Labs joke? :) Nov 07 02:11:52 http://web.mit.edu/kolya/misc/txt/dark_suckers **** ENDING LOGGING AT Wed Nov 07 02:59:59 2018