**** BEGIN LOGGING AT Sun Mar 17 02:59:57 2019 Mar 17 08:39:39 SNAPSHOT r9625 builds fine. I just made builds for my 3200acm and c7-v2 all is good Mar 17 08:40:00 That was for DonkeyHotei Mar 17 08:47:39 * Borromini hopes someone with a DIR-878 or DIR-882 can post his boot logs and dump his flash on the forums Mar 17 10:23:10 Tapper: r9625 builds likely because you already have cmake compiled Mar 17 10:23:30 my build box is offline so cannot test, can you try `make tools/cmake/clean` and then build again? Mar 17 10:25:18 meh, I'm just going to revert that cmake bump Mar 17 10:26:48 didn't dizzy check it? Mar 17 10:26:58 or does he just bump shit blindly Mar 17 10:27:46 it seems completely untested Mar 17 10:28:32 lovely Mar 17 10:28:51 well as I said it can't test, but every log I looked at failed at tools/cmake Mar 17 10:28:58 I* can't test Mar 17 10:29:10 let's see if the builds continue after the revert Mar 17 10:30:24 i can check on my desktop if you'd like Mar 17 10:30:33 already pushed the revert Mar 17 10:30:39 yeah i saw Mar 17 10:39:37 http://phase1.builds.lede-project.org/builders/sunxi%2Fcortexa53/builds/957 passed tools so that was the problem Mar 17 10:40:44 what's weird is that http://phase1.builds.lede-project.org/builders/ath79%2Fnand/builds/203/steps/tools/logs/stdio seems to have built fine Mar 17 10:41:11 maybe it's related to the host system Mar 17 10:43:56 k Mar 17 10:53:50 I don't see any build issues here Mar 17 10:55:22 ynezz: http://phase1.builds.lede-project.org/grid Mar 17 10:55:50 ahh Mar 17 10:56:33 as my buildbox is down I just reverted the commit so the snapshot images build again Mar 17 10:57:46 I need to fix serial console on that build box, damn thing keeps crashing when I'm traveling Mar 17 10:58:20 Hi stintel I built on wsl running ubuntu for my 3200cam and c7 ath79 and it worked fine Mar 17 10:59:37 must be host system related then Mar 17 11:02:05 probably too old gcc on build hosts Mar 17 11:03:38 wsl = windows subsystem for linux? Mar 17 11:12:05 stintel:building locally I did not have any issue with cmake as I'm using a more recent gcc versions in comparison to the buildbots Mar 17 11:12:24 so indeed it's related to the host system Mar 17 11:14:38 I guess we need to keep cmake at its current version, we don't want to break build for users running debian ;) Mar 17 11:14:48 ack Mar 17 11:14:50 (or centos ;)) Mar 17 11:15:48 maybe we should report it to cmake Mar 17 11:20:27 I've here Ubuntu 4.8.5-4ubuntu8~14.04.2 and it built just fine Mar 17 11:27:17 BTW, I'm collecting information about CRNG init times, please can you provide me with output of `dmesg | grep random.*done` command from your routers running latest snapshots (4.14 kernel)? Mar 17 11:27:38 I've ar9342, ipq4019 so far Mar 17 11:29:08 ynezz: is 18.06 with 4.14 ok as well? Mar 17 11:29:38 I think it's on 4.9 Mar 17 11:29:46 mt7621 is 4.14 Mar 17 11:29:55 [ 2.943770] random: fast init done Mar 17 11:29:55 [ 5.254226] random: crng init done Mar 17 11:30:11 thanks, what device is it? Mar 17 11:30:20 dir-860l rev B1 (mt7621) Mar 17 11:30:29 18.06 HEAD Mar 17 11:30:52 [ 35.261625] random: fast init done Mar 17 11:30:54 [ 108.367563] random: crng init done Mar 17 11:31:07 ^^ x86/64 APU2 18.06 HEAD (4.14 as well) Mar 17 11:32:26 seems a bit long for x86/64, i have included the hw crypto modules in the x86/64 build, don't know if that might interfere Mar 17 11:33:34 or maybe missing some hw rng module? Mar 17 11:33:46 I don't know if it's available on apu2 Mar 17 11:34:27 just exploring this rng area as the blocking getrandom is ridiculous and drives me crazy Mar 17 11:36:48 yeah i've seen your e-mails on the ML Mar 17 11:36:58 I seem to rmemeber years of arguments that led to making it blockin though. Mar 17 11:37:19 those patches about trusting cpu options seem like a better path forward though, I'd agree. Mar 17 11:38:33 the thing is, that they doesn't make much difference here Mar 17 11:40:35 I've just booted ar9342 with random.trust_cpu=on and it still takes 130s to crng init done Mar 17 11:42:10 Borromini yes mate I build openwrt on windows 10 Mar 17 11:43:09 When wsl was first released it was slow but now it's mutch faster Mar 17 11:43:12 are the build times under WSL comparable to native Linux compilation? Mar 17 11:44:19 ynezz I am not shure mate but it's not to shabby for me on a old core I7 920 clocked at 3.2 ghz Mar 17 11:44:56 I think that native Linux would be faster Mar 17 11:45:08 Because of the over heads Mar 17 11:46:36 building a os in a os in a os! Mar 17 11:47:58 Tapper: for example, how long does it take to build ath79 under WSL ? Mar 17 11:48:29 if you have such numbers of course, just curious Mar 17 11:48:30 You know what mate I am not shure, about 15 mins i think Mar 17 11:48:43 that's not too bad Mar 17 11:48:53 that is after all the tool chane is ready built Mar 17 11:49:56 I have 6 gb ram and a ssd as mane drive Mar 17 11:51:01 I need to get a new ssd tho man this thing is old now. It's a crucial m4 Mar 17 11:51:50 I got it around Mar 2011 Mar 17 11:52:53 You know I am going to look on amazon rite now. Mar 17 11:58:31 Tapper: what size are you looking for? Mar 17 11:58:45 i got a 970 1 TB at black friday prices (kinda) Mar 17 11:59:04 240 gb I think. rite now mine is 64 gb lol Mar 17 11:59:06 but the HP EX920 seems very good and reasonably priced as well e.g. Mar 17 11:59:07 eh Mar 17 11:59:23 buy at least half a terabyte i'd say Mar 17 11:59:31 was thinking about the Kingston SSD A400 Mar 17 11:59:39 £28 Mar 17 12:00:10 please don't. you have a reliable brand (crucial), buy reasonable stuff Mar 17 12:01:43 the 860 EVO costs 90 € for 500 GB now Mar 17 12:03:23 Tapper: S-ATA or NVMe? Mar 17 12:03:48 I am stuck with sata 2 on this mobo Mar 17 12:04:38 seems like it is OK https://www.hardwaresecrets.com/kingston-a400-120-gib-ssd-review/ Mar 17 12:05:47 get an MX500 500GB i'd say Mar 17 12:06:02 or 250 if you wish Mar 17 12:06:47 keep in mind smaller sizes mean lower performance, you won't notice on S-ATA 2 but you will if you keep it in newer builds Mar 17 12:10:19 I got it now Borromini I will let you know how it is. Mar 17 12:10:48 the Kingston that is. I had one befor in a laptop and it was ok Mar 17 12:11:41 I am rebuilding my pc soon anyway so then I will get a bigger NVMe one Mar 17 12:21:28 nvme is still rather expensive and you won't really notice the performance difference as a normal user Mar 17 12:22:44 it's nice to get rid of the cable clutter though :P Mar 17 12:32:20 so nvme has no measurable speedup for compilation? Mar 17 12:33:54 I understand, that compilation is more CPU then I/O bound, but still if you've 32 cores, the I/O demand could be higher as well Mar 17 12:34:46 Borromini: BTW `[ 79.990240] random: crng init done` here on apu2c, default OpenWrt config and latest master Mar 17 12:35:16 still very slow, even with random.trust_cpu=on Mar 17 12:39:56 ynezz: noted :) Mar 17 12:40:33 ynezz: it's not that much I/O i'm afraid. i have a 1 TB NVMe (and 970 is high end consumer NVMe afaik) but not a big difference Mar 17 12:40:49 i did a few clean runs between my S-ATA 600 and NVMe disks with OpenWrt Mar 17 12:41:25 it's not that much data crunching after all Mar 17 12:47:45 One of those PCE drives would help but... Mar 17 12:48:36 I saw one on a tech site that has 4000 read and 3050 rite Mar 17 12:48:56 but if the file is 20kB Mar 17 12:49:00 aye Mar 17 12:49:12 keep an eye on iotop e.g. when you're running a build Mar 17 12:51:39 So what's the best way to clone a drive? Mar 17 12:51:54 dd Mar 17 12:52:07 or just tar up the contents and unpack it Mar 17 12:52:12 dd /nul Mar 17 12:52:16 lol Mar 17 12:52:20 eh Mar 17 12:52:28 I am on windows Mar 17 12:52:56 sorry :D Mar 17 12:52:58 wsl Mar 17 12:53:01 I want to clone hole drive Mar 17 12:53:13 then reboot to live linux from usb Mar 17 12:58:09 I'm wondering when this crng init done shows up on i.mx6, it's up for 9 minutes already... Mar 17 12:58:43 eh Mar 17 13:08:56 [ 1120.394672] random: crng init done Mar 17 13:09:36 i could generate that entropy faster on paper with pencil Mar 17 13:10:39 hahaha Mar 17 13:10:59 it probably wouldn't be high quality entropy though Mar 17 13:11:06 dice Mar 17 13:11:09 after two beers? Mar 17 13:11:14 humans aren't that good at coming up with random numbers Mar 17 13:12:11 * russell-- has infinitenoise and chaoskey 'true' random noise generator dongles Mar 17 13:12:18 but then again, you can never be sure... https://dilbert.com/strip/2001-10-25 ;) Mar 17 13:13:03 russell--: how are you going to use this on router without USB? Mar 17 13:13:10 https://www.crowdsupply.com/13-37/infinite-noise-trng and https://altusmetrum.org/ChaosKey/ Mar 17 13:13:16 yeah, not. Mar 17 13:13:37 although the chaoskey has a driver in the upstream kernel Mar 17 13:14:10 almost as expensive as the router Mar 17 13:14:53 in some cases several times :) Mar 17 13:17:13 Related sidenote: I nowadays generate new device sshkeys etc with computer where I do have "guaranteed" good entropy Mar 17 13:17:37 made generating pgp keys a lot faster than wiggling the mouse and typing irregularly on the keyboard Mar 17 13:18:38 olmari: yeah, i've been generating host keys from openwrt routers on a build host for enhanced randomness Mar 17 13:18:49 for* Mar 17 13:19:39 this was after i basically published the host keys in a github repo for a few years, oops Mar 17 13:20:23 Ofcourse that won't help if device has tasks that need good rng often, but most devices I deploy doesn't:) Mar 17 13:30:57 looking at the schematics and parts needed to built chaoskey https://altusmetrum.org/ChaosKey/v1.0/noise-source.svg it makes me wonder why it costs that much (relatively) Mar 17 13:34:07 think I will give Macrium Reflect a go Mar 17 13:34:54 wish I could get faster P and P from amazon. I have to wate til the 19th now Mar 17 13:35:32 What ever happened to same day drop offs using drones? lol Mar 17 13:35:51 Well in grand scheme of things yhose are still kinda novelty items... also in general if you want to do this kind of device correctly you need verified chain and pot device in hard goop and basically make it impossible to tinker with... so plenty other stuff not directly related to conponents and that design Mar 17 13:37:35 Simple conponent and pcb layout, expensive as heck to make whole device "tinker proof" at any and all aspects Mar 17 13:57:15 can someone merge packages PR #8370 already? Mar 17 14:24:27 nvm Mar 17 16:19:08 updated openwrt/upstream, https://sdwalker.github.io/uscan/index.html Mar 17 16:28:04 stintel: that cmake fallout was caused by g++ 4.x Mar 17 16:28:19 dizzy tests his changes on bleeding edge stuff Mar 17 16:28:37 its not the first time that things broke because we use older compilers Mar 17 16:30:18 I guess generally newer is better (sans bugs), but also not always that easy to "just change" Mar 17 16:31:27 the thing is that people love to use arch or alpine or whatever else is hip to test and build their openwrt stuff Mar 17 16:32:05 while our builders use older debian releases and older gcc/g++ versions mainly to ensure that the resulting IB and SDK binaries are as widely usable as possible Mar 17 16:33:50 :-) Mar 17 16:34:07 s/people/some people/ Mar 17 18:40:10 ynezz: with WSL, CPU performance is either the same or faster than native linux Mar 17 18:41:14 but filesystem I/O is slow. I also remember make -j 4 causing BSODs when I used it before. Mar 17 19:07:39 ynezz: r9586-74ef848b96, 4.14.105: [ 2.789205] random: fast init done \n [ 12.579479] random: crng init done ipq8065/ nbg6817 Mar 17 21:13:33 jow: FYI I've here Ubuntu 14.04.2 LTS with gcc/g++ versions 4.8.5 and 4.9.4 and both can build cmake 3.14.0 just fine, maybe ccache? Mar 17 21:16:56 i have 14.04.6, and both gcc and g++ are at 4.8.4 Mar 17 21:17:45 So far stuff has compiled fine on half assed ubuntu 18.10 Mar 17 21:18:44 DonkeyHotei: you can install 4.9 Mar 17 21:19:41 Description:Ubuntu 14.04.6 LTS here, with gcc-4.8 (Ubuntu 4.8.5-4ubuntu8~14.04.2) 4.8.5 and gcc-4.9 (Ubuntu 4.9.4-2ubuntu1~14.04.1) 4.9.4 Mar 17 21:22:34 gcc-4.8 is 4.8.4-2ubuntu1~14.04.4 for me, and gcc-4.9-base is 4.9.3-0ubuntu4 Mar 17 21:23:38 ynezz: maybe Mar 17 21:23:50 DonkeyHotei: and you're not able to build cmake 3.14.0 with this versions? Mar 17 21:23:59 i have not tried Mar 17 21:24:46 my point was that your gcc pkgs are not the ones in the repo Mar 17 21:27:20 correct, I've http://ppa.launchpad.net/ubuntu-toolchain-r/test/ubuntu configured for some reason Mar 17 21:28:11 maybe for gcc-8 Mar 17 21:28:38 will debug during the next days Mar 17 21:29:13 but at the end its just another version bump just for the sake of it, tying up enourmous resources to hunt bugs Mar 17 21:29:29 I mena how meny developers have been looking into it yet? Mar 17 21:30:40 running CC="gcc-4.9" CFLAGS="-O2 -I/home/jow/devel/openwrt.git/staging_dir/host/include/" CXX="g++-4.9" CXXFLAGS="-O2 -I/home/jow/devel/openwrt.git/staging_dir/host/include/ -std=gnu++14" ./configure Mar 17 21:31:02 and the same variables + make here, results in openssl link failures when building cmake Mar 17 21:31:40 bbl Mar 17 21:33:35 I had to change symlinks to gcc/g++ 4.9 as it seems, that cmake doesn't fully honor the CC/CXX variables Mar 17 21:34:28 while running make with VERBOSE=99 I've seen hardcoded /usr/bin/g++ in the ouput instead of the g++-4.9 provided in CXX variable Mar 17 21:41:21 I mean, I've run it with CC=gcc-4.9, but in the verbose output I can see `cd /opt/devel/cmake/cmake-3.14.0/Source && /usr/bin/g++` Mar 17 21:56:07 jow: my build box is arch **** ENDING LOGGING AT Mon Mar 18 03:00:21 2019