**** BEGIN LOGGING AT Wed Nov 27 02:59:57 2019 Nov 27 07:34:03 stintel: for me, non-htt-mgt works better than htt-mgt Nov 27 07:34:57 there's some weird stuck queue with htt-mgt that grinds all traffic to a halt. I've given up on debugging it. Nov 27 09:50:21 Hello everyone. Nov 27 09:51:55 I'm trying to compile own apps for OpenWRT using bundled toolchain however I can't find headers for such libraries as libglib, libevdev etc. They are missing from toolchain even if I include these libs, neither I see them in ipks Nov 27 09:52:53 what means "bundled toolchain"? The SDK? Nov 27 09:52:54 where are you looking? Nov 27 09:55:35 jow There are either SDK and toolchain after firmware compilation. Toolchain has GCC with basic libraries + headers while SDK doesn't seem to include such files Nov 27 09:56:17 karlp I'm looking in openwrt/bin/targets/ipq40xx/generic/openwrt-toolchain-ipq40xx-generic_gcc-8.3.0_musl_eabi.Linux-x86_64.tar.bz2 in my case Nov 27 09:56:33 you're supposed to compile the required dependencies with the SDK before you compile your own application Nov 27 09:56:40 this will stage the headers appropriately Nov 27 09:57:11 unpack the sdk, run ./scripts/feeds update; ./scripts/feeds install glib2 etc. Nov 27 09:57:53 OpenWrt does not bundle package headers with its toolchain or SDK artifacts Nov 27 09:58:15 Isn't it enough to pick a package during 'make menuconfig' ? Nov 27 09:58:32 no Nov 27 09:58:38 I mean for the buildroot it is, yes Nov 27 09:59:03 but the resulting SDK archive will always only contain the minimal base set of headers and libraries to ensure reproducability Nov 27 09:59:14 regardless of what has been selected in buildroot Nov 27 10:00:16 After I executed './scripts/feeds install glib2', where am I supposed to find the headers? Nov 27 10:00:41 you need to build it Nov 27 10:00:41 that only installs a symlink so you can actually build it Nov 27 10:00:47 make package/glib2/compile Nov 27 10:07:14 Hmm, I've found headers for glib ./staging_dir/hostpkg, however I don't see libevdev there afrer I done the same steps for it Nov 27 10:07:29 *for glib at ./staging_dir/hostpkg Nov 27 10:07:49 you need to look in staging_dir/target-... I think Nov 27 10:09:04 Folder 'include' there is empty Nov 27 10:09:37 nitroshift: mornin' Nov 27 10:09:45 yo stintel Nov 27 10:10:09 what you doing? Nov 27 10:10:22 There are some results when I run 'find . -name "evdev.*h' in root of openwrt dir, however they are only in ./build_dir Nov 27 10:10:24 $dayjob Nov 27 10:10:49 test-driving codefresh container-native CI Nov 27 10:11:20 and waiting for my workstation to rebuild all 2524 packages I have installed after updating CFLAGS to match my new CPUs Nov 27 10:12:09 sounds like fun :p Nov 27 10:12:46 i'm out, going to change the wheels on my car, i'm still on summer tyres Nov 27 10:12:49 bbl Nov 27 10:13:08 same here, finally starting to make sense to put on the winter tyres here Nov 27 10:17:16 hey people! Nov 27 10:20:05 anyone tried building openwrt on alpine? Think glibc will interfere? Nov 27 10:21:16 pheoxy: I believe I tried. it could work, but if your target builds something like an ancient u-boot fork you're out of luck Nov 27 10:22:29 oh wait, https://twitter.com/mcclure111/status/1196557401710837762 fixed that Nov 27 10:22:37 at least for uboot-ar71xx Nov 27 10:26:12 stintel I checked those libevdev headers but there are just are empty files Nov 27 10:51:22 stintel: that link is a long thread on stadia? Nov 27 10:51:33 ups Nov 27 10:51:37 wrong paste buffer haha Nov 27 10:51:46 45a2771953fb351879c332105c8d270c6daed0a7 * Nov 27 10:52:19 Vavooon: maybe wait for someone who actually uses the SDK Nov 27 10:52:22 stintel I've found eventually, it appears that everything is @ openwrt/staging_dir/target-arm_cortex-a7+neon-vfpv4_musl_eabi/ and then under usr/include. Wouldn't it worth to add it somewhere on Wiki? I believe there are other devs trying to figure out where to find headers/pkg-config files Nov 27 10:53:07 ah well that's the expected location, that's what I meant when I said staging_dir/target-... Nov 27 10:53:48 feel free to add it to the wiki somewhere if you think it is useful Nov 27 10:54:32 Alright, I was looking in ./include instead of ./usr/include. Then makefiles for required packages and found actual location Nov 27 10:55:06 isn't include an empty dir? Nov 27 10:55:07 if you would use a feed to add a Makefile for the software you are trying to package, like e.g. the package feeds you can simply run "make package/yourthing/compile" and it should magically pick up all those "custom" locations Nov 27 10:56:38 To be honest it was hard for me to figure everything out since I used that SDK contains all required headers (in case of Yocto) or I just download some directories from root filesystem (in case of RPI, some other ARM boards and x86). Neither works for OpenWRT Nov 27 10:56:51 karlp It is Nov 27 10:58:06 stintel I use Cmake so I'll need to specify sysroot/lib/pkg-config files location, but now I know them Nov 27 10:58:37 karlp stintel Thanks for the help! Nov 27 10:59:20 so you are editing your CMakeLists.txt to point to that directory? that sounds ... I dunno Nov 27 10:59:50 No, I specify them in Qt creator for each device Nov 27 11:00:07 I have this in feeds.conf: src-link stintel_wip /home/stijn/Development/LEDE/lede-wip Nov 27 11:00:58 Vavooon: this is a good writeup of all the thigns you can do: https://openwrt.org/docs/guide-developer/helloworld/start Nov 27 11:01:30 I also have CMake cross-files where I specify required paths in case I build my apps from console Nov 27 11:01:31 and then in e.g. /home/stijn/Development/LEDE/lede-wip/bfgminer/Makefile: https://gist.github.com/stintel/950a73188bdef7951ba249bbf144afba Nov 27 11:01:33 stintel: "downside" of using a feed is that you have to make a package makefile for your code too. Nov 27 11:01:58 karlp: sure but it's how OpenWrt works, and no need to look up custom locations or anything and it should work for each target ;) Nov 27 11:02:01 * karlp feels making a package makefile is a _good_ thing as it's what you want eventually anyway :) Nov 27 11:05:10 I'd need to specify some paths in Qt Creator anyway, otherwise it wouldn't see dependencies headers Nov 27 11:07:02 I once made a thing in Qt Creator, I just developed it for x86 and then let the OpenWrt build system take care of the rest :) Nov 27 11:08:12 that reminds me I should probably work on that again (wanted to make a touchscreen interface to my domoticz instance to run on rpi) Nov 27 11:10:10 I have to support other devices/distributions (Yocto, Ubuntu, Raspbian) so keeping everything with one build system seems to be easier for me. Nov 27 11:11:57 the suggestion was to separate them properly. Nov 27 11:13:32 I see, I mean that then I'll need to learn how to force IDE to use OpenWrt's build system, etc. Thanks to you, it's now easier for me just to specify proper locations in Qt creator :) Nov 27 11:17:35 until you have to change it again later. Nov 27 11:17:47 but whatever floats your boat Nov 27 11:19:47 I see, I'll try to use it too, thanks Nov 27 11:21:41 BTW I just need to enable ccache in 'make menuconfig' in order to use it for compilation, correct? Nov 27 11:22:19 that won't do anything for you in qtcreator Nov 27 11:22:46 I see, I'm just talking about firmware compilation Nov 27 11:42:12 trying to get my buildbox to use a docker alpine container and getting some annoting errors. Here's the pastebin (https://pastebin.com/0uHXn7XS). Nov 27 11:42:12 trying to make a tiny build enviroment. Nov 27 11:52:04 pheoxy: that looks unrelated to alpine (or musl libc) Nov 27 11:52:24 wild guess is that it tries to open menuconfig and you don't have a tty Nov 27 11:52:33 if you want to automate, don't use menuconfig Nov 27 11:53:07 seed the .config (from config.seed or ./scripts/diffconfig.sh), then run make defconfig Nov 27 12:33:53 * ynezz I'm doomed, selling all ARM hardware Nov 27 12:34:49 ? Nov 27 12:35:17 RMK got angry Nov 27 12:36:45 http://lists.infradead.org/pipermail/openwrt-devel/2019-November/020382.html Nov 27 13:00:04 pheoxy: https://github.com/openwrt/docker/pull/8 Nov 27 13:01:27 it makes me wonder what is the size difference between Debian based SDK docker image and Alpine based one Nov 27 13:03:38 I understand, that it makes sense to use 5M Alpine image for stuff like shellcheck, but I'm unsure if it makes sense (or is worth the hassle) to use Alpine based container for complete build of OpenWrt Nov 27 13:04:04 I mean does it matter if the result is 6.05GB or 6.2GB? Nov 27 13:07:21 what would cause low rssi on QCA9886? Nov 27 13:12:54 Should I be building openwrt with sudo or doesn't it matter? I get this on root "checking whether mknod can create fifo without root privileges... configure: error: in `/source/openwrt/build_dir/host/tar-1.30':" Nov 27 13:13:24 no need root Nov 27 13:15:01 you shouldn't be building as root, yes Nov 27 13:15:14 it's bad practice Nov 27 13:18:23 given that the buildroot is contained within the buildroot, what's the appeal of trying to run it all inside docker? what's the usecase? Nov 27 13:26:58 convenience, CI, testing of different distributions, reproducible/clean environment with minimal dependencies etc. Nov 27 13:27:28 hey, maybe anyone can check why we cant get some kmod fields shown for openwrt vs normal build from git via modinfo? Last 2 comments here: Nov 27 13:27:30 https://github.com/cifsd-team/cifsd-tools/issues/39 Nov 27 13:28:16 i want the version string, but i cant get in our buildsystem, do we somehow strip those fields? Nov 27 13:37:01 nbd: ping Nov 27 13:37:07 ynezz: pong Nov 27 13:37:29 can you take a quick look please https://github.com/openwrt/openwrt/pull/2489/commits/6f0a9651b06bf5800ffc54c5bd324122b1f4d3d6 ? Nov 27 13:37:42 please push it Nov 27 13:37:48 i will solve the boot delay issue in a different way Nov 27 13:37:52 ok, thanks! Nov 27 13:38:23 i will probably change the driver code to register the phy before the firmware has been started Nov 27 14:02:32 andy2244: yes, we strip kmod metadata Nov 27 14:03:35 ah ok, the new idea is via sysfs or is there a option to preserve this i can add to the package? Nov 27 14:04:54 what's the actual goal? Nov 27 14:07:06 andy2244: https://github.com/openwrt-routing/packages/commit/1c0b31528970a181327ff9309e3efd1e8c625ff0 Nov 27 14:10:39 if this works and is "ok" than i would prefer this over the /sysfs, since i can display the version without the module being loaded Nov 27 14:11:50 karlp: just to display the module version number, so bugreports are easier and we can quickly check if userspace and kernel module match (possible abi breaks) Nov 27 14:20:11 jow: thanks this worked! Nov 27 16:49:02 4.19.85 #0 Tue Nov 26 21:16:43 2019 mips GNU/Linux Nov 27 16:49:18 where do we get that timestamp from? this image was built today from a fresh clone Nov 27 16:50:35 ah Nov 27 16:50:36 CommitDate: Tue Nov 26 22:16:43 2019 +0100 Nov 27 16:50:53 so the commit (not author) date of the last commit in UTC Nov 27 16:51:21 I understand this is for reproducible builds, but can this be disabled? Nov 27 16:51:52 I mean in my .config Nov 27 16:57:31 butwhy.gif? Nov 27 17:04:16 because it's a total mindfuck that an image built today has a kernel timestamp of another day Nov 27 17:06:05 well, now you now, so it's "ok" ? Nov 27 17:16:12 need a quick hint plz: i seem to miss something, if i use the reload button in startup all is fine, yet if i use the save&apply luci button, while i see my reload logic triggered procd is not loading the bin anymore? Nov 27 17:17:22 so whats different between the two cases in regards to procd stuff? Nov 27 17:20:34 "your reload logic" is where? Nov 27 17:20:52 reload_config won't do anything if cmdline/files haven't changed. Nov 27 17:21:15 reload is via luci Nov 27 17:21:35 not realod i mean restart Nov 27 17:21:53 andy2244: on stable or master? Sounds bug to me Nov 27 17:22:13 you mean the page that lists all the services and gives you buttons to push? that explicitly restarts. Nov 27 17:22:15 master Nov 27 17:22:26 does your save and apply actually make a change that procd can see? Nov 27 17:23:55 yes i have explicit reload code that calls my stop/start logic and i see all logs similar to the restart, yet procd does not start the bin Nov 27 17:24:17 in the save&applay code Nov 27 17:24:42 I'd need to see a lot more before I could comment further. Nov 27 17:28:48 my init code https://usercontent.irccloud-cdn.com/file/AT9xwwgt/cifsd Nov 27 17:29:02 this is my init cript Nov 27 17:29:47 if i change a value in luci on a running/connected cifsd and try save&apply cifsd breaks, since procd did not load the bin Nov 27 17:30:05 if i run the restart logic under similar conditions all looks fine Nov 27 17:30:16 does your luci stuff directly modify the smd.conf file? Nov 27 17:30:25 no Nov 27 17:31:45 as noted i see my log output for the restart call and than normal stop/start log entries, yet the userspace cifsd was not loaded by procd Nov 27 17:32:23 you need to make procd aware of the generated smb.conf Nov 27 17:32:28 the kernel module was loaded fine, so the start (procd) should have been executed and the bin should be loaded, but isnt? Nov 27 17:32:37 it will only restart the daemon process if the runtime state differs from the desired state Nov 27 17:32:52 procd_add_file Nov 27 17:32:58 even though reload_service explicitly does a stop/start? Nov 27 17:32:59 grep in other packages for examples Nov 27 17:33:24 that duality of adding reload triggers for the uci config files, but still having to track the output file is.... not all that fun to work with :) Nov 27 17:33:37 i mean i kill cifsd in the stop, so it isnt even running anymore? Nov 27 17:33:42 karlp: well the start and stop procedures do not touch the process Nov 27 17:33:44 so yes, even then Nov 27 17:34:07 so you used to overload reload_service to really make them stop/start, and now you can't? procd just knows best, always? Nov 27 17:34:26 not sure, that init script looks odd Nov 27 17:34:45 manual killing of processes does not belong in there Nov 27 17:34:48 so even if the config changes procd looks for "real" changes via procd_add_reload_trigger ? Nov 27 17:35:00 procd does not care about the config Nov 27 17:35:07 by defualt it checks if the command line changed Nov 27 17:35:08 ok get it Nov 27 17:35:18 only then it restarts the process Nov 27 17:35:32 you need to add all other change sources manually, like interfaces or config files used by the process Nov 27 17:36:17 the reload trigger for cifsd is superfluous Nov 27 17:36:39 then one for dhcp and system needs to remain Nov 27 17:37:00 karlp: for services only caring about their own uci config you do not need any reload triggers Nov 27 17:37:14 thats only required if you want to react on changes in foreign files Nov 27 17:37:36 in the end a reload trigger simply invokes /etc/init.d/foo reload Nov 27 17:37:49 so that implementation needs to ensure that reloads/restarts of the process happen, if needed Nov 27 17:38:19 i mean i see this a lot? Nov 27 17:38:20 service_triggers() Nov 27 17:38:21 { Nov 27 17:38:21 procd_add_reload_trigger acme Nov 27 17:38:21 } Nov 27 17:38:29 cargo culting Nov 27 17:39:23 or, (IMO) it didn't work originally.... Nov 27 17:39:24 or not, actually - please disregard what I said Nov 27 17:39:29 I'm sure instructions to do exactly that appear in the wiki Nov 27 17:39:34 I'm in a hurry and didn't pay attention Nov 27 17:39:36 because I'm pretty sure it definitely was required :) Nov 27 17:39:42 mhh ok thanks for the infos, need to retest this later Nov 27 17:40:03 need grab some food now :p Nov 27 17:40:10 in any case, some procd_set_param file /var/run/smb.conf or whatever it is, should cause procd to kill and restart cifsd Nov 27 17:40:24 ok will try this Nov 27 17:40:26 if and only if that file is actually changed Nov 27 17:40:42 means e.g. a whitespace only change in the uci or some option reordering would not reload cifsd Nov 27 17:40:53 since the generated smb.conf would be the same Nov 27 18:45:03 stintel: we have a patch at work to disable the reproducible builds timestamp handling, if you are interested I can send it Nov 27 18:49:20 stintel: my BTHH5A is lantiq based Nov 27 18:49:31 I used the ath10k-ct without htt FW Nov 27 18:50:38 Hauke: do you think it could be related to the 4.14 kernel ? Nov 27 18:50:55 hmm that would be strange Nov 27 18:51:00 I will try to do some testing Nov 27 18:51:46 as I said my ar71xx AP shows the problem, it is using 4.14, my ath79 with the exact same chip (according to dmesg) and the exact same firmware and mac80211 package versions does not show the problem Nov 27 18:52:30 not sure how quick I will be able to figure out creating a DTS file for my DAP-2695 so I can test it with 4.19 Nov 27 18:52:56 but if there's anything else you can think of that you want me to test .. I'd be happy to Nov 27 18:55:04 can you downgrade ath79 to kernel 4.14? Nov 27 19:00:57 Hauke: on it Nov 27 19:05:18 impressive, filed https://github.com/Motion-Project/motion/issues/1063 15m ago and already got a reply Nov 27 19:09:34 geez now I'm wondering why I never automated my builds before. I just pushed a commit that downgrades ath79 to 4.14 to my github fork of openwrt and images for my APs are already building Nov 27 19:09:43 now I can go and make dinner instead of being hungry :D Nov 27 19:35:01 CI? Nov 27 19:35:34 yeah, test-driving ovh's cds Nov 27 19:35:51 it doesn't support plain git repos though, so I'm abusing github for it Nov 27 19:40:34 :P Nov 27 19:42:24 now thinking of throwing all my diffconfigs in a separate git repo, with a custom feeds.conf so that I can include packages from other feeds in my autobuilds too Nov 27 19:42:58 to eventually have automated builds for all my devices and VMs Nov 27 19:45:56 why are you downgrading ath79? Nov 27 19:53:03 27|19:51:46 < stintel> as I said my ar71xx AP shows the problem, it is using 4.14, my ath79 with the exact same chip (according to dmesg) and the exact same firmware and mac80211 package versions does not show the problem Nov 27 19:53:04 27|19:52:30 < stintel> not sure how quick I will be able to figure out creating a DTS file for my DAP-2695 so I can test it with 4.19 Nov 27 19:53:16 27|19:55:04 < Hauke> can you downgrade ath79 to kernel 4.14? Nov 27 19:53:49 Borromini: or read the commit message: https://github.com/stintel/openwrt/commit/ca138b0369f4510e398e474aeb1dd7ba824eb991 :) Nov 27 19:54:28 k and ath79 is on 4.19 in master i suppose Nov 27 19:54:45 oh that cds is even pushing build info to githubn Nov 27 19:54:45 cool Nov 27 19:55:14 (the orange dot in front of the commit) Nov 27 20:21:22 jow: i will try to convince the devs to change the cifsd behavior i just run some tests and its really bad that it wont pickup certain smb.conf changes without fully removing or hard resetting the kernel server. Even with there own systemd scripts you can't even change the readonly option, without a reboot.... Nov 27 20:23:41 i will try to fix our scripts until this is done, but atm the easiest way seem to remap reload() simply to a restart().. at least thats what i have seen in some other packages Nov 27 21:02:27 Hauke: cannot reproduce the problem on my Unifi AP AC Pro with kernel 4.14 Nov 27 21:56:28 tsys Nov 27 21:57:42 ynezz: Sorry, just noticed your messages regarding my patches Nov 27 21:57:55 Thanks for taking a look at them Nov 27 21:58:28 I could provide some test cases I guess. Will try squeezing it into my schedule tomorrow Nov 27 22:26:52 tobleminer-tSYS: no worries, I'll have time for that next week anyway Nov 27 22:27:59 tobleminer-tSYS: no need to resend, just point me at some git repo or pastebin Nov 27 23:35:36 Hauke, stintel: I get around 450-500 MBit/s with iperf3 in both directions between an nbg6817 (ipq806x/ QCA9984, kernel 4.19, ath10k-ct 5.4, ct-htt) and a 2x2 802.11ax intel ax200 card, http://paste.debian.net/hidden/a2ef0ba3/ Nov 27 23:36:24 stintel, Hauke: that is with WPA3/ SAE and 802.11w=2 Nov 27 23:40:33 9984 = wave2? Nov 27 23:40:38 yes Nov 27 23:40:51 ok, our problems are on wave1 afaik Nov 27 23:41:11 I could also test a BTHub5 tomorrow, if needed Nov 27 23:41:30 so lantiq/ QCA9880, wave1 Nov 27 23:41:54 that is what Hauke is using afaik. could be interesting to know if you have the issue there or not Nov 27 23:45:11 I will try to look into this probably at the weekend again Nov 27 23:45:23 I do not see this blocking for 19.07-rc2 Nov 28 00:01:35 I am very eager for rc2 Nov 28 00:01:53 Just wondering if there is extra WPA3 patches Nov 28 00:07:39 how to convince openwrt-19.07 buildsys to install kernel? https://pastebin.com/NhpnpcM1 Nov 28 02:22:31 was there a recent musl update or something? A bunch of packages are failing for me (missing headers) Nov 28 02:26:11 ooooooooohhhhh I think they're failing since I don't have fortify-headers locally Nov 28 02:26:54 great...it hides problems in more than one way. **** ENDING LOGGING AT Thu Nov 28 02:59:57 2019