**** BEGIN LOGGING AT Tue Dec 06 02:59:56 2022 Dec 06 06:36:08 Morning! Dec 06 06:36:13 Tofe: Nice you got some fixes :D Dec 06 06:37:10 I thought we somehow fixed that "findMyAppId" and private/public registration already earlier when we looked into it, but maybe not fully then Dec 06 06:37:27 When we got rid of the pub/priv registration a few months ago Dec 06 07:29:12 webruntime-91-r52.4 do_compile - 9h2m9s Dec 06 07:29:23 wow. such a long compilation Dec 06 07:30:05 build ran overnight and still in progress Dec 06 07:49:26 landgraf: Depends on your builder, but yes resource intensive Dec 06 08:32:21 Morning! Dec 06 08:32:32 JaMa: test #4 started Dec 06 08:34:21 Herrie: Well. It's a side task on lenovo T15 Dec 06 08:36:17 landgraf: Well webruntime is basically full Chromium so it requires lots of CPU + memory Dec 06 08:36:19 and I'm in Prague atm. So maybe it's JaMa who toook all compute resources of the city :-D Dec 06 08:36:48 Even on my 32c/64t Threadripper it takes "long" Dec 06 08:37:07 landgraf: Just out of curiosity: What brings you here? Dec 06 08:38:14 Herrie: well. I've got job here and commuting once per week. But I'm about to give up Dec 06 08:42:32 I meant to build LuneOS Dec 06 08:43:07 landgraf: I've just switched to a more recent CPU/RAM set, because my builds were way too long (was a i5-6600, 16Gb RAM) Dec 06 08:43:41 Herrie: no. I'm building webos for pinephone to play with Dec 06 08:44:01 my job is not yocto/webos/linux related unfortunately. Dec 06 08:44:56 Tofe: T15 has i7 and 32Gb. 10 hours is too much imho. but yeah. it's cross compiling so kind of expected Dec 06 08:45:51 Last time I built pinephone with the old rig, it took around 12h iirc Dec 06 08:46:40 landgraf: you're building on a SSD, right? it makes a lot of difference Dec 06 08:46:50 yes. it's ssd Dec 06 08:48:56 https://cpu.userbenchmark.com/Compare/Intel-Core-i5-6600-vs-Intel-Core-i5-1135G7/3514vsm1286124 well, you should be in the same kind of timing, I'd say 13h Dec 06 08:49:16 chromium is really a pain for the build. And nodejs doesn't help either... Dec 06 08:49:56 landgraf: but if you want, I can upload my latest pinephone build somewhere for you to flash Dec 06 08:53:05 Tofe: I'm not in a hurry. My pinephone is 200km away from me anyway :-) Dec 06 08:53:25 ok :) Dec 06 09:05:58 landgraf: 9h is a bit too much for i7 with 32G, check tail -f on the log.do_compile file Dec 06 09:06:29 it should show around 32K ninja tasks, if you're not near the end, then something is very wrong Dec 06 09:06:47 also htop should show all cores maxed most of the time Dec 06 09:07:24 unless you've configured PARALLEL_MAKE to use too few (as it's side job after all) Dec 06 09:08:15 maybe it's the governor, like I had earlier ? it's in powersave by default, which could mean half of the max frequency Dec 06 09:09:36 JaMa: on my ongoing search for the perfect monitoring, I'm looking more carefully at "glances", which has many different modes of operation Dec 06 09:10:00 it can output to influxdb, mqtt, can be a webserver, store some history... Dec 06 09:10:37 JaMa: [35961/40997] CXX obj/third_party/blink/renderer/core/core/html_dimension.o Dec 06 09:11:25 35967/40997 Dec 06 09:11:33 so it's doing at least something Dec 06 09:12:06 landgraf: and you CPU is operating at max frequency, all the cores are being used ? Dec 06 09:12:09 uour* Dec 06 09:12:14 ah, damn. Dec 06 09:12:35 according to htop all cores are nearly 100% loaded Dec 06 09:13:34 cpu MHz: 800.016 Dec 06 09:13:36 hmmm Dec 06 09:13:39 doesn't look right Dec 06 09:14:52 that could be the governor which is "powersave" by default Dec 06 09:14:59 you can try sudo cpupower frequency-set -g ondemand Dec 06 09:15:09 (if you have cpupower installed) Dec 06 09:15:23 or withcpupower-gui, if you prefer a GUI Dec 06 09:18:47 current CPU frequency: Unable to call hardware Dec 06 09:18:49 current CPU frequency: 900 MHz (asserted by call to kernel) Dec 06 09:19:04 governor is performance Dec 06 09:19:39 that's weird Dec 06 09:20:28 yup. that's fedora :) Dec 06 09:21:45 ah, I never used that one; I've been using archlinux (and building on ubuntu) for ages Dec 06 09:23:25 google says it's the issue with intel_pstate driver Dec 06 09:29:02 ah, maybe it's not loaded ? or acting weird? Dec 06 09:30:21 dunno. # cpupower frequency-set -u 4000Mhz made system more responsive but reported frequency is 800MHz still. time to report bugs :) Dec 06 09:41:33 which T15 generation do you have? I see that there are at least 3 with different number of cores on their i7 Dec 06 09:41:55 e.g. 3rd has Intel Core i7 12700H Alder Lake with 14 cores Dec 06 09:42:23 i7-1165G7 (4C / 8T) in 2nd Dec 06 09:43:16 i7 10510U (4C) in 1st, but maybe there are other CPU options in T15 to choose as well Dec 06 09:44:03 but still 9+h is a bit too much for all of them :) Dec 06 09:44:38 for the freq, it might be throttled by cooling Dec 06 09:45:22 Tofe: I've installed glances now, will try to use it for a while as replacement for htop/atop Dec 06 10:11:04 ThinkPad T15 Gen 2i Dec 06 10:33:15 JaMa: also check all the options, there are plenty of modules you can (de)activate if the default UI doesn't please you Dec 06 10:50:05 Tofe: I was trying to find out how to show all cores, as the "per-cpu" layout is not long enough for all my cores :) Dec 06 10:50:39 in htop I can easily change it to use 4 columns for cpu cores, haven't found the same in glances yet Dec 06 11:36:00 JaMa: right now I'm building in parallel qtwebengine, webkitgtk, qtwebkit, firefox and chromium-x11... OOM in 3...2...1... Dec 06 11:37:00 JaMa: ah, not sure there are so many options for the layout in glances Dec 06 11:41:11 Tofe: interesting you got firefox at the same time as others, usually it's rust-native which takes the longest and firefox starts only after that Dec 06 11:54:46 at least it's not (yet) using all my RAM Dec 06 12:25:36 JaMa: Any thoughts on how to set "WEBOS_TARGET_MACHINE" in an easy non-hacky way? Dec 06 12:26:34 webos_machine_dep isn't hacky, it just makes everything MACHINE_ARCH Dec 06 12:27:19 the goal is to change the components not to use WEBOS_TARGET_MACHINE (or any other MACHINE specific variables) Dec 06 12:28:11 JaMa: OK, seems we just have differences in our webos_cmake.bbclass leading to this for now Dec 06 12:32:44 Running into some issues in CMake for media-resource-calculator and umediaserver-config where it doesn't have WEBOS_TARGET_MACHINE defined Dec 06 12:33:20 I.e.: https://github.com/webosose/media-resource-calculator/blob/master/calculator/CMakeLists.txt#L42 Dec 06 12:33:52 "CMake Error at calculator/CMakeLists.txt:42 (if): if given arguments: "STREQUAL" "raspberrypi3" "OR" "STREQUAL" "raspberrypi3-64" Unknown arguments specified Dec 06 12:34:57 Herrie: if we make that recipe MACHINE_ARCH, are there other recipes that DEPEND on it ? Dec 06 12:35:12 yes, e.g. umediaserver and webruntime IIRC Dec 06 12:35:20 damn, webruntime :( Dec 06 12:35:48 and it won't build for any MACHINE we currently build except qemux86* because it expects each MACHINE to have own .json files defined etc Dec 06 12:36:48 great Dec 06 12:37:09 how hard is it to just exclude it ? Dec 06 12:37:46 we've been fine so far without it Dec 06 12:38:12 09:20 < JaMa> I belive 94 could be configured to use the same deps as with 91, that might be the first step Dec 06 12:38:30 ok Dec 06 12:40:29 https://github.com/webosose/media-resource-calculator/blob/master/calculator/resource_calculator.cpp#L28 ahah, so much for defining variables in CMakeLists... Dec 06 13:05:42 0: webruntime-91-r52.4 do_compile - 14h38m40s Dec 06 13:05:45 :) Dec 06 13:07:30 amazing Dec 06 13:09:29 that T15 laptop is not doing well. RTC drifts so much that chrony gives up to sync and now CPU issue Dec 06 13:15:05 JaMa: I'm approaching the end, qtwebengine is at 90% and half chromium-x11 Dec 06 13:16:14 at least I'm pretty sure it won't trigger the OOMK Dec 06 13:45:21 JaMa: Well I'm just building for qemux86-64 now, so missing config files shouldn't be an issue as of yet, but will try to drop those recipes for now Dec 06 14:02:11 JaMa: Tofe: Sooooo. RTFM case :D `Press Fn+H to switch to performance mode. (limits CPU to 20W) Dec 06 14:03:30 * landgraf is looking how to disable this "feature" Dec 06 14:44:40 current CPU frequency: 3.30 GHz (asserted by call to kernel) Dec 06 14:44:44 better now Dec 06 14:58:22 now the plastics can melt properly :) Dec 06 15:48:51 JaMa: Running into: https://pastebin.ubuntu.com/p/4pVrSJxnn4/ Dec 06 15:49:40 Package id 0: +92.0°C Dec 06 15:49:44 no so bad Dec 06 15:49:54 and not enough to melt ABS Dec 06 15:57:08 Herrie: what are your metadata changes for webruntime? did you enable is_webos for webruntime? as media_platform_api seems to be included only with that Dec 06 15:57:50 hmm that was already enabled in 91 Dec 06 15:58:13 maybe you've removed use_neva_media=false ? Dec 06 15:59:38 JaMa: OK let me check that Dec 06 16:08:02 JaMa: Seems these were added compared to our .inc: https://github.com/webosose/meta-webosose/blob/honister/meta-webos/recipes-webos/chromium/webruntime.inc#L48-L50 Dec 06 16:08:06 Which we use for 91 Dec 06 16:08:10 Trying now with these removed Dec 06 16:10:46 https://github.com/webosose/meta-webosose/commit/a9e47a3f38dd678ee3531ba012d17e5a1737d31f Dec 06 17:12:18 JaMa: ERROR: luneos-dev-package-1.0.3+gitAUTOINC+b69bf00f60-r0 do_deploy: Required base image is not available as tar.gz image! /mnt/builds/webos-ports-env/webos-ports/tmp-glibc/deploy/images/pinephone/luneos-dev-image-pinephone.tar.gz doesn't exist. Dec 06 17:12:30 that's the issue I've mentioned in #yocto Dec 06 17:17:45 landgraf: the right target for pinephone is luneos-dev-image, not "package" Dec 06 17:18:11 but as package depends on image, you should already have everything in deploy directory Dec 06 17:19:10 maybe we should harmonize all this; currently, package is dedicated to android fastboot images Dec 06 17:19:36 JaMa: test #4 finished Dec 06 17:19:36 Tofe: MACHINE=pinephone bb luneos-dev-package Dec 06 17:19:45 MACHINE=pinephone bb luneos-dev-image Dec 06 17:19:53 oh Dec 06 17:20:05 but don't worry, it should be already built Dec 06 17:20:11 time to update docs then ) Dec 06 17:20:16 oops :) Dec 06 17:20:41 https://webos-ports.org/wiki/Build_for_Pinephone Dec 06 17:29:23 Tofe: so around 9 hours? that's more than I would expect (and the 2-bb test will take even longer, but possibly not by much in your case) Dec 06 17:30:57 5-build-8-bb should be faster thanks to less swap use Dec 06 17:33:29 JaMa: I don't think I used swap for 4-build-all Dec 06 17:33:39 (to my surprise) Dec 06 17:34:26 but let's see tomorrow for the next test Dec 06 17:34:57 you have 16t and 64G ram? Dec 06 17:38:14 yes Dec 06 17:41:45 ah ok, 4g per thread is better than what we have :) Dec 06 17:47:52 well, you just have way more threads, too Dec 06 17:49:52 I would buy another 4x32G ram, but would need to use 2666 and still might get stability issues, so I have to deal with 128G and ocassional swap when pushing it Dec 06 17:52:38 JaMa: Now getting: "../../git/src/ozone/wayland/shell/webos_shell_surface.cc:115:26: error: invalid conversion from 'void (*)(void*, wl_webos_shell_surface*)' to 'void (*)(void*, wl_webos_shell_surface*, int32_t, int32_t)' {aka 'void (*)(void*, wl_webos_shell_surface*, int, int)'} [-fpermissive]" Dec 06 17:59:12 Herrie: could that be a patch issue ? Dec 06 18:00:09 Herrie: it seems related to HandleClientSizeChanged, which I introduced in one of my patches Dec 06 18:03:31 Tofe: Ah yes could be Dec 06 18:03:41 I didn't add any patches yet Dec 06 18:17:54 Tofe: I guess we'll need: 34, 35, 36, 37, 41 and 42? https://github.com/webOS-ports/meta-webos-ports/tree/kirkstone/meta-luneos/recipes-webos/chromium/files Dec 06 18:24:03 39 is also still needed Dec 06 18:24:26 and included in meta-webosose Dec 06 18:24:52 Herrie: I guess that's correct, for our patches Dec 06 18:25:11 JaMa: Ah OK Dec 06 18:59:39 Tofe: Patches look good, if you built tested them, I say go ahead with merge **** ENDING LOGGING AT Wed Dec 07 02:59:56 2022