**** BEGIN LOGGING AT Thu Apr 29 02:59:57 2021 Apr 29 07:10:13 Morning! Apr 29 07:28:39 JaMa: When you got a minute, can you please import https://github.com/Herrie82/android_kernel_xiaomi_msm8953-1/commits/halium-9.0-mido-4.9 into https://github.com/shr-distribution/linux as branch mido/4.9/halium-9.0 ? Apr 29 07:30:06 And the same with https://github.com/Tofee/android_kernel_xiaomi_rosy/tree/halium-9.0 as rosy/3.18/halium-9.0 Apr 29 08:04:04 Morning! Apr 29 08:04:56 Morning! Apr 29 08:31:44 Tofe: Once JaMa imports the kernels I'll update those recipes and PR these changes as well: https://github.com/Herrie82/meta-smartphone/commits/herrie/fixes Apr 29 08:36:39 Tofe: BTW we don't need Bluebinder on Tenderloin? Apr 29 08:49:30 Herrie: no, it worked as-is Apr 29 08:58:14 OK Apr 29 08:58:16 Just checking Apr 29 09:00:38 JaMa: if I want to enable xz in kmod, I just do a bbappend, right ? Apr 29 09:06:29 well i've put it in luneos.inc, to me it makes more sense Apr 29 09:07:51 Herrie: now my pinephone boots again; it was because the defconfig I tried to use was really too minimalistic Apr 29 09:13:00 Tofe: Ah great! Apr 29 09:13:09 Any performance improvements in new kernel? Apr 29 09:13:45 I think it boots faster; for the rest, I had an issue with modules, so sound and UI didn't work yet Apr 29 09:14:08 and adding xz to kmod triggers a rebuild of qt :( Apr 29 09:15:08 theoretically, that kernel should handle a bit better the modem and power management Apr 29 09:29:27 That's what I understood from Twitter :P Apr 29 09:45:56 if I patch manually kmod, I manage to get wifi, sensor, UI Apr 29 09:46:22 bluetooth I just miss a firmware file (rtl_bt/rtl8723cs_xx_config-pinephone.bin) Apr 29 09:46:33 sound, I don't know yet, PA seems happy Apr 29 09:49:27 unexpected good news: I have a working rear camera in Testr Apr 29 09:49:55 Doesn't work in our camera app though, quite a shame :p Apr 29 09:53:01 Tofe: Hehe funny stuff at times with camera's Apr 29 09:53:40 Our camera app has almost never been really tested with gstreamer, so it can very well be that there's some bug there Apr 29 09:53:52 Tofe: https://github.com/webOS-ports/meta-webos-ports/pull/457 Apr 29 10:07:59 oh, great, it didn't completely rebuild Qt, there's still hope Apr 29 10:29:46 Tofe: yes, bbappend or luneos.inc both are fine Apr 29 10:29:55 Herrie: ok will do in sec, finishing qtwebengine upgrade Apr 29 10:38:49 mmh it's weird, with my xz modules it looks like "depmod -a" wasn't triggered Apr 29 10:41:50 are they built from separate recipe (not with kernel)? Apr 29 10:42:04 you do inherit module, right? it should generate postinst with depmod Apr 29 10:42:06 no, it's yocto-linux Apr 29 10:42:09 IIRC Apr 29 10:42:35 inherit kernel Apr 29 10:42:35 require recipes-kernel/linux/linux-yocto.inc Apr 29 10:42:47 that's pretty much all Apr 29 10:42:51 is it for pine? Apr 29 10:42:54 yes Apr 29 10:43:16 I can check in the logs if I see something about depmod Apr 29 10:45:12 ah, found it https://bpa.st/4UMQ Apr 29 10:45:38 what's this "99.98." ?... Apr 29 10:50:52 hmm strange, can you push your changes? I'll try it here Apr 29 10:51:11 Herrie: mido/4.9/halium-9.0 rosy/3.18/halium-9.0 pushed Apr 29 10:51:54 JaMa: yes, one moment Apr 29 10:52:24 "bb -e linux-pinephone | grep 99.98" doesn't output anything Apr 29 10:59:04 JaMa: done, https://github.com/webOS-ports/meta-pine64-luneos/pull/13 + https://github.com/webOS-ports/meta-webos-ports/pull/458 Apr 29 11:00:06 can be merged, the status isn't that bad if depmod is run manually (I just have sound not working, not sure why) Apr 29 11:01:11 JaMa: Thnx Apr 29 11:07:44 Tofe: the PACKAGECONFIG examples in luneos.inc you've followed with kmod aren't very good :/ it should be _append_pn-* probably in all these cases Apr 29 11:11:04 oh. I can fix that, one moment Apr 29 11:11:20 let me check all of them Apr 29 11:11:44 from this aspect bbappends are a bit safer to use, maybe we should move all these to .bbappends Apr 29 11:13:09 JaMa: I can move mine to a bbappend Apr 29 11:14:44 I'll force-push on my branch Apr 29 11:16:47 done Apr 29 11:17:30 nope, not done Apr 29 11:18:45 now it's better Apr 29 11:43:25 sorry, back now, got lunch irq Apr 29 11:44:22 Tofe: missing leading space in _append = " xz" Apr 29 11:45:13 JaMa: damn, I never know if I should put one or not Apr 29 11:45:46 for space separated variables always use leading space for _append and trailing space for _prepend Apr 29 11:46:40 it doesn't break anything now, because the default is empty, so with your append it will be just "xz" but once someone enables e.g. openssl in the oe-core recipe then "opensslxz" won't help :) Apr 29 11:46:54 force-pushed Apr 29 11:47:47 or sometimes there is newline at the end of default PACKAGECONFIG in oe-core, so it works, until someone reformats it to end without whitespace and then it's issue again, so to stay sane, it's always _append/_prepend responsibility to take care of the separator Apr 29 11:49:06 JaMa: but for += it's managed differently ? Apr 29 11:51:45 right, += adds the space Apr 29 11:52:39 I think they adopted the less intuitive behavior for append. Not needing a space is quite rare... Apr 29 11:53:42 ah, http://embeddedguruji.blogspot.com/2019/04/appending-and-prepending-in-yocto-vs.html explains it all nicely, and now I discover ".=" which is like append :p Apr 29 11:56:34 .=/=. are the least used ones I think, because _append/_prepend are required when you need to combine them with an override (or delayed parsing which the blog post doesn't mention at all) Apr 29 11:57:36 so .=/=. are used only in quite specific cases like BBPATH/OVERRIDES Apr 29 11:59:00 Tofe/JaMa: https://github.com/shr-distribution/meta-smartphone/pull/132 Apr 29 12:02:28 LGTM will merge after ack from Tofe Apr 29 12:04:33 Yeah let him check it first as ewll Apr 29 12:04:40 Tofe: I locally tested kernel build, was OK Apr 29 12:04:42 For Rosy Apr 29 12:04:46 So I guess it's fine there at least Apr 29 12:04:52 I didn't change any of the HCI bits yet Apr 29 12:21:02 ACK Apr 29 12:21:09 Tofe: seems to be a broken the same even with previous 5.8 kernel (and Yocto 3.4 Honister) https://paste.ubuntu.com/p/3wgsMh2GJv/ Apr 29 12:21:57 JaMa: ok, funny that I didn't see the consequence on modules loading Apr 29 12:22:30 depmod might be successful from the postinst executed during do_rootfs Apr 29 12:23:36 my guess is that this issue is from the kernel source itself (not from OE) Apr 29 12:26:14 ./linux-pinephone-standard-build/source/scripts/depmod.sh: KERNELRELEASE=99.98.$KERNELRELEASE Apr 29 12:28:25 added 10 years ago :) Apr 29 12:28:30 commit bfe5424a8b31624e7a476f959d552999f931e7c7 Apr 29 12:28:30 Author: Michal Marek Apr 29 12:28:30 Date: Thu Jun 9 16:35:41 2011 +0200 Apr 29 12:28:30 kbuild: Hack for depmod not handling X.Y versions Apr 29 12:30:22 and linux-yocto for qemux86-64 has the same, so this part is safe to ignore Apr 29 12:43:05 JaMa: ok, I'll look at do_rootfs:postinst Apr 29 12:44:19 ah, "NOTE: No Kernel Modules found, not running depmod" Apr 29 12:45:59 well, it just looks for ".ko" files, so of course it won't find it Apr 29 12:46:45 ( meta/lib/oe/rootfs.py, line 313) Apr 29 12:48:52 https://paste2.org/zMI4LyB2 should fix it, I'll testit Apr 29 12:55:41 I think it worked. Now flashing image, to be sure Apr 29 13:00:44 yes, the patch works Apr 29 13:01:10 do you want me to submit it to oe-core or will you? Apr 29 13:02:40 Well, as I'm not on the mailing or anything, it would be easier for me if you can do it; but I can do it too if you don't have time Apr 29 13:03:29 ok, will do Apr 29 13:03:56 thanks :) Apr 29 13:05:52 getting it backported to gatesgarth quickly might be more difficult, maybe add some dummy .ko file to work around this? Apr 29 13:06:40 yes, that can work Apr 29 13:06:49 or adding some other external kernel module which might produce .ko by default, does anyone need wireguard-module? :) Apr 29 13:13:09 oooh wait, I don't think it worked Apr 29 13:13:21 I saw the UI on pinephone so I thought it was ok Apr 29 13:13:37 but actually I don't have any modules loaded, and the UI is just software rendering Apr 29 13:13:53 ups, quickly v2 :) Apr 29 13:14:04 though I didn't see the message in the log Apr 29 13:14:53 so maybe depmodwrapper doesn't work, I don't know yet Apr 29 13:17:59 JaMa: when the wrapper does "exec env depmod "$1" "$2" "$3" " , it's the depmod from recipe-sysroot-native, right? Apr 29 13:19:36 ... which seems to *not* have xz support... Apr 29 13:21:43 JaMa: so if I want the native kmod to also have xz support, how should I do that ?... or is it another mistake in kmod recipe ? Apr 29 13:22:09 ah, maybe I need another bbappend for the native recipe Apr 29 13:30:23 ok, looks better. now flashing to check on actual device Apr 29 13:37:55 JaMa: the patch actually works. But if the *native* kmod doesn't also have xz support, then depmod will silently fail... Apr 29 13:41:00 ah Apr 29 13:42:37 sorry for the confusion on the ML Apr 29 13:43:33 I've updated my PR for kmod to also enable xz on the native one Apr 29 13:45:22 merged Apr 29 13:45:43 thanks Apr 29 13:46:58 I'm testing https://github.com/webOS-ports/meta-webos-ports/pull/459 now, but at least anbox will need a bit more love Apr 29 13:47:56 it's been a while since our last tests of anbox Apr 29 13:48:44 true, it fails to build with gcc-11 so I'm updating it just to fix the build Apr 29 13:50:09 they have submodules dispatched quite a bit everywhere Apr 29 13:53:00 Tofe: Well on our 4.9 kernels we should be OK to try again in general ;) Apr 29 13:53:12 morphis said that on 4.x kernels things should work a lot easier Apr 29 13:54:15 Herrie: I'll give it a try on pinephone when I will have solved the remaining issues (sound, ssh over usb, flashing from recovery) Apr 29 13:59:20 JaMa: I confirm that https://bpa.st/UE2Q also works. That way we can do a "v2" that does the same thing, but it will be less confusing for the ML :) Apr 29 14:21:19 One issue resolved: ssh via usb works. The bug was ... me. I forgot to plug the other end of the USB cable :D Apr 29 14:27:33 Tofe: Sounds like a case of needing more coffee :P Apr 29 14:36:42 Tofe: Any thoughts on this, or JaMa can merge? https://github.com/shr-distribution/meta-smartphone/pull/132 Apr 29 14:43:01 he already acked it here I believe, will merge it soon Apr 29 14:44:36 Ah OK Apr 29 14:44:44 I need more coffee too it seems Apr 29 14:45:43 hehe we all Apr 29 15:59:29 Tofe: https://github.com/Halium/halium-devices/pull/301 Apr 29 16:05:35 Herrie: PR looks good. ready to merge? Apr 29 16:05:46 Tofe: Yeah I don't see why not Apr 29 16:19:43 any other changes anyone wants to merge before I kick new builds? that meta-pinephone PR? Apr 29 16:24:53 JaMa: merged, I'm good! Apr 29 16:49:01 anbox builds as well, will add that Apr 29 17:17:26 sound is back on pinephone, it was an alsa UCM mismatch Apr 29 17:17:40 modem isn't detected yet, though Apr 29 17:30:53 Tofe: Don't you need some patched oFono or something again? Apr 29 17:37:51 Herrie: no, the modem is configured via AT commands, then ofono just works iirc Apr 29 17:37:54 Seems they have some special manager for the modem: https://github.com/sailfish-on-dontbeevil/eg25-manager-1 Apr 29 17:38:00 For SFOS Apr 29 17:38:29 yes, I'm currently updating my scripts to match the ones of pmOS, which I think are the most up-to-date Apr 29 17:38:42 JaMa: is something like this correct: RDEPENDS_append_pinephone_${PN} = "eg25-manager" Apr 29 17:39:19 (one of the pinephone's udev rule is calling a script from the eg25-manager recipe) Apr 29 17:59:32 Tofe: Ah OK Apr 29 18:00:39 Herrie: actually, thanks for your link, because I've just discovered what eg25-manager is, and that it consumes the configuration files of pmOS that I didn't understand Apr 29 18:00:58 There's a SFOS branch Apr 29 18:01:02 That works with oFono I guess Apr 29 18:01:19 Because the main branch = ModemManager which we don't use Apr 29 18:03:20 I think the master branch can do both Apr 29 18:03:26 and sfos branch is the dev branch Apr 29 18:04:12 anyway, all this is very interesting, I'll need to add it to meta-pine64-luneos Apr 29 18:19:59 damn, even if we don't want to use ModemManager, this eg25-manager has a build dependency on it... Apr 29 18:20:15 well, on libmm-glib Apr 29 18:24:18 And I don't remember the details, but I kind of remember that ModemManager and ofono did conflict... ( see https://github.com/webOS-ports/meta-webos-ports/commit/22c68384496656edb510e1a6d4df59a22376d83c ) Apr 29 18:25:04 not ofono, but connman Apr 29 18:25:31 sfos branch has some recent commits to build without libmm-glib Apr 29 18:25:54 https://github.com/sailfish-on-dontbeevil/eg25-manager-1/commit/40136c2a520d3be681aac4201a0e8ef324616140 Apr 29 18:25:56 ooh thanks, I missed this Apr 29 18:26:18 Which is not in main (yet)( Apr 29 18:26:35 great, I'll use that branch then Apr 29 18:29:48 the very last commit introduced a mistake, but the commit just before is fine for us (i.e. it'll never add modemmanager :) ) Apr 29 18:39:35 Tofe: maybe PR a fix for them :P Apr 29 18:40:19 actually I just discussed with piggz; the real upstream is https://gitlab.com/mobian1/devices/eg25-manager Apr 29 18:40:26 and it's fixed there Apr 29 18:46:47 ah ok Apr 29 19:56:39 Tofe: RDEPENDS_${PN}_append_pinephone = " eg25-manager" Apr 29 20:02:25 JaMa: ah, thanks Apr 29 23:36:07 21:35:57 Error closing trace writer: write /home/jenkins/workspace/halium-luneos-9.0-build/halium-luneos-9.0/out/build.trace.gz: no space left on device Apr 29 23:36:13 will check bonaire Apr 29 23:40:31 Herrie: Tofe: can I delete 5.1, 7.1 halium builds or do you plan to do more builds with them in future? Apr 29 23:40:34 19G halium-luneos-5.1-build Apr 29 23:40:35 50G halium-luneos-7.1-build Apr 29 23:41:03 it's not much, but whole /home has 250G so it's significant portion of that Apr 29 23:43:44 155G halium-luneos-9.0-build Apr 29 23:44:34 we might need to build 9.0 in multiple stages for various MACHINEs or just create separate jobs for them like we have for OE and rsync and cleanup between them Apr 29 23:45:26 3.9G halium-luneos-9.0-build/halium-luneos-9.0/external Apr 29 23:45:27 6.0G halium-luneos-9.0-build/halium-luneos-9.0/vendor Apr 29 23:45:27 13G halium-luneos-9.0-build/halium-luneos-9.0/prebuilts Apr 29 23:45:27 114G halium-luneos-9.0-build/halium-luneos-9.0/out Apr 29 23:46:05 1.3G halium-luneos-9.0-build/halium-luneos-9.0/out/host Apr 29 23:46:05 31G halium-luneos-9.0-build/halium-luneos-9.0/out/soong Apr 29 23:46:07 81G halium-luneos-9.0-build/halium-luneos-9.0/out/target Apr 29 23:47:14 29G halium-luneos-9.0-build/halium-luneos-9.0/out/soong/.intermediates Apr 30 00:04:43 haven't tested this, but something like this might help: https://github.com/webOS-ports/jenkins-jobs/pull/19 **** ENDING LOGGING AT Fri Apr 30 02:59:56 2021