**** BEGIN LOGGING AT Sun Jul 24 02:59:57 2022 Jul 24 08:05:19 Morning! Jul 24 08:10:10 Morning! Jul 24 08:10:10 Herrie: great ! Jul 24 08:10:17 I'll test it asap :) Jul 24 08:20:50 mmh I see the text of the boot, it seems to go quite far, but no UI Jul 24 08:21:49 and I see the usb device, but can't ssh in :/ Jul 24 08:29:32 Tofe: Well that's not a bad start to start with I guess Jul 24 08:30:06 Flashing here too Jul 24 08:30:53 In the meanwhile I found "Topaz3G-6050A2428701-MB-X01 20101012.pdf" as well. Which is fo the 4G regular TP I guess Jul 24 08:32:58 Tofe: I sent the PR's too for both meta-smartphone and meta-webos-ports Jul 24 08:36:07 I put the schematics on the regular place too Jul 24 08:36:10 ah ! after a while I could ssh in ! Jul 24 08:36:15 let's go then Jul 24 08:37:50 Tofe: Ah so it's just some delay somewhere Jul 24 08:38:18 configurator seems to eat up all the CPU Jul 24 08:38:46 Well configurator does quite a bit on first boot Jul 24 08:38:50 So that's not surprising Jul 24 08:39:11 But it shouldn't be for long Jul 24 08:39:18 surface-manager[4026]: [] [pmlog] surface-manager LSM {} (null), Could not find the Qt platform plugin "hwcomposer" in "" Jul 24 08:39:20 easy one Jul 24 08:39:58 of course it's a new file for surface-manager so I didn't have it in my rebase Jul 24 08:40:51 also, I think pulseaudio failure is getting in the way of the boot sequence here Jul 24 08:41:24 Tofe: Ah yeah that one should be easy Jul 24 08:42:59 next step: surface-manager[12116]: [] [pmlog] surface-manager LSM {} (null), Could not initialize egl display Jul 24 08:43:16 I'll just have a look in luna-next.conf Jul 24 08:45:05 there's also modprobe[16077]: modprobe: FATAL: Module bcmdhd not found in directory /lib/modules/5.6.0-luneos Jul 24 08:46:00 Apr 28 18:10:43 hammerhead surface-manager[15998]: [] [pmlog] surface-manager LSM {} (null), Failed to load EGL device integration "eglfs_kms" Jul 24 08:46:00 Apr 28 18:10:43 hammerhead surface-manager[15998]: [] [pmlog] surface-manager LSM {} (null), EGL library doesn't support Emulator extensions Jul 24 08:46:09 meh, looks like an hardcoded stuff here Jul 24 08:46:57 ah, no, there's a kernel oops in msm Jul 24 08:47:33 in msm_gem_free_work Jul 24 08:56:02 Tofe: Try a newer kernel? Jul 24 08:56:10 Tofe: Will you push the lsm fix to my branch? Jul 24 08:56:22 ok yes Jul 24 08:59:48 for the service file I'm not sure yet how to do that properly Jul 24 09:00:44 ah, but wait, pulseaudio is actually active Jul 24 09:00:51 so it should be fine for that part Jul 24 09:06:45 My flash got stuck on flashing kernel, so trying again Jul 24 09:07:45 I think I'll being a build for mainline hammerhead, it'll be much easier to debug Jul 24 09:11:14 but having a very recent mesa and an older kernel probably isn't a very good idea for this kind of WIP stuff Jul 24 09:11:29 Tofe: Upgrade to 5.16? Jul 24 09:11:44 yes, let's try that Jul 24 09:12:08 You want to have a stab at it, or should I do it? Jul 24 09:13:59 I can have a try Jul 24 09:14:35 It should be straight forward enough. Just take 5.16 and add the patches right? Jul 24 09:16:51 right Jul 24 09:35:12 mmh https://git.yoctoproject.org/yocto-kernel-cache/log/?h=yocto-5.16 doesn't exist Jul 24 09:35:50 Take master? Which seems to track 5.19 Jul 24 09:36:44 We could do 5.19 Jul 24 09:36:51 Patches don't seem that hard to rebase if needed Jul 24 09:37:42 Seems all are still needed based on my quick review Jul 24 09:37:49 None of the patches are now upstream yet Jul 24 09:52:04 I'm doing 5.15 instead, not a big fan of moving targets Jul 24 09:52:23 I mean kernel is still 5.16 Jul 24 09:56:20 Tofe: OK Jul 24 10:01:24 FWIW for 5.19 we don't need the first 2 patches anymore: https://github.com/torvalds/linux/commits/master/arch/arm/configs/qcom_defconfig Jul 24 10:02:34 ok, let me try Jul 24 10:03:29 Recipe for you: https://bpa.st/2USQ Jul 24 10:03:33 Image should be ready soon Jul 24 10:04:16 0000 not needed either ?... Jul 24 10:06:27 Well it's a cover letter Jul 24 10:10:21 Tofe: luneos-dev-package-hammerhead-mainline-5.19-20220724.zip is in the usual place, includes the LSM fix as well Jul 24 10:15:03 ah great Jul 24 10:20:16 doesn't seem to boot Jul 24 10:20:20 like, at all Jul 24 10:20:25 Yeah same here Jul 24 10:20:34 Well it uses standard defconfig v.s. our custom one Jul 24 10:20:37 That might be the issue Jul 24 10:20:46 ah yes, we do need ours Jul 24 10:24:02 Let me add that back Jul 24 10:24:10 I guess ideally I'd do a compare later on Jul 24 10:26:01 We should be able to use upstream one + extra cfg options Jul 24 10:26:41 yes, but with keeping in ming that the "snippet" approach, sometimes fails to configure properly Jul 24 10:26:46 mind* Jul 24 10:29:09 True Jul 24 10:29:22 You need a whole new image or just the kernel? Jul 24 10:29:55 I took our 5.6 defconfig, added the crypto bits. DEBUGFS was already set to Y there anyway Jul 24 10:36:40 I just built the new kernel, so I'm fine Jul 24 10:37:05 Tofe: luneos-dev-package-hammerhead-mainline-5.19-20220724v2.zip in case you need it Jul 24 10:37:12 ah, no, wait, I'll need the modules... Jul 24 10:37:19 ok I'll take your image :) Jul 24 10:37:28 Flashing here Jul 24 10:38:18 same Jul 24 10:39:25 Image is quite a bit bigger, so we'll see if it helps Jul 24 10:41:24 not booting either :( Jul 24 10:42:03 Yeah I see, same here Jul 24 10:42:07 Not sure what's up there Jul 24 10:44:02 I've reflashed with my own kernel build (i.e. should be the same), and I still don't see any usb device Jul 24 10:50:33 ok, it does boot Jul 24 10:50:52 it's just that the framebuffer isn't initialized properly I guess Jul 24 10:52:39 Tofe: Ah OK Jul 24 10:52:48 Well that could be some DTB stuff that is different Jul 24 10:53:02 and we still get Failed to load EGL device integration "eglfs_kms", but maybe not for the same reason Jul 24 10:55:21 https://paste2.org/0FBEhe13 here's the 5.19 journal Jul 24 10:57:12 point of reference for 5.6: https://paste2.org/UdL9I66h Jul 24 10:58:33 Tofe: Some of the tests might be of help to try? Jul 24 10:58:55 How do you ssh into the device? Jul 24 10:59:02 Just never did it on mainline without WiFi I think Jul 24 10:59:55 I don't see the device as usb, so no, nothing to do there Jul 24 11:02:33 How do you pull the journal then? Jul 24 11:04:42 /init_functions.sh: line 194: echo: write error: No such device Jul 24 11:05:17 I went in recovery, deleted var/log symlink, did mkdir -p var/log/journal to have it persistent Jul 24 11:05:22 in /data/luneos Jul 24 11:06:40 ok so echo "$(ls /sys/class/udc)" > $CONFIGFS/g1/UDC is failing Jul 24 11:07:06 might be just USB gadget issue then Jul 24 11:08:24 there is also the framebuffer that doesn't show the boot text, but maybe it's a consequence Jul 24 11:10:10 Tofe: Ah OK Jul 24 11:10:39 Tofe: Yeah when that works and you can get in, it's a lot easier to debug Jul 24 11:10:45 But the logs don't look that bad in general Jul 24 11:16:27 it's quite ok yes Jul 24 11:16:47 Tofe: For fb we might simply need: https://github.com/masneyb/linux/commit/d35b0a70fa97f121777ce365f719ae624070a337 Jul 24 11:16:51 Seems it's not in upstream Jul 24 11:18:30 The patch we carry seems to partially add it, just not the opp_table Jul 24 11:18:37 Not sure if that's critical though Jul 24 11:20:03 it's weird, one would think that "add support for GPU" is quite essential Jul 24 11:20:15 Yes Jul 24 11:20:18 ah, the gpu is in the DTS, ok Jul 24 11:20:24 no idea what opp is Jul 24 11:20:34 I'm still comparing the defconfig in input and in output Jul 24 11:26:12 I didn't see anything really obvious; there are some differences here and there, but I'm not sure about the meaning Jul 24 11:27:36 CONFIG_USB_CHIPIDEA_OF for instance isn't there anymore Jul 24 11:32:11 https://github.com/torvalds/linux/commit/95caa2ae70fdbf626335700d24400f47c12825ee maybe I should try enabling the MSM one Jul 24 11:38:16 No idea what this one does, but worth a try Jul 24 11:38:57 same here, but looks related to USB :p Jul 24 11:39:54 bingo ! Jul 24 11:40:35 Working? That's an easy fix then :P Jul 24 11:41:21 yes, I enabled CONFIG_USB_CHIPIDEA_MSM, IMX and GENERIC Jul 24 11:43:18 still waiting for sshd to show up Jul 24 11:43:37 it's quite long at first boot, something must be getting in the way Jul 24 11:45:42 I'm in Jul 24 11:46:50 no crash in dmesg so far Jul 24 11:47:32 Tofe: OK well that's good Jul 24 11:47:50 I guess you could try some of the egl tests that are provided? Jul 24 11:49:10 /dev/dri is empty, not good Jul 24 12:03:05 Tofe; Next step might be to compare the qcom_defconfig from 5.19 Jul 24 12:03:15 Maybe some more flags changed Jul 24 12:05:26 Let me do tht quickly in Excel :P Jul 24 12:09:22 I'm already suspecting the CONFIG_DRM_MSM_* Jul 24 12:10:02 Tofe: https://bpa.st/OTOQ Jul 24 12:10:15 These are the values differing. We have most set to Y and upstream has them to M Jul 24 12:10:24 Let me now check which ones are in upstream that we don't have Jul 24 12:11:24 but qcom_defconfig is a dev board, and doesn't have a default screen output, isn't it? Jul 24 12:12:16 Tofe: That could be indeed Jul 24 12:12:55 Upstream has these which we don't have at all: https://bpa.st/LFRQ Jul 24 12:16:35 it's my lucky day: now I see the boot text scrolling Jul 24 12:16:46 What did you do? Jul 24 12:17:16 I enabled CONFIG_DRM_MSM_MDP4=y CONFIG_DRM_MSM_MDP5=y CONFIG_DRM_MSM_DPU=y Jul 24 12:17:29 now I get things in /dev/dri Jul 24 12:18:23 https://bpa.st/I2KA different but still failing Jul 24 12:20:29 It could be you need to force some MESA variable Jul 24 12:22:57 This is for Qt5 but might be something similar? http://lists.busybox.net/pipermail/buildroot/2018-November/237334.html Jul 24 12:25:09 https://bpa.st/2USQ patches don't apply cleanly, but you can use http://dpaste.com/2GUBP7AFV Jul 24 12:28:21 should I merge both PRs? I know it's WIP but with hammerhead already broken, it doesn't make it worse :) my only concern is https://github.com/shr-distribution/meta-smartphone/pull/144/commits/35ff7d15fbaecf09d5c6f0a6db2a39f61d12a464 as it affects other MACHINEs as well Jul 24 12:29:41 Tofe: "The problem is that the build system does not select the correct eglfs backend. In my case, the solution was to export Jul 24 12:29:41 QT_QPA_EGLFS_INTEGRATION=none" Jul 24 12:32:03 JaMa: I think we should be good for the merge Jul 24 12:32:13 THere will be some defconfig updates later on etc, but yeah Jul 24 12:37:28 ok, done Jul 24 12:37:48 updated 5.19 pushed into your branch Jul 24 12:40:09 JaMa; Thnx Jul 24 12:48:25 did I miss something? I've build tested just kernel before merging, but now whole image fails with luna-sysmgr-conf.bb: Unable to get checksum for luna-sysmgr-conf SRC_URI entry luna-platform.conf: file could not be found Jul 24 12:48:50 find meta-webos-ports/ -name luna-platform.conf | grep hamm Jul 24 12:48:51 meta-webos-ports/meta-luneos/recipes-webos/luna-sysmgr-conf/luna-sysmgr-conf/hammerhead-halium/luna-platform.conf Jul 24 12:52:15 JaMa: Hmmz weird, it built OK here Jul 24 12:52:31 I build hammerhead and hammerhead-halium devices Jul 24 12:52:35 Both succeeded Jul 24 12:53:12 Seems it's indeed missing somehow, weird Jul 24 12:54:25 JaMa: Well you can just duplicate the one frm hammerhead-halium into hammerhead Jul 24 12:58:28 JaMa: https://github.com/webOS-ports/meta-webos-ports/commit/abbde2912c24d0d28b7a822e3e343ab502e645fc Jul 24 12:59:28 yes, that's what I did to start the build will merge it after langdale build is finished Jul 24 13:02:04 I pushed it for kirkstone Jul 24 13:02:08 So you can cherry pick it Jul 24 13:02:29 done Jul 24 13:17:30 sorry, was at lunch Jul 24 13:17:41 back to work, let's try the export Jul 24 13:18:38 new error: surface-manager LSM {} (null), EGL Error : Could not create the egl surface: error = 0x300b Jul 24 13:18:42 hmm I'm surprised that milla is running and I was able to ssh, but https://build.webos-ports.org/ doesn't work, that seems like just some smaller configuration issue now (last time I've tried milla was down, so I was assuming still ongoing HW issues) Jul 24 13:21:10 Tofe: And some of the gl tests in /usr/bin? Jul 24 13:21:53 To verify if it's lsm or qt related? Jul 24 13:22:34 glmark2-es2-drm shows exactly the same kind of errors as in my paste Jul 24 13:23:12 http://build.webos-ports.org/ fixed with sudo ln -snf /home/jenkins/htdocs/builds/ /media/ra_build_share/wop-build Jul 24 13:24:27 JaMa: great ! Jul 24 13:25:18 JaMa: nice Jul 24 13:25:39 mmh "GL_RENDERER: softpipe" Jul 24 13:26:12 I guess it's just some old version of builds (from before switch to ssd), but better than nothing Jul 24 13:26:37 it doesn't have hallium-9 and testing was using zeus Jul 24 13:26:50 Apr 3 2020 Jul 24 13:26:55 :/ Jul 24 13:27:14 yes, things are a bit old in there Jul 24 13:31:28 https://github.com/waydroid/gbinder-python/pull/2/ is strange, I was building it not long time ago (after few fixes in meta-webos-ports) Jul 24 13:33:09 JaMa: I bumped srcrev locally which could be cause Jul 24 13:33:38 https://github.com/webOS-ports/meta-webos-ports/commit/ed484d3bef3816110bc3f3d47fce3c70050cc978 Jul 24 13:33:53 at this point bdist_wheel was working for me Jul 24 13:34:51 https://github.com/webOS-ports/meta-webos-ports/commit/528204c26b1550f54b7b93cbc9c32352ab5b0839 with this as well Jul 24 13:37:49 there is only one new commit which wasn't in SRCREV I was using: https://github.com/waydroid/gbinder-python/commit/bcf3debcc171108d02be49e22bfbab5c15d216e2 maybe some difference in host python or something like that Jul 24 13:38:57 but it should be using python3native.bbclass, so not sure, will build test after hammerhead build is finished (I wanted to check why we need that patch for gold, imho it shouldn't be needed) Jul 24 13:41:00 JaMa: I'll PR my pending changes tonight Jul 24 13:41:24 also gbinder I updated Jul 24 13:41:33 Anyway the patch will be needed going forward Jul 24 13:56:53 yes, but setuptools will also get dropped if I understood the plan correct, so going forward they will need to switch to something else https://lists.openembedded.org/g/openembedded-core/message/166187 Jul 24 13:57:55 python is almost like node in this regard nowadays Jul 24 13:58:58 yes, both quite terrible Jul 24 14:31:05 Herrie: for me, https://github.com/masneyb/nexus-5-upstream/blob/master/out-of-tree-patches/upstream-patches/v5.16/0004-ARM-dts-qcom-msm8974-hammerhead-add-support-for-GPU.patch is incorrect : the gpu adreno node isn't located in the "otg DTS section Jul 24 14:31:11 "otg" Jul 24 14:32:12 I've changed it to top-level: " &gpu { status = "okay"; }; " and now I have other errors in dmesg when it really tries to use adreno Jul 24 14:32:25 "ocmem fdd00000.ocmem: could not lock: -5" Jul 24 14:32:57 followed by "msm_mdp fd900100.mdp: failed to load adreno gpu" Jul 24 14:33:33 so I checked OCMEM in defconfig and in DTS, but both seems ok compared to 5.6 Jul 24 14:33:55 in 5.6 we had: "ocmem fdd00000.ocmem: 8 ports, 3 regions, 2048 macros, interleaved" Jul 24 14:42:29 Tofe: Could be some stuff got moved in between 5.6 and 5.19 in dtb somehow Jul 24 14:42:43 I saw quite some dtb changes in the tree Jul 24 14:42:48 maybe Jul 24 14:43:12 You could check the log in arch/arm/boot for the 8974 dtbs Jul 24 14:43:41 Maybe something obvious got changed generically but not yet in hammerhead specific one Jul 24 14:50:31 Tofe: Stuff like this: https://github.com/torvalds/linux/commit/03110b46c99bb0c712f46bec660b1c3f674ce100 Jul 24 14:51:04 I'm on phone so not of that much help right now :P Jul 24 14:52:57 https://github.com/torvalds/linux/commits/03110b46c99bb0c712f46bec660b1c3f674ce100/arch/arm/boot/dts/qcom-msm8974.dtsi Jul 24 14:53:17 I'm looking at the same things yes Jul 24 15:00:00 https://github.com/torvalds/linux/commit/8ccecf6c710b8c048eecc65709640642e5357d6e ? Jul 24 15:02:19 I'm looking more around ocmem Jul 24 15:02:24 but maybe I'm wrong Jul 24 15:02:59 Tofe: These are pending merge: https://patchwork.kernel.org/project/linux-arm-msm/list/?series=&submitter=&state=&q=8974&archive=&delegate= Jul 24 15:03:17 Should land in 5.20 but didn't see anything related Jul 24 15:03:37 https://patchwork.kernel.org/project/linux-arm-msm/patch/20220607171848.535128-7-krzysztof.kozlowski@linaro.org/ ooh I'll try Jul 24 15:03:51 Oh well: https://patchwork.kernel.org/project/linux-arm-msm/patch/20220607171848.535128-7-krzysztof.kozlowski@linaro.org/ Jul 24 15:04:01 Ah you found the same Jul 24 15:05:37 I guess this one doesn't hurt either: https://patchwork.kernel.org/project/linux-arm-msm/patch/20220606160421.1641778-1-luca@z3ntu.xyz/ Jul 24 15:05:41 thought the ranges weren't there in v5.6 Jul 24 15:06:54 Tofe: Could be requirement was added later Jul 24 15:07:56 I've built gbinder now for qemux86-64 with langdale without your patch fine Jul 24 15:09:22 Herrie: ocmem fixed ! now, it didn't fix adreno... Jul 24 15:09:51 JaMa: weird Jul 24 15:09:57 but I get "[drm:a3xx_gpu_init] *ERROR* No memory protection without IOMMU", so it might just be a kernel parameter to pass Jul 24 15:10:28 I tried multiple times but not succeeding without the patch Jul 24 15:10:40 I'll paste my updates later so you can check Jul 24 15:13:12 ok, I'll have to guess the parameter name, but it does help, now I have a proper adreno drm I think Jul 24 15:17:05 Tofe: OK Jul 24 15:17:48 Tofe: Some clues might be found in the 5.5 patches? Jul 24 15:17:56 https://github.com/masneyb/nexus-5-upstream/blob/master/out-of-tree-patches/upstream-patches/v5.5/0000-cover-letter.patch#L21 Jul 24 15:19:10 well, the current status is actually similar to v5.6 Jul 24 15:19:18 https://github.com/masneyb/nexus-5-upstream/blob/master/out-of-tree-patches/upstream-patches/v5.5/0021-HACK-allow-running-without-IOMMU.patch Jul 24 15:19:34 for IOMMU it's solved, no pb here Jul 24 15:23:03 Herrie: it fails only in kirkstone it seems Jul 24 15:28:00 JaMa: OK Jul 24 15:28:16 Tofe: Well weird others seem to be able to boot stuff to UI? Jul 24 15:28:34 Herrie: you mean, pmOS ? Jul 24 15:28:45 Tofe: Yeah Jul 24 15:28:57 well there must be something we're missing Jul 24 15:28:57 Maybe we should compare their defconfig? Jul 24 15:29:09 Or other packaging bits for hammerhead Jul 24 15:30:45 https://gitlab.com/postmarketOS/pmaports/-/merge_requests/3159 Jul 24 15:32:20 Site doesn't load here but seems something that we might need Jul 24 15:32:34 no it's just for their spreadsheet Jul 24 15:35:23 Herrie: we're missing PACKAGECONFIG_GRAPHICS:append:pinephone = " kms gbm" in our qtbase.bbappend Jul 24 15:35:29 well for hammerhead I mean Jul 24 15:35:42 Tofe: Ah yes Jul 24 15:35:46 That could be it Jul 24 15:36:07 Easy enough to fix in general Jul 24 15:36:14 I will do that tonight when back Jul 24 15:36:27 Will be quicker vs you building I guess? Jul 24 15:36:35 we'll see :) Jul 24 15:36:39 but probably yes Jul 24 15:37:48 I can try and see I can reach my anydesk from here :P Jul 24 15:38:15 well, no worries, I can move to other stuff at home too :p Jul 24 15:38:27 my qtbase build is started, not much more to do anyway Jul 24 15:38:41 and pause can be nice too :p Jul 24 15:40:26 Tofe: True Jul 24 15:40:42 I will see if I can get a bit further with tenderloin build tonight as well Jul 24 15:40:55 I guess the hammerhead bits should give some inspiration Jul 24 15:43:30 backporting https://git.openembedded.org/openembedded-core/commit/?id=0907866325 in kirkstone fixed the gbinder as well Jul 24 15:44:47 JaMa: OK Jul 24 15:45:05 4G here is slower than a 14k4 modem with full signal :/ Jul 24 16:07:44 Herrie: ah, good ol' times Jul 24 16:08:50 Tofe: can you paste your defconfig somewhere? Jul 24 16:08:58 Will include that too then Jul 24 16:10:44 And your updated patch(es)? Jul 24 16:10:48 For dtb Jul 24 16:10:52 one moment Jul 24 16:15:13 Herrie: I still have to rebase it, but here it is https://github.com/Tofee/meta-smartphone-1/commit/34cb9bd50ba480068d1edf2642ac6bd09edb0a6f Jul 24 16:16:13 so it's 0006-Fix-some-DTS-issues-Adreno-activation-and-OCMEM.patch and ANDROID_BOOTIMG_CMDLINE that do the trick Jul 24 16:18:10 Also opened this: https://github.com/masneyb/nexus-5-upstream/issues/11 Jul 24 16:22:28 Tofe: Thnx Jul 24 16:22:40 Yeah good to discuss there Jul 24 16:23:09 Might be worth mentioning that you're on 5.19rc7 Jul 24 16:25:54 qtbase built Jul 24 16:26:44 Tofe: Not even on my way home here :P Jul 24 16:26:49 So you beat me Jul 24 16:26:59 for once, yay Jul 24 16:29:27 well, LSM starts, but the screen stays the same Jul 24 16:29:40 no drama in dmesg Jul 24 16:29:51 I'll reboot. Jul 24 16:30:29 Tofe: Maybe try the tests again? Jul 24 16:30:48 also I only rebuilt qtbase, so it's a bit fragile maybe Jul 24 16:31:13 Yeah could be Jul 24 16:31:30 I should be home in about 45 mins and can do whole image Jul 24 16:35:00 "dsi_err_worker: status=c" in dmesg, doesn't look good Jul 24 16:49:41 https://github.com/postmarketOS/linux-postmarketos/issues/2 Jul 24 16:58:18 https://www.lkml.org/lkml/2019/11/12/296 Jul 24 17:24:50 Tofe: I'll change msm.vram to 48m instead of 300m as per https://github.com/masneyb/nexus-5-upstream/blob/master/build-kernel#L95 Jul 24 17:25:30 See https://github.com/masneyb/nexus-5-upstream/commit/cce61055a568ec0edeeb17508bed74ae34b03e6b Jul 24 17:25:35 That might explain the display issues Jul 24 17:31:22 Herrie: ooh that might be it Jul 24 17:33:22 though we do setup cma to 500m Jul 24 17:34:12 mesa fails Jul 24 17:36:29 drm.h not found: https://pastebin.ubuntu.com/p/XqZcrCnbfb/ Jul 24 17:36:35 Weird it's in the halium directory though :S Jul 24 17:37:11 halium ? for mainline hammerhead ? Jul 24 17:39:33 Well it's when I'm building hte mainline one Jul 24 17:39:54 Curious it's in the work/cortexa8t2hf-neon-halium-webos-linux-gnueabi/mesa-gl though Jul 24 17:40:43 Ah wait Jul 24 17:41:28 Seems I'm missing this in my local build: https://github.com/webOS-ports/meta-webos-ports/blob/kirkstone/meta-luneui/recipes-graphics/mesa/mesa-gl_%25.bbappend#L9 Jul 24 17:51:54 TofE: some more config issues, but getting there it seems Jul 24 18:07:52 Tofe: Seems my layer update was messed up somehow Jul 24 18:07:57 Redid it now Jul 24 18:10:45 And building Jul 24 18:11:00 Out for a quick dinner but should be ready when back Jul 24 18:17:16 I'm slowly build my own in the meantime :) Jul 24 18:20:47 Tofe: OK Jul 24 18:21:01 Still think I'll beat you in the process :P Jul 24 18:24:13 hehe Jul 24 18:43:22 don't tempt me for a race :) Jul 24 18:48:22 JaMa: LOL you beat me for sure Jul 24 18:50:31 Herrie: I was able to build hammerhead's 5.6 kernel without reversed-disable-gold-linker.patch, do you still reproduce the issue with gold locally? Jul 24 18:52:23 JaMa: I think we'll switch to 5.19 soon Jul 24 18:52:34 So we can forget about gold linker Jul 24 18:53:08 I had issues with Tenderloin and gold link but it was more my build environment Jul 24 18:53:19 After I put it in Yocto it was OK Jul 24 18:59:25 JaMa: I have the same status with 5.6 and 5.19, so we'll stick with the most recent Jul 24 19:02:38 Tofe: luneos-dev-package-hammerhead-mainline-5.10-20220724v3.zip Jul 24 19:03:47 Typo in the filename: luneos-dev-package-hammerhead-mainline-5.19-20220724v3.zip Jul 24 19:06:23 After kids are in bed, I'll have another stab at the tenderloin mainline build Jul 24 19:08:58 Tofe: It boots with log on display at least right away Jul 24 19:09:09 That's more than what I saw earlier today ;) Jul 24 19:10:17 Seems stuck @ surface-manager.service after it OK's Jul 24 19:10:23 Only failure I noticed during boot was WiFi Jul 24 19:10:28 No UI though yet here Jul 24 19:11:37 So I guess that's a similar status as you have Jul 24 19:21:50 please use savedefconfig to generate minimalized defconfigs, much easier to review the changes Jul 24 19:22:54 JaMa: No idea on how to do that ;) Jul 24 19:28:36 linux-lg-hammerhead-5.19-rc7+gitrAUTOINC+1e2deb71df_9614aacfbf-r0 do_fetch: Fetcher failure: Unable to find revision 1e2deb71df8339cc0bf5f17f43f0f1a475fa1dd7 in branch master even from upstream Jul 24 19:32:06 ups my bad Jul 24 19:36:17 why do we even need kernel-meta when using static defconfig? Jul 24 19:37:04 hmm Jul 24 19:37:05 | make: Entering directory '/OE/build/luneos-langdale/webos-ports/tmp-glibc/work/hammerhead-webos-linux-gnueabi/linux-lg-hammerhead/5.19-rc7+gitrAUTOINC+f55df88ad1_1e2deb71df-r0/linux-hammerhead-standard-build' Jul 24 19:37:09 | GEN Makefile Jul 24 19:37:12 | gold linker is not supported as it is not capable of linking the kernel proper. Jul 24 19:37:15 | scripts/Kconfig.include:56: Sorry, this linker is not supported. Jul 24 19:37:17 | make[1]: *** [/OE/build/luneos-langdale/webos-ports/tmp-glibc/work-shared/hammerhead/kernel-source/scripts/kconfig/Makefile:82: savedefconfig] Error 1 Jul 24 19:39:19 JaMa: Ah when we use static defconfig we don't need kernel-meta? Jul 24 19:39:24 That's good to know Jul 24 19:40:04 I mean whole linux-yocto, biggest advantage of linux-yocto (compared to just linux.inc) is dealing with config fragments and patches Jul 24 19:55:08 JaMa: With regards to config fragments, if I understood correctly we'd still need to provide the dependent defconfig flags as well right? Jul 24 19:55:36 It seems that for the dragonboard they do some magic with a kernel config script. I.e. you pass in some options to be enabled and it will enable the dependent flags as well Jul 24 19:56:02 I,e,: https://dflund.se/~triad/krad/makefiles/apq8060.mak Jul 24 19:56:19 $(CURDIR)/scripts/config --file $(config_file) \ Jul 24 19:56:19 --enable ARCH_MSM8X60 \ etc Jul 24 19:57:52 Or Yocto does magic there as well? Jul 24 19:58:09 I now just pulled a defconfig from previously compiled kernel, but ideally would generate it on the fly Jul 24 20:12:39 JaMa; For gbinder: https://github.com/webOS-ports/meta-webos-ports/commit/e6c3ca0c14e59b8091c5107bb0d2e35ff2199268 Jul 24 20:12:54 This is what I build now, which seems OK for Waydroid in runtime on qemux86-64 Jul 24 20:13:47 JaMa: Added this one so hammerhead build is more happy: https://github.com/webOS-ports/meta-webos-ports/commit/b8fc537d82ff18af872434d9fb8c2103cf6f00a6 Jul 24 20:17:37 There's a few minor tweaks needed to kernel as well to get it to work... Jul 24 20:17:45 But Tofe might have something better later Jul 24 20:18:24 But I guess you could cherry pick this one to your linux fork for now: https://github.com/Tofee/meta-smartphone-1/blob/34cb9bd50ba480068d1edf2642ac6bd09edb0a6f/meta-lg/recipes-kernel/linux/linux-lg-hammerhead/0006-Fix-some-DTS-issues-Adreno-activation-and-OCMEM.patch Jul 24 20:21:18 I already did Jul 24 20:21:28 that's what I was building now Jul 24 20:25:28 JaMa: Ah OK Jul 24 20:25:44 Did you pick those defconfig change + cmdline change too? Jul 24 20:28:13 yes, and then savedefconfig https://github.com/shr-distribution/meta-smartphone/commit/8f990fb17290bc304a7583dbfed50a6baadc867a Jul 24 20:34:33 Ah OK Jul 24 20:34:35 Nice Jul 24 20:34:40 A lot smaller indeed Jul 24 20:38:13 fix for gold in savedefconfig https://lists.openembedded.org/g/openembedded-core/message/168465 Jul 24 20:45:33 JaMa: Nice! Jul 24 20:48:11 but if you weren't using savedefconfig then maybe there is some separate issue still Jul 24 21:01:36 JaMa: Any thoughts on this? https://bpa.st/IHBQ Jul 24 21:01:45 I tried to do the same for tenderloin as I did for hammerhead Jul 24 21:01:49 But seems it doesn't work somehow Jul 24 21:25:12 did you remove halium from tenderloin? Jul 24 21:25:44 meta-android/conf/machine/include/meta-android-halium.inc sets mesa-gl Jul 24 21:29:24 I would say so Jul 24 21:30:30 Ah I might have found it... Jul 24 21:30:57 I still had "require conf/machine/include/meta-android-halium.inc" in tenderloin.conf ;) Jul 24 21:32:14 This works better :P Jul 24 21:42:59 damn, my webruntime isn't even built yet Jul 24 21:43:48 Tofe: Well I pasted an image for you Jul 24 21:43:55 Seems my Tenderloin image is almost built as well Jul 24 21:44:11 ok, well maybe I'll try tomorrow then Jul 24 21:44:59 Tofe: With the latest image I built, I get output on screen and it seems stuck on lsm after the service loads OK Jul 24 21:46:17 This is for Hammerhead of course, so seems similar to what you had Jul 24 21:48:21 yes, looks very similar to my current results Jul 24 21:49:32 Let me see how this uImage I produced with mainline kernel behaves Jul 24 21:58:51 Nothing much it seems Jul 24 21:58:57 OK that's for another day then too Jul 24 21:58:59 Bedtime here Jul 24 22:02:01 I guess I will do 1 more try just LOL Jul 24 22:02:08 Semes I forgot to specify the dtb in machine ;) **** ENDING LOGGING AT Mon Jul 25 02:59:56 2022