**** BEGIN LOGGING AT Fri Jan 20 02:59:56 2023 Jan 20 06:45:36 Morning! Jan 20 06:45:55 Tofe: ADB doesn't see my mido somehow, so I will find some other cables this morning Jan 20 06:46:02 Already installed required drivers **** BEGIN LOGGING AT Fri Jan 20 07:11:55 2023 Jan 20 08:20:58 Morning! Jan 20 08:21:09 oh, I almost forgot to start my tissot build, one moment Jan 20 08:23:27 Herrie: I begin to really appreciate your big ssd :) Jan 20 08:36:28 ^ build finished Jan 20 08:44:56 Tofe: Yeah I know when you build multiple machines and archs it's really nice to have space Jan 20 08:45:42 yes Jan 20 08:46:10 ok I'll try and flash tissot tonight, see what happens with these shaders Jan 20 08:47:38 I have Hammerhead charged here Jan 20 08:50:42 I looked at hammerhead's qt config, and GL ES 3.1 and 3.2 are also enabled, whereas it's not supported by Adreno 330; so this might be a lead, if I'm very lucky Jan 20 08:51:44 Tofe: This is for halium or mainline? Jan 20 08:52:09 both I think, but I just checked halium Jan 20 10:03:40 Does the Halium one give you any UI or nothing at all similar to Rosy? Jan 20 10:20:38 This is my log with grep for "surface": https://bpa.st/RPZFY Jan 20 10:20:40 On mido Jan 20 10:22:04 No UI, but backlight is on it seems Jan 20 10:24:33 test_egl_config: https://paste.ubuntu.com/p/5RSJkmQQps/ Jan 20 15:36:33 Herrie: I have the same status, screen is cleaned up, stays black, backlight seems to work Jan 20 15:37:40 Herrie: also, note that I had an infinite malloc loop in the gstreamer v4l2 module, make it impossible for LSM to really start Jan 20 15:38:31 so I just renamed the .so (whose complete name I forgot) and that way I got past gstreamer init Jan 20 15:39:04 it was something like /usr/lib64/libgstmodulev4l2.so Jan 20 15:53:08 Tofe: My ADB seems to get killed after a few minutes Jan 20 15:53:16 Might be due to above too? Jan 20 15:53:43 well, I had a boot loop because of the malloc getting into OOM Jan 20 15:53:55 after a few minutes, yes Jan 20 16:22:07 Tofe: No reboot on mido, but seems stuck just Jan 20 16:22:15 So might be the same in the end Jan 20 17:42:16 Flashing tissot now Jan 20 17:55:47 same as you, seems to get stuck, or in a deep cpu loop Jan 20 17:57:10 but looking at top, it's a big memory leak in LSM actually; growing by 200Mb each second! Jan 20 17:57:33 I'll mask surface-manager Jan 20 17:58:04 My son lost Tissot so I have only Mido left for now Jan 20 17:58:27 well, looks like it's the same symptoms Jan 20 17:58:35 let me see with gdb where the leak occurs Jan 20 17:59:34 Tofe: That's good, same fix probably for all 3 Xiaomi devices then Jan 20 17:59:46 let's hope so Jan 20 17:59:57 strace might give leads too? Jan 20 18:00:05 In case some file is missing Jan 20 18:00:18 Anyway memory leak should be properly fixed Jan 20 18:00:31 well, I'll begin with gdb Jan 20 18:00:43 in hammerhead it was an obvious backtrace Jan 20 18:01:30 but here I don't see the gstreamer v4l2 module, so maybe it's something else (maybe in gstreamer too ?) Jan 20 18:05:08 JaMa said gstreamer was buggy in OSE for it's own targets even Jan 20 18:05:25 I think we don't even have all OSE's patches yet either Jan 20 18:05:30 Herrie: /usr/lib/gstreamer-1.0/libgstvideo4linux2.so still in the loop Jan 20 18:05:38 I'll rename the .so and retry Jan 20 18:06:40 UI is here ! Jan 20 18:06:54 Ah that's good Jan 20 18:07:07 Then fix might not be too hard? Jan 20 18:07:17 Though on Rosy you didn't get it? Jan 20 18:07:53 I don't have a rosy to test anymore :/ my wife uses it nowadays Jan 20 18:08:09 but the fix should be similar Jan 20 18:08:27 for someone like fakeshell who tinkers with the image directly Jan 20 18:08:46 Tofe: Yeah true Jan 20 18:08:57 I might find myself a Tissot again Jan 20 18:09:12 Oh yeah whats going on Jan 20 18:09:19 I hear fixes Jan 20 18:09:30 fakeshell: more workaround, so far Jan 20 18:10:02 Nice i love those Jan 20 18:10:02 you can rename/delete /usr/lib/gstreamer-1.0/libgstvideo4linux2.so on the Qt6 rosy image (newest one), maybe that will get you to the UI Jan 20 18:10:24 Alright will try it out today Jan 20 18:10:26 it looks like there's a nasty memory leak going on there, leading quickly to an OOM Jan 20 18:11:50 Herrie: status on tissot: wifi didn't work, but the rest looks exactly like on qemu Jan 20 18:12:14 for wifi it's probably a small regression in our module loading script, or so Jan 20 18:14:20 mmh the kernel module fails to start; could be we solved this before and I don't remember Jan 20 18:14:40 ^ "VOS ASSERT in vos_start Line 926" Jan 20 18:16:20 " After reflash, wifi works" two years ago, similar issue Jan 20 18:19:07 Tofe: Well A1 funny stuff Jan 20 18:23:16 but still, it's a good progress; it means that on hammerhead, we might revive the halium build too if we manage to restrict Qt's GLES shader version Jan 20 18:24:00 otherwise we'll stick to mainline with this device, it can be a valid choice as well Jan 20 18:28:49 Tofe: That issue might be for Tenderloin too Jan 20 18:30:24 right Jan 20 18:44:08 Yeah i saw that OOM when i first tried it Jan 20 18:44:22 It filled up 4gb in about 2.5 minutes and then lmk l Jan 20 18:44:30 Lmk killed systemd Jan 20 18:48:10 yeah, the oomk strategy isn't always great :) Jan 20 19:03:54 Herrie: for tissot, I'm definitely missing some .bin file, like you did two years ago Jan 20 19:04:14 "wcnss_service: source file do not exist /data/misc/wifi/nvbin/msm8953_293_0x00_0x01_0x00_nv.bin" Jan 20 19:06:16 Tofe: Well that could be on Halium side somehow Jan 20 19:06:52 maybe we messed up our tissot image ? or did we revert to an older image when we lost our build.webos-ports.org binaries ? Jan 20 19:07:05 Ie we picked an older Halium image due to server disk issue that doesn't have this firmware file or not on the right location Jan 20 19:07:13 we agree :) Jan 20 19:07:18 It should be minor to fix anyway Jan 20 19:07:35 yes, I just have to remember how to build a Halium image Jan 20 19:07:53 Tissot always was a bit of a Frankenimage due to it's half Treble stuff Jan 20 19:08:12 Tofe: Well I have it here on my builder I guess Jan 20 19:08:18 I have some scripts Jan 20 19:09:07 You could pull the binary from TheMuppets GitLab and push it to image as workaround Jan 20 19:09:25 I'll get to the Halium bits probably over the weekend or Monday Jan 20 19:11:52 Either here: https://gitlab.com/the-muppets/proprietary_vendor_xiaomi/-/tree/lineage-17.1/msm8953-common Jan 20 19:12:17 or here: https://gitlab.com/the-muppets/proprietary_vendor_xiaomi/-/tree/lineage-17.1/tissot Jan 20 19:12:43 we didn't update the repo in Halium ? Jan 20 19:13:49 I have some scripts too, it just the switch between devices that can be a bit confusing Jan 20 19:27:50 mido booted here to ui fwiw Jan 20 19:28:46 After deletion of the .so Jan 20 19:34:40 ok good Jan 20 19:56:16 it's weird, the halium tissot kernel doesn't seem to build Jan 20 20:17:45 damn, the Halium build fails with too recent python :/ Jan 20 20:18:15 I guess I shouldn't have upgraded my chroot to 22.04 Jan 20 20:38:24 Tofe: Can't you force python 2 for Halium build? Jan 20 20:38:40 Takes a little tinkering probably Jan 20 20:39:13 But I recall I had python2 working here on 22.04 Jan 20 20:39:26 ok, will try Jan 20 20:41:26 Anyway it might be more rewarding to look into the hammerhead or cause of gstreamer issue Jan 20 20:41:46 The fw file for tissot is really minor tbh Jan 20 20:42:00 I agree Jan 20 20:43:55 I can break my head over halium while you do some digging into lower level stuff :D Jan 20 20:44:27 well I changed a bit my PATH to have it point first on a python2 symlink; we'll see if that's enough Jan 20 21:07:30 ok, build finished; but it doesn't look like the files are there Jan 20 21:09:45 it could also be a wrong lead. Anyway, I'll now switch back to gstreamer and/or hammerhead **** ENDING LOGGING AT Sat Jan 21 02:59:56 2023