**** BEGIN LOGGING AT Sat May 12 03:00:04 2018 May 12 05:21:39 Morning! May 12 05:21:42 Tofe: Nice! May 12 07:27:22 Morning! May 12 07:41:35 Tofe: Morning! May 12 07:54:48 Morning! May 12 08:06:32 Morning! (: May 12 08:32:44 morning May 12 11:58:20 Herrie: bad news: our old Mesa doesn't support enough EGL extensions May 12 12:00:17 time to get rid of the old mesa :) May 12 12:00:48 I've finished rebasing our Qt patches for 5.11-beta3 and once my qemu upgrade is finish I'll return to looking into getting mesa work with sumo May 12 12:01:43 JaMa: great :) May 12 12:02:19 well, to be precise, it's the wayland platform for Mesa that doesn't support it May 12 12:12:26 now I'm playing with wayland 1.15 and mesa 18.0.2 May 12 12:14:37 Herrie: little status: https://paste2.org/Z2PbmNzw May 12 12:15:26 JaMa: I'm not sure how I can check what EGL extension is available for these versions May 12 12:17:30 doesn't glxinfo show them? May 12 12:18:04 can you pastebin glxinfo output from your image, then I'll upload mine a bit later when I finish the build May 12 12:18:20 let me find in which package it is first :p May 12 12:19:12 mesa-demos IIRC May 12 12:19:20 ok thanks May 12 12:19:53 ah, only exists for x11 May 12 12:20:09 makes sense, we don't have glx here actually May 12 12:20:38 strange, I though I've added few PACKAGECONFIGs to make it usable even without x11 May 12 12:21:06 maybe that's only for the newer mesa package? May 12 12:21:46 anyway, anbox gave me that info for my current system (see previous pasted link) May 12 12:22:29 I see some GL_OES_EGL_image, GL_OES_EGL_image_external, not sure if that's what anbox looks for May 12 12:22:55 ok, let me check if I get glxinfo first May 12 12:23:15 it looks for the former first, in the GLES extensions, then checks what's avaiable in the EGL extensions May 12 12:24:09 --> check is done with that code: https://bpaste.net/show/406f9ec4c015 May 12 12:24:40 and your anbox branch is what I can use to include it in my Yocto 2.6 image? May 12 12:25:26 well, it's rebased on top of Herrie's work for webosose, but I can rebase it on top of pyro, that might trigger fewer surprises May 12 12:25:36 it's quite independent anyway May 12 12:25:58 (and needs cleanup too) May 12 12:26:06 is that nizovn's anbox commit enough or do I need morphis's anbox layer as well? May 12 12:26:33 well my build with gcc-8 might show few more surprises than this rebase :) May 12 12:26:58 you need the five commits, nizovn's plus my 4, and nothing else May 12 12:27:06 ok May 12 12:27:15 will try later May 12 12:27:24 note that I had quite some hard time with gtest May 12 12:27:39 which is present twice in our layers May 12 12:28:08 time to drop the version in meta-webos-ports and use meta-oe one? May 12 12:28:41 webOS/LuneOS recipes (like db8 etc) are based on source-only gtest-1.7.0, but Anbox and a couple of depends rely on 1.8.0, linking with the binary lib May 12 12:29:22 I tried dropping gtest 1.7.0 yes, but that'll require a migration for db8 it seems (at least) May 12 12:29:45 ./meta-webos/recipes-upstreamable/gtest/gtest_%.bbappend May 12 12:29:50 I might have some solution for that :) May 12 12:30:22 :) May 12 12:31:26 Tofe: try with this https://bpaste.net/show/057dc8516dc6 May 12 12:31:52 JaMa: also tried to sketch a "compatibility" patch for cmake webos module: https://github.com/Tofee/meta-webos-ports/blob/anbox/meta-luneui/recipes-webos/cmake-modules-webos-native/0001-gtest-backward-compatibility.patch May 12 12:32:30 oooh nice May 12 12:33:09 I didn't see there was a fuse_gtest_files.py script May 12 12:33:16 I'll try this asap :) May 12 12:34:38 not sure about anbox, but this .bbappend is compatible with db8 usage May 12 12:35:53 we'll see; but it adds things, so what worked should still work May 12 12:37:42 yup May 12 12:38:46 ah, build error, that's strange -- but it's the same as when I tried with gtest 1.8 on my side May 12 12:39:02 let me see if I use the latest version of db8 May 12 12:41:30 let me check if I sit on some db8 fixes as well May 12 12:42:14 db8: link with libatomic to fix qemuarm build failure May 12 12:42:22 nothing related to gtest May 12 12:42:29 ok May 12 12:43:10 I'm still trying to determine where my version comes from: SRC_URI = "${WEBOS_PORTS_GIT_REPO}/${PN}-1;branch=webosose" but no db8-1 on our GH May 12 12:43:24 ls2-helpers: add temporary patch to fix build with gtest 1.8.0 May 12 12:43:33 this is the only gtest related fix I see in my branches May 12 12:43:54 it might be already included in your downloads directory and Herrie removed it from github May 12 12:44:10 that would be annoying :D May 12 12:44:38 but maybe he imported the branch on db8 itself May 12 12:45:18 you know what, I'll rebase my anbox stuff on top of pyro. No use in piling up issues May 12 12:45:49 agreed May 12 12:45:58 are you sure that the db8 issue was caused by gtest upgrade? May 12 12:46:42 well I can build my image with gtest 1.7.0, so it does look like the origin of the issue May 12 12:46:42 the db8 had some issues with newer components we might use in pyro, e.g. I've report.h: fix build with boost-1.65.1, RefCountTest: Fix build with gcc7, not sure if these 2 commits were included in OSE May 12 12:46:54 ok, fair enough May 12 12:48:25 https://bpaste.net/show/e18b34e15791 this kind of errors shows up May 12 13:09:37 Tofe: How is db8 affecting Anbox May 12 13:09:42 I.e. it builds OK here with OSE May 12 13:10:04 Herrie: it is indirect, through a conflict for gtest's version May 12 13:10:39 Ah ok May 12 13:10:46 I got 1.7.0 in my build May 12 13:11:02 I remember I looked at 1.8.0 at some point but errors were way over my head May 12 13:11:07 So decided to not touch it May 12 13:11:19 Tofe: I could try to build anbox at my end May 12 13:11:25 exactly, and anbox wants 1.8.0, plus that's the version in meta-oe currently so our current state isn't clean May 12 13:11:39 Just I have issues with nyx-modules May 12 13:11:56 just wait a little bit, I'm doing some cleaning May 12 13:12:09 (and checking I didn't break anything obvious) May 12 13:12:53 I.e. I dropped teh inherit of "webos_nyx_module_provider" because I don't like how it implements the modules May 12 13:13:00 The way we do it with the cmake files is nicer May 12 13:13:09 :) probably yes May 12 13:14:00 JaMa: looks like db8 from OSE is broken with gtest 1.8.0, somehow May 12 13:14:16 I.e. this is nasty: https://github.com/webosose/meta-webosose/blob/master/meta-webos/conf/distro/include/webos-preferred-providers.inc#L73 May 12 13:14:19 the one with use for pyro is fine with your bbappend May 12 13:15:16 And https://github.com/webosose/meta-webosose/blob/master/meta-webos/recipes-core/packagegroups/packagegroup-webos-extended.bb#L21 May 12 13:15:41 Herrie, JaMa: https://github.com/Tofee/meta-webos-ports/commits/anbox (based on pyro) and https://github.com/Tofee/meta-webos-ports/commits/anbox-webosose (based on webosose work) May 12 13:16:11 Tofe: interesting, then I should be seeing db8 failures in my internal webOS OSE builds as well, but I haven't seen this one May 12 13:16:48 Tofe: Tried to do a -c clean and/or -c cleanall? May 12 13:16:48 ok, will try to pull it into my build once the current one is finished (will be a while) May 12 13:16:50 it could be my OSE branch is faulty May 12 13:17:11 don't use cleanall, cleansstate is all you need May 12 13:18:09 I might need to pull the latest commits from Herries's branch May 12 13:18:16 I didn't check if there was anything new May 12 13:20:51 ok I was missing some recent work, let's try again db8 May 12 13:22:38 Tofe/JaMa: Issues with nyx-modules in do-install: https://bpaste.net/show/a81c9df9e1c5 May 12 13:22:45 All good! false alarm then :) May 12 13:23:13 Not really sure what I'm missing there tbh May 12 13:24:35 Herrie: make: *** No rule to make target 'install'. Stop. May 12 13:24:41 Tofe: Yeah May 12 13:24:47 But I'm not sure where I should define that May 12 13:25:01 Herrie: meaning the Makefile misses this target May 12 13:25:36 where are the nyx-module sources taken from, in your case? May 12 13:26:24 Tofe: See my comit a few mins ago May 12 13:26:31 in meta-wop May 12 13:26:50 ah yes ok May 12 13:27:16 herrie/webOS-OSE commit? May 12 13:28:14 JaMa: yes, and SRC_URI now points to https://github.com/herrie82/nyx-modules/tree/herrie/msm_mtp May 12 13:28:52 Herrie: ah it's a cmake makefile May 12 13:29:36 Herrie: why did you drop webos_nyx_module_provider inherit? May 12 13:30:08 without NYX_MODULES_REQUIRED there might be nothing to install and cmake didn't generate the install target May 12 13:30:50 JaMa: Well we never had NYX_MODULES_REQUIRED previously before OSE May 12 13:30:55 Need to check that one May 12 13:31:25 Just want happy with the OSE approach since it doesn't really work with multiple targets well it seems May 12 13:31:37 Happy for other suggestions on how to do this May 12 13:32:31 cmake per target is clean imho May 12 13:32:40 Where you specify what to build May 12 13:32:53 didn't we discussed this one already? yes, the OSE way is inefficient May 12 13:33:49 the cleanest option is to package modules separately and build all modules for TUNE_PKGARCH (instead of MACHINE specific list of modules) and then install the right subset from MACHINE specific packagegroup May 12 13:42:21 JaMa: Just not sure how to do that May 12 13:42:34 JaMa: I should keep the inherit in general? May 12 13:43:04 I just thought that was OSE specific to deal with the modules implementation there May 12 13:43:20 Or I misunderstood (quite likely :p) May 12 13:45:01 for the nyx-modules repo you're currently using you need to pass the NYX_MODULES_REQUIRED parameter to cmake and that's what the bbclass does May 12 13:45:32 JaMa: Ah May 12 13:45:42 That explains a lot May 12 13:46:46 and for packaging them separately I would use do_split_packages function + PACKAGES_DYNAMIC (see e.g. connman how to do that) May 12 13:47:08 then you can remove NYX_MODULES_REQUIRED from nyx-modules and always build all available plugins May 12 13:47:35 and use VIRTUAL-RUNTIME_nyx-modules to specify a list of nyx-modules- packages you need for each MACHINE May 12 13:47:49 and then include this variable in one of MACHINE_ARCH packagegroups May 12 13:47:56 I might have understood, but I clearly need more coffee before re-reading you May 12 13:49:09 or I need more coffee to explain more clearly :) May 12 13:49:36 :) May 12 15:04:44 Out of curiosity I just tried to build anbox on hammerhead, but the needed kernel modules don't build May 12 15:05:03 (binder and ashmem) May 12 16:07:55 Tofe: Wouldn't you need a specific version of those modules compatible with 3.4? May 12 16:51:51 JaMa: Seems that when I add back the "inherit webos_nyx_module_provider" stuff works May 12 16:52:23 In my device specific .cmake files I do the same as the NYX_MODULES_REQUIRES does in the .class and/or cmake modules webos ;) May 12 16:53:05 JaMa: Ah wait May 12 16:53:07 That's an old IPK May 12 16:53:14 It builds but doesn't deploy anything now May 12 16:53:18 Was cheering too early May 12 16:53:22 But at least it builds :P May 12 17:49:41 Tofe: https://github.com/Tofee/meta-webos-ports/blob/7b95f1d684b07d634b05c6f5e21ab1a39e082175/meta-luneos/recipes-core/packagegroups/packagegroup-luneos-extended.bb#L144 May 12 17:49:51 Seems that you forgot binder there? ashmem is mentioned twice May 12 17:53:18 ah, hum, ahah, yes. May 12 17:53:43 weird that it worked so well May 12 19:09:34 Tofe: I guess I'll apply your patches and try my luck on Mido ;) May 12 19:35:35 Tofe: OK started a build May 12 19:35:41 I expect dragons somewhere but we'll see May 12 20:18:53 Tofe: https://bpaste.net/show/cad80f5e03ff May 12 20:19:22 And specifically: external/android-emugl/host/libs/Translator/GLcommon/GLDispatch.cpp:26:19: fatal error: GL/gl.h: No such file or directory May 12 20:21:10 Tofe seems I have a GLES/gl.h quite often May 12 20:25:53 Well GL/gl.h also exists at some places it seems May 12 22:53:15 Herrie: yes, I know that one, it's a missing DEPENDS on virtual/libEGL **** ENDING LOGGING AT Sun May 13 03:00:06 2018