**** BEGIN LOGGING AT Sat Apr 18 02:59:57 2020 Apr 18 16:24:40 RP: I sent patches for binutils/gcc few days ago which I would like to see in dunfell eventually after release give them a shot in master-next Apr 18 17:13:30 khem, hi, sorry for the bug-report by mail Apr 18 17:34:04 otavio_: perhaps you have an idea. recipes with this line no longer build with mx6 machines: COMPATIBLE_MACHINE = "(mx6|mx7|mx8)" Apr 18 17:35:36 I get: "incompatible with machine imx6qdlsabresd (not in COMPATIBLE_MACHINE)" - also happens with the nitrogen6x for example Apr 18 17:37:18 this is with poky master & meta-freescale master & meta-freescale-3rdparty master Apr 18 17:40:26 hm. MACHINE=imx8mmevk nice bitbake -e screen 2>&1 | grep -w "^MACHINEOVERRIDES" returns: MACHINEOVERRIDES="aarch64:imx:use-nxp-bsp:mx8:mx8m:imxdrm:imxvpu:imxgpu:imxgpu2d:imxgpu3d:mx8mm:imx8mmevk" Apr 18 17:40:42 while the same but with imx6dlsabresd as machine returns: MACHINEOVERRIDES="armv7a:imx:use-mainline-bsp:imx6qdlsabresd" Apr 18 17:41:05 even though I didn't set use-mainline-bsp anywhere... has this been observed before? Apr 18 18:04:08 ah - got it. imx-base sets IMX_DEFAULT_BSP with a "=" assignment, and not "=?" . I'll submit a patch. Apr 18 18:04:19 err, I meant "?=" Apr 18 18:35:06 ant__: no problem, but I think its better to have it in somewhere public so others can benefit too Apr 18 18:36:48 I have found the initial kernel header change Apr 18 18:37:13 but I think glibc solved it differently, must try Apr 18 18:39:18 but what to do with this old recipe? I could hack it to work with old/new header checking kernel version but surely there is some better preprocessor magic Apr 18 18:39:53 I can't imagine all recipes using older input.h had to change... Apr 18 18:41:13 (my fault I was not around for too long...sigh) Apr 18 19:09:43 khem: ok, I'm trying to take a bit of a break ;-) Apr 18 19:56:43 Perhaps a little janky.. but I am trying to pass in additional compiler flags to the linux kernel based on image_features Apr 18 19:57:18 Perhaps this should be done through Kconfig.. but.. Apr 18 19:57:32 Is there a packageconfig type feature for the kernel recipe? Apr 18 20:07:41 rabbit9911: it doesn_t work like that. image_features are de facto only visible in the image recipe Apr 18 20:08:53 rabbit9911: distro_:features woukd do the trick. Apr 18 20:16:05 What if I have one piece of hardware that can run different images with different kernels? Apr 18 20:16:55 I was hoping to override the kernel version in the image config... Have not tried that yet. Apr 18 20:18:21 then you have to build for two DISTROS, in a nutshell. Apr 18 20:19:34 rabbit9911: https://twitter.com/TheYoctoJester/status/1217166071519744000 Apr 18 20:20:45 rabbit9911: your kernel recipe has totally no idea whats in the image recipe. plus, you can build the kernel freestanding (just like any other recipe) too. so your line of thinking is flawed. Apr 18 20:22:35 That makes sense... Apr 18 20:23:05 What is the method to turn on off features for the kernel? When added like this KERNEL_FEATURES_append = " test.scc" Apr 18 20:24:29 in the machine or distro config, depending on the scope Apr 18 20:25:19 rabbit9911: heres an example for the MACHINE scope: https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/conf/machine/include/qemu.inc Apr 18 20:26:32 I see. So appending to KERNEL_FEATURES actually enables it too Apr 18 20:27:49 rabbit9911: should, yes. the key point is that it has to be done in either the kernel recipe itself, or in a conf file. not in another recipe (and images recipes are, well, recipes too) Apr 18 20:44:32 Okay I will dwell on this! Thanks! **** ENDING LOGGING AT Sun Apr 19 02:59:57 2020