**** BEGIN LOGGING AT Thu Sep 30 02:59:56 2021 Sep 30 07:55:27 good morning Sep 30 17:54:50 RP: the plot thickens, even if I compile the python3 on the machine with OE image, the "benchmark" is slow Sep 30 17:55:00 including PGO Sep 30 17:55:28 marex: hmm, that does deepen the mystery Sep 30 17:55:31 RP: Python-3.9.7# CXX="/usr/bin/g++" ./configure --prefix=/home/root/build/python --enable-shared --with-system-expat --with-system-ffi --with-ensurepip=yes --enable-optimizations Sep 30 17:55:34 this ^ Sep 30 17:55:38 I just pulled that from BLFS Sep 30 17:56:05 RP: could it be python still pulls in some base pyc files or libraries ? Sep 30 17:56:20 ATH=/home/root/build/python/bin:$PATH LD_LIBRARY_PATH=/home/root/build/python/lib/ python3 bench.py Sep 30 17:56:31 this is how I ran the test after running make ; make install Sep 30 17:57:00 marex: I'm not sure, I'd have to go and look at things and I'm about to head afk Sep 30 17:57:19 RP: that's fine Sep 30 17:57:25 ttyl Sep 30 19:07:23 marex: whats version of gcc and glibc on debian and yocto versions Sep 30 19:08:56 khem: right now, I used OE-core either dunfell or master , and debian testing (current one) Sep 30 19:09:00 khem: so lets see Sep 30 19:10:05 in debian, libc6 2.32 and gcc 10.2.1 Sep 30 19:11:35 in oe, glibc 2.31 or 2.34 and gcc 9.3 or 11.2 Sep 30 19:12:11 I can also test hardknott or gatesgarth if that is closer to what is in debian testing now, but since it happens with both dunfell and master, I doubt that makes any difference Sep 30 19:42:04 right Sep 30 19:42:28 did you build python on yocto running arm box Sep 30 19:43:48 khem: no, I built it on rootfs built using oe-core and bitbake running on arm box ;-) Sep 30 19:44:06 khem: what's this yocto ? Sep 30 19:47:00 oh I am typing on phone Sep 30 19:47:10 whatever is easy Sep 30 19:47:27 I need to add Oe to dictionary here I guess Sep 30 19:52:40 khem: hehehe Sep 30 19:52:48 khem: I'm sorry Sep 30 19:53:32 khem: so anyway, yes, I built the rootfs and added development packages into it, built python ON the machine, and it was still slow, similar to the slowness of python from the OE cross build Sep 30 19:57:57 marex: if you can send an email describing your experiment to ml that would help to reproduce it Sep 30 19:59:05 khem, what do you think then about kodi, stay in meta-kodi or get imported? Sep 30 19:59:20 (I added today missing pkgconfig inherit :) Sep 30 20:20:55 ant__: if its being maintained in meta-kodi then I would say it stays there Sep 30 20:28:46 We were talking about he we need a demo image for Kodi for stands Sep 30 20:28:55 Thought s? Sep 30 20:45:54 meta-kodi is maintained. good Sep 30 20:46:12 though, there are dubious things in there Sep 30 20:47:02 I am puzzled by OECMAKE_FIND_ROOT_PATH_MODE_PROGRAM = "BOTH" Sep 30 20:47:08 that was to find git Sep 30 20:47:56 khem, this should widen the scope and include cmake rootfs and oe sysroot Sep 30 20:48:28 I've read against its use... Sep 30 20:49:45 Crofton[m], afaik kodi does build now for all desired archs Sep 30 20:50:07 provided there are gles drivers Sep 30 20:53:44 khem, finally I don't have a clear answer for users asking to add mysqclient interface to kodi Sep 30 20:54:26 for samba I'd say use the OS samba and skip kodi-smbclient but for SQL is different Sep 30 20:55:19 it's just a packageconfig option but triggers a mysql build Sep 30 21:59:58 can someoen tell me how to give in the version for an RDEPENDS ?? Sep 30 22:01:47 my recipe seems to act up everytime I specify a version Sep 30 22:02:04 * - nothing provides python3-pyserial >= 3.5 needed by python-can-3.3.4-r0.aarch64 Sep 30 23:51:12 override_: are you using rpm backend for package management Sep 30 23:51:57 ant__: yeah if meta-kodi works well then perhaps doing all kodi changes there is good place Sep 30 23:52:35 when I did cmake stuff, it was new, and OE support was also evolving so now things might be different you might now need OECMAKE_FIND_ROOT_PATH_MODE_PROGRAM anymore Sep 30 23:53:44 I think sqlite is supported see https://kodi.wiki/view/Databases Oct 01 00:03:26 https://github.com/xbmc/xbmc/blob/master/CMakeLists.txt#L187 Oct 01 00:03:56 sqlite seems not there Oct 01 00:40:06 Any idea how I'd update my .conf to get rid of "ERROR: Variable USER_CLASSES_remove contains an operation using the old override syntax. Please convert this layer/metadata before attempting to use with a newer bitbake." Oct 01 00:40:20 on honister? Oct 01 01:14:25 mrnuke: change it to USER_CLASSES:remove **** ENDING LOGGING AT Fri Oct 01 02:59:56 2021