**** BEGIN LOGGING AT Fri May 21 02:59:56 2021 May 21 07:43:01 how can I build entire rootfs with -Os? bitbake.conf sets this to -O2 or -Og, but can't see how to set it differently. May 21 07:48:09 lxc: https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#var-SELECTED_OPTIMIZATION and https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#var-BUILD_OPTIMIZATION May 21 07:49:24 you can set those in your distro config, or local.conf, but note that build scripts may override -Os with -O2 or even -O3. One way to check is to enable buildhistory and make sure all size of all built recipes changes after switch to -Os, and then fix build scripts of all recipes where size didn't reduce. May 21 07:50:31 mcfrisk thanks, so overriding SELECTED_OPTIMIZATION in distro is a valid approach? May 21 07:50:31 To do this more selectively, e.g. via bblass or per recipe: BUILD_OPTIMIZATION_class-target = "-Os -pipe" and FULL_OPTIMIZATION_class-target = "${BUILD_OPTIMIZATION} ${DEBUG_FLAGS}" May 21 07:51:07 lxc: make sure with "bitbake -e image" output, but should be May 21 08:04:43 setting BUILD_OPTIMIZATION in distro to -Os does not do it. TARGET_CFLAGS are still -O2. May 21 08:07:25 Overriding with SELECTED_OPTIMIZATION seems to take effect though. May 21 09:23:59 * RP tears hair out at autobuilder failures May 21 09:49:31 RP: wait wait wait, I hear some Bitbake operator thingies a few days before my talk... what is there to know about SRC_URI ordering? May 21 09:56:19 Hello, I have a strange error with gatesgarth, when I build an image with the previous sstate-cache: do_rootfs: Postinstall scriptlets of ['libglib-2', 'udev-hwdb', 'vim', 'glib-networking'] have failed. If the intention is to defer them to first boot, May 21 09:56:20 then please place them into pkg_postinst_ontarget_${PN} (). May 21 09:56:34 Have you any ideas? May 21 09:57:44 I use the official recipe of 'libglib-2', 'udev-hwdb', 'vim', 'glib-networking' May 21 10:01:12 qschulz: nothing too bad, just that SRC_URI += "X" inherit pypi (which does SRC_URI += "Y") differs from inherit pypi SRC_URI += "X" May 21 10:03:13 qschulz: I was wondering whether to mention the ~= and #= operators but that would just be evil :) May 21 10:06:04 (to be clear, they don't exist) May 21 10:07:01 vygu2: hard to know what the issue is without the logs about why they're failing May 21 10:07:04 I almost fainted :D May 21 10:07:35 RP: yes it differs indeed and I'll tackle it, but why was this an issue for moto-timo then? May 21 10:08:57 "kanavin | moto-timo: SRC_URI is looking odd. I think the actual tarball needs to come first, and no leading space." specifically this May 21 10:09:18 I by default always put the tarball/source code first so I never encountered that... May 21 10:09:29 is this an implicit rule? May 21 10:19:43 maybe I should actually read the whole discussion May 21 10:20:05 qschulz: the automated version upgrade checking makes an assumption it comes first May 21 10:29:46 log filtered on the vim package: https://0bin.net/paste/rCp5Zh3+#mABSMX0nQkelP3TDgNqa4mLLzJMll2JWILna43bWD5V May 21 10:31:46 RP: Mmm haven't unfortunately played with this UPSTREAM_CHECK_* thingy yet so didn't know. Thanks! In that sense, would it make sense to have bbclasses modifying SRC_URI to actually use SRC_URI_prepend? May 21 10:32:07 the same way that bbclasses should use DEPENDS_append and not DEPENDS += May 21 10:32:17 qschulz: Personally I wish things would get away from using append/prepend May 21 10:32:59 we probably should make the classes break things badly if people overrwrite rather than trying to be more and more magic May 21 10:34:16 rburton: just for info, I now have evidence that the glibc mem size thing affects qemuarm64 as well May 21 10:34:18 is it possible to set max file system size for a particular fs type? Something like this IMAGE_ROOTFS_MAXSIZE_jffs2 = ... May 21 10:37:04 RP: Mmmm, not sure you'll be happy with my slides then :p May 21 10:38:00 the issue here would be hard to detect, since it pertains to how UPSTREAM_CHECK_* checks are done May 21 10:38:20 in other cases, SRC_URI += followed by an inherit is fine (I guess) May 21 10:39:49 qschulz: well, we could improve the upstream checks too May 21 10:41:26 qschulz: _append and _prepend are really special case operators and are near impossible to override. They should only really use used rarely but the usage has grown a lot and more use encourages more use May 21 10:41:46 RP: parts of my slides :D May 21 10:42:38 what would be the safest practice then for inherit and += ?= = in classes/recipes? May 21 10:43:15 and I guess the best practice would be to have all inherit after the SUMMARY/DESCRIPTION/LICENSE/LIC_FILES_CHKSUM and all those "description" variables May 21 10:47:40 qschulz: that is usually the format and what is in the styleguide May 21 10:48:01 qschulz: we could do with being more consistent by we have so much legacy May 21 10:52:59 yeah but it's not good to have the ability to break stuff just by adding += before an inherit IMO, hence why I wanted to recommend using _append in classes, so that it's safe May 21 10:54:22 thought https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/cmake.bbclass#n4 was a good example of that :) May 21 11:12:19 qschulz: it solves one problem and creates another - you can't override the class :/ May 21 11:14:09 RP: yeah, you need _remove then with all its consequences. I don't see why would one inherit cmake and not want cmake-native in the DEPENDS but I understand this is a very specific example May 21 11:15:31 qschulz: for this case it is fine. If you make rules like "classes must use append", it becomes a problem quickly May 21 11:16:22 RP: so multilib is not the only nemesis you have :D May 21 11:16:41 qschulz: no, I suspect there are many May 21 12:17:51 Have you not ideo about my error with the above log? May 21 12:18:10 -ideo +idea May 21 12:18:23 -not +any May 21 14:21:43 hi there guys, can someone help in wpewebkit+cog integration on my image? May 21 14:23:41 i have a imx6q board, and with a vendor specific BSP May 21 14:23:45 in zeus branch May 21 14:24:03 i'm following the wpe https://github.com/Igalia/meta-webkit/wiki guid May 21 14:24:05 guide May 21 14:33:07 RP: I've been doing several multiconfig builds over the past few days with master-next, and I haven't seen fatal errors May 21 14:33:50 JPEW: cool, that is promising. I wonder how to reproduce the errors which are being reported... May 21 14:34:30 Ya, I guess thats more of an endorsement that your changes still work for our use case, not necessarily that they fix whatever the problem is May 21 14:35:47 when/where is this multiconfig talk happening? May 21 14:36:15 yates: I'm giving it, is there something specific you would like to see? May 21 14:36:36 yates: Much to tlwoerner chagrin, I *still* haven't submitted my slides :/ May 21 14:36:56 JPEW: no thank you, i just wondered where and what time. is it going to be on zoom? May 21 14:37:20 Ya, it part of the YP summit May 21 14:38:40 yates: https://www.yoctoproject.org/yocto-project-virtual-summit-2021/ May 21 14:39:02 aha. May 21 14:39:07 thanks! May 21 14:39:19 yates: it's next Tue-Wed. on day 1 we have a full day of a beginner's tutorial/introduction May 21 14:39:33 yates: on day 2 we have a full day of hands-on classes May 21 14:39:53 JPEW: and you summoned tlwoerner now... I have to hide too, still haven't submitted the slides /me hides May 21 14:39:59 on both days there are presentations and lightning talks, some more beginner oriented, some very advanced May 21 14:41:05 sounds good. i'll try to get approval from the boss to attend. May 21 14:41:56 the more beginnerish the better lol May 21 14:42:57 * tlwoerner is thrilled that on wed 10/21 of the slides were submitted. that number is now 13/21 May 21 14:43:17 not a bad ratio May 21 14:44:08 qschulz: yes, that includes the hands-on classes. they are a bit special (i know, i've presented hands-on a couple times now) May 21 14:46:46 if i remove the hands-on classes from the count, then the slides coming in were 10/18 on wed, and now 13/18 :-D May 21 14:47:14 and i'm not expecting slides for Josef's "Yocto Yocto" lightning talk ;-) May 21 14:47:24 oops "Yocto Yoga" May 21 14:49:27 There might be some, who knows :D May 21 14:50:09 <_RipFreenode_> https://youtu.be/q0cRlJs7218?t=281 May 21 14:52:29 if it's not a rickroll, i'm disappointed May 21 14:58:39 JPEW: it is helpful to know they don't regress there :) May 21 15:16:10 Is the beaglebone-yocto wic image intended to work as is on the eMMC? May 21 15:41:50 RP: small OVERRIDES question. I've never seen recipe and package names with uppercase letter(s) or underscore(s) May 21 15:42:44 I understand for the underscore, it's at least for the _pn-${PN} override because overrides are split by _ after the variable name May 21 15:43:01 qschulz: we don't allow them May 21 15:43:38 I guess to know where the variable name ends, you need a _ + a lowercase letter, but you could have uppercase later on in the string right? May 21 15:44:00 e.g. nInvaders. I remember we had issues back when I was teaching a course, but don't remember why that was an issue or what it broke May 21 15:44:25 RP: looking just for a reason, I know you're not supposed to use them May 21 15:44:46 qschulz: we don't support uppercase overrides any more - allows bitbake to be a more optimised about decinding what is/isn't an override May 21 15:44:53 but we were definitely using them in thud even (and obviously, it broke things, so I put all lowercase) May 21 15:44:57 qschulz: e.g. is URI in SRC_URI an override? May 21 15:46:01 RP: exactly May 21 15:46:04 BB_NO_NETWORK - so we have to test whether "NO" and "NETWORK" are overrides May 21 15:46:22 by mandating lowercase, it makes bitbake's life a little easier May 21 15:46:42 you just look for the first lowercase letter in the string then, and that's where the first override starts May 21 15:46:52 qschulz: the package backends don't like uppercase in package names either May 21 15:47:06 aaaah the package backends! May 21 15:47:06 qschulz: it might still not be an override but it helps limit the choices May 21 15:47:56 qschulz: the different package backends have differing requirements, I think deb is the most strict. We just said "no uppercase" for simplicity May 21 15:49:17 RP: is this enforced by bitbake now (e.g. checking the MACHINE/PN is all lowercase no undersocre)? May 21 15:51:13 RP: https://www.debian.org/doc/debian-policy/ch-controlfields.html#source indeed, debian wants all lowercase May 21 15:51:29 (and no underscore) May 21 15:52:26 thanks! May 21 15:53:48 qschulz: bitbake won't process some uppercase overrides in some cases, I can't remember if we do sanity check it May 21 16:02:15 the underscore bit me a while ago as we had a machine named sama5d3_xplained May 21 16:03:42 can I safely dd beaglebone-yocto.wic to the eMMC? May 21 16:20:18 abelloni: we probably should have more sanity tests May 21 16:27:42 there is a sanity test for that one now May 21 16:29:13 abelloni: ah, good :) May 21 16:30:31 abelloni: did we figure out anything useful to log in the ltp "hanging" case? May 21 16:37:12 answer to my question is: yes. May 21 16:41:15 abelloni: I think I did log enough to take some actions, patch coming... May 21 16:44:51 abelloni: three builds all hung in the same test... May 21 16:57:49 rburton, jonmason: the build on ubuntu1804-arm-1 is hung in a devtool test operation on a core-image-minimal eSDK :/ May 21 16:58:02 Its in python, sitting on a futex May 21 16:58:52 lovely May 21 16:59:12 a race in devtool, just what we wanted May 21 16:59:43 rburton: its sitting in pthread_cond_wait(), I just can't see which other thread there is running May 21 17:00:16 one day i'll learn how to drive gdb's python integration May 21 17:00:48 rburton: I've tried, its a nightmare to make work May 21 17:00:57 For the record its sitting in devtool sdk-install meta-extsdk-toolchain May 21 17:20:17 rburton: source /usr/share/gdb/auto-load/usr/bin/pythonXX-gdb.py is the magic in gdb but I don't have the symbols for the python it is running from the esdk May 21 17:22:43 rburton: Py_Exit() is in the backtrace so I suspect its some issue at shutdown May 21 17:26:35 * zeddii just spent three hours resolving a 5.10.38 -> preempt-rt conflict in arm/kernel/entry.S, so rburton better be nice to you RP. May 21 17:28:15 rburton: https://autobuilder.yoctoproject.org/typhoon/#/builders/97/builds/2753/steps/12/logs/stdio - I broke the lock in gdb, python aborted but we at least know what it was waiting on - python GIL May 21 17:29:56 zeddii: I remember that file. I'm trying not to now :) May 21 17:33:01 Oh BTW, on thud (haven't checked on newer releases), forcevariable is not the one with the highest precedence in OVERRIDES... it's virtclass-multilib-lib32 :D May 21 17:33:47 RP: damned gil May 21 17:34:57 tlwoerner: slides and odp sent, might be subject to small changes but should be nothing drastic May 21 17:36:46 qschulz: w00t! May 21 17:37:42 rburton, jonmason: I also just looked at this 9 hour ltp build on arm. Its stuck in proc01 test reading /proc/kmsg. The test in question just reads every file in proc May 21 17:38:09 * RP is going to disable that test May 21 17:40:14 rburton, jonmason: any interest in it or should I kill it? May 21 17:42:01 zeddii: do you know much about kvm-clocksource - could that be causing our vm lockups? May 21 17:42:23 zeddii: I have a patch in master-next to try disabling it, was just curious if you've seen/heard anything May 21 17:42:26 just opening every file in proc seems like a rather dumb thing to do May 21 17:42:40 rburton: there is one for /dev and others too May 21 17:42:54 * RP nearly disabled them too May 21 17:49:19 RP: in theory, that is the lightest weight clocksource to use, and has more accuracy for some cases. but trying without it is reasonable. May 21 17:49:46 we used it when we were doing the "real time kvm" stuff @ WindRiver May 21 18:00:28 zeddii: fair enough. I guess we run the experiment and see May 21 18:00:59 * RP kills the ltp build to let halstead get on with maint May 21 18:01:15 Thanks RP. May 21 18:14:50 * paulg notes that running ltp at all can be classified as a rather dumb thing to do.... May 21 18:48:53 Hello... I am using Yocto to build a custom distro based on core-image-minimal. We use hddimg to burn a usb flash drive , which is then used to install distro on harddisk. May 21 18:49:16 I want to add couple of empty partitions on the harddisk at the end of installation. May 21 18:49:25 How do I do that in Yocto May 21 18:50:11 I am open to using WIC. But I was not able to generate a wks which will let me build an image which can be used as live-bootable installer image like hddimg (with two extra empty partitions in the target hdd) May 21 19:04:31 python3-cffi RDEPENDS on python3-pycparser which depends on `cpp` (the pre-processor)... this seems unnecessary (and is an error when doing a non-GPLv3 build). Does anyone know if that dependency be made optional? May 21 20:13:40 Spammy McSpamface May 21 20:14:02 /!\ THIS NETWORK HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:14:03 CONNECT TO IRC.LIBERA.CHAT FOR ALL YOUR CHILD PORNOGRAPHY NEEDS May 21 20:14:16 So much spam. Set /umode +R everyone May 21 20:16:30 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:16:31 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:16:32 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:18:07 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:18:09 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:18:12 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:23:33 all the crazies are out trying to take advantage of the freenode controversy it seems. May 21 20:23:45 fantastic... May 21 20:25:46 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:25:47 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:25:48 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:26:35 halstead: around ? we need your op powers again it seems. May 21 20:35:34 Let me see.. May 21 20:40:14 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT #LIBERIA /!\ May 21 20:40:17 /!\ IRC.LIBERA.CHAT IS THE BEST IRC NETWORK /!\ May 21 20:40:20 /!\ THE JEWS HAVE TAKEN OVER FREENODE, CHATS HAVE MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:41:07 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT #LIBERIA /!\ May 21 20:41:08 /!\ IRC.LIBERA.CHAT IS THE BEST IRC NETWORK /!\ May 21 20:41:09 /!\ THE JEWS HAVE TAKEN OVER FREENODE, CHATS HAVE MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:41:25 /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT #LIBERIA /!\ May 21 20:41:26 /!\ IRC.LIBERA.CHAT IS THE BEST IRC NETWORK /!\ May 21 20:41:27 /!\ THE JEWS HAVE TAKEN OVER FREENODE, CHATS HAVE MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:44:33 /))))))))) May 21 20:44:34 //) __ __\ May 21 20:44:35 C==/_o|^|o_\ /!\ IRC.LIBERA.CHAT IS THE BEST IRC NETWORK /!\ May 21 20:44:36 | _\ ) /!\ THE JEWS HAVE TAKEN OVER FREENODE, CHATS HAVE MOVED TO IRC.LIBERA.CHAT /!\ May 21 20:44:37 \ .--- / /!\ THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT #LIBERIA /!\ May 21 20:44:41 _/`-. __.'_ /!\ JOIN #LIBERIA TODAY. THIS CHANNEL HAS MOVED TO IRC.LIBERA.CHAT #LIBERIA /!\ May 21 20:44:43 /` \`'-,._./|\ May 21 20:44:46 / \ /`\_/\/ \ May 21 20:49:51 Hopefully that helps a bit. May 21 21:07:48 If I want to use two different .wks files (one for the SD card, one for the eMMC), I must define 2 distincts machine configurations, even though the code is compatible? May 21 21:08:00 thanks, yocti May 21 21:08:24 * v2d guesses having to machine configurations sharing the same TMPDIR is the way to go. May 21 21:08:30 s/to/two/ May 21 22:53:46 v2d: not that I have done this, but you should be able to use multiconfig with twi different wks May 21 22:54:22 thank you halstead May 21 22:54:42 Sure. May 21 23:29:50 I suspect the loss of admins means freenode will become unusable fast May 21 23:41:50 #dumpsterfire2021 May 21 23:51:58 We have the channel son libera already May 22 00:23:47 moto-timo, climate change in the internet ? May 22 00:24:43 armpit: and hopefully spammers melt at +1 degree **** ENDING LOGGING AT Sat May 22 02:59:57 2021