**** BEGIN LOGGING AT Fri Apr 07 03:00:00 2017 Apr 07 10:58:08 Marex, ping Apr 07 11:17:14 hey Apr 07 11:18:38 as this channel is not machine specific, does anyone know where to get help with meta-raspberrypi? I'm having issues with a GCC compiler bug at least on arm target and would like to try out aarm64 - but I can't make the image boot Apr 07 12:34:10 different question: how can i add a deploy task depending on the machine? i have a device tree package that (obviously) installs device tree files, the package itself is only depended on raspberrypi3, but when I build a qemu image the deploy task fails .. for whatever reason. I would expect that addtasks in a package that isnt depended aren't executed. Isnt Apr 07 12:34:11 it like that? Or do i need to see if something else pulls in this package? Apr 07 13:34:00 you can start like: MACHINE=raspberrypi3 bitbake ??rootfs-image?? Apr 07 13:34:57 or edit the local.conf MACHINE ?= Apr 07 13:39:02 MarcWe: not sure what you're saying. the addtask is executed no matter if the package is depended/required or not Apr 07 13:39:33 i need to literally remove the package from my layer Apr 07 14:31:11 domme: on gcc issue, is the issue on arm (32bit) ? Apr 07 14:31:33 on making a task dependent on machine you can do Apr 07 14:31:53 do_deploy_append/prepend_raspberrypi3 Apr 07 14:32:01 khem: indeed it is Apr 07 14:32:08 to add pi3 bits Apr 07 14:32:24 khem: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80236 this one to be precise Apr 07 14:32:30 domme: ok and then you tried aarch64 and it did not boot ? Apr 07 14:32:39 khem: exactly! Apr 07 14:32:55 khem: just got the rainbow screen and didn't know how to debug Apr 07 14:33:55 domme: 64bit boots only into console Apr 07 14:34:07 khem: that is .. serial console? Apr 07 14:34:12 you have to either connect a serial cable or enable sshd Apr 07 14:34:13 so no hdmi output? Apr 07 14:34:22 ooooooh Apr 07 14:34:26 domme: yeah there were some issues Apr 07 14:34:42 I left it at that since my usecase was a router stack Apr 07 14:34:46 UI didnt matter Apr 07 14:35:23 but I am happy to get patches if someone figures the dtb changes to enable HDMI and graphics Apr 07 14:35:39 but in console mode all works fine Apr 07 14:36:09 khem: that's some really useful information, thanks a lot Apr 07 14:36:13 what is the gcc issue you are seeing ? Apr 07 14:36:19 can you explain a bit ? Apr 07 14:36:24 domme> Dominik Schmidt khem: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80236 this one to be precise Apr 07 14:36:27 this one ^ Apr 07 14:36:57 it crashes with list initializers with e.g. std::map Apr 07 14:37:08 i mean, the generated code crashes, not gcc itself Apr 07 14:38:25 hmmm O0 is not often used to compile gcc runtime, so you are really headed into a lonely trail Apr 07 14:38:57 but there might be some interest in gcc community to fix it. Apr 07 14:39:19 O0 is always useless for practical purposes, what is your usecase ? Apr 07 14:39:43 khem: reproducing the issue :) Apr 07 14:40:05 I see Apr 07 14:40:08 khem: we see the same issue with O2 for our code, but it got way harder to reproduce with that.. Apr 07 14:40:26 khem: in other words, i think the issue is simply optimized away because our example is so minimal Apr 07 14:40:46 if you make some progress on rpi3 64bit I will be interested in patches Apr 07 14:41:09 domme: yes its possible, Apr 07 14:41:52 what happens if you dont use double in map<> Apr 07 14:41:56 khem: yeah, dunno.. i'm trying to find my way around openembedded right now and probably fixing graphical output for rpi3 aarch64 is pretty much out of scope Apr 07 14:41:58 and something else Apr 07 14:42:24 khem: it depends, i think int, int does not trigger it.. and double, int does not trigger it neither.. it's completely weird Apr 07 14:42:55 (from the top of my head, no warranties on the concrete types) Apr 07 14:43:49 at O0 gcc will make lot of calls into gcc for type conversions and when you enable optimization it will optimize those away Apr 07 14:44:03 and either use internal code Apr 07 14:44:40 it could be making a call to wrong intrinsic or the intrinsic is having issues Apr 07 14:45:12 what all function calls to functions beginning with __ do you see in final binary Apr 07 14:46:12 khem: should we switch to #gcc maybe? feels offtopic here and there might be other people over there having smart ideas :) Apr 07 14:58:26 khem: btw does it matter if i build the aarch64 image with 4.9 or dev(4.10) kernel? Apr 07 17:05:21 hi all Apr 07 17:06:54 i have four pins on setop box two pin are connected with a resistor , do this mean that pins is for logic analyzer port ? Apr 07 17:59:27 domme: anything 4.9+ should work well Apr 07 18:04:46 khem: cool 👍 Apr 07 22:21:42 bluelightning: Hi! Do you know where the lsbinitscripts has upstream ? Apr 07 22:21:51 bluelightning: I sent a patch for OE-core for now ... https://patchwork.openembedded.org/patch/138961/ Apr 07 22:23:46 khem: I'm finally working on the wks patch for meta-altear Apr 07 22:23:53 khem: is meta-altera still active or is it dead ? Apr 07 22:43:28 is there way to identify IC pins on board ? Apr 07 22:43:43 Marex, ping. Hi, if you have one mom I'd briefly PM you Apr 07 22:46:20 ant_home: I have mom, even dad, why ? Apr 07 22:46:34 reactor16: yes, read the schematic and PCB layout Apr 07 22:46:46 well, let say it's 5 min for you, two nights for me Apr 07 22:46:53 the problem is there no label on pcb Apr 07 22:47:50 i could identify gnd at this time but i'm confused with scl and sda pins Apr 07 22:48:10 ant_home: and why the PM ? Apr 07 22:48:20 * Marex grumbles ... Apr 07 22:48:29 reactor16: well then read the schematic Apr 07 22:48:38 it's OT here, apart the menuconfig issues developing non-yocto kernels Apr 07 22:48:52 marex: directely from pcb ? Apr 07 22:48:55 ant_home: but this is #oe, this is non-yocto ? Apr 07 22:49:00 ant_home: what menuconfig issues ? Apr 07 22:49:29 Marex, that's a strange deps on ncurses Apr 07 22:49:41 ncurses-native is staged but not there Apr 07 22:49:57 but that's not the poin for this evening..;) Apr 07 22:50:32 ant_home: OE problem ? Apr 07 22:50:33 I have to split a kernel module in two, it's too big. I need a little advice from a expert Apr 07 22:50:51 ant_home: head -n 123 foo.c > bar.c ? Apr 07 22:51:14 ant_home: there's ##kernel for that Apr 07 22:51:25 ant_home: but if you want to take it to PM ... mhmmmm Apr 07 22:51:41 it's before sending the patch to MTD list Apr 07 22:52:25 I am just thinking it's too big, I can sho you github, it's public Apr 07 22:52:55 the maintainers are doing other things (one was RP) Apr 07 22:53:23 ant_home: ah, we can discuss that on #mtd then Apr 07 22:53:33 I'm more an OE gut than a kernel hacker :] Apr 07 22:53:35 sure Apr 07 23:01:32 * paulg is curious what makes a kernel module "too big" Apr 07 23:02:24 In the past, there were kooks that essentially made a whole SDK as a kernel module, and did EXPORT_SYMBOL on every little turd in there. Apr 07 23:03:24 it made visible a failing in symbol handling where a linear search made "insmod foo.ko" take eight minutes. Apr 07 23:03:33 anyone have schematic or datasheet for bcm73xx series ? Apr 07 23:04:12 that was 3.x vintage ; while still not an advisable use case, it doesn't fail anymore. Apr 07 23:07:01 paulg: you can encode a massive blob array into it Apr 07 23:07:08 paulg: for example ... Apr 07 23:07:49 reactor16: google harder Apr 07 23:08:05 reactor16: or poke around with a scope ... touch a pad, reboot machine, observe if there is some activity Apr 07 23:08:06 marex: its too secret Apr 07 23:08:12 reactor16: UART waveform is usually quite simple Apr 07 23:08:26 so is i2c Apr 07 23:08:32 you should be able to figure it out Apr 07 23:08:38 marex ; i will be happy if could identify uart pins Apr 07 23:09:15 but its something hard without schematic or datasheet Apr 07 23:09:49 reactor16: then use a scope ? Apr 07 23:10:34 marex what you mean by scope ? Apr 07 23:11:48 oscilloscope Apr 07 23:13:04 marex im newbie and amateur , not pro :D Apr 07 23:15:01 well there's your opportunity to learn something Apr 07 23:15:19 look up uart waveform and start poking around Apr 07 23:17:19 what i observed on this board is there are many group of four pins , which they have same propriety Apr 07 23:19:08 i mesured voltage and gnd they have the same , but not in the same place Apr 07 23:24:33 Marex: yes its active Apr 07 23:25:04 them hi Apr 07 23:27:00 khem: I have nios2 R2 datasheet, so that might be coming at some point Apr 07 23:27:17 khem: do I send you the patches or is there some ML for them ? Apr 07 23:29:01 Marex: github pull requests are preferred Apr 07 23:29:21 khem: I avoid github like plague Apr 07 23:29:50 Marex: you can use yocto ml with [meta-altera] Apr 07 23:30:05 gh pu reqs are much nicer to look Apr 07 23:30:16 khem: uh, yocto ML ? it's OE layer Apr 07 23:30:52 khem: great, so everyone should just centralize themselves on github (which is commercial entity and which has totally weird ToS) ? Apr 07 23:33:03 yocto list is general intro questions and patches for small layers. Very confusing :) Apr 07 23:36:14 * Marex rolls eyes Apr 07 23:41:29 sandy came back from walmart Apr 07 23:41:33 yachts die like flies Apr 07 23:41:39 gah Apr 07 23:44:38 * Marex rolls eyes Apr 07 23:54:18 lol Apr 08 00:35:10 Marex: I meant oe-devel ml Apr 08 00:35:53 khem: but now you're contradicting Crofton|work Apr 08 00:36:59 * Marex facepalms at the NiosII R2 Apr 08 01:16:00 Marex: as I said usually developement happened using pull requests on gh, we hardly needed mailing lists Apr 08 01:16:11 mls are so last century anyway :) Apr 08 01:20:26 khem: well yes, everything that worked is so last century ... so sad Apr 08 01:21:57 Marex: yeah just teasing you btw. Apr 08 01:24:37 khem: I'm almost asleep, so having a bit of a trouble restraining myself from totally trolling Apr 08 01:25:21 although I do have certain strong feelings about some things ... Apr 08 01:37:42 Yes, I know lot of folks have and I do too Apr 08 01:40:35 gnight **** ENDING LOGGING AT Sat Apr 08 03:00:02 2017