**** BEGIN LOGGING AT Wed Jan 29 02:59:58 2014 Jan 29 03:11:00 Hi Jan 29 03:11:16 have one doubt Jan 29 03:12:32 want to build a module for two hardware(two machine variable), will _.bb Jan 29 03:12:35 work Jan 29 03:12:36 ? Jan 29 04:04:40 pt: its impossible ot target multiple machines in a single build. what you name the recipes really doesn't matter as long as you set COMPATIBLE_MACHINE inthe recipoe so bitbake knows which machines its compatible with Jan 29 04:42:40 thanks for the info kergoth Jan 29 04:43:51 when i give _.bb, what exactly 'anything' is interpreted by bitbake? Jan 29 04:54:34 and i do not want for single build, just want to to keep the bitbake file in the build system for multiple mchines Jan 29 05:36:37 _ is the separator between recipe name and version Jan 29 05:43:20 ok, now if for both the machines versions are same. then if i specify _ will it take? Jan 29 05:47:02 i dont understand the question Jan 29 05:47:08 that would set as the version Jan 29 05:47:16 which is unlikely to be what you want Jan 29 07:59:07 good morning Jan 29 07:59:19 morning Jan 29 08:03:20 gm Jan 29 08:08:27 gm Jan 29 08:08:50 anybody having python crashes about "relocatable_binaries_preprocess"? Jan 29 08:09:00 I have an old checkout so I don't see that Jan 29 08:11:43 hi ant Jan 29 08:11:49 morgen Jan 29 08:12:29 jeez, we have kernel hackers blocked by this new python error :/ Jan 29 08:12:40 lol Jan 29 08:12:51 ant_work: it dies in /meta/recipes-devtools/quilt/quilt-native_0.61.bb Jan 29 08:13:42 o.O Jan 29 08:19:57 ant_work: try removing cache/bb_codeparser.dat* Jan 29 08:20:09 that tends to fix weird python errors most of the time Jan 29 08:26:41 * ant_work slaps stupid G*ail deciding to attach the html Jan 29 08:50:20 * ant_work got second email rejected..coffee Jan 29 09:08:33 ant_work: our internal gmail setup seems to silently disapear emails with large picture attachments Jan 29 09:08:56 so it takes a few tries when sending a scanned document Jan 29 09:12:51 finally on 3rd try ... Jan 29 09:13:35 iirc there was long ago a switch to disable rich text and stuff Jan 29 09:13:45 a global one Jan 29 09:49:08 morning all Jan 29 09:53:58 hi bluelightning, all Jan 29 10:09:52 Which hotell were you OE guys staying at during FOSDEM ? Jan 29 10:11:30 Saint Nicolas ? Jan 29 10:15:18 is anyone aware of a BSP for the marvel kirkwood SoC? Jan 29 10:15:23 * rburton thinks, at least Jan 29 10:15:40 my synology NAS has something like that inside and i need more software on it Jan 29 10:27:57 rburton: meta-kirkwood? Jan 29 10:28:50 kroon: there's normally a bunch of folks staying there Jan 29 10:29:15 kroon: I'm staying at NH Stephanie, but that's not as close to the centre of town Jan 29 10:30:01 it's always a trade off Jan 29 10:30:01 bluelightning, ok Jan 29 10:30:07 close to center or close to fosdem Jan 29 10:33:01 kroon: I always staid at Ibis off grand place Jan 29 10:35:09 mckoan: are you coming this year? Jan 29 10:35:40 bluelightning: not in the layer index so it doesn't exist ;) Jan 29 10:35:51 (google'd, thanks) Jan 29 10:35:54 hmm, I thought it was at one point... Jan 29 10:35:57 bluelightning: I have always been at Ibis, no unfortunately not this year Jan 29 10:36:37 bluelightning: since a couple of years now invest all on ELCE Jan 29 10:36:48 mckoan: ah ok, fair enough Jan 29 10:37:03 bluelightning: i want to build packages to sit alongside the existing OS, so i presume i need to extract the toolchain they use Jan 29 10:37:08 ELCE is more embedded-oriented Jan 29 10:41:57 rburton: probably Jan 29 10:42:12 mckoan: it is; I guess FOSDEM just happens to attract a lot of OE folks Jan 29 10:42:33 I looked at the list of talks for this year, to be honest there wasn't much that caught my eye Jan 29 10:42:40 pre-yocto OE was very european centric Jan 29 10:43:08 now we have more asian and portlandian developers Jan 29 10:43:11 koen: well, I still never get to any of the US conferences FWIW... Jan 29 10:43:33 koen: LOL Jan 29 10:43:38 so ELCE and FOSDEM are my opportunities to talk to people in person Jan 29 10:43:53 looks like no ELC US for me this year Jan 29 10:44:10 unless I submit a talk today *and* get budget approval Jan 29 10:44:40 koen: what a pity, ELC US is in a nice place and time this year Jan 29 10:45:00 I know Jan 29 10:45:47 koen: but you're lucky you can tavel for Linaro conf ;-) Jan 29 10:45:54 s/tavel/travel Jan 29 10:46:05 I was tempted to attend ELC on my personal dime Jan 29 10:46:13 but that would still need a presentation Jan 29 10:46:21 mckoan: yes, Macao in 5 weeks from now :) Jan 29 10:46:48 and redwood shores in september, which is a bit meh Jan 29 10:46:55 ELC was there a few years ago Jan 29 10:47:14 thanks to dvhart I know where to get a decent steak there, though Jan 29 10:48:33 hmmm Jan 29 10:48:50 with tmpfs + sstate the most time is now spent in pulling down all layer repos Jan 29 10:59:25 hi all Jan 29 11:02:44 hi pb_ Jan 29 11:10:03 morning bluelightning Jan 29 11:14:22 Hello all Jan 29 11:14:56 I remember that we added a fix that whenever bison is added flex is also added but I forgot where we added that change Jan 29 11:15:05 Can somebody point me on it Jan 29 11:18:45 hmm, I'd like to go to a couple of the Valgrind/GDB sessions, and maybe troll a couple of the Mozilla ones; but the only embedded one is Pauls talk which stands out Jan 29 11:20:40 jackmitchell: actually I was looking for Paul's talk but either I'm blind or it really isn't showing up in the FOSDEM Android app Jan 29 11:21:03 bluelightning: https://fosdem.org/2014/schedule/event/underwater_acoustics_to_opkg/ Jan 29 11:22:27 now that's a cool deployment of OE Jan 29 11:23:16 jackmitchell: aha, found it in the app too - thanks Jan 29 11:34:12 otavio: why does meta-fsl-arm/conf/machine/include/imx-base.inc contain DISTRO settings? Jan 29 12:16:13 koen: it does not ... Jan 29 12:16:22 koen: what do you consider being DISTRO Jan 29 12:20:57 hi pb Jan 29 12:21:28 hi woglinde Jan 29 12:21:29 bluelightning btw. where is our booth located? Jan 29 12:24:20 woglinde: mostly the same place we always are - but we're slightly to the left of the main doors this year (as you enter through them) Jan 29 12:25:23 we have two tables Jan 29 12:25:34 okay Jan 29 12:25:39 (the AW building, for others) Jan 29 12:50:06 hi zecke Jan 29 13:19:34 otavio: it selects hardfloat, which is a DISTRO setting Jan 29 13:19:47 otavio: especially since OE-core default is softfp Jan 29 13:24:30 otavio: and it's more annoying than than Jan 29 13:24:44 otavio: it's selecting cortexa9hf instead of armv7ahf Jan 29 13:24:55 so if I do: Jan 29 13:25:03 MACHINE=something bitbake bash Jan 29 13:25:04 and then Jan 29 13:25:13 MACHINE=wandboard bitbake bash Jan 29 13:25:28 I get bash_armv7ah.ipk and bash_cortexa9hf.ipk Jan 29 13:25:55 now, if I were using OE-core defaults it will break even worse Jan 29 13:49:58 hm intressting the next "Now we close berlios really mail" Jan 29 13:50:06 again? Jan 29 13:50:30 yes april 2014 now Jan 29 13:54:14 I guess we've moved pretty much all of bitbake off there, fortunately Jan 29 13:54:28 bl yes Jan 29 13:57:56 I need to go through the freescale bsps Jan 29 13:58:10 I've noticed some fur when I try setting tune in local.conf Jan 29 14:00:33 bluelightning, I have an odroid-xu + lcd panel that runs direct fb examples Jan 29 14:00:43 Crofton|work: cool Jan 29 14:01:02 somewhat less than I wanted, but x11 is pissing me off Jan 29 14:01:45 hmm, a clue Jan 29 14:01:57 red harry! Jan 29 14:02:03 no fb console either, which is annoying Jan 29 14:02:41 I can rotate the skull thing Jan 29 14:07:04 crofton wayland Jan 29 14:13:27 koen: it chooses the default as hardfp as is the advised by FSL for MX6. The cortexa9hf may be a problem indeed. But since the machine uses the tune shouldn't it be the default? Jan 29 14:13:45 no Jan 29 14:13:58 floating point ABI is a distro choice Jan 29 14:14:54 if you use .deb as OE output you're royally screwed when using fsl machines and others Jan 29 14:15:19 koen: well I think it is an personal opinion in this case as some machines may not support one or another float point Jan 29 14:15:43 bullshit Jan 29 14:15:48 the machines support anything Jan 29 14:15:48 koen: I wouldn't be surprise if MX6 in future drops softfp and FSL tests essentially hardfp Jan 29 14:15:55 koen: no Jan 29 14:16:01 koen: the closed drivers Jan 29 14:16:03 that's like saying "my x86_64 machine cannot run 32 bit code" Jan 29 14:16:18 koen: sorry but I insist you are wrong. Jan 29 14:16:24 otavio: fsl have done some dumb things in their time, but I don't think even they will design a cpu with no integer registers. Jan 29 14:16:40 otavio: have a look at meta-ti, they have a warning about hardfp binary only stuff in their layer Jan 29 14:16:48 koen: MX53 for example does not have AMD GPU binaries in hardfp Jan 29 14:16:52 (ignore how crap meta-ti is in general) Jan 29 14:17:10 otavio: so? Jan 29 14:17:15 it's still a distro choice Jan 29 14:18:11 koen: I think the machine can set the default Jan 29 14:19:08 koen: you are just assuming softfp is a better default Jan 29 14:19:22 koen: if I allow the distro to override it I think it is fine Jan 29 14:19:23 I'm saying that is has been a distro choice since OE started Jan 29 14:19:28 and that the OE default is softfp Jan 29 14:19:58 and a MACHINE.conf poking at DISTRO vars is bad Jan 29 14:20:02 koen: at that time hardfp was not that common, I think ... so it is natural Jan 29 14:20:06 heh Jan 29 14:20:25 when I started with OE hardfpa was standard for familiar Jan 29 14:20:28 10 years ago Jan 29 14:20:38 openzaurus used soft Jan 29 14:21:17 there was a discussion about meta-allwinner doing the same a few months ago Jan 29 14:21:28 they changed it to conform to OE Jan 29 14:21:39 so meta-fsl-arm is the lone BSP wanting to be special and breaking stuff Jan 29 14:21:56 no matter what evil binary package is used as excuse Jan 29 14:22:25 I won't decide anything right now. I think it'd be good to start a discussion about it in our mailing list Jan 29 14:22:32 so we can discuss it there Jan 29 14:22:48 for me, it doesn't matter as I can do it in my distro Jan 29 14:23:37 so let me come at the problem from a different angle Jan 29 14:23:48 angstrom defaults to hardfloat for cortexa8 and newer Jan 29 14:23:55 https://github.com/Angstrom-distribution/meta-angstrom/blob/master/conf/distro/include/arm-defaults.inc Jan 29 14:24:07 but building for meta-fsl-arm A9 machines breaks Jan 29 14:24:17 because the package arch is suddenly different Jan 29 14:24:30 so even for hardfp distros you do the wrong thing Jan 29 14:35:05 koen: so what you would suggest to solve it? Jan 29 14:35:10 brb, lunch... Jan 29 14:37:26 remove it from the .inc, add a note in the README and steal the meta-ti warning from their 3d recipes Jan 29 14:37:38 (and fix the warning to stay quiet when actually using hardfp) Jan 29 14:38:03 changing the OE default is different discussion Jan 29 14:38:20 since changing the default would only mask things like this Jan 29 14:39:38 otavio, I may have noticed what koen is talking about and will check. But I am off to FOSDEM now, and working on the road next week Jan 29 15:00:01 koen: is that to say, if I added meta-beagleboard in it's current form, then I would get armv7a, so it would be arguably better to set my default tune to cortexa8hf-neon? Jan 29 15:00:14 if I was intending of making use of neon and hf Jan 29 15:00:25 and building for beaglebone Jan 29 15:00:50 jackmitchell: the tune is about calling conventions Jan 29 15:01:02 there's no real world performance difference between the two Jan 29 15:01:17 iirc povray is 10-15% faster, that's it Jan 29 15:01:23 ok, that was how I had understood it in the past Jan 29 15:01:47 jackmitchell: copy https://github.com/Angstrom-distribution/meta-angstrom/blob/master/conf/distro/include/arm-defaults.inc and use that in your DISTRO.conf Jan 29 15:01:59 that will set every armv7 board to hf Jan 29 15:02:18 unless it thinks it's a princess like meta-fsl-arm boards Jan 29 15:02:46 koen: yeah, that was what I was going off; I have been building default all this time; and recently I was having trouble with real-time audio encoding, and this might have helped a lot... Jan 29 15:04:43 koen: I think we need a refresh-patch for dracut, maybe the first version was committed Jan 29 15:05:03 ant_work: including all my patches :) Jan 29 15:05:07 you are ahead by 2 patches isn't? Jan 29 15:05:27 * woglinde throws the tegra2 into round Jan 29 15:05:52 jackmitchell: maybe 5-10% improvement with your audio stuff Jan 29 15:06:05 jackmitchell: the main problem is that hardfp is incompatible with everything else Jan 29 15:06:36 koen: ah ok, not so mind blowing then; we got it to work but it was a constant 95%ish usages Jan 29 15:07:20 there was an unfortunate article written by debian users and of course a series of misguided posts on phoronix Jan 29 15:07:38 the article compared software emulation to using the vfp+neon Jan 29 15:07:45 geee, using the hw is faster Jan 29 15:07:53 but they claimed it was due to 'hardfp' Jan 29 15:08:05 and phoronix made a similar mistake Jan 29 15:08:26 so you can't google for this problem Jan 29 15:08:45 ok, and armv7a is neon enabled? I really need to do more reading around this Jan 29 15:09:03 or do I need to use a -neon tune to enable neon? Jan 29 15:09:04 yes Jan 29 15:09:12 except for tegra1 and marvell chips Jan 29 15:09:12 ok; not confusing at all Jan 29 15:09:21 and due to marvell ubuntu disables neon by default Jan 29 15:09:37 right, hence the duff benchmarks etc Jan 29 15:09:45 so if you lib has no runtime neon checks, no neon for you on big boy distros Jan 29 15:10:28 koen: i think you picked the wrong guy to do the oe/yocto correction post on :) Jan 29 15:10:29 ok, I ended up using opus-codec which I'm sure will have neon checks in anyway, seeing as though it is aimed specifically at that use case Jan 29 15:10:31 http://opus-codec.org/ Jan 29 15:10:48 it turned out to be pretty good actually Jan 29 15:10:49 rburton: yeah, I didn't notice it was lpapp :/ Jan 29 15:11:16 jackmitchell: runtime checks are hard to do on arm due to design issues Jan 29 15:11:30 jackmitchell: but we don't care, since we use OE :) Jan 29 15:11:39 sorry Jan 29 15:11:43 all praise OE Jan 29 15:11:46 since we run yocto on our devices Jan 29 15:11:51 I need to get it straight Jan 29 15:14:58 rburton: so OT...where would you add xinput-calibrator ? Jan 29 15:15:19 ant_work: oh yeah, need to read that :) Jan 29 15:15:44 is on my todo list...that hdd is dying ;) Jan 29 15:17:21 there, princess bride reference used Jan 29 15:21:02 koen: do you think we should reserve some mtd space for root as default (ubifs/jffs2)? Jan 29 15:21:05 inconceivable Jan 29 15:21:48 ant_work: I'm missing the context here Jan 29 15:24:16 mkfs.ubifs could use "-R, --reserved=SIZE how much space should be reserved for the super-user\n" Jan 29 15:25:56 ant_work: 5% of rootfs or 1MB whichever is bigger Jan 29 15:26:10 what xora said Jan 29 15:26:22 its just a guy feeling Jan 29 15:26:24 gut Jan 29 15:33:32 jffs2 decides at mount with rp_size= Jan 29 15:40:18 default opt *= 1024; (Kb) Jan 29 15:46:17 koen: +1 for 10^24 :) Jan 29 15:49:15 ant_work: adding xinputcalibrator to the xserver seems reasonable if the machine config says it has one Jan 29 16:07:12 * otavio is back Jan 29 16:10:55 koen: ok, I am here again ... sorry ... was out for lunch ... Jan 29 16:11:01 koen: so back to the topic Jan 29 16:11:31 koen: I am not sure I agree about softfpxhardfp setting. About tune settings I do Jan 29 16:12:47 koen: from your description in the inc. It seems mx53 for example will not have GPU working in angstrom Jan 29 16:12:58 koen: as it works only in soft-fp Jan 29 16:18:15 does angstrom etc have any special tooling to produce its public package feeds, or is it just a rsync of the deploy to something served by apache? Jan 29 16:23:13 rburton: https://github.com/Angstrom-distribution/meta-angstrom/blob/master/contrib/sort.sh Jan 29 16:23:28 rburton: https://github.com/Angstrom-distribution/meta-angstrom/blob/master/recipes-angstrom/meta/upload-packages.bb Jan 29 16:23:31 thanks koen Jan 29 16:23:57 PRSERV fixed most of the problems that tries to work around Jan 29 16:24:15 it won't upload or sort duplicate packages Jan 29 16:24:27 you can guess why I used to be so adamant about PR bumps :) Jan 29 16:26:09 oh yeah, without a PR bump feeds are pretty useless Jan 29 16:28:10 on the autobuilder I apply a patch that uploads to an internal server so I can try things out before it gets synced to the main feeds Jan 29 16:59:38 koen: how do we manage to maintain the pr server database Jan 29 16:59:51 since that will be important with pr server Jan 29 17:07:34 I would like to complain about the tool used to eject nano sims from the Nexus 5 Jan 29 17:09:21 Crofton|work: This is not a distro or machine support channel ;p Jan 29 17:09:48 Crofton|work: nexus5 uses microsim Jan 29 17:09:59 bluelightning: +1 Jan 29 17:10:06 I will compplina less, I put it in a safe place Jan 29 17:10:16 I'm just kidding ;) Jan 29 17:10:35 Crofton|work: but thanks for reminding. I have to pack that tool and charge my Nokia 101 Jan 29 17:13:50 * khem` is compiling/installing KDE framework 5 today Jan 29 17:15:49 any tips on a getting a sim that will work in Brussles and Paris? Jan 29 17:16:09 Also, if you will be at FOSDEM, please volunteer at the Stand Jan 29 17:24:01 Crofton|work: I'm sure I'll be able to lend at hand at times; I just need to figure out a schedule with regards to what talks I would like to see Jan 29 17:24:16 jackmitchell, we all have the same issu e :) Jan 29 17:24:24 and I have to help with the sdr track on Sunday Jan 29 17:24:31 morning Jan 29 17:24:43 I want to make sure that a few people do not end up spending the entire time there Jan 29 17:24:58 at the very least supplying them with food and beverages is welcome Jan 29 17:25:14 * onoffon should once take a pilgrimage to FOSDEM Jan 29 17:26:58 I would imagine I can help Sunday morning and most of Sat, but I'll have a chat with everyone once we're there; I'll be getting in at around 4pm on Friday Jan 29 17:38:28 Does one wear a nametag at FOSDEM ? Jan 29 17:38:53 Cause I havent the slightest idea of what anyone in #oe looks like.. Jan 29 17:39:56 kroon: typically no... but just come up to the OE booth and say hi :) Jan 29 17:42:11 Will do :) Jan 29 17:42:42 I wonder if there is enough interest in having clang as an alternative to gcc for cross compiler in OE Jan 29 17:43:07 khem`: I think someone was talking about that recently Jan 29 17:43:12 now that C runtime and C++ runtime are available clang/llvm suite seems to be complete Jan 29 17:43:14 on the mailing list Jan 29 17:43:31 I have been keeping it on my back burner Jan 29 17:43:39 its quite amount of work Jan 29 17:44:02 it would be ideal that it coexisted with gcc Jan 29 17:44:10 kroon: you can ask images.google.com for help :-} Jan 29 17:44:12 in SDK e.g. Jan 29 17:48:27 pb_, nah :-D think ill just ask around Jan 29 17:49:00 zibri, hey, are you going to Brussel ? Jan 29 17:49:13 I suppose I could find one of my old conference badges and wear that Jan 29 17:49:24 I think that's what hrw does sometimes... Jan 29 17:49:43 bluelightning: he wears your old badges? what a prankster. Jan 29 17:49:55 pb_: heh, but who would want to be me? ;) Jan 29 17:50:22 heh Jan 29 17:51:06 yeah, I guess the bigger risk is that you'd arrive at the oe stand and there would be 20 people there all wearing badges saying "Richard Purdie". Jan 29 17:51:39 heh Jan 29 17:52:32 futile, of course, unless they could also do the requisite northern accent Jan 29 17:56:37 and carry a worried look ;) Jan 29 17:58:13 I think RP broke the fetcher laast time he was at FOSDEM Jan 29 18:03:08 I don't think he's been while I've been there Jan 29 18:04:21 ah, obviously wise to the possibility of impersonators Jan 29 18:55:35 kroon: no, unfortunately no... i did make plans to attend, but then other things got in the way :( Jan 29 18:55:50 kroon: you? Jan 29 20:10:15 zibri, ah ok, too bad Jan 29 20:11:02 zibri, yes, flying down there on friday Jan 29 21:39:35 pb_: heh, I'm not sure why everyone would pretend to be me, that sounds rather scary **** ENDING LOGGING AT Thu Jan 30 02:59:58 2014