**** BEGIN LOGGING AT Wed Jan 07 02:59:58 2015 Jan 07 03:56:10 <_oldtopman> altker128: I'm doing a lot of work on the Edison at the moment. What do you mean by "how the recipes are generally handled"? Jan 07 03:57:50 _oldtopman: Well, OpenCV is broken in the Galileo build ; it's referring to a version that can't be found on SourceForge. I just renamed the .bb file to reflect the new version -- is that "correct" ? Jan 07 03:59:34 And, if I just didn't want OpenCV in the build, is there some "smart tool" that let's me just remove it and let me know what dependencies will break in the process? Jan 07 04:04:34 <_oldtopman> altker128: Good question. I don't know where the selected by default packages-list is located, but if you found it, you could always remove opencv. Jan 07 04:05:05 <_oldtopman> I actually had to update irssi myself a little while ago, there's a little process, but it's not too hard (as long as the script worked in the first place) Jan 07 04:05:45 <_oldtopman> altker128: Can you throw the opencv .bb file on hastebin or somesuch? Jan 07 04:06:44 _oldtopman: The file itself looks OK, the error is about not being able to download the specified version (2.4.3) Jan 07 04:07:05 but the file itself doesn't reference the version, the only place I see the version reflected is in the name (opencv-2.4.3.bb) Jan 07 04:07:15 So that's why I was asking if what I did was "kosher" Jan 07 04:07:24 <_oldtopman> Did it work? Jan 07 04:07:33 No, it failed Jan 07 04:07:36 In another way Jan 07 04:07:45 <_oldtopman> altker128: Just throw it online so I can see. Jan 07 04:07:57 I've basically almost given up on this and plan to use Debian for Galileo. Jan 07 04:09:44 _oldtopman: one sec, let me run it again and give you the output Jan 07 04:10:27 <_oldtopman> altker128: Put the .bb online too - I can't help you debug a recipe that I don't have. Jan 07 04:11:03 k Jan 07 04:11:33 (Aside, depending on places like sourceforge.net or whatever and all these other random servers IMO is a mess) Jan 07 04:11:35 http://pastebin.com/gCNejvAv Jan 07 04:12:04 http://pastebin.com/wG2pp1PL Jan 07 04:12:54 http://pastebin.com/4rSBjUE0 Jan 07 04:13:07 First is the build failure, 2nd is the directory listing of the bb file, third is the bb file Jan 07 04:13:36 I suspect it's a naming issue but seriously renaming files to get it to grab the right download (if indeed that is what one is supposed to do) seems good in theory, fails in practice. Jan 07 04:15:27 <_oldtopman> No, it's set with PV, and I don't know where *that* is set from (was set in the .bb file in the other script I looked at) Jan 07 04:15:43 <_oldtopman> You can always do PV = Jan 07 04:15:52 <_oldtopman> And then adjust the SRC_URI to a working location. Jan 07 04:16:04 <_oldtopman> Then you have to recompute the md5sum and sha256sum for the new file. Jan 07 04:16:24 <_oldtopman> But the patch is clearly required, and I don't know if it will work on a new version. Jan 07 04:16:31 Christ... Jan 07 04:16:33 <_oldtopman> So you might want to track down the original version so that the patch works. Jan 07 04:16:47 _oldtopman: I've looked everywhere, I can't find it (first thing I did) Jan 07 04:17:04 It seems like all of this "stuff" should be managed by a front-end tool of sometime (maybe it exists) Jan 07 04:17:48 <_oldtopman> altker128: I've got a spiral bound notebook with twenty-four pages of stupid crap that I never should have had to write down in the first place, so don't lecture me on the complexities of oe. Jan 07 04:18:17 <_oldtopman> At the very least, it's very straightforwards, and you only have to look here to adjust the build for opencv. Jan 07 04:18:36 <_oldtopman> You could try getting a new recipe for opencv from the oe github - it might work. Jan 07 04:18:39 I have to ask, at this point, why bother, if you have 24 pages of stuff going? Jan 07 04:19:14 Buildroot seems a better approach to me (for embedded) or something like Debian. After all Galileo / Edison are x86, shouldn't this be easy? Jan 07 04:19:28 <_oldtopman> If I gave up everytime the going got rough, I wouldn't be here today. Jan 07 04:20:10 Well, I guess it depends on your goals. If five packages change, re-doing this stuff by hand (i.e. computing hashes, etc) is crazy Jan 07 04:20:23 And error prone Jan 07 04:21:34 <_oldtopman> If you don't like oe or its systems, then you're in the wrong place, and, more importantly, you have the wrong hardware on your desk. Jan 07 04:21:52 <_oldtopman> In the meantime, have a look-see at this here: https://github.com/openembedded/meta-openembedded/blob/e751af9f900834759e59ad2b46a7118a508069e4/meta-oe/recipes-support/opencv/opencv_2.4.bb Jan 07 04:22:22 <_oldtopman> This is the newest opencv recipe. Probably tested. Ostensibly works. No idea if it'll drop in, but you could start comparing the two and see how this new version operates. Jan 07 04:23:02 kergoth (Chris) is actually an old friend , but I don't think OE is right for what I need. Jan 07 04:24:19 <_oldtopman> Very rarely would it be, I imagine. It's very much like compiling a whole distro, rather than compiling bits for a larger system. Jan 07 04:24:44 Are you creating an OE port to Edison for fun? Jan 07 04:25:09 It absolutely does depend on your goals. oe is heavyweight. it has overhead and complexity. it provides a fair bit of benefit in return, but its definitely a matter of tradeoffs. long term distro maintenance is much easier than with buildroot, but i'd absolutely use buildroot for something quick or for board bringup or similar. IMO one of the biggest benefits is sharing development effort amongst multiple distros and multiple target platforms, eases Jan 07 04:25:09 collaboration amongst disparate groups Jan 07 04:25:21 that said, i feel like shooting bitbake in the face on a weekly basis, so.. tradeoffs :) Jan 07 04:25:22 Hey there kergoth Jan 07 04:25:25 hey Jan 07 04:26:27 <_oldtopman> altker128: No, intel uses yocto/oe by default, but I need to make a lot of changes to get a display working. Just need to get a display now :) Jan 07 04:26:37 <_oldtopman> (seeing as it has no video outs) Jan 07 04:28:37 <_oldtopman> kergoth: If you don't mind me asking, who are you? People don't normally name-drop lurkers :P Jan 07 04:30:04 One of the creators of OE... Jan 07 04:30:48 yeah, one of the founders, former maintainer Jan 07 04:31:05 i apologize for some of the crap in bitbake Jan 07 04:31:07 ;) Jan 07 04:32:29 Now I'm curious, are there any front-end tools to "easily" modify/handle creating of recipes, and build configs? Jan 07 04:33:26 there's a relatively new tool in oe-core called recipetool, whose create subcommand will create a starter recipe when given a url to its sources Jan 07 04:33:39 figures out which buildsystem it's using, tries to figure out the license, adds hte src uri checksums, etc Jan 07 04:33:50 not sure if thats what you're looking for or not, but it's quite useful Jan 07 04:34:57 I'm probably going to sound like a nut job, but it'd be the right move for OE to maintain it's own repos, OpenCV 2.4.3 has all but disappeared from SourceForge Jan 07 04:35:27 yocto does maintain a source mirror Jan 07 04:35:45 Earlier I ran into an issue of x264's Git repo changing stuff around, so the MD5 hash compare failed and nothing got built. Jan 07 04:35:57 (not your fault, just saying) Jan 07 04:38:14 <_oldtopman> tbh, I rather like the "pull source from elsewhere online" idea. The project shouldn't be responsible for hosting the tens of gigs of packages, and finding other urls isn't *too* hard. Jan 07 04:38:55 <_oldtopman> Also prevents the project from tampering with the sources, or ignoring future versions "because the included one just works" Jan 07 04:41:11 <_oldtopman> Hey kergoth, does anyone build all the packages in oe? Jan 07 04:41:35 <_oldtopman> (alternatively, how do I propose updates to a recipe) Jan 07 08:37:34 morning all Jan 07 08:38:33 hi bluelightning Jan 07 08:41:09 good morning and HNY Jan 07 08:41:51 hi mckoan Jan 07 08:43:26 woglinde Jan 07 08:43:55 I am trying to install jvm on my device, alls fine but the process fails on one recipy on the do_fetch stage Jan 07 08:44:18 investigations revieled this command: Jan 07 08:44:28 -/usr/bin/env wget -t 2 -T 30 -nv --passive-ftp --no-check-certificate -P /home/griftw/w/setup-scripts/sources/downloads 'http://download.java.net/openjdk/jdk6/promoted/b18/openjdk-6-src-b18-16_feb_2010.tar.gz' Jan 07 08:44:36 but it just hangs Jan 07 08:44:46 when I enter the url into my browser it works Jan 07 08:45:02 all other downloads of the java recipies worked Jan 07 08:45:34 the recipy in question: meta-java/recipes-core/icedtea/icedtea6-native_1.8.11.bb Jan 07 08:45:41 anyway Jan 07 08:45:45 thought id mention Jan 07 08:45:49 im downloading it manually now Jan 07 08:47:04 pompomJuice sorry I can not fix your network problems Jan 07 08:47:12 actually Jan 07 08:47:20 entering that wget command worked Jan 07 08:47:38 but when bitbake runs it it errors out with: Fetch command failed with exit code 4, no output Jan 07 08:47:40 thanks anyway Jan 07 08:48:44 sounds like a firewall or proxy issue Jan 07 08:48:59 yes Jan 07 08:49:15 or http redirection Jan 07 08:49:38 actually it just seems bizarre again, extracting that wget and running it inside the console worked. It must have been network gremlins Jan 07 08:50:08 il add the .done file manually Jan 07 08:53:48 sure Jan 07 09:54:10 meh java wont compile because of this error: https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!msg/openxt/T3cuV9eDIDY/jC93xw-TjTMJ Jan 07 09:54:20 2015 issue Jan 07 09:55:19 iceadtea 10 year bug Jan 07 09:55:56 yes Jan 07 09:56:02 x) Jan 07 09:56:34 https://github.com/woglinde/meta-java/issues/81#issuecomment-68997031 Jan 07 09:56:53 feel free to update to 2.5.3 and send patches Jan 07 09:59:38 thanks Jan 07 09:59:49 thanks? Jan 07 09:59:54 for the info Jan 07 09:59:59 il just bbappend and patch Jan 07 10:00:03 I am to noob to contribute Jan 07 10:00:09 pfffff Jan 07 10:00:17 I have tried before but it consists of strange mailing lists and things I dont understand Jan 07 10:01:12 it is not harder than to get something into the linux-kernel Jan 07 10:24:52 openembedded is becoming such an awesome/easy system to use that even people like me who have no clue can use it effectively. I'm sorry I am not able to contribute. I would want to but the whole submission process is foreign to me. I simply don't know enough about open source to fix such an issue. In the mean time I am grateful for the people who do contribute. They are lifesavers. Jan 07 10:25:49 pompomJuice: are github pull requests that hard? Jan 07 10:25:59 I have done that before Jan 07 10:26:00 and most layers have #exactsteps in their README Jan 07 10:26:10 but I was told that is not the way to contribute Jan 07 10:26:17 I have done those too Jan 07 10:26:24 check the README in the layers Jan 07 10:26:24 I submitted my patch Jan 07 10:26:29 but I got no reply ever Jan 07 10:26:30 different layers have different rules Jan 07 10:26:45 so I thought I must be doing something wrong Jan 07 10:26:55 I used some script to submit the patch, that was supplied. Jan 07 10:27:11 by that time someone else fixed the bug Jan 07 10:27:35 il try again Jan 07 12:28:31 bluelightning, I'm going to ask for help with the stand, unless you are dyin gto do it :) Jan 07 12:42:22 crofton hm should I come to FOSDEM? Jan 07 12:42:56 of course! Jan 07 12:43:16 we have stand Jan 07 12:51:06 hey all Jan 07 12:51:28 crofton hm have to ask my wife Jan 07 12:52:40 has anyone experience with the nvidia tk1 board and bitbake ? Jan 07 12:53:47 pwgen? Jan 07 12:54:01 yupp Jan 07 12:54:09 its me after a long time .. Jan 07 12:54:11 as far as I know there is no layer firt the tegra stuff Jan 07 12:54:43 i tried to bitbake on it, but its reporting an arm7l architecture, that bitbake can not find. Jan 07 12:55:16 i guess my ouya layer should work ... Jan 07 12:56:13 do not think so Jan 07 12:56:28 tk1 is more advanced than tegra3 Jan 07 12:58:43 but first will get bitbake running on it . btw impressive fast. 50 mb/s on a software raid over usb.... Jan 07 12:59:12 it has sata connections Jan 07 13:00:22 yes but only one. so i connected 3 drives to the usb3.0 connection and run raid 5 on it. this gives a cheap homeserver. Jan 07 13:00:50 approx 15 watt .. for all . Jan 07 13:05:48 i am not sure if i can visit FOSDEM . my third kid is on the way Jan 07 13:14:09 Crofton|work: ok thanks; I should probably have sent something by now Jan 07 13:14:33 well, no one is paying attention to anything in Dec :) Jan 07 13:40:16 hello everyone Jan 07 14:51:13 fingers crossed that this php series is OK... Jan 07 14:52:46 fosdem stickers ordered Jan 07 14:52:53 Crofton|work: thanks Jan 07 14:52:54 before I needed rush shipping Jan 07 14:53:31 we did still have a bunch in the box at the end of last FOSDEM but I don't know if they got used up subsequently Jan 07 14:57:53 hmm Jan 07 14:57:57 that is true Jan 07 14:58:16 I am out in .us, well down to like 50 Jan 07 15:09:58 I guess we can never have too many stickers Jan 07 15:10:08 =) Jan 07 15:10:25 so you will be handing out stickers and bitbake advice at the FOSDEM OE stand, bluelightning ? :) Jan 07 15:10:30 and the ones we have been having made are quite nice size-wise Jan 07 15:10:38 mago_: yes, as usual :) Jan 07 15:10:53 mago_: will you be coming to FOSDEM? Jan 07 15:12:22 well, i'd like to go but no plan so far. i've been a few times, like the concept although i'm no a student anymore so sitting at talks for several hours with very low air quality is tough :) Jan 07 15:12:57 embedded is in a large room now Jan 07 15:14:52 i remember when I attended ELC in the US, there were talks about the different types of companies in open source. One category was the "parasites" that only use and never contribute back to the community. I'm afraid the company that employs me does that, so i'm a little ashamed to go to these open source places (even though I try to send patches and bug reports on my spare time) Jan 07 15:14:59 =) Jan 07 15:19:59 I guess that's one way of looking at things, but to be honest I'd rather stay positive Jan 07 15:20:05 we get a fair amount of contributions flowing into OE from folks who weren't under any obligation to do so, so I think things are relatively healthy in our community at least :) Jan 07 15:20:22 mago_, come and help at OE stand, that will help you atone for your sins :) Jan 07 15:20:35 hah :) Jan 07 15:20:47 there are many ways to contribute Jan 07 15:28:24 so what kind of things can i help out with? talking about OE, handing out stickers, poking bluelightning to fix my bug reports? :) Jan 07 15:29:05 mago_: a lot of people who come to the stand have never heard of OE, so a lot of it's just about being able to concisely explain what OE is and does Jan 07 15:29:43 then there are others who are attracted by the blinkenlights of the gadgets we have on the table and just want to know what they all are Jan 07 15:29:58 if we're lucky a small few of those go the step further and ask about OE :) Jan 07 15:31:28 will there be any OE talks? i'm looking at the schedule right now, doesn't seem to be updated recently Jan 07 15:32:36 I'm sure I heard there was at least one, but I didn't have a look at the schedule Jan 07 15:33:23 right, looks like the embedded devroom hasn't fully published the schedule yet Jan 07 15:38:53 yeah, I can see the stuff in pentabarf and it seems likely there will be some OE related talks in the embedded toom Jan 07 16:38:11 bluelightning: when using oe-core/oe-init-build-env to setup the build, it always assume bitbake is located in oe-core/bitbake subdir. Why is this? Wouldn't it be easier to put bitbake outside other repos, so you don't taint them with files not checked in? (even though oe-core has bitbake in its .gitignore) Jan 07 16:39:42 mago_: I think it's just been the easiest setup for people, to be honest Jan 07 16:40:16 mago_: it is possible to specify the bitbake directory; it could be that it should automatically try .. if it can't be found in the current directory Jan 07 16:40:28 how so? i would agree that if bitbake was included as a git submodule in oe-core, then it would be useful.. but it's not Jan 07 16:42:10 mago_: well, put it this way ... it's how we've done things up to this point ;) Jan 07 16:42:21 not to say that's the best way Jan 07 16:42:57 if I do "BITBAKEDIR=~/mytree/bitbake . openembedded-core/oe-init-build-env ~/mytree/build", it works Jan 07 16:43:18 so i guess i could wrap that in a shell script Jan 07 16:44:20 mago_, when you run oe-init-build-env you can specify the location of bitbake Jan 07 16:44:32 . ./oe-init... ../build ../bitbake is what I do Jan 07 16:44:57 oh. that's useful. did not know that! Jan 07 16:45:31 I hated the layout with bitbake in oe-core :) Jan 07 16:45:53 now come to fosdem and help with the stand :) Jan 07 16:46:08 hah, well, i will ask my boss if i can go Jan 07 16:46:27 tell your boss you are going and he is paying :) Jan 07 16:46:37 bitbake isn't in oe-core, it's in poky, technically, though admittedly the latter looks like the former due to its layout :) Jan 07 16:47:09 technically bitbake is separate first and in poky second ;) Jan 07 16:48:38 Crofton|work: maybe if I scream and yell on the floor "BUT! EVERYONE ELSE IS GOING!!!11!" Jan 07 16:48:57 haha Jan 07 16:48:58 :) Jan 07 16:49:06 aww Jan 07 16:50:19 * Crofton|work goes to send an email Jan 07 17:02:22 hm, weird. with the new bitbake (1.24), i get warnings that my "tmpdir is setgid", which is funny because bitbake itself created the directory. If I "chmod g-s tmp-glibc", it works.. but if bitbake doesn't want setgid bit set, it should probably avoid setting it when autocreating TMPDIR? Jan 07 17:05:20 mago_: er... normal users shouldn't be able to set that bit by default, should they? Jan 07 17:05:32 to be precise, a process running as a normal user Jan 07 17:06:00 not sure :) it does on my system Jan 07 17:06:13 if i rm -rf tmp-glibc, re-run bitbake, it creates a dir with setgid Jan 07 17:06:19 (not running as root or anything) Jan 07 17:06:25 is the parent directory setgid perhaps? Jan 07 17:06:46 hm, it is Jan 07 17:06:52 what's your umask? Jan 07 17:06:58 2 Jan 07 17:07:27 ok, I can only assume it's picked up from the parent... I don't quite know what the expected behaviour is there to be honest Jan 07 17:07:45 although, it could be because my dev tree sits on another fs mounted through fstab. I think root maybe owns the mountpoint Jan 07 19:04:55 does someone know if nvidia k1 is little or big endian ? Jan 07 23:21:01 Anyone around who knows how packages get submitted/updated? Jan 07 23:21:07 s/packages/recipes Jan 07 23:24:19 lansiir, patch sent to corresponding mailing list Jan 07 23:24:57 Corresponding? Which one should I send it to for an update in the irssi recipe? Jan 07 23:26:44 lansiir, check the README in the git tepo Jan 07 23:28:05 Ah, nice! Jan 08 01:41:07 could use a review of the diff at https://github.com/kergoth/poky/compare/kergoth:master...recipetool-python by any python folks around. ignore the individual commits, they need rebasing. Jan 08 02:28:09 <_oldtopman> Alright, patch sent to the mailing list. I had no idea git was capable of making patches and mailing them off all by itself; wow. Jan 08 02:29:43 yeah, git's pretty great :) Jan 08 02:29:49 so's mercurial, actually. both are lovely Jan 08 02:33:01 <_oldtopman> How do I know what happens to my patch, anyway? Just look at the git repo frequently, or will I get an email telling me how terrible my hack is? Jan 08 02:34:47 generally it'll get replied to if there's something problematic. but there's no guarantee as to when/if you'll geta reply or it'll get merged. if it's deemed okay, it'll end up in one of the consolidated merges and get run through the automated testing system to confirm it doesn't break anything, and then it might get merged Jan 08 02:42:17 <_oldtopman> Ah, alright. Jan 08 02:43:05 <_oldtopman> kergoth: If your recipe-maker works, I'll probably have a lot more packages to add (if I can grok the patch-addition works) Jan 08 02:43:30 its not mine, bluelightning (paul eggleton) created recipetool, i'm just adding support for python projects (distutils/setuptools) Jan 08 02:43:39 but yeah its pretty cool **** ENDING LOGGING AT Thu Jan 08 02:59:59 2015