**** BEGIN LOGGING AT Wed Jul 10 02:59:59 2013 Jul 10 06:40:06 hi, how can I add a native tool to the meta-toolchain package? Jul 10 06:41:01 I tried adding it to the task-core-standalone-sdk-target.bb, but then I get a "Cannot satisfy dependency" error Jul 10 08:11:57 morning all Jul 10 08:23:10 morning o/ Jul 10 08:31:06 night Jul 10 08:35:11 hi bluelightning, all Jul 10 08:35:22 hi pb_ Jul 10 08:48:56 morning all Jul 10 09:20:44 When I try to build SDK with "bitbake my-image -c populate-sdk" (yocto 13) an error occur: Jul 10 09:21:06 * check_conflicts_for: The following packages conflict with openssh: Jul 10 09:21:06 * check_conflicts_for: dropbear * Jul 10 09:22:04 There is a way to escape error without remove opnessh recipe? Jul 10 09:41:25 silvio_: the question is why is dropbear being pulled in? Jul 10 09:50:47 bluelightning, u right... Jul 10 09:54:18 bluelightning, I dont know maybe some package provide it... it is possible "force remove" from image? Jul 10 09:54:46 bluelightning, or check if someone provide it? Jul 10 09:55:37 silvio_: the only way to handle this kind of thing is to use buildhistory image graphs, "git grep" and/or bitbake -g to find where dependencies enter in Jul 10 09:55:52 bluelightning, thanks Jul 10 09:55:59 silvio_: http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#maintaining-build-output-quality Jul 10 09:56:03 that's the doc for buildhistory at least Jul 10 09:58:05 bluelightning, ok "packagegroup_basic -> dropbear;" Jul 10 10:00:57 silvio_: ah right, now I understand... packagegroup-basic is something added by meta-oe Jul 10 10:01:21 silvio_: to be honest it ought to be replaced by core packagegroups Jul 10 10:01:49 in any case if you look at that packagegroup it uses TASK_BASIC_SSHDAEMON to set the ssh daemon being installed Jul 10 10:02:15 you could either set that to "openssh" or "" if openssh is already being brought in through other means Jul 10 10:05:39 bluelightning,so I have to modify meta-oe, or I can set it in my layer? Jul 10 10:05:54 silvio_: you can set that in your distro config Jul 10 10:06:00 ok thanks Jul 10 10:06:38 bluelightning, I think this would work TASK_BASIC_SSHDAEMON = "openssh-sshd openssh-sftp openssh-sftp-server", thanks Jul 10 10:07:24 hi. it just occured to me... that the kernel headers package was built from the mainline kernel, not from the MACHINE kernel actually used during the build. is there any particular reason for that? so in case we use a 'custom' virtual/kernel, we need to provide our own kernel-libc-headers recipe too. is that correct? Jul 10 10:08:15 ndec: usually that's not needed, no Jul 10 10:08:33 ndec: linux-libc-headers is just there to get libc built Jul 10 10:08:49 but what if i use an old 'vendor' kernel Jul 10 10:09:20 in that case you may have to, but even then AIUI it may not be necessary Jul 10 10:09:40 AIUI eglibc at least has an OLDEST_KERNEL variable and we set that to a pretty old kernel version Jul 10 10:10:15 as a general rule, you want to build eglibc against the newest available headers. it is ok to then use an older kernel at run time. Jul 10 10:10:15 indeed, 2.6.16 Jul 10 10:10:45 like bluelightning says, there is specific control to tell eglibc what the oldest kernel is that you want to support at run time and you might need to fiddle with that per MACHINE under some circumstances. Jul 10 10:12:32 hmm. ok. i am not sure i understand it for now.. but i will digg into that. Jul 10 10:13:03 as it happens, the 'vendor' is telling me that the headers are modified in his kernel.. so that obviously requires us to provide our own header recipe. Jul 10 10:13:15 if you're using uclibc then the situation might be slightly different. if you're using klibc then the situation is probably very different, as are all situations involving klibc. Jul 10 10:13:45 well, it depends rather on what changes the vendor has made. Jul 10 10:14:05 sure. Jul 10 10:14:17 well, things are better now, klibc builds against libc-headers ;) Jul 10 10:14:19 you'd hope that, nowadays, few vendors would be deranged enough to make header changes with significant ABI impact. Jul 10 10:15:13 what's the fun of vendor kernel, if it's not to make funky things... Jul 10 10:15:15 the bitter side of using old kernels is the lack of devtmpfs and the sure troubles with udev (no cgroups,...) Jul 10 10:15:28 it's not that old in our case, it's v3.4 Jul 10 10:15:34 yeah, though you have to be on a way-old kernel to not have devtmpfs Jul 10 10:15:54 we still ship 2.6.31 in some of our production units and we definitely have it there. Jul 10 10:15:59 ah, 2.6.16 are just the headers...I misread Jul 10 10:16:25 yeah.. i guess i should try to find another job, if i had to use 2.6.16 ;-) Jul 10 10:18:40 yeah we're pretty generous there Jul 10 10:18:55 I don't know if we lose anything by having it set to such an old version Jul 10 10:19:25 last time I looked, no, not really. it varies a bit by architecture. Jul 10 10:19:37 ultimately of course this is a DISTRO decision and it's up to the distro people to pick the number they want. Jul 10 10:21:45 I don't remember offhand where the particular figure of 2.6.16 came from, pselect() maybe? Jul 10 10:29:17 bluelightning, thnanks a lot the dropebear is solved ;) "./meta-openembedded/meta-oe/recipes-core/packagegroups/packagegroup-basic.bb:# Set TASK_BASIC_SSHDAEMON = "openssh-sshd openssh-sftp openssh-sftp-server" in your DISTRO config to get openssh(d)" Jul 10 10:30:09 But I have this strange (to me) error: Jul 10 10:30:11 DEBUG: Python function do_populate_sdk finished Jul 10 10:30:11 ERROR: Function failed: populate_sdk_image Jul 10 10:30:25 pb_: that may be the root, it looks like I changed it last to the minimum version supported by glibc 2.9 (according to the commit message) Jul 10 10:32:36 http://paste.ubuntu.com/5861206/ Jul 10 10:37:12 ah right, fair enough Jul 10 11:02:01 Linaro Connect talk on OE starting now Jul 10 11:02:04 https://plus.google.com/u/0/events/c783ftb5ivumq4qn21m6rqipgjk Jul 10 11:04:12 * XorA is there in person Jul 10 11:04:17 Crofton, thanks Jul 10 11:04:50 this should be interesting Jul 10 11:05:00 maybe they will explain how it all fits togethet Jul 10 11:05:04 Crofton, they record it? Jul 10 11:05:09 or we can see the output of the marketing machine :) Jul 10 11:05:15 I think so Jul 10 11:05:24 it is a hangout on the air I believe Jul 10 11:05:25 yes, it will be on YouTube Jul 10 11:05:30 XorA, thanks Jul 10 11:05:38 * XorA gets the link from other channel Jul 10 11:05:57 cant see link Jul 10 11:06:14 is htere an official channel? Jul 10 11:06:20 hrw: youve got the YT link? Jul 10 11:06:49 rofl I am on screen! Jul 10 11:07:06 Crofton, who r u? Jul 10 11:07:17 XorA: https://www.youtube.com/watch?feature=player_embedded&v=pgYuIzoXDEk Jul 10 11:07:22 on the far right Jul 10 11:07:59 silvio_: see hrw link Jul 10 11:08:18 hi, how can I add a native tool to the meta-toolchain package? Jul 10 11:08:29 I tried adding it to the task-core-standalone-sdk-target.bb, but then I get a "Cannot satisfy dependency" error Jul 10 11:08:30 XorA, hrw: I see thanks Jul 10 11:09:39 I was here : http://www.betterembedded.it/ Jul 10 11:09:59 sorry its only Italian Jul 10 11:15:26 very interesting, thanks for link Jul 10 11:21:33 gar vid freeze for me Jul 10 11:47:22 ensc|w: 3.8.13 booting here on pxa2xx (gcc 4.8) Jul 10 11:54:26 epix, "buzzword compliance" xora, buy them both a beer for me Jul 10 11:58:08 we need a better description of distro policy Jul 10 11:58:29 90% of packages cross compile easily Jul 10 11:58:45 the minority have stupid shit that makes it hard :) Jul 10 12:00:05 ant_work: strange... Jul 10 12:00:52 I enable some debug otpions like dynamic printk and ftrace; perhaps the miscompilation is triggered by this Jul 10 12:05:04 XorA, thanks Jul 10 12:05:11 how many people attended? Jul 10 12:05:14 ensc|w: I'mbuilding linux-yocto fwiw Jul 10 12:05:46 ndec: very nice presentation! Jul 10 12:05:51 looks like the recording link is : https://www.youtube.com/watch?feature=player_embedded&v=pgYuIzoXDEk Jul 10 12:06:11 Crofton|work: yes, it is Jul 10 12:06:58 XorA, buy them all some beer and bill me Jul 10 12:30:32 ant_work: thx; that is it... the two memset() fixes make the gcc 4.8 kernel work Jul 10 12:34:20 ensc|w: thanks to Bruce ;) Jul 10 12:34:56 (and Khem) Jul 10 12:36:41 yes; I can imagine much more funny things than finding the needle in assembly code of the memory allocator Jul 10 13:30:23 someone could help me with populate_sdk? I cant understand what is wrong Jul 10 13:30:25 update-alternatives: Error: cannot register alternative opkg to /usr/local/oecore-i686/sysroots/i686-angstromsdk-linux/usr/bin/opkg since it is already registered to /usr/bin/opkg Jul 10 14:04:40 laptop slowly recovering from accidental cooking :-D Jul 10 14:05:19 bluelightning_: thanks for the feedback on the BB man, that's exactly what I was looking for. I'll ingest it and make updates and ping you again once that's complete. Jul 10 14:06:20 wmat: ok, great... again, apologies for taking so long to get that to you Jul 10 14:06:49 XorA: mm deep fried laptop **** BEGIN LOGGING AT Wed Jul 10 14:08:55 2013 Jul 10 14:12:57 Someone could give me some suggest, when I build SDK i have this error, but i cannot understand and found hw to solve, thanks Jul 10 14:13:10 update-alternatives: Error: cannot register alternative opkg to /usr/local/oecore-i686/sysroots/i686-angstromsdk-linux/usr/bin/opkg since it is already registered to /usr/bin/opkg Jul 10 14:29:00 what is the most uptodate browser layer for oe? Jul 10 14:32:56 meta-browser Jul 10 14:38:28 beer buying at ELCE is going to be rough Jul 10 14:38:42 JaMa: thanks Jul 10 14:38:46 I owe ndec some beer for his Linaro presentatoin Jul 10 14:39:25 Crofton: hehe, we shall all make you owe us beer Jul 10 14:39:39 Crofton: better start saving now :) Jul 10 14:39:45 heh Jul 10 14:39:57 we are running OE workshop now Jul 10 14:40:15 Hopefully Linaro can help educate people about making software easier to cross compile Jul 10 14:40:19 cool Jul 10 14:40:23 how is that going? Jul 10 14:41:13 if anyone there has also been to a yocto dev day, we would be interested in comparisons Jul 10 14:41:51 Crofton: quite well despite choppy network Jul 10 14:42:20 I thought I heard laughter when people mentioned the network Jul 10 14:42:44 Crofton|work: laughter? they do not have network iirc Jul 10 14:42:57 I was watching the live stream :) Jul 10 14:43:07 Crofton|work: it was pre-recorded month ago Jul 10 14:44:00 so they are spending all week at the Guinness Brewery? Jul 10 14:44:16 ewww Jul 10 14:44:29 is there any other thing to do in Dublin other then pubbing? Jul 10 14:44:32 There is real stout available in local restaurant :-D Jul 10 14:44:47 not chemical sludge of guiness Jul 10 20:03:36 for a basic sysvinit image Jul 10 20:03:42 where is rc.local equiv? Jul 10 20:25:16 Crofton|work: not sure there is one Jul 10 20:25:21 yeah Jul 10 20:25:28 dealing with it Jul 10 20:25:38 at some point in our classic build i added a local.init Jul 10 20:25:39 well dealing with people that expect one Jul 10 20:26:05 don't remember ever seeing one in poky/master Jul 10 20:26:50 wouldn't mind having the facilities to pick /etc/rc.local automatically Jul 10 20:26:59 *pick up Jul 10 20:30:33 Crofton|work: is your answer response to the "people that expect one" to add that feature? Jul 10 20:30:52 sed one of those ... Jul 10 20:31:04 well, I suggested they make a proper init script and the problem seems to be sovled Jul 10 20:31:14 of course, this is for a dirty hack Jul 10 20:31:39 the answer is systemd Jul 10 20:31:56 I suppose you cuold make an init script that looks for rc.local Jul 10 20:32:17 don't really use one in gentoo, but i thought that was typically for on-off/oddball stuff that didn't really need a full init script Jul 10 20:32:26 one-off commands, etc Jul 10 20:32:55 so what's the answer for non-systemd builds? Jul 10 20:33:18 i specifically avoided systemd in my rpi builds... Jul 10 20:34:17 since it's an optional distro_feature (i believe) shouldn't there be a more global answer? Jul 10 20:34:37 * mr_science shuts up for a while Jul 10 20:43:07 not sure Jul 10 20:51:06 kergoth`, pb_, RP ping http://comments.gmane.org/gmane.comp.handhelds.openembedded/41370 Jul 10 20:51:22 (autotools.bbclass stuff) Jul 10 20:52:28 i'm thinking about cleaning out my old tree and want to know what to do with e.g. these 3 Jul 10 21:44:37 anyone else up to autotools.bbclass? **** ENDING LOGGING AT Thu Jul 11 02:59:58 2013