**** BEGIN LOGGING AT Thu Apr 17 02:59:58 2014 Apr 17 08:44:31 good morning Apr 17 08:54:45 morning all Apr 17 09:12:58 hey bluelightning Apr 17 09:13:04 hi koen Apr 17 09:13:06 * koen stabs perl and perl modules Apr 17 09:13:21 stab them some more for me ;) Apr 17 09:13:41 I finally have git-perltools working Apr 17 09:13:49 well, send-email at least Apr 17 09:59:07 koen: over in #angstrom someone is asking if the 2013.06 repository is ever going to get restored...? Apr 17 10:00:07 it has always been there Apr 17 10:00:15 http://feeds.angstrom-distribution.org/feeds/v2013.06/ipk/eglibc/ Apr 17 10:01:29 or if that person is talking about setupscripts moving to 1.5, use the branch you want instead of master Apr 17 10:01:32 e.g. angstrom-v2013.06-yocto1.4 Apr 17 10:01:52 I'll pass that on, thanks Apr 17 10:03:34 koen: "all the directories are empty, eg http://feeds.angstrom-distribution.org/feeds/v2013.06/ipk/eglibc/cortexa8hf-vfp-neon/base/" Apr 17 10:03:56 "s/all/some important ones/" Apr 17 10:04:36 cortex* tunes have never been supported in angstrom Apr 17 10:04:45 since that means your MACHINE.conf is broken Apr 17 10:04:58 e.g. like the ones in meta-fsl-arm Apr 17 10:07:16 cortexa8 machines should use http://feeds.angstrom-distribution.org/feeds/v2013.06/ipk/eglibc/armv7ahf-vfp-neon/ Apr 17 10:07:45 the joy of people maintaining their own BSP and getting it wrong Apr 17 10:08:20 koen: do we document somewhere how people are supposed to get it right? Apr 17 10:08:50 not sure if it's documented Apr 17 10:09:10 if not, it shouldn't be too surprising that people don't know how to do it correctly... Apr 17 10:09:16 but there are a lot of email threads where we state that hardfloat/softfloat is a DISTRO setting, not a MACHINE setting Apr 17 10:09:50 yeah, and there's practically zero chance of someone who doesn't already follow the mailing lists finding those Apr 17 10:09:58 well Apr 17 10:09:59 https://github.com/Angstrom-distribution/meta-fsl-arm/commit/2db4807df84ba516aa2617df23bc8cd4eff02646 Apr 17 10:10:04 otavio refuses to fix it Apr 17 10:10:16 saying that it's "just an opinion" Apr 17 10:10:55 I'd love to add it to an "official" doc somewhere to we can slap people with it Apr 17 10:11:29 step 1 would be to make it an official angstrom doc at least Apr 17 10:12:03 angstrom is not the place to document that Apr 17 10:12:06 since it's a BSP issue Apr 17 10:12:35 but I see the value in doing it Apr 17 10:13:07 FWIW, he's saying that these instructions used to work: https://github.com/luke-jr/bfgminer/blob/bfgminer/README.ASIC#L140 Apr 17 10:13:12 so I guess something changed Apr 17 10:14:01 could be Apr 17 10:14:12 but manually poking at feeds should be avoided Apr 17 10:14:32 those feeds did exist till we forked meta-fsl-arm and fixed it Apr 17 10:14:48 dunno why it had the cortex/beagle dir there **** BEGIN LOGGING AT Thu Apr 17 13:14:40 2014 Apr 17 14:37:13 I'm facing a really weird situation that a module doesn't create its package ipk but does -dev.ipk , -dbg.ipk Apr 17 14:37:24 the module has Apr 17 14:38:22 the module is built by makefile but it's so simple that it requires 'do_install()' in .bb Apr 17 14:42:54 dbus-ptest-ptest; so thats why my image suddenly got so big... Apr 17 14:53:07 hi Apr 17 14:53:45 jama ping Apr 17 14:59:43 pong Apr 17 15:00:24 jama any plans to update krb in meta-oe? Apr 17 15:00:58 there are some CVE fixed in newer versions Apr 17 15:04:46 woglinde: not really, but patches are welcome Apr 17 15:04:54 Hmm, never noticed tzdata isn't included in default images. wonder why Apr 17 15:06:10 jama okay Apr 17 15:06:17 maybee this evening Apr 17 15:06:24 kergoth: I don't think we had tzdata in a lot of OE images so it's probably a historic thing Apr 17 15:06:33 OE classic I mean Apr 17 15:06:50 Ah, good point. It does seem like a pretty basic thing, though. I think my home router knows what timezone it's in :) Apr 17 15:06:53 * kergoth nods Apr 17 15:09:13 probably would be useful in a lot of cases yes... could depend on the size impact for others Apr 17 15:12:16 Hmm, yes, this is true. Perhaps a CORE_IMAGE_EXTRA_INSTALL snippet in local.conf to increase visibility and make it easier to enable/disable for folks not building custom images Apr 17 15:12:20 * kergoth shrugs, then yawns Apr 17 15:12:38 kergoth you have the power Apr 17 15:17:04 is there easy way to call python function (something imported in OE_IMPORTS) from shell task? Apr 17 15:17:46 in inline python, yes, but that runs at expansion time, not task execution time Apr 17 15:17:58 but, you could add a prefunc/postfunc to run a python function before/after the shell one Apr 17 15:19:07 ok, thanks Apr 17 15:30:19 moin Apr 17 15:36:09 kergoth: ah that reminds me, that's a nuance that we really ought to cover in the newly expanded bitbake manual Apr 17 15:36:32 (time when inline expressions are expanded within functions) Apr 17 15:36:38 catches a few people out Apr 17 15:37:00 kergoth: have you read over the latest bitbake manual btw? Apr 17 15:37:32 morning nerdboy Apr 17 15:37:35 Nope, didn't realize it was being worked on that heavily Apr 17 15:37:38 will have to take a look Apr 17 15:37:39 http://www.yoctoproject.org/docs/1.6/bitbake-user-manual/bitbake-user-manual.html Apr 17 15:38:13 it might be good since it does still have your name on it ;) Apr 17 15:39:55 FWIW I think Richard posted something to the bitbake list a while back about it being worked on Apr 17 16:57:35 https://github.com/sarnold/meta-alt-desktop-extras/wiki <= new bbb images, including a small hw test image with an updated ltp-ddt package Apr 17 16:57:55 just in case anyone wants to test their beaglebone black... Apr 17 16:59:20 nerdboy: interesting. you should make an entry on http://beagleboard.org/project Apr 17 17:20:51 bitbake just blew up horribly when I tried building wo a MACHINE set Apr 17 17:23:12 Crofton|work: error plz Apr 17 17:24:15 http://pastebin.com/mMWg8KbA Apr 17 17:26:57 hmm I think in setup scripts this is flagged already so you dont get to bitbake when MACHINE is not set Apr 17 17:27:04 what setup do you use ? Apr 17 17:35:33 koen: why's armv7a tunes preferred to cortexa tunes? I'm guessing it doesn't buy you much in the code generation? no cortex-only features that improve performance, so it just bloats up the feeds unnecessarily? Apr 17 17:36:28 kergoth: armv7a is Least common denom Apr 17 17:36:58 we have to stop this tune madness on arm at some point Apr 17 17:38:07 and it will also work with armv8 hopefully, we should be then able to use common packages much like x86 Apr 17 17:39:34 * kergoth nods Apr 17 17:46:29 btw if you are using orc on arm hard float: https://bugzilla.gnome.org/show_bug.cgi?id=727464 Apr 17 17:48:31 ah missing hardfp abi context save Apr 17 18:22:43 kergoth: armv7a tunes will get you the same PACKAGE_ARCH for cortex a5, a7, a8 , a8 and a15 Apr 17 18:23:19 kergoth: and I've yet to see any performance difference due to cortex vs armv7 tunes Apr 17 18:23:44 a5 and a7 might benefit from vfpv4, but so far benchmarks are inconclusive Apr 17 18:27:34 koen: VFPv4 has fused multiply accumulate that gcc might use but not sure Apr 17 18:29:41 and that is non-IEEE arithmetic Apr 17 18:31:13 so if you have -ffp-contract=fast in cmdline options and -mfpu=vfpv4 together then it might make difference Apr 17 18:31:35 khem: You're really selling this :) Apr 17 18:31:35 koen: ah, nice. we might have to think about making that switch at mentor for our customers. we're using meta-fsl-arm mostly unmodified at this point. of course, i'm not sure if we have any customers who care about package feeds, but honestly everyone should, its one of the only real field upgrade options other than reflashing :) Apr 17 18:33:12 kergoth: online package management needs a different mindset most of companies who deploy embedded systems do not have the infra or mindset Apr 17 18:33:19 heh, that's true Apr 17 18:33:35 broonie: selling what ? Apr 17 18:33:50 I expect many would be concerned about hypothetical interrupted upgrade issues, rollback, potential for being in a broken intermediate state. though most good package managers do avoid that Apr 17 18:33:53 anyone poke at ostree yet? Apr 17 18:33:56 i haven't had time Apr 17 18:34:03 khem: The benefits. Apr 17 18:34:13 broonie: there arent any Apr 17 18:34:21 khem: Quite. Apr 17 18:35:01 once I did experiment with tuning to machine and tuning to generic arch and infact many benchmarks did better Apr 17 18:35:19 when tuned to generic arch Apr 17 18:35:39 real workloads are different than benchmarks sometimes benchmarks are misleading Apr 17 18:35:47 true Apr 17 18:35:51 Sounds like the old thing with -Os frequently being quicker than things that optimised for speed. Apr 17 18:36:13 isn't that often the case due to cache pipelines? Apr 17 18:36:18 * kergoth pokes his memory Apr 17 18:36:22 Yes. Apr 17 18:36:22 most common optimization paths are best in gcc when its optimizing Apr 17 18:36:45 if you give it too many dimentions it sometimes pessimises Apr 17 18:36:53 after all its heuristics Apr 17 18:38:55 but sometimes there are few packages which can take advantages of specifics of processor then enable it surgicallu Apr 17 18:38:58 y Apr 17 18:40:56 RP_: can you remind me, a build completely from sstate won't make available intermediate siginfos, so you need to bitbake -S as well if you want enough information to be available to later builds to use -S printdiff, correct? Apr 17 18:55:38 it's kind of sad how often i end up needing to pkill -f bitbake Apr 17 18:55:44 though admittedly less often that it used to be :) **** ENDING LOGGING AT Fri Apr 18 02:59:59 2014