**** BEGIN LOGGING AT Wed Nov 20 02:59:58 2013 Nov 20 07:46:02 Yocto is a development platform for embedded system? Nov 20 07:54:39 you can say so, basically Nov 20 08:19:49 would be interesting to "visualize" the poky commit history... Nov 20 08:20:04 * nerdboy messing with graphviz Nov 20 08:26:47 i guess there are some tools to visualize git commits Nov 20 08:31:35 I usually peek at http://cgit.openembedded.org/ Nov 20 08:31:55 yeah, made bash script to feed data to some of the tools Nov 20 08:32:28 tried git-big-picture but it wasn't quite what i wanted Nov 20 08:33:09 here's an early one that's readable: http://www.gentoogeek.org/files/nerdboy-map.png Nov 20 08:33:25 but not quite what i'm shooting for Nov 20 08:34:42 that's one author extracted from a gentoo git repo, so it's a big bunch of unconnected nodes Nov 20 08:35:03 heh, just like gentoo devs... Nov 20 08:37:10 and i didn't choose the colors, either, i swear... Nov 20 08:41:05 maybe I'm biased for stats so I don't see it..what is the purpose of such visualization? Nov 20 08:41:41 just looking for an interesting ways to present it Nov 20 08:42:07 the usual 'useful' visualizations are i.e. ohloh and github Nov 20 08:42:41 if i need to answer a question like "where did this crap come from?" then most of the graphical git front-ends have a decent view of branching, merging, etc Nov 20 08:43:19 i don't use that stuff to do any work, just view things Nov 20 08:43:33 when looking for culprits I usually check the history of the branch or of one particual file Nov 20 08:43:50 *particular Nov 20 08:43:59 this is more of an exercise in "how the heck can you display a large commit history?" Nov 20 08:44:45 not sure how useful it'll be, but we'll see... Nov 20 08:45:11 cgit with more lines per page ;) Nov 20 08:45:39 this isn't about doing work... ;) Nov 20 08:45:52 i get enough of that with git already Nov 20 08:46:20 yes but those kind of graphs lacks the time on one axis..how do you reorder things? Nov 20 08:47:34 that's topo order Nov 20 08:47:53 sort of close to time, oldest in the bottom right i think Nov 20 08:54:40 if you can't tell by the commit messages, those are all my commits from 2003 to 2012 Nov 20 08:55:31 the changelog usually gets the details for my stuff... Nov 20 08:57:18 apparently i save the technical terms (eg, brain-fart) for commits Nov 20 08:58:37 nerdboy: Have you tried "gource" for visualizing the Git history? Nov 20 08:59:39 nope, hadn't heard about until now... Nov 20 09:00:18 i will though, thanks Nov 20 09:01:31 so far just basic data extraction and stealing ideas from the graphviz gallery Nov 20 09:16:22 hey, there are a couple of ebuilds for it to Nov 20 09:16:26 *too even Nov 20 09:16:50 and we're amazingly current for once Nov 20 09:17:04 figures, since it's a dev tool... Nov 20 09:24:33 okay, installed 0.40, will play with it tomorrow Nov 20 09:27:25 nerdboy: you see, the time is in the frames, in another dimension Nov 20 09:28:30 not sure i can display n-dimensional space Nov 20 09:28:40 you cannot visualize the full history in one graph Nov 20 09:28:57 here it is 3d but it could have been 2d Nov 20 09:29:12 impressive anyway Nov 20 09:29:34 looks like the NASA simulations about the universe ;) Nov 20 09:29:51 the gource screehshots? Nov 20 09:30:00 the videos Nov 20 09:30:04 ah Nov 20 09:30:21 i got work in a few hours... Nov 20 09:31:06 ah, I remember wher I've seen it in the past Nov 20 09:31:11 http://marcin.juszkiewicz.com.pl/2010/02/23/gource-openembedded/ Nov 20 09:31:50 night Nov 20 09:32:14 gn Nov 20 10:17:09 morning all Nov 20 10:18:38 bluelightning: morning! Nov 20 10:19:30 hi g1zer0 Nov 20 11:45:18 JaMa: you mentioned that our systemd is broken with our kernel? Nov 20 11:45:29 JaMa: presumably this is the cgroups change Nov 20 11:46:04 JaMa: i'm seeing services that are starting fine be declared as timing out by systemd and aborted (used to work), is this the sort of breakage you've seen? Nov 20 11:46:16 rburton: systemd version in dora release requires kernel 3.0+ Nov 20 11:46:58 rburton: I've one device with 2.6.39 kernel, so I had to downgrade systemd or backport cgroups.procs from 3.0 Nov 20 11:47:04 ah Nov 20 11:47:11 rburton: in the end I've downgraded systemd for dora and upgraded kernel for master Nov 20 11:47:55 more details in http://trac.shr-project.org/trac/ticket/2100 if you're interested Nov 20 11:48:29 right, so not a problem with base oe-core as dora has 3.x Nov 20 11:48:57 correct Nov 20 11:49:23 back to having no idea why this is failing then Nov 20 12:05:08 that reminds me I meant to add this enhancement bug ages ago: https://bugzilla.yoctoproject.org/show_bug.cgi?id=5574 Nov 20 12:05:09 Bug 5574: enhancement, Undecided, ---, saul.wold, NEW , Add kernel version / configuration check mechanism Nov 20 14:29:55 I'm migrating from poky-5.0.1 onto 10.0.0 and am confused by how to setup the bblayers / conf files to have it build my kernel. I successfully ran yocto-bsp and told it where my kernel git repo resides, but it won't build a kernel image with 'source oe-init-build-env meta-ilnx; bitbake core-image-minimal' Nov 20 14:30:24 I get a bootable filesystem image, but no kernel inside the tmp/work or tmp/deploy. Nov 20 14:33:11 with 5.0.1, all I had to do to get a kernel built was to add 'kernel26' to MACHINE_FEATURES. And, have a DEPENDS = "virtual/kernel" inside the task file. Nov 20 14:48:59 Regarding the lighttpd recipe, it will produce lighttpd package and many more modules (like lighttpd-module-fastcgi) I would like to know how to select which module to package in my image ? I have the same question for the libxml2 package. There is a libxml2-dev package produced by libxml2 (available in build/tmp/deploy/rpm directory) but how to put it in my image ? Nov 20 14:51:33 weebet: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#usingpoky-extend-customimage Nov 20 14:52:37 T0mW: I don't think you want to use "meta-ilnx" as the argument for oe-init-build-env, since that's meant to be the name of the build directory you want to use/create Nov 20 14:53:02 uh, ok Nov 20 14:53:03 T0mW: by default you'd give no argument and it would create/use a directory called "build" Nov 20 14:53:14 same as the old 5.0.1 Nov 20 14:53:22 then, you'd need to add an entry in bblayers.conf pointing to the directory containing your BSP layer Nov 20 14:53:41 inside build/conf/bblayers ? Nov 20 14:54:19 Thank you bluelightning, I understand how to create a custom image, but for these case I don't understand how to add those package since bitbake will tell me that it don't know these package Nov 20 14:54:28 Docs on yocto side are a bit confusing. I'll go back to build dir then. Nov 20 14:57:43 T0mW: correct Nov 20 14:58:14 T0mW: if you can make any suggestions for concrete improvements we'd definitely appreciate them Nov 20 14:58:27 LOL Nov 20 14:58:45 "if you don't like the coffee, then you make it" heh Nov 20 14:59:14 T0mW: well, we're not asking you to do the work, just to point out specifically what's missing/needs changing ;) Nov 20 14:59:27 yeah, I know. Nov 20 14:59:55 weebet: are you trying bitbake ? that won't work since these are package names, not recipe names Nov 20 15:00:08 bluelightning, T0mW knows how the game works :) Nov 20 15:00:09 weebet: one way or another you need to get the package name into IMAGE_INSTALL for your image Nov 20 15:00:20 Crofton|work: cool :) Nov 20 15:00:34 I'm trying to follow the recipie guidelines this time around. Before, I just went in a modified the recipes rather than use the .bbappend stuff. So, I have a lot of reading to do yet. Nov 20 15:01:01 sounds good Nov 20 15:01:02 brb Nov 20 15:01:09 bluelightning: I can install libxml2 in my image, but libxml2-dev will not be copied. (same for lighttpd-module-fastcgi) Nov 20 15:01:20 * T0mW goes for coffee Nov 20 15:01:34 you need to tell it to install the -dev packages also Nov 20 15:01:58 if I do this, bitbake will complain it don't know how to do that Nov 20 15:36:17 Hi all, my first attempt at using yocoto for freescale devices as a replacment for LTIB, I believe have everything installed and I have MACHINE set to imx28evk in local.conf, but every time i run bitbake fsl-image-test I get a sanity checker warning "Please set a valid MACHINE in your local.conf or enviroment", I have also set the MACHINE in my environment but the same warning persists. any ideas? I have searched but don't seem to fin Nov 20 15:37:13 did you run the yocto-bsp tool? Nov 20 15:38:24 T0mW: whats that? all i've ran is ". ./setup-environment build" as per one of the guides Nov 20 15:39:10 D4v33: http://www.yoctoproject.org/docs/latest/bsp-guide/bsp-guide.html#using-the-yocto-projects-bsp-tools Nov 20 15:39:37 D4v33: what distro are you using? Nov 20 15:39:57 D4v33: it creates a meta- tree and the machine config file resides in that tree. Nov 20 15:40:10 D4v33: that error means it can't find the machine definition, which probably means you're missing the layer that defines that machine Nov 20 15:42:37 T0mW: rburton: ok, I don't know what distribution I'm using as I've not seen anything relating to one in the guides I've been using which problably mean I don't have one and I need to read the yocoto documentation rather than the freescale quick start guides Nov 20 15:45:07 D4v33: the important thing is do you actually have the meta-fsl-* layers installed and in your conf/bblayers.conf Nov 20 15:46:19 I have the meta-fsl* directories and bblayers.conf file Nov 20 15:46:42 and they are listed inside bblayers? Nov 20 15:47:01 build/conf/bblayers.conf Nov 20 15:47:25 no Nov 20 15:47:31 there we go Nov 20 15:47:34 add them :) Nov 20 15:47:43 hee Nov 20 15:48:02 D4v33: made the same mistake myself this morning Nov 20 15:48:19 under BBLAYERS orBBLAYERS_NON_REMOVABLE Nov 20 15:48:21 ? Nov 20 15:48:43 BBLAYERS Nov 20 15:54:09 right added the entries into BBLAYER but the same error happens Nov 20 15:55:22 D4v33: bitbake-layers show-layers will tell you precisely all the layers bitbake is using. can be handy to debug such problems Nov 20 15:57:29 bitbake-layers comes up with the same message about setting MACHINE Nov 20 15:57:42 yeah, your machine isn't valid, so it can't do anything Nov 20 16:00:25 The machine(s) I've set all come from the README. Nov 20 16:00:47 I suppose it now a question for the imx community forum Nov 20 16:01:55 D4v33: MACHINE needs to be one of the .conf files in conf/machine folder of the BSP layer. Nov 20 16:02:09 so you can check all machines by looking at the layer directly. Nov 20 16:03:28 or export MACHINE="valid-machine" before running bb Nov 20 16:04:02 Crofton|work: D4v33 doesn't know atm what valid-machine is ;-) Nov 20 16:04:05 or should be Nov 20 16:04:19 i've set MACHINE in build/conf/local.conf and that name matches one in sources/meta-fsl-arm/conf/machine Nov 20 16:04:43 * Crofton|work multitasks too much Nov 20 16:04:50 D4v33: if you're setting MACHINE to what the readme in teh bsp layer says, then either the layer isn'jt correctly in BBLAYERS, or the bsp layer readme is incorrect. if the former, best to double check bblayers.conf, if the latter, contact the bsp layer maintainers Nov 20 16:04:54 * kergoth yawns Nov 20 16:05:04 D4v33: for example, I created a layer for my machine using the yocto-bsp: 'ilnx'. After answering a few questions, it created a directory tree 'meta-ilnx'. inside that tree is a machine conf, 'meta-ilnx/conf/machine/ilnx.conf'. I edited that file, as it defaulted to qemux86 instead of 'arm', then edited build/conf/local.conf to change 'MACHINE=ilnx'. Nov 20 16:06:25 hang on something new is happening.... I think it was the export that I was missing, thanks Crofton|work. its now parsing recipes! Nov 20 16:06:28 T0mW: but in his situation, D4v33 already has the BSP layer for the machine he wants, so it's a slightly different situation Nov 20 16:06:42 the export is not needed. Nov 20 16:06:46 D4v33: setting in local.conf will just work Nov 20 16:06:54 only if you don't set MACHINE in local.conf ;-) Nov 20 16:07:17 depends how you set machine Nov 20 16:07:26 sonds like a type in local.conf Nov 20 16:08:20 great thanks guys I think i'm on to the next issue now... ERROR No recipes available for: Nov 20 16:08:20 /home/dave/fsl-community-bsp/sources/meta-fsl-demos/recipes-graphics/xinput-calibrator/pointercal-xinput_0.0.bbappend Nov 20 16:08:26 yeah Nov 20 16:08:42 that's in oe-core Nov 20 16:09:07 so you've either managed to delete meta/ from your bblayers, or your meta/ is too old Nov 20 16:10:01 I need to "restart" the git fetch sequence for my kernel source from my repository. I know 'bibake -c clean linux-yocto-custom' won't do that, what command do I use to blow away the cached git source? Nov 20 16:10:12 D4v33: sounds like you want to moan at the freescale quickstart guide Nov 20 16:10:17 T0mW: cleanall == clean + remove downloaded soruces Nov 20 16:10:21 T0mW: cleanall Nov 20 16:10:58 ah! thanks. Nov 20 16:11:18 but if you just need to restart a fetch, that should just happen if it were interrupted Nov 20 16:11:19 kergoth: 10.0.0 has come a long way, nice system. Nov 20 16:11:56 rburton: I sure will, thanks everyone :) Nov 20 16:12:44 rburton: cleanall worked, I forgot to 'make mrproper' in the source. Nov 20 16:13:12 * T0mW hugs poky Nov 20 16:34:31 Thinking more about the sanity check stuff: Right now, the only options are sanity checks enabled-and-fatal, or disabled. Nov 20 16:34:54 What about a setting like SANITY_CHECKS_WARN_ONLY = "1" Nov 20 16:34:56 ? Nov 20 16:36:03 seebs: that would work, but the bitbake code itself shouldn't know about that variable Nov 20 16:36:54 Maybe it should set BB_ONLY_SHOW_ENV = "1" and leave it up to whatever other code is looking at the data to decide what implications that has. Nov 20 16:39:37 right, that would be a reasonable alternative Nov 20 16:39:57 the conversation above referring to bitbake-layers suggests that that might also be another place where turning off sanity checks would be helpful Nov 20 16:40:48 Yeah. Nov 20 16:40:54 *thinks* Nov 20 16:41:14 Okay, so, the basic issue is, we don't want bitbake to know specifically about oe-core's internals. Nov 20 16:41:26 So we don't want it setting a value that's only meaningful in terms of sanity-checks, or whatever. Nov 20 16:41:55 right Nov 20 16:42:25 I'm struggling to think of a term that would express that in a meaningful but generic manner... Nov 20 16:42:28 But perhaps it would make sense for us to specifically define some value like BB_NO_EARLY_ABORT that a hypothetical non-oe-core thing could also use to handle cases where we really don't want any early events to trigger aborts if they can possibly avoid it, because we want to do something unusual. Nov 20 16:43:10 So we're not specifically saying "disable the oe-core sanity checks", but more generally "if any build system using bitbake is doing something in early parsing that could error out, please try to degrade gracefully so we can do something with the parsed data." Nov 20 16:43:37 ... Although really, it should be BB_EARLY_ABORT, and default to "1", because negation is bad for debugging. Nov 20 16:44:20 kergoth: RP: any thoughts here? Nov 20 16:44:55 seebs: I think we should change bitbake so it explicitly runs the sanity checks with a specific event Nov 20 16:45:02 seebs: then that wouldn't happen with -e Nov 20 16:45:41 That would work too. Nov 20 16:46:35 The cd %s problem was the one that I really couldn't get around in any coherent way. I think the trigger for that is WR-specific, in our case. Nov 20 16:46:48 There's a task being run VERY early, like, "during bitbake -p". Which I'm not sure is a good idea at all. Nov 20 16:47:26 But with the effect that if the directory name for the task has a ${} in it, bitbake -e doesn't get to run. Whee. Nov 20 16:51:38 rburton: can you check if http://patchwork.openembedded.org/patch/57019/ is still needed for new librsvg version? Nov 20 16:55:48 seebs: I merged that bit Nov 20 16:56:41 It took me something like an hour to figure out what was failing there, which I felt silly about once I'd found it. Nov 20 16:56:55 seebs: sometimes the way it happens Nov 20 16:57:03 I was expecting that the single quotes would get me to the next failure, which would be even more complicated, but no, it was just that one line. Nov 20 16:57:34 seebs: I sent out a patch yesterday and it caused a ton of interesting problems. The logging printed the right things in all the places. The actually code to do stuff was in the wrong place in the loop though :) Nov 20 16:57:53 That's awesome. Nov 20 16:58:12 I have actually gotten significantly better about massive unintended side-effects in the last couple of years. Nov 20 16:58:28 The big thing is, if I start thinking I'm tired and not sure I understand what I'm doing, I *stop*. Nov 20 16:58:58 seebs: In this case I knew what I wanted it to do, and the logs said it was doing them. Its just it wasn;t :) Nov 20 16:59:40 hmm, lightning again, lighting up the windows Nov 20 16:59:58 Yeah. Logging mismatching actual events can be a nightmare to debug. Nov 20 17:00:18 It sort of exacerbates the existing tendency to assume that if you already thought of or fixed a given bug, you must not be seeing that particular bug. Nov 20 17:00:47 One of the best security bugs I ever saw involved a case where the devs couldn't figure out what was happening, because it was something they *knew* they had written, and tested, the code for. Nov 20 17:01:29 Cryptographically signed authentication tokens. The crypto checks would confirm whether a token matched or did not match its signature. And it even had code to verify that there *was* a valid signature. Nov 20 17:01:43 The authentication checks then checked for invalid signatures. Nov 20 17:01:51 So if you could manufacture a token with no signature at all, they'd accept it. Nov 20 17:02:11 It turns out that "has a valid signature" and "does not have an invalid signature" are not the same thing. Nov 20 17:03:30 seebs: indeed, sounds like fun :) Nov 20 17:04:03 It took a couple of weeks until one of the customers affected by mysterious account compromises happened to be a security researcher who Got Very Curious. Nov 20 17:05:02 ... BTW, if I am going to mention security, I should mention a thing that may someday affect people: Nov 20 17:05:20 The popular website "tvtropes" keeps all passwords in plaintext, and has plaintext username/password pairs in its login cookies. Nov 20 17:05:40 I dunno how many people here read books, watch movies, or otherwise interact with any sort of media, but I bet a few do. Nov 20 17:09:40 RP: 6af097c8809cdd395bfb3aeadac0bc7c9106b79c removes FREESMARTPHONE_GIT. Wouldn't it have been nice to fix meta-oe first to not use it (or define it itself)? Nov 20 17:10:05 Saur: I had assumed someone was taking care of this :( Nov 20 17:10:31 RP: Maybe they are, but if so, no one has pushed a fix for it yet... Nov 20 17:14:52 So, on that bitbake/sanity-check thing... There's already a bb.event.SanityCheck. Which is fired, and which appears to be processed, only it appears that it fires a SanityCheckPassed() unconditionally (assuming, I guess, that the fatal() will stop it before it gets there otherwise). Nov 20 17:15:16 But nothing appears to fire that event in bitbake right now. Nov 20 17:15:48 So it looks like adding that, and then removing the ConfigParsed branch in sanity.bbclass, would probably do the trick. Nov 20 17:25:08 Hi Nov 20 17:25:18 moin Nov 20 17:25:35 Is there a machine file for a Marvell pxa166 in yocto? Nov 20 17:27:06 if so, probably in one of the "extra" layers? Nov 20 17:27:35 did you check the layer index and/or google? Nov 20 17:29:30 mr_science where can I find the extra layers? Nov 20 17:30:12 http://www.openembedded.org/wiki/LayerIndex Nov 20 17:38:11 probably I should use xscale architecture ... armv5te Nov 20 17:41:18 i'm pretty sure there's more than one marvell arm5 machine def Nov 20 17:41:36 just need the right kernel drivers, etc... Nov 20 17:42:24 mr_science : thank you...I've to go now bye Nov 20 17:43:37 shouldn't be too difficult to make one if needed, as long as you have the kernel source and/or firmware, & bootloader Nov 20 20:28:06 vmeson: So what should we be doing to get busybox compiling again if not using compiler-specific patches? Nov 20 21:09:18 vmeson: ping Nov 20 21:10:41 mtahmed: patch it and test with both compilers. Nov 20 21:11:10 also you might talk / check with the llvmlinux guys. Nov 20 21:11:22 gotta run. back after dinner. Nov 20 21:33:27 seebs: correct, we got half way through a conversion Nov 20 21:33:39 seebs: the graphical UIs do use the event iirc Nov 20 21:34:50 Maybe it would make sense to give the SanityCheck() event an optional parameter that would change the behavior on failure. Nov 20 21:34:59 With the default being to do what it does now. Nov 20 21:35:40 Then bitbake could send a SanityCheck(continue_on_failure), and would know whether it wanted to continue even if the checks failed. Nov 20 21:35:46 Or alternatively, just not send it in some cases. Nov 20 21:36:04 But my experience has often been, if I'm running bitbake -e, I *do* want the sanity check output, I just don't want that to be all I get. Nov 20 21:41:38 jero Nov 20 21:42:25 seebs: That would seem like a reasonable way to enhance it Nov 20 21:42:59 I spent a bit of time staring, and concluded that I am not sure I actually know where the SanityCheck event would *go*. Nov 20 21:44:41 *thinks* Probably in parseConfiguration(), because otherwise we have to spawn it from multiple locations, after calls to parseConfiguration(). Nov 20 21:58:44 Okay the disruptive part of the upgrade is starting shortly. Nov 20 23:26:30 RP, I'm suspicious that the code is not going to be quite of usable quality, but I have patches where bitbake uses SanityCheck events to make sanity checks happen, and "bitbake -p" dies quickly from sanity check failures, and "bitbake -e" runs, also producing a warning message about the failures. Nov 21 00:17:01 halstead, ping Nov 21 00:17:36 dvhart, Hi. I'm just getting the last systems up. I'll be available after. Nov 21 00:17:46 ack Nov 21 00:19:35 dvhart, maintenance window was supposed to end at 4. Nov 21 00:19:52 halstead, no problem, do your thing Nov 21 00:37:08 sgw_, Ready for you to test. Nov 21 00:40:23 halstead: let me know what you want Nov 21 00:40:45 sgw_, If there is a MUT/master ready to try lets start there. Nov 21 00:41:11 Sure I can give MUT a try Nov 21 00:41:40 halstead: is this now running on the new masterab instead of ab01? Nov 21 00:42:17 sgw_, Not yet. I want to test with all the hardware upgrades. Then lets switch masters. Nov 21 00:43:03 Ok, so after this build then? Nov 21 00:47:42 sgw_, I can be ready in 6 hours. Nov 21 00:49:50 halstead: it can wait for the morning, you don't have to stay up for this. Nov 21 00:57:20 sgw_, Okay. **** ENDING LOGGING AT Thu Nov 21 02:59:59 2013