**** BEGIN LOGGING AT Thu Aug 27 02:59:58 2015 Aug 27 03:04:16 hmm, does use of VPATH cause problems for the -dbg code which grabs source files? Aug 27 03:04:24 seems like it should be looking in both B and S for relative paths Aug 27 03:04:34 kergoth: you got a link to the repo generator? Aug 27 03:06:02 it's in codebench svn, p2_repogenerator. could also look at the yaml based bits fahad did (iirc) which uses the stock p2 stuff Aug 27 03:06:44 darknighte: http://svr-sje-svn-01.sje.mentorg.com:8080/svn/esd/tools/csl/p2_repogenerator/trunk Aug 27 03:08:29 gnu-devel -n -U -h i686-pc-linux-gnu sgxx/trunk/arm-none-linux-gnueabi-sourceryg++ init/build_p2_repogenerator from my shell history on build5 in the server farm seems promising, but it requires a gnu-checkout Aug 27 03:08:47 actually, it looks like i have a tarball with built binaries Aug 27 03:09:16 dhirc: i would probably enable these too Aug 27 03:09:20 # CONFIG_SND_HRTIMER is not set Aug 27 03:09:20 # CONFIG_SND_DYNAMIC_MINORS is not set Aug 27 03:09:32 lemme look at usb... Aug 27 03:11:48 i would make EHCI not modules, and either similar for OHCI or disable it Aug 27 03:12:34 usb tends to be crucial on these things so just compile it in Aug 27 03:12:53 er, wrong channel Aug 27 03:13:05 ? Aug 27 03:13:08 kergoth: crud. Aug 27 03:13:32 CONFIG_USB_EHCI_HCD=y? Aug 27 03:13:38 other stuff can be modules, just not the primary controller stuff Aug 27 03:13:47 dhirc: yes Aug 27 03:13:57 plus omap support Aug 27 03:14:54 some of these things do almost everything over usb bus... Aug 27 03:15:06 look at rpi model b Aug 27 03:16:11 * nerdboy would've paid another 89 cents for a real network chip... Aug 27 03:19:20 * nerdboy getting tennis-spectator-neck bouncing between channels Aug 27 04:09:51 nerdboy: It seems my fragment is not being applied when I look at /proc/config.gz.. I have create a meta layer (meta-bbb-via) and have added a .bbappend (http://pastebin.com/Za2YgJEr), and the cfg (http://pastebin.com/DrkeaSTG). My local and bblayers are http://pastebin.com/Rr27Gvxx and http://pastebin.com/gdQsUhj0. What am I doing wrong here? Aug 27 05:31:28 i don't see anything jumping out... Aug 27 05:31:35 hang on a sec Aug 27 05:35:46 https://bpaste.net/raw/df8dab1c86ca Aug 27 05:36:26 that's it Aug 27 05:36:44 try adding machine to that append Aug 27 05:36:58 dhirc: ^^ Aug 27 05:42:04 I have a recipe which requires a variable be set in another recipe… how would I do that? FWIW it is one of the variables in https://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/recipes-bsp/bootfiles/rpi-config_git.bb that I need to set. Aug 27 05:42:41 Or should I be appending to that file (config.txt) directly from my recipe instead of trying to go through a variable? Aug 27 05:46:37 nemequ, using a local file to your recipe is probably the best practice Aug 27 05:47:51 redengin, I have to add something to a file generated by that recipe (config.txt), AFAIK there is no other way to do what I need. Aug 27 05:48:03 i take it it's not one of the options there now? Aug 27 05:48:42 nerdboy, nope, it's i2c… i have a patch to add it, which I'm planning on submitting as soon as I get everything working Aug 27 05:49:23 add i2c as option to what? Aug 27 05:49:27 can you just chain in for now and use sed like the current recipe, but with your i2c variable? Aug 27 05:49:37 works when you load kernel modules Aug 27 05:49:42 nerdboy, https://paste.fedoraproject.org/259926/06545531/ Aug 27 05:50:02 sorry, did I miss something when i accidentally closed the channel? Aug 27 05:50:13 oh, the dt option Aug 27 05:50:33 nm, channel logs Aug 27 05:50:47 that could go up to to az guy Aug 27 05:51:25 * nerdboy steals it for his fork Aug 27 05:53:42 nemequ, how about just using bbappend and adding a predeploy task before deploy Aug 27 05:54:11 and append to the file? is that considered to be a good practice? Aug 27 05:54:56 nemequ, err, make that a postdeploy :) since deploy is copying the file in Aug 27 05:54:58 i'm planning on trying to get this integrated in meta-raspberrypi, so i want to do it right and i have no issue with modifying the existing recipe, if that matters... Aug 27 05:57:21 nemequ, how about "addtask do_postdeploy after do_deploy before do_package Aug 27 05:57:45 just patch yours with that Aug 27 05:58:10 enable the dt stuff, etc... Aug 27 05:58:17 if you can live with a local recipe branch, thats probably best, doesn't rpi need this support anyway? Aug 27 05:58:37 they just tell you to enable it manually Aug 27 05:58:47 nerdboy, you mean the patch i posted? that's what I'm doing, and setting ENABLE_I2C in my local.conf… Aug 27 05:59:11 great, let me know how that goes... Aug 27 05:59:14 yeah, it's probably something that should be there, but i'd like to avoid making people enable it manually Aug 27 05:59:25 nerdboy, it goes fine, i've been doing it for weeks. Aug 27 05:59:34 nerdboy, manually when? upon runtime? Aug 27 05:59:34 no, that particular patch Aug 27 06:00:03 manually edit config.txt or use the config tool i guess Aug 27 06:00:33 ah, cuz there could always be boot issues with a bunch of unserviceable interrupts enabled Aug 27 06:00:35 dt stuff is "relatively" new... Aug 27 06:01:21 last i checked, the readme did not say anything about stable... Aug 27 06:01:22 whats dt stand for? Aug 27 06:01:28 device tree Aug 27 06:01:29 redengin, device tree Aug 27 06:02:05 * nerdboy 's fork is a bit older than weeks Aug 27 06:02:22 forked it from the original guy Aug 27 06:02:50 https://github.com/VCTLabs/meta-raspberrypi Aug 27 06:03:30 what does config.txt do differently than the kernel config? Aug 27 06:03:48 i.e. do you need i2c to boot? Aug 27 06:04:07 no, it's runtime config Aug 27 06:04:36 k, the "*-bootfiles" made me think this is pre-kernel Aug 27 06:05:19 so, then is this just another path to setting the kernel config? Aug 27 06:05:39 nope Aug 27 06:05:53 nothing to do with kernel config Aug 27 06:06:17 but you do need the right stuff enabled for the runtime options to mean anything Aug 27 06:06:39 it's convenience Aug 27 06:07:24 so what do these settings do that the kernel doesn't? Aug 27 06:07:52 maybe read the comments? Aug 27 06:08:19 the full config.txt is mostly comments Aug 27 06:10:08 redengin, https://www.raspberrypi.org/documentation/configuration/device-tree.md#part3 may help Aug 27 06:10:15 so, my googling says this config.txt is the stand-in bios Aug 27 06:11:05 that actually sounds like a pretty good description to me… Aug 27 06:11:48 http://elinux.org/RPiconfig Aug 27 06:17:29 so basically, I have a recipe to add support for the PiFace RTC (http://www.piface.org.uk/products/piface_clock/), which requires i2c. in order to do that i need to write a line to config.txt, which is provided by https://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/recipes-bsp/bootfiles/rpi-config_git.bb… what is the *right* way to do that? Aug 27 06:18:22 that patch i posted earlier works okay, but requires (afaik) intervention in local.conf, which I would like to avoid… i want to put something in my piface-rtc recipe, if possible, to just automatically enable it Aug 27 06:20:29 the postdeploy thing seems feasible, but it also seems like a bit of a hack so i'm not sure it would be accepted Aug 27 06:21:08 does the rpi bsp kernel config already enable i2c? Aug 27 06:22:06 nope Aug 27 06:22:49 without adding something to config.txt the kernel can't communicate with the clock :( Aug 27 06:23:04 then I'd just append the kernel config to enable i2c and not let the user via config.txt disable it Aug 27 06:24:51 i don't think that will work. based on https://www.raspberrypi.org/documentation/configuration/device-tree.md it sounds like the firmware needs to know to enable config.txt, not just the kernel Aug 27 06:24:56 ^ @ redengin Aug 27 06:25:38 see the second paragraph on that page Aug 27 06:26:48 err, the firmware needs to know to enable *i2c*, not just the kernel Aug 27 06:27:46 sounds like rpi isn't sharing its device code, so it must do it itself Aug 27 06:28:53 without knowing more about the rpi yocto architecture I really don't have any good suggestions beyond the postdeploy Aug 27 06:29:33 you should just submit your patch as an addition to do_deploy and I'm sure they'd be happy with it Aug 27 06:29:58 ok, i'll do that Aug 27 06:30:10 thanks for the help redengin, nerdboy Aug 27 06:31:09 could for and do pull request Aug 27 06:31:28 nemequ, you may want to change it to a sed one liner to follow suit with their recipe Aug 27 06:31:45 or open issue here: https://github.com/agherzan/meta-raspberrypi Aug 27 06:32:10 *fork even Aug 27 06:32:15 or spork maybe Aug 27 06:32:36 nerdboy, using seed would require a placeholder in config.txt Aug 27 06:33:07 "#enable_i2c" or something so i could s/#enable_i2c/dtparam=i2c=on/ Aug 27 06:33:15 nemequ, I assumed that a stand in for bios would already have explicit defaults, but perhaps I assume too much Aug 27 06:33:54 well that config file is part of the meta-raspberrypi repo, so i could add a placeholder… Aug 27 06:34:35 I'd definitely suggest adding explicit defaults in a config.txt.patch with your pull request Aug 27 06:34:57 yeah, that seems reasonable Aug 27 06:39:20 so, whats the rpi community doing these days? I'm always looking for a reason to write some code Aug 27 08:25:46 morning all Aug 27 08:56:22 moin Aug 27 08:57:26 question: I have a package which in its bitbake recipe needs a hook for preupgrade. Aug 27 08:59:35 is there any way to create this? Aug 27 09:15:30 JYDawg: take a look at prerm, postrm, preinst hooks, https://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#new-recipe-post-installation-scripts Aug 27 09:26:59 bboozzoo: looking at the code ... no reference to upgrade. Aug 27 09:27:02 bluelightning: ping Aug 27 09:27:58 bluelightning: I would like to spend some effort on https://bugzilla.yoctoproject.org/show_bug.cgi?id=6787. If you can guide me on the same, like how to move ahead with the task, it would be great help for me. Aug 27 09:27:59 Bug 6787: enhancement, Medium, 2.1, unassigned, NEW , Cancelling builds at any point in the build process Aug 27 09:31:46 sujith_h: I think there are several challenges with this one Aug 27 09:32:24 sujith_h: the first is that only one controlling UI can connect to the bitbake server at once, so it's not possible to connect to the running server separately and tell it to stop Aug 27 09:34:09 sujith_h: the other I guess is that bitbake might not be running at that point at all (it might still be fetching metadata) Aug 27 09:35:19 sujith_h: the first problem might be easy to work around by just sending SIGINT to the bitbake UI process Aug 27 09:35:33 if you want to force stop you'd just send it twice Aug 27 09:35:53 it's not the ideal solution though and I'm not sure how practical it will be Aug 27 09:36:45 bluelightning: I was thinking about the same. Like an button to stop bitbake process which would send SIGINT Aug 27 09:36:58 bluelightning: but as you said right, its a work around Aug 27 11:03:26 Hi folks, I'm porting python-gst1.0 to yocto, would you apply a patch on poky? Aug 27 11:09:52 good question Aug 27 11:10:38 assuming you effectively mean "would we put python-gst in OE-Core", it depends on whether that needs to be aligned closely with gstreamer itself Aug 27 11:11:57 ./poky/meta/recipes-devtools/python/python-gst Aug 27 11:11:59 I see it there Aug 27 11:12:14 so I would like to give you a ./poky/meta/recipes-devtools/python/python-gst1.0_1.5.2.bb Aug 27 11:12:28 or whatever, isn't that already part of core? Aug 27 11:15:28 LocutusOfBorg1: there is a python-gst in meta-multimedia (part of the meta-openembedded repo), it's 0.10.x though Aug 27 11:15:43 LocutusOfBorg1: I'd probably go for adding it to meta-multimedia Aug 27 11:17:26 so would you accept a recipe? Aug 27 11:19:28 well, it's not really up to me :) Aug 27 11:20:05 but I would say since we have python-gst in meta-multimedia, the right place to add the 1.0 version is meta-multimedia rather than the core Aug 27 11:22:38 (in case it's not clear, poky is really a combination of several different components; the recipes under meta/ come from OE-Core and poky just pulls those in verbatim - so any additions/changes to those are sent to OE-Core first and then they automatically come into the poky repository; however that wouldn't apply here if the recipe was instead added to meta-multimedia, that's completely separate) Aug 27 11:22:55 in case you haven't seen it: http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded Aug 27 11:23:00 so where does this come from? Aug 27 11:23:02 http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-devtools/python/python-gst?h=fido Aug 27 11:23:48 seems like that's in fido but not master Aug 27 11:23:55 I was looking at master Aug 27 11:24:07 oh... sorry, I always work with fido sure Aug 27 11:24:10 :) Aug 27 11:24:25 right, just looked at the history - it was moved out because it's the obsolete 0.10.x version Aug 27 11:24:29 strange to find something in fido and not in master... Aug 27 11:24:30 I'm honestly not sure now Aug 27 11:24:51 rburton: where do you think python-gst 1.0 ought to go? OE-Core or meta-multimedia? Aug 27 11:25:13 anyway I'm writing the recipe, if you have a place, I'll be so happy to contribute Aug 27 11:26:34 LocutusOfBorg1: ok, that's great... I'm sure there will be a home for it :) Aug 27 11:43:45 I have a weird problem and don't know how to start debug it ... Aug 27 11:44:35 I have a recipe "foo" - this contains several packages: foo-bar, foo-runtime, foo-extra, foo-doc, ... Aug 27 11:45:13 for some kind of reason, bitbake doesn't find foo-bar and so depending packages fail Aug 27 11:45:23 bluelightning: doesn't new python-gst need meta-gir? Aug 27 12:21:07 rburton: possibly, not sure Aug 27 12:22:38 rburton, not sure, for now I need to write a recipe for python-pyygobject3 Aug 27 12:37:34 yeah that's where gir comes in i'm sure Aug 27 14:23:49 hi guys Aug 27 14:24:10 I am new to yocto and I am trying to build my first image Aug 27 14:24:35 as a target I've picked up raspberrypi Aug 27 14:24:52 I have problem building it Aug 27 14:25:05 my host OS is Ubuntu12 Aug 27 14:25:29 and I am facing problems building dpkg (dpkg_1.17.21) Aug 27 14:25:47 ERROR: Task 192 (virtual:native:/home/a10053/yocto/poky/meta/recipes-devtools/dpkg/dpkg_1.17.21.bb, do_compile) failed with exit code '1' Aug 27 14:26:33 but the thing is that after running 'bitbake -c compile dpkg' everythings is just fine Aug 27 14:27:16 any hints why it is not working while building whole image running 'bitbake rpi-basic-image' while is fine with previous command? Aug 27 14:30:30 marcin_: can you use pastebin to show us more of the error message? Aug 27 14:30:58 error is here: /usr/local/include/bson.h:63:2: error: #error Must compile with c99 or define MONGO_HAVE_STDINT, MONGO_HAVE_UNISTD, MONGO_USE__INT64, or MONGO_USE_LONG_INT. Aug 27 14:31:33 my question is why this is working while compiling just this one single package Aug 27 14:31:46 and not when building whole image Aug 27 14:32:05 also I am wondering how to force c99 for building dpkg Aug 27 14:32:43 should I configure this in 'poky/meta/recipes-devtools/dpkg/dpkg_1.17.21.bb' ? Aug 27 14:33:10 also if do_compile is not defined where this one is inherited from? Aug 27 14:33:37 typically it's -std=c99 You can add that to the CFLAGS in the recipe if necessary.. but I find that very odd that dpkg does do that itself if it actually needs c99 Aug 27 14:33:40 marcin_: is dpkg-native actually being built when building rpi-basic-image? Aug 27 14:34:24 marcin_: I'm not sure it would be by default Aug 27 14:37:14 is anyone successfully using SYSTEMD_AUTO_ENABLE ? I suspect it is broken, and can't find a working reference .. so if someone has one, I'd appreciate the link. Aug 27 14:38:29 see this for full log: http://pastebin.com/NqrJQhyu Aug 27 14:40:21 looks like package is taken to be build Aug 27 15:02:38 huh how does bind-dev work Aug 27 15:03:07 its full of headers but no libraries anywhere Aug 27 15:03:20 oh, static libs only? Aug 27 15:20:20 zeddii: i've used it in the past -- it just arranges to run systemctl enable / disable in the postinst/prerm. if your service file doesn't have the needed wantedby/etc, the enable/disable might not do what you want.. Aug 27 15:30:09 how to add C flags to be used for compilation of some given package? Aug 27 15:33:07 marcin_: CFLAGS_append = " --desired-option" - you'd put that in the recipe itself (or a bbappend if you're applying it to an existing recipe in someone else's layer that you're using) Aug 27 15:33:32 marcin_: another alternative is to do CFLAGS_append_pn-recipename = ... at the configuration level Aug 27 15:37:36 bluelightning: thanks Aug 27 15:38:22 when yocto goes to create a DEB package of the kernel, where are the contents of the DESCRIPTION field sourced from? Aug 27 15:38:27 and what about defining some env variables for specific package? Aug 27 15:42:06 is do_configure and export good place to do so? Aug 27 15:42:18 marcin_: you can certainly do it that way yes Aug 27 15:43:14 daleg_: the DESCRIPTION metadata variable. Aug 27 15:46:15 kergoth: right, of course. I'm trying to hunt down where that is, because yocto (rather, python) is bombing when creating my kernel deb because it's blindly trying to re-encode a UTF-8 string into a UTF-8 string, which python can't cope with and the run bombs with a stacktrace. Aug 27 15:46:42 as with any metadata variable, you can inspect what it's set to, and how it got set that way, with bitbake -e Aug 27 15:46:47 bitbake -e yourrecipe Aug 27 15:47:11 cool, thanks. I'm fumbling around in the dark here given that I've really only been working with yocto for a few days now. Aug 27 16:06:09 hi, I added a line in the preprocess() block in my image recipe, namely "ln -s /lib/foo ${IMAGE_ROOTFS}/lib/bar", but the link fails to appear in my resulting image. I'm wondering where I can find the log file for this step so I can find out what's gone wrong. Aug 27 16:08:41 xerent: are you doing anything to ensure the preprocess() function is being called within the recipe? Aug 27 16:12:50 I guess not :) Aug 27 16:13:39 maybe IMAGE_PREPROCESS_COMMAND += "preprocess; " Aug 27 16:13:41 would be a good idea. Aug 27 16:13:43 thanks! Aug 27 16:13:52 right yes that's what I was getting at :) Aug 27 16:13:59 * xerent hugs bluelightning Aug 27 16:14:13 :D Aug 27 16:14:29 having so much fun working overtime to set up a new iMX6 board Aug 27 16:30:39 * nerdboy waves at the outgoing shift... Aug 27 17:41:22 alright, I found the source of my problem with yocto bombing while packaging a 4.1.5 kernel Aug 27 17:41:42 it's an issue with the description in one of the kernel module packages Aug 27 17:41:58 "cafe-nand kernel module; NAND flash driver for OLPC CAFÉ chip" Aug 27 17:42:30 this gets fed into line 331 of package_deb.bbclass Aug 27 17:42:40 ctrlfile.write('%s\n' % unicode(textwrap.fill(description.strip(), width=74, initial_indent=' ', subsequent_indent=' '))) Aug 27 17:43:23 unicode() was born without a frontal lobe, so if it's fed a unicode string, it freaks out and kills itself Aug 27 17:43:38 THANKS PYTHON Aug 27 17:44:00 that would be a recipe with non-ascii in it, didn't you say you took care of that? :) Aug 27 17:44:21 that was the "orinal out of range" failure? Aug 27 17:44:29 *ordinal even Aug 27 17:44:35 kergoth: I'm using a recipe-kernel that was generated using the 'yocto-kernel' command Aug 27 17:44:41 nerdboy: yep Aug 27 17:44:51 awesome Aug 27 17:45:11 so the backstory is this: Aug 27 17:45:16 1) I want a 4.1.5 kernel Aug 27 17:45:24 the trac version was a commit message somewhere with a non-ascii char in it Aug 27 17:45:29 2) I use 'yocto-kernel' to generate a recipe for that Aug 27 17:45:49 3) build bombs with unicode poison sprinkled somewhere in the mix Aug 27 17:45:53 lesson: bitbake and oe need fixing to not suck with unicode in general Aug 27 17:46:18 bitbake is not alone there... Aug 27 17:46:21 now, the question is where is this "cafe-nand kernel module; NAND flash driver for OLPC CAFÉ chip" string coming from? Aug 27 17:46:41 good question Aug 27 17:46:42 kconf in the kernel source? Aug 27 17:46:51 time for some recursive greps Aug 27 17:46:52 commit msg? or config description? Aug 27 17:47:38 basically what this means, is that there's multiple vectors into the build system that unicode can be injected Aug 27 17:47:47 and if that happens... poof Aug 27 17:47:54 you lose, good day sir Aug 27 17:48:25 * nerdboy also loves the random source file with comment block containing some eastern european char and system thinks file is wrong iso type Aug 27 17:48:53 tends to also make things blow up strangely Aug 27 17:49:11 yes, like baník pýčo :) Aug 27 17:49:35 * nerdboy high-fives JaMa Aug 27 17:49:47 no funny names in comments please Aug 27 17:50:19 you know, this goes back to the golden rule of C compilers: never use unicode in anything ever Aug 27 17:50:26 not in comments, at least Aug 27 17:50:36 daleg_: that's a definite problem, have you oepned a bug in the yocto bugzilla yet? Aug 27 17:50:38 probably should Aug 27 17:50:51 yup Aug 27 17:51:00 kergoth: I plan to, I want to nail down the exact source of this particular issue first so as to have a complete example Aug 27 17:51:11 that's admirable Aug 27 17:51:15 and sadly uncommon Aug 27 17:51:35 but the overall theme is: many vectors for unicode to get into The System; and if that happens, good luck Aug 27 17:51:52 like training lockheed engineers to use the PR template? Aug 27 17:52:07 :) Aug 27 17:52:10 that sounds unpleasant Aug 27 17:52:19 1) what happened, 2) which LAN/system... Aug 27 17:52:50 finally got it into clearquest and they still ignored it Aug 27 17:54:54 PR scrubs were at least "interesting"... Aug 27 17:56:22 ok so who is familiar with the yocto kernel source trees? Aug 27 17:56:40 a recursive grep of it found the culprit location: Aug 27 17:56:42 ./pkgdata/runtime/kernel-module-cafe-nand:DESCRIPTION_kernel-module-cafe-nand: cafe-nand kernel module; NAND flash driver for OLPC CAF\xc3\x89 chip Aug 27 17:57:01 so whatever makes the file: pkgdata/runtime/kernel-module-cafe-nand:DESCRIPTION_kernel-module-cafe-nand Aug 27 17:57:16 er pkgdata/runtime/kernel-module-cafe-nand Aug 27 17:57:34 * nerdboy points at OLPC devs Aug 27 17:57:45 looks like their patch... Aug 27 17:58:25 maybe they don't enable this module in their kernel defconfig and don't run into this Aug 27 17:59:25 who knows, this sausage factory is already turning me into a vegetarian Aug 27 18:01:08 likely drivers/mtd/nand/cafe_nand.c from the kernel source? Aug 27 18:01:22 * Driver for One Laptop Per Child ‘CAFÉ’ controller, aka Marvell 88ALP01 Aug 27 18:02:27 ok, so basically, who or whatever generates those pkgdata files needs to sanitize the strings and ASCII-ify them first Aug 27 18:03:56 or else when Joe Random Kernel Dev spits a new module onto the conveyor belt leading to Torvalds' desk and it has unicode in its description, we're going to have issues Aug 27 18:05:20 or, as a short term thing Aug 27 18:05:37 we can pass on UnicodeDecodeError exceptions Aug 27 18:05:48 because really, do we care? Aug 27 18:09:35 or just try not to encode them ever Aug 27 18:09:49 because we make a lot of assumptions by even attempting to do it Aug 27 18:10:04 whereas if we don't, we're just passing what's there through Aug 27 18:10:23 we might want to just blindly pass them on, but it depends on the package manager. rpm doesn't assume anything about the field contents, they're passed as is according to fray, but not sure about debian Aug 27 18:11:53 maybe this is a deb requirement. but it's still silly that we assume that the string isn't already utf-8 given how shitty python 2's ability to deal with char encoding is Aug 27 18:15:18 um, i would imagine they use it a lot on certain hardware... Aug 27 18:16:23 olpc 1.0 runs off that thing Aug 27 18:16:41 er, boots to nand anyway Aug 27 18:18:59 if was going to stuff into my overlay i would probably enforce filtering and spit out one-and-only-one known encoding Aug 27 18:19:35 but that would be a consensus of "1" so not too hard to convince myself... Aug 27 18:20:19 we really need to make the transition to python3 and move to using unicode everywhere internally Aug 27 18:20:49 Any thoughts on whether to keep wireless-tools as well as iw in packagegroup-base? I'm thinking we probably should keep both until we *verify* that every network configuration mechanism can use the latter Aug 27 18:21:55 * nerdboy votes for python3 and max verification Aug 27 18:22:12 that be one of my "V"'s... Aug 27 18:26:01 wow ok Aug 27 18:26:04 what do you know Aug 27 18:26:11 this fix went in just days ago: Aug 27 18:26:12 http://git.yoctoproject.org/cgit.cgi/poky/commit/meta/classes/package_deb.bbclass?id=1019d8c802d426e0692f3b5939fa34fb46ff002d Aug 27 18:27:34 I'm cherry-picking this over to my pocky branch and trying it Aug 27 18:35:28 success Aug 27 18:38:29 ok, so to recap - a few days ago a fix for all this debian package unicode crap was committed back to master; I'm a fido user, and missed it. I cherry-picked it over to my fido tree, it works, the world seems ok now Aug 27 18:40:20 nice Aug 27 18:43:03 lessons learned: python char encoding is still the pit of despair it has always been; and check master. Aug 27 18:47:47 Fetcher failure: Fetch command failed with exit code 4, output: Aug 27 18:47:50 is the issue Aug 27 18:47:55 whats the reason ? Aug 27 18:54:43 you didn't include the reason Aug 27 18:54:56 summarizing or grabbing part of a log doesn't give us enough info to tell you anything Aug 27 18:54:58 put the log in a pastebin Aug 27 19:01:07 daleg_: what is your requirement to be on release branch? Aug 27 19:01:23 er, more context maybe... Aug 27 19:01:31 being new to yocto, I guess it made sense? Aug 27 19:01:34 should I be on master? Aug 27 19:01:50 probably Aug 27 19:02:11 I am making fido 1.8 build Aug 27 19:02:11 then what good are the releases then? Aug 27 19:02:14 and got error Aug 27 19:02:24 stablize on your end as you reach your "target" goal Aug 27 19:02:42 WARNING: Failed to fetch URL https://github.com/downloads/mozilla/rhino/rhino1_7R4.zip, attempting MIRRORS if available ERROR: Fetcher failure: Fetch command failed with exit code 4, output: OpenSSL: error:14077410:SSL routines:SSL23_GET_SERVER_HELLO:sslv3 alert handshake failure Unable to establish SSL connection. Aug 27 19:02:53 ERROR: Function failed: Fetcher failure for URL: 'https://github.com/downloads/mozilla/rhino/rhino1_7R4.zip'. Unable to fetch URL from any source Aug 27 19:03:18 SSLv3 nice Aug 27 19:03:20 * nerdboy really an outsider, not inner circle Aug 27 19:04:18 they should be stable snapshots pretty much, fixing/adding specific things to that release, also matching docs release Aug 27 19:04:33 you can't sprint forever... Aug 27 19:04:54 OTOH, stuff is changing fast upstream, kernel/u-boot/runtime Aug 27 19:05:25 "vendors" like freescale migrating off their old forks and pushing stuff up Aug 27 19:05:39 can't capture it all at once Aug 27 19:06:00 * nerdboy waits for comments/corrections Aug 27 19:06:06 daleg_: if you're new to yocto, running off a release branch makes perfect sense Aug 27 19:06:11 master changes, a lot Aug 27 19:06:29 but I want to use master Aug 27 19:06:30 lot of work goes into each one trying to make something release-able Aug 27 19:06:53 master is usually "safe" Aug 27 19:07:05 nerdboy: tell that to the autobuilder ;) Aug 27 19:07:18 usually, yeah. but it's the churn that can be hard to work against Aug 27 19:07:22 could hit build failures anywhere depending on image/machine combo Aug 27 19:07:52 i ain't complainin'... Aug 27 19:07:55 every day your entire platform rebuilds and suddenly the toolchain changes underneath you and your code starts emitting new errors... Aug 27 19:07:58 master works for me Aug 27 19:08:13 that's normal Aug 27 19:08:14 I mean, my OS is currently build using Buildroot right now. I'm exploring Yocto as a replacement for Buildroot because I want to have a dev variant of my OS... so my goal right now is not so much to track Yocto master, but to use a Yocto release (Fido, in my case) to get a OS image that is the same as what I currently have with Buildroot Aug 27 19:08:23 * nerdboy used to gentoo ~ Aug 27 19:08:36 nerdboy: depends on how much effort you want to put into tracking oe-core vs how much you want it to just work Aug 27 19:08:57 like i said, works for me Aug 27 19:09:05 daleg_: especially if you're new to yocto, use the fido branch Aug 27 19:09:10 we use master, but not directly. we use specific revisions, locked on a weekly basis, i update after testing to make sure nothing explodes Aug 27 19:09:12 i choose when i want o pull and/or fix Aug 27 19:09:24 yeah, fido is good for a stable baseline Aug 27 19:09:33 i can at least keep myself stable for a little while Aug 27 19:10:06 the the medication runs out Aug 27 19:10:11 *then even Aug 27 19:11:48 daleg_: for it usually goes like this, every few months or so (or when i get a new something i can build for) Aug 27 19:13:47 pull on master, build/fix/pull to get it going, test things on the same (upstream) commit for a while (push/pull from my own repo), spit something out, next Aug 27 19:14:01 that's generally what I do with buildroot Aug 27 19:14:16 there ya go Aug 27 19:14:52 i do push back fixes when i have something Aug 27 19:15:08 they seem to tolerate that from me Aug 27 19:15:43 * nerdboy crashed a couple dev mtgs even Aug 27 19:57:03 otavio: getting a couple of boot errors with console-image-full Aug 27 19:57:14 *imx28 Aug 27 19:57:26 besides the lockdep error Aug 27 19:58:02 repeated after first-boot, so i don't think it will fix itself Aug 27 20:00:59 https://bpaste.net/raw/21ec97a4a59a <= second boot from latest image build (last commit is a week old i think) Aug 27 20:01:33 still, i guess i expected more trouble, so Aug 27 20:01:39 you get a cookie Aug 27 20:02:51 core-image-full-cmdline-imx28evk-20150826050446 Aug 27 20:23:15 * nerdboy waves at belen1 Aug 27 20:50:09 my rootfs is readonly right after building yocto, does anyone know what this could possibly be? Aug 27 20:51:09 booted with "ro" on the kernel command line perhaps? Aug 27 21:16:05 paulg, grub is configured to be ro Aug 27 21:18:01 well, there is your answer then. Aug 27 21:19:05 you can fix it on the fly temporarily with... Aug 27 21:19:10 mount -nv -o remount,rw / Aug 27 21:19:29 then update grub etc. Aug 27 21:19:58 yes I can make it to rw but, shouldn't yocto like make it rw by default? Aug 27 21:23:33 Hi, I enabled libc-backtrace in DISTRO_FEATURES and now glibc fails to build with core-image-minimal Aug 27 21:24:10 cc1: all warnings being treated as errors Aug 27 21:24:10 /home/ubuntu/poky/builds/core-image-minimal-beagle/tmp/work/cortexa8hf-vfp-neon-poky-linux-gnueabi/glibc/2.21-r0/build-arm-poky-linux-gnueabi/sysd-rules:225: recipe for target '/home/ubuntu/poky/builds/core-image-minimal-beagle/tmp/work/cortexa8hf-vfp-neon-poky-linux-gnueabi/glibc/2.21-r0/build-arm-poky-linux-gnueabi/misc/gethostid.o' failed Aug 27 21:24:10 Aug 27 21:38:19 realBigfoot, normally sysV init or systemdisease is responsible for flipping it to rw during bootup. Aug 27 21:38:49 paulg, I see, i am on systemd maybe i have compiled it with wrong tag? Aug 27 21:39:44 I've not used systemdisease enough yet to know much of how it operates. (Not looking forward to it either....) Aug 27 21:40:15 paulg, ahhaha i see Aug 27 21:41:44 guess i'll have to someday, seeing as it has taken on the "we are the borg; prepare to be assimilated" approach in deployment everywhere. :-( Aug 27 21:42:36 paulg, yeah it looks like everything will have systemd after all Aug 27 21:43:42 * paulg wonders if it spread to android too Aug 27 21:44:32 ljose: did you build from scratch Aug 27 21:44:41 hahahaha Aug 27 21:44:51 paulg, it might Aug 27 21:44:56 lols Aug 27 21:45:04 khem: nope just rebuild a previous image Aug 27 21:45:49 BTW, does anybody know if -mapcs-frame is required to get backtraces with ARMv7l Aug 27 21:55:05 ljose: now a days it should provided you have compiled right ABI specific unwind tables into glibc as well Aug 27 21:55:28 in the past it apcs backtrace support was iffy Aug 27 21:55:40 and framepointer based was what worked always Aug 27 21:57:05 khem`: I using glibc backtrace function and building with -rdynamic but getting bogus stacks Aug 27 21:58:02 should I build with -funwind-tables too? Aug 27 22:42:33 depends on readonlyroot options no? Aug 27 22:42:53 should be a setting for that now... Aug 27 22:54:30 there is, read-only-rootfs image feature Aug 27 22:56:11 * nerdboy needed 3 years ago when stuck on oe-classic Aug 27 22:56:26 *that feature even Aug 27 22:57:17 i kinda hacked it, but it wasn't enough we still had to hack stuff even more... Aug 27 22:57:57 those commits were helpful at least Aug 27 23:38:28 nerdboy: I am going a little crazy getting this to work. I have has zero luck with the fragments. So tried creating a .config and copied it to meta-ti/recipes-kernel/linux/linux-ti-staging-4.1/defconfig, but with no luck. Is this the correct spot to put a whole config? Really want to try and get this sorted today.. The other thing is, how can I check to see what config bitbake is using before building the whole image? Aug 27 23:39:22 dhirc: config fragments? Aug 27 23:39:28 yeha Aug 27 23:39:30 yeah Aug 27 23:39:54 denix: Backstory, trying to enable USB audio on the beagle bone black. Aug 27 23:40:43 I added my own meta layer with the fragments, but when I check /proc/config.gz they are not changing Aug 27 23:41:01 dhirc: you can just append your config fragments to KERNEL_CONFIG_FRAGMENTS list Aug 27 23:41:09 dhirc: here's an example - http://git.yoctoproject.org/cgit/cgit.cgi/meta-ti/tree/recipes-kernel/linux/linux-ti-staging_4.1.bb#n62 Aug 27 23:42:04 is there a way to share the downloads folder kind of like you can share the shared state? Aug 27 23:42:44 The reasoning is so you could do a clean build even if the Internet was down and also to speed up new builds for clients on the same local network Aug 27 23:43:15 denix: This looks promising.. Is there a way to check what config bitbake is working with? I don't want to have to wait for a complete build to check the config is correct. Wasted a lot of time doing this! Aug 27 23:44:40 dhirc: run do_configure and check the resulting .config in build/ Aug 27 23:48:20 aurence1 Aug 27 23:48:31 s Aug 27 23:51:48 denix: How do you run "do_configure"? I am pretty green Aug 27 23:52:16 dhirc: bitbake virtual/kernel -c configure Aug 27 23:52:22 thanks Aug 28 00:02:18 denix: KERNEL_CONFIG_DIR = "${S}/ti_config_fragments" - is see {S} means "source" , but where exactly is this directory? Also, do I need to append anything to KERNEL_CONFIG_FRAGMENTS? The examples does KERNEL_CONFIG_FRAGMENTS_append_ti43x Aug 28 00:37:07 do_configure didn't appear to alter the .config. But when I just did a build it seems to have been updated. Maybe another step in the build process modifies the .config with the fragments? Aug 28 00:38:36 That's a lie sorry, it isn't being updated at all. Grepped wrong file Aug 28 00:49:03 the one i showed you last night works Aug 28 00:49:28 did you add _beaglebone to BLAHBLAH_append ? Aug 28 00:50:06 i'd have to look at the code to be sure (and that is really kergoth's job ;) Aug 28 00:50:42 but it should accept your "files" dir config Aug 28 00:51:29 but if the recipe has a */linux dir instead you can just put it there and forget the extra files_blahblah stuff Aug 28 00:52:02 easiest to use existing patch dir as long as it applies to machine Aug 28 00:56:44 nerdboy: I am stumped and it is obvious that I am doing something stupid. I thought I would try denix's approach so I added "KERNEL_CONFIG_FRAGMENTS_append_beaglebone = " ${THISDIR}/usb_bbb.cfg " but when I do 'bitbake core-image-minimal -c configure' it is still not picked up Aug 28 00:57:10 I have tried with KERNEL_CONFIG_FRAGMENTS = .. and KERNEL_CONFIG_FRAGMENTS_append as well Aug 28 00:57:41 wth did i paste last night? Aug 28 00:58:09 ran out of scroll... Aug 28 00:58:31 same Aug 28 01:01:14 okay, this patch works with previous 3.14 recipe, should still work Aug 28 01:02:36 actually it might still be current for yocto beaglebone kernel Aug 28 01:02:48 https://bpaste.net/show/4b6959a7e448 Aug 28 01:03:35 note the correct dir is linux-yocto ${THISDIR}/${PN Aug 28 01:04:02 drop ohci.cfg there Aug 28 01:04:27 *whatever your config is Aug 28 01:07:56 https://bpaste.net/show/a11ae4e3b2eb Aug 28 01:08:57 that was basically a hands-on example for crash course, so i'd kinda like to know why it blows up for you... Aug 28 01:22:45 nerdboy: Where abouts is the .config generated after running bitbake core-image-minimal -c configure? Aug 28 01:32:37 it would have to be in the build dir at that point, no? Aug 28 01:32:54 work/git maybe? Aug 28 01:33:25 why do people ask me to memorize capitals? i always hated that... Aug 28 01:34:54 sorry, gentoo is $workdir/git Aug 28 01:35:36 ls poky-std/build-bbb/tmp/work/beaglebone-poky-linux-gnueabi/linux-ti-staging/4.1.6+gitAUTOINC+e8f07af182-r22b/ Aug 28 01:35:57 there is a build dir in there... Aug 28 01:36:55 called build even Aug 28 01:37:19 but it's empty unless you turn off rm_work Aug 28 01:37:33 except during build obviously Aug 28 02:09:11 dhirc: did try the -c configure option? Aug 28 02:12:07 looks kinda like that? https://bpaste.net/show/0b54badc10c1 Aug 28 02:45:04 I'm trying to modify a file from another package, and I don't get any errors but the contents aren't changed. here is what I have: https://paste.fedoraproject.org/260470/40729857/ (see the do_deploy part) Aug 28 02:45:29 if i change the file name i get an error message about the file not existing, so i'm pretty sure it is run Aug 28 02:45:50 and if i run that command from the command line it works Aug 28 02:45:54 anyone have any ideas? **** ENDING LOGGING AT Fri Aug 28 02:59:58 2015