**** BEGIN LOGGING AT Fri Mar 08 02:59:56 2019 Mar 08 05:07:32 New news from stackoverflow: How to get static library in sdk? Mar 08 07:23:10 Is there a way to "package feed" the SDK generated by Yocto? Mar 08 07:23:48 Insted of exporting the big SDK package. Can I somehow distribute it with yum or apt-get? Mar 08 07:30:58 Holk: there shuold be nothing that actvely forbits it, but we do not have a packaging mechanism for the sdk/esdk artifacts at this time. Mar 08 07:32:31 Holk: OTOH, it becomes pretty complicated then to install multiple sdks Mar 08 07:32:56 I see Mar 08 07:33:04 Do you think its possible? Mar 08 07:33:15 you either end up with massive naming duplication, or clashes Mar 08 07:33:23 possible, sure. Mar 08 07:33:51 But not recommended? :P Mar 08 07:34:37 well, define recommended. lets say, obviously nobody had a usecase for it until now that was so important that it was actually done Mar 08 07:35:21 Holk: one thing to think about is that you basically have, in the worst case, a sdk for each machine-distro-image triplet that you work on Mar 08 07:36:34 Holk: so its not "the SDK" that you would package, but a multitude Mar 08 07:37:52 Well since there is no official support I guess I will have a hard time convincing my boss its a good idea Mar 08 07:38:02 It was just a thought I had the other day Mar 08 07:38:50 Our project wants to distribute the SDK to other sites which are in other countries and they thing the SDK blob is too big Mar 08 07:40:18 Holk: well, based on what do you think that a packaged sdk would be smaller? Mar 08 07:41:06 it would still need to contain the same thinks, so it would merely be like some shrinkwrap for easier shipping/installing. Mar 08 07:42:29 I personally dont have a problem with the size. Its just the management spitting out the word simple and cheap in every sentense and now I have to come up with some kind of solution Mar 08 07:43:13 Thanks for you time LetoThe2nd Mar 08 07:43:16 your* Mar 08 07:43:54 Holk: you basically have three sizes. 1) toolchain only. small, but doesn give you the benefits of libraries etc. matching the image. 2) sdk: mid-size, offers toolchain and everything to compile for a given target. 3) esdk: biggest, offers the full-blown shbang to do OE development for a given target Mar 08 09:26:06 Hi everyone Mar 08 09:26:27 I don't thing everyone is currently online/active Mar 08 09:26:30 But hi! Mar 08 09:26:32 hehe :D Mar 08 09:27:02 I'm having some problems with building a 4.4 kernel with thud 2.6.1 Mar 08 09:27:28 is 4.4 unsupported? Mar 08 09:27:56 I can't upgrade my kernel, due to nonexistant drivers Mar 08 09:28:12 I'm using the default poky distro Mar 08 09:30:10 I'm using this kernel: github.com/rockchip-linux/kernel Mar 08 09:31:30 klemen: please define "problems" (optimally by providing the logs in a pastebin) Mar 08 09:32:15 LetoThe2nd, just a second, let me start the build Mar 08 09:41:10 it's taking a bit longer than expected Mar 08 09:42:01 we're in no hurry :) Mar 08 09:51:04 LetoThe2nd, here's the output: https://pastebin.com/xcmjisuP Mar 08 09:51:48 I should say, that I had success with building the same kernel with rocko 2.4.4 Mar 08 09:53:34 klemen: hm Mar 08 09:53:55 hard to say, but i'd guess that 4.4 breaks with the new gcc Mar 08 09:54:12 klemen: you could try to nail gcc down to something older Mar 08 09:55:41 LetoThe2nd, I see.. how can I do that? Mar 08 09:56:15 klemen: just checked, its actually not that simple. sumo came with gcc7, thud with gcc8, and no secondary choices for both Mar 08 09:57:25 klemen: yet this is fixed in the LTS of 4.4: https://cdn.kernel.org/pub/linux/kernel/v4.x/ChangeLog-4.4.145 Mar 08 09:57:32 klemen: so thats your answer Mar 08 09:58:51 LetoThe2nd, so I just upgrade to 4.4.145 and I'll be fine? :-) Mar 08 09:59:22 klemen: if that maybe-frankenkernel updates easily to 4.4.145 or newer, then yes. Mar 08 10:00:35 LetoThe2nd, okay. thank you for your help! :-) Mar 08 10:00:42 klemen: actually https://github.com/rockchip-linux/kernel/blob/release-4.4/Makefile doesn't look that bad. Mar 08 10:00:55 so my guess is that just the layer is badly maintained. Mar 08 10:01:16 klemen: have fun then. Mar 08 10:01:35 LetoThe2nd, will do, thanks :-) Mar 08 10:08:28 New news from stackoverflow: How can i add 'GTK+3' in my toolchain (Yocto SDK) Mar 08 10:15:07 in multilib configuration the variables can only be MACHINEOVERRIDES dependent? Cannot be dependent from TUNE? Mar 08 10:17:25 if I want a variable to have different value in 32 and 64 bit configuration. The only I found is the MACHINEOVERRIDES which adds a "common" suffix to the MACHINE variable Mar 08 13:05:31 Hi, I have another newbie question but I'm not sure how to phrase it. When i install for example : fsl-image-qt5 I get two directories with fonts : /usr/share/fonts and /usr/lib/fonts. For my image i need the /usr/lib/fonts but i cannot figure out which recipe that adds it, so that i can port it to my image. I'm using qt 5.6.2 so fonts should be included as I understand it(removed in 5.7) Mar 08 13:06:59 wildwillie: oe-pkgutil should be able to tell you which recipe provides a certain path Mar 08 13:14:29 rsync command is not included in HOSTTOOLS. should this be included in bitbake.conf or change the recipes that execute that? Mar 08 13:17:02 I mean do we care to see rsync upstream in HOSTTOOLS? Mar 08 13:20:40 LetoThe2nd: Did you mean oe-pkgdata-util? Mar 08 13:22:44 wildwillie: yeah Mar 08 13:37:01 Hi, I'm having some problems getting the console output on HDMI. I'm building core-image-base with poky thud 2.6.1 Mar 08 13:37:22 I do get the yocto project splash screen Mar 08 13:37:26 so HDMI seems to work Mar 08 13:48:44 hi all is there a way to force uninative ? the build system doesn't like my glibc version Mar 08 13:57:28 LetoThe2nd: I ended up forcing it to use another font dir since i dont need that many anway. But thanks that is very usefull script!! Mar 08 14:13:00 rizzitello: INHERIT += "uninative" Mar 08 14:13:14 in the local.conf? Mar 08 14:13:55 rizzitello: or your distro Mar 08 14:30:03 Is it possible to use some sort of dummy server for X11? That is entirely blank. I was here asking about using custom themes a few days ago, and I have realized that it would be even better to have nothing. My qt application will launch at boot but takes 1-2 seconds in which the desktop will show Mar 08 14:30:47 I'm using the default theme atm but it feels unnecessery Mar 08 14:31:24 sato has logic to not switch to X until it is ready Mar 08 14:31:29 where ready is up to you Mar 08 14:31:55 not sure how a dummy X server will help Mar 08 14:32:02 assuming you want to see something at some point Mar 08 14:32:23 maybe not the best idea, i just want it to be black until my application boots up Mar 08 14:32:31 how will a dummy x server help? Mar 08 14:32:56 If i have no X server i cannot start the application Mar 08 14:33:05 how will a dummy x server help? Mar 08 14:33:19 (you can't move apps between X servers) Mar 08 14:34:34 what Sato used to do, but doesnt appear to right now, is start X with the novt flag so it doesn't switch the active vt on startup Mar 08 14:34:43 so when it starts the user still sees the splash or whatever Mar 08 14:35:00 then when sato is fully loaded, it fires a dbus signal, and the init script does a chvt Mar 08 14:35:19 voila, splash->app without any loading Mar 08 14:35:26 not sure why this is commented out right now, but it's not hard Mar 08 14:35:31 that is exactly what i want to achieve Mar 08 14:35:59 -novtswitch Mar 08 14:47:41 rburton: Sorry but what sato recipe had this function? Mar 08 14:55:08 All i can find is theme settings Mar 08 14:55:28 its spread around. the xserver-nodm init script used to wait but its now commented out. your app needs to send a dbus signal, matchbox-desktop does that Mar 08 14:58:24 # Required-Start: $local_fs $remote_fs dbus in combo with matchbox-desktop Mar 08 14:58:42 Alright thanks again, I'll poke around a bit :) Mar 08 15:09:11 * kergoth yawns Mar 08 15:17:36 rburton: ok, the mbwm now shows compile-time compositing is enabled, but i STILL am not getting transparency. is there something further to do, e.g., some command line option to give? Mar 08 15:19:33 "mbwm --help" does not show any composite enable options Mar 08 15:29:39 rburton: ? Mar 08 16:26:28 anyone? Mar 08 16:27:33 * yates burps rudely Mar 08 16:36:06 On sumo, with reproducible builds, I (still) see that the .pyc is not used. python3 -v -c "import time" gives "bytecode is stale for 'encodings'". Resulting in slow python execution. Mar 08 16:36:17 This is for a read-only filesystem. Mar 08 17:24:27 yates: are you sure your app is actually doing transparency correctly Mar 08 17:24:37 like have you seen it work with another WM Mar 08 17:26:37 of course mbwm basically fullscreens the windows so not sure what your expecting to be transparent Mar 08 17:27:01 popups? Mar 08 17:28:18 [ Mar 08 17:31:01 Hmm, anyone with experience withd evtool upgrade? I'm getting "unable to find revision ... in branch master even from upstream" with the new srcrev, even though i explicitly specified a non-master branch on the command-line Mar 08 17:31:06 thud, not master Mar 08 17:31:10 devtool, that is Mar 08 18:06:09 rburton: yes, exact same code works with transparency under fedora Mar 08 18:06:52 i think fullscreen is orthogonal to transparent.. Mar 08 18:07:38 but what i'm doing is creating a widget with rounded corners where the corners show the main window's background Mar 08 18:09:01 The phrase "rounded corners" does not really represent the full scope of what needed to be done. A more complete explanation would be "Make a widget with a gradient background with rounded corners which are transparent and show the main frame's background through the corners, when the main frame's background is itself a gradient." Mar 08 18:09:43 so you can't cheat with tricks like setting the widget's background color to the parent window's background color.. Mar 08 18:14:06 oh in the window itself Mar 08 18:14:16 thats nothing to do with window manager level transparency Mar 08 18:14:29 unles the widget is a popup window Mar 08 18:15:12 i'd look at what wx is doing Mar 08 18:15:39 ie is the wx you build using gtk2 or something stupid Mar 08 18:29:55 Can someone with a Raspberry PI 3 B+ test an image for me? I have it working fine with RPI 3 b, but someone said the image doesn't work well with RPI 3 B+. I don't have any issue with this on my RPI 3 B+. The issue is related to bcm2835_sdhost . Mar 08 19:19:16 RP: finally my problem with the ??= PACKAGECONFIG yesterday was a fluke on my end, I had a rogue PACKAGECONFIG += "manpages" inside my bbappend for tools. Mar 08 19:44:43 armpit, RP, rburton AB is ready for new work. Mar 08 19:47:28 we sho uld really stop using ??= and ?= for PACKAGECONFIG, it's just way too error prone Mar 08 19:47:36 we shouldn't support setting it globally anyway Mar 08 19:47:43 anyone who wants to adjust it uses recipe overrides Mar 08 19:53:48 halstead, thanks Mar 08 19:53:58 :) Mar 08 20:30:20 kergoth: I wasn't expecting the += to nuke the ??= ... anyway might be a rookie mistake Mar 08 20:31:10 += is an immediate operation, using it sets it globally. then later on when the recipe is parsed, the defaults arent applied, because it was already set globally. we shuld just stop using ?=/??= for it in the recipes and it'd be a non-issue. can still use _append/_prepend/_remove with the pn- override Mar 08 20:31:27 but yeah, it's a common enough mistake with variables of that sort. IMAGE_INSTALL has the same problem in some cases Mar 08 20:31:38 which is part of why CORE_IMAGE_EXTRA_INSTALL exists Mar 08 22:22:00 halstead: thanks! Mar 08 22:22:12 psrcode: ah, glad you figured it out at least... Mar 08 22:22:33 zeddii: thanks for the patches, I'll put them in for testing. Looks like we have a regression with the kernel-devsrc patch Mar 08 22:23:32 oh ? which arch ? I can test it here tonight before I go poof for a week. Mar 08 22:31:48 i hate it when projects never merge back their tags to master. i know why they don't, the overhead is a pain, but it's still pretty irritating at times. have to muck with merge-base, can't make accurate use of git-describe, etc Mar 08 22:40:58 zeddii: qemumips: https://autobuilder.yoctoproject.org/typhoon/#/builders/60/builds/372/steps/7/logs/step1b Mar 08 22:41:04 zeddii: its with 4.9 Mar 08 22:41:13 er, 4.19 Mar 08 22:41:42 zeddii: also qemumips64: https://autobuilder.yoctoproject.org/typhoon/#/builders/74/builds/376/steps/7/logs/step1b Mar 08 22:42:48 zeddii: and edgrouter: https://autobuilder.yoctoproject.org/typhoon/#/builders/62/builds/376/steps/7/logs/step1b Mar 08 22:43:13 zeddii: I think you just need to tweak the mips syscall install you added to fail gracefully Mar 08 22:45:31 zeddii: I suspect I can sort it if you're out of time Mar 08 22:49:24 damnit. I meant to make them all graceful. I meant to loop back to that. Mar 08 22:49:38 I should have time. will send you an email at the end of my night here, to let you know either way. Mar 08 22:51:21 RP: yah. I see it now, will send a patch. leave that one with me Mar 08 22:51:38 * zeddii scolds himself. I thought those were around in the old releases, should have checked. Mar 08 22:52:06 zeddii: its easily done Mar 08 22:53:16 zeddii: my current plan is to go 5.0 everywhere except for qemuarm and qemuarm64 and to switch qemuarm to the v7 for M3 Mar 08 22:53:34 zeddii: that would then let us figure out 5.0 for qemuarm* for 2.7 final Mar 08 22:53:46 zeddii: worst case they'd have 4.19 Mar 08 23:05:28 agreed. **** ENDING LOGGING AT Sat Mar 09 03:00:34 2019