**** BEGIN LOGGING AT Sat Jan 04 02:59:58 2014 Jan 04 19:19:46 when using meta-qt5, how to i tell bitbake to pick qtbase_git.bb instead of qtbase_5.1.1.git? adding things like QT_VERSION = "git" or PREFERRED_VERSION_qtbase = "git" to local.conf doesn't seem to work Jan 04 19:26:09 hasselmm: use real PV, see example in meta-qt5/conf/distro/include/qt5-versions.inc Jan 04 19:26:20 you also need to include this file in your local.conf Jan 04 19:29:49 jama: oh, i have to include that file. ok, let's see if i can figure that out. a straight "require qt5-versions" obviously fails Jan 04 19:30:50 require conf/distro/include/qt5-versions.inc Jan 04 19:30:58 QT_VERSION = "5.1.0+git%" Jan 04 19:33:14 JaMa: too obvious. thanks! Jan 04 19:33:26 why do you want to use git recipes? Jan 04 19:33:35 are you aware that there is branch with 5.2.0 recipes? Jan 04 19:34:12 JaMa: i am using jansa/qt5-5.2.0, but there are no 5.2.0 recipes yet Jan 04 19:34:16 unless you want something newer than 5.2.0 I would strongly recomment using that branch (otherwise you will have to rebase quite a few especially qtbase patches) Jan 04 19:34:34 ok then Jan 04 19:35:32 btw, would you mind adding machine specific settings to your layer? i've seen .bbappend files for imx5/6 and some ti chips Jan 04 19:35:53 probably there also are .bbappend files for the rpi, which i just didn't find yet Jan 04 19:36:00 they should stay in machine specifil layer :) Jan 04 19:36:06 ok :-) Jan 04 19:36:59 JaMa: other thing: are you working with someone already to get you massive number of qt patches upstreamed? Jan 04 19:37:22 otherwise i might poke some collegues of me to maybe spend their qt-contrib time Jan 04 19:37:50 hasselmm: let me know how 5.2.0 works for you (I did mostly just build tests and haven't found time yet to update our images to test it in runtime too) Jan 04 19:38:28 hasselmm: I got some offer from digia and I've tried to upstream some of them earlier, but there was minimal progress on that front Jan 04 19:38:41 hasselmm: so any help with upstreaming them would be highly appreciated Jan 04 19:39:05 ok, i'll poke some guys on monday then Jan 04 19:41:09 ...compiling will take some time. only have my private laptop here. Jan 04 19:41:25 hasselmm: fwiw: I have discussed "linux-oe-g++ platform" with upstream devs and Oswald was the only one on channel who understood why we want to do something like that Jan 04 19:41:48 hasselmm: but still it's too different to other specs platforms so it isn't good candidate for upstreaming Jan 04 19:42:03 JaMa: honestly it also took me some time to understand linux-oe-g++ platform :-) Jan 04 19:42:30 already started my own minimal recipes :-D Jan 04 19:42:42 hasselmm: -external-hostbindir is a lot better for upstreaming (and it was already sort of approved by Oswald) but in 5.2.0 I've noticed that he added something similar for android builds Jan 04 19:42:52 just to realize that i got closer and closer to your recipes Jan 04 19:43:12 JaMa: -external-hostbindir definitely makes sense imho Jan 04 19:43:24 hasselmm: so it should be consolidated a bit with that added functionality (I've just rebased my solution, because that was what was well tested in OE environment) Jan 04 19:46:31 JaMa: actually i am slightly surprised to see no digia commits in your layer. after all their "qt enterprise for embedded" (aka. boot2qt) is based on yocto Jan 04 19:46:58 well. some yocto images with a manually built qt on top. Jan 04 19:48:38 hasselmm: well honestly I'm surprised that, quite a lot of people shown interest in this layer and only few people actually sent some patches back Jan 04 19:49:06 hasselmm: and I don't expect that it works as-is so good for everybody, that thay don't need any patches on top of it Jan 04 20:12:54 well, cross-compiling qt5 by hand isn't that hard... so either your layer _is_ good enough, or they just do what digia did and plumb a hand-built qt on top Jan 04 20:17:26 did qt5 make qmake less corss-unfriendly? Jan 04 20:30:05 I wouldn't say that, just different Jan 04 20:30:27 most issues I had in meta-qt5 was with sstate and "relocation" of tools Jan 04 21:29:24 early qt/arm such as on the zaurus wasn't bad performance-wise... Jan 04 21:29:56 what about today? do you need a quad-core with a bunch of ram? Jan 04 22:06:35 to do what.. Jan 04 22:39:59 nerdboy: depends on what you do. plain qwidgets or qml has a pretty low memory footprint, compared to html5 platforms like webkit or chromium Jan 04 22:41:10 nerdboy: have a project where the pretty complex qml ui takes less than a second for startup on single core imx6 with 128 MB ram Jan 04 22:41:58 most important is having a good GPU those days. if that one sucks even 16 cores won't save you Jan 04 22:42:55 things that make you go "hmmm..." Jan 04 22:44:09 well, and even a nifty CPU+GPU won't safe you, if you messup your video driver setup Jan 04 22:45:09 i pretty much stay far away from vendor graphics driver blobs on x86 Jan 04 22:45:19 we once had a desparate customer were a nifty OMAP only managed to render something like 6 FPS because they entirely messup their framebuffer setup... Jan 04 22:45:23 can't really do that on arm... Jan 04 22:45:26 ...and then even threw compositing X.org on top Jan 04 22:45:48 we gave them 18 FPS. more we couldn't do with their nearby deadlines Jan 04 22:45:55 got a new trimslice build with all opensource tegra stuff Jan 04 22:46:04 and still even that platform should have done 60 FPS easily Jan 04 22:46:28 works okay in X, now building some other stuff for video/audio testing Jan 04 22:46:43 nerdboy: really depends on the SOC what's best Jan 04 22:47:04 and the state of various opensource projects Jan 04 22:47:21 imx5.3 + qt works best with X11. the eglfs drivers are total junk. for imx6 it's the opposite Jan 04 22:47:33 too many things to test/play with... Jan 04 22:47:59 yup. needs thoughtful testing Jan 04 22:48:16 oh, probably time to update the efikamx then Jan 04 22:56:35 freedreno is supposed to be in pretty good shape but i don't have any qualcomm devices... Jan 04 23:18:25 oh bitbake! what are do doing to my ccache!? Jan 04 23:18:33 why are do deleting it before each build? Jan 04 23:18:50 how do you expect it to be useful when doing such nonsense? Jan 04 23:19:28 * hasselmm cries for the 21G of precompiled files just lost Jan 04 23:40:48 hasselmm: AFAIK, ccache isn't enabled by default... **** ENDING LOGGING AT Sun Jan 05 02:59:58 2014