**** BEGIN LOGGING AT Fri Feb 22 02:59:56 2019
Feb 22 07:32:44 <yocti> New news from stackoverflow: Yocto - exclude files from -dev package <https://stackoverflow.com/questions/54822100/yocto-exclude-files-from-dev-package>
Feb 22 09:33:07 <yocti> New news from stackoverflow: In Yocto, how to read a git repository via ssh on a nonstandard port <https://stackoverflow.com/questions/54823823/in-yocto-how-to-read-a-git-repository-via-ssh-on-a-nonstandard-port>
Feb 22 10:40:22 <yohboy> hi there, i'm trying to add .gdbinit in my build while building gdb but i'm falling in some trouble with this error: "Files/directories were installed but not shipped in any package". I'm definitively missing something but don't know what, someone have any input please ? Here my bbappend and full error : https://pastebin.com/cTr950DU
Feb 22 10:48:40 <rburton> kanavin: reckon you can get a gdkpixbuf upgrade in as part of the meson switch?
Feb 22 10:52:04 <kanavin> rburton, yep, I'll send it as a separate patch though
Feb 22 10:52:13 <kanavin> also looking into glib
Feb 22 10:52:35 <kanavin> also, any chance of looking at virgl patchset?
Feb 22 10:52:43 <kanavin> it's been sitting in master-next for a looong time
Feb 22 10:52:50 <rburton> kanavin: note that buildhistory shows we used to disable the x11 bit of gdkpixbuf globally
Feb 22 10:52:57 <rburton> yeah browsing next this morning
Feb 22 11:04:39 <rburton> kanavin: your submitted patch is a backport now :)
Feb 22 11:04:49 <kanavin> which one?
Feb 22 11:04:55 <rburton> both
Feb 22 11:05:01 <kanavin> which both? :D
Feb 22 11:05:42 <rburton> kanavin: if you're shipping the 0.7.0 tag then i prefer naming the recipe virglrenderer_0.7.0.bb because its not a git snapshot
Feb 22 11:06:04 <rburton> (fetching via fetch being an implementation detail, we don't do virglrenderer_http.bb and set PV in the recipe)
Feb 22 11:06:21 <kanavin> yep
Feb 22 11:10:48 <rburton> kanavin: should qemu-nativesdk be built with gtk+?
Feb 22 11:10:56 <rburton> if it was built with sdl before then it had a UI
Feb 22 11:11:08 <rburton> so surely just removing sdl but not adding gtk+ makes it headless?
Feb 22 11:18:34 <kanavin> rburton, that's a decision made through local.conf, just like for native: https://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?h=master-next&id=e3374bf545ce8ec8c4e78b607365b865af465366
Feb 22 11:18:52 <kanavin> previously there was confusion because sdl was enabled from two different places
Feb 22 11:18:56 <rburton> right
Feb 22 11:19:01 <rburton> maybe mention that in the commit too
Feb 22 11:19:27 <kanavin> yep
Feb 22 11:20:25 <kanavin> I still hope someone figures out how to make virgl work with sdl. I tried everything and wasted an awful amount of time on it, and it would just refuse to start, or start, then crash on virgl guest driver init
Feb 22 11:20:44 <kanavin> or start, but show blank screen instead of sato
Feb 22 11:21:05 <kanavin> (the last one is by the way how fedora's and suse's packaged qemu behaves too, so it has to be an upstream issue)
Feb 22 11:21:46 <rburton> hm i wonder if thats related to the blank screen when decoding video using dri2 under sato
Feb 22 11:22:06 <rburton> with a compositing wm it works but without one dri2 would just see black
Feb 22 11:26:49 <kanavin> oh and btw: proprietary nvidia driver on the host is fully incompatible
Feb 22 11:26:55 <kanavin> not surprising
Feb 22 11:27:34 <kanavin> nouveau works well nowadays, although nvidia keeps its performance much lower than their blob by refusing to provide some necessary firmwares
Feb 22 11:27:50 <kanavin> I guess they see it as just a way to get some kind of output so that people can switch to the blob
Feb 22 11:33:32 <yocti> New news from stackoverflow: yocto: Add .gdbinit in build while building gdb <https://stackoverflow.com/questions/54825939/yocto-add-gdbinit-in-build-while-building-gdb>
Feb 22 11:36:49 <rburton> kanavin: what's the performance hit? what did you test it with?
Feb 22 11:36:53 <rburton> (glxgears doesn't count ;)
Feb 22 11:37:03 <kanavin> rburton, glmark2, let me find the outcomes
Feb 22 11:37:16 <rburton> glmark2 is pretty lame too tbh.
Feb 22 11:37:22 <rburton> as i've discovered recently
Feb 22 11:37:33 <rburton> 2000fps on a nuc with bad gpu
Feb 22 11:38:09 <kanavin> rburton, score 297 on the host (nouveau), 131 in qemu
Feb 22 11:38:38 <kanavin> there were a couple of tests that had a particularly notable slowdown (terrain is one(
Feb 22 11:38:40 <rburton> the uniengine haven benchmark runs at ~10fps on my nuc in normal linux and is easy to install
Feb 22 11:38:52 <rburton> yeah terrain is the only test there that actually hits the gpu in any meaningful way
Feb 22 11:39:03 <kanavin> rburton, I tried uniengine, it gives me like 7fps on the host, so I didn't bother
Feb 22 11:40:16 <kanavin> rburton, google has been working on improving virgl performance, but it's not yet in a release. uniengine in particular has gotten better.
Feb 22 11:40:19 <kanavin> let me find a link
Feb 22 11:40:51 <kanavin> https://www.phoronix.com/scan.php?page=news_item&px=Mesa-19.1-Virgl-Transfer-Queue
Feb 22 11:40:53 <kanavin> there you go
Feb 22 11:42:00 <rburton> cool
Feb 22 11:43:24 <kanavin> rburton, performance is not the primary point however. the primary point is being able to run things that require 'hardware' gl to begin with.
Feb 22 11:44:00 <rburton> sure but "comparable" performance indicates that its not just doing software GL, which sort of works
Feb 22 11:44:04 <rburton> for limited values of sort of
Feb 22 11:44:15 <rburton> (if you enable llvmpipe)
Feb 22 11:49:27 <kanavin> it's basically establishing a link to the host GL (which can be software or hardware, depending on host's mesa drivers and hardware), there is some loss of performance due to transfer overhead (which is being worked on)
Feb 22 11:49:40 <kanavin> but otherwise, it's certainly not doing software gl in the guest
Feb 22 14:35:31 <kanavin> rburton, all fixed :)
Feb 22 14:43:29 <rburton> kanavin: woo
Feb 22 14:44:08 <rburton> kanavin: can you file a bug about the loaders cache cross thing?
Feb 22 14:44:29 <rburton> kanavin: i keep on promising to do a gtk-from-git ci thingy to catch stuff like that
Feb 22 14:44:52 <kanavin> rburton, jku has filed it looooooong time ago
Feb 22 14:45:00 <rburton> maybe a prod then :)
Feb 22 14:45:41 <kanavin> https://bugzilla.gnome.org/show_bug.cgi?id=779057
Feb 22 14:45:43 <yocti> Bug 779057: was not found.
Feb 22 14:45:59 <kanavin> basically upstream doesn't know what to do about it
Feb 22 14:46:22 <kanavin> the best we can do for now is to run native versions of the respective tools and hope it works
Feb 22 14:47:09 <rburton> upstream claim its fixed https://gitlab.gnome.org/GNOME/gdk-pixbuf/commit/e191c75683e6e1db639a0e4213a169fdb3a805ea
Feb 22 14:47:26 <rburton> via disabling stuff, mostly
Feb 22 14:47:47 <kanavin> rburton, that fix is not yet in a release
Feb 22 14:48:00 <kanavin> once they release it, we can remove our patching
Feb 22 14:48:20 <kanavin> I kinda considered disabling, but it would mean all of the ptests would be gone, for one thing
Feb 22 14:49:08 <kanavin> if not meson.is_cross_build()
Feb 22 14:49:08 <kanavin>   subdir('tests')
Feb 22 14:49:08 <kanavin>   subdir('thumbnailer')
Feb 22 14:49:08 <kanavin> endif
Feb 22 14:49:14 <kanavin> can you really call this
Feb 22 14:49:18 <kanavin> 'fixed'? :)
Feb 22 14:49:22 <rburton> no
Feb 22 14:49:24 <rburton> worked around
Feb 22 14:49:25 <kanavin> yep
Feb 22 15:28:02 <rburton> armpit: is rocko still getting patches?
Feb 22 15:28:27 <rburton> hm looks like no
Feb 22 16:40:55 <armpit> RP, finally got around to sumo pull request. should be good for next point release
Feb 22 17:13:23 <RP> armpit: I was wondering if I missed it! :)
Feb 22 17:15:50 <armpit> RP, nope... I was just slow
Feb 22 17:17:29 <Crofton> too much beer
Feb 22 17:23:10 <khem> JPEW:http://ix.io/1BLi have you seen these kind of errors
Feb 22 17:24:16 <khem> RP: are plans to switch qemuarm to v7 on for 2.7 ?
Feb 22 17:24:46 <georgem> seems I missed the boat with my sumo CVE patches for systemd. doh
Feb 22 17:26:03 <RP> khem: jonmason is supposed to be doing it
Feb 22 17:26:20 <RP> jonmason: We're about to hit feature freeze so I'm getting concerned :/
Feb 22 17:26:44 <khem> jonmason: whats blocking it ?
Feb 22 17:31:09 <RP> khem: time :(
Feb 22 17:31:58 <Crofton> I can drive over and beat him
Feb 22 17:41:49 <gsalazar> Hi, I'm using the meta-ros layer from https://github.com/bmwcarit/meta-ros/ and hit an issue where it cannot compile rostime due to missing lpthread. I followed the steps in the wiki, has anyone hit this problem before?
Feb 22 17:59:37 <khem> gsalazar: add CLFAGS += "-pthread" to recipe
Feb 22 17:59:49 <khem> CFLAGS I mean but you get it
Feb 22 18:06:54 <JPEW> khem: Ya, I had to blacklist pkg-config
Feb 22 18:07:02 <JPEW> I need to update the master blacklist :(
Feb 22 18:10:02 <gsalazar> khem: to the meta-ros recipe?
Feb 22 18:10:13 <khem> gsalazar:yes
Feb 22 18:10:16 <khem> JPEW:OK
Feb 22 18:11:12 <JPEW> khem: Here is my current blacklist: ICECC_USER_PACKAGE_BL = "pkgconfig json-c glib-2.0 bison gettext"
Feb 22 18:12:56 <khem> I added that to my site.conf with ICECC_USER_PACKAGE_BL_append
Feb 22 18:14:52 <gsalazar> khem: thanks
Feb 22 18:15:37 <jonmason> khem: I have FB working on qemuarmv7.  I just can't get X working
Feb 22 18:15:44 <jonmason> that is the blocker
Feb 22 18:16:00 <jonmason> I'll send out patches to the mailing list asking for help
Feb 22 18:23:10 <gsalazar> khem: changing the recipe didn't fix the issue, it asks: "Did you compile project 'cpp_common'?"
Feb 22 18:43:56 <khem> hmm
Feb 22 18:44:11 <khem> you need to add -pthread to linker command
Feb 22 18:44:18 <khem> so either via recipe or via makefile
Feb 22 18:44:28 <khem> whatever the package respects use that approach
Feb 22 19:35:12 <yocti> New news from stackoverflow: bitbake BBPATH confusion <https://stackoverflow.com/questions/54833674/bitbake-bbpath-confusion>
Feb 23 01:25:38 <halstead> AB maintenance complete for the week.
**** ENDING LOGGING AT Sat Feb 23 02:59:56 2019