**** BEGIN LOGGING AT Sat Oct 27 03:00:02 2012 Oct 27 03:25:27 morning! Oct 27 03:25:32 morning penk! Oct 27 03:25:36 * PatrickC is about to go AFK again lol Oct 27 03:26:06 PatrickC: what timezone are you lol Oct 27 03:26:13 Central Oct 27 03:28:45 * LarrySteeze kicks PatrickC to the cub Oct 27 03:28:48 * LarrySteeze kicks PatrickC to the curb Oct 27 03:28:54 LarrySteeze! Oct 27 03:29:03 Me! Oct 27 03:29:04 been forever Oct 27 03:29:15 yeah, I've been busy Oct 27 03:29:27 how's phoenix Oct 27 03:29:39 goin good... Oct 27 03:29:40 PM? Oct 27 03:29:48 in half an hour it'll be AM Oct 27 03:29:57 haha Oct 27 03:30:07 woah... traffic Oct 27 03:30:15 invalidopcode: no one asked you Oct 27 03:30:16 conversation... what is this? Oct 27 03:30:30 invalidopcode there's always traffic when I'm around Oct 27 03:30:32 PatrickC: who said I was talking to you? Oct 27 03:30:41 I've just been away for a couple weeks :-( Oct 27 03:30:57 invalidopcode: you get the email about tomorrow? Oct 27 03:31:02 s/email/invite/ Oct 27 03:31:02 PatrickC meant: invalidopcode: you get the invite about tomorrow? Oct 27 03:31:18 PatrickC: I did not... should I have? Oct 27 03:59:57 friday night, all the webos guys are out partying? Oct 27 04:47:32 probably Oct 27 05:37:41 EricBlade1: ping? Oct 27 06:03:38 PatrickC: sup! Oct 27 06:04:05 nvm, was gonna ask you if you wanted to try something.. but i tried it and it failed miserably :P Oct 27 06:05:36 oh :D bbl Oct 27 06:48:59 morning Oct 27 06:53:37 Stskeeps: my kernel should _not_ enable TLS, isn't it ? Oct 27 06:53:49 nah Oct 27 06:54:00 it's just that the bionic c library we'll abuse shouldn't even remotely touch the TLS area Oct 27 06:54:05 so to aid debugging, we add traps Oct 27 06:54:50 oki; so between gingerbread and ics there are other entry points where tls is used Oct 27 06:55:15 (because, well, your libc.so works for the GNex !) Oct 27 06:57:17 mmmh if my egl blob from nvidia needs TLS, it's a dead-end, then, isn't it Oct 27 06:58:09 that's fine, the opengl slots are in a safe place from glibc pov Oct 27 07:00:00 so it's not always fatal if __init_tls is called ? I'm a bit lost Oct 27 07:01:06 practically it won't matter, but this is all a bit of black magic :) Oct 27 07:01:11 I'll have a second look at the modifications you did on bionic_tls.h, maybe I'll understand better Oct 27 07:01:25 you have to visualize the TLS areas of glibc and bionic side by side Oct 27 07:01:44 the location of the opengl TLS slots in bionic fits in an area reserved in glibc for future expansion Oct 27 07:02:25 oh... that's touchy :) Oct 27 07:03:40 the problem is that somebody when doing android had a brilliant idea of hardcoding this crap Oct 27 07:03:43 :P Oct 27 07:04:03 hence, bionic_tls.h patch Oct 27 07:05:15 correct, where we move what we acn Oct 27 07:05:18 can Oct 27 07:05:19 ok, so my pb is only that I introduced the "*0 = -1" thing, but it's not supposed to work with that part of the patch, it's only for debug Oct 27 07:06:08 how did you make 0 == -1 ? Oct 27 07:07:11 EricBlade: in a modified Bool algebra, applied to the bionic libc Oct 27 07:07:24 Tofe: where do you patch this, btw? Oct 27 07:07:28 more seriouisy: it's "*(int*)0==-1" Oct 27 07:08:00 oh. i thought you assigned -1 to 0 Oct 27 07:08:49 Stskeeps: I took the ICS source from linaro, because they have clear build intructions Oct 27 07:09:13 so the patch is applied in bionic/libc directory Oct 27 07:09:24 right Oct 27 07:09:38 I then start a minimal build of ICS, and stop it when libc is built Oct 27 07:09:55 Also I forced it to take _my_ kernel Oct 27 07:11:30 one other thing is that I now build the "ics" part of libhybris, but I don't know if that really matters Oct 27 07:12:59 All this, just because libc from morphis was built with NEON ^^ Oct 27 07:24:50 still fails, but goes further. now I see some pthread_condattr_xxx = -6, and it segfaults Oct 27 07:26:08 Stskeeps: where did you get the glibc slot numbers ? maybe I fixed it wrongly Oct 27 07:27:03 there's no such thing :) just structures Oct 27 07:27:17 maybe the nvidia stack uses pthread calls we don't support yet Oct 27 07:30:42 well, the exact crash seems to be in pthread_cond_timedwait, I'd say. Or maybe pthread_mutexattr_setpshared. But it sure crashes :) Oct 27 07:38:44 Tofe: what exact instruction pointer does gdb say? Oct 27 07:53:26 Stskeeps: oh, wait -- I don't have gdb on my tablet yet :p Oct 27 07:53:43 good idea to get Oct 27 07:54:12 yep, seems so, that will need a partial build of webos-image... Oct 27 08:28:31 ah, weird: webos-dev-image doesn't build out of the box... Oct 27 08:29:05 I have a conflict between mesa-dri and my libgles-a500, both provide egl Oct 27 08:29:31 How can I know the dependency graph leading to mesa-dri ? Oct 27 08:30:48 Tofe: bitbake -g Oct 27 08:30:54 okay Oct 27 08:58:48 "webos-dev-image.do_build" -> "mesa-dri.do_build" Oct 27 09:00:00 Tofe: you selected the correct provider for libgles/libegl? Oct 27 09:00:05 in your machine configuration? Oct 27 09:00:22 ah, I'll have to have a second look at that Oct 27 09:08:25 JaMa: meta-webos updated? Oct 27 09:09:19 ah ok, I understand; as qt4-webos didn't failed to find the opengl libs when building webos-dev-image, I added virtual/libglesv2 and virtual/libglesv1 to its dependencies. But the latter wasn't speficied in my preferred providers, thus leading to mesa-dri. But qt4-webos still doesn't find the opengl libs... this is weird, as there was not pb with the webos-image. Oct 27 09:09:34 s/didn't// Oct 27 09:09:35 Tofe meant: ah ok, I understand; as qt4-webos failed to find the opengl libs when building webos-dev-image, I added virtual/libglesv2 and virtual/libglesv1 to its dependencies. But the latter wasn't speficied in my preferred providers, thus leading to mesa-dri. But ... Oct 27 09:09:55 morphis: not yet, but new PR sent :) https://github.com/openwebos/meta-webos/pull/44 Oct 27 09:10:07 morphis: jansa/master-next if you want it sooner Oct 27 09:10:22 no problem, I will wait Oct 27 09:11:30 anyone interested in builds for 64bit target? Oct 27 09:12:55 yes, danny is out, hope they get the point and update their reference oe-core very soon Oct 27 09:13:23 JaMa: btw. I will do a recipe for connman 1.9 very soon as thats what we want to base our work regarding network on Oct 27 09:13:30 will submit it to oe-core as well Oct 27 09:13:36 but we have to maintain it in meta-webos Oct 27 09:13:58 until we switch to oe-core/master again... Oct 27 09:15:04 yes Oct 27 09:30:28 ok, I now understand the weak point. I'm missing coffee. Oct 27 11:18:49 Roy__: ShiftyAxel: I've updated local-builds.inc example again to make it even easier https://github.com/webOS-ports/webos-ports-setup/commit/8cbc42fb3cfb5028ea4f37c9252d7653e503252a Oct 27 11:21:11 morphis: updated meta-webos pushed Oct 27 11:22:22 JaMa: great, will pull and try compiling it Oct 27 11:25:17 JaMa: I will change repository for connman-qt in the near future, should I bump PE or we're not yet interested in backward compatiblity Oct 27 11:26:42 I'm OK without PE bump Oct 27 11:27:25 ok Oct 27 11:27:39 btw when luna-sysmgr is built for qemux86-64 it fails to find affinity.h, looking at webos-ports/master diff it's maybe caused by our changes? Oct 27 11:28:04 yes, it can Oct 27 11:28:40 we have to adjust for the latest commit in luna-symsgr Oct 27 11:28:45 it changes the way we have to build it Oct 27 11:29:14 do you already bother with backward compatibility ? :) Oct 27 11:29:24 * JaMa starting another qemux86-64 build Oct 27 11:29:33 Tofe: it's more like upgrade path on target device Oct 27 11:29:41 ah ok Oct 27 11:29:48 Tofe: so that devs doing opkg upgrade wont miss that change Oct 27 11:30:05 yes, that's right Oct 27 11:41:44 JaMa: updating meta-webos requires recompile qt4/webkit? Oct 27 11:44:33 yes, but only due to PR bump Oct 27 11:44:53 webkit-supplemental is only recipe with real changes Oct 27 11:45:28 s/real/important/ Oct 27 11:45:28 JaMa meant: webkit-supplemental is only recipe with important changes Oct 27 11:45:29 ok Oct 27 11:50:50 aaarg, how much more space does the -dev image need ? Oct 27 11:51:25 Tofe: the image itself? Oct 27 11:51:44 all the fetch/compilation stuff... Oct 27 11:51:49 a lot Oct 27 11:51:58 I see that ^^ Oct 27 11:51:58 40GB should be enough Oct 27 11:52:03 but not much more then normal image Oct 27 11:52:20 ShiftyAxel: thanks for the four icons :) Oct 27 11:53:30 welcome :D Oct 27 11:53:43 for those who haven't seen them: https://twitter.com/shiftyaxel/status/262159657869660161 Oct 27 11:57:28 ShiftyAxel++ Oct 27 12:22:33 Stskeeps: I now have more information: it crash in libpthread, in pthread_mutex_lock, after a call made by libhybris to glibc Oct 27 12:23:01 Tofe: alright, is this ICS or JB, btw? Oct 27 12:23:04 ics Oct 27 12:23:06 ok Oct 27 12:23:17 and it has the patch and replaced bionic libc.so etc? Oct 27 12:23:32 yes, all of that Oct 27 12:23:44 ok Oct 27 12:24:14 so the TLS slots are supposed to being at index "5" in bionic, after the patch. Oct 27 12:24:34 (well, after the GL ones, that's to say.) Oct 27 12:24:58 yes Oct 27 12:25:02 maybe, simply, we should let more space for glibc TLS ? Oct 27 12:25:13 and begin, let's say, at index 30 ? Oct 27 12:25:17 do you have a __errno patch in the bionic? Oct 27 12:25:43 the ERRNO define has been changed in the way of your patch Oct 27 12:25:52 but maybe you're talking of another one Oct 27 12:26:13 __errno.c has not been changed on my side Oct 27 12:26:41 but it returns the TLS_SLOT_ERRNO index, so I guessed it was ok Oct 27 12:26:48 well, there's a more recent patch that also modifies __errno.c and libhybris Oct 27 12:26:54 but im unsure if that's integrated into OWO yet Oct 27 12:27:02 ah, I didn't know about this one. Oct 27 12:27:15 it is on github ? Oct 27 12:27:23 yes Oct 27 12:29:38 Oh ! I see, I'm missing some improvements :) I'll apply the new patch Oct 27 12:30:19 the "abort" are the unsupported ones ? Oct 27 12:30:54 yes Oct 27 12:30:58 well Oct 27 12:30:59 rather Oct 27 12:31:05 "debug" :) Oct 27 12:31:09 "if this gets called, it's seriously messed up Oct 27 12:31:38 I see you changed the slots, too Oct 27 12:31:50 that's an interesting change Oct 27 12:32:49 more consistent, I would say. I would even have reserved more than 2 slots, "just in case" Oct 27 12:33:05 ok, let's go Oct 27 12:53:03 little bit different this time: crashes in common/hooks.c:36 Oct 27 12:53:33 but it's nearly the same thing, mutex_lock Oct 27 12:56:44 Stskeeps: I must say I don't understand the change for TLS_SLOT_ERRNO, as it is quite contradictory with the previous version Oct 27 15:20:24 http://www.geeks.com/details.asp?invtid=FB359UAR-FB&utm_source=geekmail&utm_medium=daily_html&utm_content=FB359UAR-FB_Product&utm_campaign=NewArrivals_27oct12 Oct 27 16:05:46 bitbake is nerving me; it doesn't want to rebuild a package whose .bb file has been modified Oct 27 16:06:20 "bb -f", "bb -C fetch", removing stamps... what else ? would say Georges. Oct 27 16:06:31 OEBasic or OEBasicHash? what modification? Oct 27 16:06:36 Tofe: -c cleansstate Oct 27 16:08:47 JaMa: I modified a git SRCURI source Oct 27 16:08:58 no, not exactly: a SRCREV Oct 27 16:09:28 and is it used in PV? Oct 27 16:09:50 nope Oct 27 16:09:51 e.g. with SRCPV Oct 27 16:10:45 it's just a fixed PV from the name of the bb file Oct 27 16:13:22 ah, forcing "unpack" helped a bit Oct 27 16:15:16 oh, I was using the wrong libhybris repository... well well well Oct 27 16:16:57 Stskeeps: you do know that your makefile is a bit wrong concerning includes, when it come to ics ? Oct 27 16:22:22 where is that dkirker dude when you need him Oct 27 17:20:27 sigh, I'm getting "webos-ports-env/webos-ports/tmp-eglibc/sysroots/x86_64-linux/usr/bin/qmake-palm: No such file or directory" when configuring connman Oct 27 17:21:18 hmm, is a dep missing? because it hadn't built qmake-webos-native yet Oct 27 17:28:06 I did "bb qmake-webos-native" then re-started the build and it seems to have worked around the connman configure issue Oct 27 17:29:04 jacques: connman-qt? Oct 27 17:30:38 JaMa, let me double-check Oct 27 17:31:00 JaMa, yes, connman-qt - sorry Oct 27 17:31:11 adding it now Oct 27 17:31:47 thanks Oct 27 17:32:06 thanks for reporting it **** ENDING LOGGING AT Sun Oct 28 03:00:02 2012