**** BEGIN LOGGING AT Sun Oct 28 03:00:03 2012 Oct 28 10:24:21 morning Oct 28 10:26:38 Stskeeps: I've found a hint with gdb: it calls my_pthread_mutex_lock with __mutex=, but there is no init that is done before ==> the mutex isn't initialized. Oct 28 10:38:17 yeah, supposedly static initializers are not done correctly Oct 28 10:38:21 there's a patch waiting for that i think Oct 28 10:45:57 static ? you mean the ones that are done within bionic ? Oct 28 10:50:32 http://publib.boulder.ibm.com/infocenter/iseries/v5r4/index.jsp?topic=%2Fapis%2Fusers_61.htm , see the PTHREAD_MUTEX_INITIALIZER Oct 28 10:52:32 ah, yes, that's not nice Oct 28 10:53:28 and it's used in several places in bionic... Oct 28 11:52:28 Stskeeps: I don't think the problem is related to static init Oct 28 11:53:07 because it is already handled in the hook Oct 28 12:31:04 Stskeeps: maybe I was wrong: it tries to use a mutex statically initialized with PTHREAD_RECURSIVE_MUTEX_INITIALIZER Oct 28 13:13:19 Stskeeps: the current management of the mutexattr seems wrong: it should be wrapped too, because the mutex attributes values are not equal between bionic and eglibc Oct 28 13:15:29 :nod: Oct 28 13:18:19 I may have missed something, though, I'll double check, it's a bit complicated Oct 28 13:18:53 sure is, :) Oct 28 13:39:59 you're really lucky that the GNex opengl libraries don't use at all those special init of mutex... Oct 28 13:52:36 you iterate by findign out what breaks :) Oct 28 13:52:51 yes :) Oct 28 13:55:04 ok, I've passed over this problem, now, let's go to next segfault :) Oct 28 14:41:53 ok, now I think I'm stumbling against a badly initialized Tegra2 Oct 28 14:46:16 much better :) Oct 28 14:48:42 woot, system menu scaling is underway Oct 28 17:33:06 Stskeeps: yes and no, I have nearly no backtrace Oct 28 17:33:59 Tofe: what was the first issue you hit? some munmap thing? Oct 28 17:34:04 or was that somebody else Oct 28 17:36:44 the first issue was that some statically initialized mutex were used; but statically initialized with the "recursive" flag, not the normal one, which is ok with current libhybris Oct 28 17:37:06 i'm switching to ICS myself now, too, so :P Oct 28 17:37:46 I've forked you libhybris on my github, if you want to take back some modifications Oct 28 17:37:56 but it's not really heavily tested :) Oct 28 17:38:08 github://Tofee/libhybris Oct 28 17:38:34 thanks Oct 28 17:39:14 also you are missing a "bionic_tls.h" file in the ics subdirectory Oct 28 17:39:20 (I think) Oct 28 17:40:52 probably yeah Oct 28 17:44:29 with bitbake, what we find in the package/lib/.debug directly is the debug, or non-stripped, version of the binaries ? Oct 28 17:44:42 s/directly/directory/ Oct 28 17:44:42 Tofe meant: with bitbake, what we find in the package/lib/.debug directory is the debug, or non-stripped, version of the binaries ? Oct 28 18:00:05 Answer: *no* Oct 28 18:05:57 ah, irony... it crashes on an "strerror" call... Oct 28 18:17:20 good morning Oct 28 18:19:25 morning! Oct 28 18:23:14 oh wow Oct 28 18:23:34 valgrind is a lot more badass than i thought, just gave me exact function names on where things went south Oct 28 18:23:53 :) Oct 28 18:24:09 Stskeeps: we have a pb with strerror, I think. Oct 28 18:24:14 http://pastie.org/5128489 Oct 28 18:25:00 impressive, yes Oct 28 18:43:15 Stskeeps: ah, it now crash on a mutex_destroy with a mutex that is malloc'ed, but not by our init... Oct 28 18:44:30 ah, no, not necessarily malloc'ed. Oct 28 19:04:36 Ah ! it now goes further: an assert in main.c :) Oct 28 19:05:36 much better Oct 28 19:05:44 switch 32 to 16 or 24, as a start Oct 28 19:05:49 using test_egl first? Oct 28 19:05:55 I think it is 16 for tegra2 Oct 28 19:05:59 yes Oct 28 19:05:59 ok Oct 28 19:06:11 couldn't find an EGL surface Oct 28 19:06:35 :nod: Oct 28 19:14:27 okay, one step further :) Oct 28 19:14:52 got a surface? Oct 28 19:15:22 well... not yet Oct 28 19:15:32 chooseConfig failed Oct 28 19:16:06 progress Oct 28 19:16:21 that's weird, I was pretty sure I read that it was the surface that failed earlier Oct 28 19:17:13 maybe I was debugging test_glesv2. I'll have to re-check that. Oct 28 19:17:36 start with egl first, then gles :) Oct 28 19:26:07 note that I encountered a bug in the nvidia driver: it tried to destroy a non-initialized mutex... Oct 28 19:26:49 ok, eglChooseConfig fails both for 32bit and 16bit. I'll try 24bit. Oct 28 19:27:14 oh nice, now that i'm using ICS instead, my demos suddenly have no font problems Oct 28 19:27:17 and touch works correctly Oct 28 19:27:30 and i don't have to restart every time i do a fullscreen egl \o/ Oct 28 19:27:57 :) Oct 28 19:33:54 same with 24... There must be something wrong. Oct 28 19:34:13 I'll see that tomorrow Oct 28 19:34:21 goonight ! Oct 28 19:44:28 system menu is DONE Oct 28 19:44:41 stick a fork in it Oct 28 19:44:41 there shall be a tweet in due time Oct 28 19:44:59 the luna UI is pretty much finished now in regards to scaling... Oct 28 19:45:27 next up: fix bottom-of-screen notifications Oct 28 20:00:43 whats the hardware status on gnex? Oct 28 20:00:53 as in what's still missing Oct 28 20:04:31 I believe wifi is being worked Oct 28 20:04:48 bluetooth... unsure, and we don't have a user interface to it yet Oct 28 20:04:55 check the wiki? Oct 28 21:38:31 https://twitter.com/shiftyaxel/status/262669617766486017 Oct 28 21:57:33 ShiftyAxel, awesomenessivity Oct 28 22:05:21 :D Oct 28 22:05:34 now working on making the keyboard and bottom-of-screen notifications compatible Oct 28 22:05:37 no problem Oct 28 22:05:48 this is goanna be great when it's done Oct 28 22:52:15 * EricBlade builds the webkit WTF component and giggles Oct 28 22:55:15 libwtf? Oct 29 00:12:25 dwc-: hmm. if there's not a lib by that name there needs to be one. Oct 29 00:24:17 no kidding **** ENDING LOGGING AT Mon Oct 29 03:00:01 2012