Wednesday, 2022-12-14

T42<adampigg> @slavamon ping :)09:30
v01andhi all18:44
v01andi want to understand - how libhybris is working18:45
v01andi build aosp 10 for Pixel4 and deploy linux sysroot to phone18:46
v01andnow i try run test_egl from hybris project, but it cannot resolve some functions from bionic18:48
v01anddoes anyone have the same experience?18:49
T42<NotKit> v01and: hi. not if this is correct group, but at minimum you need some patches to AOSP libs so you can dlopen Android libs with libhybris18:51
T42<NotKit> see 0003, 0006 at and 0003 at
T42<elros34> I am not sure test_egl works for anybody so better use test_egl_configs18:55
v01andI tried with these patches too.18:55
v01andthe same result and test_egl_configs18:55
v01andthere is a lot of18:56
v01andHOOKS hooks.c:3518 (__hybris_get_hooked_symbol) DEBUG: Could not find a hook for symbol eglLockSurfaceKHR HOOKS hooks.c:3518 (__hybris_get_hooked_symbol) DEBUG: Could not find a hook for symbol eglUnlockSurfaceKHR HOOKS hooks.c:3518 (__hybris_get_hooked_symbol) DEBUG: Could not find a hook for symbol eglCreateImageKHR HOOKS hooks.c:3518 (__hybris_get_hooked_symbol) DEBUG: Could not find a hook for symbol eglDestroyImag18:56
T42<NotKit> that is just debug message and not an actual issue18:58
T42<NotKit> start with test_dlopen, then test_egl_configs as Elros suggested18:59
v01andtest_dlopen is ok, it show base adress without crash18:59
T42<Ayman_rgab> in this manner what could cause test_egl_configs return error like this :
T42<NotKit> try to strace test_egl_configs to see if it is missing access to some paths from sysroot19:02
v01andEGL_PLATFORM=hwcomposer test_egl_configs Starting test (EGL_PLATFORM=hwcomposer) Segmentation fault19:06
v01andwithout hwcomposer - the same19:07
v01anddo I need some kernel patches?...19:16
T42<elros34> strace it, segementation fault itself doesn't tell much19:18
v01andso, ltrace show this
v01andand the strace log
v01andsorry there is a lot of my debug messages19:35
T42<NotKit> no idea why it crashes from the log, seems to be somewhere at linking steps19:38
T42<NotKit> you could try to mount Android system from inside Linux root to compare against19:39
v01andnope, nothing changes20:17
v01andwhat a librhybris configure flags do you use?...20:19
v01andi got some backtrace from gdb
v01andcrash always in soinfo::relocate<plain_reloc_iterator>...20:33
v01anddo i have a wrong c++ compiler or flags?... or it cause wrong ENV....20:35
piggzmal: @slavamon: spot the difference
T42<b100dian> pick me! pick me! no disconnectreason?21:33
T42<b100dian> or is it ngfd ;)21:34
piggzyeah, no disconnct signal21:35
piggznow ...why21:35
piggzcauses a 10 second delay on pinephone before UI closes call21:36
piggzthink i found the problem!21:45
T42<b100dian> I need to restart some debugging in this area too, I have some calls (once in a blue moon) where vibration doesn't start, is this just ofono logging what you're showing?21:46
T42<b100dian> s/start/stop21:46
T42<b100dian> This is great news, can't wait for your insight(s)!21:46
piggzthis is pinephone specific,
piggzmal: ^^21:59
v01andif some still have an interest in my problem23:01
v01andi just rollback hybris from master to 33f1c76ce515879d0c04e5f601806b228e89d829, and disable debug options23:01
v01andlooks like now it works well23:02
v01andthanks for your help23:15
malthe changes after that commit are quite small23:18
malare you sure the issue wasn't actually the debug options?23:19
malwhich android base are you using?23:19
malah, aosp1023:20
v01andmaybe it was cause of  "--enable-tracing" options, i still have "--enable-debug"23:22

Generated by 2.17.1 by Marius Gedminas - find it at!