**** BEGIN LOGGING AT Wed Jun 27 03:00:02 2018 Jun 27 04:37:50 @vanyasem, So does it mean that ubuntu touch now can be installed on nexus 5x Jun 27 04:50:25 No, this is about running Arch Linux on it. Jun 27 04:57:58 @leoaslan, afaik Ubuntu Touch uses Mir, not X11 Jun 27 06:54:04 How do people detect whether they have CAF or non-CAF? Jun 27 06:54:22 I wanna document PM rootfs and add a short explanation Jun 27 06:54:24 Simple question, is it nexus device? Jun 27 06:54:24 https://github.com/Halium/docs/pull/104#pullrequestreview-132290677 Jun 27 06:54:42 hey there :) Jun 27 06:54:48 if it's non-nexus Qualcomm device, then it's CAF device Jun 27 06:55:00 ok. thx Jun 27 07:38:14 @bhushanshah, Yeah, we should. Jun 27 07:40:32 Rebuilding/patching glibc is a PITA and must be done for each rootfs, so probably the patching GLES to stop using an static TLS slot would be the best option here. Jun 27 07:40:40 Even if that means a small performance impact. Jun 27 07:54:54 @slp1605: how often would the dynamic TLS slot be allocated ? once per thread creation ? at each TLS access ? Jun 27 07:55:12 @slp1605 Do you know how can one do that? or do you maybe have PoC patch in mind? Jun 27 07:55:43 in general Rebuilding glibc is nada for PlaMo usecase.. :/ Jun 27 07:55:53 I mean I can do that but *shrug8 Jun 27 07:56:05 [Edit] I mean I can do that but *shrug* Jun 27 07:56:42 @Tofe, Once, while linking the binary for execution. The problem is the couple of additional instructions needed for jumping through the PLT. Jun 27 07:57:03 @bhushanshah, Sure, I can give it a try a post a patch to the github issue. Jun 27 07:57:54 The ugly thing is that i'd require maintaining a branch/patch for GLES. Jun 27 07:57:59 *it'd Jun 27 07:58:23 you mean android_Frameworks-native? Jun 27 07:58:41 @slp1605: ok I see; seems like a good compromise, that can be optimized later on if we find a better solution Jun 27 07:58:55 probably, I'm not sure which repo holds it Jun 27 07:59:50 Tofe: yup, projects can decide patching glibc instead and avoid the branch/patch. Jun 27 08:00:38 @slp1605, https://github.com/Halium/android_frameworks_native/tree/halium-7.1/opengl/libs Jun 27 08:03:05 @bhushanshah, I see, thanks! Jun 27 08:04:19 Also, please checkout the libdsycall branch in both repos (bionic and frameworks_native) Jun 27 08:06:18 @slp1605: so if I understand correctly, dynamic TLS would mean using pthread_once/pthread_getspecific/pthread_setspecific ? Jun 27 08:06:57 (Like in Qt when disabling TLS: https://github.com/qt/qtbase/blob/5.11/src/corelib/thread/qthread_unix.cpp#L183 ) Jun 27 08:10:17 Tofe: that's an option, but in most arches using the attribute __thread for the variable is more efficient Jun 27 08:11:10 I've debugging this in ARM, and it's using the PLT Jun 27 08:11:19 *debugged Jun 27 08:13:26 ok, but then, I'm lost: isn't that what bionic and glibc are using today? Jun 27 08:18:26 Tofe: glibc does that, bionic defines a number of static TLS slots and happily assumes no one is going to use them. Jun 27 08:18:39 bionic-- Jun 27 08:26:06 @slp1605: ok Jun 27 08:54:55 @dohniks very nice to see PM in the docs :-) Jun 27 08:56:31 Indeed! If you know of anything to be added please report or pr Jun 27 11:56:52 slp1605: would your patch be alike the one for i386? https://github.com/ubports/android_frameworks_native/commit/50f14f28fbc607b2d352d4c274a9ca5201ce68c8 Jun 27 13:04:14 Tofe: FWIW, I tried the variation of that patch ... Didn't really help Jun 27 13:04:15 At least for halium 7 though Jun 27 13:08:42 ah, ok. But remember that (at least for me) the call stack points to __set_errno, not to __get_tls Jun 27 13:26:52 @vanyasem, 😍 Jun 27 17:03:38 @Tofe, I committed a patch based on it to Halium; it's slightly different Jun 27 17:20:00 @ilyaishere: do you have a link to the commit? Jun 27 17:22:51 @Tofe, Yeah: https://github.com/Halium/android_frameworks_native/pull/2/files … It turns out it hasn't been merged yet Jun 27 17:23:31 ok thanks, will have a look at it Jun 27 17:23:52 Maybe makes sense to adapt this change to work on arm as well, Tofe, can you give it shot and see if things gets fixed? Jun 27 17:25:48 bshah: yes, but it'll have to wait a bit, I'm working on other things currently and I don't have a halium-5.1 repo ready Jun 27 17:26:00 Okay Jun 27 17:26:28 Wait actually.. Jun 27 17:26:42 @vanyasem tested this right? Jun 27 17:27:01 define this Jun 27 17:27:18 @ilyaishere, . Jun 27 17:27:28 Variation of it Jun 27 17:27:35 yeah, the one you sent me in PM Jun 27 17:27:48 it hasn't fixed the issue with new Qt Jun 27 17:28:38 Maybe though, it changes things with my bionic and android_frameworks_native branch + your patch Jun 27 17:30:46 Ah no hm Jun 27 18:01:57 (Video, 26s) https://irc.ubports.com/wg9DAMJ4.mp4 Jun 27 18:02:36 check this out :P Jun 27 18:02:59 Tf is that music Jun 27 18:03:24 oh i forgot to mute the sound Jun 27 18:03:24 i apologize Jun 27 19:47:29 @vanyasem, the stuff i sent you...did it work ? Jun 27 19:47:35 @abhishek_01, i replied in the OT, no? Jun 27 19:47:40 do you want me to forward that here? Jun 27 19:48:36 @vanyasem, ah ... nvm .. i usually dont go through the chat logs of ot .. :P Jun 27 19:48:40 btw nice job Jun 27 19:49:03 well you replied to that message Jun 27 19:49:03 saying that you're glad you could help Jun 27 19:49:03 :P Jun 27 19:50:13 i need sleep .. lol Jun 27 19:50:15 good night Jun 27 19:50:31 good night my brother Jun 27 20:06:35 Wellington Alves was added by: Gilmars Jun 27 23:09:41 mateosalta was added by: mateosalta **** ENDING LOGGING AT Thu Jun 28 03:00:02 2018