**** BEGIN LOGGING AT Wed Jan 11 03:00:02 2017 Jan 11 06:15:21 DougReeder: Sorry it's been hectic past 2 days at the new job. Been thrown in at the deep end right away. Let me have a look in the next hr or so Jan 11 17:58:56 Morning Jan 11 17:59:16 I'm back from a little trip, I'll try to catch up Jan 11 18:02:07 Tofe: Morning & welcome back! Jan 11 18:02:08 granpc: hi Jan 11 18:02:13 granpc: we recently did quite some changes regarding how we use the android drivers, I don't know if you noticed Jan 11 18:02:15 hello Jan 11 18:02:20 no I honestly have no idea Jan 11 18:02:33 I was working on ubuntu touch but the lack of functional docs got on my nerves and I decided to try luneos Jan 11 18:02:33 Herrie: thanks :) Jan 11 18:02:59 (plus it seems a lot easier to upload debug symbols to the target on luneos, which is nice!) Jan 11 18:03:17 but yeah I'm basically new to luneos. been watching updates for a while but never poked at the code Jan 11 18:03:31 granpc: well, the idea is that we now use a very minimal build for everything that needs bionic, and which can be hooked by libhybris Jan 11 18:03:55 yeah, an android build without java userspace stuff I guess Jan 11 18:03:57 and we use that as an "overlay" over the original android stsrem Jan 11 18:04:12 system* Jan 11 18:04:12 oh right, I noticed that Jan 11 18:04:28 I sorta disabled it for now, I figured if I just built the whole android system with the bionic patches it'd have the same result Jan 11 18:04:30 granpc: it's even more minimal than that, but that's the idea Jan 11 18:04:48 yes, it's just more difficult :) Jan 11 18:05:26 so the bizarre thing is that I did just that Jan 11 18:05:39 and it still crashes in glibc functions (say, strlower) that use locale stuff Jan 11 18:05:59 so I thought maybe my patches weren't being applied properly/android wasn't rebuilding/whatever, but I disassembled my own libc and it does seem to be using the new TLS constants Jan 11 18:06:00 looks like missing libhybris hooks, somehow Jan 11 18:06:21 what hooks could be missing to cause that? Jan 11 18:06:24 I'll trace it and figure out what's going wrong Jan 11 18:06:47 one moment Jan 11 18:07:38 the list is in hooks.c, iirc Jan 11 18:07:52 yeah Jan 11 18:07:57 I do know at least some of those hooks are working Jan 11 18:08:05 because I had to add one to kill the mali-hist-dump thread Jan 11 18:08:12 (just mali things, I guess.) Jan 11 18:08:16 but it doesn't make sense that only a part of them works and not the rest Jan 11 18:08:35 yeah I'm only thinking that maybe it's trying to hook something that is built as an internal symbol here or something Jan 11 18:08:56 otherwise it doesn't make much sense Jan 11 18:09:12 could be. errno is properly rerouoted ? Jan 11 18:09:35 hmm, I have actually noticed some oddities with errno Jan 11 18:10:00 that one is a bit tricky Jan 11 18:10:11 yeah it doesn't work on arm64 at all Jan 11 18:10:27 I read in a few places that you could take that one out and it'd work just fine Jan 11 18:10:32 but I did try a 32 bit build with that patch in Jan 11 18:11:08 what I find surprising is that the same build works well with ubuntu touch, if I read properly Jan 11 18:11:28 it works well, for some values of well Jan 11 18:11:38 ok :) Jan 11 18:11:39 it didn't render or anything (not even test_hwcomposer) Jan 11 18:11:43 but test_sensors worked fine without crashing Jan 11 18:11:47 whereas here it crashes with glibc stuff Jan 11 18:11:51 so I guess both builds have different issues Jan 11 18:11:57 oh, ok Jan 11 18:12:05 I'd rather fix the luneos one for now Jan 11 18:12:17 seems easier to troubleshoot and I sorta miss webos Jan 11 18:12:56 but I have no idea what could be causing those crashes other than missing bionic patches (which I'm pretty sure can't be the case) Jan 11 18:13:05 It could be a header mismatch Jan 11 18:13:38 the android-headers thing? Jan 11 18:13:44 yes Jan 11 18:13:58 you know, I actually have no idea where I pulled my headers from Jan 11 18:14:06 so I'll try that tonight. thank you! Jan 11 18:14:12 I recently reworked it to make it use the ubuntu headers, which are more up to date Jan 11 18:14:33 I guess I could just copy them from the ubuntu build? Jan 11 18:16:53 well, I guess the ubuntu build use their ubuntu-headers repo, so actually it could as well be as simple to just update your luneos sources to match the latest testing and migrate the header thingy Jan 11 18:17:20 how recent was this? Jan 11 18:17:27 I think I pulled luneos like 4-5 days ago Jan 11 18:19:13 it's... around that time :P Jan 11 18:19:25 I migrated maguro and mako devices (google nexus and nexus 4) to use Android 5.1.1 drivers Jan 11 18:19:42 And made it more easy to do the same for future devices Jan 11 18:19:49 hmm. I copied basically everything from the mako tree Jan 11 18:19:52 and that was using cm-12.1 Jan 11 18:20:05 ah great Jan 11 18:20:25 ok then you're already using ubuntu headers API 22 Jan 11 18:20:35 yeah Jan 11 18:21:29 so, it still could be a mismatch :p if the manufacturer had the great idea to patch the headers a bit, for their own driver needs Jan 11 18:22:11 I don't think it happens often, but I know the embedded mobile world if full of little hacks, so why not Jan 11 18:22:25 no yeah, it probably happened Jan 11 18:22:28 this is mediatek we're talking about.. Jan 11 18:22:38 I'll take a look at that, thank you! Jan 11 18:23:26 here it would mean comparing the ubuntu headers API 22 with the one you pulled for building CM12.1 Jan 11 18:23:26 good luck :) Jan 11 18:23:40 thanks, I will keep you updated Jan 11 18:23:51 ok! Jan 11 18:34:52 Tofe: Can you try http://build.webos-ports.org/luneos-testing/images/maguro/luneos-dev-package-maguro.zip on your Maguro? Jan 11 18:34:58 It was bootlooping on my Maguro I think Jan 11 18:35:03 With CM 12.1 build on it Jan 11 18:35:27 I was also testing various Google images for Mako on 5.1 and 5.1.1 none worked properly except for the CM 12.1 one Jan 11 18:35:35 I'll try it, I'm still in the train Jan 11 18:36:22 That's strange; in addition I'm pretty sure the drivers are almost the same anyway Jan 11 18:37:40 Tofe: Yeah I just did some quick tests I didn't have much time to look into the specifics to be honest Jan 11 18:40:38 DougReeder: Had a look into your question, but I'm not really sure what you mean? I don't see the "new thread search field"? Jan 11 18:41:00 Herrie: I would prefer that our build works with google images, if any Jan 11 19:19:56 Tofe: Yeah me too.... Jan 11 19:20:07 Especially seeing CM's status currently Jan 11 19:56:11 granpc: I came across this one: https://github.com/mer-hybris/android/pull/27/files Jan 11 19:56:26 You migth need that as well if you don't have it already Jan 12 00:31:52 Herrie: thanks. that seems to be in there already Jan 12 00:32:14 I suppose the arm64 parts wouldn't have built at all without it. still a good find though, thank you! Jan 12 00:34:40 I diffed the android headers I'm using against the ones I extracted with extract-headers.sh, these are the resutls: http://paste2.org/2HHtz0Cn Jan 12 00:34:52 so probably not a header thing /cc Tofe Jan 12 00:35:58 it's sorta weird that the wifi headers aren't in my extracted headers but I think they were just in a different path Jan 12 00:36:06 but other than that they're the same so probably not incompatible **** ENDING LOGGING AT Thu Jan 12 03:00:01 2017