**** BEGIN LOGGING AT Sun May 06 03:00:04 2018
May 06 14:30:42 <MonkeyofDoom>  1 file changed, 55 insertions(+), 2800 deletions(-)
May 06 14:30:55 <MonkeyofDoom> that's a nicer LED driver :)
May 06 14:31:08 <MonkeyofDoom> now to see if I can compile it
May 06 14:53:29 <Beiriannydd> I figured out why I couldn't chroot to the linux partition from Android: the chroot is aarch64 and android is armhf
May 06 15:02:20 <Beiriannydd> thas
May 06 15:02:54 <Beiriannydd> that isnt it.
May 06 15:19:10 <MonkeyofDoom> ../drivers/misc/mediatek/ccci_util/ccci_util_lib_load_img.c:900:46: error: ' Cipher chekc fail
May 06 15:19:14 <MonkeyofDoom> very nice, mtk
May 06 15:24:32 <NeKit> MonkeyofDoom, what do you use to compile it?
May 06 15:25:28 <NeKit> Beiriannydd, Android is aarch64/armhf multilib
May 06 15:27:25 <MonkeyofDoom> NeKit: copied config.gz from running debian, "make -j5 ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- O=build"
May 06 15:27:37 <Beiriannydd> NeKit: It has all the hallmarks of missing libraries: chroot: exec /bin/sh: no such file or directory
May 06 15:28:17 <NeKit> MonkeyofDoom, I mean, what compiler
May 06 15:28:30 <NeKit> Beiriannydd, did you try Linux Deploy? It should just work
May 06 15:28:52 <NeKit> you can configure it to use real partition
May 06 15:29:47 <MonkeyofDoom> gcc 7.3.0
May 06 15:29:49 <Beiriannydd> if I did that correctly, it starts, sets up mounts and then fails to start a terminal
May 06 15:30:01 <MonkeyofDoom> from the Arch repositories
May 06 15:31:17 <NeKit> Beiriannydd, I think it won't start terminal by default, but you can configure it to provide CLI binary and use su; linuxdeploy shell in terminal
May 06 15:33:03 <NeKit> MonkeyofDoom, the one used in Android is aarch64-linux-android-4.9. It will likely require changes to compile with newer gcc
May 06 15:36:56 <MonkeyofDoom> I'm making those presently
May 06 15:56:25 <Beiriannydd> NeKit: thanks so much. I just had to find the binary directory for linuxdeploy as it wasnt in the path
May 06 15:57:11 <NeKit> it has option to make it available in path in settings
May 06 16:04:40 <Beiriannydd> the mount partition was configured for /run/connman/resolv.conf so couldn't look anything up.  I pointed it at 1.1.1.1 manually - seems fine now
May 06 16:46:48 <MonkeyofDoom> ../drivers/input/keyboard/mediatek/mt6797/hal_kpd.c:21:10: fatal error: mt_soc_afe_control.h: No such file or directory
May 06 16:47:59 <MonkeyofDoom> and ../sound/soc/mediatek/mt_soc_audio_6797/mt_soc_pcm_routing.c:61:10: fatal error: auddrv_underflow_mach.h: No such file or directory
May 06 16:57:01 <MonkeyofDoom> ../fs/sdcardfs/main.c:365:34: error: expected ')' before 'SDCARDFS_VERSION'
May 06 16:57:19 <MonkeyofDoom> ???
May 06 17:00:31 <MonkeyofDoom> hm, seems to be defined in a the nearby Makefile
May 06 17:55:56 <MonkeyofDoom> wow it compiled
May 06 17:58:51 <NeKit> what output do you have?
May 06 17:59:23 <NeKit> in theory you need Image.gz-dtb and pack that into boot.img with initramfs extracted from Debian boot.img
May 06 17:59:30 <MonkeyofDoom> a 200MB vmlinux :|
May 06 18:00:03 <NeKit> did you make vmlinux?
May 06 18:02:48 <MonkeyofDoom> no, just `make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu`
May 06 18:08:20 <MonkeyofDoom> oh, "Image.gz-dtb" is the target
May 06 18:29:26 <MonkeyofDoom> NeKit: how do I extract the initramfs from the debian boot.img?
May 06 18:29:45 <NeKit> https://github.com/vianney/bootimgtool
May 06 18:30:08 <MonkeyofDoom> ah, thanks
May 06 19:05:48 <adam_b> as an experiment I just flashed a different kernel to the recovery partition and then booted with the esc key as would normally give me a dead android, it booted debian linux as hoped and gave me the expected uname -a, a reboot to android of course put it back to a dead android so its only useful for short term tests before deciding the new one is safe
May 06 19:06:15 <MonkeyofDoom> nice!
May 06 19:06:35 <MonkeyofDoom> "flashed a different kernel to the recovery partition" with the flashtool?
May 06 19:06:44 <adam_b> note added to DebianTP wiki page as to how I did that
May 06 19:06:51 * MonkeyofDoom looks
May 06 19:06:54 <adam_b> just a sudo dd
May 06 19:07:07 <MonkeyofDoom> word
May 06 19:11:02 <MonkeyofDoom> sigh, my kernel does not seem to boot
May 06 19:11:58 <adam_b> wonder if this means that if you pop the android kernel on the 'boot' and debian kernel on 'linux_boot' you'll effectively swap the default key combinations?
May 06 19:12:09 <MonkeyofDoom> I wouldn't be surprised
May 06 19:12:14 <adam_b> damm was inteneding to swap the names on those
May 06 19:12:28 <MonkeyofDoom> mm
May 06 19:12:54 <adam_b> still you clearly understood my meaning
May 06 19:13:11 <adam_b> so how do you debug a failing kernel build?
May 06 19:13:47 <MonkeyofDoom> ideally, by watching debubbing output via a serial TTY during early boot
May 06 19:14:04 <adam_b> you tried plugging in a usb cable to see if stuff comes out?
May 06 19:15:05 <MonkeyofDoom> I'm going to try building with the aarch64-linux-android-4.9 toolchain
May 06 19:15:55 <adam_b> is that the know to work way NeKit does it?
May 06 19:16:18 <NeKit> <adam_b> wonder if this means that if you pop the android kernel on the 'boot' and debian kernel on 'linux_boot' you'll effectively swap the default key combinations? - yes
May 06 19:16:37 <NeKit> MonkeyofDoom, does it reboot or freeze?
May 06 19:17:26 <MonkeyofDoom> reboots
May 06 19:17:32 <NeKit> adam_b, but don't flash Linux kernel to recovery partition, lk doesn't init modem when booting recovery
May 06 19:17:56 <MonkeyofDoom> ah, that's good to know
May 06 19:18:03 <adam_b> NeKit: interesting, it 'worked' in that i was able to login and check kernel version
May 06 19:18:06 <NeKit> if it reboots and you can force it into recovery, you can try checking /proc/last_kmsg
May 06 19:18:08 <adam_b> didnt try more than that
May 06 19:20:08 <adam_b> we need to map the fourth key combination to a new boot partition name, as per the Bootloader wiki page
May 06 19:21:28 <NeKit> Esc+side key?
May 06 19:21:40 <adam_b> yep
May 06 19:22:20 <adam_b> unless there was a secret third version of the bootloader that has any side button activity launching debian
May 06 19:23:34 <adam_b> in which case it would just cause even more confusion for newbies
May 06 19:24:21 <NeKit> I'm confused by those key combos myself
May 06 19:24:42 <adam_b> you've presumably the original pre 6th march bootloader?
May 06 19:24:48 <NeKit> yes
May 06 19:25:10 <adam_b> so of the four combinations which are duplicates for you?
May 06 19:25:41 <adam_b> or could you just add another key table to that wiki page with what they are pre 6th march: https://github.com/gemian/gemini-keyboard-apps/wiki/Bootloader
May 06 19:25:46 <NeKit> I think it has only Esc for Linux and side for recovery (3 combos)
May 06 19:26:04 <adam_b> but with two buttons there are four possible options to press them
May 06 19:26:19 <NeKit> yeah
May 06 19:26:28 <NeKit> actually, lk has functions for text output, adding something like "booting Android/Linux/recovery" would be nice in case of 4 combos
May 06 19:27:33 <NeKit> (not very nice text output though by default)
May 06 19:27:36 <adam_b> so if you press esc and side which wins out of linux or recovery
May 06 19:28:38 <NeKit> recovery
May 06 19:29:35 <adam_b> the splash screens try to tell you what to expect, you get the linux wording appear for a split second before the animated android startup, or it stays there till x takes over for linux, or you don't even see it for recovery as its replaced by the dead android so quickly
May 06 19:32:20 <adam_b> I've updated that wiki page, if anyone who's downloaded more recently gets another combination let us know
May 06 19:34:34 <AuroraAvenue> So where can I get a gemini - umpc ?
May 06 19:37:01 <canny[m]1> AuroraAvenue: On the Indiegogo page: https://www.indiegogo.com/projects/gemini-pda-android-linux-keyboard-mobile-device--2
May 06 19:41:30 <MonkeyofDoom> ooh, I think it booted with my patched kernel (android-4.9 toolchain)
May 06 19:43:36 <MonkeyofDoom> yep :D
May 06 19:43:46 <adam_b> LED's a blinking?
May 06 19:44:07 <MonkeyofDoom> well, the first change I made was to not light the blue near caps-lock in driver init
May 06 19:44:26 <MonkeyofDoom> and did a rewrite of the handling of the aw9120_operation procfs entry handling
May 06 19:44:36 <MonkeyofDoom> which I now need to fix because it's buggy
May 06 19:44:45 <adam_b> I think the idea of doing that was to tell you that the thing had booted successfully?
May 06 19:45:05 <adam_b> I'd be inclined to keep that but have a service that turns it off at the end of the boot
May 06 19:45:31 <MonkeyofDoom> it could program it with an animation that fades it out
May 06 19:45:43 <adam_b> sounds even better
May 06 19:46:12 <NeKit> it's really buggy if building with newer gcc breaks the boot
May 06 19:46:23 <MonkeyofDoom> the mediatek code is fucking horrifying
May 06 19:46:28 <MonkeyofDoom> I'm not surprised new gcc kills it
May 06 19:46:30 <adam_b> we could have different colours for each varient, so android can keep the blue
May 06 19:47:18 <adam_b> we could have red or purple for deb/sf etc
May 06 19:47:22 <MonkeyofDoom> NeKit: do you think it's possible to compile some files with old gcc and some with new, then link together and get a booting kernel?
May 06 19:47:33 <MonkeyofDoom> to figure out which files are being miscompiled and need fixing
May 06 19:48:00 <NeKit> I doubt, maybe getting info why it crashes is easier in case of kernel panic
May 06 19:50:02 <NeKit> could you upload the changes you did for it to build with 7.3? I will probably try myself inside postmarketOS
May 06 19:50:12 <MonkeyofDoom> sure, one sec
May 06 19:50:37 <MonkeyofDoom> the 4.9 also builds/boots with those changes (except one where I added a #pragma to disable a warning and 4.9 doesn't even have said warning)
May 06 19:52:32 <NeKit> it's just that their toolchain is known to produce working kernels
May 06 19:57:57 <MonkeyofDoom> NeKit: oof, may be a minute, I need to properly stage all these commits
May 06 19:58:11 <NeKit> no hurry
May 06 19:58:29 <MonkeyofDoom> is there a way to use grep to pick which hunks of `git add -p` to stage?
May 06 20:00:57 <NeKit> doubt it
May 06 20:01:04 <MonkeyofDoom> heh
May 06 20:14:09 <Beiriannydd> NeKit: thanks for your help, now I have easy access into my linux.  Now I only wish that the kernel included the necessary cgroups to be able to use docker but I can live without it for now
May 06 20:14:38 <NeKit> you can build custom Android kernel if you want
May 06 20:14:44 <Beiriannydd> oooh
May 06 20:15:11 <NeKit> but it's a bit troublesome for now as you see :)
May 06 20:15:49 <Beiriannydd> heh I see
May 06 20:16:33 <Beiriannydd> they also said that they had a 4.x kernel working, is that ready for Oreo?
May 06 20:19:22 <Beiriannydd> I guess the best way to dev for the kernel would be vm on PC?
May 06 20:21:55 <Beiriannydd> I have linux experience, just zero on Android and only arm on SBCs
May 06 20:23:02 <NeKit> the best way is likely Linux on PC, but VM should work as well
May 06 20:24:53 <Beiriannydd> Is there a good howto that you would recommend to get started?
May 06 20:42:37 <MonkeyofDoom> NeKit: https://github.com/mdgem/kernel-3.18-geminipda
May 06 20:42:50 <MonkeyofDoom> led changes are not uploaded yet, I need to debug a little
May 06 20:43:35 <MonkeyofDoom> that tree should build with android-4.9 and upstream gcc 7.3, but only the former boots for me
May 06 20:44:04 <MonkeyofDoom> (using .config from `zcat config.gz` on the debian kernel)
May 06 23:17:05 <MonkeyofDoom> ok, have 8 brightness levels on all leds working with much cleaner code now
May 06 23:17:17 <MonkeyofDoom> (for aw9120)
May 06 23:17:33 <MonkeyofDoom> kernel interface isn't changed much yet though
May 06 23:18:35 <MonkeyofDoom> a better kernel interface will expose 256 brightness values over pwm as well as the 8 (electrical) current values
May 06 23:18:45 <MonkeyofDoom> in sysfs
May 06 23:19:04 <MonkeyofDoom> and hopefully with LED trigger support as well so they can be tied to I/O or such
May 06 23:35:19 <MonkeyofDoom> pushed
May 06 23:40:19 <Beiriannydd> would a diff between standard mtk 3.18 and gemini be useful?
May 06 23:41:48 <MonkeyofDoom> probably so
May 06 23:44:53 <Beiriannydd> the android kernel is kind of big
May 06 23:45:01 <Beiriannydd> the repo I mean
May 06 23:46:15 <Beiriannydd> in order to sanely do this, I should just do the branches we are interested in
**** ENDING LOGGING AT Mon May 07 03:00:00 2018