Messages

31

Gemini PDA - Linux / left/middle/right mouse clicks and F1-F12 and Insert keystrokes

« on: July 17, 2018, 10:59:22 am »

Problem is getting F11 and F12...

I read somewhere that the lid switches are on F11 & F12 which may or may not be true, not sure how you'd go about testing it though.

That isn't the case any more https://github.com/gemian/gemini-linux-kern...47c8967963d6b0f

32

Gemini PDA - Linux / PL2303 serial adapter drivers?

« on: July 17, 2018, 08:40:49 am »

Anyways I am still hoping for a 4.X kernel, which hopefully fixes the sd card performance issue...

There's an sd card performance issue?

I've been using F2FS on mine, but it's been perfect

33

Gemini PDA - Linux / System load

« on: July 13, 2018, 05:54:13 am »

Kernel sees whatever the mediatek driver does to turn off cores as them being under full load, so subtract roughly 10 to get the true value

34

Gemini PDA - Linux / poweroff, reboot, pm-suspend, and pm-hibernate in DebianTP2

« on: July 11, 2018, 03:48:27 pm »

You need to use repowerd for this, none of the other linux utilities for power management work very well/at all

Wish this wasn't the case, it doesn't suit PDAs/laptops, seems to be more designed for phones. I want the gemini to be 100% awake when open, even when idle, and 100% asleep when closed, and it's a real ballache to make that happen with repowerd. It's a fairly complex state machine that fights against a binary lid=on/off system in many ways

Very interested if someone manages to get locking and suspend working without repowerd, as the screen brightness without repowerd problem is already solved

35

Gemini PDA - Linux / Incredible amounts of kernel logs

« on: July 11, 2018, 08:45:06 am »

Well, to clarify what I was after, I wanted to be able to use dmesg -w without getting drowned in stuff, and journalctl -f without getting inundated.
On a "normal" debian system I have never had these amounts of logs. Also I was thinking that with the amount of logging there must be a little too much disk access being generated, which would reduce battery life...

My above suggestion hides kernel logs from journalctl, but doesn't help with dmesg. In my kernel fork you can find a couple of commits that hack out the worst offenders https://github.com/lukefor/gemini-linux-kernel-3.18

You may also want to disable logd, which is the worst for CPU usage - see https://www.oesf.org/forum/index.php?showtopic=35245

36

Gemini PDA - Linux / Annoying dynamic contrast effect

« on: July 10, 2018, 01:15:19 pm »

Hacked the above fix into repowerd to allow brightness control to still work, anyone interested can find it in this commit: https://github.com/lukefor/repowerd/commit/...c4b82487db3e750

37

Gemini PDA - Linux / Annoying dynamic contrast effect

« on: July 08, 2018, 07:42:39 am »

So it turns out disabling AAL breaks sleep (i.e. screen does not wake up when opening lid, everything else does)

Fortunately there is a simple fix - instead of disabling it in the kernel, add a cron to pause aal - it seems this usermode binary does all the heavy lifting. It autorestarts if killed, so pausing is the only option

Code: [Select]

killall -STOP /vendor/bin/aal

38

Gemini PDA - Linux / Incredible amounts of kernel logs

« on: July 08, 2018, 07:14:28 am »

You should be able to

Code: [Select]

cat /proc/sys/kernel/printk w x y z
where you substitute the letters for the desired numbers below

0 - emergency              
1 - alert          
2 - critical  
3 - error  
4 - warning
5 - notice
6 - informational
7 - debug

Or you can put quiet in your kernel command line (or loglevel=X)

Or you can

Code: [Select]

sudo dmesg -n 1 or dmesg -D  (off) and dmesg -E (on) to toggle console messages  on/off

Or you create / edit /etc/sysctl.conf and add

Code: [Select]

kernel.printk = w x y z substituting the numbers from above.

You can install systemd from stretch-backports, and then set ReadKMsg=no (I think that setting may already be in place by default)

I wouldn't install systemd just to get functionality that already exists several times over. But if you do then be aware that recent versions of systemd no longer read /etc/sysctl.conf and look in /etc/sysctl.d/*.conf

All of those only affect what is printed to TTYs, which in this case would be the serial console

It's not clear whether the OP is referring to that, to dmesg, or to the systemd journal

39

Gemini PDA - Linux / Incredible amounts of kernel logs

« on: July 07, 2018, 06:16:30 pm »

You can install systemd from stretch-backports, and then set ReadKMsg=no (I think that setting may already be in place by default)

40

Gemini PDA - Linux / Annoying dynamic contrast effect

« on: July 07, 2018, 01:45:30 pm »

Solved!

Simply set this in the kernel config:

Code: [Select]

CONFIG_MTK_AAL_SUPPORT=n
Nice increase in image quality with this change (things look closer to how they should), god only knows what mediatek were thinking with this

41

Gemini PDA - Hardware / MediaTek USB-UART on Gemini-PDA

« on: July 04, 2018, 06:31:39 pm »

I've just built one of these without cutting wires etc by buying a few things from eBay:

https://rover.ebay.com/rover/0/0/0?mpre=htt...%2F282074575442
https://rover.ebay.com/rover/0/0/0?mpre=htt...%2F262202365883
https://rover.ebay.com/rover/0/0/0?mpre=htt...%2F162036110446

Pics:

https://imgkk.com/i/nj6z.jpg
https://imgkk.com/i/48qa.jpg
https://imgkk.com/i/370p.jpg
https://imgkk.com/i/spc6.jpg

Really impressed with the level of output, should really help me solve my shutdown freeze and make kernel changes non blindly

42

Gemini PDA - Linux / Upgrading to TP2 from TP1 without losing data

« on: July 04, 2018, 03:20:17 pm »

Sorry, I'm not sure I understand you.

I was asking about upgrading from https://github.com/gemian/gemini-keyboard-apps/wiki/DebianTP to current Gemian (DebianTP2 probably?).

As for the GPU acceleration, the Mali GPU is/was not used in DebianTP(1), it is apparent as the desktop environment is not as fast as it should be.

GPU acceleration is present in TP1 - apps that use ES 2 work perfectly. The lag you are experiencing is due to a significant bottleneck in the fork of glamor used - you can disable glamor and get a much more responsive DE (e.g. scrolling goes from <1fps to >30fps), but at the expense of all GPU acceleration and the login screen

43

Gemini PDA - Linux / Annoying dynamic contrast effect

« on: June 30, 2018, 11:09:01 am »

There seems to be some sort of dynamic contrast baked into the kernel/display, similar to what Intel GPUs do by default on laptops

See here: https://www.youtube.com/watch?v=cys7tbrVlNw (it's not the camera)

It's particularly noticeable in the shade of black/grey on the LXQT menu button and clock when focusing and unfocusing a terminal window

Unless I'm mistaken this is Miravision, which seems to live in /drivers/misc/mediatek/video/common/color20, but I've had no luck disabling it via the obvious config flags (CONFIG_MTK_PQ_COLOR_MODE etc)

Has anyone else noticed this/tried to disable it?

44

Gemini PDA - Linux / 2% power drain per hour on a sleeping device is not acceptable

« on: June 30, 2018, 07:15:51 am »

So something I changed has fixed it! Just left the gemini overnight with a load of applications running and it only dropped 1% and didn't freeze!

I suspect it is the logind dbus removal that did it, but that isn't a totally harmless change as I now have to restart repowerd on login. Needs more investigation...

I also removed power button/esc key handling from gemian-lock - this seems to be one of the reasons that can freeze the gemini in suspend similar to xscreensaver

45

Gemini PDA - Linux / 2% power drain per hour on a sleeping device is not acceptable

« on: June 29, 2018, 06:16:42 pm »

I think I've just had a successful suspend by disabling part of the LogindSessionTracker and also adding a script to disable wifi via networkmanager before sleeping (my hacks are here), at least in terms of not losing any battery % at all over 30 mins and no increase in process CPU times from when I closed the lid

You can identify a proper suspend by some of the log info printed by the kernel in dmesg (though unmodified kernel is hilariously verbose and it's difficult to pick apart), or also by the fact that the device will be completely cool to the touch (burning 2% an hour produces faint heat)