reMarkable Paper Pro support

[NoaHimesaka1873]

What works:

• Everything except restarting KOReader

What's janky:

• Shutdown: rmpp-qtfb-shim and qtfb dies before it can display shutdown screen, resulting in whatever was displayed before staying
• Refresh: controlled by xochitl, (reMarkable's proprietary UI/UX and controls screen) will change in future after qtfb and shim gets ability to control it.

What doesn't work:

• restart KOReader: it doesn't even try to lol

Dependencies required to run:

• XOVI
• QTFB
• rmpp-qtfb-shim

You need to set a few environmental variables when launching, like this:
LD_PRELOAD=(path to your qtfb-shim.so) QTFB_SHIM_INPUT=false QTFB_SHIM_MODEL=false QTFB_SHIM_MODE=RGB565 KO_DONT_GRAB_INPUT=1 ./reader.lua

Closes #12856

---

This change is 

[Frenzie]

lgtm, just a couple of nits and minor questions

Updating l10n/ is fine but I suspect done by mistake.

[NoaHimesaka1873]

Yep, l10n/ update is a mistake which I don't know how to remove lol

[pazos]

Frontlight: documentation is janky so couldn't figure out how to implement

I'm not sure what documentation :p .

It should be pretty straightforward (assuming only brightness, no warmth). Have a look at https://github.com/koreader/koreader/blob/master/frontend/device/sysfs_light.lua

Then on your device def you tell if has light and has warmth and include frontlight_settings override (mostly sysfs nodes, but also min and max if they're different). For instance: https://github.com/koreader/koreader/blob/master/frontend/device/kobo/device.lua#L323-L329

[NoaHimesaka1873]

@pazos One problem:

1. rM uses 4 as off, 0 as on.
2. brightness range is 0-2047 instead of 0-100 which that code assumes. (that range check seem like debug only?)

[NoaHimesaka1873]

I can't really figure out how it works even after reading all the devices' examples, e.g. why the hell other devices call kobolight? What functions are actually necessary to implement?

It's just simple frontlight system, but the unnecessarily complex PowerD implementation is making me confused. It just need to do these things (base path is /sys/class/backlight/rm_frontlight/):

• To turn frontlight on, write 0 to bl_power
• To turn frontlight off, write 4 to bl_power
• To read intensity, read brightness
• To write intensity, write to brightness (0-2047)

I can't really figure this out, so I need help from someone else.

[pazos]

No problem. Lights don't need to be part of this PR.

why the hell other devices call kobolight?

currently there's no device that:

1. has brightness only, no warmth
2. updates the brightness by writting to sysfs nodes

kobolight uses a ioctl system call instead of writting to sysfs. It turns out it is widely used outside kobo too.

What functions are actually necessary to implement?

We need to handle all these bits in device specific powerd implementation. If you want to read one that doesn't hurt your brain you can look at https://github.com/koreader/koreader/blob/master/frontend/device/cervantes/powerd.lua

If you look at this chunk:

koreader/frontend/device/cervantes/powerd.lua

Lines 43 to 80 in a2e0642

	function CervantesPowerD:init()
	-- Default values in case self:_syncLightOnStart() does not find
	-- any previously saved setting (and for unit tests where it will
	-- not be called)
	self.hw_intensity = 20
	self.initial_is_fl_on = true
	if self.device:hasFrontlight() then
	if self.device:hasNaturalLight() then
	local nl_config = G_reader_settings:readSetting("natural_light_config")
	if nl_config then
	for key, val in pairs(nl_config) do
	self.device.frontlight_settings[key] = val
	end
	end
	-- Does this device's NaturalLight use a custom scale?
	self.fl_warmth_min = self.device.frontlight_settings.nl_min or self.fl_warmth_min
	self.fl_warmth_max = self.device.frontlight_settings.nl_max or self.fl_warmth_max
	-- If this device has a mixer, we can use the ioctl for brightness control, as it's much lower latency.
	if self.device:hasNaturalLightMixer() then
	local kobolight = require("ffi/kobolight")
	local ok, light = pcall(kobolight.open)
	if ok then
	self.device.frontlight_settings.frontlight_ioctl = light
	end
	end
	self.fl = SysfsLight:new(self.device.frontlight_settings)
	self:_syncLightOnStart()
	else
	local kobolight = require("ffi/kobolight")
	local ok, light = pcall(kobolight.open)
	if ok then
	self.fl = light
	self:_syncLightOnStart()
	end
	end
	end
	end

You'll see that it uses kobolight for brightness on both branches (hasNaturalLight or not!) but that doesn't mean the remarkable needs to do the same.

It should be enough to fill the same powerd functions as cervantes has, that take extra logic when the device has frontlight.

But, no problem. The rest of the work is solid. This is just peanuts. We can take care later 👍

[NoaHimesaka1873]

@pazos Thanks to your comment, I actually managed to push frontlight support!

[pazos]

@pazos Thanks to your comment, I actually managed to push frontlight support!

Great! I think you're now on the mood to fix the last "doesn't work" from OP :)

restart KOReader: likely because it can't LD_PRELOAD rmpp-qtfb-shim again

You can add canRestart = no to your device definition to prevent users from trying it.

[NoaHimesaka1873]

Sonce it's 4am here now, I'll work on it after I get some sleep

[NoaHimesaka1873]

@pazos Thanks to your comment, I actually managed to push frontlight support!

Great! I think you're now on the mood to fix the last "doesn't work" from OP :)

restart KOReader: likely because it can't LD_PRELOAD rmpp-qtfb-shim again

You can add canRestart = no to your device definition to prevent users from trying it.

canRestart = no also removes the option to exit KOReader so I'll leave it like that.

[NoaHimesaka1873]

On second look, it doesn't seem like it even tries to relaunch itself.

[pazos]

canRestart = no also removes the option to exit KOReader so I'll leave it like that.

Are you sure?. Then the android version would lack the exit button, too

[Frenzie]

Yeah, only the restart menu entry has that condition. :-)

[NoaHimesaka1873]

Yeah, only the restart menu entry has that condition. :-)

Weird, when I test it, exit KOReader option is also gone too :(

[NoaHimesaka1873]

first pic is with canRestart set to no and second pic is without that thing.

[Frenzie]

Something about the behavior must've broken since it was originally implemented. Drat.

[pazos]

Just give it a quick look.

It turns out that, until now, every device that cannot restart (android and macos) has defined hasExitOptions = no. That makes sense as in those devices it isn't possible to reboot/poweroff/suspend from the device.

canRestart was only tested paired with hasExitOptions. The intent is to replace the exit_menu with the exit callback directly.

But canRestart does indeed remove the exit entry on a device that hasExitOptions and the on_hold callback never fires.

[Frenzie]

Indeed, it still works on Android.

It looks like this doesn't work as intended anymore however, even when removing the suspend/reboot/poweroff. For the sake of this PR that doesn't matter, but it should become:

if not Device:canRestart() and not Device:canSuspend() and not Device:canReboot() and not Device:canPowerOff() then
    exit_settings.exit_menu = exit_settings.exit
    exit_settings.exit = nil
    exit_settings.restart_koreader = nil
end

koreader/frontend/ui/elements/common_exit_menu_table.lua

Lines 27 to 57 in 9687034

	if not Device:canRestart() then
	exit_settings.exit_menu = exit_settings.exit
	exit_settings.exit = nil
	exit_settings.restart_koreader = nil
	end
	if Device:canSuspend() then
	exit_settings.sleep = {
	text = _("Sleep"),
	callback = function()
	UIManager:suspend()
	end,
	}
	end
	if Device:canReboot() then
	exit_settings.reboot = {
	text = _("Reboot the device"),
	keep_menu_open = true,
	callback = function()
	UIManager:askForReboot()
	end
	}
	end
	if Device:canPowerOff() then
	exit_settings.poweroff = {
	text = _("Power off"),
	keep_menu_open = true,
	callback = function()
	UIManager:askForPowerOff()
	end
	}
	end

[Commodore64user]

emulator with canRestart = no

[NoaHimesaka1873]

ironed out some more quirks to make it more smooth

[benoit-pierre]

Assuming everything will be squashed on merge, to fix the l10n module:

▸ git -C l10n checkout f00869ff7fb187bea212382636fe0877948ea7ae
▸ git commit -m 'revert unintended l10n changes' l10n

Also, now that the base PR was merged, please bump it as part of this PR:

▸ git -C base checkout fa825ad8edcdcef18b93c7575a9870437a6c4e7a
▸ git commit -m 'bump base' base

[NoaHimesaka1873]

@benoit-pierre Done!

[pazos]

@NoaHimesaka1873: we would need some instructions in https://github.com/koreader/koreader/wiki/Installation-on-ReMarkable

Could you add a new entry for the ReMarkable Paper Pro just below the manual installation?

The rest of the instructions also deserve some work, but first lets see how different the installation and launch of the program is on the rMPP

[Frenzie]

No need for parentheses here either. ;-)

[NoaHimesaka1873]

@pazos I was planning to do that after it was merged, but yeah I'll add it asap.

[NoaHimesaka1873]

Done! It's quite complicated though.

[Frenzie]

Were you ready here or still testing/tweaking final details?

[NoaHimesaka1873]

@Frenzie Now ready!
Though some change (all related to screen refresh) will come in the future after qtfb and shim can control refresh. Don't know when it will be implemented so leave as is and let xochitl control refresh for now.

[NoaHimesaka1873]

Also, current merge blocker is container upgrade to 22.04 to facilitate building remarkable-aarch64 target.

[Frenzie]

That's not a merge blocker, that's a nightly blocker. ;-)

[NoaHimesaka1873]

I made PR at koreader/virdevenv#117