[RFC] Remarkable port #1023

tcrs · 2019-12-19T20:43:26Z

I have a remarkable (https://remarkable.com/). The builtin PDF reader was too slow so I did a quick port of koreader. This is mostly based on the Sony PRS port.

I've never worked on koreader before so would appreciate any comments etc.

I display stuff could be improved. I think the remarkable has a modern eink display with reagl and stuff so could probably use better paths for partial updates etc (xochitl, the interface the remarkable ships with, does a good job of low-latency not too flashy updates). I don't really know much about the eink driver stuff so I've just left it alone now things seem to be working...

There's also a pull request open over in koreader/koreader for the other bits and bobs.

This change is

pazos · 2019-12-19T21:32:20Z

input/input-remarkable.h

+#define _KO_INPUT_REMARKABLE_H
+
+static void generateFakeEvent(int pipefd[2]) {
+	// TODO anything required here? Sony PRS port monitors battery charge etc


You can generate fake events that will be forwarded to the frontend as normal input events. There are a few defined in https://github.com/koreader/koreader-base/blob/master/input/input.c#L36-L41 but you can define more.

Normally is used to notify the program hardware changes, and we use libue to get that info from udev.

Once you have your fake events ready you can react to them in UIManager, but first do note that names are slightly different in the frontend. See how they're named in https://github.com/koreader/koreader/blob/f1f75c5cb0fa063c35f49a84b2e025f1d292ea01/frontend/device/input.lua#L176

pazos · 2019-12-19T21:34:23Z

Makefile.defs

@@ -151,6 +151,10 @@ else ifeq ($(TARGET), sony-prstux)
 	CHOST?=arm-linux-gnueabihf
 	export SONY_PRSTUX=1
 	export USE_LJ_WPACLIENT=1
+else ifeq ($(TARGET), remarkable)
+	CHOST?=arm-kobo-linux-gnueabihf


Are you using the kobo toolchain from https://github.com/koreader/koxtoolchain?

The Kobo tc should work fine but may be targeting somewhat older features.

Ah yea I had forgotten about that, I did use the kobo toolchain from koxtoolchain. I started using the toolchain provided by remarkable but it doesn't have a static libstdc++ library (which the koreader build seems to require).

I will take a look at creating a custom toolchain, I think I compared the kobo/remarkable SoCs and decided that there wasn't much difference in the hardware really, but the remarkable is running a much newer version of the Linux kernel so possibly it's worth building a custom toolchain for that reason.

remarkable's cpuinfo is exactly the same as the Kobo Forma. Not sure about kernel/libc versions in these devices and if is possible to make a new target for both and/or if makes sense. Pinging @NiLuJe

Oh, then you're in luck ;).

It's not plugged in the koxtoolchain frontend, but the 1.24 branch of my ct-ng tree has a kobomk7 sample, which should be much closer to that ;).

You'll just have to tweak the minimum kernel version (or not, as the Mk7 one is lower and still 4.1.x), and double-check the glibc version, as I'm not sure what poky 2.1.3 was using.

Probably something along those lines, since that's similar the the current Kindle (lab126) TC, IIRC.

Err, now that I think about it, probably best to take it as inspiration only, because it's using mainline GCC instead of the final Linaro GCC 7.5 release, and GCC 8/9 (possibly 10) is still suffering from severe neon performance regressions.

pazos

At first glance everything looks good. I dunno about that Archlinux change with luarocks will break older debian/ubuntu luarocks versions. Pinging @NiLuJe

Since the Remarkable runs plain linux you might want to port FBInk too. FBInk is used to give feedback on certain actions, like OTA updates. It shouldn't take you too long.

The same with the toolchain. You can pick one of the availables or create one for your device.

BTW, welcome aboard 😄

Frenzie · 2019-12-19T21:40:18Z

For reference, the other PR is koreader/koreader#5697.

Makefile

Frenzie · 2019-12-19T21:43:43Z

Makefile.defs

@@ -151,6 +151,10 @@ else ifeq ($(TARGET), sony-prstux)
 	CHOST?=arm-linux-gnueabihf
 	export SONY_PRSTUX=1
 	export USE_LJ_WPACLIENT=1
+else ifeq ($(TARGET), remarkable)
+	CHOST?=arm-kobo-linux-gnueabihf


The Kobo tc should work fine but may be targeting somewhat older features.

Frenzie · 2019-12-19T22:33:49Z

Makefile.third

@@ -682,7 +682,7 @@ $(LUA_SPORE_ROCK): $(THIRDPARTY_DIR)/lua-Spore/*.*
 		$(CMAKE) $(CMAKE_FLAGS) -DOUTPUT_DIR="$(CURDIR)/$(OUTPUT_DIR)" \
 		-DLUA_SPORE_VER=$(LUA_SPORE_VER) -DLD="$(CC)" \
 		-DCC="$(CC)" -DCFLAGS="$(CFLAGS) -I$(LUAJIT_DIR)/src" \
-		-DLUAROCKS=$(if $(DARWINHOST),"luarocks --lua-dir=/usr/local/opt/lua@5.1",luarocks) \
+		-DLUAROCKS=$(if $(DARWINHOST),"luarocks --lua-dir=/usr/local/opt/lua@5.1","luarocks --lua-version=5.1") \


This looks potentially suspicious; does it work on 2.x or is that just 3.x?

That's 3.x only, IIRC.

(For ref., I still patch my local copy w/ http://trac.ak-team.com/trac/browser/niluje/Configs/trunk/Kindle/Misc/koreader-luarocks-3.patch because I'm way too lazy to figure out what's got its panties in a twist ;)).

NiLuJe · 2019-12-20T03:08:20Z

IIRC, there's some fancy undocumented (on their side) flags for their specific tweaks to waveform shenanigans.

Most of it is detailed in the libremarkable sources, if memory serves ;).

In any case, for a start, that should be perfectly fine enough. I don't actually recall the extent of their tweaks and if they'd actually be useful here anyway ;).
(I seem to remember a fancier A2 which we probably won't ever care for, for instance. If, on the other hand, there's a fancier REAGL and/or making it behave requires some hoop-jumping like on some platforms (hi, Kindle), that'd be something worth looking into for later ;)).

Take that with a grain of salt, I've never ever seen a Remarkable, and I've only ever quickly skimmed libremarkable's code ;).

tcrs · 2020-01-27T21:08:46Z

I've added event generation for USB plug/unplug and battery charging/discharging. Looks like there is a delay of up to a minute or two for the battery status to change after plug/unplug (watching the uevents with udevadm monitor shows the same behaviour so I'm pretty sure it's not something wrong in the input code - don't know if this is normal on other devices or if it's specific to the remarkable.

Frenzie · 2020-01-27T22:16:30Z

You mean like this? I never really looked into the raw stuff.

koreader-base/input/input-kobo.h

Lines 44 to 56 in 26bdb0b

    
           if (uev.action == UEVENT_ACTION_ADD 
        
                   && uev.devpath 
        
                   && (UE_STR_EQ(uev.devpath, KOBO_USB_DEVPATH_PLUG) 
        
                       || UE_STR_EQ(uev.devpath, KOBO_USB_DEVPATH_HOST))) { 
        
               ev.code = CODE_FAKE_CHARGING; 
        
           } else if (uev.action == UEVENT_ACTION_REMOVE 
        
                   && uev.devpath 
        
                   && (UE_STR_EQ(uev.devpath, KOBO_USB_DEVPATH_PLUG) 
        
                       || UE_STR_EQ(uev.devpath, KOBO_USB_DEVPATH_HOST))) { 
        
               ev.code = CODE_FAKE_NOT_CHARGING; 
        
           } else { 
        
               continue; 
        
           }

36771d5

NiLuJe · 2020-01-27T22:17:35Z

If the plug/unplug events themselves are on time, I wouldn't worry about it too much ;).

But speaking of battery, there is sometimes a different state for "plugged and charging" vs. "plugged and not charging". Not sure if that's relevant here, though ;).

tcrs · 2020-01-28T20:27:26Z

I generate the plug/unplug events from the charger, which seem to be instantaneous. The charging/not-charging comes from the battery device, and its state seems slow to change. So the plugged and charging status are separate. The state usually goes something like:

plugged
...delay..
charging
...
unplugged
...delay...
not-charging

But you could get plugged+not-charging if the battery is full probably.

Looks like kindle & kobo have quite different device setup for usb plug/unplug.

I also noticed that the Sony PRS port has a custom input generation function in input-sony-prstux.h (which I started from to make the remarkable one), but never opens the fake_event input in its frontend device lua file so I don't think it will ever actually be used (in koreader at least).

pazos · 2020-01-29T14:27:42Z

I also noticed that the Sony PRS port has a custom input generation function in input-sony-prstux.h (which I started from to make the remarkable one), but never opens the fake_event input in its frontend device lua file so I don't think it will ever actually be used (in koreader at least).

It is in koreader/koreader#4567

Having support for the SonyPRSTux platform is great. But the support wasn't intended for other people outside its contributor. We can safely assume the number of users are <= 1. Maybe I'm wrong, but we lack:

toolchain definition
platform instructions to install the software over OEM.

So the only person who is able to give feedback about its state is @v01d and we're not in a hurry to merge PRs because we're sure the author of the port is able to figure out and fix things, even when the fixes never reach this repo.

ffi-cdecl/include/mxcfb-remarkable.h

ffi-cdecl/mxcfb_remarkable_decl.c

NiLuJe · 2020-02-01T00:30:34Z

ffi/framebuffer_mxcfb.lua

@@ -441,6 +452,12 @@ local function refresh_pocketbook(fb, refreshtype, waveform_mode, x, y, w, h)
    return mxc_update(fb, C.MXCFB_SEND_UPDATE, refarea, refreshtype, waveform_mode, x, y, w, h)
 end

+local function refresh_remarkable(fb, refreshtype, waveform_mode, x, y, w, h)
+    local refarea = ffi.new("struct mxcfb_update_data[1]")
+    refarea[0].temp = C.TEMP_USE_AMBIENT


To match xochitl:

Suggested change

refarea[0].temp = C.TEMP_USE_AMBIENT

if waveform_mode == C.WAVEFORM_MODE_DU then

refarea[0].temp = C.TEMP_USE_REMARKABLE

else

refarea[0].temp = C.TEMP_USE_AMBIENT

end

NiLuJe · 2020-02-01T00:34:22Z

ffi/framebuffer_mxcfb.lua

+        self.waveform_fast = C.WAVEFORM_MODE_DU
+        self.waveform_ui = C.WAVEFORM_MODE_AUTO
+        self.waveform_flashui = self.waveform_ui
+        self.waveform_full = C.WAVEFORM_MODE_GC16
+        self.waveform_partial = C.WAVEFORM_MODE_AUTO
+        self.waveform_night = C.WAVEFORM_MODE_GC16
+        self.waveform_flashnight = self.waveform_night
+        self.night_is_reagl = false


To match xochitl:

Suggested change

self.waveform_fast = C.WAVEFORM_MODE_DU

self.waveform_ui = C.WAVEFORM_MODE_AUTO

self.waveform_flashui = self.waveform_ui

self.waveform_full = C.WAVEFORM_MODE_GC16

self.waveform_partial = C.WAVEFORM_MODE_AUTO

self.waveform_night = C.WAVEFORM_MODE_GC16

self.waveform_flashnight = self.waveform_night

self.night_is_reagl = false

self.waveform_fast = C.WAVEFORM_MODE_DU

self.waveform_ui = C.WAVEFORM_MODE_GL16

self.waveform_flashui = C.WAVEFORM_MODE_GC16

self.waveform_full = C.WAVEFORM_MODE_GC16

self.waveform_partial = C.WAVEFORM_MODE_GL16

self.waveform_night = C.WAVEFORM_MODE_GC16

self.waveform_flashnight = self.waveform_night

self.night_is_reagl = false

ffi/mxcfb_remarkable_h.lua

tcrs · 2020-02-01T10:36:56Z

Thanks @NiLuJe for sorting out the waveform mapping & related stuff.

I can remove the luarocks --lua-version change as it sounds like that isn't suitable for the build setup expected (I'll keep it locally for my machine). Is it OK to rebase the branch this pull request is from?

Frenzie · 2020-02-01T10:44:32Z

Should be fine, here at least. :-)

Mostly copied from sony-prstux port

Assumes the koreader systemd unit is installed correctly. Launched koreader when the middle (home) button is held for 3 seconds. Inspired by https://github.com/dixonary/button-capture-reMarkable but significantly simpler.

Port not yet upstream

USB Plug/Unplug, charge status change. Seems to take a minute or two from USB Unplug to the battery state changing to discharging in sysfs.

Rather than reusing kobo toolchain. Built from koxtoolchain.

To get remarkable support

After some testing NiLuJe has figured out the proper mapping Removes a bunch of modes which are just unusable

As recommended by NiLuJe.

tcrs · 2020-02-01T12:19:11Z

Rebased on latest master, removed --lua-version build change

Frenzie · 2020-02-01T14:19:18Z

Makefile

@@ -29,7 +29,8 @@ all: $(OUTPUT_DIR)/libs $(if $(ANDROID),,$(LUAJIT)) \
 		$(if $(or $(DARWIN),$(WIN32),$(ANDROID),$(UBUNTUTOUCH),$(APPIMAGE)),,$(OUTPUT_DIR)/dropbear) \
 		$(if $(or $(KINDLE),$(KOBO),$(CERVANTES)),$(OUTPUT_DIR)/sftp-server,) \
 		$(if $(or $(DARWIN),$(WIN32)),,$(OUTPUT_DIR)/tar) \
-		$(if $(or $(KINDLE),$(KOBO),$(CERVANTES)),$(OUTPUT_DIR)/fbink,) \
+		$(if $(or $(KINDLE),$(KOBO),$(CERVANTES),$(REMARKABLE)),$(OUTPUT_DIR)/fbink,) \


That being said, my order comment applies more broadly.

Frenzie · 2020-02-01T14:19:34Z

Makefile

@@ -95,7 +97,7 @@ libs: \
 $(OUTPUT_DIR)/libs/libkoreader-input.so: input/*.c input/*.h $(if $(KINDLE),$(POPEN_NOSHELL_LIB),)
 	@echo "Building koreader input module..."
 	$(CC) $(DYNLIB_CFLAGS) -I$(POPEN_NOSHELL_DIR) -I./input \
-		$(if $(KOBO),-DKOBO,) $(if $(KINDLE),-DKINDLE,) $(if $(POCKETBOOK),-DPOCKETBOOK,) $(if $(SONY_PRSTUX),-DSONY_PRSTUX,) $(if $(CERVANTES),-DCERVANTES,)\
+		$(if $(KOBO),-DKOBO,) $(if $(KINDLE),-DKINDLE,) $(if $(POCKETBOOK),-DPOCKETBOOK,) $(if $(REMARKABLE),-DREMARKABLE,) $(if $(SONY_PRSTUX),-DSONY_PRSTUX,) $(if $(CERVANTES),-DCERVANTES,)\


Frenzie · 2020-02-01T14:20:18Z

thirdparty/fbink/CMakeLists.txt

        set(INSTALL_CMD3 ${CMAKE_COMMAND} -E rename ${SOURCE_DIR}/Release/fbdepth ${BINARY_DIR}/fbdepth)
    endif()
 endif()

 ko_write_gitclone_script(
    GIT_CLONE_SCRIPT_FILENAME
    https://github.com/NiLuJe/FBInk.git
-    tags/v1.20.3
+    origin/master


We really do prefer tags or specific commits.

master is a moving target. In this case we can prod @NiLuJe to create a new tag, generally speaking just use a hash like b06bd399611875fe98b52e5fdadd6525872dde5c.

I've just tagged v1.21.0 ;).

Thanks, updated to use v1.21.0

NiLuJe · 2020-02-01T14:50:09Z

Yeah, I'll tag today or tomorrow ;).

…

On Sat, Feb 1, 2020, 15:21 Frans de Jonge ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In thirdparty/fbink/CMakeLists.txt <#1023 (comment)> : > set(INSTALL_CMD3 ${CMAKE_COMMAND} -E rename ${SOURCE_DIR}/Release/fbdepth ${BINARY_DIR}/fbdepth) endif() endif() ko_write_gitclone_script( GIT_CLONE_SCRIPT_FILENAME https://github.com/NiLuJe/FBInk.git - tags/v1.20.3 + origin/master master is a moving target. In this case we can prod @NiLuJe <https://github.com/NiLuJe> to create a new tag, generally speaking just use a hash like b06bd399611875fe98b52e5fdadd6525872dde5c. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1023?email_source=notifications&email_token=AAA3KZXJZT2JL276IS3ATM3RAWAO7A5CNFSM4J5PTVF2YY3PNVWWK3TUL52HS4DFWFIHK3DMKJSXC5LFON2FEZLWNFSXPKTDN5WW2ZLOORPWSZGOCT4VZPQ#discussion_r373782418>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA3KZRVOBMC2OURSNAIW6TRAWAO7ANCNFSM4J5PTVFQ> .

Includes remarkable support

As on Kobo

Frenzie

Nothing jumps out at me; I'll leave merging to @NiLuJe ;-)

NiLuJe · 2020-02-02T18:25:12Z

And we're cooking with gas! ;).

Thanks!

* Re-enable HW dithering on Kindle (koreader/koreader-base#1034) * Update SQLite to 3.31.1 (koreader/koreader-base#1035) * reMarkable port (koreader/koreader-base#1023) * zsync2 (koreader/koreader-base#1036)

@chrox

* Switch to zsync2 Requires koreader/koreader-base#1036 * Simplify FBInk syntax The weird-ass workarounds for -s's subopts handling are no longer needed w/ FBInk >= 1.21.0 * Update base * Re-enable HW dithering on Kindle (koreader/koreader-base#1034) * Update SQLite to 3.31.1 (koreader/koreader-base#1035) * reMarkable port (koreader/koreader-base#1023) * zsync2 (koreader/koreader-base#1036) * zsync2 means we can finally have nice things (OpenStack backed storage) * We also no longer need that insane workaround on ARM * And comment out @chrox's mirror, which appears to be down. * Warn that a malformed URL will horribly blow up in fun and interesting ways

@chrox

* Switch to zsync2 Requires koreader/koreader-base#1036 * Simplify FBInk syntax The weird-ass workarounds for -s's subopts handling are no longer needed w/ FBInk >= 1.21.0 * Update base * Re-enable HW dithering on Kindle (koreader/koreader-base#1034) * Update SQLite to 3.31.1 (koreader/koreader-base#1035) * reMarkable port (koreader/koreader-base#1023) * zsync2 (koreader/koreader-base#1036) * zsync2 means we can finally have nice things (OpenStack backed storage) * We also no longer need that insane workaround on ARM * And comment out @chrox's mirror, which appears to be down. * Warn that a malformed URL will horribly blow up in fun and interesting ways

tcrs changed the title ~~Remarkable~~ [RFC] Remarkable port Dec 19, 2019

pazos reviewed Dec 19, 2019

View reviewed changes

Frenzie reviewed Dec 19, 2019

View reviewed changes

NiLuJe reviewed Feb 1, 2020

View reviewed changes

ffi-cdecl/include/mxcfb-remarkable.h Show resolved Hide resolved

NiLuJe reviewed Feb 1, 2020

View reviewed changes

ffi-cdecl/mxcfb_remarkable_decl.c Show resolved Hide resolved

NiLuJe reviewed Feb 1, 2020

View reviewed changes

ffi/mxcfb_remarkable_h.lua Show resolved Hide resolved

tcrs and others added 11 commits February 1, 2020 12:16

Remarkable port

2110ba7

Mostly copied from sony-prstux port

Add additional waveform modes from libremarkable

cae614f

Fix tab indentation, order remarkable before sony-prs

20fa50d

remarkable: no longer using lj-wpaclient

e0f8848

remarkable: add simple button listener launcher

2e7ad6e

Assumes the koreader systemd unit is installed correctly. Launched koreader when the middle (home) button is held for 3 seconds. Inspired by https://github.com/dixonary/button-capture-reMarkable but significantly simpler.

remarkable: include fbink & fbdepth

a5171d9

Port not yet upstream

remarkable: add fake event generation

03f936e

USB Plug/Unplug, charge status change. Seems to take a minute or two from USB Unplug to the battery state changing to discharging in sysfs.

remarkable: switch to remarkable toolchain

bd27cc4

Rather than reusing kobo toolchain. Built from koxtoolchain.

remarkable: build FBInk from master branch

7c5303c

To get remarkable support

remarkable: update mxcfb ffi to use header from FBInk

8d0967e

After some testing NiLuJe has figured out the proper mapping Removes a bunch of modes which are just unusable

remarkable: update waveform modes based on FBInk

96ed087

As recommended by NiLuJe.

tcrs force-pushed the remarkable branch from 6575e44 to 96ed087 Compare February 1, 2020 12:18

Frenzie reviewed Feb 1, 2020

View reviewed changes

tcrs added 3 commits February 2, 2020 10:02

Use v1.21.0 fbink version

33be5c8

Includes remarkable support

Reorder some lists alphabetically

ff73e9c

remarkable: strip fbdepth

0f1f08a

As on Kobo

Frenzie approved these changes Feb 2, 2020

View reviewed changes

NiLuJe merged commit 45347f7 into koreader:master Feb 2, 2020

Frenzie mentioned this pull request Apr 12, 2020

Bookeen Saga ereader. Not sure where to start. koreader/koreader#6043

Open

11 tasks

[RFC] Remarkable port #1023

[RFC] Remarkable port #1023

Conversation

tcrs commented Dec 19, 2019 • edited by Frenzie

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

NiLuJe Dec 28, 2019 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

NiLuJe Dec 29, 2019 • edited

Choose a reason for hiding this comment

pazos left a comment

Choose a reason for hiding this comment

Frenzie commented Dec 19, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

NiLuJe commented Dec 20, 2019 • edited

tcrs commented Jan 27, 2020

Frenzie commented Jan 27, 2020

NiLuJe commented Jan 27, 2020 • edited

tcrs commented Jan 28, 2020

pazos commented Jan 29, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tcrs commented Feb 1, 2020

Frenzie commented Feb 1, 2020

tcrs commented Feb 1, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

NiLuJe commented Feb 1, 2020 via email

Frenzie left a comment

Choose a reason for hiding this comment

NiLuJe commented Feb 2, 2020

tcrs commented Dec 19, 2019 •

edited by Frenzie

NiLuJe Dec 28, 2019 •

edited

NiLuJe Dec 29, 2019 •

edited

NiLuJe commented Dec 20, 2019 •

edited

NiLuJe commented Jan 27, 2020 •

edited