Koreader running on Kobos (all of them, hopefully!)

[mementomori]

sed -n '/ntx/,+2p'

im learning a lot of neat unix commands here.

[Frenzie]

I don't know if sed is neat. I think it becomes horribly complicated rather quickly. :P But sed is always available and if all you need is something like grep plus the next n lines it's pretty neat indeed.

[roger64]

Hi

Forgive me if this post seems rude. It really was not my intention and I waited quite a long time before sending it.

I was using koreader some months ago and enjoyed it (who cares?). I thought it was more reactive, more handy (folders), generaly speaking more pleasant. I know there has been recently a lot of work on some very technical matters (for me). But there are two basic things that keep me away from using koreader:

1. - It does not have the equivalent of TypeGenius for side-loaded fonts. I had it for some weeks and it was lost after one update.
2. - Hyphenation quality is under average. We have good hyphenation dictionaries in our language, how come we cannot use them?

For these two basic reasons, I switched back to the main Kobo reading software some two or three months ago. Can some progress be expected on these two points above in the next months?

[hawhill]

Maybe - development is driven only by motivation. I know that not everyone is a programmer, but those who are, I encourage to take a look: It's easy to contribute!

That said, a few words on the two issues you raised:

- TypeGenius: I've seen that for the first time now. You're missing it on an e-ink screen? I think I need to compare a few screenshots to see what it does on a technical level. That one should not be too hard (but still considerable work, so don't hold your breath), since I think there's some groundwork already within the freetype font rendering engine that we're using (and I bet that's what the Kobo software is using, too).

- Hyphenation: Well, when there are good free hyphenation dictionaries, we certainly like to include them in the distribution! Can you give pointers? I think they'd be welcome for CoolReader, too.

[Ken Maltby]

Originally Posted by hawhill

Maybe - development is driven only by motivation. I know that not everyone is a programmer, but those who are, I encourage to take a look: It's easy to contribute!

That said, a few words on the two issues you raised:

- TypeGenius: I've seen that for the first time now. You're missing it on an e-ink screen? I think I need to compare a few screenshots to see what it does on a technical level. That one should not be too hard (but still considerable work, so don't hold your breath), since I think there's some groundwork already within the freetype font rendering engine that we're using (and I bet that's what the Kobo software is using, too).

- Hyphenation: Well, when there are good free hyphenation dictionaries, we certainly like to include them in the distribution! Can you give pointers? I think they'd be welcome for CoolReader, too.

As I understand it ( in other words, what I've read, not something I've worked on) freetype 2 can both work with a font's hinting and apply a process to generate some on it's own. "The open-source FreeType 2 font rendering engine uses an auto-hinter when such hinting data are not present..." There is also some antialiasing applied, I think.

How are the Weight and Sharpness adjustments of TypeGenius any different/superior to those adjustments in KOreader? It may only take some adjustments to the FreeType renderer to meet your needs. One thing to keep in mind though, is that the font itself normally needs the hinting included, if the best results are to be achieved. Kobo's TypeGenius works with Kobo supplied fonts.

KOreader has the ability to use hyphenation dictionaries, which is probably the biggest step. Finding a better un-copyrighted hyphenation dictionary, is what would be needed.

Luck;
Ken

P.S.; On the screensaver front; the suspend to book cover image seems to work with .epub but if I turn that off trying to suspend puts me back in the KSM menu. (So I'll double check the default image file and the path to it.) I moved the image file and changed the path, and it no longer crashes on suspend. Now in a .pdb file with or without the use the cover set all I get is the Suspended message. So everything before that must have failed, don't see why yet.

[WS64]

Can we edit the existing hyphenation dictionaries?
One problem I come across especially in the German version is that it tries to create very small words with just one or two letters on one side of the hyphen, which usually is wrong.
In my opinion it would help a lot already to set up the minimum of letters each side of a hyphen should have.

While we're at it, can we sort the hyphenation dictionaries? I tried and failed to do so...

[roger64]

@hawhill, Ken Maltby, WS64

Thanks for your replies.

I need a little time to provide you online with the most up to date free hyphenation dictionnaries in my language (French). I do this a.s.a.p.

For the side-loaded font with my Kobo Glo, I use regularly Linux Libertine otf. I am used to add a little weight to it and I had installed the -deprecated- Kobo Patcher v. 07 which allowed to replicate the advanced Type Genius features to side-loaded fonts. I still use v. 2.8.1.

Now, I think GeoffR patcher does the same trick (sharpness and weight). The problem for him is that he has to issue a new patch for every Kobo version, because his patch is not official. This should not happen for koreader.

[roger64]

1. - French hyphenation dictionary

Here is the download link for the main dictionary.
http://www.dicollecte.org/download.php?prj=fr

On this page, you can download (flagged with a green star) "Césures 3.0." which will provide you with the "hyph-fr-v3.0" folder. Within the file "README_hyph_fr.txt" are the mail coordinates of the authors of the three successives versions.

This hyphenation dictionary works with a software named "hyphen" which is part of the Hunspell project. It has been converted from a dictionary for Latex and further improved by several persons. I have no other information.

2. - About Kobo hyphenation

Here are also some information published by Joël Goguen about improving hyphenation for a Kobo plugin.
https://github.com/jgoguen/calibre-k...er#hyphenation

Other information and related files can be found on MobileRead. I hope this can be of some help.

[mementomori]

Hello,

Does anyone know whether we can call this using some sort of a function pointer?

~/Desktop/kobo-update-3.11.0/koboroot/usr/local/Kobo$ arm-linux-gnueabihf-objdump -T -C ./libnickel.so.1.0.0 |grep -i "imx508" |grep -i "suspend"
004ec9d8 g DF .text 00000010 Base iMX508::suspend()
or maybe its this one
0071374c g DF .text 00000024 Base AutomationSurface::sleep(int)
or this one?
00758d20 g DF .text 00000028 Base DevicePowerWorkflowManager::requestSleep()

[hawhill]

Ah, now we're getting at something. Yes, we can. Either by linking binary code, or by using LuaJIT+FFI. My first tests would be using the latter. On the weekend, I can try a bit with my Kobo Mini. I guess the API is the same there.

It's C++ code, so we might have to jump through a few hoops with the calling convention. And iMX508 seems to be a class that we might have to set up an object for before using that method. A good start, but lots to fiddle with.

[mementomori]

yeah, the trick does work, i tried out on some sample codes.. At first I tried to do something like this.. (the code for the library is below, compiled using:

g++ -Wall -shared -fPIC -o libdynamic.so dynamic.cpp

and the main is compiled using:

g++ main.cpp -o go -L. -ldynamic -ldl

for void dynamic::dynamicClass::method01()

#include <dlfcn.h>
    handle = dlopen("/usr/local/lib/libdynamic.so",RTLD_LAZY);
    void (* functionpointer_method01)(void*);
    functionpointer_method01 = (void (*)()) dlsym(handle, "_ZN7dynamic12dynamicClass8method01Ev");

which works but the object wasn't created, so this might be good for static methods only. Also i had to insert an extra argument for my function pointers and I have no idea why.

e.g

    void method02(int a)

becomes

    void method02(int X,int a);

and i dont know what X is for. Passing 0 to it causes segmentation fault and passing an address to it does not change the address' value. Any idea on why this happens, I'll be happy to hear.

this approach seems to work, but as you mentioned it, will need some labor:

/* 23 October 2014: The Lost Header Exploit */

/*
~/projects/dynamictest$ objdump -T /usr/local/lib/libdynamic.so

/usr/local/lib/libdynamic.so:     file format elf32-i386

DYNAMIC SYMBOL TABLE:
00000000      DF *UND*	00000000  GLIBCXX_3.4 _ZNSolsEi
00000000      DF *UND*	00000000  GLIBC_2.1.3 __cxa_atexit
00000000  w   D  *UND*	00000000              __gmon_start__
00000000  w   D  *UND*	00000000              _Jv_RegisterClasses
00000000      DF *UND*	00000000  GLIBCXX_3.4 _ZNSt8ios_base4InitC1Ev
00000000      DF *UND*	00000000  GLIBCXX_3.4 _ZNSt8ios_base4InitD1Ev
00000000  w   D  *UND*	00000000              _ITM_deregisterTMCloneTable
00000000      DF *UND*	00000000  GLIBCXX_3.4 _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc
00000000  w   D  *UND*	00000000              _ITM_registerTMCloneTable
00000000      DO *UND*	00000000  GLIBCXX_3.4 _ZSt4cout
00000000      DF *UND*	00000000  GLIBCXX_3.4 _ZNSolsEPFRSoS_E
00000000      DF *UND*	00000000  GLIBCXX_3.4 _ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_
00000000  w   DF *UND*	00000000  GLIBC_2.1.3 __cxa_finalize
00000962 g    DF .text	00000047  Base        _ZN7dynamic12dynamicClass8method02Ev
0000090c g    DF .text	0000000d  Base        _ZN7dynamic12dynamicClassC2Ei
0000090c g    DF .text	0000000d  Base        _ZN7dynamic12dynamicClassC1Ei
00002030 g    D  .bss	00000000  Base        _end
0000202c g    D  .data	00000000  Base        _edata
00000920 g    DF .text	00000042  Base        _ZN7dynamic12dynamicClass8method01Ev
0000091a g    DF .text	00000005  Base        _ZN7dynamic12dynamicClassD1Ev
0000202c g    D  .bss	00000000  Baseg++ main2.cpp -o go -L. -ldynamic -ldl        __bss_start
0000091a g    DF .text	00000005  Base        _ZN7dynamic12dynamicClassD2Ev
00000a76 g    DF .text	00000089  Base        _ZN7dynamic12dynamicClass8method05Eii
00000734 g    DF .init	00000000  Base        _init
00000b74 g    DF .fini	00000000  Base        _fini
00000a0e g    DF .text	00000068  Base        _ZN7dynamic12dynamicClass8method04Ei
000009aa g    DF .text	00000063  Base        _ZN7dynamic12dynamicClass8method03Ei
*/

// rearranging base based on address and demangling, i get: 
// -----------------------------------------
// dynamic::dynamicClass::dynamicClass(int)
// dynamic::dynamicClass::~dynamicClass()
// dynamic::dynamicClass::~dynamicClass()
// dynamic::dynamicClass::method01()
// dynamic::dynamicClass::method02()
// dynamic::dynamicClass::method03(int)
// dynamic::dynamicClass::method04(int)
// dynamic::dynamicClass::method05(int, int)
// -----------------------------------------
/* note: why dont i get the return value? */


#include <iostream>

namespace dynamic
{
    class dynamicClass
    {
	public:
	dynamicClass(int);
	~dynamicClass();
	void method01();
	void method02();
	void method03(int);
	void method04(int);
	void method05(int,int);
    };
}

using namespace std;
int main ()
{
    dynamic::dynamicClass oDynamicClass(1);
    oDynamicClass.method01();
    oDynamicClass.method02();
    oDynamicClass.method03(1);
    oDynamicClass.method04(1);
    oDynamicClass.method05(1,2);
    return (0);
}



/* outputs: 
  method01
  method02
  method03(1)
  method04(1)
  method05(1,2)
*/

I think being able to call libNickel subroutines would be a great help in Kobo development, not just for KoReader.

Complete source:
https://www.dropbox.com/s/1idg3q2gql...er.tar.gz?dl=0

[gontofe]

Originally Posted by Ken Maltby

On the screensaver front; the suspend to book cover image seems to work with .epub but if I turn that off trying to suspend puts me back in the KSM menu. (So I'll double check the default image file and the path to it.)

Ken, Chrox has fixed something around the screensaver in 880 (for me, crashing back to KSM on suspend for books with no cover/use this cover disabled), see if this fixes anything for you too.

[Ken Maltby]

Originally Posted by gontofe

Ken, Chrox has fixed something around the screensaver in 880 (for me, crashing back to KSM on suspend for books with no cover/use this cover disabled), see if this fixes anything for you too.

I no longer had the crash to KSM, after I redid my designated screensaver image file and path.

With 880 I can suspend to an .epub cover page, but not to a .pdb cover. For .pdb files I get the suspended Message no matter what. With .epub I also get the Suspended Message when the use this cover is disabled. The thing is there are two conditions before the Suspended Message, both must have failed. One is for my designated image file, the other is for a folder of image files. The defaults variable KOBO_SCREEN_SAVER is used in screensaver.lua
But it is not working, as I would expect.

I redid both Defaults and the Persistent Defaults, and now my designated image (cracked.png) is showing up, instead of the Suspended Message with a .pdb file. With an .epub and the use this book's cover disabled, I am now getting the cracked.png image. With the use this book's cover enabled, I get the cover image.

So, with redoing both Defaults file entries my designated image is again available, working as it should.

Luck;
Ken

[Ebri]

Somewhere during KSM install I had to factory reset, and install everything (I remained with advboot for now), the system is more or less the same as it was this morning (except for upgrading from firmware 3.8 to 3.11), but earlier when I long pressed a word, the dictionary came up with the English - Hungarian one first, and I can skip to the English-English. But now with the same names (I thought the order of lookup is the order of filenames, but I am definitely not right) the English-English dictionary pops up. Is there a way to set my two language dictionary as default?

[hawhill]

Originally Posted by mementomori

yeah, the trick does work, i tried out on some sample codes.. [...]
Also i had to insert an extra argument for my function pointers and I have no idea why.

That would be a pointer to the class instance (i.e. object) data. I need to refresh my memories regarding C++ calling convention. I think this is too much for this thread, which is absolutely overloaded with information, discussion and so forth in any case. Plus, learning to use Nickel's API for suspending would be of huge benefit for other projects, too. So if you continue your venture (please do!), I think we should discuss this in another forum thread for now. Feel free to start one! This is, after all, the Developer's Corner, not the Software Announce and Support forum, even if it looks different at times :-)

I'll not get to start working on this before the weekend.