GUI Launcher for Kindle Touch

[shinew]

Thank you for the documentation!

Regarding the # of hacks available, the Kindle is such a specialized(underpowered and fragile) device that it already does what it's supposed to do pretty well, so IMO there isn't much need to have hundreds of useful & practical hacks anyway.

I think it'll be really useful for those who're knowledgeable and willing enough to dive into low-level hacking, to document & share whatever they've discovered in a centralized place such as a wiki or something. At the moment I don't think there is lack of interests, but rather the commitment required for combing through and figuring out stuff is still pretty steep without much documentation.

[yifanlu]

Ok, whatever, I released the installer. We now have 3 hacks for the Touch!

Note that I should call this release "beta" since I only have myself as a tester until now. I think adding new menu options should be easy to figure out if read the two example extensions. I've also did a nice writeup a couple posts ago about Java menu hooks.

[geekmaster]

Personally, I think the limiting factor for previous touches was the difficulty in using the 5-way controller to navigate.

It is my belief that the touchscreen and the "open" GUI environment will cross a threshold making app development available to "lesser geeks" (of which there are many).

My eventual plan is to add one of those fancy new drag-n-drop GUI "programming" environments that are becoming popular. You can build useful GUI scripts/plug-ins as easy as playing with Legos, and they can do useful things too.

[amirza]

I've just installed your launcher, and it works perfectly.
I'm amazed that amazon has not been able to deliver a fix for this issue,.

Thanks for all your efforts...

[ixtab]

Wow yifanlu,

thanks for all the amazing stuff you're producing!

The jailbreak came just in time (1 day before I got my KT), and since then, I have been poking around on the device. I actually did install your launcher yesterday but didn't get around to code anything for it yet. I'll take a look at the installer now and test-drive a bit.

As to my own intentions, besides playing around I was looking a bit closer at two things:
- WPA-PAE encryption. I use this, and by now I do have a working shell script with wpa_cli etc. (and an ugly hack in the pillow/JS UI to invoke it) for my own setup.
- Localization. (JS UI for now, Java parts later).

My idea is that it should be (is) relatively simple to modify the UI, and its functionality, using JS itself. Strings and functions -- essentially, arbitrary parts of the JS subsystem -- can be modified at will by simple script-src'ing other JS scripts from the HTML. But there must be a better, less intrusive way (i.e. ideally without modifying the root fs at all).

BTW, I think the launcher is a great opportunity to make the patch/hack/package/whatever-management that you were discussing in other threads both extremely easy and user-friendly (graphical UI anyone?)

I'm still trying to understand how the overall architecture is plugged together, so some insight on the following questions is appreciated. So far, my understanding is that it's a conglomerate of native Linux apps, the Java (booklets etc.) part providing much of the UI, but also pillow / JavaScript parts. All of these seem to communicate via dbus and/or lirc.

Is this correct? Can someone provide more insight into the architecture? (like, what is lirc in the first place?...)

Which method are you guys using to quickly check the results of some changes? The most reliable method I found so far is to "killall cvm" or "killall pillowd", which seems to reload everything except Linux itself. Do you know of more fine-grained/faster ways (e.g. when putting new files in the classpath, or changing a JS file). Are there any ways of debugging?

Two final things: if anyone is interested in the localization and/or WPA-EAP part, let me know.

And: how do you guys consolidate all the information from the forums? I know forums are a great way to communicate and share, but the info tends to get scattered around. Sticky threads are a workaround, but wouldn't a Wiki make sense for aggregating the infos in a structured way(or does one exist already)?

[yifanlu]

Originally Posted by ixtab

Wow yifanlu,

thanks for all the amazing stuff you're producing!

The jailbreak came just in time (1 day before I got my KT), and since then, I have been poking around on the device. I actually did install your launcher yesterday but didn't get around to code anything for it yet. I'll take a look at the installer now and test-drive a bit.

As to my own intentions, besides playing around I was looking a bit closer at two things:
- WPA-PAE encryption. I use this, and by now I do have a working shell script with wpa_cli etc. (and an ugly hack in the pillow/JS UI to invoke it) for my own setup.
- Localization. (JS UI for now, Java parts later).

My idea is that it should be (is) relatively simple to modify the UI, and its functionality, using JS itself. Strings and functions -- essentially, arbitrary parts of the JS subsystem -- can be modified at will by simple script-src'ing other JS scripts from the HTML. But there must be a better, less intrusive way (i.e. ideally without modifying the root fs at all).

BTW, I think the launcher is a great opportunity to make the patch/hack/package/whatever-management that you were discussing in other threads both extremely easy and user-friendly (graphical UI anyone?)

I'm still trying to understand how the overall architecture is plugged together, so some insight on the following questions is appreciated. So far, my understanding is that it's a conglomerate of native Linux apps, the Java (booklets etc.) part providing much of the UI, but also pillow / JavaScript parts. All of these seem to communicate via dbus and/or lirc.

Is this correct? Can someone provide more insight into the architecture? (like, what is lirc in the first place?...)

Which method are you guys using to quickly check the results of some changes? The most reliable method I found so far is to "killall cvm" or "killall pillowd", which seems to reload everything except Linux itself. Do you know of more fine-grained/faster ways (e.g. when putting new files in the classpath, or changing a JS file). Are there any ways of debugging?

Two final things: if anyone is interested in the localization and/or WPA-EAP part, let me know.

And: how do you guys consolidate all the information from the forums? I know forums are a great way to communicate and share, but the info tends to get scattered around. Sticky threads are a workaround, but wouldn't a Wiki make sense for aggregating the infos in a structured way(or does one exist already)?

That's a lot of questions, I'll try to answer them. The Kindle "OS" or whatever you call it starts out when the Kernel calls upstart (not init.d anymore). Upstart loads the low level linux components. Lipc is this thing amazon wrote to allow all components to talk to one another. Everything in the Kindle is moduler. In theory, I can write a python implementation of the media player and set a lipc event handler and property reader and music will play through python. Pillows are the HTML5/JS component of the system. It's weird how they decided to make some parts of the OS HTML. My guess is that either 1) they wanted the whole thing to be HTML and halfway through realized it wouldn't work and fell back to Java, but didn't have time to rewrite everything or 2) they wanted the whole thing is Java but didn't have time to rewrite everything and then quickly whipped up some HTML stuff they were doing for the browser anyways. Either way, pillows are registered into the appdb (a sqlite3 database) which also contains a listing of book handlers (java reader plugins) and download handlers (unused). Overall, it's VERY messy. I use upstart to restart the framework "restart framework" from shell would do it. Using upstart is the best way because it allows you to mock a reboot of the system without actually rebooting. You can debug native apps with the gdb that is built in. You can debug Java using JDB remote by enabling debugging and using usbnetwork. I agree that we need a wiki or something, but I'm afraid that it'll be just me writing everything and I don't have the time for that.

[ixtab]

Originally Posted by yifanlu

That's a lot of questions, I'll try to answer them.

Yeah sorry, I realized that after posting.

Originally Posted by yifanlu

The Kindle "OS" or whatever you call it starts out when the Kernel calls upstart (not init.d anymore). Upstart loads the low level linux components. Lipc is this thing amazon wrote to allow all components to talk to one another. Everything in the Kindle is moduler. In theory, I can write a python implementation of the media player and set a lipc event handler and property reader and music will play through python. Pillows are the HTML5/JS component of the system. It's weird how they decided to make some parts of the OS HTML. My guess is that either 1) they wanted the whole thing to be HTML and halfway through realized it wouldn't work and fell back to Java, but didn't have time to rewrite everything or 2) they wanted the whole thing is Java but didn't have time to rewrite everything and then quickly whipped up some HTML stuff they were doing for the browser anyways. Either way, pillows are registered into the appdb (a sqlite3 database) which also contains a listing of book handlers (java reader plugins) and download handlers (unused). Overall, it's VERY messy.

Cool, thanks for the explanation! It gets slightly clearer now. I'll look around some more...

Originally Posted by yifanlu

I use upstart to restart the framework "restart framework" from shell would do it. Using upstart is the best way because it allows you to mock a reboot of the system without actually rebooting. You can debug native apps with the gdb that is built in. You can debug Java using JDB remote by enabling debugging and using usbnetwork.

Again, thanks. I'll try to stay away from the native part as long as I can (too unfamiliar with both gdb and arm assembly), but JDB sounds like a plan to me.

Originally Posted by yifanlu

I agree that we need a wiki or something, but I'm afraid that it'll be just me writing everything and I don't have the time for that.

I totally see your point, and of course that's not the intention of it. I think that for exchange and brand new stuff, the forum is the best place. But ideally, the relevant infos should be aggregated somewhere in a more or less structured manner. Anyway, I just realized that I possibly didn't have to look any further than https://wiki.mobileread.com/. I'll try to get something on there, and we'll see how it goes.

So, finally back on topic with the launcher. A few remarks, as tested with my device after installation and a reboot:

- the installer as such works fine.
- the loop-mount of the configuration file (which is an excellent idea BTW!) does not always seem to work reliably here. I cannot reproduce the case, but in some cases the mount just vanished. As a side note, when issuing "restart framework" as you mentioned above, it may also be mounted multiple times. Furthermore, it will also never appear on the 'settings' and 'experimental' page, but only within ebooks. My personal suggestion is for this particular case to simply not enable the plugin at all, but to just leave the home screen booklet patched. IMO, this avoids confusion ("Launcher" only available on Home screen. Ican remember and live with that) as well as other complications. Just a thought though.

- usbnetwork toggling works. a small glitch: after being connected via usbNetwork and manually issuing 'reboot', the label says "Stop Usbnetwork" (instead of "Start") when invoked the first time (i.e. when actually still disabled). Selecting this once enables usbNetwork connectivity, and afterwards, the labels are correct. This is reproducible, so I guess it just writes a state file somewhere which is re-read with the wrong state after reboot(?)

- orientation changes don't work here at all. When using any landscape orientation, showlog -f says:
111220:221736 winmgr[1073]: W ligl:fbioctl::Waiting for marker

whereas for portrait inverted, I always get:
111220:221739 winmgr[1073]: E lua:invlock::Invalid global lock: D
111220:221740 winmgr[1073]: W ligl:fbioctl::Waiting for marker

On my device, this also happens when manually invoking "ipc-set-prop com.lab126.winmgr orientationLock [UDLR]". Nothing happens, ever.

OK, I hope that's of some use. Thanks again :-)

[yifanlu]

Originally Posted by ixtab

Yeah sorry, I realized that after posting.


Cool, thanks for the explanation! It gets slightly clearer now. I'll look around some more...



Again, thanks. I'll try to stay away from the native part as long as I can (too unfamiliar with both gdb and arm assembly), but JDB sounds like a plan to me.



I totally see your point, and of course that's not the intention of it. I think that for exchange and brand new stuff, the forum is the best place. But ideally, the relevant infos should be aggregated somewhere in a more or less structured manner. Anyway, I just realized that I possibly didn't have to look any further than https://wiki.mobileread.com/. I'll try to get something on there, and we'll see how it goes.

So, finally back on topic with the launcher. A few remarks, as tested with my device after installation and a reboot:

- the installer as such works fine.
- the loop-mount of the configuration file (which is an excellent idea BTW!) does not always seem to work reliably here. I cannot reproduce the case, but in some cases the mount just vanished. As a side note, when issuing "restart framework" as you mentioned above, it may also be mounted multiple times. Furthermore, it will also never appear on the 'settings' and 'experimental' page, but only within ebooks. My personal suggestion is for this particular case to simply not enable the plugin at all, but to just leave the home screen booklet patched. IMO, this avoids confusion ("Launcher" only available on Home screen. Ican remember and live with that) as well as other complications. Just a thought though.

- usbnetwork toggling works. a small glitch: after being connected via usbNetwork and manually issuing 'reboot', the label says "Stop Usbnetwork" (instead of "Start") when invoked the first time (i.e. when actually still disabled). Selecting this once enables usbNetwork connectivity, and afterwards, the labels are correct. This is reproducible, so I guess it just writes a state file somewhere which is re-read with the wrong state after reboot(?)

- orientation changes don't work here at all. When using any landscape orientation, showlog -f says:
111220:221736 winmgr[1073]: W ligl:fbioctl::Waiting for marker

whereas for portrait inverted, I always get:
111220:221739 winmgr[1073]: E lua:invlock::Invalid global lock: D
111220:221740 winmgr[1073]: W ligl:fbioctl::Waiting for marker

On my device, this also happens when manually invoking "ipc-set-prop com.lab126.winmgr orientationLock [UDLR]". Nothing happens, ever.

OK, I hope that's of some use. Thanks again :-)

I enabled it in books because I might want to run a launcher action in a book. I'll have to fix the mounting thing. It was supposed to be a temporary fix but I never got around to making something full. Usbnetwork only updates the labels if you use Launcher. If you turn usbnetwork on/off any other way, it won't update. You can do a soft link from usbnetwork.sh to the one in extensions/usbnetwork if you want and then it will always update the label. Rotations only work in books and in the web browser.

[redblade]

Hi Yifanlu,
thanks a lot for all your effort to open kindle to developers.
Could you add some note to your opensource project, GuiLauncher, on how to generate install.bin from your java sources? Or if you prefer, some explaination in this forum.

Thanks

[miguelos]

Launcher works like a charm
one minor flaw, though :
screen orientation causes on-screen-keyboard to cover buttons in "go to page or location" popup window (Cancel/Page/Location) - causing it unable to jump there. Have to jump in portrait mode, then rotate screen..

interestingly pdfs in this rotation look quite well - one display of page is divided to 3 steps (fist 47%, second 47%, last 6% of the page). If only finds where to remove margins..

[yifanlu]

I can't really help how amazon coded screen rotations.

I'm working on the update creator tool. It'll take a while.

[baraka17360]

Hi Yifanlu,

Thanks a lot for a great "Launcher". However, "...Houston, we have a problem..."

Here are my first impressions-experiences of using the landscape mode:

1. At the certain point the screen stops reacting to the touch navigation. It reacts on 'menu' touch, but I cannot move anyway - not the next/previous page, not the next/previous chapter. It is a consistent bug.

2. To make it work again I have to go to Home and start the book again - and here (!) another bug!. It loses the bookmark! Suddenly it shows absolutely different place in the book that I was reading and I have to move back page by page (and it's becoming "non-responding" to touch again). The arrow "back" does not put me back to my bookmark either. It is also a consistent bug.

3. When I return to my book from "Home", left 2/3rds of the screen are black - I have to go to the next page to get rid of it.

So unfortunately, had to go back to 'portrait' mode. Do not know how well portrait work with launcher.

I would test it further and will let you know.

[yifanlu]

Again, this is an undocumented built-in feature. I did not make the landscape mode. See my previous post.

[ixtab]

Hi all,

just an update on the "putting together the information in a Wiki" discussion above: I have now created an initial version at https://wiki.mobileread.com/wiki/Kindle_Touch_Hacking

This is still very incomplete, but at least the outline of what I was roughly imagining is there. Let me know if it makes any sense. If you think it does, you're more than welcome to enhance the page right there! :-)

[yifanlu]

Originally Posted by ixtab

Hi all,

just an update on the "putting together the information in a Wiki" discussion above: I have now created an initial version at https://wiki.mobileread.com/wiki/Kindle_Touch_Hacking

This is still very incomplete, but at least the outline of what I was roughly imagining is there. Let me know if it makes any sense. If you think it does, you're more than welcome to enhance the page right there! :-)

Nice job on that. I've added a request for an Android port for the touch. I want to work on this and find any people who are willing to help.