Database Info

Recompile and overwrite applications in system/app

As an experiment I am trying to rebuild some standard android applications and replace them in system/app on the G1. I have been through all the steps to get the source code and build for the dream platform and have built the various .apk files of interest (e.g. AlarmClock.apk, Browser.apk etc)
To put the files on the device I delete the old .apk and .odex files and copy my newly built .apk file on to the device. However when I try to run the application it crashes with the following message.
The application Alarm Clock (process com.android.alarmclock) has stopped unexpectedly. Please try again.
I know that replacing the applications like this is possible, because the AutoRotating Browser build works fine when copies over in this manner.
I'm using JF1.31 (RC8)
My initial reaction was that I was not signing the applications properly but having read some posts I think the default built .apk should have the right key already in it.
Another theory I have is that perhaps the applications from the head of the source tree are not compatible with the RC8 (or RC30) Android OS releases. Can anyone tell me how to get the source tree which corresponds to this baseline, I've done some reading around but cannot figure it out. I presume I need to do a repo init -u git://android.git.kernel.org/platofrm/manifest.git -b BASELINE but I can't figure out what BASELINE should be.
Many thanks in advance for any help you can give me!!!

There are some branches in android sources:
master
cupcake
release-1.0
Apps from the first two will not run on default G1, you need to reinstall a whole system. I think by default, following google docs you'll get master. So you need to download a release-1.0 sources.
I may be wrong, but that is what I'm see from my experience.

Thanks for that, I'll get the 1.0 branch downloaded and have a go with that.
Cheers for your help!

I was also trying to recompile some of the built-in apps, specifically the browser, but I can't even get it to build. I get a bunch of import errors, stating that it can't find some of the android libraries, such as android.net.http.AndroidHttpClient, android.os.AsyncTask, etc. I've got the android.jar from the SDK in my build path, and it finds some of them, such as android.webkit.URLUtil.
Can anyone shed some light on what I need to do to get it to see the missing libraries? Thanks.

UndeadCretin said:
Thanks for that, I'll get the 1.0 branch downloaded and have a go with that.
Cheers for your help!
Click to expand...
Click to collapse
There are around a dozen build breaks in release-1.0... all of them are due to missing header #includes in various .c and .h files. So, when it doesn't work, don't give up. Fix the breaks and everything will build properly.

Are you resigning the .apk files? Cuz you have to do that for them to work correctly.

Koush said:
There are around a dozen build breaks in release-1.0... all of them are due to missing header #includes in various .c and .h files. So, when it doesn't work, don't give up. Fix the breaks and everything will build properly.
Click to expand...
Click to collapse
Yep I fixed these problems but I have now hit upon the following problem:
(unknown): error 17: Field android.hardware.SensorManager.LIGHT_NO_MOON has changed value from 0.0010f to 0.001f
******************************
You have tried to change the API from what has been previously released in
an SDK. Please fix the errors listed above.
******************************
I've been in and modified SensorManager back to 0.0010f and that let me build get further but I hit the same error again later in the build.
Given that release-1.0 should be a stable branch is it normal to get all these build issues?

Managed to fix the java issue by modifying public_api.xml. Then hit several more C++ problems which I fixed and finally I can build the lot!
Just tried building the AlarmClock application and running on the G1 and it works fine. Thanks everyone for your help!

>Managed to fix the java issue by modifying public_api.xml. Then hit several more C++ problems which I fixed and finally I can build the lot!
Can you write, what did you fix?

^ Agreed, let us know which files need modifying and what needs doing, i've been trying to get my release-1.0 build root working too!
Alternatively, UndeadCretin, could you build the firmware (release-1.0) with a modified framework-res i can send you?

Ok, I managed to compile it without any editing of xml.
Just added stdlib, string, vector headers to dozen of cpp/h.

worry said:
>Managed to fix the java issue by modifying public_api.xml. Then hit several more C++ problems which I fixed and finally I can build the lot!
Can you write, what did you fix?
Click to expand...
Click to collapse
To fix the java issue, I modified frameworks/base/core/java/android/hardware/SensorManager to change the LIGHT_NO_MOON value to 0.0010f (from 0.001f) and in out/target/common/obj/PACKAGING I modified the <field name="LIGHT_NO_MOON" to have value-"0.0010f">
After this there were several other c++ files which were missing relevant includes. I'm afraid I didn't keep a note of these so cannot provide much detail but mostly they were missing one of the following
#include "stdlib.h"
#include "string.h"
#include "stdio.h"
I think one file needed the following include
#include <string>
and there were a couple of other files that needed other includes. The best way to find these is to google for the function name that isn't building and you should be able to find the appropriate include (that's how I did it).
Hope that helps a bit!

were you able to repo sync after adding the local_manifest.xml?

ximonx said:
were you able to repo sync after adding the local_manifest.xml?
Click to expand...
Click to collapse
I did try that previously but it didn't work. I don't think the relevant files for the dream build are available in the release-1.0 branch. This wasn't a problem for me since I'm only interested in building the applications which work fine with the generic build.

I would like to do the same for the mms application. Could you give me the steps or a link how to do it? I mean do I need the whole sources from android platform to do it? How can I just compile one application?

Phlogiston said:
I would like to do the same for the mms application. Could you give me the steps or a link how to do it? I mean do I need the whole sources from android platform to do it? How can I just compile one application?
Click to expand...
Click to collapse
I downloaded the whole Android source (the release-1.0 branch) and compiled the lot. It may be possible to just build the individual application but I do not know how. It is not vital to build for the dream platform if you only care about the applications since they will work fine with the generic build.
So the basic steps to start are:
Get yourself a Linux or Mac OS platform (I use Ubuntu running in VMWare on my XP box).
Follow the instructions here: http://source.android.com/download but when you come to repo init add the flag -b release-1.0
Fix various build problems

When recompiling individual apps to replace system apps is there a way of just building a single application or does the entire thing need making?

ximonx said:
When recompiling individual apps to replace system apps is there a way of just building a single application or does the entire thing need making?
Click to expand...
Click to collapse
My experience is that you have to do the whole thing if you are building from source. There is one way I know of to get around this, which is to use baksmali and smali.

Just to be clear, making the entire thing = build from source root?

ximonx said:
Just to be clear, making the entire thing = build from source root?
Click to expand...
Click to collapse
If you are asking me--yes, that's what I mean. Make sure to build for dream-open as the target (it's generic by default).

[Resolved] Compiling Android USB Gadget drivers and insmod

Hi everyone,
I have been trying for quite a while now to compile HID Gadget support for my galaxy S3 and think i may be getting close. My current problem is that when i run insmod g_hid.ko i receive the following
Code:
insmod: init_module 'g_hid.ko' failed (Device or resource busy)
. The weird thing is that next time the screen turns off or if i attempt to insert it again the phone crashes (kernel panic i assume). I get this same error when i try and insmod g_zero.ko (although this one doesn't kill the phone). Can anyone tell me how i would attempt to debug this? Or even better, has anyone got g_hid working on there phone and if so, what modifications/kernel configs did you use?
Also, when compiling the kernel, in "USB gadget support" if i select any option other than Android Gadget in USB Gadget Drivers my kernel fails to build. Has anyone got any USB gadgets to compile and run and if so what procedure/modifications did you have to make. Currently i am compiling my kernel with android gadget selected as the usb gadget driver and then changing it to HID module and running make module_prepair and make module to get my .ko files (which may be why i am getting the above error).
If anyone can provide some insite and advice it would be greatly appreciated.
Thank you
Adrian

After months of trying, it appears i have got it solved. Unfortunately its not the most graceful solution and currently it disables MTP but i dont care. To fix it i searched every file in the kernel for mentions of CONFIG_USB_G_ANDROID and everywhere it was involved in an #if defined statement i added another section like so
Code:
#if defined(CONFIG_USB_ANDROID) || defined(CONFIG_USB_G_ANDROID) || defined(CONFIG_USB_S3C_OTGD)
In this case i added USB_S3C_OTG. As i said, it isn't pretty but it works.
Hope this helps someone in the future coz there is very little info on the net about this
Enjoy
Adrian

Code sample
I asked a question on Stack Overflow about exactly this issue. There is a small group of people who seem quite interested in this issue, and there really is not a lot of information on the net.
I would very much appreciate if you could explain in a little more detail what steps you took to get this to work for you. You could probably answer the Stack Overflow question more decisively than the current answers too: stack-overflow questions/9805731 - is-it-possible-to-program-android-to-act-as-physical-usb-keyboard. (I can't post the link as this forum blocks it.)
darcpyro said:
After months of trying, it appears i have got it solved. Unfortunately its not the most graceful solution and currently it disables MTP but i dont care. To fix it i searched every file in the kernel for mentions of CONFIG_USB_G_ANDROID and everywhere it was involved in an #if defined statement i added another section like so
Code:
#if defined(CONFIG_USB_ANDROID) || defined(CONFIG_USB_G_ANDROID) || defined(CONFIG_USB_S3C_OTGD)
In this case i added USB_S3C_OTG. As i said, it isn't pretty but it works.
Hope this helps someone in the future coz there is very little info on the net about this
Enjoy
Adrian
Click to expand...
Click to collapse

No worries, im happy to help. The funny thing is by reading the answers you got it started me on the right path to finding a solution. Im not sure how this will translate to other devices firmware but in the case of the galaxy S3 it was i went threw every file and added the
Code:
|| defined(CONFIG_USB_S3C_OTGD)
to all if defined statement that made any mention of CONFIG_USB_GADGET so that the code inside would be compiled. This was so that in the .config file for the kernel CONFIG_USB_GADGET would be set to yes and would then allow the hid driver to be compiled. Basically every time i got an error i saying something wasn't compiled or wasn't declared i went to that file, found where it was declared and added that if defined statement. I would upload my working source code but it would take weeks on my internet connection. If there is a specific file you want to see then let me know and i will post its contents.
Good luck
Adrian

darcpyro said:
No worries, im happy to help. The funny thing is by reading the answers you got it started me on the right path to finding a solution. Im not sure how this will translate to other devices firmware but in the case of the galaxy S3 it was i went threw every file and added the
Code:
|| defined(CONFIG_USB_S3C_OTGD)
to all if defined statement that made any mention of CONFIG_USB_GADGET so that the code inside would be compiled. This was so that in the .config file for the kernel CONFIG_USB_GADGET would be set to yes and would then allow the hid driver to be compiled. Basically every time i got an error i saying something wasn't compiled or wasn't declared i went to that file, found where it was declared and added that if defined statement. I would upload my working source code but it would take weeks on my internet connection. If there is a specific file you want to see then let me know and i will post its contents.
Good luck
Adrian
Click to expand...
Click to collapse
Hi Adrian
did you manage to use your android phone as a keyboard with USB_H_HID module ? I'm trying to do the same thing and I would be interested if you have more informations. For now, I had cyanogendmod source for Samsung Galaxy S III and I've compiled the default 10.1 ROM.
Yesterday, I tried to modify kernel options to include USB_H_HID in the kernel but the build fails on some android.c error.
See you
--
Thus0

Hi Thus0
I did manage to get it to work late last year (Just so you know that it can be done).From memory i basically ran grep over every file in the modules directory looking for CONFIG_USB_GADGET and then edited those files to say something like:
#if defined(CONFIG_USB_ANDROID) || defined(CONFIG_USB_G_ANDROID) || defined(CONFIG_USB_S3C_OTGD)
This way it would compile the USB Gadget modules as well. Let em know if you have any more questions and ill try and help.
Adrian

Huh – is it possible to make apk file for root devices?

Shir_man said:
Huh – is it possible to make apk file for root devices?
Click to expand...
Click to collapse
No sorry, as far as i know a custom kernal is compiled

darcpyro said:
No sorry, as far as i know a custom kernal is compiled
Click to expand...
Click to collapse
Is it possible to only use a driver and to load it with insmod in a stock rom? What modification require a custom kernel?

i was looking for a solution and ou,d something for the nexus 7
https://play.google.com/store/apps/details?id=remote.hid.keyboard.client
open xource here with kernel patch
https://www.google.com/url?q=https:...t&sa=D&usg=AFQjCNGztTd4-U4gHvd4DLzdg_qBxLb7gw
maybe someone can use it

Winfile

https://github.com/Microsoft/winfile was open source recently, so I built it for rt.

It doesnt ran under WinRT8.1 with same error like WinDjView "number 1566 didn't fnd in the library". Rebuilded new one WinDjView with leaked MFC from VS2001beta works well (with some limitation). Please research this bug. I need an 2window file manager for my RT. How can I help ya?

any chance to get the error as shown in the English ui? would be more familiar to me.
I don't think this project uses mfc.
my guess is some linked dll has an ordinal of 1566 which moved in the dll between rt8.0 and rt8.1. if i get some time I'll recreate the libs from custom def and avoid including ordinal values, so the linker will be forced to load by symbol name.
a note to remind myself how to do that:
lib.exe /def:libfile.def /out:libfile.lib

Please look at there its my opened issue
https://github.com/Microsoft/winfile/issues/6
Unfortunately even when I install English UI some parts of UI still in prenstalled language (Russian in my case) ecen when it sets to English and reboot. This error is still in Russain and Im not sure how it sounds correctly in English.

hooddy said:
Please look at there its my opened issue
https://github.com/Microsoft/winfile/issues/6
Unfortunately even when I install English UI some parts of UI still in prenstalled language (Russian in my case) ecen when it sets to English and reboot. This error is still in Russain and Im not sure how it sounds correctly in English.
Click to expand...
Click to collapse
Privet! Yes, thanks, I am in that github thread also.
Please try this one if you have the time and willingness.
TLDR; Made libs without ordinals to force the system to load libraries and symbols by name. Based upon your error message containing a numerical value and not knowing whether the system will fail if incorrect ordinal or will retry by name.
Steps to produce are different from these: http://www.wanderinghuman.com/blog/archives/000123.html
In addition to that workspace, here's the changes:
Linked without each library individually to see which symbols were used
For each library
created a library.def file with the missing symbols obtained from linked errors
used this command: lib /def:library.def /out:library.lib /MACHINE:ARM
If First time;Observed linker error on resulting library about 0x1c0 (MACHINE:ARM) and remembered that's the Windows CE ARM thumb1
Hex edited library.lib to replace 0x1c0 with 0x1c4 (THUMB2) where relevant
end
Adjusted linker properties to hard code my libraries, so it didn't search default libraries.
Linked and produced what is attached here.
Edit, had typo in a def file, so fixing binary after fixing it/

smb_gaiden said:
any chance to get the error as shown in the English ui? would be more familiar to me.
I don't think this project uses mfc.
my guess is some linked dll has an ordinal of 1566 which moved in the dll between rt8.0 and rt8.1. if i get some time I'll recreate the libs from custom def and avoid including ordinal values, so the linker will be forced to load by symbol name.
a note to remind myself how to do that:
lib.exe /def:libfile.def /out:libfile.lib
Click to expand...
Click to collapse
NEVER LINK WITH ORDINALS, JUST NEVER. That made me lose a metric ton of time reporting stuff which was already ported to RT otherwise. It's a very bad idea, it's worse having builds with ordinals than no builds at all; a fixed version is attached by the way.
Why are you doing it by generating .libs? that's not required anymore, just add the ARM desktop SDK in individual components on VSinstaller

black_blob said:
NEVER LINK WITH ORDINALS, JUST NEVER. That made me lose a metric ton of time reporting stuff which was already ported to RT otherwise. It's a very bad idea, it's worse having builds with ordinals than no builds at all; a fixed version is attached by the way
Click to expand...
Click to collapse
Thanks for the build. I miss some dll to run it. Can ya share it please?
And please share your Far manager build. I really miss them.

smb_gaiden said:
Privet! Yes, thanks, I am in that github thread also.
Please try this one if you have the time and willingness.
TLDR; Made libs without ordinals to force the system to load libraries and symbols by name. Based upon your error message containing a numerical value and not knowing whether the system will fail if incorrect ordinal or will retry by name.
Steps to produce are different from these: http://www.wanderinghuman.com/blog/archives/000123.html
In addition to that workspace, here's the changes:
Linked without each library individually to see which symbols were used
For each library
created a library.def file with the missing symbols obtained from linked errors
used this command: lib /def:library.def /out:library.lib /MACHINE:ARM
If First time;Observed linker error on resulting library about 0x1c0 (MACHINE:ARM) and remembered that's the Windows CE ARM thumb1
Hex edited library.lib to replace 0x1c0 with 0x1c4 (THUMB2) where relevant
end
Adjusted linker properties to hard code my libraries, so it didn't search default libraries.
Linked and produced what is attached here.
Edit, had typo in a def file, so fixing binary after fixing it/
Click to expand...
Click to collapse
Privet! Ill try this new build and it's not workd with the same error as early was. What about all you write there I not really undasttod what can I do for you to improve it Im sorry for my stupidility(((

black_blob
can you tell me more about that please?

hooddy said:
Privet! Ill try this new build and it's not workd with the same error as early was. What about all you write there I not really undasttod what can I do for you to improve it Im sorry for my stupidility(((
Click to expand...
Click to collapse
you'll have to install the ucrt compatibility update
smb_gaiden said:
black_blob
can you tell me more about that please?
Click to expand...
Click to collapse
Windows 10 SDKs for VS2017 have the full set of ARM .libs these days. Set the SDK version to 10.0.16299

black_blob said:
you'll have to install the ucrt compatibility update
Click to expand...
Click to collapse
Where I can get it? Some msu file please. I try to obtain this dll form powershell and dot.net at github with no luck.

@black_blob thank you
@hooddy don't try to understand, just some development information.

smb_gaiden said:
@black_blob thank you
@hooddy don't try to understand, just some development information.
Click to expand...
Click to collapse
I thinks it's an https://support.microsoft.com/en-us/help/3118401/update-for-universal-c-runtime-in-windows I didn't get it over WU. Can ya share direct link for KB3118401 and KB2919355 msu for WinRT please?

Maybe this can help to build proper
https://github.com/Microsoft/winfile/issues/102

Well how about working RT build? Nope?

hooddy said:
Well how about working RT build? Nope?
Click to expand...
Click to collapse
Ended up choosing the nuclear option, and that seems to work alright.

black_blob said:
Ended up choosing the nuclear option, and that seems to work alright.
Click to expand...
Click to collapse
Ya make my day! Thank ya really much!

black_blob said:
Ended up choosing the nuclear option, and that seems to work alright.
Click to expand...
Click to collapse
Oh cool, you saved my future self some time! I ordered a surface RT in order to run 8.1 and leave my main one undisturbed. However, still have some work to do elsewhere, so can you elaborate on the nuclear option in order to save my future self more time? Thank you very much!

Static linking for everything + LTCG + removal of all unused code. It ends up working so well that it has the same size as the binary with outside depends.

I got my new (used) surface rt a few days ago, spent 3 days discovering the service pack update black screen error, installing all the service packs, then using the tool provided by Qiangong et. all.
Reproduced the error with my Winfile and realised that there was one critical library that was being linked with ordinals. Drum roll... kernel32.lib due to not using /NODEFAULTLIBS.
Anyway, it has been corrected and attached is a dynamically linked winfile.exe which runs on Windows RT 8.1. It has already been signed.

Need help building a ROM [SOLVED]

So I began with this whole endeavour like a week ago and while I've learnt a couple of things I'm still hitting my head against a wall. The issues seem to be regarding the Kernel I'm using, someone mentioned that I'm supposed to be using the source Kerner of the ROM but isn't the Kernel device specific?
I'm also getting some errors regarding neverallows, a user suggested to ignore them but as far as I understand you should not be doing that.
So any pointers, suggestions or tips?
Here are my sources as well as the errors I'm getting.
Device tree: https://github.com/celepharn/device_xiaomi_vince/tree/havoc-11
Vendor tree: https://github.com/celepharn/vendor_xiaomi_vince/tree/havoc-11
Kernel Source: https://github.com/celepharn/kernel_dark_ages_vince
ROM Source: https://github.com/Havoc-OS
Haste or Dogbin URL (here are the errors as displayed in the terminal): https://del.dog/wiryckoful.txt

look like the compilation stop because of sepolicy usually you just remove and it would compile

ineedroot69 said:
look like the compilation stop because of sepolicy usually you just remove and it would compile
Click to expand...
Click to collapse
Remove what exactly? I'm pretty new to this.

celephrn said:
Remove what exactly? I'm pretty new to this.
Click to expand...
Click to collapse
i uploaded the compile log on pastebin so it's much easier to read
on line 493 its where it stop
also notice on line 499 it say " Please fix the policy "

ineedroot69 said:
i uploaded the compile log on pastebin so it's much easier to read
on line 493 its where it stop
Click to expand...
Click to collapse
Ahh you mean in the domain.te and coredomain.te files? I'm doing that right now, will get back to you when I'm through. Thanks!

ineedroot69 said:
i uploaded the compile log on pastebin so it's much easier to read
on line 493 its where it stop
also notice on line 499 it say " Please fix the policy "
Click to expand...
Click to collapse
Did that, and got error in
#error "Please don't include <linux/compiler-gcc.h> directly, include <linux/compiler.h> instead."
it's a line in the compiler-gcc.h file located in /havoc/kernel/xiaomi/vince/include/linux
Any clues on how to proceed? I'm going to try with clang to see if it works.
Edit: here's the code in said file.

celephrn said:
Did that, and got error in
#error "Please don't include <linux/compiler-gcc.h> directly, include <linux/compiler.h> instead."
it's a line in the compiler-gcc.h file located in /havoc/kernel/xiaomi/vince/include/linux
Any clues on how to proceed? I'm going to try with clang to see if it works.
Edit: here's the code in said file.
Click to expand...
Click to collapse
usually sepolicy are easy to fix since they point out line # of code and the file location causing issue on your case thou look like you have allow and not allow parameter conflict
also i forgot to say .te file are viewable as text so gedit works on them

ineedroot69 said:
usually sepolicy are easy to fix since they point out line # of code and the file location causing issue on your case thou look like you have allow and not allow parameter conflict
also i forgot to say .te file are viewable as text so gedit works on them
Click to expand...
Click to collapse
I did that, then I got some other errors and the suggestion was to switch to clang-proton, which I did. I got to the part of kernel image building only to fail once again.
These are my updated sources:
Device tree: https://github.com/celepharn/device_xiaomi_vince/tree/havoc-11
Vendor tree: https://github.com/celepharn/vendor_xiaomi_vince/tree/havoc-11
Kernel Source: https://github.com/celepharn/kernel_dark_ages_vince
ROM Source: https://github.com/Havoc-OS
Haste or Dogbin URL (here are the errors as displayed in the terminal): https://del.dog/locorfyvas
I'm certain the issue is building the kernel image, the Image.gz file but don't know how else to proceed, is the kernel I'm using outdated or something?
I'm also getting issues with lines of qseecomi.h located in /kernel/xiaomi/vince/include/soc/qcom, for example in line 108 stating I should put packed after "struct" but such instruction is not clear to me, perhaps would it be something like this?
__struct packed qsee_apps_region_info_ireq {
uint32_t qsee_cmd_id;
uint32_t addr;
uint32_t size;
};

Try to hire a Good developer to solve this issue! Good luck buddy!

mehtamanisha169 said:
Try to hire a Good developer to solve this issue! Good luck buddy!
Click to expand...
Click to collapse
I got it working, the build is now complete. Thanks!

Build Android 11 from source

Hi, I'm trying to build Android 11. When I run the command
repo sync
I obtai the error
fatal: remove-project element specifies non-existent project: platform/prebuilts/gcc/darwin-x86/mips/mips64el-linux-android-4.9
What it means? How can I resolve it?
Many thanks

Check this report from GitHub. It doesn't look like a big deal. Maybe missing resources/files that you probably don't need if you are building an OS for a phone.
I see they simple removed a line from local_manifests/untracked.xml of which contains the error.

Hi, many thanks for your reply. I have removed more than 10 lines from the xml files.
Is it a normal behaviour? Why the files are not correct?

denisgottardello said:
Hi, many thanks for your reply. I have removed more than 10 lines from the xml files.
Is it a normal behaviour? Why the files are not correct?
Click to expand...
Click to collapse
I didn't find a direct answer. But I've searched the errors, and found that your error on the OP has a darwin-x86 on it.
So I did Google that, and found that it is an Operating System that has something to do with Mac OS, released in 2000, still active.
So as the errors of other builders. I noticed that they're building an OS for mobile phones, which does not include, for an example, darwin-x86. So I asked myself .. What's the point of including an arch that you won't need?
I think when you're building the OS, you do export the arch you're targeting, right? So did it include darwin-x86?