The reason I am on the stance is because in all actuality there is not one kernel being used that is stock. I believe all kernel's TickerGuy, IssacJ, Mantera, Whyzor and b_randon's kernels are all from merged CAF source code. I know there is lots of issues with all those kernel's and deep sleep and Wifi but the stock kernel deep sleep works fine with Wifi. I propose that if you can upgrade the 100% stock kernel everything should work properly as a lot of the hardware issues have been resolved in the code, and or can be fixed through the code.
The kernel is the spinal cord of the whole OS which I know you know. So in fixing that everything else should come together in my logic. Of course no one ever said my logic is sound
When I say stock I mean the source code that Motorola gave us which is here
https://github.com/MTDEV-KERNEL/MOTO-KERNEL
That repo was taken from kernel download that Motorola gave us located here
Triumph WX435 - Browse /WX435/CAYM_D_0_1010_2039_BP_2037_CDA_1024 at SourceForge.net
The WX_435_Kernel is Whyzor's kernel which was again taken from CAF source code and diffed between our stock Motorola Kernel. My thought process is to roll with the stock kernel and get it working properly with CM9 and then added in changes for the overclock and stuff later. Again since I am no kernel guru this is just my thought process.
The reason I am on the stance is because in all actuality there is not one kernel being used that is stock. I believe all kernel's TickerGuy, IssacJ, Mantera, Whyzor and b_randon's kernels are all from merged CAF source code. I know there is lots of issues with all those kernel's and deep sleep and Wifi but the stock kernel deep sleep works fine with Wifi. I propose that if you can upgrade the 100% stock kernel everything should work properly as a lot of the hardware issues have been resolved in the code, and or can be fixed through the code.
The kernel is the spinal cord of the whole OS which I know you know. So in fixing that everything else should come together in my logic. Of course no one ever said my logic is sound
If you want, you can check out my github, I have a custom stock kernel that has everything already in it, plus updated touch screen drivers. It is the Sharp kernel, but it is our kernel source modified. I don't know what the differences in GB or ICS kernels from Froyo, but I know ueventd was added for GB. Hope this helps.
). dorragaray installed patches in our kernel to stop the ongoing SIGV error's. While this worked fine in his kernel it did not work in ours. I really believe that if we could upgrade our stock kernel it would solve all our issues. Sadly I don't know enough about kernels to solve this and I think if we had a kernel wiz or some one that understands the Android kernel source we might be able to solve the blasted camera.
This is where I am stuck I have not seen any documentation what is different from one kernel rev from Froyo to GB to ICS to JB. I am really not sure what it would take. I would assume updating specific settings for KGSL, PMEM and other features of the kernel are needed but I am not sure where to look for this info.
The problem right now with getting dorragaray involved with this project is there is just to many possible issues on what is wrong. I really truly believe all the code dorragaray gave us for the CameraHAL is complete and stable, I think the issues lies with in the kernel (I will state this till I am blue in the face
I have gone back and forth rewriting this post with different information for the last 2 hours as I learn more. I think I've finally come full circle.
Found a difference in msm_camera.h in libcamera2/camerahal and the kernel. Got excited, but it looks like it isn't being used. It's in native_get_picture() in QualcommCameraHardware.cpp, and it probably won't work if anyone tries to use it
Then I took a closer look at your logcat, and saw the window is NULL message, and got excited again. But, I traced it back through the android source and it looks like that is what they do for QCOM_HARDWARE.
I don't know why I couldn't find some of your logcat messages yesterday, I see them all fine now. But, I realized that the data parameter of QualcommCameraHardware::runFrameThread() is really an msm_frame, which is where the difference is in msm_camera.h. So I'm back to where I was, this is certainly a problem. Here is the diff, I believe the kernel version is correct:
[HIGH]
diff -u ./hardware/qcom/libcamera2/msm_camera.h ./kernel/motorola/triumph/include/media/msm_camera.h
--- ./hardware/qcom/libcamera2/msm_camera.h 2012-11-26 19:35:09.714678299 -0700
+++ ./kernel/motorola/triumph/include/media/msm_camera.h 2012-11-26 19:35:22.370519193 -0700
@@ -584,8 +584,8 @@
int stcam_quality_ind;
uint32_t stcam_conv_value;
- struct ion_allocation_data * ion_alloc;
- struct ion_fd_data * fd_data;
+ struct ion_allocation_data ion_alloc;
+ struct ion_fd_data fd_data;
};
[/HIGH]
