[hlte][2014/DEC/07][CM11/12] Yank555.lu kernel v1.1c-beta6

[capnkrunch]

I'm about finished setting up my Nexus, rooted, all apps on it ... let's see how it goes

Just a preliminary, are the file sizes 'fine' ? I also use gcc 4.7.

JP.

Send from my Nexus 5 running stock rooted AOSP.

My boot.img is within a couple kb of yours. I've been using the prebuilt 4.7 from AOSP recently but I have the same issue with Linaro 4.7 and I've tried cherry-picking commits from leankernel and/or temasek to allow 4.8 with the same result.

Awesome to hear M8 is stable, I've been considering switching to a TW based rom but no one really wants TW.. Hopefully Carbon's next release is similarly reliable to M8.

[yank555]

Very strange thing ... I use Google's prebuilt 4.7 toolchain for all my kernels (so far).

I don't use linaro personally, had made a blind test compiling the very same kernel with Google's 4.4.3, 4.7 and linaro 4.7.something back then, and labled the kernels A, B, C, and opened a poll.

As long as the test was blind, Google's prebuilt 4.7 was far far far ahead, then come nothing and then came Google's 4.4.3 and shortly followed by Linaro.

Then, mid-time poll, I published which kernel was compiled with which toolchain, so it was no more a blind test, and linaro took up to finish second right behind Google's 4.7, 4.4.3 was left dead.

So much for marketing !

Well I will be heading to test M8 as soon as I'm done transferring all my crap into the Nexus ... just flinging in that SD was so much faster

But if that's the price to pay for peace of mind, I gladly will !

JP.

[Raver099]

I have the same issue as before on M8: reboot when making a call with my (older) Bluetooth car kit. I didn't have any issue yet with other usage, but changed to Temaseks rom now, since the Bluetooth calling was working before on his rom......

To be continued.

[capnkrunch]

So I decided to try building the AOSP kernel for my manta and that boots fine so I think it's probably not an issue with my build environment. So I reconfigured and rebuilt your kernel and wrote stderr to a text file. I also noticed a couple warnings from stdout so I copied those as well.

EDIT: Before you have to ask I also ran build_kernel.sh on the new zImage. The resulting boot.img was also within a few kb of yours and still failed to boot.

My make commands after adding GCC 4.7 to my path were:

Code: Select all

export ARCH=arm
export CROSS_COMPILE=arm-eab-
make Yank555.lu_CM11-light_defconfig VARIANT_DEFCONFIG=Yank555.lu_CM11-light_defconfig 2> config_err.txt
make 2> compile_err.txt

The warnings from compiling were:

Code: Select all

# stderr from compile
defined(@array) is deprecated at kernel/timeconst.pl line 373.
        (Maybe you should just omit the defined()?)
 
# Warnings in stdout from compile
arch/arm/kernel/return_address.c:62:2: warning: #warning "TODO: return_address should use unwind tables" [-Wcpp]
/tmp/cccd2n7W.s: Assembler messages:
/tmp/cccd2n7W.s:167: Warning: swp{b} use is deprecated for this architecture
/tmp/cccd2n7W.s:193: Warning: swp{b} use is deprecated for this architecture

There were lots of warnings for config so they are here: http://pastebin.com/AAS80adc.

EDIT: that's pretty interesting about the different compilers. It'd nice to see someone do a focused study on the various toolchains. Every comparison I've seen runs a couple benchmarks once on both kernels which is ridiculous, considering the benchmarks for android are so inconsisent. The differences they tend to show you could get just by running the benchmark twice on the same system in the same conditions. But who cares about facts when Linaro "feels" snappier. Ridiculous.

[btbamzao]

Damn beta 10 is running beautifully on Will03's slim ROM.

Sent from my SM-N900T using Tapatalk

[yank555]

Damn beta 10 is running beautifully on Will03's slim ROM.

Sent from my SM-N900T using Tapatalk

Crap, with starting my kernel for the Nexus 5 I still haven't come around giving it a test drive

JP.

Send from my Nexus 5 running stock rooted AOSP.

[yank555]

So I decided to try building the AOSP kernel for my manta and that boots fine so I think it's probably not an issue with my build environment. So I reconfigured and rebuilt your kernel and wrote stderr to a text file. I also noticed a couple warnings from stdout so I copied those as well.

EDIT: Before you have to ask I also ran build_kernel.sh on the new zImage. The resulting boot.img was also within a few kb of yours and still failed to boot.

My make commands after adding GCC 4.7 to my path were:

Code: Select all

export ARCH=arm
export CROSS_COMPILE=arm-eab-
make Yank555.lu_CM11-light_defconfig VARIANT_DEFCONFIG=Yank555.lu_CM11-light_defconfig 2> config_err.txt
make 2> compile_err.txt

The warnings from compiling were:

Code: Select all

# stderr from compile
defined(@array) is deprecated at kernel/timeconst.pl line 373.
        (Maybe you should just omit the defined()?)
 
# Warnings in stdout from compile
arch/arm/kernel/return_address.c:62:2: warning: #warning "TODO: return_address should use unwind tables" [-Wcpp]
/tmp/cccd2n7W.s: Assembler messages:
/tmp/cccd2n7W.s:167: Warning: swp{b} use is deprecated for this architecture
/tmp/cccd2n7W.s:193: Warning: swp{b} use is deprecated for this architecture

There were lots of warnings for config so they are here: http://pastebin.com/AAS80adc.

There is no need to make the defconfig, it's a plain flat one in my case.

Try this :

cp arch/arm/config/Yank555.lu_CM11-light_defconfig .config
make clean
./build_kernel.sh

My script takes care of everything, as long as you point to a valid temp folder to build the ramfs.

As for the warning during compilation, those I get as well, those come on full stock sources. If you try this on a 3.0.y (like i9300), the warnings list is pages, so 3.4/y is pretty cleaned up in that regard )

EDIT: that's pretty interesting about the different compilers. It'd nice to see someone do a focused study on the various toolchains. Every comparison I've seen runs a couple benchmarks once on both kernels which is ridiculous, considering the benchmarks for android are so inconsisent. The differences they tend to show you could get just by running the benchmark twice on the same system in the same conditions. But who cares about facts when Linaro "feels" snappier. Ridiculous.

Yes benchmarks are very insonsistent, I consider them at best as a stability test as they produce load. But there's a better app in that regard called stability test

JP.

Send from my Nexus 5 running stock rooted AOSP.

[capnkrunch]

There is no need to make the defconfig, it's a plain flat one in my case.

Try this :

cp arch/arm/config/Yank555.lu_CM11-light_defconfig .config
make clean
./build_kernel.sh

My script takes care of everything, as long as you point to a valid temp folder to build the ramfs.

As for the warning during compilation, those I get as well, those come on full stock sources. If you try this on a 3.0.y (like i9300), the warnings list is pages, so 3.4/y is pretty cleaned up in that regard )

Tried this still failed to boot . On CM's wiki I saw this

However, as of 10.1 CyanogenMod does not build with 64bit host binaries on some systems. You will have to set something like BUILD_HOST_32bit=1 in your environment when building on such 64bit system.

so I tried export BUILD_HOST_32bit=1 before building, still failed. When I get the chance I'm going to install Virtual Box on my brother's or dad's computer and see if I can build it successfully there. In the meantime, any chance you could compile a frandom module for us (https://github.com/Ryuinferno/frandom-android)? Also, how's the N5 been treating you so far?

One more off topic comment, CarbonRom 7-13 seems to be nice and stable so far.

[yank555]

There is no need to make the defconfig, it's a plain flat one in my case.

Try this :

cp arch/arm/config/Yank555.lu_CM11-light_defconfig .config
make clean
./build_kernel.sh

My script takes care of everything, as long as you point to a valid temp folder to build the ramfs.

As for the warning during compilation, those I get as well, those come on full stock sources. If you try this on a 3.0.y (like i9300), the warnings list is pages, so 3.4/y is pretty cleaned up in that regard )

Tried this still failed to boot . On CM's wiki I saw this

However, as of 10.1 CyanogenMod does not build with 64bit host binaries on some systems. You will have to set something like BUILD_HOST_32bit=1 in your environment when building on such 64bit system.

so I tried export BUILD_HOST_32bit=1 before building, still failed. When I get the chance I'm going to install Virtual Box on my brother's or dad's computer and see if I can build it successfully there. In the meantime, any chance you could compile a frandom module for us (https://github.com/Ryuinferno/frandom-android)? Also, how's the N5 been treating you so far?

One more off topic comment, CarbonRom 7-13 seems to be nice and stable so far.

My build-box is a VM which is setup only for the purpose of compiling Android kernels, it's 'easier' that way

Time is a bit of an issue atm Why would you want a frandom module ?? (Not that I really believe in any benefit, my kernels have a fix for normal random generator which keeps it from starving without reducing entropy nor keeping the device awake)

The Nexus, well, it's the main reason my time is limited, trying to finish my basic kernel for it

JP.

Send from my Nexus 5 running stock rooted AOSP on Yank555.lu v1.0-alpha2 kernel.

[herk718]

Hey yank being that your kernel is stable on practically any build, any plans on future updates still for us Note 3 users? Thanks in advance.