[Letux-kernel] LetuxOS: kernel: letux-5.5-rc1

[H. Nikolaus Schaller]

> Am 12.12.2019 um 10:06 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
> 
> 
>> Am 11.12.2019 um 19:00 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
>> 
>> 
>>> Am 11.12.2019 um 17:03 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
>>> 
>>> 
>>>> Am 10.12.2019 um 22:05 schrieb Andreas Kemnade <andreas at kemnade.info>:
>>>> 
>>>> On Tue, 10 Dec 2019 21:51:52 +0100
>>>> "H. Nikolaus Schaller" <hns at goldelico.com> wrote:
>>>> 
>>>>>> Am 10.12.2019 um 20:57 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
>>>>>> 
>>>>>> 
>>>>>>> Am 10.12.2019 um 19:01 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
>>>>>>> 
>>>>>>> Next, I'll try to build with your defconfig.  
>>>>> 
>>>>> I have simply copied it to arch/arm/configs/letux_defconfig. Then it will
>>>>> be taken by my scripts.
>>>>> 
>>>>>> 
>>>>>> Hm:
>>>>>> 
>>>>>> LD [M]  lib/crypto/libdes.o - due to: lib/crypto/des.o
>>>>>> Building modules, stage 2.
>>>>>> MODPOST 436 modules - due to target is PHONY
>>>>>> ERROR: "input_unregister_polled_device" [drivers/iio/industrialio.ko] undefined!
>>>>>> ERROR: "input_register_polled_device" [drivers/iio/industrialio.ko] undefined!
>>>>>> ERROR: "devm_input_allocate_polled_device" [drivers/iio/industrialio.ko] undefined!
>>>>>> make[3]: *** [__modpost] Error 1
>>>>>> make[2]: *** [modules] Error 2
>>>>>> make[1]: *** [__build_one_by_one] Error 2
>>>>>> make: *** [sub-make] Error 2
>>>>>> 
>>>>>> $head->scratch()  
>>>>> 
>>>>> Ok, I have disabled the IIO_INPUT_BRIDGE.
>>>>> 
>>>>> Now I can compile
>>>>> ...
>>>>> ...
>>>>> ...
>>>>> and it boots. Also without display. There is no td028ttec1 panel driver in
>>>>> modprobe -c. Indeed, CONFIG_DRM_PANEL_TPO_TD028TTEC1=m is not in your config.
>>>>> CONFIG_DRM_OMAP_PANEL_TPO_TD028TTEC1=m is but that does no longer exist in
>>>>> v5.5-rc1.
>>>>> 
>>>> well, no need for that on the letux 3704.
>>> 
>>> Indeed... I think the missing panel is a secondary bug.
>>> 
>>> I think I have boiled it down to
>>> 
>>> CONFIG_ARCH_BCM=y or n
>>> 
>>> (or one of the subconfigs enabled by that).
>>> 
>>> Will need a clean test though.
>> 
>> Well, unfortunately this is not sufficient alone.
>> Seems to need more than one config to be disabled :(
>> 
>> Unfortunately I have no good strategy to bisect this.
>> 
>> Maybe I could take the good letux_defconfig that I have
>> found, do a git reset and then by git add -p create
>> small chunks of changes. And then a git bisect on this
>> sequence?
> 
> Yes, that works. And seems to confirm the finding:
> 
> 86b283cbdaf695009ee6d967e7314162839c696f is the first bad commit
> commit 86b283cbdaf695009ee6d967e7314162839c696f
> Author: H. Nikolaus Schaller <hns at goldelico.com>
> Date:   Thu Dec 12 06:43:21 2019 +0100
> 
>    autosplit
> 
> arch/arm/configs/letux_defconfig | 2 ++
> 1 file changed, 2 insertions(+)
> iMac:master hns$ git show 86b283cbdaf695009ee6d967e7314162839c696f
> commit 86b283cbdaf695009ee6d967e7314162839c696f (refs/bisect/bad)
> Author: H. Nikolaus Schaller <hns at goldelico.com>
> Date:   Thu Dec 12 06:43:21 2019 +0100
> 
>    autosplit
> 
> diff --git a/arch/arm/configs/letux_defconfig b/arch/arm/configs/letux_defconfig
> index 4d9c6c029e8b..e75831642f26 100644
> --- a/arch/arm/configs/letux_defconfig
> +++ b/arch/arm/configs/letux_defconfig
> @@ -25,6 +25,8 @@ CONFIG_BLK_DEV_INITRD=y
> CONFIG_CC_OPTIMIZE_FOR_SIZE=y
> CONFIG_SLAB=y
> CONFIG_PROFILING=y
> +CONFIG_ARCH_BCM=y
> +CONFIG_ARCH_BCM2835=y
> CONFIG_ARCH_MXC=y
> CONFIG_POWER_AVS_OMAP=y
> CONFIG_POWER_AVS_OMAP_CLASS3=y
> iMac:master hns$ 
> 
> Well, I did start with a good config and this change makes it a bad one.
> But the good one may have changed other configs.
> 
> I think I should reverse the good/bad tagging and start with original
> letux-5.5-rc1 bad defconfig until it starts to work.
> 
> The basic principle is (well, I had to abort the loop manually):
> 
> GOOD=5aaa71815c46
> BAD=letux-5.5-rc1
> git checkout -B working-branch $GOOD
> ARCH=arm make letux_defconfig savedefconfig
> mv defconfig arch/arm/configs/letux_defconfig
> git commit -m 'good' -a
> git checkout $BAD -- arch/arm/configs/letux_defconfig
> git reset HEAD
> ARCH=arm make letux_defconfig savedefconfig
> mv defconfig arch/arm/configs/letux_defconfig
> 
> while true
> do
> (echo s; echo s; echo s; echo s; echo s; echo y; echo q) | git add -p
> git commit -m 'autosplit' | break
> done

I now did run the bisect in reverse direction and it also says that the
first problematic config is CONFIG_ARCH_BCM && CONFIG_ARCH_BCM2835.

But removing just this from the current letux_defconfig doesn't
seem to help.

Therefore I can only assume that reducing to savedefconfig and expanding
back gives a different result for some default settings.

Well, just

diff --git a/arch/arm/configs/letux_defconfig b/arch/arm/configs/letux_defconfig
index 91d0d6934ca6..c45ae1bae650 100644
--- a/arch/arm/configs/letux_defconfig
+++ b/arch/arm/configs/letux_defconfig
@@ -1,12 +1,13 @@
 #
+# Automatically generated file; DO NOT EDIT.
 # Linux/arm 5.5.0-rc1 Kernel Configuration
 #
 
 #
-# Compiler: arm-linux-gnueabi-gcc (GCC) 4.9.2
+# Compiler: gcc (MacPorts gcc49 4.9.4_5) 4.9.4
 #
 CONFIG_CC_IS_GCC=y
-CONFIG_GCC_VERSION=40902
+CONFIG_GCC_VERSION=40904
 CONFIG_CLANG_VERSION=0
 CONFIG_CC_CAN_LINK=y
 CONFIG_CC_HAS_ASM_GOTO=y
@@ -130,6 +131,7 @@ CONFIG_GENERIC_SCHED_CLOCK=y
 #
 # end of Scheduler features
 
+CONFIG_CC_HAS_INT128=y
 CONFIG_CGROUPS=y
 # CONFIG_MEMCG is not set
 # CONFIG_BLK_CGROUP is not set
iMac:master hns$ 

So I'll try again to disable CONFIG_ARCH_BCM && CONFIG_ARCH_BCM2835.

BR,
Nikolaus