Re: OE-Core status with gcc9

["Yu, Mingli" <mingli.yu@windriver.com>]

On 2019年05月27日 15:53, Khem Raj wrote:
> On Sun, May 26, 2019 at 10:17 AM Khem Raj <raj.khem@gmail.com> wrote:
>>
>> On Mon, May 13, 2019 at 8:39 PM Khem Raj <raj.khem@gmail.com> wrote:
>>>
>>> On Sun, May 12, 2019 at 1:14 PM Khem Raj <raj.khem@gmail.com> wrote:
>>>>
>>>> On Sun, May 12, 2019 at 6:21 AM Richard Purdie
>>>> <richard.purdie@linuxfoundation.org> wrote:
>>>>>
>>>>> I've run some tests and think we're now down to one remaining issue
>>>>> with gcc 9:
>>>>>
>>>>> https://autobuilder.yoctoproject.org/typhoon/#/builders/60/builds/579
>>>>> https://autobuilder.yoctoproject.org/typhoon/#/builders/55/builds/145
>>>>>
>>>>
>>>> Assembler is not liking what gcc9 is generating
>>>>
>>>> HOSTCC scripts/kconfig/conf.o
>>>> /tmp/ccnMW5be.s: Assembler messages:
>>>> /tmp/ccnMW5be.s: Internal error (Segmentation fault).
>>>>
>>>> I wonder if its a common case or specific here?
>>>> I will try to reproduce it as well
>>>>
>>>
>>> I am able to get a .s file which crashes assembler on
>>> qemu. Same file assembles ok with mips cross assembler
>>> so next I want to get hold of a mips hardware and try
>>> it on the real h/w, meanwhile if someone has easy access
>>> to mips and can load core-image-sato-sdk on it then I can
>>> provide the asm file which crashes assembler
>>>
>> got some more time to narrow it down.
>> works with musl+gcc-9
>> works with glibc+clang-8
>> works with musl+Clang-8
>>
>> so only glibc+gcc-9 combo is one which exhibits the issue
>>
>
> I have sent a workaound for this which we can try out.
> https://patchwork.openembedded.org/patch/161674/

Failed to build llvm with gcc9 for qemuppc as below:
| collect2: fatal error: ld terminated with signal 11 [Segmentation 
fault], core dumped

Thanks,

>
>>>>> which is a kernel "make scripts" segfaulting on target on mips and
>>>>> hence failing to build kernel modules. If we can get that one fixed
>>>>> we're good from an OE-Core perspective to move to gcc 9.
>>>>>
>>>>> I have to recognise the work Khem has put in getting us this close!
>>>>>
>>>>> Cheers,
>>>>>
>>>>> Richard
>>>>>