Re: [OE-core] [PATCH] buildtools-extended-tarball: add nativesdk-glibc

["Jeremy Puhlman" <jpuhlman@mvista.com>]

On 5/11/2020 7:30 PM, Kang Kai wrote:
> On 2020/5/10 下午9:40, Richard Purdie wrote:
>> On Wed, 2020-05-06 at 09:54 +0800, kai wrote:
>>> From: Kai Kang <kai.kang@windriver.com>
>>>
>>> It requires gcc 5.0 via OE-Core rev abc741a. On centos 7, the gcc
>>> version is too low then it has to build with buildtools-extended-
>>> tarball
>>> which provides nativesdk-gcc.
>>>
>>> But it fails to build nspr-native:
>>>
>>>> gcc   abstract.o -Xlinker -L../../dist/lib -lplc4 -L../../dist/lib
>>>> -lnspr4 -lpthread -o abstract
>>>> /PATH/TO/x86_64-wrlinuxsdk-linux/bin/ld: /lib64/librt.so.1:
>>>> undefined reference to `__clock_getcpuclockid@GLIBC_PRIVATE'
>>>> /PATH/TO/x86_64-wrlinuxsdk-linux/bin/ld: /lib64/librt.so.1:
>>>> undefined reference to `__clock_nanosleep@GLIBC_PRIVATE'
>>>> /PATH/TO/x86_64-wrlinuxsdk-linux/bin/ld: /lib64/librt.so.1:
>>>> undefined reference to `__clock_settime@GLIBC_PRIVATE'
>>>> /PATH/TO/x86_64-wrlinuxsdk-linux/bin/ld: /lib64/librt.so.1:
>>>> undefined reference to `__clock_getres@GLIBC_PRIVATE'
>>>> collect2: error: ld returned 1 exit status
>>>> make: *** [Makefile:379: abstract] Error 1
>>> Add nativesdk-glibc to buildtools-extended-tarball. And it increases
>>> size of buildtools-extended-tarball about 3K from 48356989 to
>>> 48360329.
>>>
>>> Signed-off-by: Kai Kang <kai.kang@windriver.com>
>>> ---
>>>   meta/recipes-core/meta/buildtools-extended-tarball.bb | 1 +
>>>   1 file changed, 1 insertion(+)
>> This doesn't make sense. Does it mean we need a new uninative version
>> instead?
>>
>> Cheers,
>>
>> Richard
>>
> ```This works in dunfell, as of right now. This should have been fixed 
> with:
>
> ```https://patchwork.openembedded.org/patch/171584/
>
> Hi Jeremy,
>
> I can't receive mails from maillist for now. So reply here.
>
> I suppose when I build oe-core master on ubunu 16.04, it already 
> contains the commit. I'll double  check it.
Yeah it is in master right now. I am not entirely sure why you would be 
seeing it. There was some discussion on
the list about twiddling with the location of /etc/ld.so.conf and 
relocating it in the generation of sdks in general, but
I have not been able to poke at it today. If that went in, it might have 
caused the issue. Basically what I saw when I was
looking in to it, was the sdk ld.so.conf was not getting loaded(in the 
case of the patch because it was looking for etc/etc), but
if it got moved or something similar, it might cause a similar issue.

If you are still seeing the issue on master, strace the link of the file 
with -f and look at what is going on with the load of ld.so.conf and
that might explain what is going on. Also make sure that ld.so.conf is 
still there and has reasonable content.


-- 
Jeremy A. Puhlman
jpuhlman@mvista.com