[OE-core] Reproducible build problem with BB_NUMBER_THREADS=8

Ulf Samuelsson openembedded-core at emagii.com
Thu Dec 1 23:53:22 UTC 2011


On 2011-12-02 00:08, Khem Raj wrote:
> On (01/12/11 16:59), Ulf Samuelsson wrote:
>> 2011-12-01 15:12, Cliff Brake skrev:
>>> On Thu, Dec 1, 2011 at 7:48 AM, Richard Purdie
>>> <richard.purdie at linuxfoundation.org>   wrote:
>>>
>>>> When you restart the build is the problem persistent or does it work the
>>>> second time?
>>> I set BB_NUMBER_THREADS  = "2" (was 4 when failing)
>>>
>>> I tried to restart the build and cleanall just the failing package,
>>> and it still failed.  I then ran:
>>>
>>> bitbake -c cleanall gcc-cross gcc-cross-initial gcc-cross-intermediate
>>> gcc-runtime
>>>
>>> and after that, everything worked.
>>>
>>>> Does someone have a complete console log for a build that failed with
>>>> this they could share?
>>> I'll restart the build and collect one.  So you just want the output
>>> of the console for bitbake?
>>>
>>> Thanks,
>>> Cliff
>>>
>> Here is my buildlog, creating the failure.
>> The C++ compiler cannot find "string", even though it is in
>> "<sysroot>/usr/include/c++"
>>
> This probably might have something to do with-gxx-include settings
> I will see if I can reproduce it here
I have seen the problem on the following machine
Core i7 980X (hexa core) @ 3.33 GHz
12 GB RAM
Ubuntu 11.10 x64
Build on Intel 510 120 GB SSD disk
PARALLEL_MAKE = "24" | "4"
BB_NUMBER_THREADS = "24" | "4".

I have not seen the problem on the following machine
Core i7 920 @ 2.67 GHz
4 GB RAM.
Ubuntu 11.04 x64
WD Black 1TB SATA-II
PARALLEL_MAKE = "4"
BB_NUMBER_THREADS = "4".


The problem comes in two flavours.
The one in the log, but also when building "tiff".
"tiff" bombs with another C++ include.

Someone else had a similar problem, so I cleaned out the build directory,
built tiff, and then console-image, which completed.

I changed my build script to first build tiff, and then console-image
for the next clean build. This time the c++ include problem reappeared.

I guess there is some race condition somewhere.

BR
Ulf Samuelsson

> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core at lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core


-- 
Best Regards
Ulf Samuelsson
eMagii





More information about the Openembedded-core mailing list