9.2-STABLE #0 r259072 -> 9.3 buildworld fails

[Volodymyr Kostyrko]

On 05.12.2014 01:26, Joseph Olatt wrote:
> On Thu, Dec 04, 2014 at 10:26:02PM +0200, Volodymyr Kostyrko wrote:
>>>>> Anybody else on the list seeing anything similar?
>>>>>
>>>>
>>>> Me. I have a derelict mother on VIA KT600 chipset with some memory
>>>> problems. Sometimes it faults. Sometimes it tries to kill fs. Most
>>>> times running compilation again works.
>>>>
>>>> Does it always breaks on the same command with the same output?
>>>
>>>
>>> Yes. Ran it twice and both times it stopped at the same place with the
>>> same error.
>>
>> So this has nothing to do with bad memory. Try building
>> WITH_CLANG_IS_CC or even WITHOUT_GCC. Those should be added to the
>> /etc/src.conf:
>>
>> WITH_CLANG_IS_CC=yes
>> WITHOUT_GCC=yes
>>
>> Meanwhile, can you name your processor or list what is in your
>> /etc/make.conf?
>
> I think you are correct here. I wondered why GCC was being invoked.
> Unfortunately, I'm unable to test your suggestion. I've put another hard
> disk in system, installed 10.1-RELEASE and updated the sources to
> 10-STABLE and am currently building world without any issues.

Good luck then. :)

> /etc/make.conf was completely empty (I had checked that).
>
> The processor is: i386 (an old Dell Inspiron 8200! FreeBSD allows
> these old laptops to continue to live and be of service...)

The question is rather about processor type. GCC is famous for 
incorrectly generating code for some old chips assuming SSE2 support for 
ones that can only do SSE (like some Athlons).

-- 
Sphinx of black quartz judge my vow.