Re: Clang error make buildworld

On 05/05/11 15:46, Olivier Smedts wrote:
> 2011/5/5 O. Hartmann<ohartman_at_zedat.fu-berlin.de>:
>> On 05/04/11 16:20, Dimitry Andric wrote:
>>>
>>> On 2011-05-04 15:44, Manfred Antar wrote:
>>> ...
>>>>
>>>> src.conf:
>>>>
>>>> WITHOUT_DYNAMICROOT=yes
>>>> WITH_IDEA=yes
>>>> .if !defined(CC) || ${CC} == "cc"
>>>> CC=clang
>>>> .endif
>>>> .if !defined(CXX) || ${CXX} == "c++"
>>>> CXX=clang++
>>>> .endif
>>>> #Don't die on warnings
>>>> NO_WERROR=
>>>> WERROR=
>>>
>>> Aha. Please move the clang-related stuff to make.conf instead, e.g.
>>> this fragment:
>>>
>>> .if !defined(CC) || ${CC} == "cc"
>>> CC=clang
>>> .endif
>>> .if !defined(CXX) || ${CXX} == "c++"
>>> CXX=clang++
>>> .endif
>>> #Don't die on warnings
>>> NO_WERROR=
>>> WERROR=
>>>
>>
>>
>> On a notebook (DELL Latitude E6510) I tried compiling world with CLANG. So
>> far, so good. It worked. But after rebooting I got a strange misbehaviour of
>> the xdm login window (black/white instead of coloured), but this was only
>> some superficial symptome. The whole system seems to be corrupted. Hitting
>> tab key results like hitting exit in the console. The gcc 4.2.1 system
>> compiler isn't capable of producing binaries, see message below. At this
>> very moment, the box isn't usable anymore, I can't even compile a world with
>> cc (see error below, that was generated by trying to compile a kernel and
>> I'm really confused why cc is used instead of clang).
>>
>> Well, the boxes I reported errors from prior to this are desktop systems
>> with nVidia (Fermi based) graphics boards using a driver BLOB 270.XX.XX
>> which is also used by the notebook.
>>
>> The desktop boxes uses C2D based intel chips, the notebook uses a Core-i5
>> based chip. All systems got compiled with option
>>
>> CPUTYPE?=native
>
> Can you try without CPUTYPE "native", or with another value ?
> "native" is not a supported value in /usr/share/mk/bsd.cpu.mk
>
> With gcc I used :
> CPUTYPE?=core2
> CFLAGS=-O2 -pipe -march=native
> NO_CPU_CFLAGS=yes
> COPTFLAGS=-O2 -pipe -march=native
> NO_CPU_COPTFLAGS=yes
>
> So that /usr/share/mk/bsd.cpu.mk could set the right variables and I
> could set my own "-march" value in CFLAGS for gcc.
>
> But now for HEAD (which has a newer gcc and clang) I use :
> CPUTYPE?=core2
> CFLAGS=-O2 -pipe -march=core2
> NO_CPU_CFLAGS=yes
> COPTFLAGS=-O2 -pipe -march=core2
> NO_CPU_COPTFLAGS=yes
>
> Because with clang, -march=native often breaks buildworld, while
> -march=core2 is ok.
>
> First, try to see if you buildworld is still broken with a different
> (or empty!) make.conf.
>

Well I would like to to as suggested, but I can not even build a 
system/kernel anymore. Using clang, the build process dies when it comes 
to "rpcgen" as shown below, it uses "cc" (fixed) and cc doen't work 
properly anymore.


>> I guess the first compilation with CLANG "destroyed" the base' system
>> compiler, at this moment I'm incapable of switching back. Floating like a
>> dead man in the water.
>>
>>
>> Any suggestions?
>>
>> Regards and thanks in advance,
>> Oliver
>> ---
>> awk -f /usr/src/sys/tools/usbdevs2h.awk /usr/src/sys/dev/usb/usbdevs -h
>> awk -f /usr/src/sys/tools/usbdevs2h.awk /usr/src/sys/dev/usb/usbdevs -d
>> rpcgen -hM /usr/src/sys/kgssapi/gssd.x | grep -v pthread.h>  gssd.h
>> cc1: internal compiler error: Bus error: 10
>> Please submit a full bug report,
>> with preprocessed source if appropriate.
>> See<URL:http://gcc.gnu.org/bugs.html>  for instructions.
>> rpcgen -c /usr/src/sys/kgssapi/gssd.x -o gssd_xdr.c
>> cc1: internal compiler error: Bus error: 10
>> Please submit a full bug report,
>> with preprocessed source if appropriate.
>> See<URL:http://gcc.gnu.org/bugs.html>  for instructions.
>> *** Error code 1
>>
>> Stop in /usr/obj/usr/src/sys/MUNIN.
>> *** Error code 1
>>
>> Stop in /usr/src.
>> *** Error code 1
>>
>> Stop in /usr/src.