Re: 9.0/i386 build failure

On Sat, Nov 05, 2011 at 02:20:03PM +0100, Dimitry Andric wrote:
> On 2011-11-04 20:09, Michael W. Lucas wrote:
> > I suspect I'm building on a system that's too old, but it's worth
> > asking.
> > 
> > FreeBSD eyeball.lodden.com 9.0-CURRENT FreeBSD 9.0-CURRENT #0: Sat Aug 29 00:31:14 EDT 2009     mwlucas_at_stretchlimo.blackhelicopters.org:/usr/obj/usr/src/sys/GENERIC  i386
> > 
> > csup today.  no /etc/src.conf, /etc/make.conf only contains a perl
> > definition.
> > 
> > ...
> > c++ -O2 -pipe -I/usr/src/usr.bin/clang/tblgen/../../../contrib/llvm/include -I/usr/src/usr.bin/clang/tblgen/../../../contrib/llvm/tools/clang/include -I/usr/src/usr.bin/clang/tblgen/../../../contrib/llvm/utils/TableGen -I. -I/usr/src/usr.bin/clang/tblgen/../../../contrib/llvm/../../lib/clang/include -DLLVM_ON_UNIX -DLLVM_ON_FREEBSD -D__STDC_LIMIT_MACROS -D__STDC_CONSTANT_MACROS -DLLVM_HOSTTRIPLE=\"i386-unknown-freebsd9.0\" -I/usr/obj/usr/src/tmp/legacy/usr/include  -static -L/usr/obj/usr/src/tmp/legacy/usr/lib -o tblgen ARMDecoderEmitter.o AsmMatcherEmitter.o AsmWriterEmitter.o AsmWriterInst.o CallingConvEmitter.o CodeEmitterGen.o CodeGenDAGPatterns.o CodeGenInstruction.o CodeGenRegisters.o CodeGenTarget.o DAGISelEmitter.o DAGISelMatcher.o DAGISelMatcherEmitter.o DAGISelMatcherGen.o DAGISelMatcherOpt.o DisassemblerEmitter.o EDEmitter.o FastISelEmitter.o FixedLenDecoderEmitter.o InstrEnumEmitter.o InstrInfoEmitter.o IntrinsicEmitter.o PseudoLoweringEmitter.o RegisterInfoEmi
> tter.o SetTheory.o StringMatcher.o SubtargetEmitter.o TGValueTypes.o TableGen.o X86DisassemblerTables.o X86RecognizableInstr.o /usr/obj/usr/src/tmp/usr/src/usr.bin/clang/tblgen/../../../lib/clang/libllvmtablegen/libllvmtablegen.a /usr/obj/usr/src/tmp/usr/src/usr.bin/clang/tblgen/../../../lib/clang/libllvmsupport/libllvmsupport.a -legacy
> > /usr/obj/usr/src/tmp/usr/src/usr.bin/clang/tblgen/../../../lib/clang/libllvmsupport/libllvmsupport.a(Atomic.o)(.text+0x25): In function `llvm::sys::AtomicIncrement(unsigned int volatile*)':
> > : undefined reference to `__sync_add_and_fetch_4'
> 
> Hm, the only way I can imagine this happening, is then you compile your
> system with -march=i386.  In that case the atomic primitives, such as
> __sync_add_and_fetch, are called as external functions, which would lead
> to a link error like the above.
> 
> I tried building 9-STABLE from a 9-CURRENT box that was not updated
> since April 2011, and from a fairly recent 8-STABLE, but both did not
> exhibit this issue.
> 
> Are you using any special other settings, e.g. in your environment, or
> on the make command line?  In case, it would be handy if you post the
> full buildlog somewhere, so we can see which flags have been used to
> compile the llvmsupport library.
The system gcc was changed to assume march=486 some time ago.
I suppose that the current-9 system is before the change, see r198344.