[llvm-dev] Module with a Target-incompatible DataLayout attached
edA-qa mort-ora-y via llvm-dev
producing the error:
leaf: /opt/llvm/lib/CodeGen/MachineFunction.cpp:108:
llvm::MachineFunction::MachineFunction(const llvm::Function*, const
llvm::TargetMachine&, unsigned int, llvm::MachineModuleInfo&): Assertion
`TM.isCompatibleDataLayout(getDataLayout()) && "Can't create a
MachineFunction using a Module with a " "Target-incompatible DataLayout
attached\n"' failed.
I looked at my code and noticed I had a line commented out since it
didn't work in 3.8, but I'm guessing may be the source of the issue:
llvm::legacy::PassManager fpm;
//MIGRATE: fpm.add( new llvm::DataLayout( *data_layout ) );
bool ret = machine->addPassesToEmitFile( fpm, out.os(),
llvm::TargetMachine::CGFT_ObjectFile );
STATE_CHECK( !ret );
fpm.run( *module );
Where `data_layout` was created as:
data_layout.reset( new llvm::DataLayout( module ) );
What do I have to do to avoid the error message? Note that I'm not
trying to get a different data layout, at the moment I just want the
default data layout for the system.
If helpful, here is more context to the object generation code:
std::string triple_name = llvm::sys::getDefaultTargetTriple();
llvm::Triple triple(llvm::Triple::normalize(triple_name));
llvm::Target const * target = llvm::TargetRegistry::lookupTarget(
"x86-64", triple, err_str );
STATE_CHECK( target, err_str );
llvm::TargetOptions topts;
llvm::TargetMachine * machine = target->createTargetMachine(
triple.getTriple(),
"" /*cpu*/, "" /*featurestr*/, topts,
gopts.pic ? llvm::Reloc::PIC_ : llvm::Reloc::Default,
llvm::CodeModel::Default,
llvm::CodeGenOpt::Default );
llvm::tool_output_file out( (base_path + ".o").c_str(),
llvm::sys::fs::OpenFlags::F_None );
llvm::legacy::PassManager fpm;
//MIGRATE: fpm.add( new llvm::DataLayout( *data_layout ) );
bool ret = machine->addPassesToEmitFile( fpm, out.os(),
llvm::TargetMachine::CGFT_ObjectFile );
STATE_CHECK( !ret );
fpm.run( *module );
out.keep();
--
edA-qa mort-ora-y
http://mortoray.com/
Leaf - the language we always wanted
http://leaflang.org/
_______________________________________________
LLVM Developers mailing list
llvm...@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
Mehdi Amini via llvm-dev
--
Mehdi