Extended Description

For this code LLVM generates code that doesn't use tailcall optimization.

struct token
{
char const* tok_start;
char const* tok_end;
int tok_type;
unsigned identifier_hash;
};

token f();

token g()
{
return f();
}

Generated code:

g(): # @​g()
pushq %rbx
movq %rdi, %rbx
callq f()
movq %rbx, %rax
popq %rbx
retq

Expected code:

g(): # @​g()
jmpq f()