[lldb-dev] Question about LLVMCDisassembler

[Tatyana Krasnukha via lldb-dev]

Want to answer my second question: in that case LLVMCDisassembler will be created with m_is_valid = false as it should, but what can happen in create{Target}MCAsmInfo is UB nevertheless. Suppose it is not good, that's why I created revision D41584.

From: lldb-dev [mailto:lldb-dev-bounces at lists.llvm.org] On Behalf Of Tatyana Krasnukha via lldb-dev
Sent: Friday, 22 December, 2017 9:06 PM
To: lldb-dev at lists.llvm.org
Subject: [lldb-dev] Question about LLVMCDisassembler

Hello,

Could someone answer two questions about this code, please
DisassemblerLLVMC::LLVMCDisassembler::LLVMCDisassembler starting from line 897:

  m_reg_info_ap.reset(curr_target->createMCRegInfo(triple));

  m_subtarget_info_ap.reset(
      curr_target->createMCSubtargetInfo(triple, cpu, features_str));

  std::unique_ptr<llvm::MCRegisterInfo> reg_info(
      curr_target->createMCRegInfo(triple));
  m_asm_info_ap.reset(curr_target->createMCAsmInfo(*reg_info, triple));


  1.  Why it cannot reuse m_reg_info_ap  for createMCAsmInfo instead of creation reg_info again?
  2.  It seems that createMCRegInfo can return nullptr. Is it safe to dereference reg_info in createMCAsmInfo (line 904)?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-dev/attachments/20171226/63043941/attachment.html>