[Courgette] Refactor: Add and use Instruction*Receptor classes; call ParseFile() in 2 passes.

courgette/disassembler_elf_32.h

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef COURGETTE_DISASSEMBLER_ELF_32_H_
 #define COURGETTE_DISASSEMBLER_ELF_32_H_
 
 #include <stddef.h>
 #include <stdint.h>
 
 #include <memory>
 #include <string>
 #include <vector>
 
 #include "base/macros.h"
 #include "courgette/disassembler.h"
 #include "courgette/image_utils.h"
 #include "courgette/memory_allocator.h"
 #include "courgette/types_elf.h"
 
 namespace courgette {
 
 class AssemblyProgram;
+class InstructionReceptor;
 
 // A Courgette disassembler for 32-bit ELF files. This is only a partial
 // implementation that admits subclasses for the architecture-specific parts of
 // 32-bit ELF file processing. Specifically:
 // - RelToRVA() processes entries in ELF relocation table.
 // - ParseRelocationSection() verifies the organization of the ELF relocation
 //   table.
 // - ParseRel32RelocsFromSection() finds branch targets by looking for relative
 //   branch/call opcodes in the particular architecture's machine code.
 class DisassemblerElf32 : public Disassembler {
@@ skipping 12 matching lines @@
 
     void set_relative_target(RVA relative_target) {
       relative_target_ = relative_target;
     }
     void set_file_offset(FileOffset file_offset) { file_offset_ = file_offset; }
 
     // Computes the relative jump's offset from the op in p.
     virtual CheckBool ComputeRelativeTarget(const uint8_t* op_pointer) = 0;
 
     // Emits the assembly instruction corresponding to |label|.
-    virtual CheckBool EmitInstruction(AssemblyProgram* program,
-                                      Label* label) = 0;
+    virtual CheckBool EmitInstruction(Label* label,
+                                      InstructionReceptor* receptor) = 0;
 
     // Returns the size of the instruction containing the RVA.
     virtual uint16_t op_size() const = 0;
 
     // Comparator for sorting, which assumes uniqueness of RVAs.
     static bool IsLessThanByRVA(const std::unique_ptr<TypedRVA>& a,
                                 const std::unique_ptr<TypedRVA>& b) {
       return a->rva() < b->rva();
     }
 
     // Comparator for sorting, which assumes uniqueness of file offsets.
     static bool IsLessThanByFileOffset(const std::unique_ptr<TypedRVA>& a,
                                        const std::unique_ptr<TypedRVA>& b) {
       return a->file_offset() < b->file_offset();
     }
 
-  private:
+   private:
     const RVA rva_;
     RVA relative_target_ = kNoRVA;
     FileOffset file_offset_ = kNoFileOffset;
   };
 
   // Visitor/adaptor to translate RVA to target RVA. This is the ELF
   // counterpart to RvaVisitor_Rel32 that uses TypedRVA.
   class Elf32RvaVisitor_Rel32 :
   public VectorRvaVisitor<std::unique_ptr<TypedRVA>> {
    public:
@@ skipping 10 matching lines @@
 
  public:
   DisassemblerElf32(const uint8_t* start, size_t length);
 
   ~DisassemblerElf32() override { }
 
   // Disassembler interfaces.
   RVA FileOffsetToRVA(FileOffset file_offset) const override;
   FileOffset RVAToFileOffset(RVA rva) const override;
   RVA PointerToTargetRVA(const uint8_t* p) const override;
-  virtual ExecutableType kind() const override = 0;
+  ExecutableType kind() const override = 0;
   bool ParseHeader() override;
   bool Disassemble(AssemblyProgram* target) override;
 
   virtual e_machine_values ElfEM() const = 0;
 
   CheckBool IsValidTargetRVA(RVA rva) const WARN_UNUSED_RESULT;
 
   // Converts an ELF relocation instruction into an RVA.
   virtual CheckBool RelToRVA(Elf32_Rel rel, RVA* result)
     const WARN_UNUSED_RESULT = 0;
@@ skipping 40 matching lines @@
   }
 
   const Elf32_Phdr* ProgramSegmentHeader(Elf32_Half id) const {
     assert(id >= 0 && id < ProgramSegmentHeaderCount());
     return program_header_table_ + id;
   }
 
   // Misc address space helpers
 
   CheckBool RVAsToFileOffsets(const std::vector<RVA>& rvas,
-                              std::vector<FileOffset>* file_offsets);
+                              std::vector<FileOffset>* file_offsets) const;
 
   CheckBool RVAsToFileOffsets(
-      std::vector<std::unique_ptr<TypedRVA>>* typed_rvas);
+      std::vector<std::unique_ptr<TypedRVA>>* typed_rvas) const;
 
   // Parsing code for Disassemble().
 
   virtual CheckBool ParseRelocationSection(const Elf32_Shdr* section_header,
-                                           AssemblyProgram* program)
+                                           InstructionReceptor* receptor) const
       WARN_UNUSED_RESULT = 0;
 
   virtual CheckBool ParseRel32RelocsFromSection(const Elf32_Shdr* section)
       WARN_UNUSED_RESULT = 0;
 
   // Disassembler interfaces.
   RvaVisitor* CreateAbs32TargetRvaVisitor() override;
   RvaVisitor* CreateRel32TargetRvaVisitor() override;
   void RemoveUnusedRel32Locations(AssemblyProgram* program) override;
 
-  CheckBool ParseFile(AssemblyProgram* target) WARN_UNUSED_RESULT;
+  CheckBool ParseFile(AssemblyProgram* target,
+                      InstructionReceptor* receptor) const WARN_UNUSED_RESULT;
 
   CheckBool ParseProgbitsSection(
       const Elf32_Shdr* section_header,
       std::vector<FileOffset>::iterator* current_abs_offset,
       std::vector<FileOffset>::iterator end_abs_offset,
       std::vector<std::unique_ptr<TypedRVA>>::iterator* current_rel,
       std::vector<std::unique_ptr<TypedRVA>>::iterator end_rel,
-      AssemblyProgram* program) WARN_UNUSED_RESULT;
+      AssemblyProgram* program,
+      InstructionReceptor* receptor) const WARN_UNUSED_RESULT;
 
   CheckBool ParseSimpleRegion(FileOffset start_file_offset,
                               FileOffset end_file_offset,
-                              AssemblyProgram* program) WARN_UNUSED_RESULT;
+                              InstructionReceptor* receptor) const
+      WARN_UNUSED_RESULT;
 
   CheckBool ParseAbs32Relocs() WARN_UNUSED_RESULT;
 
   CheckBool CheckSection(RVA rva) WARN_UNUSED_RESULT;
 
   // Extracts all rel32 TypedRVAs. Does not sort the result.
   CheckBool ParseRel32RelocsFromSections() WARN_UNUSED_RESULT;
 
   const Elf32_Ehdr* header_;
 
   Elf32_Half section_header_table_size_;
 
   // Section header table, ordered by section id.
   std::vector<Elf32_Shdr> section_header_table_;
 
   // An ordering of |section_header_table_|, sorted by file offset.
   std::vector<Elf32_Half> section_header_file_offset_order_;
 
   const Elf32_Phdr* program_header_table_;
   Elf32_Half program_header_table_size_;
 
   // Pointer to string table containing section names.
   const char* default_string_section_;
   size_t default_string_section_size_;
 
-  std::vector<RVA> abs32_locations_;
-  std::vector<std::unique_ptr<TypedRVA>> rel32_locations_;
+  // Sorted abs32 and reel32 RVAs. These are mutable because ParseFile() needs
+  // to sort these by file offsets.
+  mutable std::vector<RVA> abs32_locations_;
+  mutable std::vector<std::unique_ptr<TypedRVA>> rel32_locations_;
 
+ private:
   DISALLOW_COPY_AND_ASSIGN(DisassemblerElf32);
 };
 
 }  // namespace courgette
 
 #endif  // COURGETTE_DISASSEMBLER_ELF_32_H_