hexrays.hpp

Go to the documentation of this file.

/*!
 *      Hex-Rays Decompiler project
 *      Copyright (c) 1990-2023 Hex-Rays
 *      ALL RIGHTS RESERVED.
 * \mainpage
 *      There are 2 representations of the binary code in the decompiler:
 *        - microcode: processor instructions are translated into it and then
 *                     the decompiler optimizes and transforms it
 *        - ctree:     ctree is built from the optimized microcode and represents
 *                     AST-like tree with C statements and expressions. It can
 *                     be printed as C code.
 *
 *      Microcode is represented by the following classes:
 *       - mba_t keeps general info about the decompiled code and
 *                     array of basic blocks. usually mba_t is named 'mba'
 *       - mblock_t    a basic block. includes list of instructions
 *       - minsn_t     an instruction. contains 3 operands: left, right, and
 *                     destination
 *       - mop_t       an operand. depending on its type may hold various info
 *                     like a number, register, stack variable, etc.
 *       - mlist_t     list of memory or register locations; can hold vast areas
 *                     of memory and multiple registers. this class is used
 *                     very extensively in the decompiler. it may represent
 *                     list of locations accessed by an instruction or even
 *                     an entire basic block. it is also used as argument of
 *                     many functions. for example, there is a function
 *                     that searches for an instruction that refers to a mlist_t.
 
 *      See https://www.hex-rays.com/blog/microcode-in-pictures for some pictures.
 *
 *      Ctree is represented by:
 *       - cfunc_t     keeps general info about the decompiled code, including a
 *                     pointer to mba_t. deleting cfunc_t will delete
 *                     mba_t too (however, decompiler returns cfuncptr_t,
 *                     which is a reference counting object and deletes the
 *                     underlying function as soon as all references to it go
 *                     out of scope). cfunc_t has 'body', which represents the
 *                     decompiled function body as cinsn_t.
 *       - cinsn_t     a C statement. can be a compound statement or any other
 *                     legal C statements (like if, for, while, return,
 *                     expression-statement, etc). depending on the statement
 *                     type has pointers to additional info. for example, the
 *                     'if' statement has poiner to cif_t, which holds the
 *                     'if' condition, 'then' branch, and optionally 'else'
 *                     branch. Please note that despite of the name cinsn_t
 *                     we say "statements", not "instructions". For us
 *                     instructions are part of microcode, not ctree.
 *       - cexpr_t     a C expression. is used as part of a C statement, when
 *                     necessary. cexpr_t has 'type' field, which keeps the
 *                     expression type.
 *       - citem_t     a base class for cinsn_t and cexpr_t, holds common info
 *                     like the address, label, and opcode.
 *       - cnumber_t   a constant 64-bit number. in addition to its value also
 *                     holds information how to represent it: decimal, hex, or
 *                     as a symbolic constant (enum member). please note that
 *                     numbers are represented by another class (mnumber_t)
 *                     in microcode.
 
 *      See https://www.hex-rays.com/blog/hex-rays-decompiler-primer
 *      for some pictures and more details.
 *
 *      Both microcode and ctree use the following class:
 *       - lvar_t      a local variable. may represent a stack or register
 *                     variable. a variable has a name, type, location, etc.
 *                     the list of variables is stored in mba->vars.
 *       - lvar_locator_t holds a variable location (vdloc_t) and its definition
 *                     address.
 *       - vdloc_t     describes a variable location, like a register number,
 *                     a stack offset, or, in complex cases, can be a mix of
 *                     register and stack locations. very similar to argloc_t,
 *                     which is used in ida. the differences between argloc_t
 *                     and vdloc_t are:
 *                       - vdloc_t never uses ARGLOC_REG2
 *                       - vdloc_t uses micro register numbers instead of
 *                         processor register numbers
 *                       - the stack offsets are never negative in vdloc_t, while
 *                         in argloc_t there can be negative offsets
 *
 *      The above are the most important classes in this header file. There are
 *      many auxiliary classes, please see their definitions in the header file.
 *
 *      See also the description of \ref vmpage.
 *
 */
 
#ifndef __HEXRAYS_HPP
#define __HEXRAYS_HPP
 
#include <pro.h>
#include <fpro.h>
#include <ida.hpp>
#include <idp.hpp>
#include <gdl.hpp>
#include <ieee.h>
#include <loader.hpp>
#include <kernwin.hpp>
#include <typeinf.hpp>
#include <deque>
#include <queue>
 
/*!
 * \page vmpage Virtual Machine used by Microcode
 *      We can imagine a virtual micro machine that executes microcode.
 *      This virtual micro machine has many registers.
 *      Each register is 8 bits wide. During translation of processor
 *      instructions into microcode, multibyte processor registers are mapped
 *      to adjacent microregisters. Processor condition codes are also
 *      represented by microregisters. The microregisters are grouped
 *      into following groups:
 *       - 0..7: condition codes
 *       - 8..n: all processor registers (including fpu registers, if necessary)
 *               this range may also include temporary registers used during
 *               the initial microcode generation
 *       - n.. : so called kernel registers; they are used during optimization
 *               see is_kreg()
 *
 *      Each micro-instruction (minsn_t) has zero to three operands.
 *      Some of the possible operands types are:
 *        - immediate value
 *        - register
 *        - memory reference
 *        - result of another micro-instruction
 *
 *      The operands (mop_t) are l (left), r (right), d (destination).
 *      An example of a microinstruction:
 *
 *              add r0.4, #8.4, r2.4
 *
 *      which means 'add constant 8 to r0 and place the result into r2'.
 *      where
 *       - the left operand is 'r0', its size is 4 bytes (r0.4)
 *       - the right operand is a constant '8', its size is 4 bytes (#8.4)
 *       - the destination operand is 'r2', its size is 4 bytes (r2.4)
 *      Note that 'd' is almost always the destination but there are exceptions.
 *      See mcode_modifies_d(). For example, stx does not modify 'd'.
 *      See the opcode map below for the list of microinstructions and their
 *      operands. Most instructions are very simple and do not need
 *      detailed explanations. There are no side effects in microinstructions.
 *
 *      Each operand has a size specifier. The following sizes can be used in
 *      practically all contexts: 1, 2, 4, 8, 16 bytes. Floating types may have
 *      other sizes. Functions may return objects of arbitrary size, as well as
 *      operations upon UDT's (user-defined types, i.e. are structs and unions).
 *
 *      Memory is considered to consist of several segments.
 *      A memory reference is made using a (selector, offset) pair.
 *      A selector is always 2 bytes long. An offset can be 4 or 8 bytes long,
 *      depending on the bitness of the target processor.
 *      Currently the selectors are not used very much. The decompiler tries to
 *      resolve (selector, offset) pairs into direct memory references at each
 *      opportunity and then operates on mop_v operands. In other words,
 *      while the decompiler can handle segmented memory models, internally
 *      it still uses simple linear addresses.
 *
 *      The following memory regions are recognized:
 *        - GLBLOW   global memory: low part, everything below the stack
 *        - LVARS    stack: local variables
 *        - RETADDR  stack: return address
 *        - SHADOW   stack: shadow arguments
 *        - ARGS     stack: regular stack arguments
 *        - GLBHIGH  global memory: high part, everything above the stack
 *      Any stack region may be empty. Objects residing in one memory region
 *      are considered to be completely distinct from objects in other regions.
 *      We allocate the stack frame in some memory region, which is not
 *      allocated for any purposes in IDA. This permits us to use linear addresses
 *      for all memory references, including the stack frame.
 *
 *      If the operand size is bigger than 1 then the register
 *      operand references a block of registers. For example:
 *
 *              ldc   #1.4, r8.4
 *
 *      loads the constant 1 to registers 8, 9, 10, 11:
 *
 *               #1  ->  r8
 *               #0  ->  r9
 *               #0  ->  r10
 *               #0  ->  r11
 *
 *      This example uses little-endian byte ordering.
 *      Big-endian byte ordering is supported too. Registers are always little-
 *      endian, regardless of the memory endianness.
 *
 *      Each instruction has 'next' and 'prev' fields that are used to form
 *      a doubly linked list. Such lists are present for each basic block (mblock_t).
 *      Basic blocks have other attributes, including:
 *        - dead_at_start: list of dead locations at the block start
 *        - maybuse:  list of locations the block may use
 *        - maybdef:  list of locations the block may define (or spoil)
 *        - mustbuse: list of locations the block will certainly use
 *        - mustbdef: list of locations the block will certainly define
 *        - dnu:      list of locations the block will certainly define
 *                    but will not use (registers or non-aliasable stkack vars)
 *
 *      These lists are represented by the mlist_t class. It consists of 2 parts:
 *        - rlist_t: list of microregisters (possibly including virtual stack locations)
 *        - ivlset_t: list of memory locations represented as intervals
 *                    we use linear addresses in this list.
 *      The mlist_t class is used quite often. For example, to find what an operand
 *      can spoil, we build its 'maybe-use' list. Then we can find out if this list
 *      is accessed using the is_accessed() or is_accessed_globally() functions.
 *
 *      All basic blocks of the decompiled function constitute an array called
 *      mba_t (array of microblocks). This is a huge class that has too
 *      many fields to describe here (some of the fields are not visible in the sdk)
 *      The most importants ones are:
 *        - stack frame: frregs, stacksize, etc
 *        - memory: aliased, restricted, and other ranges
 *        - type: type of the current function, its arguments (argidx) and
 *                local variables (vars)
 *        - natural: array of pointers to basic blocks. the basic blocks
 *                   are also accessible as a doubly linked list starting from 'blocks'.
 *        - bg: control flow graph. the graph gives access to the use-def
 *                   chains that describe data dependencies between basic blocks
 *
 *   Facilities for debugging decompiler plugins:
 *      Many decompiler objects have a member function named dstr().
 *      These functions create a text representation of the object and return
 *      a pointer to it. They are very convenient to use in a debugger instead of
 *      inspecting class fields manually. The mba_t object does not have the
 *      dstr() function because its text representation very long. Instead, we
 *      provide the mba_t::dump_mba() and mba_t::dump() functions.
 *
 *      To ensure that your plugin manipulates the microcode in a correct way,
 *      please call mba_t::verify() before returning control to the decompiler.
 *
 */
 
#ifdef __NT__
#pragma warning(push)
#pragma warning(disable:4062) // enumerator 'x' in switch of enum 'y' is not handled
#pragma warning(disable:4265) // virtual functions without virtual destructor
#endif
 
#define hexapi                ///< Public functions are marked with this keyword
 
// Warning suppressions for PVS Studio:
//-V:2:654 The condition '2' of loop is always true.
//-V::719  The switch statement does not cover all values
//-V:verify:678
//-V:chain_keeper_t:690 copy ctr will be generated
//-V:add_block:656 call to the same function
//-V:add:792 The 'add' function located to the right of the operator '|' will be called regardless of the value of the left operand
//-V:sub:792 The 'sub' function located to the right of the operator '|' will be called regardless of the value of the left operand
//-V:intersect:792 The 'intersect' function located to the right of the operator '|' will be called regardless of the value of the left operand
// Lint suppressions:
//lint -sem(mop_t::_make_cases, custodial(1))
//lint -sem(mop_t::_make_pair, custodial(1))
//lint -sem(mop_t::_make_callinfo, custodial(1))
//lint -sem(mop_t::_make_insn, custodial(1))
//lint -sem(mop_t::make_insn, custodial(1))
 
// Microcode level forward definitions:
class mop_t;            // microinstruction operand
class mop_pair_t;       // pair of operands.      example, :(edx.4,eax.4).8
class mop_addr_t;       // address of an operand. example: &global_var
class mcallinfo_t;      // function call info.    example: <cdecl:"int x" #10.4>.8
class mcases_t;         // jump table cases.      example: {0 => 12, 1 => 13}
class minsn_t;          // microinstruction
class mblock_t;         // basic block
class mba_t;            // array of blocks, represents microcode for a function
class codegen_t;        // helper class to generate the initial microcode
class mbl_graph_t;      // control graph of microcode
struct vdui_t;          // widget representing the pseudocode window
struct hexrays_failure_t; // decompilation failure object, is thrown by exceptions
struct mba_stats_t;     // statistics about decompilation of a function
struct mlist_t;         // list of memory and register locations
struct voff_t;          // value offset (microregister number or stack offset)
typedef std::set<voff_t> voff_set_t;
struct vivl_t;          // value interval (register or stack range)
typedef int mreg_t;     ///< Micro register
 
// Ctree level forward definitions:
struct cfunc_t;         // result of decompilation, the highest level object
struct citem_t;         // base class for cexpr_t and cinsn_t
struct cexpr_t;         // C expression
struct cinsn_t;         // C statement
struct cblock_t;        // C statement block (sequence of statements)
struct cswitch_t;       // C switch statement
struct carg_t;          // call argument
struct carglist_t;      // vector of call arguments
 
typedef std::set<ea_t> easet_t;
typedef std::set<minsn_t *> minsn_ptr_set_t;
typedef std::set<qstring> strings_t;
typedef qvector<minsn_t*> minsnptrs_t;
typedef qvector<mop_t*> mopptrs_t;
typedef qvector<mop_t> mopvec_t;
typedef qvector<uint64> uint64vec_t;
typedef qvector<mreg_t> mregvec_t;
typedef qrefcnt_t<cfunc_t> cfuncptr_t;
 
// Function frames must be smaller than this value, otherwise
// the decompiler will bail out with MERR_HUGESTACK
#define MAX_SUPPORTED_STACK_SIZE 0x100000 // 1MB
 
//-------------------------------------------------------------------------
// Original version of macro DEFINE_MEMORY_ALLOCATION_FUNCS
// (uses decompiler-specific memory allocation functions)
#define HEXRAYS_PLACEMENT_DELETE void operator delete(void *, void *) {}
#define HEXRAYS_MEMORY_ALLOCATION_FUNCS()                          \
  void *operator new  (size_t _s) { return hexrays_alloc(_s); }    \
  void *operator new[](size_t _s) { return hexrays_alloc(_s); }    \
  void *operator new(size_t /*size*/, void *_v) { return _v; }     \
  void operator delete  (void *_blk) { hexrays_free(_blk); }       \
  void operator delete[](void *_blk) { hexrays_free(_blk); }       \
  HEXRAYS_PLACEMENT_DELETE
 
void *hexapi hexrays_alloc(size_t size);
void hexapi  hexrays_free(void *ptr);
 
typedef uint64 uvlr_t;
typedef int64 svlr_t;
enum { MAX_VLR_SIZE = sizeof(uvlr_t) };
const uvlr_t MAX_VALUE = uvlr_t(-1);
const svlr_t MAX_SVALUE = svlr_t(uvlr_t(-1) >> 1);
const svlr_t MIN_SVALUE = ~MAX_SVALUE;
 
enum cmpop_t
{ // the order of comparisons is the same as in microcode opcodes
  CMP_NZ,
  CMP_Z,
  CMP_AE,
  CMP_B,
  CMP_A,
  CMP_BE,
  CMP_GT,
  CMP_GE,
  CMP_LT,
  CMP_LE,
};
 
//-------------------------------------------------------------------------
// value-range class to keep possible operand value(s).
class valrng_t
{
protected:
  int flags;
#define VLR_TYPE 0x0F     // valrng_t type
#define   VLR_NONE   0x00 //   no values
#define   VLR_ALL    0x01 //   all values
#define   VLR_IVLS   0x02 //   union of disjoint intervals
#define   VLR_RANGE  0x03 //   strided range
#define   VLR_SRANGE 0x04 //   strided range with signed bound
#define   VLR_BITS   0x05 //   known bits
#define   VLR_SECT   0x06 //   intersection of sub-ranges
                          //   each sub-range should be simple or union
#define   VLR_UNION  0x07 //   union of sub-ranges
                          //   each sub-range should be simple or
                          //   intersection
#define   VLR_UNK    0x08 //   unknown value (like 'null' in SQL)
  int size;               // operand size: 1..8 bytes
                          // all values must fall within the size
  union
  {
    struct                // VLR_RANGE/VLR_SRANGE
    {                     // values that are between VALUE and LIMIT
                          // and conform to: value+stride*N
      uvlr_t value;       // initial value
      uvlr_t limit;       // final value
                          // we adjust LIMIT to be on the STRIDE lattice
      svlr_t stride;      // stride between values
    };
    struct                // VLR_BITS
    {
      uvlr_t zeroes;      // bits known to be clear
      uvlr_t ones;        // bits known to be set
    };
    char reserved[sizeof(qvector<int>)];
                          // VLR_IVLS/VLR_SECT/VLR_UNION
  };
  void hexapi clear(void);
  void hexapi copy(const valrng_t &r);
  valrng_t &hexapi assign(const valrng_t &r);
 
public:
  explicit valrng_t(int size_ = MAX_VLR_SIZE)
    : flags(VLR_NONE), size(size_), value(0), limit(0), stride(0) {}
  valrng_t(const valrng_t &r) { copy(r); }
  ~valrng_t(void) { clear(); }
  valrng_t &operator=(const valrng_t &r) { return assign(r); }
  void swap(valrng_t &r) { qswap(*this, r); }
  DECLARE_COMPARISONS(valrng_t);
  DEFINE_MEMORY_ALLOCATION_FUNCS()
 
  void set_none(void) { clear(); }
  void set_all(void) { clear(); flags = VLR_ALL; }
  void set_unk(void) { clear(); flags = VLR_UNK; }
  void hexapi set_eq(uvlr_t v);
  void hexapi set_cmp(cmpop_t cmp, uvlr_t _value);
 
  // reduce size
  // it takes the low part of size NEW_SIZE
  // it returns "true" if size is changed successfully.
  // e.g.: valrng_t vr(2); vr.set_eq(0x1234);
  //       vr.reduce_size(1);
  //       uvlr_t v; vr.cvt_to_single_value(&v);
  //       assert(v == 0x34);
  bool hexapi reduce_size(int new_size);
 
  // Perform intersection or union or inversion.
  // \return did we change something in THIS?
  bool hexapi intersect_with(const valrng_t &r);
  bool hexapi unite_with(const valrng_t &r);
  void hexapi inverse(); // works for VLR_IVLS only
 
  bool empty(void) const { return flags == VLR_NONE; }
  bool all_values(void) const { return flags == VLR_ALL; }
  bool is_unknown(void) const { return flags == VLR_UNK; }
  bool hexapi has(uvlr_t v) const;
 
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr(void) const;
 
  bool hexapi cvt_to_single_value(uvlr_t *v) const;
  bool hexapi cvt_to_cmp(cmpop_t *cmp, uvlr_t *val, bool strict) const;
 
  int get_size() const { return size; }
  static uvlr_t max_value(int size_)
  {
    return size_ == MAX_VLR_SIZE
         ? MAX_VALUE
         : (uvlr_t(1) << (size_ * 8)) - 1;
  }
  static uvlr_t min_svalue(int size_)
  {
    return size_ == MAX_VLR_SIZE
         ? MIN_SVALUE
         : (uvlr_t(1) << (size_ * 8 - 1));
  }
  static uvlr_t max_svalue(int size_)
  {
    return size_ == MAX_VLR_SIZE
         ? MAX_SVALUE
         : (uvlr_t(1) << (size_ * 8 - 1)) - 1;
  }
  uvlr_t max_value()  const { return max_value(size);  }
  uvlr_t min_svalue() const { return min_svalue(size); }
  uvlr_t max_svalue() const { return max_svalue(size); }
};
DECLARE_TYPE_AS_MOVABLE(valrng_t);
 
//-------------------------------------------------------------------------
// Are we looking for 'must access' or 'may access' information?
// 'must access' means that the code will always access the specified location(s)
// 'may access' means that the code may in some cases access the specified location(s)
// Example:     ldx cs.2, r0.4, r1.4
//      MUST_ACCESS: r0.4 and r1.4, usually displayed as r0.8 because r0 and r1 are adjacent
//      MAY_ACCESS: r0.4 and r1.4, and all aliasable memory, because
//                  ldx may access any part of the aliasable memory
typedef int maymust_t;
const maymust_t
  // One of the following two bits should be specified:
  MUST_ACCESS = 0x00, // access information we can count on
  MAY_ACCESS  = 0x01, // access information we should take into account
  // Optionally combined with the following bits:
  MAYMUST_ACCESS_MASK = 0x01,
 
  ONE_ACCESS_TYPE = 0x20,      // for find_first_use():
                               // use only the specified maymust access type
                               // (by default it inverts the access type for def-lists)
  INCLUDE_SPOILED_REGS = 0x40, // for build_def_list() with MUST_ACCESS:
                               // include spoiled registers in the list
  EXCLUDE_PASS_REGS = 0x80,    // for build_def_list() with MAY_ACCESS:
                               // exclude pass_regs from the list
  FULL_XDSU = 0x100,           // for build_def_list():
                               // if xds/xdu source and targets are the same
                               // treat it as if xdsu redefines the entire destination
  WITH_ASSERTS = 0x200,        // for find_first_use():
                               // do not ignore assertions
  EXCLUDE_VOLATILE = 0x400,    // for build_def_list():
                               // exclude volatile memory from the list
  INCLUDE_UNUSED_SRC = 0x800,  // for build_use_list():
                               // do not exclude unused source bytes for m_and/m_or insns
  INCLUDE_DEAD_RETREGS = 0x1000, // for build_def_list():
                               // include dead returned registers in the list
  INCLUDE_RESTRICTED = 0x2000,// for MAY_ACCESS: include restricted memory
  CALL_SPOILS_ONLY_ARGS = 0x4000;// for build_def_list() & MAY_ACCESS:
                               // do not include global memory into the
                               // spoiled list of a call
 
inline THREAD_SAFE bool is_may_access(maymust_t maymust)
{
  return (maymust & MAYMUST_ACCESS_MASK) != MUST_ACCESS;
}
 
//-------------------------------------------------------------------------
/// \defgroup MERR_ Microcode error codes
//@{
enum merror_t
{
  MERR_OK        = 0,   ///< ok
  MERR_BLOCK     = 1,   ///< no error, switch to new block
  MERR_INTERR    = -1,  ///< internal error
  MERR_INSN      = -2,  ///< cannot convert to microcode
  MERR_MEM       = -3,  ///< not enough memory
  MERR_BADBLK    = -4,  ///< bad block found
  MERR_BADSP     = -5,  ///< positive sp value has been found
  MERR_PROLOG    = -6,  ///< prolog analysis failed
  MERR_SWITCH    = -7,  ///< wrong switch idiom
  MERR_EXCEPTION = -8,  ///< exception analysis failed
  MERR_HUGESTACK = -9,  ///< stack frame is too big
  MERR_LVARS     = -10, ///< local variable allocation failed
  MERR_BITNESS   = -11, ///< 16-bit functions cannot be decompiled
  MERR_BADCALL   = -12, ///< could not determine call arguments
  MERR_BADFRAME  = -13, ///< function frame is wrong
  MERR_UNKTYPE   = -14, ///< undefined type %s (currently unused error code)
  MERR_BADIDB    = -15, ///< inconsistent database information
  MERR_SIZEOF    = -16, ///< wrong basic type sizes in compiler settings
  MERR_REDO      = -17, ///< redecompilation has been requested
  MERR_CANCELED  = -18, ///< decompilation has been cancelled
  MERR_RECDEPTH  = -19, ///< max recursion depth reached during lvar allocation
  MERR_OVERLAP   = -20, ///< variables would overlap: %s
  MERR_PARTINIT  = -21, ///< partially initialized variable %s
  MERR_COMPLEX   = -22, ///< too complex function
  MERR_LICENSE   = -23, ///< no license available
  MERR_ONLY32    = -24, ///< only 32-bit functions can be decompiled for the current database
  MERR_ONLY64    = -25, ///< only 64-bit functions can be decompiled for the current database
  MERR_BUSY      = -26, ///< already decompiling a function
  MERR_FARPTR    = -27, ///< far memory model is supported only for pc
  MERR_EXTERN    = -28, ///< special segments cannot be decompiled
  MERR_FUNCSIZE  = -29, ///< too big function
  MERR_BADRANGES = -30, ///< bad input ranges
  MERR_BADARCH   = -31, ///< current architecture is not supported
  MERR_DSLOT     = -32, ///< bad instruction in the delay slot
  MERR_STOP      = -33, ///< no error, stop the analysis
  MERR_CLOUD     = -34, ///< cloud: %s
  MERR_MAX_ERR   = 34,
  MERR_LOOP      = -35, ///< internal code: redo last loop (never reported)
};
//@}
 
/// Get textual description of an error code
/// \param out  the output buffer for the error description
/// \param code \ref MERR_
/// \param mba  the microcode array
/// \return the error address
 
ea_t hexapi get_merror_desc(qstring *out, merror_t code, mba_t *mba);
 
//-------------------------------------------------------------------------
// List of microinstruction opcodes.
// The order of setX and jX insns is important, it is used in the code.
 
// Instructions marked with *F may have the FPINSN bit set and operate on fp values
// Instructions marked with +F must have the FPINSN bit set. They always operate on fp values
// Other instructions do not operate on fp values.
 
enum mcode_t
{
  m_nop    = 0x00, // nop                       // no operation
  m_stx    = 0x01, // stx  l,    {r=sel, d=off} // store register to memory     *F
  m_ldx    = 0x02, // ldx  {l=sel,r=off}, d     // load register from memory    *F
  m_ldc    = 0x03, // ldc  l=const,     d       // load constant
  m_mov    = 0x04, // mov  l,           d       // move                         *F
  m_neg    = 0x05, // neg  l,           d       // negate
  m_lnot   = 0x06, // lnot l,           d       // logical not
  m_bnot   = 0x07, // bnot l,           d       // bitwise not
  m_xds    = 0x08, // xds  l,           d       // extend (signed)
  m_xdu    = 0x09, // xdu  l,           d       // extend (unsigned)
  m_low    = 0x0A, // low  l,           d       // take low part
  m_high   = 0x0B, // high l,           d       // take high part
  m_add    = 0x0C, // add  l,   r,      d       // l + r -> dst
  m_sub    = 0x0D, // sub  l,   r,      d       // l - r -> dst
  m_mul    = 0x0E, // mul  l,   r,      d       // l * r -> dst
  m_udiv   = 0x0F, // udiv l,   r,      d       // l / r -> dst
  m_sdiv   = 0x10, // sdiv l,   r,      d       // l / r -> dst
  m_umod   = 0x11, // umod l,   r,      d       // l % r -> dst
  m_smod   = 0x12, // smod l,   r,      d       // l % r -> dst
  m_or     = 0x13, // or   l,   r,      d       // bitwise or
  m_and    = 0x14, // and  l,   r,      d       // bitwise and
  m_xor    = 0x15, // xor  l,   r,      d       // bitwise xor
  m_shl    = 0x16, // shl  l,   r,      d       // shift logical left
  m_shr    = 0x17, // shr  l,   r,      d       // shift logical right
  m_sar    = 0x18, // sar  l,   r,      d       // shift arithmetic right
  m_cfadd  = 0x19, // cfadd l,  r,    d=carry   // calculate carry    bit of (l+r)
  m_ofadd  = 0x1A, // ofadd l,  r,    d=overf   // calculate overflow bit of (l+r)
  m_cfshl  = 0x1B, // cfshl l,  r,    d=carry   // calculate carry    bit of (l<<r)
  m_cfshr  = 0x1C, // cfshr l,  r,    d=carry   // calculate carry    bit of (l>>r)
  m_sets   = 0x1D, // sets  l,          d=byte  SF=1          Sign
  m_seto   = 0x1E, // seto  l,  r,      d=byte  OF=1          Overflow of (l-r)
  m_setp   = 0x1F, // setp  l,  r,      d=byte  PF=1          Unordered/Parity        *F
  m_setnz  = 0x20, // setnz l,  r,      d=byte  ZF=0          Not Equal               *F
  m_setz   = 0x21, // setz  l,  r,      d=byte  ZF=1          Equal                   *F
  m_setae  = 0x22, // setae l,  r,      d=byte  CF=0          Unsigned Above or Equal *F
  m_setb   = 0x23, // setb  l,  r,      d=byte  CF=1          Unsigned Below          *F
  m_seta   = 0x24, // seta  l,  r,      d=byte  CF=0 & ZF=0   Unsigned Above          *F
  m_setbe  = 0x25, // setbe l,  r,      d=byte  CF=1 | ZF=1   Unsigned Below or Equal *F
  m_setg   = 0x26, // setg  l,  r,      d=byte  SF=OF & ZF=0  Signed Greater
  m_setge  = 0x27, // setge l,  r,      d=byte  SF=OF         Signed Greater or Equal
  m_setl   = 0x28, // setl  l,  r,      d=byte  SF!=OF        Signed Less
  m_setle  = 0x29, // setle l,  r,      d=byte  SF!=OF | ZF=1 Signed Less or Equal
  m_jcnd   = 0x2A, // jcnd   l,         d       // d is mop_v or mop_b
  m_jnz    = 0x2B, // jnz    l, r,      d       // ZF=0          Not Equal               *F
  m_jz     = 0x2C, // jz     l, r,      d       // ZF=1          Equal                   *F
  m_jae    = 0x2D, // jae    l, r,      d       // CF=0          Unsigned Above or Equal *F
  m_jb     = 0x2E, // jb     l, r,      d       // CF=1          Unsigned Below          *F
  m_ja     = 0x2F, // ja     l, r,      d       // CF=0 & ZF=0   Unsigned Above          *F
  m_jbe    = 0x30, // jbe    l, r,      d       // CF=1 | ZF=1   Unsigned Below or Equal *F
  m_jg     = 0x31, // jg     l, r,      d       // SF=OF & ZF=0  Signed Greater
  m_jge    = 0x32, // jge    l, r,      d       // SF=OF         Signed Greater or Equal
  m_jl     = 0x33, // jl     l, r,      d       // SF!=OF        Signed Less
  m_jle    = 0x34, // jle    l, r,      d       // SF!=OF | ZF=1 Signed Less or Equal
  m_jtbl   = 0x35, // jtbl   l, r=mcases        // Table jump
  m_ijmp   = 0x36, // ijmp       {r=sel, d=off} // indirect unconditional jump
  m_goto   = 0x37, // goto   l                  // l is mop_v or mop_b
  m_call   = 0x38, // call   l          d       // l is mop_v or mop_b or mop_h
  m_icall  = 0x39, // icall  {l=sel, r=off} d   // indirect call
  m_ret    = 0x3A, // ret
  m_push   = 0x3B, // push   l
  m_pop    = 0x3C, // pop               d
  m_und    = 0x3D, // und               d       // undefine
  m_ext    = 0x3E, // ext  in1, in2,  out1      // external insn, not microcode *F
  m_f2i    = 0x3F, // f2i    l,    d       int(l) => d; convert fp -> integer   +F
  m_f2u    = 0x40, // f2u    l,    d       uint(l)=> d; convert fp -> uinteger  +F
  m_i2f    = 0x41, // i2f    l,    d       fp(l)  => d; convert integer -> fp   +F
  m_u2f    = 0x42, // i2f    l,    d       fp(l)  => d; convert uinteger -> fp  +F
  m_f2f    = 0x43, // f2f    l,    d       l      => d; change fp precision     +F
  m_fneg   = 0x44, // fneg   l,    d       -l     => d; change sign             +F
  m_fadd   = 0x45, // fadd   l, r, d       l + r  => d; add                     +F
  m_fsub   = 0x46, // fsub   l, r, d       l - r  => d; subtract                +F
  m_fmul   = 0x47, // fmul   l, r, d       l * r  => d; multiply                +F
  m_fdiv   = 0x48, // fdiv   l, r, d       l / r  => d; divide                  +F
#define m_max 0x49 // first unused opcode
};
 
/// Must an instruction with the given opcode be the last one in a block?
/// Such opcodes are called closing opcodes.
/// \param mcode instruction opcode
/// \param including_calls should m_call/m_icall be considered as the closing opcodes?
/// If this function returns true, the opcode cannot appear in the middle
/// of a block. Calls are a special case: unknown calls (\ref is_unknown_call)
/// are considered as closing opcodes.
 
THREAD_SAFE bool hexapi must_mcode_close_block(mcode_t mcode, bool including_calls);
 
 
/// May opcode be propagated?
/// Such opcodes can be used in sub-instructions (nested instructions)
/// There is a handful of non-propagatable opcodes, like jumps, ret, nop, etc
/// All other regular opcodes are propagatable and may appear in a nested
/// instruction.
 
THREAD_SAFE bool hexapi is_mcode_propagatable(mcode_t mcode);
 
 
// Is add or sub instruction?
inline THREAD_SAFE bool is_mcode_addsub(mcode_t mcode) { return mcode == m_add || mcode == m_sub; }
// Is xds or xdu instruction? We use 'xdsu' as a shortcut for 'xds or xdu'
inline THREAD_SAFE bool is_mcode_xdsu(mcode_t mcode) { return mcode == m_xds || mcode == m_xdu; }
// Is a 'set' instruction? (an instruction that sets a condition code)
inline THREAD_SAFE bool is_mcode_set(mcode_t mcode) { return mcode >= m_sets && mcode <= m_setle; }
// Is a 1-operand 'set' instruction? Only 'sets' is in this group
inline THREAD_SAFE bool is_mcode_set1(mcode_t mcode) { return mcode == m_sets; }
// Is a 1-operand conditional jump instruction? Only 'jcnd' is in this group
inline THREAD_SAFE bool is_mcode_j1(mcode_t mcode) { return mcode == m_jcnd; }
// Is a conditional jump?
inline THREAD_SAFE bool is_mcode_jcond(mcode_t mcode) { return mcode >= m_jcnd && mcode <= m_jle; }
// Is a 'set' instruction that can be converted into a conditional jump?
inline THREAD_SAFE bool is_mcode_convertible_to_jmp(mcode_t mcode) { return mcode >= m_setnz && mcode <= m_setle; }
// Is a conditional jump instruction that can be converted into a 'set'?
inline THREAD_SAFE bool is_mcode_convertible_to_set(mcode_t mcode) { return mcode >= m_jnz && mcode <= m_jle; }
// Is a call instruction? (direct or indirect)
inline THREAD_SAFE bool is_mcode_call(mcode_t mcode) { return mcode == m_call || mcode == m_icall; }
// Must be an FPU instruction?
inline THREAD_SAFE bool is_mcode_fpu(mcode_t mcode) { return mcode >= m_f2i; }
// Is a commutative instruction?
inline THREAD_SAFE bool is_mcode_commutative(mcode_t mcode)
{
  return mcode == m_add
      || mcode == m_mul
      || mcode == m_or
      || mcode == m_and
      || mcode == m_xor
      || mcode == m_setz
      || mcode == m_setnz
      || mcode == m_cfadd
      || mcode == m_ofadd;
}
// Is a shift instruction?
inline THREAD_SAFE bool is_mcode_shift(mcode_t mcode)
{
  return mcode == m_shl
      || mcode == m_shr
      || mcode == m_sar;
}
// Is a kind of div or mod instruction?
inline THREAD_SAFE bool is_mcode_divmod(mcode_t op)
{
  return op == m_udiv || op == m_sdiv || op == m_umod || op == m_smod;
}
// Is an instruction with the selector/offset pair?
inline THREAD_SAFE bool has_mcode_seloff(mcode_t op)
{
  return op == m_ldx || op == m_stx || op == m_icall || op == m_ijmp;
}
 
// Convert setX opcode into corresponding jX opcode
// This function relies on the order of setX and jX opcodes!
inline THREAD_SAFE mcode_t set2jcnd(mcode_t code)
{
  return mcode_t(code - m_setnz + m_jnz);
}
 
// Convert setX opcode into corresponding jX opcode
// This function relies on the order of setX and jX opcodes!
inline THREAD_SAFE mcode_t jcnd2set(mcode_t code)
{
  return mcode_t(code + m_setnz - m_jnz);
}
 
// Negate a conditional opcode.
// Conditional jumps can be negated, example: jle -> jg
// 'Set' instruction can be negated, example: seta -> setbe
// If the opcode cannot be negated, return m_nop
THREAD_SAFE mcode_t hexapi negate_mcode_relation(mcode_t code);
 
 
// Swap a conditional opcode.
// Only conditional jumps and set instructions can be swapped.
// The returned opcode the one required for swapped operands.
// Example "x > y" is the same as "y < x", therefore swap(m_jg) is m_jl.
// If the opcode cannot be swapped, return m_nop
 
THREAD_SAFE mcode_t hexapi swap_mcode_relation(mcode_t code);
 
// Return the opcode that performs signed operation.
// Examples: jae -> jge; udiv -> sdiv
// If the opcode cannot be transformed into signed form, simply return it.
 
THREAD_SAFE mcode_t hexapi get_signed_mcode(mcode_t code);
 
 
// Return the opcode that performs unsigned operation.
// Examples: jl -> jb; xds -> xdu
// If the opcode cannot be transformed into unsigned form, simply return it.
 
THREAD_SAFE mcode_t hexapi get_unsigned_mcode(mcode_t code);
 
// Does the opcode perform a signed operation?
inline THREAD_SAFE bool is_signed_mcode(mcode_t code) { return get_unsigned_mcode(code) != code; }
// Does the opcode perform a unsigned operation?
inline THREAD_SAFE bool is_unsigned_mcode(mcode_t code) { return get_signed_mcode(code) != code; }
 
 
// Does the 'd' operand gets modified by the instruction?
// Example: "add l,r,d" modifies d, while instructions
// like jcnd, ijmp, stx does not modify it.
// Note: this function returns 'true' for m_ext but it may be wrong.
// Use minsn_t::modifies_d() if you have minsn_t.
 
THREAD_SAFE bool hexapi mcode_modifies_d(mcode_t mcode);
 
 
// Processor condition codes are mapped to the first microregisters
// The order is important, see mop_t::is_cc()
const mreg_t mr_none  = mreg_t(-1);
const mreg_t mr_cf    = mreg_t(0);      // carry bit
const mreg_t mr_zf    = mreg_t(1);      // zero bit
const mreg_t mr_sf    = mreg_t(2);      // sign bit
const mreg_t mr_of    = mreg_t(3);      // overflow bit
const mreg_t mr_pf    = mreg_t(4);      // parity bit
const int    cc_count = mr_pf - mr_cf + 1; // number of condition code registers
const mreg_t mr_cc    = mreg_t(5);       // synthetic condition code, used internally
const mreg_t mr_first = mreg_t(8);       // the first processor specific register
 
//-------------------------------------------------------------------------
/// Operand locator.
/// It is used to denote a particular operand in the ctree, for example,
/// when the user right clicks on a constant and requests to represent it, say,
/// as a hexadecimal number.
struct operand_locator_t
{
private:
  // forbid the default constructor, force the user to initialize objects of this class.
  operand_locator_t(void) {}
public:
  ea_t ea;              ///< address of the original processor instruction
  int opnum;            ///< operand number in the instruction
  operand_locator_t(ea_t _ea, int _opnum) : ea(_ea), opnum(_opnum) {}
  DECLARE_COMPARISONS(operand_locator_t);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
};
 
//-------------------------------------------------------------------------
/// Number representation.
/// This structure holds information about a number format.
struct number_format_t
{
  flags_t flags32 = 0;    ///< low 32bit of flags (for compatibility)
  char opnum;             ///< operand number: 0..UA_MAXOP
  char props = 0;         ///< properties: combination of NF_ bits (\ref NF_)
/// \defgroup NF_ Number format property bits
/// Used in number_format_t::props
//@{
#define NF_FIXED    0x01  ///< number format has been defined by the user
#define NF_NEGDONE  0x02  ///< temporary internal bit: negation has been performed
#define NF_BINVDONE 0x04  ///< temporary internal bit: inverting bits is done
#define NF_NEGATE   0x08  ///< The user asked to negate the constant
#define NF_BITNOT   0x10  ///< The user asked to invert bits of the constant
#define NF_VALID    0x20  ///< internal bit: stroff or enum is valid
                          ///< for enums: this bit is set immediately
                          ///< for stroffs: this bit is set at the end of decompilation
//@}
  uchar serial = 0;       ///< for enums: constant serial number
  char org_nbytes = 0;    ///< original number size in bytes
  qstring type_name;      ///< for stroffs: structure for offsetof()\n
                          ///< for enums: enum name
  flags64_t flags = 0;    ///< ida flags, which describe number radix, enum, etc
  /// Contructor
  number_format_t(int _opnum=0) : opnum(char(_opnum)) {}
  /// Get number radix
  /// \return 2,8,10, or 16
  int get_radix() const { return ::get_radix(flags, opnum); }
  /// Is number representation fixed?
  /// Fixed representation cannot be modified by the decompiler
  bool is_fixed() const { return props != 0; }
  /// Is a hexadecimal number?
  bool is_hex() const { return ::is_numop(flags, opnum) && get_radix() == 16; }
  /// Is a decimal number?
  bool is_dec() const { return ::is_numop(flags, opnum) && get_radix() == 10; }
  /// Is a octal number?
  bool is_oct() const { return ::is_numop(flags, opnum) && get_radix() == 8; }
  /// Is a symbolic constant?
  bool is_enum() const { return ::is_enum(flags, opnum); }
  /// Is a character constant?
  bool is_char() const { return ::is_char(flags, opnum); }
  /// Is a structure field offset?
  bool is_stroff() const { return ::is_stroff(flags, opnum); }
  /// Is a number?
  bool is_numop() const { return !is_enum() && !is_char() && !is_stroff(); }
  /// Does the number need to be negated or bitwise negated?
  /// Returns true if the user requested a negation but it is not done yet
  bool needs_to_be_inverted() const
  {
    return (props & (NF_NEGATE|NF_BITNOT)) != 0      // the user requested it
        && (props & (NF_NEGDONE|NF_BINVDONE)) == 0;  // not done yet
  }
  // symbolic constants and struct offsets cannot easily change
  // their sign or size without a cast. only simple numbers can do that.
  // for example, by modifying the expression type we can convert:
  // 10u -> 10
  // but replacing the type of a symbol constant would lead to an inconsistency.
  bool has_unmutable_type() const
  {
    return (props & NF_VALID) != 0 && (is_stroff() || is_enum());
  }
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
};
 
// Number formats are attached to (ea,opnum) pairs
typedef std::map<operand_locator_t, number_format_t> user_numforms_t;
 
//-------------------------------------------------------------------------
/// Base helper class to convert binary data structures into text.
/// Other classes are derived from this class.
struct vd_printer_t
{
  qstring tmpbuf;
  int hdrlines;         ///< number of header lines (prototype+typedef+lvars)
                        ///< valid at the end of print process
  /// Print.
  /// This function is called to generate a portion of the output text.
  /// The output text may contain color codes.
  /// \return the number of printed characters
  /// \param indent  number of spaces to generate as prefix
  /// \param format  printf-style format specifier
  /// \return length of printed string
  AS_PRINTF(3, 4) virtual int hexapi print(int indent, const char *format, ...);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
};
 
/// Helper class to convert cfunc_t into text.
struct vc_printer_t : public vd_printer_t
{
  const cfunc_t *func;          ///< cfunc_t to generate text for
  char lastchar;                ///< internal: last printed character
  /// Constructor
  vc_printer_t(const cfunc_t *f) : func(f), lastchar(0) {}
  /// Are we generating one-line text representation?
  /// \return \c true if the output will occupy one line without line breaks
  virtual bool idaapi oneliner(void) const newapi { return false; }
};
 
/// Helper class to convert binary data structures into text and put into a file.
struct file_printer_t : public vd_printer_t
{
  FILE *fp;                     ///< Output file pointer
  /// Print.
  /// This function is called to generate a portion of the output text.
  /// The output text may contain color codes.
  /// \return the number of printed characters
  /// \param indent  number of spaces to generate as prefix
  /// \param format  printf-style format specifier
  /// \return length of printed string
  AS_PRINTF(3, 4) int hexapi print(int indent, const char *format, ...) override;
  /// Constructor
  file_printer_t(FILE *_fp) : fp(_fp) {}
};
 
/// Helper class to convert cfunc_t into a text string
struct qstring_printer_t : public vc_printer_t
{
  bool with_tags;               ///< Generate output with color tags
  qstring &s;                   ///< Reference to the output string
  /// Constructor
  qstring_printer_t(const cfunc_t *f, qstring &_s, bool tags)
    : vc_printer_t(f), with_tags(tags), s(_s) {}
  /// Print.
  /// This function is called to generate a portion of the output text.
  /// The output text may contain color codes.
  /// \return the number of printed characters
  /// \param indent  number of spaces to generate as prefix
  /// \param format  printf-style format specifier
  /// \return length of the printed string
  AS_PRINTF(3, 4) int hexapi print(int indent, const char *format, ...) override;
};
 
//-------------------------------------------------------------------------
/// \defgroup type Type string related declarations
/// Type related functions and class.
//@{
 
/// Print the specified type info.
/// This function can be used from a debugger by typing "tif->dstr()"
 
const char *hexapi dstr(const tinfo_t *tif);
 
 
/// Verify a type string.
/// \return true if type string is correct
 
bool hexapi is_type_correct(const type_t *ptr);
 
 
/// Is a small structure or union?
/// \return true if the type is a small UDT (user defined type).
///              Small UDTs fit into a register (or pair or registers) as a rule.
 
bool hexapi is_small_udt(const tinfo_t &tif);
 
 
/// Is definitely a non-boolean type?
/// \return true if the type is a non-boolean type (non bool and well defined)
 
bool hexapi is_nonbool_type(const tinfo_t &type);
 
 
/// Is a boolean type?
/// \return true if the type is a boolean type
 
bool hexapi is_bool_type(const tinfo_t &type);
 
 
/// Is a pointer or array type?
inline THREAD_SAFE bool is_ptr_or_array(type_t t)
{
  return is_type_ptr(t) || is_type_array(t);
}
 
/// Is a pointer, array, or function type?
inline THREAD_SAFE bool is_paf(type_t t)
{
  return is_ptr_or_array(t) || is_type_func(t);
}
 
/// Is struct/union/enum definition (not declaration)?
inline THREAD_SAFE bool is_inplace_def(const tinfo_t &type)
{
  return type.is_decl_complex() && !type.is_typeref();
}
 
/// Calculate number of partial subtypes.
/// \return number of partial subtypes. The bigger is this number, the uglier is the type.
 
int hexapi partial_type_num(const tinfo_t &type);
 
 
/// Get a type of a floating point value with the specified width
/// \returns type info object
/// \param width width of the desired type
 
tinfo_t hexapi get_float_type(int width);
 
 
/// Create a type info by width and sign.
/// Returns a simple type (examples: int, short) with the given width and sign.
/// \param srcwidth size of the type in bytes
/// \param sign sign of the type
 
tinfo_t hexapi get_int_type_by_width_and_sign(int srcwidth, type_sign_t sign);
 
 
/// Create a partial type info by width.
/// Returns a partially defined type (examples: _DWORD, _BYTE) with the given width.
/// \param size size of the type in bytes
 
tinfo_t hexapi get_unk_type(int size);
 
 
/// Generate a dummy pointer type
///  \param ptrsize size of pointed object
///  \param isfp is floating point object?
 
tinfo_t hexapi dummy_ptrtype(int ptrsize, bool isfp);
 
 
/// Get type of a structure field.
/// This function performs validity checks of the field type. Wrong types are rejected.
/// \param mptr structure field
/// \param type pointer to the variable where the type is returned. This parameter can be nullptr.
/// \return false if failed
 
bool hexapi get_member_type(const member_t *mptr, tinfo_t *type);
 
 
/// Create a pointer type.
/// This function performs the following conversion: "type" -> "type*"
/// \param type object type.
/// \return "type*". for example, if 'char' is passed as the argument,
//          the function will return 'char *'
 
tinfo_t hexapi make_pointer(const tinfo_t &type);
 
 
/// Create a reference to a named type.
/// \param name type name
/// \return type which refers to the specified name. For example, if name is "DWORD",
///             the type info which refers to "DWORD" is created.
 
tinfo_t hexapi create_typedef(const char *name);
 
 
/// Create a reference to an ordinal type.
/// \param n ordinal number of the type
/// \return type which refers to the specified ordinal. For example, if n is 1,
///             the type info which refers to ordinal type 1 is created.
 
inline tinfo_t create_typedef(int n)
{
  tinfo_t tif;
  tif.create_typedef(nullptr, n);
  return tif;
}
 
/// Type source (where the type information comes from)
enum type_source_t
{
  GUESSED_NONE,  // not guessed, specified by the user
  GUESSED_WEAK,  // not guessed, comes from idb
  GUESSED_FUNC,  // guessed as a function
  GUESSED_DATA,  // guessed as a data item
  TS_NOELL   = 0x8000000, // can be used in set_type() to avoid merging into ellipsis
  TS_SHRINK  = 0x4000000, // can be used in set_type() to prefer smaller arguments
  TS_DONTREF = 0x2000000, // do not mark type as referenced (referenced_types)
  TS_MASK    = 0xE000000, // all high bits
};
 
 
/// Get a global type.
/// Global types are types of addressable objects and struct/union/enum types
/// \param id address or id of the object
/// \param tif buffer for the answer
/// \param guess what kind of types to consider
/// \return success
 
bool hexapi get_type(uval_t id, tinfo_t *tif, type_source_t guess);
 
 
/// Set a global type.
/// \param id address or id of the object
/// \param tif new type info
/// \param source where the type comes from
/// \param force true means to set the type as is, false means to merge the
///        new type with the possibly existing old type info.
/// \return success
 
bool hexapi set_type(uval_t id, const tinfo_t &tif, type_source_t source, bool force=false);
 
//@}
 
//-------------------------------------------------------------------------
// We use our own class to store argument and variable locations.
// It is called vdloc_t that stands for 'vd location'.
// 'vd' is the internal name of the decompiler, it stands for 'visual decompiler'.
// The main differences between vdloc and argloc_t:
//   ALOC_REG1: the offset is always 0, so it is not used. the register number
//              uses the whole ~VLOC_MASK field.
//   ALOCK_STKOFF: stack offsets are always positive because they are based on
//              the lowest value of sp in the function.
class vdloc_t : public argloc_t
{
  int regoff(void); // inaccessible & undefined: regoff() should not be used
public:
  // Get the register number.
  // This function works only for ALOC_REG1 and ALOC_REG2 location types.
  // It uses all available bits for register number for ALOC_REG1
  int reg1(void) const { return atype() == ALOC_REG2 ? argloc_t::reg1() : get_reginfo(); }
 
  // Set vdloc to point to the specified register without cleaning it up.
  // This is a dangerous function, use set_reg1() instead unless you understand
  // what it means to cleanup an argloc.
  void _set_reg1(int r1) { argloc_t::_set_reg1(r1, r1>>16); }
 
  // Set vdloc to point to the specified register.
  void set_reg1(int r1) { cleanup_argloc(this); _set_reg1(r1); }
 
  // Use member functions of argloc_t for other location types.
 
  // Return textual representation.
  // Note: this and all other dstr() functions can be used from a debugger.
  // It is much easier than to inspect the memory contents byte by byte.
  const char *hexapi dstr(int width=0) const;
  DECLARE_COMPARISONS(vdloc_t);
  bool hexapi is_aliasable(const mba_t *mb, int size) const;
};
 
/// Print vdloc.
/// Since vdloc does not always carry the size info, we pass it as NBYTES..
void hexapi print_vdloc(qstring *vout, const vdloc_t &loc, int nbytes);
 
//-------------------------------------------------------------------------
/// Do two arglocs overlap?
bool hexapi arglocs_overlap(const vdloc_t &loc1, size_t w1, const vdloc_t &loc2, size_t w2);
 
/// Local variable locator.
/// Local variables are located using definition ea and location.
/// Each variable must have a unique locator, this is how we tell them apart.
struct lvar_locator_t
{
  vdloc_t location;     ///< Variable location.
  ea_t defea;           ///< Definition address. Usually, this is the address
                        ///< of the instruction that initializes the variable.
                        ///< In some cases it can be a fictional address.
 
  lvar_locator_t(void) : defea(BADADDR) {}
  lvar_locator_t(const vdloc_t &loc, ea_t ea) : location(loc), defea(ea) {}
  /// Get offset of the varialbe in the stack frame.
  /// \return a non-negative value for stack variables. The value is
  ///         an offset from the bottom of the stack frame in terms of
  ///         vd-offsets.
  ///         negative values mean error (not a stack variable)
  sval_t get_stkoff(void) const
  {
    return location.is_stkoff() ? location.stkoff() : -1;
  }
  /// Is variable located on one register?
  bool is_reg1(void) const { return  location.is_reg1(); }
  /// Is variable located on two registers?
  bool is_reg2(void) const { return  location.is_reg2(); }
  /// Is variable located on register(s)?
  bool is_reg_var(void) const { return location.is_reg(); }
  /// Is variable located on the stack?
  bool is_stk_var(void) const { return location.is_stkoff(); }
  /// Is variable scattered?
  bool is_scattered(void) const { return location.is_scattered(); }
  /// Get the register number of the variable
  mreg_t get_reg1(void) const { return location.reg1(); }
  /// Get the number of the second register (works only for ALOC_REG2 lvars)
  mreg_t get_reg2(void) const { return location.reg2(); }
  /// Get information about scattered variable
  const scattered_aloc_t &get_scattered(void) const { return location.scattered(); }
        scattered_aloc_t &get_scattered(void)       { return location.scattered(); }
  DECLARE_COMPARISONS(lvar_locator_t);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  // Debugging: get textual representation of a lvar locator.
  const char *hexapi dstr() const;
};
 
/// Definition of a local variable (register or stack) #var #lvar
class lvar_t : public lvar_locator_t
{
  friend class mba_t;
  int flags;                    ///< \ref CVAR_
/// \defgroup CVAR_ Local variable property bits
/// Used in lvar_t::flags
//@{
#define CVAR_USED    0x00000001 ///< is used in the code?
#define CVAR_TYPE    0x00000002 ///< the type is defined?
#define CVAR_NAME    0x00000004 ///< has nice name?
#define CVAR_MREG    0x00000008 ///< corresponding mregs were replaced?
#define CVAR_NOWD    0x00000010 ///< width is unknown
#define CVAR_UNAME   0x00000020 ///< user-defined name
#define CVAR_UTYPE   0x00000040 ///< user-defined type
#define CVAR_RESULT  0x00000080 ///< function result variable
#define CVAR_ARG     0x00000100 ///< function argument
#define CVAR_FAKE    0x00000200 ///< fake variable (return var or va_list)
#define CVAR_OVER    0x00000400 ///< overlapping variable
#define CVAR_FLOAT   0x00000800 ///< used in a fpu insn
#define CVAR_SPOILED 0x00001000 ///< internal flag, do not use: spoiled var
#define CVAR_MAPDST  0x00002000 ///< other variables are mapped to this var
#define CVAR_PARTIAL 0x00004000 ///< variable type is partialy defined
#define CVAR_THISARG 0x00008000 ///< 'this' argument of c++ member functions
#define CVAR_FORCED  0x00010000 ///< variable was created by an explicit request
                                ///< otherwise we could reuse an existing var
#define CVAR_REGNAME 0x00020000 ///< has a register name (like _RAX): if lvar
                                ///< is used by an m_ext instruction
#define CVAR_NOPTR   0x00040000 ///< variable cannot be a pointer (user choice)
#define CVAR_DUMMY   0x00080000 ///< dummy argument (added to fill a hole in
                                ///< the argument list)
#define CVAR_NOTARG  0x00100000 ///< variable cannot be an input argument
#define CVAR_AUTOMAP 0x00200000 ///< variable was automatically mapped
#define CVAR_BYREF   0x00400000 ///< the address of the variable was taken
#define CVAR_INASM   0x00800000 ///< variable is used in instructions translated
                                ///< into __asm {...}
#define CVAR_UNUSED  0x01000000 ///< user-defined __unused attribute
                                ///< meaningful only if: is_arg_var() && !mba->final_type
#define CVAR_SHARED  0x02000000 ///< variable is mapped to several chains
//@}
 
public:
  qstring name;          ///< variable name.
                         ///< use mba_t::set_nice_lvar_name() and
                         ///< mba_t::set_user_lvar_name() to modify it
  qstring cmt;           ///< variable comment string
  tinfo_t tif;           ///< variable type
  int width = 0;         ///< variable size in bytes
  int defblk = -1;       ///< first block defining the variable.
                         ///< 0 for args, -1 if unknown
  uint64 divisor = 0;    ///< max known divisor of the variable
 
  lvar_t(void) : flags(CVAR_USED) {}
  lvar_t(const qstring &n, const vdloc_t &l, ea_t e, const tinfo_t &t, int w, int db)
    : lvar_locator_t(l, e), flags(CVAR_USED), name(n), tif(t), width(w), defblk(db)
  {
  }
  // Debugging: get textual representation of a local variable.
  const char *hexapi dstr() const;
 
  /// Is the variable used in the code?
  bool used(void)  const { return (flags & CVAR_USED) != 0; }
  /// Has the variable a type?
  bool typed(void) const { return (flags & CVAR_TYPE) != 0; }
  /// Have corresponding microregs been replaced by references to this variable?
  bool mreg_done(void) const { return (flags & CVAR_MREG) != 0; }
  /// Does the variable have a nice name?
  bool has_nice_name(void) const { return (flags & CVAR_NAME) != 0; }
  /// Do we know the width of the variable?
  bool is_unknown_width(void) const { return (flags & CVAR_NOWD) != 0; }
  /// Has any user-defined information?
  bool has_user_info(void) const
  {
    return (flags & (CVAR_UNAME|CVAR_UTYPE|CVAR_NOPTR|CVAR_UNUSED)) != 0
        || !cmt.empty();
  }
  /// Has user-defined name?
  bool has_user_name(void) const { return (flags & CVAR_UNAME) != 0; }
  /// Has user-defined type?
  bool has_user_type(void) const { return (flags & CVAR_UTYPE) != 0; }
  /// Is the function result?
  bool is_result_var(void) const { return (flags & CVAR_RESULT) != 0; }
  /// Is the function argument?
  bool is_arg_var(void) const { return (flags & CVAR_ARG) != 0; }
  /// Is the promoted function argument?
  bool hexapi is_promoted_arg(void) const;
  /// Is fake return variable?
  bool is_fake_var(void) const { return (flags & CVAR_FAKE) != 0; }
  /// Is overlapped variable?
  bool is_overlapped_var(void) const { return (flags & CVAR_OVER) != 0; }
  /// Used by a fpu insn?
  bool is_floating_var(void) const { return (flags & CVAR_FLOAT) != 0; }
  /// Is spoiled var? (meaningful only during lvar allocation)
  bool is_spoiled_var(void) const { return (flags & CVAR_SPOILED) != 0; }
  /// Variable type should be handled as a partial one
  bool is_partialy_typed(void) const { return (flags & CVAR_PARTIAL) != 0; }
  /// Variable type should not be a pointer
  bool is_noptr_var(void) const { return (flags & CVAR_NOPTR) != 0; }
  /// Other variable(s) map to this var?
  bool is_mapdst_var(void) const { return (flags & CVAR_MAPDST) != 0; }
  /// Is 'this' argument of a C++ member function?
  bool is_thisarg(void) const { return (flags & CVAR_THISARG) != 0; }
  /// Is a forced variable?
  bool is_forced_var(void) const { return (flags & CVAR_FORCED) != 0; }
  /// Has a register name? (like _RAX)
  bool has_regname(void) const { return (flags & CVAR_REGNAME) != 0; }
  /// Is variable used in an instruction translated into __asm?
  bool in_asm(void) const { return (flags & CVAR_INASM) != 0; }
  /// Is a dummy argument (added to fill a hole in the argument list)
  bool is_dummy_arg(void) const { return (flags & CVAR_DUMMY) != 0; }
  /// Is a local variable? (local variable cannot be an input argument)
  bool is_notarg(void) const { return (flags & CVAR_NOTARG) != 0; }
  /// Was the variable automatically mapped to another variable?
  bool is_automapped(void) const { return (flags & CVAR_AUTOMAP) != 0; }
  /// Was the address of the variable taken?
  bool is_used_byref(void) const { return (flags & CVAR_BYREF) != 0; }
  /// Was declared as __unused by the user? See CVAR_UNUSED
  bool is_decl_unused(void) const { return (flags & CVAR_UNUSED) != 0; }
  /// Is lvar mapped to several chains
  bool is_shared(void) const { return (flags & CVAR_SHARED) != 0; }
  void set_used(void) { flags |= CVAR_USED; }
  void clear_used(void) { flags &= ~CVAR_USED; }
  void set_typed(void) { flags |= CVAR_TYPE; clr_noptr_var(); }
  void set_non_typed(void) { flags &= ~CVAR_TYPE; }
  void clr_user_info(void) { flags &= ~(CVAR_UNAME|CVAR_UTYPE|CVAR_NOPTR); }
  void set_user_name(void) { flags |= CVAR_NAME|CVAR_UNAME; }
  void set_user_type(void) { flags |= CVAR_TYPE|CVAR_UTYPE; }
  void clr_user_type(void) { flags &= ~CVAR_UTYPE; }
  void clr_user_name(void) { flags &= ~CVAR_UNAME; }
  void set_mreg_done(void) { flags |= CVAR_MREG; }
  void clr_mreg_done(void) { flags &= ~CVAR_MREG; }
  void set_unknown_width(void) { flags |= CVAR_NOWD; }
  void clr_unknown_width(void) { flags &= ~CVAR_NOWD; }
  void set_arg_var(void) { flags |= CVAR_ARG; }
  void clr_arg_var(void) { flags &= ~(CVAR_ARG|CVAR_THISARG); }
  void set_fake_var(void) { flags |= CVAR_FAKE; }
  void clr_fake_var(void) { flags &= ~CVAR_FAKE; }
  void set_overlapped_var(void) { flags |= CVAR_OVER; }
  void clr_overlapped_var(void) { flags &= ~CVAR_OVER; }
  void set_floating_var(void) { flags |= CVAR_FLOAT; }
  void clr_floating_var(void) { flags &= ~CVAR_FLOAT; }
  void set_spoiled_var(void) { flags |= CVAR_SPOILED; }
  void clr_spoiled_var(void) { flags &= ~CVAR_SPOILED; }
  void set_mapdst_var(void) { flags |= CVAR_MAPDST; }
  void clr_mapdst_var(void) { flags &= ~CVAR_MAPDST; }
  void set_partialy_typed(void) { flags |= CVAR_PARTIAL; }
  void clr_partialy_typed(void) { flags &= ~CVAR_PARTIAL; }
  void set_noptr_var(void) { flags |= CVAR_NOPTR; }
  void clr_noptr_var(void) { flags &= ~CVAR_NOPTR; }
  void set_thisarg(void) { flags |= CVAR_THISARG; }
  void clr_thisarg(void) { flags &= ~CVAR_THISARG; }
  void set_forced_var(void) { flags |= CVAR_FORCED; }
  void clr_forced_var(void) { flags &= ~CVAR_FORCED; }
  void set_dummy_arg(void) { flags |= CVAR_DUMMY; }
  void clr_dummy_arg(void) { flags &= ~CVAR_DUMMY; }
  void set_notarg(void) { clr_arg_var(); flags |= CVAR_NOTARG; }
  void clr_notarg(void) { flags &= ~CVAR_NOTARG; }
  void set_automapped(void) { flags |= CVAR_AUTOMAP; }
  void clr_automapped(void) { flags &= ~CVAR_AUTOMAP; }
  void set_used_byref(void) { flags |= CVAR_BYREF; }
  void clr_used_byref(void) { flags &= ~CVAR_BYREF; }
  void set_decl_unused(void) { flags |= CVAR_UNUSED; }
  void clr_decl_unused(void) { flags &= ~CVAR_UNUSED; }
  void set_shared(void) { flags |= CVAR_SHARED; }
  void clr_shared(void) { flags &= ~CVAR_SHARED; }
 
  /// Do variables overlap?
  bool has_common(const lvar_t &v) const
  {
    return arglocs_overlap(location, width, v.location, v.width);
  }
  /// Does the variable overlap with the specified location?
  bool has_common_bit(const vdloc_t &loc, asize_t width2) const
  {
    return arglocs_overlap(location, width, loc, width2);
  }
  /// Get variable type
  const tinfo_t &type(void) const { return tif; }
  tinfo_t &type(void) { return tif; }
 
  /// Check if the variable accept the specified type.
  /// Some types are forbidden (void, function types, wrong arrays, etc)
  bool hexapi accepts_type(const tinfo_t &t, bool may_change_thisarg=false);
  /// Set variable type
  /// Note: this function does not modify the idb, only the lvar instance
  /// in the memory. For permanent changes see modify_user_lvars()
  /// Also, the variable type is not considered as final by the decompiler
  /// and may be modified later by the type derivation.
  /// In some cases set_final_var_type() may work better, but it does not
  /// do persistent changes to the database neither.
  /// \param t new type
  /// \param may_fail if false and type is bad, interr
  /// \return success
  bool hexapi set_lvar_type(const tinfo_t &t, bool may_fail=false);
 
  /// Set final variable type.
  void set_final_lvar_type(const tinfo_t &t)
  {
    set_lvar_type(t);
    set_typed();
  }
 
  /// Change the variable width.
  /// We call the variable size 'width', it is represents the number of bytes.
  /// This function may change the variable type using set_lvar_type().
  /// \param w new width
  /// \param svw_flags combination of SVW_... bits
  /// \return success
  bool hexapi set_width(int w, int svw_flags=0);
#define SVW_INT   0x00 // integer value
#define SVW_FLOAT 0x01 // floating point value
#define SVW_SOFT  0x02 // may fail and return false;
                       // if this bit is not set and the type is bad, interr
 
  /// Append local variable to mlist.
  /// \param mba ptr to the current mba_t
  /// \param lst list to append to
  /// \param pad_if_scattered if true, append padding bytes in case of scattered lvar
  void hexapi append_list(const mba_t *mba, mlist_t *lst, bool pad_if_scattered=false) const;
 
  /// Is the variable aliasable?
  /// \param mba ptr to the current mba_t
  /// Aliasable variables may be modified indirectly (through a pointer)
  bool is_aliasable(const mba_t *mba) const
  {
    return location.is_aliasable(mba, width);
  }
 
};
DECLARE_TYPE_AS_MOVABLE(lvar_t);
 
/// Vector of local variables
struct lvars_t : public qvector<lvar_t>
{
  /// Find input variable at the specified location.
  /// \param argloc variable location
  /// \param _size variable size
  /// \return -1 if failed, otherwise the index into the variables vector.
  int find_input_lvar(const vdloc_t &argloc, int _size) { return find_lvar(argloc, _size, 0); }
 
 
  /// Find stack variable at the specified location.
  /// \param spoff offset from the minimal sp
  /// \param width variable size
  /// \return -1 if failed, otherwise the index into the variables vector.
  int hexapi find_stkvar(sval_t spoff, int width);
 
 
  /// Find variable at the specified location.
  /// \param ll variable location
  /// \return pointer to variable or nullptr
  lvar_t *hexapi find(const lvar_locator_t &ll);
 
 
  /// Find variable at the specified location.
  /// \param location variable location
  /// \param width variable size
  /// \param defblk definition block of the lvar. -1 means any block
  /// \return -1 if failed, otherwise the index into the variables vector.
  int hexapi find_lvar(const vdloc_t &location, int width, int defblk=-1) const;
};
 
/// Saved user settings for local variables: name, type, comment.
struct lvar_saved_info_t
{
  lvar_locator_t ll;            ///< Variable locator
  qstring name;                 ///< Name
  tinfo_t type;                 ///< Type
  qstring cmt;                  ///< Comment
  ssize_t size;                 ///< Type size (if not initialized then -1)
  int flags;                    ///< \ref LVINF_
/// \defgroup LVINF_ saved user lvar info property bits
/// Used in lvar_saved_info_t::flags
//@{
#define LVINF_KEEP   0x0001     ///< preserve saved user settings regardless of vars
                                ///< for example, if a var loses all its
                                ///< user-defined attributes or even gets
                                ///< destroyed, keep its lvar_saved_info_t.
                                ///< this is used for ephemeral variables that
                                ///< get destroyed by macro recognition.
#define LVINF_FORCE  0x0002     ///< force allocation of a new variable.
                                ///< forces the decompiler to create a new
                                ///< variable at ll.defea
#define LVINF_NOPTR  0x0004     ///< variable type should not be a pointer
#define LVINF_NOMAP  0x0008     ///< forbid automatic mapping of the variable
#define LVINF_UNUSED 0x0010     ///< unused argument, corresponds to CVAR_UNUSED
//@}
  lvar_saved_info_t(void) : size(BADSIZE), flags(0) {}
  bool has_info(void) const
  {
    return !name.empty()
        || !type.empty()
        || !cmt.empty()
        || is_forced_lvar()
        || is_noptr_lvar()
        || is_nomap_lvar();
  }
  bool operator==(const lvar_saved_info_t &r) const
  {
    return name == r.name
        && cmt == r.cmt
        && ll == r.ll
        && type == r.type;
  }
  bool operator!=(const lvar_saved_info_t &r) const { return !(*this == r); }
  bool is_kept(void) const { return (flags & LVINF_KEEP) != 0; }
  void clear_keep(void) { flags &= ~LVINF_KEEP; }
  void set_keep(void) { flags |= LVINF_KEEP; }
  bool is_forced_lvar(void) const { return (flags & LVINF_FORCE) != 0; }
  void set_forced_lvar(void) { flags |= LVINF_FORCE; }
  void clr_forced_lvar(void) { flags &= ~LVINF_FORCE; }
  bool is_noptr_lvar(void) const { return (flags & LVINF_NOPTR) != 0; }
  void set_noptr_lvar(void) { flags |= LVINF_NOPTR; }
  void clr_noptr_lvar(void) { flags &= ~LVINF_NOPTR; }
  bool is_nomap_lvar(void) const { return (flags & LVINF_NOMAP) != 0; }
  void set_nomap_lvar(void) { flags |= LVINF_NOMAP; }
  void clr_nomap_lvar(void) { flags &= ~LVINF_NOMAP; }
  bool is_unused_lvar(void) const { return (flags & LVINF_UNUSED) != 0; }
  void set_unused_lvar(void) { flags |= LVINF_UNUSED; }
  void clr_unused_lvar(void) { flags &= ~LVINF_UNUSED; }
};
DECLARE_TYPE_AS_MOVABLE(lvar_saved_info_t);
typedef qvector<lvar_saved_info_t> lvar_saved_infos_t;
 
/// Local variable mapping (is used to merge variables)
typedef std::map<lvar_locator_t, lvar_locator_t> lvar_mapping_t;
 
/// All user-defined information about local variables
struct lvar_uservec_t
{
  /// User-specified names, types, comments for lvars. Variables without
  /// user-specified info are not present in this vector.
  lvar_saved_infos_t lvvec;
 
  /// Local variable mapping (used for merging variables)
  lvar_mapping_t lmaps;
 
  /// Delta to add to IDA stack offset to calculate Hex-Rays stack offsets.
  /// Should be set by the caller before calling save_user_lvar_settings();
  uval_t stkoff_delta;
 
  /// Various flags. Possible values are from \ref ULV_
  int ulv_flags;
/// \defgroup ULV_ lvar_uservec_t property bits
/// Used in lvar_uservec_t::ulv_flags
//@{
#define ULV_PRECISE_DEFEA 0x0001        ///< Use precise defea's for lvar locations
//@}
 
  lvar_uservec_t(void) : stkoff_delta(0), ulv_flags(ULV_PRECISE_DEFEA) {}
  void swap(lvar_uservec_t &r)
  {
    lvvec.swap(r.lvvec);
    lmaps.swap(r.lmaps);
    std::swap(stkoff_delta, r.stkoff_delta);
    std::swap(ulv_flags, r.ulv_flags);
  }
  void clear()
  {
    lvvec.clear();
    lmaps.clear();
    stkoff_delta = 0;
    ulv_flags = ULV_PRECISE_DEFEA;
  }
  bool empty() const
  {
    return lvvec.empty()
        && lmaps.empty()
        && stkoff_delta == 0
        && ulv_flags == ULV_PRECISE_DEFEA;
  }
 
  /// find saved user settings for given var
  lvar_saved_info_t *find_info(const lvar_locator_t &vloc)
  {
    for ( lvar_saved_infos_t::iterator p=lvvec.begin(); p != lvvec.end(); ++p )
    {
      if ( p->ll == vloc )
        return p;
    }
    return nullptr;
  }
 
  /// Preserve user settings for given var
  void keep_info(const lvar_t &v)
  {
    lvar_saved_info_t *p = find_info(v);
    if ( p != nullptr )
      p->set_keep();
  }
};
 
/// Restore user defined local variable settings in the database.
/// \param func_ea entry address of the function
/// \param lvinf ptr to output buffer
/// \return success
 
bool hexapi restore_user_lvar_settings(lvar_uservec_t *lvinf, ea_t func_ea);
 
 
/// Save user defined local variable settings into the database.
/// \param func_ea entry address of the function
/// \param lvinf user-specified info about local variables
 
void hexapi save_user_lvar_settings(ea_t func_ea, const lvar_uservec_t &lvinf);
 
 
/// Helper class to modify saved local variable settings.
struct user_lvar_modifier_t
{
  /// Modify lvar settings.
  /// Returns: true-modified
  virtual bool idaapi modify_lvars(lvar_uservec_t *lvinf) = 0;
};
 
/// Modify saved local variable settings.
/// \param entry_ea         function start address
/// \param mlv              local variable modifier
/// \return true if modified variables
 
bool hexapi modify_user_lvars(ea_t entry_ea, user_lvar_modifier_t &mlv);
 
 
/// Modify saved local variable settings of one variable.
/// \param func_ea          function start address
/// \param info             local variable info attrs
/// \param mli_flags        bits that specify which attrs defined by INFO are to be set
/// \return true if modified, false if invalid MLI_FLAGS passed
 
bool hexapi modify_user_lvar_info(
        ea_t func_ea,
        uint mli_flags,
        const lvar_saved_info_t &info);
 
/// \defgroup MLI_ user info bits
//@{
#define MLI_NAME        0x01 ///< apply lvar name
#define MLI_TYPE        0x02 ///< apply lvar type
#define MLI_CMT         0x04 ///< apply lvar comment
#define MLI_SET_FLAGS   0x08 ///< set LVINF_... bits
#define MLI_CLR_FLAGS   0x10 ///< clear LVINF_... bits
//@}
 
 
/// Find a variable by name.
/// \param out              output buffer for the variable locator
/// \param func_ea          function start address
/// \param varname          variable name
/// \return success
/// Since VARNAME is not always enough to find the variable, it may decompile
/// the function.
 
bool hexapi locate_lvar(
        lvar_locator_t *out,
        ea_t func_ea,
        const char *varname);
 
 
/// Rename a local variable.
/// \param func_ea          function start address
/// \param oldname          old name of the variable
/// \param newname          new name of the variable
/// \return success
/// This is a convenience function.
/// For bulk renaming consider using modify_user_lvars.
 
inline bool rename_lvar(
        ea_t func_ea,
        const char *oldname,
        const char *newname)
{
  lvar_saved_info_t info;
  if ( !locate_lvar(&info.ll, func_ea, oldname) )
    return false;
  info.name = newname;
  return modify_user_lvar_info(func_ea, MLI_NAME, info);
}
 
//-------------------------------------------------------------------------
/// User-defined function calls
struct udcall_t
{
  qstring name;         // name of the function
  tinfo_t tif;          // function prototype
  DECLARE_COMPARISONS(udcall_t)
  {
    int code = ::compare(name, r.name);
    if ( code == 0 )
      code = ::compare(tif, r.tif);
    return code;
  }
 
  bool empty() const { return name.empty() && tif.empty(); }
};
 
// All user-defined function calls (map address -> udcall)
typedef std::map<ea_t, udcall_t> udcall_map_t;
 
/// Restore user defined function calls from the database.
/// \param udcalls ptr to output buffer
/// \param func_ea entry address of the function
/// \return success
 
bool hexapi restore_user_defined_calls(udcall_map_t *udcalls, ea_t func_ea);
 
 
/// Save user defined local function calls into the database.
/// \param func_ea entry address of the function
/// \param udcalls user-specified info about user defined function calls
 
void hexapi save_user_defined_calls(ea_t func_ea, const udcall_map_t &udcalls);
 
 
/// Convert function type declaration into internal structure
/// \param udc    - pointer to output structure
/// \param decl   - function type declaration
/// \param silent - if TRUE: do not show warning in case of incorrect type
/// \return success
 
bool hexapi parse_user_call(udcall_t *udc, const char *decl, bool silent);
 
 
/// try to generate user-defined call for an instruction
/// \return \ref MERR_ code:
///   MERR_OK      - user-defined call generated
///   else         - error (MERR_INSN == inacceptable udc.tif)
 
merror_t hexapi convert_to_user_call(const udcall_t &udc, codegen_t &cdg);
 
 
//-------------------------------------------------------------------------
/// Generic microcode generator class.
/// An instance of a derived class can be registered to be used for
/// non-standard microcode generation. Before microcode generation for an
/// instruction all registered object will be visited by the following way:
///   if ( filter->match(cdg) )
///     code = filter->apply(cdg);
///   if ( code == MERR_OK )
///     continue;     // filter generated microcode, go to the next instruction
struct microcode_filter_t
{
  /// check if the filter object is to be applied
  /// \return success
  virtual bool match(codegen_t &cdg) = 0;
 
  /// generate microcode for an instruction
  /// \return MERR_... code:
  ///   MERR_OK      - user-defined microcode generated, go to the next instruction
  ///   MERR_INSN    - not generated - the caller should try the standard way
  ///   else         - error
  virtual merror_t apply(codegen_t &cdg) = 0;
};
 
/// register/unregister non-standard microcode generator
/// \param filter  - microcode generator object
/// \param install - TRUE - register the object, FALSE - unregister
/// \return success
bool hexapi install_microcode_filter(microcode_filter_t *filter, bool install=true);
 
//-------------------------------------------------------------------------
/// Abstract class: User-defined call generator
/// derived classes should implement method 'match'
class udc_filter_t : public microcode_filter_t
{
  udcall_t udc;
 
public:
  ~udc_filter_t() { cleanup(); }
 
  /// Cleanup the filter
  /// This function properly clears type information associated to this filter.
  void hexapi cleanup(void);
 
  /// return true if the filter object should be applied to given instruction
  virtual bool match(codegen_t &cdg) override = 0;
 
  bool hexapi init(const char *decl);
  virtual merror_t hexapi apply(codegen_t &cdg) override;
 
  bool empty(void) const { return udc.empty(); }
};
 
//-------------------------------------------------------------------------
typedef size_t mbitmap_t;
const size_t bitset_width = sizeof(mbitmap_t) * CHAR_BIT;
const size_t bitset_align = bitset_width - 1;
const size_t bitset_shift = 6;
 
/// Bit set class. See https://en.wikipedia.org/wiki/Bit_array
class bitset_t
{
  mbitmap_t *bitmap;    ///< pointer to bitmap
  size_t high;          ///< highest bit+1 (multiply of bitset_width)
 
public:
  bitset_t(void) : bitmap(nullptr), high(0) {}
  hexapi bitset_t(const bitset_t &m);          // copy constructor
  ~bitset_t(void)
  {
    qfree(bitmap);
    bitmap = nullptr;
  }
  void swap(bitset_t &r)
  {
    std::swap(bitmap, r.bitmap);
    std::swap(high, r.high);
  }
  bitset_t &operator=(const bitset_t &m) { return copy(m); }
  bitset_t &hexapi copy(const bitset_t &m);    // assignment operator
  bool hexapi add(int bit);                    // add a bit
  bool hexapi add(int bit, int width);         // add bits
  bool hexapi add(const bitset_t &ml);         // add another bitset
  bool hexapi sub(int bit);                    // delete a bit
  bool hexapi sub(int bit, int width);         // delete bits
  bool hexapi sub(const bitset_t &ml);         // delete another bitset
  bool hexapi cut_at(int maxbit);              // delete bits >= maxbit
  void hexapi shift_down(int shift);           // shift bits down
  bool hexapi has(int bit) const;       // test presence of a bit
  bool hexapi has_all(int bit, int width) const; // test presence of bits
  bool hexapi has_any(int bit, int width) const; // test presence of bits
  void print(
        qstring *vout,
        int (*get_bit_name)(qstring *out, int bit, int width, void *ud)=nullptr,
        void *ud=nullptr) const;
  const char *hexapi dstr() const;
  bool hexapi empty(void) const;        // is empty?
  int hexapi count(void) const;         // number of set bits
  int hexapi count(int bit) const;      // get number set bits starting from 'bit'
  int hexapi last(void) const;          // get the number of the last bit (-1-no bits)
  void clear(void) { high = 0; }        // make empty
  void hexapi fill_with_ones(int maxbit);
  bool hexapi fill_gaps(int total_nbits);
  bool hexapi has_common(const bitset_t &ml) const; // has common elements?
  bool hexapi intersect(const bitset_t &ml);    // intersect sets. returns true if changed
  bool hexapi is_subset_of(const bitset_t &ml) const; // is subset of?
  bool includes(const bitset_t &ml) const { return ml.is_subset_of(*this); }
  void extract(intvec_t &out) const;
  DECLARE_COMPARISONS(bitset_t);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  class iterator
  {
    friend class bitset_t;
    int i;
  public:
    iterator(int n=-1) : i(n) {}
    bool operator==(const iterator &n) const { return i == n.i; }
    bool operator!=(const iterator &n) const { return i != n.i; }
    int operator*(void) const { return i; }
  };
  typedef iterator const_iterator;
  iterator itat(int n) const { return iterator(goup(n)); }
  iterator begin(void) const { return itat(0); }
  iterator end(void)   const { return iterator(high); }
  int front(void)      const { return *begin(); }
  int back(void)       const { return *end(); }
  void inc(iterator &p, int n=1) const { p.i = goup(p.i+n); }
private:
  int hexapi goup(int reg) const;
};
DECLARE_TYPE_AS_MOVABLE(bitset_t);
typedef qvector<bitset_t> array_of_bitsets;
 
//-------------------------------------------------------------------------
template <class T>
struct ivl_tpl  // an interval
{
  ivl_tpl() = delete;
public:
  T off;
  T size;
  ivl_tpl(T _off, T _size) : off(_off), size(_size) {}
  bool valid() const { return last() >= off; }
  T end() const { return off + size; }
  T last() const { return off + size - 1; }
 
  DEFINE_MEMORY_ALLOCATION_FUNCS()
};
 
//-------------------------------------------------------------------------
typedef ivl_tpl<uval_t> uval_ivl_t;
struct ivl_t : public uval_ivl_t
{
private:
  typedef ivl_tpl<uval_t> inherited;
 
public:
  ivl_t(uval_t _off=0, uval_t _size=0) : inherited(_off,_size) {}
  bool empty(void) const { return size == 0; }
  void clear(void) { size = 0; }
  void print(qstring *vout) const;
  const char *hexapi dstr(void) const;
 
  bool extend_to_cover(const ivl_t &r) // extend interval to cover 'r'
  {
    uval_t new_end = end();
    bool changed = false;
    if ( off > r.off )
    {
      off = r.off;
      changed = true;
    }
    if ( new_end < r.end() )
    {
      new_end = r.end();
      changed = true;
    }
    if ( changed )
      size = new_end - off;
    return changed;
  }
  void intersect(const ivl_t &r)
  {
    uval_t new_off = qmax(off, r.off);
    uval_t new_end = end();
    if ( new_end > r.end() )
      new_end = r.end();
    if ( new_off < new_end )
    {
      off = new_off;
      size = new_end - off;
    }
    else
    {
      size = 0;
    }
  }
 
  // do *this and ivl overlap?
  bool overlap(const ivl_t &ivl) const
  {
    return interval::overlap(off, size, ivl.off, ivl.size);
  }
  // does *this include ivl?
  bool includes(const ivl_t &ivl) const
  {
    return interval::includes(off, size, ivl.off, ivl.size);
  }
  // does *this contain off2?
  bool contains(uval_t off2) const
  {
    return interval::contains(off, size, off2);
  }
 
  DECLARE_COMPARISONS(ivl_t);
  static const ivl_t allmem;
#define ALLMEM ivl_t::allmem
};
DECLARE_TYPE_AS_MOVABLE(ivl_t);
 
//-------------------------------------------------------------------------
struct ivl_with_name_t
{
  ivl_t ivl;
  const char *whole;            // name of the whole interval
  const char *part;             // prefix to use for parts of the interval (e.g. sp+4)
  ivl_with_name_t(): ivl(0, BADADDR), whole("<unnamed inteval>"), part(nullptr) {}
  DEFINE_MEMORY_ALLOCATION_FUNCS()
};
 
//-------------------------------------------------------------------------
template <class Ivl, class T>
class ivlset_tpl // set of intervals
{
public:
  typedef qvector<Ivl> bag_t;
 
protected:
  bag_t bag;
  bool verify(void) const;
  // we do not store the empty intervals in bag so size == 0 denotes
  // MAX_VALUE<T>+1, e.g. 0x100000000 for uint32
  static bool ivl_all_values(const Ivl &ivl) { return ivl.off == 0 && ivl.size == 0; }
 
public:
  ivlset_tpl(void) {}
  ivlset_tpl(const Ivl &ivl) { if ( ivl.valid() ) bag.push_back(ivl); }
  DEFINE_MEMORY_ALLOCATION_FUNCS()
 
  void swap(ivlset_tpl &r) { bag.swap(r.bag); }
  const Ivl &getivl(int idx) const { return bag[idx]; }
  const Ivl &lastivl(void) const { return bag.back(); }
  size_t nivls(void) const { return bag.size(); }
  bool empty(void) const { return bag.empty(); }
  void clear(void) { bag.clear(); }
  void qclear(void) { bag.qclear(); }
  bool all_values() const { return nivls() == 1 && ivl_all_values(bag[0]); }
  void set_all_values() { clear(); bag.push_back(Ivl(0, 0)); }
  bool single_value() const { return nivls() == 1 && bag[0].size == 1; }
  bool single_value(T v) const { return single_value() && bag[0].off == v; }
 
  bool operator==(const Ivl &v) const { return nivls() == 1 && bag[0] == v; }
  bool operator!=(const Ivl &v) const { return !(*this == v); }
 
  typedef typename bag_t::iterator iterator;
  typedef typename bag_t::const_iterator const_iterator;
  const_iterator begin(void) const { return bag.begin(); }
  const_iterator end(void)   const { return bag.end(); }
  iterator begin(void) { return bag.begin(); }
  iterator end(void)   { return bag.end(); }
};
 
//-------------------------------------------------------------------------
/// Set of address intervals.
/// Bit arrays are efficient only for small sets. Potentially huge
/// sets, like memory ranges, require another representation.
/// ivlset_t is used for a list of memory locations in our decompiler.
typedef ivlset_tpl<ivl_t, uval_t> uval_ivl_ivlset_t;
struct ivlset_t : public uval_ivl_ivlset_t
{
  typedef ivlset_tpl<ivl_t, uval_t> inherited;
  ivlset_t() {}
  ivlset_t(const ivl_t &ivl) : inherited(ivl) {}
  bool hexapi add(const ivl_t &ivl);
  bool add(ea_t ea, asize_t size) { return add(ivl_t(ea, size)); }
  bool hexapi add(const ivlset_t &ivs);
  bool hexapi addmasked(const ivlset_t &ivs, const ivl_t &mask);
  bool hexapi sub(const ivl_t &ivl);
  bool sub(ea_t ea, asize_t size) { return sub(ivl_t(ea, size)); }
  bool hexapi sub(const ivlset_t &ivs);
  bool hexapi has_common(const ivl_t &ivl, bool strict=false) const;
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr(void) const;
  asize_t hexapi count(void) const;
  bool hexapi has_common(const ivlset_t &ivs) const;
  bool hexapi contains(uval_t off) const;
  bool hexapi includes(const ivlset_t &ivs) const;
  bool hexapi intersect(const ivlset_t &ivs);
 
  DECLARE_COMPARISONS(ivlset_t);
 
};
DECLARE_TYPE_AS_MOVABLE(ivlset_t);
typedef qvector<ivlset_t> array_of_ivlsets;
//-------------------------------------------------------------------------
// We use bitset_t to keep list of registers.
// This is the most optimal storage for them.
class rlist_t : public bitset_t
{
public:
  rlist_t(void) {}
  rlist_t(const rlist_t &m) : bitset_t(m)
  {
  }
  rlist_t(mreg_t reg, int width) { add(reg, width); }
  ~rlist_t(void) {}
  rlist_t &operator=(const rlist_t &) = default;
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr() const;
};
DECLARE_TYPE_AS_MOVABLE(rlist_t);
 
//-------------------------------------------------------------------------
// Microlist: list of register and memory locations
struct mlist_t
{
  rlist_t reg;         // registers
  ivlset_t mem;        // memory locations
 
  mlist_t(void) {}
  mlist_t(const ivl_t &ivl) : mem(ivl) {}
  mlist_t(mreg_t r, int size) : reg(r, size) {}
 
  void swap(mlist_t &r) { reg.swap(r.reg); mem.swap(r.mem); }
  bool hexapi addmem(ea_t ea, asize_t size);
  bool add(mreg_t r, int size) { return add(mlist_t(r, size)); } // also see append_def_list()
  bool add(const rlist_t &r)   { return reg.add(r); }
  bool add(const ivl_t &ivl)   { return add(mlist_t(ivl)); }
  bool add(const mlist_t &lst) { return reg.add(lst.reg) | mem.add(lst.mem); }
  bool sub(mreg_t r, int size) { return sub(mlist_t(r, size)); }
  bool sub(const ivl_t &ivl)   { return sub(mlist_t(ivl)); }
  bool sub(const mlist_t &lst) { return reg.sub(lst.reg) | mem.sub(lst.mem); }
  asize_t count(void) const { return reg.count() + mem.count(); }
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr() const;
  bool empty(void) const { return reg.empty() && mem.empty(); }
  void clear(void) { reg.clear(); mem.clear(); }
  bool has(mreg_t r) const { return reg.has(r); }
  bool has_all(mreg_t r, int size) const { return reg.has_all(r, size); }
  bool has_any(mreg_t r, int size) const { return reg.has_any(r, size); }
  bool has_memory(void) const { return !mem.empty(); }
  bool has_allmem(void) const { return mem == ALLMEM; }
  bool has_common(const mlist_t &lst) const { return reg.has_common(lst.reg) || mem.has_common(lst.mem); }
  bool includes(const mlist_t &lst) const { return reg.includes(lst.reg) && mem.includes(lst.mem); }
  bool intersect(const mlist_t &lst) { return reg.intersect(lst.reg) | mem.intersect(lst.mem); }
  bool is_subset_of(const mlist_t &lst) const { return lst.includes(*this); }
 
  DECLARE_COMPARISONS(mlist_t);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
};
DECLARE_TYPE_AS_MOVABLE(mlist_t);
typedef qvector<mlist_t> mlistvec_t;
DECLARE_TYPE_AS_MOVABLE(mlistvec_t);
 
//-------------------------------------------------------------------------
/// Get list of temporary registers.
/// Tempregs are temporary registers that are used during code generation.
/// They do not map to regular processor registers. They are used only to
/// store temporary values during execution of one instruction.
/// Tempregs may not be used to pass a value from one block to another.
/// In other words, at the end of a block all tempregs must be dead.
const mlist_t &hexapi get_temp_regs(void);
 
/// Is a kernel register?
/// Kernel registers are temporary registers that can be used freely.
/// They may be used to store values that cross instruction or basic block
/// boundaries. Kernel registers do not map to regular processor registers.
/// See also \ref mba_t::alloc_kreg()
bool hexapi is_kreg(mreg_t r);
 
/// Map a processor register to a microregister.
/// \param reg   processor register number
/// \return microregister register id or mr_none
mreg_t hexapi reg2mreg(int reg);
 
/// Map a microregister to a processor register.
/// \param reg   microregister number
/// \param width size of microregister in bytes
/// \return processor register id or -1
int hexapi mreg2reg(mreg_t reg, int width);
 
/// Get the microregister name.
/// \param out   output buffer, may be nullptr
/// \param reg   microregister number
/// \param width size of microregister in bytes. may be bigger than the real
///              register size.
/// \param ud    reserved, must be nullptr
/// \return width of the printed register. this value may be less than
///         the WIDTH argument.
 
int hexapi get_mreg_name(qstring *out, mreg_t reg, int width, void *ud=nullptr);
 
//-------------------------------------------------------------------------
/// User defined callback to optimize individual microcode instructions
struct optinsn_t
{
  /// Optimize an instruction.
  /// \param blk current basic block. maybe nullptr, which means that
  ///            the instruction must be optimized without context
  /// \param ins instruction to optimize; it is always a top-level instruction.
  ///            the callback may not delete the instruction but may
  ///            convert it into nop (see mblock_t::make_nop). to optimize
  ///            sub-instructions, visit them using minsn_visitor_t.
  ///            sub-instructions may not be converted into nop but
  ///            can be converted to "mov x,x". for example:
  ///               add x,0,x => mov x,x
  ///            this callback may change other instructions in the block,
  ///            but should do this with care, e.g. to no break the
  ///            propagation algorithm if called with OPTI_NO_LDXOPT.
  /// \param optflags combination of \ref OPTI_ bits
  /// \return number of changes made to the instruction.
  ///         if after this call the instruction's use/def lists have changed,
  ///         you must mark the block level lists as dirty (see mark_lists_dirty)
  virtual int idaapi func(mblock_t *blk, minsn_t *ins, int optflags) = 0;
};
 
/// Install an instruction level custom optimizer
/// \param opt an instance of optinsn_t. cannot be destroyed before calling
///        remove_optinsn_handler().
void hexapi install_optinsn_handler(optinsn_t *opt);
 
/// Remove an instruction level custom optimizer
bool hexapi remove_optinsn_handler(optinsn_t *opt);
 
/// User defined callback to optimize microcode blocks
struct optblock_t
{
  /// Optimize a block.
  /// This function usually performs the optimizations that require analyzing
  /// the entire block and/or its neighbors. For example it can recognize
  /// patterns and perform conversions like:
  /// b0:                                 b0:
  ///    ...                                 ...
  ///    jnz x, 0, @b2      =>               jnz x, 0, @b2
  /// b1:                                 b1:
  ///    add x, 0, y                         mov x, y
  ///    ...                                 ...
  /// \param blk Basic block to optimize as a whole.
  /// \return number of changes made to the block. See also mark_lists_dirty.
  virtual int idaapi func(mblock_t *blk) = 0;
};
 
/// Install a block level custom optimizer.
/// \param opt an instance of optblock_t. cannot be destroyed before calling
///        remove_optblock_handler().
void hexapi install_optblock_handler(optblock_t *opt);
 
/// Remove a block level custom optimizer
bool hexapi remove_optblock_handler(optblock_t *opt);
 
 
//-------------------------------------------------------------------------
// abstract graph interface
class simple_graph_t : public gdl_graph_t
{
public:
  qstring title;
  bool colored_gdl_edges;
private:
  friend class iterator;
  virtual int goup(int node) const newapi;
};
 
//-------------------------------------------------------------------------
// Since our data structures are quite complex, we use the visitor pattern
// in many of our algorthims. This functionality is available for plugins too.
// https://en.wikipedia.org/wiki/Visitor_pattern
 
// All our visitor callbacks return an integer value.
// Visiting is interrupted as soon an the return value is non-zero.
// This non-zero value is returned as the result of the for_all_... function.
// If for_all_... returns 0, it means that it successfully visited all items.
 
/// The context info used by visitors
struct op_parent_info_t
{
  mba_t *mba;          // current microcode
  mblock_t *blk;       // current block
  minsn_t *topins;     // top level instruction (parent of curins or curins itself)
  minsn_t *curins;     // currently visited instruction
  op_parent_info_t(
        mba_t *_mba=nullptr,
        mblock_t *_blk=nullptr,
        minsn_t *_topins=nullptr)
    : mba(_mba), blk(_blk), topins(_topins), curins(nullptr) {}
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  bool really_alloc(void) const;
};
 
/// Micro instruction visitor.
/// See mba_t::for_all_topinsns, minsn_t::for_all_insns,
///     mblock_::for_all_insns, mba_t::for_all_insns
struct minsn_visitor_t : public op_parent_info_t
{
  minsn_visitor_t(
        mba_t *_mba=nullptr,
        mblock_t *_blk=nullptr,
        minsn_t *_topins=nullptr)
    : op_parent_info_t(_mba, _blk, _topins) {}
  virtual int idaapi visit_minsn(void) = 0;
};
 
/// Micro operand visitor.
/// See mop_t::for_all_ops, minsn_t::for_all_ops, mblock_t::for_all_insns,
///     mba_t::for_all_insns
struct mop_visitor_t : public op_parent_info_t
{
  mop_visitor_t(
        mba_t *_mba=nullptr,
        mblock_t *_blk=nullptr,
        minsn_t *_topins=nullptr)
    : op_parent_info_t(_mba, _blk, _topins), prune(false) {}
  /// Should skip sub-operands of the current operand?
  /// visit_mop() may set 'prune=true' for that.
  bool prune;
  virtual int idaapi visit_mop(mop_t *op, const tinfo_t *type, bool is_target) = 0;
};
 
/// Scattered mop: visit each of the scattered locations as a separate mop.
/// See mop_t::for_all_scattered_submops
struct scif_visitor_t
{
  virtual int idaapi visit_scif_mop(const mop_t &r, int off) = 0;
};
 
// Used operand visitor.
// See mblock_t::for_all_uses
struct mlist_mop_visitor_t
{
  minsn_t *topins;
  minsn_t *curins;
  bool changed;
  mlist_t *list;
  mlist_mop_visitor_t(void): topins(nullptr), curins(nullptr), changed(false), list(nullptr) {}
  virtual int idaapi visit_mop(mop_t *op) = 0;
};
 
//-------------------------------------------------------------------------
/// Instruction operand types
 
typedef uint8 mopt_t;
const mopt_t
  mop_z   = 0,  ///< none
  mop_r   = 1,  ///< register (they exist until MMAT_LVARS)
  mop_n   = 2,  ///< immediate number constant
  mop_str = 3,  ///< immediate string constant (user representation)
  mop_d   = 4,  ///< result of another instruction
  mop_S   = 5,  ///< local stack variable (they exist until MMAT_LVARS)
  mop_v   = 6,  ///< global variable
  mop_b   = 7,  ///< micro basic block (mblock_t)
  mop_f   = 8,  ///< list of arguments
  mop_l   = 9,  ///< local variable
  mop_a   = 10, ///< mop_addr_t: address of operand (mop_l, mop_v, mop_S, mop_r)
  mop_h   = 11, ///< helper function
  mop_c   = 12, ///< mcases
  mop_fn  = 13, ///< floating point constant
  mop_p   = 14, ///< operand pair
  mop_sc  = 15; ///< scattered
 
const int NOSIZE = -1; ///< wrong or unexisting operand size
 
//-------------------------------------------------------------------------
/// Reference to a local variable. Used by mop_l
struct lvar_ref_t
{
  /// Pointer to the parent mba_t object.
  /// Since we need to access the 'mba->vars' array in order to retrieve
  /// the referenced variable, we keep a pointer to mba_t here.
  /// Note: this means this class and consequently mop_t, minsn_t, mblock_t
  ///       are specific to a mba_t object and cannot migrate between
  ///       them. fortunately this is not something we need to do.
  ///       second, lvar_ref_t's appear only after MMAT_LVARS.
  mba_t *const mba;
  sval_t off;           ///< offset from the beginning of the variable
  int idx;              ///< index into mba->vars
  lvar_ref_t(mba_t *m, int i, sval_t o=0) : mba(m), off(o), idx(i) {}
  lvar_ref_t(const lvar_ref_t &r) : mba(r.mba), off(r.off), idx(r.idx) {}
  lvar_ref_t &operator=(const lvar_ref_t &r)
  {
    off = r.off;
    idx = r.idx;
    return *this;
  }
  DECLARE_COMPARISONS(lvar_ref_t);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  void swap(lvar_ref_t &r)
  {
    std::swap(off, r.off);
    std::swap(idx, r.idx);
  }
  lvar_t &hexapi var(void) const;       ///< Retrieve the referenced variable
};
 
//-------------------------------------------------------------------------
/// Reference to a stack variable. Used for mop_S
struct stkvar_ref_t
{
  /// Pointer to the parent mba_t object.
  /// We need it in order to retrieve the referenced stack variable.
  /// See notes for lvar_ref_t::mba.
  mba_t *const mba;
 
  /// Offset to the stack variable from the bottom of the stack frame.
  /// It is called 'decompiler stkoff' and it is different from IDA stkoff.
  /// See a note and a picture about 'decompiler stkoff' below.
  sval_t off;
 
  stkvar_ref_t(mba_t *m, sval_t o) : mba(m), off(o) {}
  DECLARE_COMPARISONS(stkvar_ref_t);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  void swap(stkvar_ref_t &r)
  {
    std::swap(off, r.off);
  }
  /// Retrieve the referenced stack variable.
  /// \param p_off if specified, will hold IDA stkoff after the call.
  /// \return pointer to the stack variable
  member_t *hexapi get_stkvar(uval_t *p_off=nullptr) const;
};
 
//-------------------------------------------------------------------------
/// Scattered operand info. Used for mop_sc
struct scif_t : public vdloc_t
{
  /// Pointer to the parent mba_t object.
  /// Some operations may convert a scattered operand into something simpler,
  /// (a stack operand, for example). We will need to create stkvar_ref_t at
  /// that moment, this is why we need this pointer.
  /// See notes for lvar_ref_t::mba.
  mba_t *mba;
 
  /// Usually scattered operands are created from a function prototype,
  /// which has the name information. We preserve it and use it to name
  /// the corresponding local variable.
  qstring name;
 
  /// Scattered operands always have type info assigned to them
  /// because without it we won't be able to manipulte them.
  tinfo_t type;
 
  scif_t(mba_t *_mba, tinfo_t *tif, qstring *n=nullptr) : mba(_mba)
  {
    if ( n != nullptr )
      n->swap(name);
    tif->swap(type);
  }
  scif_t &operator =(const vdloc_t &loc)
  {
    *(vdloc_t *)this = loc;
    return *this;
  }
};
 
//-------------------------------------------------------------------------
/// An integer constant. Used for mop_n
/// We support 64-bit values but 128-bit values can be represented with mop_p
struct mnumber_t : public operand_locator_t
{
  uint64 value;
  uint64 org_value;     // original value before changing the operand size
  mnumber_t(uint64 v, ea_t _ea=BADADDR, int n=0)
    : operand_locator_t(_ea, n), value(v), org_value(v) {}
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  DECLARE_COMPARISONS(mnumber_t)
  {
    if ( value < r.value )
      return -1;
    if ( value > r.value )
      return -1;
    return 0;
  }
  // always use this function instead of manually modifying the 'value' field
  void update_value(uint64 val64)
  {
    value = val64;
    org_value = val64;
  }
};
 
//-------------------------------------------------------------------------
/// Floating point constant. Used for mop_fn
/// For more details, please see the ieee.h file from IDA SDK.
struct fnumber_t
{
  fpvalue_t fnum;       ///< Internal representation of the number
  int nbytes;           ///< Original size of the constant in bytes
  operator       uint16 *(void)       { return fnum.w; }
  operator const uint16 *(void) const { return fnum.w; }
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr() const;
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  DECLARE_COMPARISONS(fnumber_t)
  {
    return ecmp(fnum, r.fnum);
  }
};
 
//-------------------------------------------------------------------------
/// \defgroup SHINS_ Bits to control how we print instructions
//@{
#define SHINS_NUMADDR 0x01 ///< display definition addresses for numbers
#define SHINS_VALNUM  0x02 ///< display value numbers
#define SHINS_SHORT   0x04 ///< do not display use-def chains and other attrs
#define SHINS_LDXEA   0x08 ///< display address of ldx expressions (not used)
//@}
 
//-------------------------------------------------------------------------
/// How to handle side effect of change_size()
/// Sometimes we need to create a temporary operand and change its size in order
/// to check some hypothesis. If we revert our changes, we do not want that the
/// database (global variables, stack frame, etc) changes in any manner.
enum side_effect_t
{
  NO_SIDEFF,          ///< change operand size but ignore side effects
                      ///< if you decide to keep the changed operand,
                      ///< handle_new_size() must be called
  WITH_SIDEFF,        ///< change operand size and handle side effects
  ONLY_SIDEFF,        ///< only handle side effects
  ANY_REGSIZE = 0x80, ///< any register size is permitted
  ANY_FPSIZE = 0x100, ///< any size of floating operand is permitted
};
 
//-------------------------------------------------------------------------
/// A microinstruction operand.
/// This is the smallest building block of our microcode.
/// Operands will be part of instructions, which are then grouped into basic blocks.
/// The microcode consists of an array of such basic blocks + some additional info.
class mop_t
{
  void hexapi copy(const mop_t &rop);
public:
  /// Operand type.
  mopt_t t;
 
  /// Operand properties.
  uint8 oprops;
#define OPROP_IMPDONE 0x01 ///< imported operand (a pointer) has been dereferenced
#define OPROP_UDT     0x02 ///< a struct or union
#define OPROP_FLOAT   0x04 ///< possibly floating value
#define OPROP_CCFLAGS 0x08 ///< mop_n: a pc-relative value
                           ///< mop_a: an address obtained from a relocation
                           ///< else: value of a condition code register (like mr_cc)
#define OPROP_UDEFVAL 0x10 ///< uses undefined value
#define OPROP_LOWADDR 0x20 ///< a low address offset
 
  /// Value number.
  /// Zero means unknown.
  /// Operands with the same value number are equal.
  uint16 valnum;
 
  /// Operand size.
  /// Usually it is 1,2,4,8 or NOSIZE but for UDTs other sizes are permitted
  int size;
 
  /// The following union holds additional details about the operand.
  /// Depending on the operand type different kinds of info are stored.
  /// You should access these fields only after verifying the operand type.
  /// All pointers are owned by the operand and are freed by its destructor.
  union
  {
    mreg_t r;           // mop_r   register number
    mnumber_t *nnn;     // mop_n   immediate value
    minsn_t *d;         // mop_d   result (destination) of another instruction
    stkvar_ref_t *s;    // mop_S   stack variable
    ea_t g;             // mop_v   global variable (its linear address)
    int b;              // mop_b   block number (used in jmp,call instructions)
    mcallinfo_t *f;     // mop_f   function call information
    lvar_ref_t *l;      // mop_l   local variable
    mop_addr_t *a;      // mop_a   variable whose address is taken
    char *helper;       // mop_h   helper function name
    char *cstr;         // mop_str utf8 string constant, user representation
    mcases_t *c;        // mop_c   cases
    fnumber_t *fpc;     // mop_fn  floating point constant
    mop_pair_t *pair;   // mop_p   operand pair
    scif_t *scif;       // mop_sc  scattered operand info
  };
  // -- End of data fields, member function declarations follow:
 
  void set_impptr_done(void) { oprops |= OPROP_IMPDONE; }
  void set_udt(void)         { oprops |= OPROP_UDT; }
  void set_undef_val(void)   { oprops |= OPROP_UDEFVAL; }
  void set_lowaddr(void)     { oprops |= OPROP_LOWADDR; }
  bool is_impptr_done(void) const { return (oprops & OPROP_IMPDONE) != 0; }
  bool is_udt(void)         const { return (oprops & OPROP_UDT) != 0; }
  bool probably_floating(void) const { return (oprops & OPROP_FLOAT) != 0; }
  bool is_undef_val(void)   const { return (oprops & OPROP_UDEFVAL) != 0; }
  bool is_lowaddr(void)     const { return (oprops & OPROP_LOWADDR) != 0; }
  bool is_ccflags(void) const
  {
    return (oprops & OPROP_CCFLAGS) != 0
        && (t == mop_l || t == mop_v || t == mop_S || t == mop_r);
  }
  bool is_pcval(void) const
  {
    return t == mop_n && (oprops & OPROP_CCFLAGS) != 0;
  }
  bool is_glbaddr_from_fixup() const
  {
    return is_glbaddr() && (oprops & OPROP_CCFLAGS) != 0;
  }
 
  mop_t(void) { zero(); }
  mop_t(const mop_t &rop) { copy(rop); }
  mop_t(mreg_t _r, int _s) : t(mop_r), oprops(0), valnum(0), size(_s), r(_r) {}
  mop_t &operator=(const mop_t &rop) { return assign(rop); }
  mop_t &hexapi assign(const mop_t &rop);
  ~mop_t(void)
  {
    erase();
  }
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  void zero() { t = mop_z; oprops = 0; valnum = 0; size = NOSIZE; nnn = nullptr; }
  void hexapi swap(mop_t &rop);
  void hexapi erase(void);
  void erase_but_keep_size(void) { int s2 = size; erase(); size = s2; }
 
  void hexapi print(qstring *vout, int shins_flags=SHINS_SHORT|SHINS_VALNUM) const;
  const char *hexapi dstr() const; // use this function for debugging
 
  //-----------------------------------------------------------------------
  // Operand creation
  //-----------------------------------------------------------------------
  /// Create operand from mlist_t.
  /// Example: if LST contains 4 bits for R0.4, our operand will be
  ///          (t=mop_r, r=R0, size=4)
  /// \param mba pointer to microcode
  /// \param lst list of locations
  /// \param fullsize mba->fullsize
  /// \return success
  bool hexapi create_from_mlist(mba_t *mba, const mlist_t &lst, sval_t fullsize);
 
  /// Create operand from ivlset_t.
  /// Example: if IVS contains [glbvar..glbvar+4), our operand will be
  ///          (t=mop_v, g=&glbvar, size=4)
  /// \param mba pointer to microcode
  /// \param ivs set of memory intervals
  /// \param fullsize mba->fullsize
  /// \return success
  bool hexapi create_from_ivlset(mba_t *mba, const ivlset_t &ivs, sval_t fullsize);
 
  /// Create operand from vdloc_t.
  /// Example: if LOC contains (type=ALOC_REG1, r=R0), our operand will be
  ///          (t=mop_r, r=R0, size=_SIZE)
  /// \param mba pointer to microcode
  /// \param loc location
  /// \param _size operand size
  /// Note: this function cannot handle scattered locations.
  /// \return success
  void hexapi create_from_vdloc(mba_t *mba, const vdloc_t &loc, int _size);
 
  /// Create operand from scattered vdloc_t.
  /// Example: if LOC is (ALOC_DIST, {EAX.4, EDX.4}) and TYPE is _LARGE_INTEGER,
  /// our operand will be
  ///          (t=mop_sc, scif={EAX.4, EDX.4})
  /// \param mba pointer to microcode
  /// \param name name of the operand, if available
  /// \param type type of the operand, must be present
  /// \param loc a scattered location
  /// \return success
  void hexapi create_from_scattered_vdloc(
        mba_t *mba,
        const char *name,
        tinfo_t type,
        const vdloc_t &loc);
 
  /// Create operand from an instruction.
  /// This function creates a nested instruction that can be used as an operand.
  /// Example: if m="add x,y,z", our operand will be (t=mop_d,d=m).
  /// The destination operand of 'add' (z) is lost.
  /// \param m instruction to embed into operand. may not be nullptr.
  void hexapi create_from_insn(const minsn_t *m);
 
  /// Create an integer constant operand.
  /// \param _value value to store in the operand
  /// \param _size size of the value in bytes (1,2,4,8)
  /// \param _ea   address of the processor instruction that made the value
  /// \param opnum operand number of the processor instruction
  void hexapi make_number(uint64 _value, int _size, ea_t _ea=BADADDR, int opnum=0);
 
  /// Create a floating point constant operand.
  /// \param bytes pointer to the floating point value as used by the current
  ///              processor (e.g. for x86 it must be in IEEE 754)
  /// \param _size number of bytes occupied by the constant.
  /// \return success
  bool hexapi make_fpnum(const void *bytes, size_t _size);
 
  /// Create a register operand without erasing previous data.
  /// \param reg  micro register number
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void _make_reg(mreg_t reg)
  {
    t = mop_r;
    r = reg;
  }
  void _make_reg(mreg_t reg, int _size)
  {
    t = mop_r;
    r = reg;
    size = _size;
  }
  /// Create a register operand.
  void make_reg(mreg_t reg) { erase(); _make_reg(reg); }
  void make_reg(mreg_t reg, int _size) { erase(); _make_reg(reg, _size); }
 
  /// Create a local variable operand.
  /// \param mba pointer to microcode
  /// \param idx index into mba->vars
  /// \param off offset from the beginning of the variable
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void _make_lvar(mba_t *mba, int idx, sval_t off=0)
  {
    t = mop_l;
    l = new lvar_ref_t(mba, idx, off);
  }
 
  /// Create a global variable operand without erasing previous data.
  /// \param ea  address of the variable
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void hexapi _make_gvar(ea_t ea);
  /// Create a global variable operand.
  void hexapi make_gvar(ea_t ea);
 
  /// Create a stack variable operand.
  /// \param mba pointer to microcode
  /// \param off decompiler stkoff
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void _make_stkvar(mba_t *mba, sval_t off)
  {
    t = mop_S;
    s = new stkvar_ref_t(mba, off);
  }
  void make_stkvar(mba_t *mba, sval_t off) { erase(); _make_stkvar(mba, off); }
 
  /// Create pair of registers.
  /// \param loreg register holding the low part of the value
  /// \param hireg register holding the high part of the value
  /// \param halfsize the size of each of loreg/hireg
  void hexapi make_reg_pair(int loreg, int hireg, int halfsize);
 
  /// Create a nested instruction without erasing previous data.
  /// \param ins pointer to the instruction to encapsulate into the operand
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  /// See also create_from_insn, which is higher level
  void _make_insn(minsn_t *ins);
  /// Create a nested instruction.
  void make_insn(minsn_t *ins) { erase(); _make_insn(ins); }
 
  /// Create a block reference operand without erasing previous data.
  /// \param blknum block number
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void _make_blkref(int blknum)
  {
    t = mop_b;
    b = blknum;
  }
  /// Create a global variable operand.
  void make_blkref(int blknum) { erase(); _make_blkref(blknum); }
 
  /// Create a helper operand.
  /// A helper operand usually keeps a built-in function name like "va_start"
  /// It is essentially just an arbitrary identifier without any additional info.
  void hexapi make_helper(const char *name);
 
  /// Create a constant string operand.
  void _make_strlit(const char *str)
  {
    t = mop_str;
    cstr = ::qstrdup(str);
  }
  void _make_strlit(qstring *str) // str is consumed
  {
    t = mop_str;
    cstr = str->extract();
  }
 
  /// Create a call info operand without erasing previous data.
  /// \param fi callinfo
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void _make_callinfo(mcallinfo_t *fi)
  {
    t = mop_f;
    f = fi;
  }
 
  /// Create a 'switch cases' operand without erasing previous data.
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void _make_cases(mcases_t *_cases)
  {
    t = mop_c;
    c = _cases;
  }
 
  /// Create a pair operand without erasing previous data.
  /// Note: this function does not erase the previous contents of the operand;
  ///       call erase() if necessary
  void _make_pair(mop_pair_t *_pair)
  {
    t = mop_p;
    pair = _pair;
  }
 
  //-----------------------------------------------------------------------
  // Various operand tests
  //-----------------------------------------------------------------------
  bool empty(void) const { return t == mop_z; }
  /// Is a register operand?
  /// See also get_mreg_name()
  bool is_reg(void) const { return t == mop_r; }
  /// Is the specified register?
  bool is_reg(mreg_t _r) const { return t == mop_r && r == _r; }
  /// Is the specified register of the specified size?
  bool is_reg(mreg_t _r, int _size) const { return t == mop_r && r == _r && size == _size; }
  /// Is a list of arguments?
  bool is_arglist(void) const { return t == mop_f; }
  /// Is a condition code?
  bool is_cc(void) const { return is_reg() && r >= mr_cf && r < mr_first; }
  /// Is a bit register?
  /// This includes condition codes and eventually other bit registers
  static bool hexapi is_bit_reg(mreg_t reg);
  bool is_bit_reg(void) const { return is_reg() && is_bit_reg(r); }
  /// Is a kernel register?
  bool is_kreg(void) const;
  /// Is a block reference to the specified block?
  bool is_mob(int serial) const { return t == mop_b && b == serial; }
  /// Is a scattered operand?
  bool is_scattered(void) const { return t == mop_sc; }
  /// Is address of a global memory cell?
  bool is_glbaddr() const;
  /// Is address of the specified global memory cell?
  bool is_glbaddr(ea_t ea) const;
  /// Is address of a stack variable?
  bool is_stkaddr() const;
  /// Is a sub-instruction?
  bool is_insn(void) const { return t == mop_d; }
  /// Is a sub-instruction with the specified opcode?
  bool is_insn(mcode_t code) const;
  /// Has any side effects?
  /// \param include_ldx_and_divs consider ldx/div/mod as having side effects?
  bool has_side_effects(bool include_ldx_and_divs=false) const;
  /// Is it possible for the operand to use aliased memory?
  bool hexapi may_use_aliased_memory(void) const;
 
  /// Are the possible values of the operand only 0 and 1?
  /// This function returns true for 0/1 constants, bit registers,
  /// the result of 'set' insns, etc.
  bool hexapi is01(void) const;
 
  /// Does the high part of the operand consist of the sign bytes?
  /// \param nbytes number of bytes that were sign extended.
  ///               the remaining size-nbytes high bytes must be sign bytes
  /// Example: is_sign_extended_from(xds.4(op.1), 1) -> true
  ///          because the high 3 bytes are certainly sign bits
  bool hexapi is_sign_extended_from(int nbytes) const;
 
  /// Does the high part of the operand consist of zero bytes?
  /// \param nbytes number of bytes that were zero extended.
  ///               the remaining size-nbytes high bytes must be zero
  /// Example: is_zero_extended_from(xdu.8(op.1), 2) -> true
  ///          because the high 6 bytes are certainly zero
  bool hexapi is_zero_extended_from(int nbytes) const;
 
  /// Does the high part of the operand consist of zero or sign bytes?
  bool is_extended_from(int nbytes, bool is_signed) const
  {
    if ( is_signed )
      return is_sign_extended_from(nbytes);
    else
      return is_zero_extended_from(nbytes);
  }
 
  //-----------------------------------------------------------------------
  // Comparisons
  //-----------------------------------------------------------------------
  /// Compare operands.
  /// This is the main comparison function for operands.
  /// \param rop     operand to compare with
  /// \param eqflags combination of \ref EQ_ bits
  bool hexapi equal_mops(const mop_t &rop, int eqflags) const;
  bool operator==(const mop_t &rop) const { return  equal_mops(rop, 0); }
  bool operator!=(const mop_t &rop) const { return !equal_mops(rop, 0); }
 
  /// Lexographical operand comparison.
  /// It can be used to store mop_t in various containers, like std::set
  bool operator <(const mop_t &rop) const { return lexcompare(rop) < 0; }
  friend int lexcompare(const mop_t &a, const mop_t &b) { return a.lexcompare(b); }
  int hexapi lexcompare(const mop_t &rop) const;
 
  //-----------------------------------------------------------------------
  // Visiting operand parts
  //-----------------------------------------------------------------------
  /// Visit the operand and all its sub-operands.
  /// This function visits the current operand as well.
  /// \param mv        visitor object
  /// \param type      operand type
  /// \param is_target is a destination operand?
  int hexapi for_all_ops(
        mop_visitor_t &mv,
        const tinfo_t *type=nullptr,
        bool is_target=false);
 
  /// Visit all sub-operands of a scattered operand.
  /// This function does not visit the current operand, only its sub-operands.
  /// All sub-operands are synthetic and are destroyed after the visitor.
  /// This function works only with scattered operands.
  /// \param sv        visitor object
  int hexapi for_all_scattered_submops(scif_visitor_t &sv) const;
 
  //-----------------------------------------------------------------------
  // Working with mop_n operands
  //-----------------------------------------------------------------------
  /// Retrieve value of a constant integer operand.
  /// These functions can be called only for mop_n operands.
  /// See is_constant() that can be called on any operand.
  uint64 value(bool is_signed) const { return extend_sign(nnn->value, size, is_signed); }
  int64 signed_value(void) const { return value(true); }
  uint64 unsigned_value(void) const { return value(false); }
  void update_numop_value(uint64 val)
  {
    nnn->update_value(extend_sign(val, size, false));
  }
 
  /// Retrieve value of a constant integer operand.
  /// \param out pointer to the output buffer
  /// \param is_signed should treat the value as signed
  /// \return true if the operand is mop_n
  bool hexapi is_constant(uint64 *out=nullptr, bool is_signed=true) const;
 
  bool is_equal_to(uint64 n, bool is_signed=true) const
  {
    uint64 v;
    return is_constant(&v, is_signed) && v == n;
  }
  bool is_zero(void) const { return is_equal_to(0, false); }
  bool is_one(void) const { return is_equal_to(1, false); }
  bool is_positive_constant(void) const
  {
    uint64 v;
    return is_constant(&v, true) && int64(v) > 0;
  }
  bool is_negative_constant(void) const
  {
    uint64 v;
    return is_constant(&v, true) && int64(v) < 0;
  }
 
  //-----------------------------------------------------------------------
  // Working with mop_S operands
  //-----------------------------------------------------------------------
  /// Retrieve the referenced stack variable.
  /// \param p_off if specified, will hold IDA stkoff after the call.
  /// \return pointer to the stack variable
  member_t *get_stkvar(uval_t *p_off) const { return s->get_stkvar(p_off); }
 
  /// Get the referenced stack offset.
  /// This function can also handle mop_sc if it is entirely mapped into
  /// a continuous stack region.
  /// \param p_off the output buffer
  /// \return success
  bool hexapi get_stkoff(sval_t *p_off) const;
 
  //-----------------------------------------------------------------------
  // Working with mop_d operands
  //-----------------------------------------------------------------------
  /// Get subinstruction of the operand.
  /// If the operand has a subinstruction with the specified opcode, return it.
  /// \param code desired opcode
  /// \return pointer to the instruction or nullptr
  const minsn_t *get_insn(mcode_t code) const;
        minsn_t *get_insn(mcode_t code);
 
  //-----------------------------------------------------------------------
  // Transforming operands
  //-----------------------------------------------------------------------
  /// Make the low part of the operand.
  /// This function takes into account the memory endianness (byte sex)
  /// \param width the desired size of the operand part in bytes
  /// \return success
  bool hexapi make_low_half(int width);
 
  /// Make the high part of the operand.
  /// This function takes into account the memory endianness (byte sex)
  /// \param width the desired size of the operand part in bytes
  /// \return success
  bool hexapi make_high_half(int width);
 
  /// Make the first part of the operand.
  /// This function does not care about the memory endianness
  /// \param width the desired size of the operand part in bytes
  /// \return success
  bool hexapi make_first_half(int width);
 
  /// Make the second part of the operand.
  /// This function does not care about the memory endianness
  /// \param width the desired size of the operand part in bytes
  /// \return success
  bool hexapi make_second_half(int width);
 
  /// Shift the operand.
  /// This function shifts only the beginning of the operand.
  /// The operand size will be changed.
  /// Examples: shift_mop(AH.1, -1) -> AX.2
  ///           shift_mop(qword_00000008.8, 4) -> dword_0000000C.4
  ///           shift_mop(xdu.8(op.4), 4) -> #0.4
  ///           shift_mop(#0x12345678.4, 3) -> #12.1
  /// \param offset shift count (the number of bytes to shift)
  /// \return success
  bool hexapi shift_mop(int offset);
 
  /// Change the operand size.
  /// Examples: change_size(AL.1, 2) -> AX.2
  ///           change_size(qword_00000008.8, 4) -> dword_00000008.4
  ///           change_size(xdu.8(op.4), 4) -> op.4
  ///           change_size(#0x12345678.4, 1) -> #0x78.1
  /// \param nsize  new operand size
  /// \param sideff may modify the database because of the size change?
  /// \return success
  bool hexapi change_size(int nsize, side_effect_t sideff=WITH_SIDEFF);
  bool double_size(side_effect_t sideff=WITH_SIDEFF) { return change_size(size*2, sideff); }
 
  /// Move subinstructions with side effects out of the operand.
  /// If we decide to delete an instruction operand, it is a good idea to
  /// call this function. Alternatively we should skip such operands
  /// by calling mop_t::has_side_effects()
  /// For example, if we transform: jnz x, x, @blk => goto @blk
  /// then we must call this function before deleting the X operands.
  /// \param blk  current block
  /// \param top  top level instruction that contains our operand
  /// \param moved_calls pointer to the boolean that will track if all side
  ///                    effects get handled correctly. must be false initially.
  /// \return false failed to preserve a side effect, it is not safe to
  ///               delete the operand
  ///         true  no side effects or successfully preserved them
  bool hexapi preserve_side_effects(
        mblock_t *blk,
        minsn_t *top,
        bool *moved_calls=nullptr);
 
  /// Apply a unary opcode to the operand.
  /// \param mcode   opcode to apply. it must accept 'l' and 'd' operands
  ///                but not 'r'. examples: m_low/m_high/m_xds/m_xdu
  /// \param ea      value of minsn_t::ea for the newly created insruction
  /// \param newsize new operand size
  /// Example: apply_ld_mcode(m_low) will convert op => low(op)
  void hexapi apply_ld_mcode(mcode_t mcode, ea_t ea, int newsize);
  void apply_xdu(ea_t ea, int newsize) { apply_ld_mcode(m_xdu, ea, newsize); }
  void apply_xds(ea_t ea, int newsize) { apply_ld_mcode(m_xds, ea, newsize); }
};
DECLARE_TYPE_AS_MOVABLE(mop_t);
 
/// Pair of operands
class mop_pair_t
{
public:
  mop_t lop;            ///< low operand
  mop_t hop;            ///< high operand
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
};
 
/// Address of an operand (mop_l, mop_v, mop_S, mop_r)
class mop_addr_t : public mop_t
{
public:
  int insize;   // how many bytes of the pointed operand can be read
  int outsize;  // how many bytes of the pointed operand can be written
 
  mop_addr_t(): insize(NOSIZE), outsize(NOSIZE) {}
  mop_addr_t(const mop_addr_t &ra)
    : mop_t(ra), insize(ra.insize), outsize(ra.outsize) {}
  mop_addr_t(const mop_t &ra, int isz, int osz)
    : mop_t(ra), insize(isz), outsize(osz) {}
 
  mop_addr_t &operator=(const mop_addr_t &rop)
  {
    *(mop_t *)this = mop_t(rop);
    insize = rop.insize;
    outsize = rop.outsize;
    return *this;
  }
  int lexcompare(const mop_addr_t &ra) const
  {
    int code = mop_t::lexcompare(ra);
    return code    != 0          ? code
         : insize  != ra.insize  ? (insize-ra.insize)
         : outsize != ra.outsize ? (outsize-ra.outsize)
         :                         0;
  }
};
 
/// A call argument
class mcallarg_t : public mop_t // #callarg
{
public:
  ea_t ea = BADADDR;            ///< address where the argument was initialized.
                                ///< BADADDR means unknown.
  tinfo_t type;                 ///< formal argument type
  qstring name;                 ///< formal argument name
  argloc_t argloc;              ///< ida argloc
  uint32 flags = 0;             ///< FAI_...
 
  mcallarg_t() {}
  mcallarg_t(const mop_t &rarg) : mop_t(rarg) {}
  void copy_mop(const mop_t &op) { *(mop_t *)this = op; }
  void hexapi print(qstring *vout, int shins_flags=SHINS_SHORT|SHINS_VALNUM) const;
  const char *hexapi dstr() const;
  void hexapi set_regarg(mreg_t mr, int sz, const tinfo_t &tif);
  void set_regarg(mreg_t mr, const tinfo_t &tif)
  {
    set_regarg(mr, tif.get_size(), tif);
  }
  void set_regarg(mreg_t mr, char dt, type_sign_t sign = type_unsigned)
  {
    int sz = get_dtype_size(dt);
    set_regarg(mr, sz, get_int_type_by_width_and_sign(sz, sign));
  }
  void make_int(int val, ea_t val_ea, int opno = 0)
  {
    type = tinfo_t(BTF_INT);
    make_number(val, inf_get_cc_size_i(), val_ea, opno);
  }
  void make_uint(int val, ea_t val_ea, int opno = 0)
  {
    type = tinfo_t(BTF_UINT);
    make_number(val, inf_get_cc_size_i(), val_ea, opno);
  }
};
DECLARE_TYPE_AS_MOVABLE(mcallarg_t);
typedef qvector<mcallarg_t> mcallargs_t;
 
/// Function roles.
/// They are used to calculate use/def lists and to recognize functions
/// without using string comparisons.
enum funcrole_t
{
  ROLE_UNK,                  ///< unknown function role
  ROLE_EMPTY,                ///< empty, does not do anything (maybe spoils regs)
  ROLE_MEMSET,               ///< memset(void *dst, uchar value, size_t count);
  ROLE_MEMSET32,             ///< memset32(void *dst, uint32 value, size_t count);
  ROLE_MEMSET64,             ///< memset64(void *dst, uint64 value, size_t count);
  ROLE_MEMCPY,               ///< memcpy(void *dst, const void *src, size_t count);
  ROLE_STRCPY,               ///< strcpy(char *dst, const char *src);
  ROLE_STRLEN,               ///< strlen(const char *src);
  ROLE_STRCAT,               ///< strcat(char *dst, const char *src);
  ROLE_TAIL,                 ///< char *tail(const char *str);
  ROLE_BUG,                  ///< BUG() helper macro: never returns, causes exception
  ROLE_ALLOCA,               ///< alloca() function
  ROLE_BSWAP,                ///< bswap() function (any size)
  ROLE_PRESENT,              ///< present() function (used in patterns)
  ROLE_CONTAINING_RECORD,    ///< CONTAINING_RECORD() macro
  ROLE_FASTFAIL,             ///< __fastfail()
  ROLE_READFLAGS,            ///< __readeflags, __readcallersflags
  ROLE_IS_MUL_OK,            ///< is_mul_ok
  ROLE_SATURATED_MUL,        ///< saturated_mul
  ROLE_BITTEST,              ///< [lock] bt
  ROLE_BITTESTANDSET,        ///< [lock] bts
  ROLE_BITTESTANDRESET,      ///< [lock] btr
  ROLE_BITTESTANDCOMPLEMENT, ///< [lock] btc
  ROLE_VA_ARG,               ///< va_arg() macro
  ROLE_VA_COPY,              ///< va_copy() function
  ROLE_VA_START,             ///< va_start() function
  ROLE_VA_END,               ///< va_end() function
  ROLE_ROL,                  ///< rotate left
  ROLE_ROR,                  ///< rotate right
  ROLE_CFSUB3,               ///< carry flag after subtract with carry
  ROLE_OFSUB3,               ///< overflow flag after subtract with carry
  ROLE_ABS,                  ///< integer absolute value
  ROLE_3WAYCMP0,             ///< 3-way compare helper, returns -1/0/1
  ROLE_3WAYCMP1,             ///< 3-way compare helper, returns 0/1/2
  ROLE_WMEMCPY,              ///< wchar_t *wmemcpy(wchar_t *dst, const wchar_t *src, size_t n)
  ROLE_WMEMSET,              ///< wchar_t *wmemset(wchar_t *dst, wchar_t wc, size_t n)
  ROLE_WCSCPY,               ///< wchar_t *wcscpy(wchar_t *dst, const wchar_t *src);
  ROLE_WCSLEN,               ///< size_t wcslen(const wchar_t *s)
  ROLE_WCSCAT,               ///< wchar_t *wcscat(wchar_t *dst, const wchar_t *src)
  ROLE_SSE_CMP4,             ///< e.g. _mm_cmpgt_ss
  ROLE_SSE_CMP8,             ///< e.g. _mm_cmpgt_sd
};
 
/// \defgroup FUNC_NAME_ Well known function names
//@{
#define FUNC_NAME_MEMCPY   "memcpy"
#define FUNC_NAME_WMEMCPY  "wmemcpy"
#define FUNC_NAME_MEMSET   "memset"
#define FUNC_NAME_WMEMSET  "wmemset"
#define FUNC_NAME_MEMSET32 "memset32"
#define FUNC_NAME_MEMSET64 "memset64"
#define FUNC_NAME_STRCPY   "strcpy"
#define FUNC_NAME_WCSCPY   "wcscpy"
#define FUNC_NAME_STRLEN   "strlen"
#define FUNC_NAME_WCSLEN   "wcslen"
#define FUNC_NAME_STRCAT   "strcat"
#define FUNC_NAME_WCSCAT   "wcscat"
#define FUNC_NAME_TAIL     "tail"
#define FUNC_NAME_VA_ARG   "va_arg"
#define FUNC_NAME_EMPTY    "$empty"
#define FUNC_NAME_PRESENT  "$present"
#define FUNC_NAME_CONTAINING_RECORD "CONTAINING_RECORD"
//@}
 
 
// the default 256 function arguments is too big, we use a lower value
#undef MAX_FUNC_ARGS
#define MAX_FUNC_ARGS 64
 
/// Information about a call
class mcallinfo_t               // #callinfo
{
public:
  ea_t callee;                  ///< address of the called function, if known
  int solid_args;               ///< number of solid args.
                                ///< there may be variadic args in addtion
  int call_spd;                 ///< sp value at call insn
  int stkargs_top;              ///< first offset past stack arguments
  cm_t cc;                      ///< calling convention
  mcallargs_t args;             ///< call arguments
  mopvec_t retregs;             ///< return register(s) (e.g., AX, AX:DX, etc.)
                                ///< this vector is built from return_regs
  tinfo_t return_type;          ///< type of the returned value
  argloc_t return_argloc;       ///< location of the returned value
 
  mlist_t return_regs;          ///< list of values returned by the function
  mlist_t spoiled;              ///< list of spoiled locations (includes return_regs)
  mlist_t pass_regs;            ///< passthrough registers: registers that depend on input
                                ///< values (subset of spoiled)
  ivlset_t visible_memory;      ///< what memory is visible to the call?
  mlist_t dead_regs;            ///< registers defined by the function but never used.
                                ///< upon propagation we do the following:
                                ///<   - dead_regs += return_regs
                                ///<   - retregs.clear() since the call is propagated
  int flags;                    ///< combination of \ref FCI_... bits
/// \defgroup FCI_ Call properties
//@{
#define FCI_PROP    0x001       ///< call has been propagated
#define FCI_DEAD    0x002       ///< some return registers were determined dead
#define FCI_FINAL   0x004       ///< call type is final, should not be changed
#define FCI_NORET   0x008       ///< call does not return
#define FCI_PURE    0x010       ///< pure function
#define FCI_NOSIDE  0x020       ///< call does not have side effects
#define FCI_SPLOK   0x040       ///< spoiled/visible_memory lists have been
                                ///< optimized. for some functions we can reduce them
                                ///< as soon as information about the arguments becomes
                                ///< available. in order not to try optimize them again
                                ///< we use this bit.
#define FCI_HASCALL 0x080       ///< A function is an synthetic helper combined
                                ///< from several instructions and at least one
                                ///< of them was a call to a real functions
#define FCI_HASFMT  0x100       ///< A variadic function with recognized
                                ///< printf- or scanf-style format string
#define FCI_EXPLOCS 0x400       ///< all arglocs are specified explicitly
//@}
  funcrole_t role;              ///< function role
  type_attrs_t fti_attrs;       ///< extended function attributes
 
  mcallinfo_t(ea_t _callee=BADADDR, int _sargs=0)
    : callee(_callee), solid_args(_sargs), call_spd(0), stkargs_top(0),
      cc(CM_CC_INVALID), flags(0), role(ROLE_UNK) {}
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  int hexapi lexcompare(const mcallinfo_t &f) const;
  bool hexapi set_type(const tinfo_t &type);
  tinfo_t hexapi get_type(void) const;
  bool is_vararg(void) const { return is_vararg_cc(cc); }
  void hexapi print(qstring *vout, int size=-1, int shins_flags=SHINS_SHORT|SHINS_VALNUM) const;
  const char *hexapi dstr() const;
};
 
/// List of switch cases and targets
class mcases_t                  // #cases
{
public:
  casevec_t values;             ///< expression values for each target
  intvec_t targets;             ///< target block numbers
 
  void swap(mcases_t &r) { values.swap(r.values); targets.swap(r.targets); }
  DECLARE_COMPARISONS(mcases_t);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  bool empty(void) const { return targets.empty(); }
  size_t size(void) const { return targets.size(); }
  void resize(int s) { values.resize(s); targets.resize(s); }
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr() const;
};
 
//-------------------------------------------------------------------------
/// Value offset (microregister number or stack offset)
struct voff_t
{
  sval_t off;         ///< register number or stack offset
  mopt_t type;        ///< mop_r - register, mop_S - stack, mop_z - undefined
 
  voff_t() : off(-1), type(mop_z) {}
  voff_t(mopt_t _type, sval_t _off) : off(_off), type(_type) {}
  voff_t(const mop_t &op) : off(-1), type(mop_z)
  {
    if ( op.is_reg() || op.t == mop_S )
      set(op.t, op.is_reg() ? op.r : op.s->off);
  }
 
  void set(mopt_t _type, sval_t _off) { type = _type; off = _off; }
  void set_stkoff(sval_t stkoff)      { set(mop_S, stkoff); }
  void set_reg(mreg_t mreg)           { set(mop_r, mreg); }
  void undef()                        { set(mop_z, -1); }
 
  bool defined()      const { return type != mop_z; }
  bool is_reg()       const { return type == mop_r; }
  bool is_stkoff()    const { return type == mop_S; }
  mreg_t get_reg()    const { QASSERT(51892, is_reg()); return off; }
  sval_t get_stkoff() const { QASSERT(51893, is_stkoff()); return off; }
 
  void inc(sval_t delta)              { off += delta; }
  voff_t add(int width) const         { return voff_t(type, off+width); }
  sval_t diff(const voff_t &r) const  { QASSERT(51894, type == r.type); return off - r.off; }
 
  DECLARE_COMPARISONS(voff_t)
  {
    int code = ::compare(type, r.type);
    return code != 0 ? code : ::compare(off, r.off);
  }
};
 
//-------------------------------------------------------------------------
/// Value interval (register or stack range)
struct vivl_t : voff_t
{
  int size;     ///< Interval size in bytes
 
  vivl_t(mopt_t _type = mop_z, sval_t _off = -1, int _size = 0)
    : voff_t(_type, _off), size(_size) {}
  vivl_t(const class chain_t &ch);
  vivl_t(const mop_t &op) : voff_t(op), size(op.size) {}
 
  // Make a value interval
  void set(mopt_t _type, sval_t _off, int _size = 0)
    { voff_t::set(_type, _off); size = _size; }
  void set(const voff_t &voff, int _size)
    { set(voff.type, voff.off, _size); }
  void set_stkoff(sval_t stkoff, int sz = 0) { set(mop_S, stkoff, sz); }
  void set_reg   (mreg_t mreg,   int sz = 0) { set(mop_r, mreg,   sz); }
 
  /// Extend a value interval using another value interval of the same type
  /// \return success
  bool hexapi extend_to_cover(const vivl_t &r);
 
  /// Intersect value intervals the same type
  /// \return size of the resulting intersection
  uval_t hexapi intersect(const vivl_t &r);
 
  /// Do two value intervals overlap?
  bool overlap(const vivl_t &r) const
  {
    return type == r.type
        && interval::overlap(off, size, r.off, r.size);
  }
  /// Does our value interval include another?
  bool includes(const vivl_t &r) const
  {
    return type == r.type
        && interval::includes(off, size, r.off, r.size);
  }
 
  /// Does our value interval contain the specified value offset?
  bool contains(const voff_t &voff2) const
  {
    return type == voff2.type
        && interval::contains(off, size, voff2.off);
  }
 
  // Comparisons
  DECLARE_COMPARISONS(vivl_t)
  {
    int code = voff_t::compare(r);
    return code; //return code != 0 ? code : ::compare(size, r.size);
  }
  bool operator==(const mop_t &mop) const
  {
    return type == mop.t && off == (mop.is_reg() ? mop.r : mop.s->off);
  }
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr() const;
};
 
//-------------------------------------------------------------------------
/// ud (use->def) and du (def->use) chain.
/// We store in chains only the block numbers, not individual instructions
/// See https://en.wikipedia.org/wiki/Use-define_chain
class chain_t : public intvec_t // sequence of block numbers
{
  voff_t k;             ///< Value offset of the chain.
                        ///< (what variable is this chain about)
 
public:
  int width;            ///< size of the value in bytes
  int varnum;           ///< allocated variable index (-1 - not allocated yet)
  uchar flags;          ///< combination \ref CHF_ bits
/// \defgroup CHF_ Chain properties
//@{
#define CHF_INITED     0x01 ///< is chain initialized? (valid only after lvar allocation)
#define CHF_REPLACED   0x02 ///< chain operands have been replaced?
#define CHF_OVER       0x04 ///< overlapped chain
#define CHF_FAKE       0x08 ///< fake chain created by widen_chains()
#define CHF_PASSTHRU   0x10 ///< pass-thru chain, must use the input variable to the block
#define CHF_TERM       0x20 ///< terminating chain; the variable does not survive across the block
//@}
  chain_t() : width(0), varnum(-1), flags(CHF_INITED) {}
  chain_t(mopt_t t, sval_t off, int w=1, int v=-1)
    : k(t, off), width(w), varnum(v), flags(CHF_INITED) {}
  chain_t(const voff_t &_k, int w=1)
    : k(_k), width(w), varnum(-1), flags(CHF_INITED) {}
  void set_value(const chain_t &r)
    { width = r.width; varnum = r.varnum; flags = r.flags; *(intvec_t *)this = (intvec_t &)r; }
  const voff_t &key() const { return k; }
  bool is_inited(void) const { return (flags & CHF_INITED) != 0; }
  bool is_reg(void) const { return k.is_reg(); }
  bool is_stkoff(void) const { return k.is_stkoff(); }
  bool is_replaced(void) const { return (flags & CHF_REPLACED) != 0; }
  bool is_overlapped(void) const { return (flags & CHF_OVER) != 0; }
  bool is_fake(void) const { return (flags & CHF_FAKE) != 0; }
  bool is_passreg(void) const { return (flags & CHF_PASSTHRU) != 0; }
  bool is_term(void) const { return (flags & CHF_TERM) != 0; }
  void set_inited(bool b) { setflag(flags, CHF_INITED, b); }
  void set_replaced(bool b) { setflag(flags, CHF_REPLACED, b); }
  void set_overlapped(bool b) { setflag(flags, CHF_OVER, b); }
  void set_term(bool b) { setflag(flags, CHF_TERM, b); }
  mreg_t get_reg() const { return k.get_reg(); }
  sval_t get_stkoff() const { return k.get_stkoff(); }
  bool overlap(const chain_t &r) const
    { return k.type == r.k.type && interval::overlap(k.off, width, r.k.off, r.width); }
  bool includes(const chain_t &r) const
    { return k.type == r.k.type && interval::includes(k.off, width, r.k.off, r.width); }
  const voff_t endoff() const { return k.add(width); }
 
  bool operator<(const chain_t &r) const { return key() < r.key(); }
 
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr() const;
  /// Append the contents of the chain to the specified list of locations.
  void hexapi append_list(const mba_t *mba, mlist_t *list) const;
  void clear_varnum(void) { varnum = -1; set_replaced(false); }
};
 
//-------------------------------------------------------------------------
#if defined(__NT__)
#  ifdef _DEBUG
#    define SIZEOF_BLOCK_CHAINS  32
#else
#    define SIZEOF_BLOCK_CHAINS  24
#  endif
#elif defined(__MAC__)
#  define SIZEOF_BLOCK_CHAINS  32
#else
#  define SIZEOF_BLOCK_CHAINS  56
#endif
 
/// Chains of one block.
/// Please note that this class is based on std::set and it must be accessed
/// using the block_chains_begin(), block_chains_find() and similar functions.
/// This is required because different compilers use different implementations
/// of std::set. However, since the size of std::set depends on the compilation
/// options, we replace it with a byte array.
class block_chains_t
{
  size_t body[SIZEOF_BLOCK_CHAINS/sizeof(size_t)]; // opaque std::set, uncopyable
public:
  /// Get chain for the specified register
  /// \param reg   register number
  /// \param width size of register in bytes
  const chain_t *get_reg_chain(mreg_t reg, int width=1) const
    { return get_chain((chain_t(mop_r, reg, width))); }
  chain_t *get_reg_chain(mreg_t reg, int width=1)
    { return get_chain((chain_t(mop_r, reg, width))); }
 
  /// Get chain for the specified stack offset
  /// \param off   stack offset
  /// \param width size of stack value in bytes
  const chain_t *get_stk_chain(sval_t off, int width=1) const
    { return get_chain(chain_t(mop_S, off, width)); }
  chain_t *get_stk_chain(sval_t off, int width=1)
    { return get_chain(chain_t(mop_S, off, width)); }
 
  /// Get chain for the specified value offset.
  /// \param k     value offset (register number or stack offset)
  /// \param width size of value in bytes
  const chain_t *get_chain(const voff_t &k, int width=1) const
    { return get_chain(chain_t(k, width)); }
  chain_t *get_chain(const voff_t &k, int width=1)
    { return (chain_t*)((const block_chains_t *)this)->get_chain(k, width); }
 
  /// Get chain similar to the specified chain
  /// \param ch    chain to search for. only its 'k' and 'width' are used.
  const chain_t *hexapi get_chain(const chain_t &ch) const;
  chain_t *get_chain(const chain_t &ch)
    { return (chain_t*)((const block_chains_t *)this)->get_chain(ch); }
 
  void hexapi print(qstring *vout) const;
  const char *hexapi dstr() const;
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
};
//-------------------------------------------------------------------------
/// Chain visitor class
struct chain_visitor_t
{
  block_chains_t *parent;          ///< parent of the current chain
  chain_visitor_t(void) : parent(nullptr) {}
  virtual int idaapi visit_chain(int nblock, chain_t &ch) = 0;
};
 
//-------------------------------------------------------------------------
/// Graph chains.
/// This class represents all ud and du chains of the decompiled function
typedef qvector<block_chains_t> block_chains_vec_t;
class graph_chains_t : public block_chains_vec_t
{
  int lock;             ///< are chained locked? (in-use)
public:
  graph_chains_t(void) : lock(0) {}
  ~graph_chains_t(void) { QASSERT(50444, !lock); }
  /// Visit all chains
  /// \param cv chain visitor
  /// \param gca_flags combination of GCA_ bits
  int hexapi for_all_chains(chain_visitor_t &cv, int gca_flags);
  /// \defgroup GCA_ chain visitor flags
  //@{
#define GCA_EMPTY  0x01 ///< include empty chains
#define GCA_SPEC   0x02 ///< include chains for special registers
#define GCA_ALLOC  0x04 ///< enumerate only allocated chains
#define GCA_NALLOC 0x08 ///< enumerate only non-allocated chains
#define GCA_OFIRST 0x10 ///< consider only chains of the first block
#define GCA_OLAST  0x20 ///< consider only chains of the last block
  //@}
  /// Are the chains locked?
  /// It is a good idea to lock the chains before using them. This ensures
  /// that they won't be recalculated and reallocated during the use.
  /// See the \ref chain_keeper_t class for that.
  bool is_locked(void) const { return lock != 0; }
  /// Lock the chains
  void acquire(void) { lock++; }
  /// Unlock the chains
  void hexapi release(void);
  void swap(graph_chains_t &r)
  {
    qvector<block_chains_t>::swap(r);
    std::swap(lock, r.lock);
  }
};
//-------------------------------------------------------------------------
/// Microinstruction class #insn
class minsn_t
{
  void hexapi init(ea_t _ea);
  void hexapi copy(const minsn_t &m);
public:
  mcode_t opcode;       ///< instruction opcode
  int iprops;           ///< combination of \ref IPROP_ bits
  minsn_t *next;        ///< next insn in doubly linked list. check also nexti()
  minsn_t *prev;        ///< prev insn in doubly linked list. check also previ()
  ea_t ea;              ///< instruction address
  mop_t l;              ///< left operand
  mop_t r;              ///< right operand
  mop_t d;              ///< destination operand
 
  /// \defgroup IPROP_ instruction property bits
  //@{
  // bits to be used in patterns:
#define IPROP_OPTIONAL  0x0001 ///< optional instruction
#define IPROP_PERSIST   0x0002 ///< persistent insn; they are not destroyed
#define IPROP_WILDMATCH 0x0004 ///< match multiple insns
 
  // instruction attributes:
#define IPROP_CLNPOP    0x0008 ///< the purpose of the instruction is to clean stack
                               ///< (e.g. "pop ecx" is often used for that)
#define IPROP_FPINSN    0x0010 ///< floating point insn
#define IPROP_FARCALL   0x0020 ///< call of a far function using push cs/call sequence
#define IPROP_TAILCALL  0x0040 ///< tail call
#define IPROP_ASSERT    0x0080 ///< assertion: usually mov #val, op.
                               ///< assertions are used to help the optimizer.
                               ///< assertions are ignored when generating ctree
 
  // instruction history:
#define IPROP_SPLIT     0x0700 ///< the instruction has been split:
#define IPROP_SPLIT1    0x0100 ///<   into 1 byte
#define IPROP_SPLIT2    0x0200 ///<   into 2 bytes
#define IPROP_SPLIT4    0x0300 ///<   into 4 bytes
#define IPROP_SPLIT8    0x0400 ///<   into 8 bytes
#define IPROP_COMBINED  0x0800 ///< insn has been modified because of a partial reference
#define IPROP_EXTSTX    0x1000 ///< this is m_ext propagated into m_stx
#define IPROP_IGNLOWSRC 0x2000 ///< low part of the instruction source operand
                               ///< has been created artificially
                               ///< (this bit is used only for 'and x, 80...')
#define IPROP_INV_JX    0x4000 ///< inverted conditional jump
#define IPROP_WAS_NORET 0x8000 ///< was noret icall
#define IPROP_MULTI_MOV 0x10000 ///< the minsn was generated as part of insn that moves multiple registers
                                ///< (example: STM on ARM may transfer multiple registers)
 
                                ///< bits that can be set by plugins:
#define IPROP_DONT_PROP 0x20000 ///< may not propagate
#define IPROP_DONT_COMB 0x40000 ///< may not combine this instruction with others
#define IPROP_MBARRIER  0x80000 ///< this instruction acts as a memory barrier
                                ///< (instructions accessing memory may not be reordered past it)
#define IPROP_UNMERGED 0x100000 ///< 'goto' instruction was transformed info 'call'
  //@}
 
  bool is_optional(void)     const { return (iprops & IPROP_OPTIONAL)  != 0; }
  bool is_combined(void)     const { return (iprops & IPROP_COMBINED)  != 0; }
  bool is_farcall(void)      const { return (iprops & IPROP_FARCALL)   != 0; }
  bool is_cleaning_pop(void) const { return (iprops & IPROP_CLNPOP)    != 0; }
  bool is_extstx(void)       const { return (iprops & IPROP_EXTSTX)    != 0; }
  bool is_tailcall(void)     const { return (iprops & IPROP_TAILCALL)  != 0; }
  bool is_fpinsn(void)       const { return (iprops & IPROP_FPINSN)    != 0; }
  bool is_assert(void)       const { return (iprops & IPROP_ASSERT)    != 0; }
  bool is_persistent(void)   const { return (iprops & IPROP_PERSIST)   != 0; }
  bool is_wild_match(void)   const { return (iprops & IPROP_WILDMATCH) != 0; }
  bool is_propagatable(void) const { return (iprops & IPROP_DONT_PROP) == 0; }
  bool is_ignlowsrc(void)    const { return (iprops & IPROP_IGNLOWSRC) != 0; }
  bool is_inverted_jx(void)  const { return (iprops & IPROP_INV_JX)    != 0; }
  bool was_noret_icall(void) const { return (iprops & IPROP_WAS_NORET) != 0; }
  bool is_multimov(void)     const { return (iprops & IPROP_MULTI_MOV) != 0; }
  bool is_combinable(void)   const { return (iprops & IPROP_DONT_COMB) == 0; }
  bool was_split(void)       const { return (iprops & IPROP_SPLIT)     != 0; }
  bool is_mbarrier(void)     const { return (iprops & IPROP_MBARRIER)  != 0; }
  bool was_unmerged(void)    const { return (iprops & IPROP_UNMERGED)  != 0; }
 
  void set_optional(void) { iprops |= IPROP_OPTIONAL; }
  void hexapi set_combined(void);
  void clr_combined(void) { iprops &= ~IPROP_COMBINED; }
  void set_farcall(void)  { iprops |= IPROP_FARCALL; }
  void set_cleaning_pop(void) { iprops |= IPROP_CLNPOP; }
  void set_extstx(void)   { iprops |= IPROP_EXTSTX; }
  void set_tailcall(void) { iprops |= IPROP_TAILCALL; }
  void clr_tailcall(void) { iprops &= ~IPROP_TAILCALL; }
  void set_fpinsn(void)   { iprops |= IPROP_FPINSN; }
  void clr_fpinsn(void)   { iprops &= ~IPROP_FPINSN; }
  void set_assert(void)   { iprops |= IPROP_ASSERT; }
  void clr_assert(void)   { iprops &= ~IPROP_ASSERT; }
  void set_persistent(void) { iprops |= IPROP_PERSIST; }
  void set_wild_match(void) { iprops |= IPROP_WILDMATCH; }
  void clr_propagatable(void) { iprops |= IPROP_DONT_PROP; }
  void set_ignlowsrc(void) { iprops |= IPROP_IGNLOWSRC; }
  void clr_ignlowsrc(void) { iprops &= ~IPROP_IGNLOWSRC; }
  void set_inverted_jx(void) { iprops |= IPROP_INV_JX; }
  void set_noret_icall(void) { iprops |= IPROP_WAS_NORET; }
  void clr_noret_icall(void) { iprops &= ~IPROP_WAS_NORET; }
  void set_multimov(void) { iprops |= IPROP_MULTI_MOV; }
  void clr_multimov(void) { iprops &= ~IPROP_MULTI_MOV; }
  void set_combinable(void) { iprops &= ~IPROP_DONT_COMB; }
  void clr_combinable(void) { iprops |= IPROP_DONT_COMB; }
  void set_mbarrier(void) { iprops |= IPROP_MBARRIER; }
  void set_unmerged(void) { iprops |= IPROP_UNMERGED; }
  void set_split_size(int s)
  { // s may be only 1,2,4,8. other values are ignored
    iprops &= ~IPROP_SPLIT;
    iprops |= (s == 1 ? IPROP_SPLIT1
             : s == 2 ? IPROP_SPLIT2
             : s == 4 ? IPROP_SPLIT4
             : s == 8 ? IPROP_SPLIT8 : 0);
  }
  int get_split_size(void) const
  {
    int cnt = (iprops & IPROP_SPLIT) >> 8;
    return cnt == 0 ? 0 : 1 << (cnt-1);
  }
 
  /// Constructor
  minsn_t(ea_t _ea) { init(_ea); }
  minsn_t(const minsn_t &m) { next = prev = nullptr; copy(m); } //-V1077 uninitialized: opcode, iprops, ea
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
 
  /// Assignment operator. It does not copy prev/next fields.
  minsn_t &operator=(const minsn_t &m) { copy(m); return *this; }
 
  /// Swap two instructions.
  /// The prev/next fields are not modified by this function
  /// because it would corrupt the doubly linked list.
  void hexapi swap(minsn_t &m);
 
  /// Generate insn text into the buffer
  void hexapi print(qstring *vout, int shins_flags=SHINS_SHORT|SHINS_VALNUM) const;
 
  /// Get displayable text without tags in a static buffer
  const char *hexapi dstr() const;
 
  /// Change the instruction address.
  /// This function modifies subinstructions as well.
  void hexapi setaddr(ea_t new_ea);
 
  /// Optimize one instruction without context.
  /// This function does not have access to the instruction context (the
  /// previous and next instructions in the list, the block number, etc).
  /// It performs only basic optimizations that are available without this info.
  /// \param optflags combination of \ref OPTI_ bits
  /// \return number of changes, 0-unchanged
  /// See also mblock_t::optimize_insn()
  int optimize_solo(int optflags=0) { return optimize_subtree(nullptr, nullptr, nullptr, nullptr, optflags); }
  /// \defgroup OPTI_ optimization flags
  //@{
#define OPTI_ADDREXPRS 0x0001 ///< optimize all address expressions (&x+N; &x-&y)
#define OPTI_MINSTKREF 0x0002 ///< may update minstkref
#define OPTI_COMBINSNS 0x0004 ///< may combine insns (only for optimize_insn)
#define OPTI_NO_LDXOPT 0x0008 ///< the function is called after the
                              ///< propagation attempt, we do not optimize
                              ///< low/high(ldx) in this case
  //@}
 
  /// Optimize instruction in its context.
  /// Do not use this function, use mblock_t::optimize()
  int hexapi optimize_subtree(
        mblock_t *blk,
        minsn_t *top,
        minsn_t *parent,
        ea_t *converted_call,
        int optflags=OPTI_MINSTKREF);
 
  /// Visit all instruction operands.
  /// This function visits subinstruction operands as well.
  /// \param mv operand visitor
  /// \return non-zero value returned by mv.visit_mop() or zero
  int hexapi for_all_ops(mop_visitor_t &mv);
 
  /// Visit all instructions.
  /// This function visits the instruction itself and all its subinstructions.
  /// \param mv instruction visitor
  /// \return non-zero value returned by mv.visit_mop() or zero
  int hexapi for_all_insns(minsn_visitor_t &mv);
 
  /// Convert instruction to nop.
  /// This function erases all info but the prev/next fields.
  /// In most cases it is better to use mblock_t::make_nop(), which also
  /// marks the block lists as dirty.
  void hexapi _make_nop(void);
 
  /// Compare instructions.
  /// This is the main comparison function for instructions.
  /// \param m       instruction to compare with
  /// \param eqflags combination of \ref EQ_ bits
  bool hexapi equal_insns(const minsn_t &m, int eqflags) const; // intelligent comparison
  /// \defgroup EQ_ comparison bits
  //@{
#define EQ_IGNSIZE 0x0001      ///< ignore source operand sizes
#define EQ_IGNCODE 0x0002      ///< ignore instruction opcodes
#define EQ_CMPDEST 0x0004      ///< compare instruction destinations
#define EQ_OPTINSN 0x0008      ///< optimize mop_d operands
  //@}
 
  /// Lexographical comparison
  /// It can be used to store minsn_t in various containers, like std::set
  bool operator <(const minsn_t &ri) const { return lexcompare(ri) < 0; }
  int hexapi lexcompare(const minsn_t &ri) const;
 
  //-----------------------------------------------------------------------
  // Call instructions
  //-----------------------------------------------------------------------
  /// Is a non-returing call?
  /// \param flags combination of NORET_... bits
  bool hexapi is_noret_call(int flags=0);
#define NORET_IGNORE_WAS_NORET_ICALL 0x01 // ignore was_noret_icall() bit
#define NORET_FORBID_ANALYSIS        0x02 // forbid additional analysis
 
  /// Is an unknown call?
  /// Unknown calls are calls without the argument list (mcallinfo_t).
  /// Usually the argument lists are determined by mba_t::analyze_calls().
  /// Unknown calls exist until the MMAT_CALLS maturity level.
  /// See also \ref mblock_t::is_call_block
  bool is_unknown_call(void) const { return is_mcode_call(opcode) && d.empty(); }
 
  /// Is a helper call with the specified name?
  /// Helper calls usually have well-known function names (see \ref FUNC_NAME_)
  /// but they may have any other name. The decompiler does not assume any
  /// special meaning for non-well-known names.
  bool hexapi is_helper(const char *name) const;
 
  /// Find a call instruction.
  /// Check for the current instruction and its subinstructions.
  /// \param with_helpers consider helper calls as well?
  minsn_t *hexapi find_call(bool with_helpers=false) const;
 
  /// Does the instruction contain a call?
  bool contains_call(bool with_helpers=false) const { return find_call(with_helpers) != nullptr; }
 
  /// Does the instruction have a side effect?
  /// \param include_ldx_and_divs consider ldx/div/mod as having side effects?
  ///                    stx is always considered as having side effects.
  /// Apart from ldx/std only call may have side effects.
  bool hexapi has_side_effects(bool include_ldx_and_divs=false) const;
 
  /// Get the function role of a call
  funcrole_t get_role(void) const { return d.is_arglist() ? d.f->role : ROLE_UNK; }
  bool is_memcpy(void) const { return get_role() == ROLE_MEMCPY; }
  bool is_memset(void) const { return get_role() == ROLE_MEMSET; }
  bool is_alloca(void) const { return get_role() == ROLE_ALLOCA; }
  bool is_bswap (void) const { return get_role() == ROLE_BSWAP;  }
  bool is_readflags (void) const { return get_role() == ROLE_READFLAGS;  }
 
  //-----------------------------------------------------------------------
  // Misc
  //-----------------------------------------------------------------------
  /// Does the instruction have the specified opcode?
  /// This function searches subinstructions as well.
  /// \param mcode opcode to search for.
  bool contains_opcode(mcode_t mcode) const { return find_opcode(mcode) != nullptr; }
 
  /// Find a (sub)insruction with the specified opcode.
  /// \param mcode opcode to search for.
  const minsn_t *find_opcode(mcode_t mcode) const { return (CONST_CAST(minsn_t*)(this))->find_opcode(mcode); }
  minsn_t *hexapi find_opcode(mcode_t mcode);
 
  /// Find an operand that is a subinsruction with the specified opcode.
  /// This function checks only the 'l' and 'r' operands of the current insn.
  /// \param[out] other pointer to the other operand
  ///             (&r if we return &l and vice versa)
  /// \param op   opcode to search for
  /// \return &l or &r or nullptr
  const minsn_t *hexapi find_ins_op(const mop_t **other, mcode_t op=m_nop) const;
  minsn_t *find_ins_op(mop_t **other, mcode_t op=m_nop) { return CONST_CAST(minsn_t*)((CONST_CAST(const minsn_t*)(this))->find_ins_op((const mop_t**)other, op)); }
 
  /// Find a numeric operand of the current instruction.
  /// This function checks only the 'l' and 'r' operands of the current insn.
  /// \param[out] other pointer to the other operand
  ///             (&r if we return &l and vice versa)
  /// \return &l or &r or nullptr
  const mop_t *hexapi find_num_op(const mop_t **other) const;
  mop_t *find_num_op(mop_t **other) { return CONST_CAST(mop_t*)((CONST_CAST(const minsn_t*)(this))->find_num_op((const mop_t**)other)); }
 
  bool is_mov(void) const { return opcode == m_mov || (opcode == m_f2f && l.size == d.size); }
  bool is_like_move(void) const { return is_mov() || is_mcode_xdsu(opcode) || opcode == m_low; }
 
  /// Does the instruction modify its 'd' operand?
  /// Some instructions (e.g. m_stx) do not modify the 'd' operand.
  bool hexapi modifies_d(void) const;
  bool modifies_pair_mop(void) const { return d.t == mop_p && modifies_d(); }
 
  /// Is the instruction in the specified range of instructions?
  /// \param m1 beginning of the range in the doubly linked list
  /// \param m2 end of the range in the doubly linked list (excluded, may be nullptr)
  /// This function assumes that m1 and m2 belong to the same basic block
  /// and they are top level instructions.
  bool hexapi is_between(const minsn_t *m1, const minsn_t *m2) const;
 
  /// Is the instruction after the specified one?
  /// \param m the instruction to compare against in the list
  bool is_after(const minsn_t *m) const { return m != nullptr && is_between(m->next, nullptr); }
 
  /// Is it possible for the instruction to use aliased memory?
  bool hexapi may_use_aliased_memory(void) const;
 
  /// Serialize an instruction
  /// \param b the output buffer
  /// \return the serialization format that was used to store info
  int hexapi serialize(bytevec_t *b) const;
 
  /// Deserialize an instruction
  /// \param bytes pointer to serialized data
  /// \param nbytes number of bytes to deserialize
  /// \param format_version serialization format version. this value is returned by minsn_t::serialize()
  /// \return success
  bool hexapi deserialize(const uchar *bytes, size_t nbytes, int format_version);
 
};
 
/// Skip assertions forward
const minsn_t *hexapi getf_reginsn(const minsn_t *ins);
/// Skip assertions backward
const minsn_t *hexapi getb_reginsn(const minsn_t *ins);
inline minsn_t *getf_reginsn(minsn_t *ins) { return CONST_CAST(minsn_t*)(getf_reginsn(CONST_CAST(const minsn_t *)(ins))); }
inline minsn_t *getb_reginsn(minsn_t *ins) { return CONST_CAST(minsn_t*)(getb_reginsn(CONST_CAST(const minsn_t *)(ins))); }
 
//-------------------------------------------------------------------------
/// Basic block types
enum mblock_type_t
{
  BLT_NONE = 0, ///< unknown block type
  BLT_STOP = 1, ///< stops execution regularly (must be the last block)
  BLT_0WAY = 2, ///< does not have successors (tail is a noret function)
  BLT_1WAY = 3, ///< passes execution to one block (regular or goto block)
  BLT_2WAY = 4, ///< passes execution to two blocks (conditional jump)
  BLT_NWAY = 5, ///< passes execution to many blocks (switch idiom)
  BLT_XTRN = 6, ///< external block (out of function address)
};
 
// Maximal bit range
#define MAXRANGE bitrange_t(0, USHRT_MAX)
 
//-------------------------------------------------------------------------
/// Microcode of one basic block.
/// All blocks are part of a doubly linked list. They can also be addressed
/// by indexing the mba->natural array. A block contains a doubly linked list
/// of instructions, various location lists that are used for data flow
/// analysis, and other attributes.
class mblock_t
{
  friend class codegen_t;
  DECLARE_UNCOPYABLE(mblock_t)
  void hexapi init(void);
public:
  mblock_t *nextb;              ///< next block in the doubly linked list
  mblock_t *prevb;              ///< previous block in the doubly linked list
  uint32 flags;                 ///< combination of \ref MBL_ bits
  /// \defgroup MBL_ Basic block properties
  //@{
#define MBL_PRIV        0x0001  ///< private block - no instructions except
                                ///< the specified are accepted (used in patterns)
#define MBL_NONFAKE     0x0000  ///< regular block
#define MBL_FAKE        0x0002  ///< fake block
#define MBL_GOTO        0x0004  ///< this block is a goto target
#define MBL_TCAL        0x0008  ///< aritifical call block for tail calls
#define MBL_PUSH        0x0010  ///< needs "convert push/pop instructions"
#define MBL_DMT64       0x0020  ///< needs "demote 64bits"
#define MBL_COMB        0x0040  ///< needs "combine" pass
#define MBL_PROP        0x0080  ///< needs 'propagation' pass
#define MBL_DEAD        0x0100  ///< needs "eliminate deads" pass
#define MBL_LIST        0x0200  ///< use/def lists are ready (not dirty)
#define MBL_INCONST     0x0400  ///< inconsistent lists: we are building them
#define MBL_CALL        0x0800  ///< call information has been built
#define MBL_BACKPROP    0x1000  ///< performed backprop_cc
#define MBL_NORET       0x2000  ///< dead end block: doesn't return execution control
#define MBL_DSLOT       0x4000  ///< block for delay slot
#define MBL_VALRANGES   0x8000  ///< should optimize using value ranges
#define MBL_KEEP       0x10000  ///< do not remove even if unreachable
  //@}
  ea_t start;                   ///< start address
  ea_t end;                     ///< end address
                                ///< note: we cannot rely on start/end addresses
                                ///<       very much because instructions are
                                ///<       propagated between blocks
  minsn_t *head;                ///< pointer to the first instruction of the block
  minsn_t *tail;                ///< pointer to the last instruction of the block
  mba_t *mba;                   ///< the parent micro block array
  int serial;                   ///< block number
  mblock_type_t type;           ///< block type (BLT_NONE - not computed yet)
 
  mlist_t dead_at_start;        ///< data that is dead at the block entry
  mlist_t mustbuse;             ///< data that must be used by the block
  mlist_t maybuse;              ///< data that may  be used by the block
  mlist_t mustbdef;             ///< data that must be defined by the block
  mlist_t maybdef;              ///< data that may  be defined by the block
  mlist_t dnu;                  ///< data that is defined but not used in the block
 
  sval_t maxbsp;                ///< maximal sp value in the block (0...stacksize)
  sval_t minbstkref;            ///< lowest stack location accessible with indirect
                                ///< addressing (offset from the stack bottom)
                                ///< initially it is 0 (not computed)
  sval_t minbargref;            ///< the same for arguments
 
  intvec_t predset;             ///< control flow graph: list of our predecessors
                                ///< use npred() and pred() to access it
  intvec_t succset;             ///< control flow graph: list of our successors
                                ///< use nsucc() and succ() to access it
 
  // the exact size of this class is not documented, there may be more fields
  char reserved[];
 
  void mark_lists_dirty(void) { flags &= ~MBL_LIST; request_propagation(); }
  void request_propagation(void) { flags |= MBL_PROP; }
  bool needs_propagation(void) const { return (flags & MBL_PROP) != 0; }
  void request_demote64(void) { flags |= MBL_DMT64; }
  bool lists_dirty(void) const { return (flags & MBL_LIST) == 0; }
  bool lists_ready(void) const { return (flags & (MBL_LIST|MBL_INCONST)) == MBL_LIST; }
  int make_lists_ready(void) // returns number of changes
  {
    if ( lists_ready() )
      return 0;
    return build_lists(false);
  }
 
  /// Get number of block predecessors
  int npred(void) const { return predset.size(); } // number of xrefs to the block
  /// Get number of block successors
  int nsucc(void) const { return succset.size(); } // number of xrefs from the block
  // Get predecessor number N
  int pred(int n) const { return predset[n]; }
  // Get successor number N
  int succ(int n) const { return succset[n]; }
 
  mblock_t(void) = delete;
  virtual ~mblock_t(void);
  HEXRAYS_MEMORY_ALLOCATION_FUNCS()
  bool empty(void) const { return head == nullptr; }
 
  /// Print block contents.
  /// \param vp print helpers class. it can be used to direct the printed
  ///           info to any destination
  void hexapi print(vd_printer_t &vp) const;
 
  /// Dump block info.
  /// This function is useful for debugging, see mba_t::dump for info
  void hexapi dump(void) const;
  AS_PRINTF(2, 0) void hexapi vdump_block(const char *title, va_list va) const;
  AS_PRINTF(2, 3) void dump_block(const char *title, ...) const
  {
    va_list va;
    va_start(va, title);
    vdump_block(title, va);
    va_end(va);
  }
 
  //-----------------------------------------------------------------------
  // Functions to insert/remove insns during the microcode optimization phase.
  // See codegen_t, microcode_filter_t, udcall_t classes for the initial
  // microcode generation.
  //-----------------------------------------------------------------------
  /// Insert instruction into the doubly linked list
  /// \param nm new instruction
  /// \param om existing instruction, part of the doubly linked list
  ///           if nullptr, then the instruction will be inserted at the beginning
  ///           of the list
  /// NM will be inserted immediately after OM
  /// \return pointer to NM
  minsn_t *hexapi insert_into_block(minsn_t *nm, minsn_t *om);
 
  /// Remove instruction from the doubly linked list
  /// \param m instruction to remove
  /// The removed instruction is not deleted, the caller gets its ownership
  /// \return pointer to the next instruction
  minsn_t *hexapi remove_from_block(minsn_t *m);
 
  //-----------------------------------------------------------------------
  // Iterator over instructions and operands
  //-----------------------------------------------------------------------
  /// Visit all instructions.
  /// This function visits subinstructions too.
  /// \param mv instruction visitor
  /// \return zero or the value returned by mv.visit_insn()
  /// See also mba_t::for_all_topinsns()
  int hexapi for_all_insns(minsn_visitor_t &mv);
 
  /// Visit all operands.
  /// This function visit subinstruction operands too.
  /// \param mv operand visitor
  /// \return zero or the value returned by mv.visit_mop()
  int hexapi for_all_ops(mop_visitor_t &mv);
 
  /// Visit all operands that use LIST.
  /// \param list ptr to the list of locations. it may be modified:
  ///             parts that get redefined by the instructions in [i1,i2)
  ///             will be deleted.
  /// \param i1   starting instruction. must be a top level insn.
  /// \param i2   ending instruction (excluded). must be a top level insn.
  /// \param mmv  operand visitor
  /// \return zero or the value returned by mmv.visit_mop()
  int hexapi for_all_uses(
        mlist_t *list,
        minsn_t *i1,
        minsn_t *i2,
        mlist_mop_visitor_t &mmv);
 
  //-----------------------------------------------------------------------
  // Optimization functions
  //-----------------------------------------------------------------------
  /// Optimize one instruction in the context of the block.
  /// \param m pointer to a top level instruction
  /// \param optflags combination of \ref OPTI_ bits
  /// \return number of changes made to the block
  /// This function may change other instructions in the block too.
  /// However, it will not destroy top level instructions (it may convert them
  /// to nop's). This function performs only intrablock modifications.
  /// See also minsn_t::optimize_solo()
  int hexapi optimize_insn(minsn_t *m, int optflags=OPTI_MINSTKREF|OPTI_COMBINSNS);
 
  /// Optimize a basic block.
  /// Usually there is no need to call this function explicitly because the
  /// decompiler will call it itself if optinsn_t::func or optblock_t::func
  /// return non-zero.
  /// \return number of changes made to the block
  int hexapi optimize_block(void);
 
  /// Build def-use lists and eliminate deads.
  /// \param kill_deads do delete dead instructions?
  /// \return the number of eliminated instructions
  /// Better mblock_t::call make_lists_ready() rather than this function.
  int hexapi build_lists(bool kill_deads);
 
  /// Remove a jump at the end of the block if it is useless.
  /// This function preserves any side effects when removing a useless jump.
  /// Both conditional and unconditional jumps are handled (and jtbl too).
  /// This function deletes useless jumps, not only replaces them with a nop.
  /// (please note that \optimize_insn does not handle useless jumps).
  /// \return number of changes made to the block
  int hexapi optimize_useless_jump(void);
 
  //-----------------------------------------------------------------------
  // Functions that build with use/def lists. These lists are used to
  // reprsent list of registers and stack locations that are either modified
  // or accessed by microinstructions.
  //-----------------------------------------------------------------------
  /// Append use-list of an operand.
  /// This function calculates list of locations that may or must be used
  /// by the operand and appends it to LIST.
  /// \param list    ptr to the output buffer. we will append to it.
  /// \param op      operand to calculate the use list of
  /// \param maymust should we calculate 'may-use' or 'must-use' list?
  ///                see \ref maymust_t for more details.
  /// \param mask    if only part of the operand should be considered,
  ///                a bitmask can be used to specify which part.
  ///                example: op=AX,mask=0xFF means that we will consider only AL.
  void hexapi append_use_list(
        mlist_t *list,
        const mop_t &op,
        maymust_t maymust,
        bitrange_t mask=MAXRANGE) const;
 
  /// Append def-list of an operand.
  /// This function calculates list of locations that may or must be modified
  /// by the operand and appends it to LIST.
  /// \param list    ptr to the output buffer. we will append to it.
  /// \param op      operand to calculate the def list of
  /// \param maymust should we calculate 'may-def' or 'must-def' list?
  ///                see \ref maymust_t for more details.
  void hexapi append_def_list(
        mlist_t *list,
        const mop_t &op,
        maymust_t maymust) const;
 
  /// Build use-list of an instruction.
  /// This function calculates list of locations that may or must be used
  /// by the instruction. Examples:
  ///   "ldx ds.2, eax.4, ebx.4", may-list: all aliasable memory
  ///   "ldx ds.2, eax.4, ebx.4", must-list: empty
  /// Since LDX uses EAX for indirect access, it may access any aliasable
  /// memory. On the other hand, we cannot tell for sure which memory cells
  /// will be accessed, this is why the must-list is empty.
  /// \param ins     instruction to calculate the use list of
  /// \param maymust should we calculate 'may-use' or 'must-use' list?
  ///                see \ref maymust_t for more details.
  /// \return the calculated use-list
  mlist_t hexapi build_use_list(const minsn_t &ins, maymust_t maymust) const;
 
  /// Build def-list of an instruction.
  /// This function calculates list of locations that may or must be modified
  /// by the instruction. Examples:
  ///   "stx ebx.4, ds.2, eax.4", may-list: all aliasable memory
  ///   "stx ebx.4, ds.2, eax.4", must-list: empty
  /// Since STX uses EAX for indirect access, it may modify any aliasable
  /// memory. On the other hand, we cannot tell for sure which memory cells
  /// will be modified, this is why the must-list is empty.
  /// \param ins     instruction to calculate the def list of
  /// \param maymust should we calculate 'may-def' or 'must-def' list?
  ///                see \ref maymust_t for more details.
  /// \return the calculated def-list
  mlist_t hexapi build_def_list(const minsn_t &ins, maymust_t maymust) const;
 
  //-----------------------------------------------------------------------
  // The use/def lists can be used to search for interesting instructions
  //-----------------------------------------------------------------------
  /// Is the list used by the specified instruction range?
  /// \param list list of locations. LIST may be modified by the function:
  ///             redefined locations will be removed from it.
  /// \param i1   starting instruction of the range (must be a top level insn)
  /// \param i2   end instruction of the range (must be a top level insn)
  ///             i2 is excluded from the range. it can be specified as nullptr.
  ///             i1 and i2 must belong to the same block.
  /// \param maymust should we search in 'may-access' or 'must-access' mode?
  bool is_used(mlist_t *list, const minsn_t *i1, const minsn_t *i2, maymust_t maymust=MAY_ACCESS) const
    { return find_first_use(list, i1, i2, maymust) != nullptr; }
 
  /// Find the first insn that uses the specified list in the insn range.
  /// \param list list of locations. LIST may be modified by the function:
  ///             redefined locations will be removed from it.
  /// \param i1   starting instruction of the range (must be a top level insn)
  /// \param i2   end instruction of the range (must be a top level insn)
  ///             i2 is excluded from the range. it can be specified as nullptr.
  ///             i1 and i2 must belong to the same block.
  /// \param maymust should we search in 'may-access' or 'must-access' mode?
  /// \return pointer to such instruction or nullptr.
  ///         Upon return LIST will contain only locations not redefined
  ///         by insns [i1..result]
  const minsn_t *hexapi find_first_use(mlist_t *list, const minsn_t *i1, const minsn_t *i2, maymust_t maymust=MAY_ACCESS) const;
  minsn_t *find_first_use(mlist_t *list, minsn_t *i1, const minsn_t *i2, maymust_t maymust=MAY_ACCESS) const
  {
    return CONST_CAST(minsn_t*)(find_first_use(list,
                                               CONST_CAST(const minsn_t*)(i1),
                                               i2,
                                               maymust));
  }
 
  /// Is the list redefined by the specified instructions?
  /// \param list list of locations to check.
  /// \param i1   starting instruction of the range (must be a top level insn)
  /// \param i2   end instruction of the range (must be a top level insn)
  ///             i2 is excluded from the range. it can be specified as nullptr.
  ///             i1 and i2 must belong to the same block.
  /// \param maymust should we search in 'may-access' or 'must-access' mode?
  bool is_redefined(
        const mlist_t &list,
        const minsn_t *i1,
        const minsn_t *i2,
        maymust_t maymust=MAY_ACCESS) const
  {
    return find_redefinition(list, i1, i2, maymust) != nullptr;
  }
 
  /// Find the first insn that redefines any part of the list in the insn range.
  /// \param list list of locations to check.
  /// \param i1   starting instruction of the range (must be a top level insn)
  /// \param i2   end instruction of the range (must be a top level insn)
  ///             i2 is excluded from the range. it can be specified as nullptr.
  ///             i1 and i2 must belong to the same block.
  /// \param maymust should we search in 'may-access' or 'must-access' mode?
  /// \return pointer to such instruction or nullptr.
  const minsn_t *hexapi find_redefinition(
        const mlist_t &list,
        const minsn_t *i1,
        const minsn_t *i2,
        maymust_t maymust=MAY_ACCESS) const;
  minsn_t *find_redefinition(
        const mlist_t &list,
        minsn_t *i1,
        const minsn_t *i2,
        maymust_t maymust=MAY_ACCESS) const
  {
    return CONST_CAST(minsn_t*)(find_redefinition(list,
                                                  CONST_CAST(const minsn_t*)(i1),
                                                  i2,
                                                  maymust));
  }
 
  /// Is the right hand side of the instruction redefined the insn range?
  /// "right hand side" corresponds to the source operands of the instruction.
  /// \param ins instruction to consider
  /// \param i1   starting instruction of the range (must be a top level insn)
  /// \param i2   end instruction of the range (must be a top level insn)
  ///             i2 is excluded from the range. it can be specified as nullptr.
  ///             i1 and i2 must belong to the same block.
  bool hexapi is_rhs_redefined(const minsn_t *ins, const minsn_t *i1, const minsn_t *i2) const;
 
  /// Find the instruction that accesses the specified operand.
  /// This function search inside one block.
  /// \param op     operand to search for
  /// \param parent ptr to ptr to a top level instruction.
  ///               denotes the beginning of the search range.
  /// \param mend   end instruction of the range (must be a top level insn)
  ///               mend is excluded from the range. it can be specified as nullptr.
  ///               parent and mend must belong to the same block.
  /// \param fdflags combination of \ref FD_ bits
  /// \return       the instruction that accesses the operand. this instruction
  ///               may be a sub-instruction. to find out the top level
  ///               instruction, check out *p_i1.
  ///               nullptr means 'not found'.
  minsn_t *hexapi find_access(
        const mop_t &op,
        minsn_t **parent,
        const minsn_t *mend,
        int fdflags) const;
  /// \defgroup FD_ bits for mblock_t::find_access
  //@{
#define FD_BACKWARD 0x0000  ///< search direction
#define FD_FORWARD  0x0001  ///< search direction
#define FD_USE      0x0000  ///< look for use
#define FD_DEF      0x0002  ///< look for definition
#define FD_DIRTY    0x0004  ///< ignore possible implicit definitions
                            ///< by function calls and indirect memory access
  //@}
 
  // Convenience functions:
  minsn_t *find_def(
        const mop_t &op,
        minsn_t **p_i1,
        const minsn_t *i2,
        int fdflags)
  {
    return find_access(op, p_i1, i2, fdflags|FD_DEF);
  }
  minsn_t *find_use(
        const mop_t &op,
        minsn_t **p_i1,
        const minsn_t *i2,
        int fdflags)
  {
    return find_access(op, p_i1, i2, fdflags|FD_USE);
  }
 
  /// Find possible values for a block.
  /// \param res     set of value ranges
  /// \param vivl    what to search for
  /// \param vrflags combination of \ref VR_ bits
  bool hexapi get_valranges(
        valrng_t *res,
        const vivl_t &vivl,
        int vrflags) const;
 
  /// Find possible values for an instruction.
  /// \param res     set of value ranges
  /// \param vivl    what to search for
  /// \param m       insn to search value ranges at. \sa VR_ bits
  /// \param vrflags combination of \ref VR_ bits
  bool hexapi get_valranges(
        valrng_t *res,
        const vivl_t &vivl,
        const minsn_t *m,
        int vrflags) const;
 
  /// \defgroup VR_ bits for get_valranges
  //@{
#define VR_AT_START 0x0000    ///< get value ranges before the instruction or
                              ///< at the block start (if M is nullptr)
#define VR_AT_END   0x0001    ///< get value ranges after the instruction or
                              ///< at the block end, just after the last
                              ///< instruction (if M is nullptr)
#define VR_EXACT    0x0002    ///< find exact match. if not set, the returned
                              ///< valrng size will be >= vivl.size
  //@}
 
  /// Erase the instruction (convert it to nop) and mark the lists dirty.
  /// This is the recommended function to use because it also marks the block
  /// use-def lists dirty.
  void make_nop(minsn_t *m) { m->_make_nop(); mark_lists_dirty(); }
 
  /// Calculate number of regular instructions in the block.
  /// Assertions are skipped by this function.
  /// \return Number of non-assertion instructions in the block.
  size_t hexapi get_reginsn_qty(void) const;
 
  bool is_call_block(void) const { return tail != nullptr && is_mcode_call(tail->opcode); }
  bool is_unknown_call(void) const { return tail != nullptr && tail->is_unknown_call(); }
  bool is_nway(void) const { return type == BLT_NWAY; }
  bool is_branch(void) const { return type == BLT_2WAY && tail->d.t == mop_b; }
  bool is_simple_goto_block(void) const
  {
    return get_reginsn_qty() == 1
        && tail->opcode == m_goto
        && tail->l.t == mop_b;
  }
  bool is_simple_jcnd_block() const
  {
    return is_branch()
        && npred() == 1
        && get_reginsn_qty() == 1
        && is_mcode_convertible_to_set(tail->opcode);
  }
};
//-------------------------------------------------------------------------
/// Warning ids
enum warnid_t
{
  WARN_VARARG_REGS,   ///<  0 cannot handle register arguments in vararg function, discarded them
  WARN_ILL_PURGED,    ///<  1 odd caller purged bytes %d, correcting
  WARN_ILL_FUNCTYPE,  ///<  2 invalid function type '%s' has been ignored
  WARN_VARARG_TCAL,   ///<  3 cannot handle tail call to vararg
  WARN_VARARG_NOSTK,  ///<  4 call vararg without local stack
  WARN_VARARG_MANY,   ///<  5 too many varargs, some ignored
  WARN_ADDR_OUTARGS,  ///<  6 cannot handle address arithmetics in outgoing argument area of stack frame -- unused
  WARN_DEP_UNK_CALLS, ///<  7 found interdependent unknown calls
  WARN_ILL_ELLIPSIS,  ///<  8 erroneously detected ellipsis type has been ignored
  WARN_GUESSED_TYPE,  ///<  9 using guessed type %s;
  WARN_EXP_LINVAR,    ///< 10 failed to expand a linear variable
  WARN_WIDEN_CHAINS,  ///< 11 failed to widen chains
  WARN_BAD_PURGED,    ///< 12 inconsistent function type and number of purged bytes
  WARN_CBUILD_LOOPS,  ///< 13 too many cbuild loops
  WARN_NO_SAVE_REST,  ///< 14 could not find valid save-restore pair for %s
  WARN_ODD_INPUT_REG, ///< 15 odd input register %s
  WARN_ODD_ADDR_USE,  ///< 16 odd use of a variable address
  WARN_MUST_RET_FP,   ///< 17 function return type is incorrect (must be floating point)
  WARN_ILL_FPU_STACK, ///< 18 inconsistent fpu stack
  WARN_SELFREF_PROP,  ///< 19 self-referencing variable has been detected
  WARN_WOULD_OVERLAP, ///< 20 variables would overlap: %s
  WARN_ARRAY_INARG,   ///< 21 array has been used for an input argument
  WARN_MAX_ARGS,      ///< 22 too many input arguments, some ignored
  WARN_BAD_FIELD_TYPE,///< 23 incorrect structure member type for %s::%s, ignored
  WARN_WRITE_CONST,   ///< 24 write access to const memory at %a has been detected
  WARN_BAD_RETVAR,    ///< 25 wrong return variable
  WARN_FRAG_LVAR,     ///< 26 fragmented variable at %s may be wrong
  WARN_HUGE_STKOFF,   ///< 27 exceedingly huge offset into the stack frame
  WARN_UNINITED_REG,  ///< 28 reference to an uninitialized register has been removed: %s
  WARN_FIXED_MACRO,   ///< 29 fixed broken macro-insn
  WARN_WRONG_VA_OFF,  ///< 30 wrong offset of va_list variable
  WARN_CR_NOFIELD,    ///< 31 CONTAINING_RECORD: no field '%s' in struct '%s' at %d
  WARN_CR_BADOFF,     ///< 32 CONTAINING_RECORD: too small offset %d for struct '%s'
  WARN_BAD_STROFF,    ///< 33 user specified stroff has not been processed: %s
  WARN_BAD_VARSIZE,   ///< 34 inconsistent variable size for '%s'
  WARN_UNSUPP_REG,    ///< 35 unsupported processor register '%s'
  WARN_UNALIGNED_ARG, ///< 36 unaligned function argument '%s'
  WARN_BAD_STD_TYPE,  ///< 37 corrupted or unexisting local type '%s'
  WARN_BAD_CALL_SP,   ///< 38 bad sp value at call
  WARN_MISSED_SWITCH, ///< 39 wrong markup of switch jump, skipped it
  WARN_BAD_SP,        ///< 40 positive sp value %a has been found
  WARN_BAD_STKPNT,    ///< 41 wrong sp change point
  WARN_UNDEF_LVAR,    ///< 42 variable '%s' is possibly undefined
  WARN_JUMPOUT,       ///< 43 control flows out of bounds
  WARN_BAD_VALRNG,    ///< 44 values range analysis failed
  WARN_BAD_SHADOW,    ///< 45 ignored the value written to the shadow area of the succeeding call
  WARN_OPT_VALRNG,    ///< 46 conditional instruction was optimized away because %s
  WARN_RET_LOCREF,    ///< 47 returning address of temporary local variable '%s'
  WARN_BAD_MAPDST,    ///< 48 too short map destination '%s' for variable '%s'
  WARN_BAD_INSN,      ///< 49 bad instruction
  WARN_ODD_ABI,       ///< 50 encountered odd instruction for the current ABI
  WARN_UNBALANCED_STACK, ///< 51 unbalanced stack, ignored a potential tail call
 
  WARN_OPT_VALRNG2,   ///< 52 mask 0x%X is shortened because %s <= 0x%X"
 
  WARN_OPT_VALRNG3,   ///< 53 masking with 0X%X was optimized away because %s <= 0x%X
  WARN_OPT_USELESS_JCND, ///< 54 simplified comparisons for '%s': %s became %s
  WARN_MAX,           ///< may be used in notes as a placeholder when the
                      ///< warning id is not available
};
 
/// Warning instances
struct hexwarn_t
{
  ea_t ea;            ///< Address where the warning occurred
  warnid_t id;        ///< Warning id
  qstring text;       ///< Fully formatted text of the warning
  DECLARE_COMPARISONS(hexwarn_t)
  {
    if ( ea < r.ea )
      return -1;
    if ( ea > r.ea )
      return 1;
    if ( id < r.id )
      return -1;
    if ( id > r.id )
      return 1;
    return strcmp(text.c_str(), r.text.c_str());
  }
};
DECLARE_TYPE_AS_MOVABLE(hexwarn_t);
typedef qvector<hexwarn_t> hexwarns_t;
 
//-------------------------------------------------------------------------
/// Microcode maturity levels
enum mba_maturity_t
{
  MMAT_ZERO,         ///< microcode does not exist
  MMAT_GENERATED,    ///< generated microcode
  MMAT_PREOPTIMIZED, ///< preoptimized pass is complete
  MMAT_LOCOPT,       ///< local optimization of each basic block is complete.
                     ///< control flow graph is ready too.
  MMAT_CALLS,        ///< detected call arguments
  MMAT_GLBOPT1,      ///< performed the first pass of global optimization
  MMAT_GLBOPT2,      ///< most global optimization passes are done
  MMAT_GLBOPT3,      ///< completed all global optimization. microcode is fixed now.
  MMAT_LVARS,        ///< allocated local variables
};
 
//-------------------------------------------------------------------------
enum memreg_index_t  ///< memory region types
{
  MMIDX_GLBLOW,      ///< global memory: low part
  MMIDX_LVARS,       ///< stack: local variables
  MMIDX_RETADDR,     ///< stack: return address
  MMIDX_SHADOW,      ///< stack: shadow arguments
  MMIDX_ARGS,        ///< stack: regular stack arguments
  MMIDX_GLBHIGH,     ///< global memory: high part
};
 
//-------------------------------------------------------------------------
/// Ranges to decompile. Either a function or an explicit vector of ranges.
struct mba_ranges_t
{
  func_t *pfn = nullptr; ///< function to decompile. if not null, then function mode.
  rangevec_t ranges;     ///< snippet mode: ranges to decompile.
                         ///< function mode: list of outlined ranges
  mba_ranges_t(func_t *_pfn=nullptr) : pfn(_pfn) {}
  mba_ranges_t(const rangevec_t &r) : ranges(r) {}
  ea_t start(void) const { return (pfn != nullptr ? *pfn : ranges[0]).start_ea; }
  bool empty(void) const { return pfn == nullptr && ranges.empty(); }
  void clear(void) { pfn = nullptr; ranges.clear(); }
  bool is_snippet(void) const { return pfn == nullptr; }
  bool hexapi range_contains(ea_t ea) const;
  bool is_fragmented(void) const
  {
    int n_frags = ranges.size();
    if ( pfn != nullptr )
      n_frags += pfn->tailqty + 1;
    return n_frags > 1;
  }
};
 
/// Item iterator of arbitrary rangevec items
struct range_item_iterator_t
{
  const rangevec_t *ranges = nullptr;
  const range_t *rptr = nullptr;       // pointer into ranges
  ea_t cur = BADADDR;                  // current address
  bool set(const rangevec_t &r);
  bool next_code(void);
  ea_t current(void) const { return cur; }
};
 
/// Item iterator for mba_ranges_t
struct mba_item_iterator_t
{
  range_item_iterator_t rii;
  func_item_iterator_t fii;
  bool func_items_done = true;
  bool set(const mba_ranges_t &mbr)
  {
    bool ok = false;
    if ( mbr.pfn != nullptr )
    {
      ok = fii.set(mbr.pfn);
      if ( ok )
        func_items_done = false;
    }
    if ( rii.set(mbr.ranges) )
      ok = true;
    return ok;
  }
  bool next_code(void)
  {
    bool ok = false;
    if ( !func_items_done )
    {
      ok = fii.next_code();
      if ( !ok )
        func_items_done = true;
    }
    if ( !ok )
      ok = rii.next_code();
    return ok;
  }
  ea_t current(void) const
  {
    return func_items_done ? rii.current() : fii.current();
  }
};
 
/// Chunk iterator of arbitrary rangevec items
struct range_chunk_iterator_t
{
  const range_t *rptr = nullptr;          // pointer into ranges
  const range_t *rend = nullptr;
  bool set(const rangevec_t &r) { rptr = r.begin(); rend = r.end(); return rptr != rend; }
  bool next(void) { return ++rptr != rend; }
  const range_t &chunk(void) const { return *rptr; }
};
 
/// Chunk iterator for mba_ranges_t
struct mba_range_iterator_t
{
  range_chunk_iterator_t rii;
  func_tail_iterator_t fii;     // this is used if rii.rptr==nullptr
  bool is_snippet(void) const { return rii.rptr != nullptr; }
  bool set(const mba_ranges_t &mbr)
  {
    if ( mbr.is_snippet() )
      return rii.set(mbr.ranges);
    else
      return fii.set(mbr.pfn);
  }
  bool next(void)
  {
    if ( is_snippet() )
      return rii.next();
    else
      return fii.next();
  }
  const range_t &chunk(void) const
  {
    return is_snippet() ? rii.chunk() : fii.chunk();
  }
};
 
//-------------------------------------------------------------------------
/// Array of micro blocks representing microcode for a decompiled function.
/// The first micro block is the entry point, the last one is the exit point.
/// The entry and exit blocks are always empty. The exit block is generated
/// at MMAT_LOCOPT maturity level.
class mba_t
{
  DECLARE_UNCOPYABLE(mba_t)
  uint32 flags;
  uint32 flags2;
 
public:
                     // bits to describe the microcode, set by the decompiler
#define MBA_PRCDEFS  0x00000001 ///< use precise defeas for chain-allocated lvars
#define MBA_NOFUNC   0x00000002 ///< function is not present, addresses might be wrong
#define MBA_PATTERN  0x00000004 ///< microcode pattern, callinfo is present
#define MBA_LOADED   0x00000008 ///< loaded gdl, no instructions (debugging)
#define MBA_RETFP    0x00000010 ///< function returns floating point value
#define MBA_SPLINFO  0x00000020 ///< (final_type ? idb_spoiled : spoiled_regs) is valid
#define MBA_PASSREGS 0x00000040 ///< has mcallinfo_t::pass_regs
#define MBA_THUNK    0x00000080 ///< thunk function
#define MBA_CMNSTK   0x00000100 ///< stkvars+stkargs should be considered as one area
 
                     // bits to describe analysis stages and requests
#define MBA_PREOPT   0x00000200 ///< preoptimization stage complete
#define MBA_CMBBLK   0x00000400 ///< request to combine blocks
#define MBA_ASRTOK   0x00000800 ///< assertions have been generated
#define MBA_CALLS    0x00001000 ///< callinfo has been built
#define MBA_ASRPROP  0x00002000 ///< assertion have been propagated
#define MBA_SAVRST   0x00004000 ///< save-restore analysis has been performed
#define MBA_RETREF   0x00008000 ///< return type has been refined
#define MBA_GLBOPT   0x00010000 ///< microcode has been optimized globally
#define MBA_LVARS0   0x00040000 ///< lvar pre-allocation has been performed
#define MBA_LVARS1   0x00080000 ///< lvar real allocation has been performed
#define MBA_DELPAIRS 0x00100000 ///< pairs have been deleted once
#define MBA_CHVARS   0x00200000 ///< can verify chain varnums
 
                     // bits that can be set by the caller:
#define MBA_SHORT    0x00400000 ///< use short display
#define MBA_COLGDL   0x00800000 ///< display graph after each reduction
#define MBA_INSGDL   0x01000000 ///< display instruction in graphs
#define MBA_NICE     0x02000000 ///< apply transformations to c code
#define MBA_REFINE   0x04000000 ///< may refine return value size
#define MBA_WINGR32  0x10000000 ///< use wingraph32
#define MBA_NUMADDR  0x20000000 ///< display definition addresses for numbers
#define MBA_VALNUM   0x40000000 ///< display value numbers
 
#define MBA_INITIAL_FLAGS  (MBA_INSGDL|MBA_NICE|MBA_CMBBLK|MBA_REFINE\
        |MBA_PRCDEFS|MBA_WINGR32|MBA_VALNUM)
 
#define MBA2_LVARNAMES_OK  0x00000001 ///< may verify lvar_names?
#define MBA2_LVARS_RENAMED 0x00000002 ///< accept empty names now?
#define MBA2_OVER_CHAINS   0x00000004 ///< has overlapped chains?
#define MBA2_VALRNG_DONE   0x00000008 ///< calculated valranges?
#define MBA2_IS_CTR        0x00000010 ///< is constructor?
#define MBA2_IS_DTR        0x00000020 ///< is destructor?
#define MBA2_ARGIDX_OK     0x00000040 ///< may verify input argument list?
#define MBA2_NO_DUP_CALLS  0x00000080 ///< forbid multiple calls with the same ea
#define MBA2_NO_DUP_LVARS  0x00000100 ///< forbid multiple lvars with the same ea
#define MBA2_UNDEF_RETVAR  0x00000200 ///< return value is undefined
#define MBA2_ARGIDX_SORTED 0x00000400 ///< args finally sorted according to ABI
                                      ///< (e.g. reverse stkarg order in Borland)
#define MBA2_CODE16_BIT    0x00000800 ///< the code16 bit removed
#define MBA2_STACK_RETVAL  0x00001000 ///< the return value is on the stack
#define MBA2_HAS_OUTLINES  0x00002000 ///< calls to outlined code have been inlined
#define MBA2_NO_FRAME      0x00004000 ///< do not use function frame info (only snippet mode)
#define MBA2_PROP_COMPLEX  0x00008000 ///< allow propagation of more complex variable definitions
 
#define MBA2_DONT_VERIFY   0x80000000 ///< Do not verify microcode. This flag
                                      ///< is recomended to be set only when
                                      ///< debugging decompiler plugins
 
#define MBA2_INITIAL_FLAGS  (MBA2_LVARNAMES_OK|MBA2_LVARS_RENAMED)
 
#define MBA2_ALL_FLAGS    0x0000FFFF
 
  bool precise_defeas(void) const { return (flags & MBA_PRCDEFS) != 0; }
  bool optimized(void)      const { return (flags & MBA_GLBOPT) != 0; }
  bool short_display(void)  const { return (flags & MBA_SHORT ) != 0; }
  bool show_reduction(void) const { return (flags & MBA_COLGDL) != 0; }
  bool graph_insns(void)    const { return (flags & MBA_INSGDL) != 0; }
  bool loaded_gdl(void)     const { return (flags & MBA_LOADED) != 0; }
  bool should_beautify(void)const { return (flags & MBA_NICE  ) != 0; }
  bool rtype_refined(void)  const { return (flags & MBA_RETREF) != 0; }
  bool may_refine_rettype(void) const { return (flags & MBA_REFINE) != 0; }
  bool use_wingraph32(void) const { return (flags & MBA_WINGR32) != 0; }
  bool display_numaddrs(void) const { return (flags & MBA_NUMADDR) != 0; }
  bool display_valnums(void) const { return (flags & MBA_VALNUM) != 0; }
  bool is_pattern(void)     const { return (flags & MBA_PATTERN) != 0; }
  bool is_thunk(void)       const { return (flags & MBA_THUNK) != 0; }
  bool saverest_done(void)  const { return (flags & MBA_SAVRST) != 0; }
  bool callinfo_built(void) const { return (flags & MBA_CALLS) != 0; }
  bool really_alloc(void)   const { return (flags & MBA_LVARS0) != 0; }
  bool lvars_allocated(void)const { return (flags & MBA_LVARS1) != 0; }
  bool chain_varnums_ok(void)const { return (flags & MBA_CHVARS) != 0; }
  bool returns_fpval(void)  const { return (flags & MBA_RETFP) != 0; }
  bool has_passregs(void)   const { return (flags & MBA_PASSREGS) != 0; }
  bool generated_asserts(void) const { return (flags & MBA_ASRTOK) != 0; }
  bool propagated_asserts(void) const { return (flags & MBA_ASRPROP) != 0; }
  bool deleted_pairs(void) const { return (flags & MBA_DELPAIRS) != 0; }
  bool common_stkvars_stkargs(void) const { return (flags & MBA_CMNSTK) != 0; }
  bool lvar_names_ok(void) const { return (flags2 & MBA2_LVARNAMES_OK) != 0; }
  bool lvars_renamed(void) const { return (flags2 & MBA2_LVARS_RENAMED) != 0; }
  bool has_over_chains(void) const { return (flags2 & MBA2_OVER_CHAINS) != 0; }
  bool valranges_done(void) const { return (flags2 & MBA2_VALRNG_DONE) != 0; }
  bool argidx_ok(void) const { return (flags2 & MBA2_ARGIDX_OK) != 0; }
  bool argidx_sorted(void) const { return (flags2 & MBA2_ARGIDX_SORTED) != 0; }
  bool code16_bit_removed(void) const { return (flags2 & MBA2_CODE16_BIT) != 0; }
  bool has_stack_retval(void) const { return (flags2 & MBA2_STACK_RETVAL) != 0; }
  bool has_outlines(void) const { return (flags2 & MBA2_HAS_OUTLINES) != 0; }
  bool is_ctr(void) const { return (flags2 & MBA2_IS_CTR) != 0; }
  bool is_dtr(void) const { return (flags2 & MBA2_IS_DTR) != 0; }
  bool is_cdtr(void) const { return (flags2 & (MBA2_IS_CTR|MBA2_IS_DTR)) != 0; }
  bool prop_complex(void) const { return (flags2 & MBA2_PROP_COMPLEX) != 0; }
  int  get_mba_flags(void) const { return flags; }
  int  get_mba_flags2(void) const { return flags2; }
  void set_mba_flags(int f) { flags |= f; }
  void clr_mba_flags(int f) { flags &= ~f; }
  void set_mba_flags2(int f) { flags2 |= f; }
  void clr_mba_flags2(int f) { flags2 &= ~f; }
  void clr_cdtr(void) { flags2 &= ~(MBA2_IS_CTR|MBA2_IS_DTR); }
  int calc_shins_flags(void) const
  {
    int shins_flags = 0;
    if ( short_display() )
      shins_flags |= SHINS_SHORT;
    if ( display_valnums() )
      shins_flags |= SHINS_VALNUM;
    if ( display_numaddrs() )
      shins_flags |= SHINS_NUMADDR;
    return shins_flags;
  }
 
/*
                     +-----------+ <- inargtop
                     |   prmN    |
                     |   ...     | <- minargref
                     |   prm0    |
                     +-----------+ <- inargoff
                     |shadow_args|
                     +-----------+
                     |  retaddr  |
     frsize+frregs   +-----------+ <- initial esp  |
                     |  frregs   |                 |
           +frsize   +-----------+ <- typical ebp  |
                     |           |  |              |
                     |           |  | fpd          |
                     |           |  |              |
                     |  frsize   | <- current ebp  |
                     |           |                 |
                     |           |                 |
                     |           |                 | stacksize
                     |           |                 |
                     |           |                 |
                     |           | <- minstkref    |
 stkvar base off 0   +---..      |                 |    | current
                     |           |                 |    | stack
                     |           |                 |    | pointer
                     |           |                 |    | range
                     |tmpstk_size|                 |    | (what getspd() returns)
                     |           |                 |    |
                     |           |                 |    |
                     +-----------+ <- minimal sp   |    | offset 0 for the decompiler (vd)
 
  There is a detail that may add confusion when working with stack variables.
  The decompiler does not use the same stack offsets as IDA.
  The picture above should explain the difference:
  - IDA stkoffs are displayed on the left, decompiler stkoffs - on the right
  - Decompiler stkoffs are always >= 0
  - IDA stkoff==0 corresponds to stkoff==tmpstk_size in the decompiler
  - See stkoff_vd2ida and stkoff_ida2vd below to convert IDA stkoffs to vd stkoff
 
*/
 
  // convert a stack offset used in vd to a stack offset used in ida stack frame
  sval_t hexapi stkoff_vd2ida(sval_t off) const;
  // convert a ida stack frame offset to a stack offset used in vd
  sval_t hexapi stkoff_ida2vd(sval_t off) const;
  sval_t argbase() const
  {
    return retsize + stacksize;
  }
  static vdloc_t hexapi idaloc2vd(const argloc_t &loc, int width, sval_t spd);
  vdloc_t hexapi idaloc2vd(const argloc_t &loc, int width) const;
 
  static argloc_t hexapi vd2idaloc(const vdloc_t &loc, int width, sval_t spd);
  argloc_t hexapi vd2idaloc(const vdloc_t &loc, int width) const;
 
  bool is_stkarg(const lvar_t &v) const
  {
    return v.is_stk_var() && v.get_stkoff() >= inargoff;
  }
  member_t *get_stkvar(sval_t vd_stkoff, uval_t *poff) const;
  // get lvar location
  argloc_t get_ida_argloc(const lvar_t &v) const
  {
    return vd2idaloc(v.location, v.width);
  }
  mba_ranges_t mbr;
  ea_t entry_ea = BADADDR;
  ea_t last_prolog_ea = BADADDR;
  ea_t first_epilog_ea = BADADDR;
  int qty = 0;                  ///< number of basic blocks
  int npurged = -1;             ///< -1 - unknown
  cm_t cc = CM_CC_UNKNOWN;      ///< calling convention
  sval_t tmpstk_size = 0;       ///< size of the temporary stack part
                                ///< (which dynamically changes with push/pops)
  sval_t frsize = 0;            ///< size of local stkvars range in the stack frame
  sval_t frregs = 0;            ///< size of saved registers range in the stack frame
  sval_t fpd = 0;               ///< frame pointer delta
  int pfn_flags = 0;            ///< copy of func_t::flags
  int retsize = 0;              ///< size of return address in the stack frame
  int shadow_args = 0;          ///< size of shadow argument area
  sval_t fullsize = 0;          ///< Full stack size including incoming args
  sval_t stacksize = 0;         ///< The maximal size of the function stack including
                                ///< bytes allocated for outgoing call arguments
                                ///< (up to retaddr)
  sval_t inargoff = 0;          ///< offset of the first stack argument;
                                ///< after fix_scattered_movs() INARGOFF may
                                ///< be less than STACKSIZE