abg-reporter-priv.cc

// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// -*- Mode: C++ -*-
//
// Copyright (C) 2017-2024 Red Hat, Inc.
//
// Author: Dodji Seketeli
 
#include <libgen.h>
#include <algorithm>
#include "abg-comparison-priv.h"
#include "abg-reporter-priv.h"
 
namespace abigail
{
 
namespace comparison
{
 
/// Convert a number in bits into a number in bytes.
///
/// @param bits the number in bits to convert.
///
/// @return the number @p bits converted into bytes.
uint64_t
convert_bits_to_bytes(size_t bits)
{return bits / 8;}
 
/// Emit a numerical value to an output stream.
///
/// Depending on the current @ref diff_context, the number is going to
/// be emitted either in decimal or hexadecimal base.
///
/// @param value the value to emit.
///
/// @param ctxt the current diff context.
///
/// @param out the output stream to emit the numerical value to.
void
emit_num_value(uint64_t value, const diff_context& ctxt, ostream& out)
{
  if (ctxt.show_hex_values())
    out << std::hex << std::showbase ;
  else
    out << std::dec;
  out << value << std::dec << std::noshowbase;
}
 
/// Convert a bits value into a byte value if the current diff context
/// instructs us to do so.
///
/// @param bits the bits value to convert.
///
/// @param ctxt the current diff context to consider.
///
/// @return the resulting bits or bytes value, depending on what the
/// diff context instructs us to do.
uint64_t
maybe_convert_bits_to_bytes(uint64_t bits, const diff_context& ctxt)
{
  if (ctxt.show_offsets_sizes_in_bits())
    return bits;
  return convert_bits_to_bytes(bits);
}
 
/// Emit a message showing the numerical change between two values, to
/// a given output stream.
///
/// The function emits a message like
///
///      "XXX changes from old_bits to new_bits (in bits)"
///
/// or
///
///      "XXX changes from old_bits to new_bits (in bytes)"
///
/// Depending on if the current diff context instructs us to emit the
/// change in bits or bytes.  XXX is the string content of the @p what
/// parameter.
///
/// @param what the string that tells us what the change represents.
/// This is the "XXX" we refer to in the explanation above.
///
/// @param old_bits the initial value (which changed) in bits.
///
/// @param new_bits the final value (resulting from the change or @p
/// old_bits) in bits.
///
/// @param ctxt the current diff context to consider.
///
/// @param out the output stream to send the change message to.
///
/// @param show_bits_or_byte if this is true, then the message is
/// going to precise if the changed value is in bits or bytes.
/// Otherwise, no mention of that is made.
void
show_numerical_change(const string& what,
              uint64_t      old_bits,
              uint64_t      new_bits,
              const diff_context& ctxt,
              ostream&      out,
              bool show_bits_or_byte)
{
  bool can_convert_bits_to_bytes = (old_bits % 8 == 0 && new_bits % 8 == 0);
  uint64_t o = can_convert_bits_to_bytes
    ? maybe_convert_bits_to_bytes(old_bits, ctxt)
    : old_bits;
  uint64_t n = can_convert_bits_to_bytes
    ? maybe_convert_bits_to_bytes(new_bits, ctxt)
    : new_bits;
  string bits_or_bytes =
    (!can_convert_bits_to_bytes || ctxt.show_offsets_sizes_in_bits ())
    ? "bits"
    : "bytes";
 
  out << what << " changed from ";
  emit_num_value(o, ctxt, out);
  out << " to ";
  emit_num_value(n, ctxt, out);
  if (show_bits_or_byte)
    {
      out << " (in ";
      out << bits_or_bytes;
      out << ")";
    }
}
 
/// Emit a message showing the value of a numerical value representing
/// a size or an offset, preceded by a string.  The message is ended
/// by a part which says if the value is in bits or bytes.
///
/// @param what the string prefix of the message to emit.
///
/// @param value the numerical value to emit.
///
/// @param ctxt the diff context to take into account.
///
/// @param out the output stream to emit the message to.
void
show_offset_or_size(const string&   what,
            uint64_t        value,
            const diff_context& ctxt,
            ostream&        out)
{
  uint64_t v = value;
  bool can_convert_bits_to_bytes = (value % 8 == 0);
  if (can_convert_bits_to_bytes)
    v = maybe_convert_bits_to_bytes(v, ctxt);
  string bits_or_bytes =
    (!can_convert_bits_to_bytes || ctxt.show_offsets_sizes_in_bits())
    ? "bits"
    : "bytes";
 
  if (!what.empty())
    out << what << " ";
  emit_num_value(v, ctxt, out);
  out << " (in " << bits_or_bytes << ")";
}
 
/// Emit a message showing the value of a numerical value representing
/// a size or an offset.  The message is ended by a part which says if
/// the value is in bits or bytes.
///
/// @param value the numerical value to emit.
///
/// @param ctxt the diff context to take into account.
///
/// @param out the output stream to emit the message to.
void
show_offset_or_size(uint64_t        value,
            const diff_context& ctxt,
            ostream&        out)
{show_offset_or_size("", value, ctxt, out);}
 
/// Stream a string representation for a member function.
///
/// @param ctxt the current diff context.
///
/// @param mem_fn the member function to stream
///
/// @param out the output stream to send the representation to
void
represent(const diff_context& ctxt,
      method_decl_sptr mem_fn,
      ostream& out)
{
  if (!mem_fn || !is_member_function(mem_fn))
    return;
 
  method_decl_sptr meth =
    dynamic_pointer_cast<method_decl>(mem_fn);
  ABG_ASSERT(meth);
 
  out << "'" << mem_fn->get_pretty_representation() << "'";
  report_loc_info(meth, ctxt, out);
  if (get_member_function_is_virtual(mem_fn))
    {
 
      ssize_t voffset = get_member_function_vtable_offset(mem_fn);
      ssize_t biggest_voffset =
    is_class_type(meth->get_type()->get_class_type())->
    get_biggest_vtable_offset();
      if (voffset > -1)
    {
      out << ", virtual at voffset ";
      emit_num_value(get_member_function_vtable_offset(mem_fn),
             ctxt, out);
      out << "/";
      emit_num_value(biggest_voffset, ctxt, out);
    }
    }
 
  if (ctxt.show_linkage_names()
      && (mem_fn->get_symbol()))
    {
      out << "    {"
      << mem_fn->get_symbol()->get_id_string()
      << "}";
    }
  out << "\n";
}
 
/// Stream a string representation for a data member.
///
/// @param d the data member to stream
///
/// @param ctxt the current diff context.
///
/// @param out the output stream to send the representation to
///
/// @param indent the indentation string to use for the change report.
void
represent_data_member(var_decl_sptr d,
              const diff_context_sptr& ctxt,
              ostream& out,
              const string& indent)
{
  if (!is_data_member(d)
      || (!get_member_is_static(d) && !get_data_member_is_laid_out(d)))
    return;
 
  out << indent
      << "'"
      << d->get_pretty_representation(/*internal=*/false,
                      /*qualified_name=*/false)
      << "'";
  if (!get_member_is_static(d))
    {
      // Do not emit offset information for data member of a union
      // type because all data members of a union are supposed to be
      // at offset 0.
      if (!is_union_type(d->get_scope()))
    show_offset_or_size(", at offset",
                get_data_member_offset(d),
                *ctxt, out);
      report_loc_info(d, *ctxt, out);
    }
  out << "\n";
}
 
/// If a given @ref var_diff node carries a data member change in
/// which the offset of the data member actually changed, then emit a
/// string (to an output stream) that represents that offset change.
///
/// For instance, if the offset of the data member increased by 32
/// bits then the string emitted is going to be "by +32 bits".
///
/// If, on the other hand, the offset of the data member decreased by
/// 64 bits then the string emitted is going to be "by -64 bits".
///
/// This function is a sub-routine used by the reporting system.
///
/// @param diff the diff node that potentially carries the data member
/// change.
///
/// @param ctxt the context in which the diff is being reported.
///
/// @param out the output stream to emit the string to.
void
maybe_show_relative_offset_change(const var_diff_sptr &diff,
                  diff_context& ctxt,
                  ostream&  out)
{
  if (!ctxt.show_relative_offset_changes())
    return;
 
  var_decl_sptr o = diff->first_var();
  var_decl_sptr n = diff->second_var();
 
  uint64_t first_offset = get_data_member_offset(o),
    second_offset = get_data_member_offset(n);
 
  string sign;
  uint64_t change = 0;
  if (first_offset < second_offset)
    {
      sign = "+";
      change = second_offset - first_offset;
    }
  else if (first_offset > second_offset)
    {
      sign = "-";
      change = first_offset - second_offset;
    }
  else
    return;
 
  if (!ctxt.show_offsets_sizes_in_bits())
    change = convert_bits_to_bytes(change);
 
  string bits_or_bytes = ctxt.show_offsets_sizes_in_bits()
    ? "bits"
    : "bytes";
 
  out << " (by " << sign;
  emit_num_value(change, ctxt, out);
  out << " " << bits_or_bytes << ")";
}
 
/// If a given @ref var_diff node carries a hange in which the size of
/// the variable actually changed, then emit a string (to an output
/// stream) that represents that size change.
///
/// For instance, if the size of the variable increased by 32 bits
/// then the string emitted is going to be "by +32 bits".
///
/// If, on the other hand, the size of the variable decreased by 64
/// bits then the string emitted is going to be "by -64 bits".
///
/// This function is a sub-routine used by the reporting system.
///
/// @param diff the diff node that potentially carries the variable
/// change.
///
/// @param ctxt the context in which the diff is being reported.
///
/// @param out the output stream to emit the string to.
void
maybe_show_relative_size_change(const var_diff_sptr &diff,
                diff_context&       ctxt,
                ostream&        out)
{
  if (!ctxt.show_relative_offset_changes())
    return;
 
  var_decl_sptr o = diff->first_var();
  var_decl_sptr n = diff->second_var();
 
  uint64_t first_size = get_var_size_in_bits(o),
    second_size = get_var_size_in_bits(n);
 
  string sign;
  uint64_t change = 0;
  if (first_size < second_size)
    {
      sign = "+";
      change = second_size - first_size;
    }
  else if (first_size > second_size)
    {
      sign = "-";
      change = first_size - second_size;
    }
  else
    return;
 
  if (!ctxt.show_offsets_sizes_in_bits())
    change = convert_bits_to_bytes(change);
 
  string bits_or_bytes = ctxt.show_offsets_sizes_in_bits()
    ? "bits"
    : "bytes";
 
  out << " (by " << sign;
  emit_num_value(change, ctxt, out);
  out << " " << bits_or_bytes << ")";
}
 
/// Represent the changes carried by an instance of @ref var_diff that
/// represent a difference between two class data members.
///
/// @param diff diff the diff node to represent.
///
/// @param ctxt the diff context to use.
///
/// @param local_only if true, only display local changes.
///
/// @param out the output stream to send the representation to.
///
/// @param indent the indentation string to use for the change report.
void
represent(const var_diff_sptr   &diff,
      diff_context_sptr ctxt,
      ostream&      out,
      const string& indent,
      bool          local_only)
{
  if (!ctxt->get_reporter()->diff_to_be_reported(diff.get()))
    return;
 
  const var_decl_sptr o = diff->first_var();
  const var_decl_sptr n = diff->second_var();
  const bool o_anon = !!is_anonymous_data_member(o);
  const bool n_anon = !!is_anonymous_data_member(n);
  const bool is_strict_anonymous_data_member_change = o_anon && n_anon;
  const string o_name = (is_data_member_of_anonymous_class_or_union(o)
             ? o->get_name()
             : o->get_qualified_name());
  const string n_name = (is_data_member_of_anonymous_class_or_union(n)
             ? n->get_name()
             : n->get_qualified_name());
  const uint64_t o_size = get_var_size_in_bits(o);
  const uint64_t n_size = get_var_size_in_bits(n);
  const uint64_t o_offset = get_data_member_offset(o);
  const uint64_t n_offset = get_data_member_offset(n);
  const string o_pretty_representation =
    o->get_pretty_representation(/*internal=*/false, /*qualified_name=*/false);
  // no n_pretty_representation here as it's only needed in a couple of places
  const bool show_size_offset_changes = ctxt->get_allowed_category()
                    & SIZE_OR_OFFSET_CHANGE_CATEGORY;
 
  // Has the main diff text been output?
  bool emitted = false;
  // Are we continuing on a new line? (implies emitted)
  bool begin_with_and = false;
  // Have we reported a size change already?
  bool size_reported = false;
 
  //----------------------------------------------------------------
  // First we'll try to emit a report about the type change of this
  // var_decl_diff.
  //
  // In the context of that type change report, we need to keep in
  // mind that because we want to emit specific (useful) reports about
  // anonymous data member changes, we'll try to detect the various
  // scenarii that involve anonymous data member changes.
  //
  // Then, as a fallback method, we'll emit a more generic type change
  // report for the other generic type changes.
  //----------------------------------------------------------------
 
  if (is_strict_anonymous_data_member_change)
    {
      const string n_pretty_representation =
    n->get_pretty_representation(/*internal=*/false,
                     /*qualified_name=*/false);
      const type_base_sptr o_type = o->get_type(), n_type = n->get_type();
      if (o_pretty_representation != n_pretty_representation)
    {
      show_offset_or_size(indent + "anonymous data member at offset",
                  o_offset, *ctxt, out);
 
      out << " changed from:\n"
          << indent << "  " << o_pretty_representation << "\n"
          << indent << "to:\n"
          << indent << "  " << n_pretty_representation << "\n";
 
      begin_with_and = true;
      emitted = true;
    }
      else if (get_type_name(o_type) != get_type_name(n_type)
           && is_decl(o_type) && is_decl(n_type)
           && is_decl(o_type)->get_is_anonymous()
           && is_decl(n_type)->get_is_anonymous())
    {
      out << indent << "while looking at anonymous data member '"
          << o_pretty_representation << "':\n"
          << indent << "the internal name of that anonymous data member"
               " changed from:\n"
          << indent << " " << get_type_name(o_type) << "\n"
          << indent << "to:\n"
          << indent << " " << get_type_name(n_type) << "\n"
          << indent << " This is usually due to "
          << "an anonymous member type being added or removed from "
          << "the containing type\n";
 
      begin_with_and = true;
      emitted = true;
    }
    }
  else if (filtering::has_anonymous_data_member_change(diff))
    {
      ABG_ASSERT(o_anon != n_anon);
      // So we are looking at a non-anonymous data member change from
      // or to an anonymous data member.
      const string n_pretty_representation =
    n->get_pretty_representation(/*internal=*/false,
                     /*qualified_name=*/false);
      out << indent << (o_anon ? "anonymous " : "")
      << "data member " << o_pretty_representation;
      show_offset_or_size(" at offset", o_offset, *ctxt, out);
      out << " became " << (n_anon ? "anonymous " : "")
      << "data member '" << n_pretty_representation << "'\n";
 
      begin_with_and = true;
      emitted = true;
    }
 
  //
  // If we haven't succeeded in emitting a specific type change report
  // (mainly related to anonymous data data member changes) then let's
  // try to emit a more generic report about the type change.
  //
  // This is the fallback method outlined in the comment at the
  // beginning of this section.
  //
  if (!emitted)
    if (const diff_sptr d = diff->type_diff())
      {
    if (ctxt->get_reporter()->diff_to_be_reported(d.get()))
      {
        if (local_only)
          out << indent << "type '"
          << get_pretty_representation(o->get_type())
          << "' of '"
          << (o_anon ?
              string("anonymous data member")
              : o->get_qualified_name())
          << "' changed";
        else
          out << indent
          << "type of '"<< (o_anon ? "anonymous data member ": "")
          << o_pretty_representation << "' changed";
 
        if (d->currently_reporting())
          out << ", as being reported\n";
        else if (d->reported_once())
          out << ", as reported earlier\n";
        else
          {
        out << ":\n";
        d->report(out, indent + "  ");
          }
 
        begin_with_and = true;
        emitted = true;
        size_reported = true;
      }
      }
 
  //
  // Okay, now we are done with report type changes.  Let's report the
  // other potential kinds of changes.
  //
 
  if (!filtering::has_anonymous_data_member_change(diff) && o_name != n_name)
    {
      if (filtering::has_harmless_name_change(o, n)
      && !(ctxt->get_allowed_category()
           & HARMLESS_DECL_NAME_CHANGE_CATEGORY))
    ;
      else
    {
      if (begin_with_and)
        {
          out << indent << "and ";
          begin_with_and = false;
        }
      else if (!emitted)
        out << indent;
      else
        out << ", ";
      out << "name of '" << o_name << "' changed to '" << n_name << "'";
      report_loc_info(n, *ctxt, out);
      emitted = true;
    }
    }
 
  if (get_data_member_is_laid_out(o)
      != get_data_member_is_laid_out(n))
    {
      if (begin_with_and)
    {
      out << indent << "and ";
      begin_with_and = false;
    }
      else if (!emitted)
    out << indent << "'" << o_pretty_representation << "' ";
      else
    out << ", ";
      if (get_data_member_is_laid_out(o))
    out << "is no more laid out";
      else
    out << "now becomes laid out";
      emitted = true;
    }
  if (show_size_offset_changes)
    {
      if (o_offset != n_offset)
    {
      if (begin_with_and)
        {
          out << indent << "and ";
          begin_with_and = false;
        }
      else if (!emitted)
        {
          out << indent;
          if (is_strict_anonymous_data_member_change)
        out << "anonymous data member ";
          out << "'" << o_pretty_representation << "' ";
        }
      else
        out << ", ";
 
      show_numerical_change("offset", o_offset, n_offset, *ctxt, out);
      maybe_show_relative_offset_change(diff, *ctxt, out);
      emitted = true;
    }
 
      if (!size_reported && o_size != n_size)
    {
      if (begin_with_and)
        {
          out << indent << "and ";
          begin_with_and = false;
        }
      else if (!emitted)
        {
          out << indent;
          if (is_strict_anonymous_data_member_change)
        out << "anonymous data member ";
          out << "'" << o_pretty_representation << "' ";
        }
      else
        out << ", ";
 
      show_numerical_change("size", o_size, n_size, *ctxt, out);
      maybe_show_relative_size_change(diff, *ctxt, out);
      emitted = true;
    }
    }
  if (o->get_binding() != n->get_binding())
    {
      if (begin_with_and)
    {
      out << indent << "and ";
      begin_with_and = false;
    }
      else if (!emitted)
    out << indent << "'" << o_pretty_representation << "' ";
      else
    out << ", ";
      out << "elf binding changed from " << o->get_binding()
      << " to " << n->get_binding();
      emitted = true;
    }
  if (o->get_visibility() != n->get_visibility())
    {
      if (begin_with_and)
    {
      out << indent << "and ";
      begin_with_and = false;
    }
      else if (!emitted)
    out << indent << "'" << o_pretty_representation << "' ";
      else
    out << ", ";
      out << "visibility changed from " << o->get_visibility()
      << " to " << n->get_visibility();
      emitted = true;
    }
  if ((ctxt->get_allowed_category() & ACCESS_CHANGE_CATEGORY)
      && (get_member_access_specifier(o)
      != get_member_access_specifier(n)))
    {
      if (begin_with_and)
    {
      out << indent << "and ";
      begin_with_and = false;
    }
      else if (!emitted)
    out << indent << "'" << o_pretty_representation << "' ";
      else
    out << ", ";
 
      out << "access changed from '"
      << get_member_access_specifier(o)
      << "' to '"
      << get_member_access_specifier(n) << "'";
      emitted = true;
    }
  if (get_member_is_static(o)
      != get_member_is_static(n))
    {
      if (begin_with_and)
    {
      out << indent << "and ";
      begin_with_and = false;
    }
      else if (!emitted)
    out << indent << "'" << o_pretty_representation << "' ";
      else
    out << ", ";
 
      if (get_member_is_static(o))
    out << "is no more static";
      else
    out << "now becomes static";
      emitted = true;
    }
 
  if (begin_with_and)
    // do nothing as begin_with_and implies emitted
    ;
  else if (!emitted)
    // We appear to have fallen off the edge of the map.
    out << indent << "'" << o_pretty_representation
    << "' has *some* difference - please report as a bug";
  else
    {
      ;// do nothing
    }
  emitted = true;
 
  if (!begin_with_and)
    out << "\n";
}
 
/// Represent the changes carried by an instance of @ref subrange_diff
/// that represent a difference between two ranges.
///
/// @param diff diff the diff node to represent.
///
/// @param ctxt the diff context to use.
///
/// @param local_only if true, only display local changes.
///
/// @param out the output stream to send the representation to.
///
/// @param indent the indentation string to use for the change report.
void
represent(const subrange_diff&      d,
      const diff_context_sptr   ctxt,
      ostream&          out,
      const string&     indent,
      bool              local_only)
{
  array_type_def::subrange_sptr o = d.first_subrange();
  array_type_def::subrange_sptr n = d.second_subrange();
  string oor = o->get_pretty_representation();
  string nr = n->get_pretty_representation();
  string on = o->get_name();
  string nn = n->get_name();
  int64_t olb = o->get_lower_bound();
  int64_t nlb = n->get_lower_bound();
  int64_t oub = o->get_upper_bound();
  int64_t nub = n->get_upper_bound();
 
    if (on != nn)
    {
      out << indent << "name of range changed from '"
      << on << "' to '" << nn << "'\n";
    }
 
  if (olb != nlb)
    {
      out << indent << "lower bound of range '"
      << on
      << "' change from '";
      emit_num_value(olb, *ctxt, out);
      out << "' to '";
      emit_num_value(nlb, *ctxt, out);
      out << "'\n";
    }
 
  if (oub != nub)
    {
      out << indent << "upper bound of range '"
      << on
      << "' change from '";
      emit_num_value(oub, *ctxt, out);
      out << "' to '";
      emit_num_value(nub, *ctxt, out);
      out << "'\n";
    }
 
  if (!local_only)
    {
      diff_sptr dif = d.underlying_type_diff();
      if (dif && dif->to_be_reported())
    {
      // report range underlying type changes
      out << indent << "underlying type of range '"
          << oor << "' changed:\n";
      dif->report(out, indent + "  ");
    }
    }
}
 
/// Report the size and alignment changes of a type.
///
/// @param first the first type to consider.
///
/// @param second the second type to consider.
///
/// @param ctxt the content of the current diff.
///
/// @param out the output stream to report the change to.
///
/// @param indent the string to use for indentation.
void
report_size_and_alignment_changes(type_or_decl_base_sptr    first,
                  type_or_decl_base_sptr    second,
                  diff_context_sptr     ctxt,
                  ostream&          out,
                  const string&         indent)
{
  type_base_sptr f = dynamic_pointer_cast<type_base>(first),
    s = dynamic_pointer_cast<type_base>(second);
 
  if (!s || !f)
    return;
 
  class_or_union_sptr first_class = is_class_or_union_type(first),
    second_class = is_class_or_union_type(second);
 
  if (filtering::has_class_decl_only_def_change(first_class, second_class))
    // So these two classes differ only by the fact that one is the
    // declaration-only form of the second.  The declaration-only class
    // is of unknown size (recorded as 0) and it is not meaningful to
    // report a size change.
    return;
 
  unsigned fs = f->get_size_in_bits(), ss = s->get_size_in_bits(),
    fa = f->get_alignment_in_bits(), sa = s->get_alignment_in_bits();
  array_type_def_sptr first_array = is_array_type(is_type(first)),
    second_array = is_array_type(is_type(second));
  unsigned fdc = first_array ? first_array->get_dimension_count(): 0,
    sdc = second_array ? second_array->get_dimension_count(): 0;
 
  if (ctxt->get_allowed_category() & SIZE_OR_OFFSET_CHANGE_CATEGORY)
    {
      if (fs != ss || fdc != sdc)
    {
      if (first_array && second_array)
        {
          // We are looking at size or alignment changes between two
          // arrays ...
          out << indent << "array type size changed from ";
          if (first_array->is_non_finite())
        out << "\'unknown\'";
          else
        emit_num_value(first_array->get_size_in_bits(), *ctxt, out);
          out << " to ";
          if (second_array->is_non_finite())
        out << "\'unknown\'";
          else
        emit_num_value(second_array->get_size_in_bits(), *ctxt, out);
          out << "\n";
 
          if (sdc != fdc)
        {
          out << indent + "  "
              << "number of dimensions changed from "
              << fdc
              << " to "
              << sdc
              << "\n";
        }
          array_type_def::subranges_type::const_iterator i, j;
          for (i = first_array->get_subranges().begin(),
             j = second_array->get_subranges().begin();
           (i != first_array->get_subranges().end()
            && j != second_array->get_subranges().end());
           ++i, ++j)
        {
          if ((*i)->get_length() != (*j)->get_length())
            {
              out << indent
              << "array type subrange "
              << i - first_array->get_subranges().begin() + 1
              << " changed length from ";
 
              if ((*i)->is_non_finite())
            out << "\'unknown\'";
              else
            out << (*i)->get_length();
 
              out << " to ";
 
              if ((*j)->is_non_finite())
            out << "\'unknown\'";
              else
            out << (*j)->get_length();
              out << "\n";
            }
        }
        } // end if (first_array && second_array)
      else if (fs != ss)
        {
          out << indent;
          show_numerical_change("type size", fs, ss, *ctxt, out);
          out << "\n";
        }
    } // end if (fs != ss || fdc != sdc)
      else
    if (ctxt->show_relative_offset_changes())
      {
        out << indent << "type size hasn't changed\n";
      }
    }
  if ((ctxt->get_allowed_category() & SIZE_OR_OFFSET_CHANGE_CATEGORY)
      && (fa != sa))
    {
      out << indent;
      show_numerical_change("type alignment", fa, sa, *ctxt, out,
                /*show_bits_or_bytes=*/false);
      out << "\n";
    }
}
 
/// @param tod the type or declaration to emit loc info about
///
/// @param ctxt the content of the current diff.
///
/// @param out the output stream to report the change to.
///
/// @return true iff something was reported.
bool
report_loc_info(const type_or_decl_base_sptr& tod,
        const diff_context& ctxt,
        ostream &out)
{
  if (!ctxt.show_locs())
    return false;
 
  decl_base_sptr decl = is_decl(tod);
 
  if (!decl)
    return false;
 
  location loc;
  translation_unit* tu = get_translation_unit(decl);
 
  if (tu && (loc = decl->get_location()))
  {
    string path;
    unsigned line, column;
 
    loc.expand(path, line, column);
    //tu->get_loc_mgr().expand_location(loc, path, line, column);
    path = basename(const_cast<char*>(path.c_str()));
 
    out  << " at " << path << ":" << line << ":" << column;
 
    return true;
  }
  return false;
}
 
/// Report the name, size and alignment changes of a type.
///
/// @param first the first type to consider.
///
/// @param second the second type to consider.
///
/// @param ctxt the content of the current diff.
///
/// @param out the output stream to report the change to.
///
/// @param indent the string to use for indentation.
void
report_name_size_and_alignment_changes(decl_base_sptr       first,
                       decl_base_sptr       second,
                       diff_context_sptr    ctxt,
                       ostream&         out,
                       const string&        indent)
{
  string fn = first->get_qualified_name(),
    sn = second->get_qualified_name();
 
  if (!(first->get_is_anonymous() && second->get_is_anonymous())
      && fn != sn)
    {
      if (!(ctxt->get_allowed_category() & HARMLESS_DECL_NAME_CHANGE_CATEGORY)
      && filtering::has_harmless_name_change(first, second))
    // This is a harmless name change.  but then
    // HARMLESS_DECL_NAME_CHANGE_CATEGORY doesn't seem allowed.
    ;
      else
    {
      out << indent;
      if (is_type(first))
        out << "type";
      else
        out << "declaration";
      out << " name changed from '" << fn << "' to '" << sn << "'";
      out << "\n";
    }
    }
 
  report_size_and_alignment_changes(first, second, ctxt, out, indent);
}
 
/// Output the header preceding the the report for
/// insertion/deletion/change of a part of a class.  This is a
/// subroutine of class_diff::report.
///
/// @param out the output stream to output the report to.
///
/// @param number the number of insertion/deletion to refer to in the
/// header.
///
/// @param num_filtered the number of filtered changes.
///
/// @param k the kind of diff (insertion/deletion/change) we want the
/// head to introduce.
///
/// @param section_name the name of the sub-part of the class to
/// report about.
///
/// @param indent the string to use as indentation prefix in the
/// header.
void
report_mem_header(ostream& out,
          size_t number,
          size_t num_filtered,
          diff_kind k,
          const string& section_name,
          const string& indent)
{
  size_t net_number = number - num_filtered;
  string change;
  char colon_or_semi_colon = ':';
 
  switch (k)
    {
    case del_kind:
      change = (number > 1) ? "deletions" : "deletion";
      break;
    case ins_kind:
      change = (number > 1) ? "insertions" : "insertion";
      break;
    case subtype_change_kind:
    case change_kind:
      change = (number > 1) ? "changes" : "change";
      break;
    }
 
  if (net_number == 0)
    {
      out << indent << "no " << section_name << " " << change;
      colon_or_semi_colon = ';';
    }
  else if (net_number == 1)
    out << indent << "1 " << section_name << " " << change;
  else
    out << indent << net_number << " " << section_name
    << " " << change;
 
  if (num_filtered)
    out << " (" << num_filtered << " filtered)";
  out << colon_or_semi_colon << "\n";
}
 
/// Output the header preceding the the report for
/// insertion/deletion/change of a part of a class.  This is a
/// subroutine of class_diff::report.
///
/// @param out the output stream to output the report to.
///
/// @param k the kind of diff (insertion/deletion/change) we want the
/// head to introduce.
///
/// @param section_name the name of the sub-part of the class to
/// report about.
///
/// @param indent the string to use as indentation prefix in the
/// header.
void
report_mem_header(ostream& out,
          diff_kind k,
          const string& section_name,
          const string& indent)
{
  string change;
 
  switch (k)
    {
    case del_kind:
      change = "deletions";
      break;
    case ins_kind:
      change = "insertions";
      break;
    case subtype_change_kind:
    case change_kind:
      change = "changes";
      break;
    }
 
  out << indent << "there are " << section_name << " " << change << ":\n";
}
 
/// Report the differences in access specifiers and static-ness for
/// class members.
///
/// @param decl1 the first class member to consider.
///
/// @param decl2 the second class member to consider.
///
/// @param out the output stream to send the report to.
///
/// @param indent the indentation string to use for the report.
///
/// @return true if something was reported, false otherwise.
bool
maybe_report_diff_for_member(const decl_base_sptr&  decl1,
                 const decl_base_sptr&  decl2,
                 const diff_context_sptr&   ctxt,
                 ostream&           out,
                 const string&      indent)
 
{
  bool reported = false;
  if (!is_member_decl(decl1) || !is_member_decl(decl2))
    return reported;
 
  string decl1_repr = decl1->get_pretty_representation(),
    decl2_repr = decl2->get_pretty_representation();
 
  if (get_member_is_static(decl1) != get_member_is_static(decl2))
    {
      bool lost = get_member_is_static(decl1);
      out << indent << "'" << decl1_repr << "' ";
      if (report_loc_info(decl2, *ctxt, out))
    out << " ";
      if (lost)
    out << "became non-static";
      else
    out << "became static";
      out << "\n";
      reported = true;
    }
  if ((ctxt->get_allowed_category() & ACCESS_CHANGE_CATEGORY)
      && (get_member_access_specifier(decl1)
      != get_member_access_specifier(decl2)))
    {
      out << indent << "'" << decl1_repr << "' access changed from '"
      << get_member_access_specifier(decl1)
      << "' to '"
      << get_member_access_specifier(decl2)
      << "'\n";
      reported = true;
    }
  return reported;
}
 
/// Report the differences between two generic variables.
///
/// @param decl1 the first version of the variable.
///
/// @param decl2 the second version of the variable.
///
/// @param ctxt the context of the diff.
///
/// @param out the output stream to emit the change report to.
///
/// @param indent the indentation prefix to emit.
///
/// @return true if any text has been emitted to the output stream.
bool
maybe_report_diff_for_variable(const decl_base_sptr&    decl1,
                   const decl_base_sptr&    decl2,
                   const diff_context_sptr& ctxt,
                   ostream&     out,
                   const string&        indent)
{
  bool reported = false;
 
  var_decl_sptr var1 = is_var_decl(decl1);
  var_decl_sptr var2 = is_var_decl(decl2);
 
  if (!var1 || !var2)
    return reported;
 
  if (filtering::is_var_1_dim_unknown_size_array_change(var1, var2))
    {
      uint64_t var_size_in_bits = var1->get_symbol()->get_size() * 8;
 
      out << indent;
      show_offset_or_size("size of variable symbol (",
              var_size_in_bits, *ctxt, out);
      out << ") hasn't changed\n"
      << indent << "but it does have a harmless type change\n";
      reported = true;
    }
 
  return reported;
}
 
/// Report the difference between two ELF symbols, if there is any.
///
/// @param symbol1 the first symbol to consider.
///
/// @param symbol2 the second symbol to consider.
///
/// @param ctxt the diff context.
///
/// @param the output stream to emit the report to.
///
/// @param indent the indentation string to use.
void
maybe_report_diff_for_symbol(const elf_symbol_sptr& symbol1,
                 const elf_symbol_sptr& symbol2,
                 const diff_context_sptr&   ctxt,
                 ostream&           out,
                 const string&      indent)
{
  if (!symbol1 || !symbol2 || symbol1 == symbol2)
    return;
 
  if (symbol1->get_size() != symbol2->get_size())
    {
      out << indent;
      show_numerical_change("size of symbol",
                symbol1->get_size(),
                symbol2->get_size(),
                *ctxt, out,
                /*show_bits_or_bytes=*/false);
      out << "\n";
    }
 
  if (symbol1->get_name() != symbol2->get_name())
    {
      out << indent << "symbol name changed from "
      << symbol1->get_name()
      << " to "
      << symbol2->get_name()
      << "\n";
    }
 
  if (symbol1->get_type() != symbol2->get_type())
    {
      out << indent << "symbol type changed from '"
      << symbol1->get_type()
      << "' to '"
      << symbol2->get_type()
      << "'\n";
    }
 
  if (symbol1->is_public() != symbol2->is_public())
    {
      out << indent << "symbol became ";
      if (symbol2->is_public())
    out << "exported";
      else
    out << "non-exported";
      out << "\n";
    }
 
  if (symbol1->is_defined() != symbol2->is_defined())
    {
      out << indent << "symbol became ";
      if (symbol2->is_defined())
    out << "defined";
      else
    out << "undefined";
      out << "\n";
    }
 
  if (symbol1->get_version() != symbol2->get_version())
    {
      out << indent << "symbol version changed from "
      << symbol1->get_version().str()
      << " to "
      << symbol2->get_version().str()
      << "\n";
    }
 
  const abg_compat::optional<uint32_t>& crc1 = symbol1->get_crc();
  const abg_compat::optional<uint32_t>& crc2 = symbol2->get_crc();
  if (crc1 != crc2)
    {
      const std::string none = "(none)";
      out << indent << "CRC (modversions) changed from "
      << std::showbase << std::hex;
      if (crc1.has_value())
    out << crc1.value();
      else
    out << none;
      out << " to ";
      if (crc2.has_value())
    out << crc2.value();
      else
    out << none;
      out << std::noshowbase << std::dec
      << "\n";
    }
 
  const abg_compat::optional<std::string>& ns1 = symbol1->get_namespace();
  const abg_compat::optional<std::string>& ns2 = symbol2->get_namespace();
  if (ns1 != ns2)
    {
      const std::string none = "(none)";
      out << indent << "namespace changed from ";
      if (ns1.has_value())
    out << "'" << ns1.value() << "'";
      else
    out << none;
      out << " to ";
      if (ns2.has_value())
    out << "'" << ns2.value() << "'";
      else
    out << none;
      out << "\n";
    }
}
 
/// For a given symbol, emit a string made of its name and version.
/// The string also contains the list of symbols that alias this one.
///
/// @param out the output string to emit the resulting string to.
///
/// @param indent the indentation string to use before emitting the
/// resulting string.
///
/// @param symbol the symbol to emit the representation string for.
///
/// @param sym_map the symbol map to consider to look for aliases of
/// @p symbol.
void
show_linkage_name_and_aliases(ostream& out,
                  const string& indent,
                  const elf_symbol& symbol,
                  const string_elf_symbols_map_type& sym_map)
{
  out << indent << symbol.get_id_string();
  string aliases =
    symbol.get_aliases_id_string(sym_map,
                 /*include_symbol_itself=*/false);
  if (!aliases.empty())
    out << ", aliases " << aliases;
}
 
/// Report changes about types that are not reachable from global
/// functions and variables, in a given @param corpus_diff.
///
/// @param d the corpus_diff to consider.
///
/// @param s the statistics of the changes, after filters and
/// suppressions are reported.  This is typically what is returned by
/// corpus_diff::apply_filters_and_suppressions_before_reporting().
///
/// @param indent the indendation string (usually a string of white
/// spaces) to use for indentation during the reporting.
///
/// @param out the output stream to emit the report to.
void
maybe_report_unreachable_type_changes(const corpus_diff& d,
                      const corpus_diff::diff_stats &s,
                      const string& indent,
                      ostream& out)
{
  const diff_context_sptr& ctxt = d.context();
 
  if (!(ctxt->show_unreachable_types()
    && (!d.priv_->deleted_unreachable_types_.empty()
        || !d.priv_->added_unreachable_types_.empty()
        || !d.priv_->changed_unreachable_types_.empty())))
    // The user either doesn't want us to show changes about
    // unreachable types or there are not such changes.
    return;
 
  // Handle removed unreachable types.
  if (s.net_num_removed_unreachable_types() == 1)
    out << indent
    << "1 removed type unreachable from any public interface:\n\n";
  else if (s.net_num_removed_unreachable_types() > 1)
    out << indent
    << s.net_num_removed_unreachable_types()
    << " removed types unreachable from any public interface:\n\n";
 
  vector<type_base_sptr> sorted_removed_unreachable_types;
  sort_string_type_base_sptr_map(d.priv_->deleted_unreachable_types_,
                 sorted_removed_unreachable_types);
  bool emitted = false;
  for (vector<type_base_sptr>::const_iterator i =
     sorted_removed_unreachable_types.begin();
       i != sorted_removed_unreachable_types.end();
       ++i)
    {
      if (d.priv_->deleted_unreachable_type_is_suppressed((*i).get()))
    continue;
 
      out << indent << "  [D] '" << get_pretty_representation(*i) << "'";
      report_loc_info(*i, *ctxt, out);
      out << "\n";
      emitted = true;
    }
  if (emitted)
    out << "\n";
 
  // Handle changed unreachable types!
  if (s.net_num_changed_unreachable_types() == 1)
    out << indent
    << "1 changed type unreachable from any public interface:\n\n";
  else if (s.net_num_changed_unreachable_types() > 1)
    out << indent
    << s.net_num_changed_unreachable_types()
    << " changed types unreachable from any public interface:\n\n";
 
  diff_sptrs_type sorted_diff_sptrs;
  sort_string_diff_sptr_map(d.priv_->changed_unreachable_types_,
                sorted_diff_sptrs);
  for (diff_sptrs_type::const_iterator i = sorted_diff_sptrs.begin();
       i != sorted_diff_sptrs.end();
       ++i)
    {
      diff_sptr diff = *i;
      if (!diff || !diff->to_be_reported())
    continue;
 
      string repr = diff->first_subject()->get_pretty_representation();
 
      out << indent << "  [C] '" << repr << "' changed:\n";
      diff->report(out, indent + "    ");
      // Extra spacing.
      out << "\n";
    }
  // Changed types have extra spacing already. No new line here.
 
  // Handle added unreachable types.
  if (s.net_num_added_unreachable_types() == 1)
    out << indent
    << "1 added type unreachable from any public interface:\n\n";
  else if (s.net_num_added_unreachable_types() > 1)
    out << indent
    << s.net_num_added_unreachable_types()
    << " added types unreachable from any public interface:\n\n";
 
  vector<type_base_sptr> sorted_added_unreachable_types;
  sort_string_type_base_sptr_map(d.priv_->added_unreachable_types_,
                 sorted_added_unreachable_types);
  emitted = false;
  for (vector<type_base_sptr>::const_iterator i =
     sorted_added_unreachable_types.begin();
       i != sorted_added_unreachable_types.end();
       ++i)
    {
      if (d.priv_->added_unreachable_type_is_suppressed((*i).get()))
    continue;
 
      out << indent << "  [A] '" << get_pretty_representation(*i) << "'";
      report_loc_info(*i, *ctxt, out);
      out << "\n";
      emitted = true;
    }
  if (emitted)
    out << "\n";
}
 
/// If a given diff node impacts some public interfaces, then report
/// about those impacted interfaces on a given output stream.
///
/// @param d the diff node to get the impacted interfaces for.
///
/// @param out the output stream to report to.
///
/// @param indent the white space string to use for indentation.
void
maybe_report_interfaces_impacted_by_diff(const diff *d,
                     ostream    &out,
                     const string   &indent)
{
  const diff_context_sptr &ctxt = d->context();
  const corpus_diff_sptr &corp_diff = ctxt->get_corpus_diff();
  if (!corp_diff)
    return;
 
  if (!ctxt->show_impacted_interfaces())
    return;
 
  const diff_maps &maps = corp_diff->get_leaf_diffs();
  artifact_sptr_set_type* impacted_artifacts =
    maps.lookup_impacted_interfaces(d);
  if (impacted_artifacts == 0)
    return;
 
  if (impacted_artifacts->empty())
    return;
 
  vector<type_or_decl_base_sptr> sorted_impacted_interfaces;
  sort_artifacts_set(*impacted_artifacts, sorted_impacted_interfaces);
 
  size_t num_impacted_interfaces = impacted_artifacts->size();
  if (num_impacted_interfaces == 1)
    out << indent << "one impacted interface:\n";
  else
    out << indent << num_impacted_interfaces << " impacted interfaces:\n";
 
  string cur_indent = indent + "  ";
  vector<type_or_decl_base_sptr>::const_iterator it;
  for (it = sorted_impacted_interfaces.begin();
       it != sorted_impacted_interfaces.end();
       ++it)
    {
      out << cur_indent << get_pretty_representation(*it) << "\n";
    }
}
 
/// If a given diff node impacts some public interfaces, then report
/// about those impacted interfaces on standard output.
///
/// @param d the diff node to get the impacted interfaces for.
///
/// @param out the output stream to report to.
///
/// @param indent the white space string to use for indentation.
void
maybe_report_interfaces_impacted_by_diff(const diff_sptr    &d,
                     ostream        &out,
                     const string       &indent)
{return maybe_report_interfaces_impacted_by_diff(d.get(), out, indent);}
 
/// Tests if the diff node is to be reported.
///
/// @param p the diff to consider.
///
/// @return true iff the diff is to be reported.
bool
reporter_base::diff_to_be_reported(const diff *d) const
{return d && d->to_be_reported();}
 
/// Report about data members replaced by an anonymous data member
/// without changing the overall bit-layout of the class or union in
/// an ABI-meaningful way.
///
/// @param d the diff to consider.
///
/// @param out the output stream to emit the change report to.
///
/// @param indent the indentation string to use.
void
maybe_report_data_members_replaced_by_anon_dm(const class_or_union_diff &d,
                          ostream           &out,
                          const string      &indent)
{
  const diff_context_sptr& ctxt = d.context();
 
  if ((ctxt->get_allowed_category() & HARMLESS_DATA_MEMBER_CHANGE_CATEGORY)
      && !d.data_members_replaced_by_adms().empty())
    {
      // Let's detect all the data members that are replaced by
      // members of the same anonymous data member and report them
      // in one go.
      for (changed_var_sptrs_type::const_iterator i =
         d.ordered_data_members_replaced_by_adms().begin();
       i != d.ordered_data_members_replaced_by_adms().end();)
    {
      // This contains the data members replaced by the same
      // anonymous data member.
      vector<var_decl_sptr> dms_replaced_by_same_anon_dm;
      dms_replaced_by_same_anon_dm.push_back(i->first);
      // This contains the anonymous data member that replaced the
      // data members in the variable above.
      var_decl_sptr anonymous_data_member = i->second;
      // Let's look forward to see if the subsequent data
      // members were replaced by members of
      // anonymous_data_member.
      for (++i;
           i != d.ordered_data_members_replaced_by_adms().end()
         && *i->second == *anonymous_data_member;
           ++i)
        dms_replaced_by_same_anon_dm.push_back(i->first);
 
      bool several_data_members_replaced =
        dms_replaced_by_same_anon_dm.size() > 1;
 
      out << indent << "data member";
      if (several_data_members_replaced)
        out << "s";
 
      bool first_data_member = true;
      for (vector<var_decl_sptr>::const_iterator it =
         dms_replaced_by_same_anon_dm.begin();
           it != dms_replaced_by_same_anon_dm.end();
           ++it)
        {
          string name = (*it)->get_qualified_name();
          if (!first_data_member)
        out << ",";
          out << " '" << name << "'";
          first_data_member = false;
        }
 
      if (several_data_members_replaced)
        out << " were ";
      else
        out << " was ";
 
      out << "replaced by anonymous data member:\n"
          << indent + "  "
          << "'"
          << anonymous_data_member->get_pretty_representation()
          << "'\n";
    }
    }
}
 
/// Report about the base classes of a class having been re-ordered.
///
/// @param d the class diff to consider.
///
/// @param out the output stream to report the change to.
///
/// @param indent the indentation string to use.
void
maybe_report_base_class_reordering(const class_diff &d,
                   ostream      &out,
                   const string &indent)
{
  if (d.moved_bases().empty())
    return;
 
  class_decl_sptr first = d.first_class_decl(),
    second = d.second_class_decl();
 
  ABG_ASSERT(!first->get_base_specifiers().empty());
  ABG_ASSERT(!second->get_base_specifiers().empty());
 
  out << indent << "base classes of '"
      << first->get_pretty_representation()
      << "' are re-ordered from: ";
 
  vector<class_decl_sptr> classes = {first, second};
  unsigned nb_classes_seen = 0;
  for (auto &klass : classes)
    {
      if (nb_classes_seen >= 1)
    out << " to: ";
      out << "'";
      bool needs_comma = false;
      for (auto &b : klass->get_base_specifiers())
    {
      if (needs_comma)
        out << ", ";
      if (b->get_is_virtual())
        out << "virtual ";
      out << b->get_base_class()->get_qualified_name();
      needs_comma = true;
    }
      out << "'";
      nb_classes_seen++;
    }
  if (nb_classes_seen)
    out << "\n";
}
} // Namespace comparison
} // end namespace abigail