/// <summary>
/// Analyzes uncategorized regions of the file to see if they fit common patterns.
///
/// This is re-run after most changes to the project, so we don't want to do anything
/// crazily expensive.
/// </summary>
/// <returns>True on success.</returns>
public void AnalyzeUncategorized()
{
// TODO(someday): we can make this faster. The data doesn't change, so we
// only need to do a full scan once, when the file is first loaded. We can
// create a TypedRangeSet for runs of identical bytes, using the byte value
// as the type. A second TypedRangeSet would identify runs of ASCII chars,
// with different types for high/low ASCII (and PETSCII?). AnalyzeRange() would
// then just need to find the intersection with the sets, which should be
// significantly faster. We would need to re-do the scan if the parameters
// for things like min match length change.
FormatDescriptor oneByteDefault = FormatDescriptor.Create(1,
FormatDescriptor.Type.Default, FormatDescriptor.SubType.None);
FormatDescriptor.DebugPrefabBump(-1);
// If it hasn't been identified as code or data, set the "data" flag to
// give it a positive identification as data. (This should be the only
// place outside of CodeAnalysis that sets this flag.) This isn't strictly
// necessary, but it helps us assert things when pieces start moving around.
for (int offset = 0; offset < mAnattribs.Length; offset++)
{
Anattrib attr = mAnattribs[offset];
if (attr.IsInlineData)
{
// While we're here, add a default format descriptor for inline data
// that doesn't have one. We don't try to analyze it otherwise.
if (attr.DataDescriptor == null)
{
mAnattribs[offset].DataDescriptor = oneByteDefault;
FormatDescriptor.DebugPrefabBump();
}
}
else if (!attr.IsInstruction)
{
mAnattribs[offset].IsData = true;
}
}
mDebugLog.LogI("Analyzing uncategorized data...");
int startOffset = -1;
for (int offset = 0; offset < mAnattribs.Length;)
{
// We want to find a contiguous series of offsets which are not known
// to hold code or data. We stop if we encounter a user-defined label
// or format descriptor.
Anattrib attr = mAnattribs[offset];
if (attr.IsInstruction || attr.IsInlineData || attr.IsDataStart)
{
// Instruction, inline data, or formatted data known to be here. Analyze
// previous chunk, then advance past this.
if (startOffset >= 0)
{
AnalyzeRange(startOffset, offset - 1);
startOffset = -1;
}
if (attr.IsInstruction)
{
// Because of embedded instructions, we can't simply leap forward.
offset++;
}
else
{
Debug.Assert(attr.Length > 0);
offset += attr.Length;
}
}
else if (attr.Symbol != null || mProject.HasCommentOrNote(offset))
{
// In an uncategorized area, but we want to break at this byte
// so the user or auto label doesn't get buried in the middle of
// a large chunk.
//
// This is similar to, but independent of, GroupedOffsetSetFromSelected()
// in ProjectView. This is for auto-detection, the other is for user
// selection. It's best if the two behave similarly though.
if (startOffset >= 0)
{
AnalyzeRange(startOffset, offset - 1);
}
startOffset = offset;
offset++;
}
else
{
// This offset is uncategorized, keep gathering.
if (startOffset < 0)
{
startOffset = offset;
}
offset++;
// Check to see if the address has changed from the previous entry.
if (offset < mAnattribs.Length &&
mAnattribs[offset - 1].Address + 1 != mAnattribs[offset].Address)
{
// Must be an ORG here. Scan previous region.
AnalyzeRange(startOffset, offset - 1);
startOffset = -1;
}
}
}
if (startOffset >= 0)
{
AnalyzeRange(startOffset, mAnattribs.Length - 1);
}
}
/// <summary>
/// Formats the address map for viewing.
/// </summary>
public static string GenerateString(DisasmProject project, Formatter formatter)
{
AddressMap addrMap = project.AddrMap;
bool showBank = !project.CpuDef.HasAddr16;
StringBuilder sb = new StringBuilder();
int depth = 0;
int prevOffset = -1;
int prevAddr = 0;
int lastEndOffset = -1;
sb.AppendLine("Address region map for \"" + project.DataFileName + "\"");
sb.Append(CRLF);
IEnumerator <AddressChange> iter = addrMap.AddressChangeIterator;
while (iter.MoveNext())
{
AddressChange change = iter.Current;
if (change.IsStart)
{
//if (change.Offset == lastEndOffset) {
// // Extra vertical space for a START following an END at the same offset.
// PrintDepthLines(sb, depth, true);
// sb.Append(CRLF);
// lastEndOffset = -1;
//}
if (prevOffset >= 0 && change.Offset != prevOffset)
{
// Start of region at new offset. Output address info for space
// between previous start or end.
PrintAddressInfo(sb, formatter, depth, prevAddr,
change.Offset - prevOffset, showBank);
}
// Start following end, or start following start after a gap.
sb.Append(formatter.FormatOffset24(change.Offset));
PrintDepthLines(sb, depth, false);
sb.Append("+- " + "start");
if (change.IsSynthetic)
{
sb.Append(" (auto-generated)");
}
else
{
// If there's a label here, show it.
Anattrib attr = project.GetAnattrib(change.Offset);
if (attr.Symbol != null && !string.IsNullOrEmpty(attr.Symbol.Label))
{
sb.Append(" '");
sb.Append(attr.Symbol.GenerateDisplayLabel(formatter));
sb.Append("'");
}
}
if (change.Region.HasValidPreLabel)
{
sb.Append(" pre='");
sb.Append(change.Region.PreLabel);
sb.Append("'");
}
sb.Append(CRLF);
prevOffset = change.Offset;
prevAddr = change.Address;
depth++;
}
else
{
Debug.Assert(prevOffset >= 0);
depth--;
if (change.Offset + 1 != prevOffset)
{
// End of region at new offset. Output address info for space
// between previous start or end.
PrintAddressInfo(sb, formatter, depth + 1, prevAddr,
change.Offset + 1 - prevOffset, showBank);
}
sb.Append(formatter.FormatOffset24(change.Offset));
PrintDepthLines(sb, depth, false);
sb.Append("+- " + "end");
if (change.Region.IsFloating)
{
sb.Append(" (floating)");
}
//PrintAddress(sb, formatter, change.Address, showBank);
//sb.Append(")");
sb.Append(CRLF);
// Add a blank line, but with the depth lines.
if (depth > 0)
{
PrintDepthLines(sb, depth, true);
}
sb.Append(CRLF);
// Use offset+1 here so it lines up with start records.
prevOffset = lastEndOffset = change.Offset + 1;
prevAddr = change.Address;
}
}
Debug.Assert(depth == 0);
return(sb.ToString());
}
/// <summary>
/// Analyzes instruction operands and Address data descriptors to identify references
/// to offsets within the file.
///
/// Instructions with format descriptors are left alone. Instructions with
/// operand offsets but no descriptor will have a descriptor generated
/// using the label at the target offset; if the target offset is unlabeled,
/// a unique label will be generated. Data descriptors with type=Address are
/// handled the same way.
///
/// In some cases, such as a reference to the middle of an instruction, we will
/// label a nearby location instead.
///
/// This should be called after code analysis has run, user labels and format
/// descriptors have been applied, and platform/project symbols have been merged
/// into the symbol table.
/// </summary>
/// <returns>True on success.</returns>
public void AnalyzeDataTargets()
{
mDebugLog.LogI("Analyzing data targets...");
for (int offset = 0; offset < mAnattribs.Length; offset++)
{
Anattrib attr = mAnattribs[offset];
if (attr.IsInstructionStart)
{
if (attr.DataDescriptor != null)
{
// It's being shown as numeric, or as a reference to some other symbol.
// Either way there's nothing further for us to do. (Technically we
// would want to treat it like the no-descriptor case if the type was
// numeric/Address, but we don't allow that for instructions.)
Debug.Assert(attr.DataDescriptor.FormatSubType !=
FormatDescriptor.SubType.Address);
continue;
}
int operandOffset = attr.OperandOffset;
if (operandOffset >= 0)
{
// This is an offset reference: a branch or data access instruction whose
// target is inside the file. Create a FormatDescriptor for it, and
// generate a label at the target if one is not already present.
SetDataTarget(offset, attr.Length, operandOffset);
}
// We advance by a single byte, rather than .Length, in case there's
// an instruction embedded inside another one.
}
else if (attr.DataDescriptor != null)
{
// We can't check IsDataStart / IsInlineDataStart because the bytes might
// still be uncategorized. If there's a user-specified format, check it
// to see if it's an address.
FormatDescriptor dfd = attr.DataDescriptor;
// Is this numeric/Address?
if ((dfd.FormatType == FormatDescriptor.Type.NumericLE ||
dfd.FormatType == FormatDescriptor.Type.NumericBE) &&
dfd.FormatSubType == FormatDescriptor.SubType.Address)
{
// Treat like an absolute address. Convert the operand
// to an address, then resolve the file offset.
int address = RawData.GetWord(mFileData, offset, dfd.Length,
(dfd.FormatType == FormatDescriptor.Type.NumericBE));
if (dfd.Length < 3)
{
// Bank not specified by data, add current program bank. Not always
// correct, but should be often enough. In most cases we'd just
// assume a correct data bank register, but here we need to find
// a file offset, so we have to assume data bank == program bank
// (unless we find a good way to track the data bank register).
address |= attr.Address & 0x7fff0000;
}
int operandOffset = mProject.AddrMap.AddressToOffset(offset, address);
if (operandOffset >= 0)
{
SetDataTarget(offset, dfd.Length, operandOffset);
}
}
// For other formats, we don't need to do anything. Numeric/Address is
// the only one that represents an offset reference. Numeric/Symbol
// is a name reference. The others are just data.
// There shouldn't be any data items inside other data items, so we
// can just skip forward.
offset += mAnattribs[offset].DataDescriptor.Length - 1;
}
}
}
