private bool CheckXRefStreamOffset(IRandomAccessRead source, long startXRefOffset, bool isLenient, CosObjectPool pool)
{
// repair mode isn't available in non-lenient mode
if (!isLenient || startXRefOffset == 0)
{
return(true);
}
// seek to offset-1
source.Seek(startXRefOffset - 1);
int nextValue = source.Read();
// the first character has to be a whitespace, and then a digit
if (ReadHelper.IsWhitespace(nextValue))
{
ReadHelper.SkipSpaces(source);
if (ReadHelper.IsDigit(source))
{
try
{
// it's a XRef stream
ObjectHelper.ReadObjectNumber(source);
ObjectHelper.ReadGenerationNumber(source);
ReadHelper.ReadExpectedString(source, "obj", true);
// check the dictionary to avoid false positives
PdfDictionary dict = dictionaryParser.Parse(source, baseParser, pool);
source.Seek(startXRefOffset);
if (dict.IsType(CosName.XREF))
{
return(true);
}
}
catch (Exception ex)
{
log.Error("Couldn't read the xref stream object.", ex);
// there wasn't an object of a xref stream
source.Seek(startXRefOffset);
}
}
}
return(false);
}
private void BfSearchForXRefStreams(IInputBytes bytes)
{
if (bfSearchXRefStreamsOffsets != null)
{
return;
}
// a pdf may contain more than one /XRef entry
bfSearchXRefStreamsOffsets = new List <long>();
var startOffset = bytes.CurrentOffset;
bytes.Seek(MinimumSearchOffset);
// search for XRef streams
var objString = " obj";
while (bytes.MoveNext() && !bytes.IsAtEnd())
{
if (!ReadHelper.IsString(bytes, "xref"))
{
continue;
}
// search backwards for the beginning of the stream
long newOffset = -1;
long xrefOffset = bytes.CurrentOffset;
bool objFound = false;
for (var i = 1; i < 40; i++)
{
if (objFound)
{
break;
}
long currentOffset = xrefOffset - (i * 10);
if (currentOffset > 0)
{
bytes.Seek(currentOffset);
for (int j = 0; j < 10; j++)
{
if (ReadHelper.IsString(bytes, objString))
{
long tempOffset = currentOffset - 1;
bytes.Seek(tempOffset);
var generationNumber = bytes.Peek();
// is the next char a digit?
if (generationNumber.HasValue && ReadHelper.IsDigit(generationNumber.Value))
{
tempOffset--;
bytes.Seek(tempOffset);
// is the digit preceded by a space?
if (ReadHelper.IsSpace(bytes.CurrentByte))
{
int length = 0;
bytes.Seek(--tempOffset);
while (tempOffset > MinimumSearchOffset && ReadHelper.IsDigit(bytes.CurrentByte))
{
bytes.Seek(--tempOffset);
length++;
}
if (length > 0)
{
bytes.MoveNext();
newOffset = bytes.CurrentOffset;
}
}
}
objFound = true;
break;
}
currentOffset++;
bytes.MoveNext();
}
}
}
if (newOffset > -1)
{
bfSearchXRefStreamsOffsets.Add(newOffset);
}
bytes.Seek(xrefOffset + 5);
}
bytes.Seek(startOffset);
}
public bool TryParse(IRandomAccessRead source, long offset, bool isLenientParsing, CosObjectPool pool, out CrossReferenceTablePartBuilder builder)
{
builder = null;
var tableStartOffset = source.GetPosition();
if (source.Peek() != 'x')
{
return(false);
}
var xref = ReadHelper.ReadString(source);
if (!xref.Trim().Equals("xref"))
{
return(false);
}
// check for trailer after xref
var str = ReadHelper.ReadString(source);
byte[] b = OtherEncodings.StringAsLatin1Bytes(str);
source.Rewind(b.Length);
if (str.StartsWith("trailer"))
{
log.Warn("skipping empty xref table");
return(false);
}
builder = new CrossReferenceTablePartBuilder
{
Offset = offset,
XRefType = CrossReferenceType.Table
};
// Tables can have multiple sections. Each starts with a starting object id and a count.
while (true)
{
if (!TableSubsectionDefinition.TryRead(log, source, out var subsectionDefinition))
{
log.Warn($"Unexpected subsection definition in the cross-reference table at offset {offset}");
if (isLenientParsing)
{
break;
}
return(false);
}
var currentObjectId = subsectionDefinition.FirstNumber;
ReadHelper.SkipSpaces(source);
for (var i = 0; i < subsectionDefinition.Count; i++)
{
if (source.IsEof() || ReadHelper.IsEndOfName((char)source.Peek()))
{
break;
}
if (source.Peek() == 't')
{
break;
}
//Ignore table contents
var currentLine = ReadHelper.ReadLine(source);
var splitString = currentLine.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
if (splitString.Length < 3)
{
log.Warn("invalid xref line: " + currentLine);
break;
}
// This supports the corrupt table as reported in PDFBOX-474 (XXXX XXX XX n)
if (splitString[splitString.Length - 1].Equals(InUseEntry))
{
try
{
var objectOffset = long.Parse(splitString[0]);
if (objectOffset >= tableStartOffset && objectOffset <= source.GetPosition())
{
// PDFBOX-3923: offset points inside this table - that can't be good
throw new InvalidOperationException(
$"Object offset {objectOffset} is within its own cross-reference table for object {currentObjectId}");
}
var generation = int.Parse(splitString[1]);
builder.Add(currentObjectId, generation, objectOffset);
}
catch (FormatException e)
{
throw new InvalidOperationException("Bad", e);
}
}
else if (!splitString[2].Equals(FreeEntry))
{
throw new InvalidOperationException(
$"Corrupt cross-reference table entry for object {currentObjectId}. The indicator was not 'n' or 'f' but {splitString[2]}.");
}
currentObjectId++;
ReadHelper.SkipSpaces(source);
}
ReadHelper.SkipSpaces(source);
if (!ReadHelper.IsDigit(source))
{
break;
}
}
if (!TryParseTrailer(source, isLenientParsing, pool, out var trailer))
{
throw new InvalidOperationException($"Something went wrong trying to read the XREF table at {offset}.");
}
builder.Dictionary = trailer;
builder.Previous = trailer.GetLongOrDefault(CosName.PREV);
return(true);
}
private bool TryParseTrailer(IRandomAccessRead source, bool isLenientParsing, CosObjectPool pool, out PdfDictionary trailer)
{
trailer = null;
// parse the last trailer.
var trailerOffset = source.GetPosition();
// PDFBOX-1739 skip extra xref entries in RegisSTAR documents
if (isLenientParsing)
{
int nextCharacter = source.Peek();
while (nextCharacter != 't' && ReadHelper.IsDigit(nextCharacter))
{
if (source.GetPosition() == trailerOffset)
{
// warn only the first time
//LOG.warn("Expected trailer object at position " + trailerOffset
// + ", keep trying");
}
ReadHelper.ReadLine(source);
nextCharacter = source.Peek();
}
}
if (source.Peek() != 't')
{
return(false);
}
//read "trailer"
long currentOffset = source.GetPosition();
string nextLine = ReadHelper.ReadLine(source);
if (!nextLine.Trim().Equals("trailer"))
{
// in some cases the EOL is missing and the trailer immediately
// continues with "<<" or with a blank character
// even if this does not comply with PDF reference we want to support as many PDFs as possible
// Acrobat reader can also deal with this.
if (nextLine.StartsWith("trailer"))
{
// we can't just unread a portion of the read data as we don't know if the EOL consist of 1 or 2 bytes
int len = "trailer".Length;
// jump back right after "trailer"
source.Seek(currentOffset + len);
}
else
{
return(false);
}
}
// in some cases the EOL is missing and the trailer continues with " <<"
// even if this does not comply with PDF reference we want to support as many PDFs as possible
// Acrobat reader can also deal with this.
ReadHelper.SkipSpaces(source);
PdfDictionary parsedTrailer = dictionaryParser.Parse(source, baseParser, pool);
trailer = parsedTrailer;
ReadHelper.SkipSpaces(source);
return(true);
}
private void BfSearchForXRefStreams()
{
if (bfSearchXRefStreamsOffsets == null)
{
// a pdf may contain more than one /XRef entry
bfSearchXRefStreamsOffsets = new List <long>();
long originOffset = source.GetPosition();
source.Seek(MinimumSearchOffset);
// search for XRef streams
var objString = " obj";
while (!source.IsEof())
{
if (ReadHelper.IsString(source, "xref"))
{
// search backwards for the beginning of the stream
long newOffset = -1;
long xrefOffset = source.GetPosition();
bool objFound = false;
for (int i = 1; i < 40 && !objFound; i++)
{
long currentOffset = xrefOffset - (i * 10);
if (currentOffset > 0)
{
source.Seek(currentOffset);
for (int j = 0; j < 10; j++)
{
if (ReadHelper.IsString(source, objString))
{
long tempOffset = currentOffset - 1;
source.Seek(tempOffset);
int genId = source.Peek();
// is the next char a digit?
if (ReadHelper.IsDigit(genId))
{
tempOffset--;
source.Seek(tempOffset);
if (ReadHelper.IsSpace(source))
{
int length = 0;
source.Seek(--tempOffset);
while (tempOffset > MinimumSearchOffset && ReadHelper.IsDigit(source))
{
source.Seek(--tempOffset);
length++;
}
if (length > 0)
{
source.Read();
newOffset = source.GetPosition();
}
}
}
objFound = true;
break;
}
else
{
currentOffset++;
source.Read();
}
}
}
}
if (newOffset > -1)
{
bfSearchXRefStreamsOffsets.Add(newOffset);
}
source.Seek(xrefOffset + 5);
}
source.Read();
}
source.Seek(originOffset);
}
}
private void bfSearchForObjects(IRandomAccessRead source)
{
bfSearchForLastEOFMarker(source);
bfSearchCOSObjectKeyOffsets = new Dictionary <CosObjectKey, long>();
long originOffset = source.GetPosition();
long currentOffset = MINIMUM_SEARCH_OFFSET;
long lastObjectId = long.MinValue;
int lastGenID = int.MinValue;
long lastObjOffset = long.MinValue;
char[] objString = " obj".ToCharArray();
char[] endobjString = "endobj".ToCharArray();
bool endobjFound = false;
do
{
source.Seek(currentOffset);
if (ReadHelper.IsString(source, "obj"))
{
long tempOffset = currentOffset - 1;
source.Seek(tempOffset);
int genID = source.Peek();
// is the next char a digit?
if (ReadHelper.IsDigit(genID))
{
genID -= 48;
tempOffset--;
source.Seek(tempOffset);
if (ReadHelper.IsSpace(source))
{
while (tempOffset > MINIMUM_SEARCH_OFFSET && ReadHelper.IsSpace(source))
{
source.Seek(--tempOffset);
}
bool objectIDFound = false;
while (tempOffset > MINIMUM_SEARCH_OFFSET && ReadHelper.IsDigit(source))
{
source.Seek(--tempOffset);
objectIDFound = true;
}
if (objectIDFound)
{
source.Read();
long objectId = ObjectHelper.ReadObjectNumber(source);
if (lastObjOffset > 0)
{
// add the former object ID only if there was a subsequent object ID
bfSearchCOSObjectKeyOffsets[new CosObjectKey(lastObjectId, lastGenID)] = lastObjOffset;
}
lastObjectId = objectId;
lastGenID = genID;
lastObjOffset = tempOffset + 1;
currentOffset += objString.Length - 1;
endobjFound = false;
}
}
}
}
else if (ReadHelper.IsString(source, "endobj"))
{
endobjFound = true;
currentOffset += endobjString.Length - 1;
}
currentOffset++;
} while (currentOffset < lastEOFMarker && !source.IsEof());
if ((lastEOFMarker < long.MaxValue || endobjFound) && lastObjOffset > 0)
{
// if the pdf wasn't cut off in the middle or if the last object ends with a "endobj" marker
// the last object id has to be added here so that it can't get lost as there isn't any subsequent object id
bfSearchCOSObjectKeyOffsets[new CosObjectKey(lastObjectId, lastGenID)] = lastObjOffset;
}
// reestablish origin position
source.Seek(originOffset);
}