private CosBase ReadNormalObjectStream(IRandomAccessRead reader, CosBase currentBase, long offset,
bool isLenientParsing,
out string endObjectKey)
{
if (currentBase is PdfDictionary dictionary)
{
PdfRawStream stream = streamParser.Parse(reader, dictionary, isLenientParsing, null);
currentBase = stream;
}
else
{
// this is not legal
// the combination of a dict and the stream/endstream
// forms a complete stream object
throw new InvalidOperationException($"Stream not preceded by dictionary (offset: {offset}).");
}
ReadHelper.SkipSpaces(reader);
endObjectKey = ReadHelper.ReadLine(reader);
// we have case with a second 'endstream' before endobj
if (!endObjectKey.StartsWith("endobj") && endObjectKey.StartsWith("endstream"))
{
endObjectKey = endObjectKey.Substring(9).Trim();
if (endObjectKey.Length == 0)
{
// no other characters in extra endstream line
// read next line
endObjectKey = ReadHelper.ReadLine(reader);
}
}
return(currentBase);
}
public long CheckXRefOffset(long startXRefOffset, bool isLenientParsing)
{
// repair mode isn't available in non-lenient mode
if (!isLenientParsing)
{
return(startXRefOffset);
}
source.Seek(startXRefOffset);
ReadHelper.SkipSpaces(source);
if (source.Peek() == 'x' && ReadHelper.IsString(source, "xref"))
{
return(startXRefOffset);
}
if (startXRefOffset > 0)
{
if (CheckXRefStreamOffset(source, startXRefOffset, true, pool))
{
return(startXRefOffset);
}
return(CalculateXRefFixedOffset(startXRefOffset));
}
// can't find a valid offset
return(-1);
}
private void BruteForceSearchForEndOfFileMarker(IInputBytes source)
{
if (lastEndOfFileMarker != null)
{
return;
}
long startOffset = source.CurrentOffset;
source.Seek(MINIMUM_SEARCH_OFFSET);
while (!source.IsAtEnd())
{
// search for EOF marker
if (ReadHelper.IsString(source, "%%EOF"))
{
long tempMarker = source.CurrentOffset;
if (tempMarker >= source.Length)
{
lastEndOfFileMarker = tempMarker;
break;
}
try
{
source.Seek(tempMarker + 5);
// check if the following data is some valid pdf content
// which most likely indicates that the pdf is linearized,
// updated or just cut off somewhere in the middle
ReadHelper.SkipSpaces(source);
ObjectHelper.ReadObjectNumber(source);
ObjectHelper.ReadGenerationNumber(source);
}
catch (Exception)
{
// save the EOF marker as the following data is most likely some garbage
lastEndOfFileMarker = tempMarker;
}
}
source.MoveNext();
}
source.Seek(startOffset);
// no EOF marker found
if (lastEndOfFileMarker == null)
{
lastEndOfFileMarker = long.MaxValue;
}
}
private CosBase ParseObjectFromFile(long offset, IRandomAccessRead reader,
CosObjectKey key,
CosObjectPool pool,
bool isLenientParsing)
{
reader.Seek(offset);
var objectNumber = ObjectHelper.ReadObjectNumber(reader);
var objectGeneration = ObjectHelper.ReadGenerationNumber(reader);
ReadHelper.ReadExpectedString(reader, "obj", true);
if (objectNumber != key.Number || objectGeneration != key.Generation)
{
throw new InvalidOperationException($"Xref for {key} points to object {objectNumber} {objectGeneration} at {offset}");
}
ReadHelper.SkipSpaces(reader);
var baseObject = baseParser.Parse(reader, pool);
var endObjectKey = ReadHelper.ReadString(reader);
var atStreamStart = string.Equals(endObjectKey, "stream");
if (atStreamStart)
{
var streamStartBytes = OtherEncodings.StringAsLatin1Bytes(endObjectKey);
reader.Rewind(streamStartBytes.Length);
baseObject = ReadNormalObjectStream(reader, baseObject, offset, isLenientParsing, out endObjectKey);
}
if (!string.Equals(endObjectKey, "endobj"))
{
var message =
$"Object ({objectNumber}:{objectGeneration}) at offset {offset} does not end with \'endobj\' but with \'{endObjectKey}\'";
if (isLenientParsing)
{
log.Warn(message);
}
else
{
throw new InvalidOperationException(message);
}
}
return(baseObject);
}
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 bfSearchForLastEOFMarker(IRandomAccessRead source)
{
if (lastEOFMarker == null)
{
long originOffset = source.GetPosition();
source.Seek(MINIMUM_SEARCH_OFFSET);
while (!source.IsEof())
{
// search for EOF marker
if (ReadHelper.IsString(source, "%%EOF"))
{
long tempMarker = source.GetPosition();
source.Seek(tempMarker + 5);
try
{
// check if the following data is some valid pdf content
// which most likely indicates that the pdf is linearized,
// updated or just cut off somewhere in the middle
ReadHelper.SkipSpaces(source);
ObjectHelper.ReadObjectNumber(source);
ObjectHelper.ReadGenerationNumber(source);
}
catch (InvalidOperationException exception)
{
// save the EOF marker as the following data is most likely some garbage
lastEOFMarker = tempMarker;
}
}
source.Read();
}
source.Seek(originOffset);
// no EOF marker found
if (lastEOFMarker == null)
{
lastEOFMarker = long.MaxValue;
}
}
}
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);
}
