Try to repair xref offset by looking for all startxref and fix #1040

Author: BobLd

src/UglyToad.PdfPig.Core/ReadHelper.cs

@@ -189,7 +189,34 @@ public static bool IsString(IInputBytes bytes, string s)
 
             return found;
         }
-        
+
+
+        /// <summary>
+        /// Whether the given string is at this position in the input.
+        /// Resets to the current offset once read.
+        /// </summary>
+        public static bool IsString(IInputBytes bytes, ReadOnlySpan<byte> s)
+        {
+            bool found = true;
+
+            var startOffset = bytes.CurrentOffset;
+
+            foreach (var c in s)
+            {
+                if (bytes.CurrentByte != c)
+                {
+                    found = false;
+                    break;
+                }
+
+                bytes.MoveNext();
+            }
+
+            bytes.Seek(startOffset);
+
+            return found;
+        }
+
         /// <summary>
         /// Read a long from the input.
         /// </summary>
@@ -252,14 +279,6 @@ public static int ReadInt(IInputBytes bytes)
                 throw new PdfDocumentFormatException($"Error: Expected an integer type at offset {bytes.CurrentOffset}, instead got \'{OtherEncodings.BytesAsLatin1String(intBytes)}\'");
             }
         }
-        
-        /// <summary>
-        /// Whether the given character is a space.
-        /// </summary>
-        public static bool IsSpace(int c)
-        {
-            return c == ' ';
-        }
 
         /// <summary>
         /// Whether the given character value is a valid hex value.

src/UglyToad.PdfPig.Tests/Integration/GithubIssuesTests.cs

@@ -7,6 +7,21 @@
 
     public class GithubIssuesTests
     {
+        [Fact]
+        public void Issue1040()
+        {
+            var path = IntegrationHelpers.GetSpecificTestDocumentPath("pdfpig-issue-1040.pdf");
+
+            using (var document = PdfDocument.Open(path, new ParsingOptions() { UseLenientParsing = true}))
+            {
+                var page1 = document.GetPage(1);
+                Assert.NotEmpty(page1.Letters);
+
+                var page2 = document.GetPage(2);
+                Assert.NotEmpty(page2.Letters);
+            }
+        }
+        
         [Fact]
         public void Issue1013()
         {

src/UglyToad.PdfPig.Tests/Integration/SpecificTestDocuments/pdfpig-issue-1040.pdf

@@ -1,7 +1,6 @@
 namespace UglyToad.PdfPig.Parser.FileStructure
 {
     using System;
-    using System.Collections.Generic;
     using Core;
     using Tokenization.Scanner;
     using Tokens;
@@ -26,7 +25,7 @@ internal static class FileTrailerParser
         /// </summary>
         private const int EndOfFileSearchRange = 2048;
 
-        private static ReadOnlySpan<byte> StartXRefBytes => "startxref"u8;
+        internal static ReadOnlySpan<byte> StartXRefBytes => "startxref"u8;
 
         public static long GetFirstCrossReferenceOffset(IInputBytes bytes, ISeekableTokenScanner scanner, bool isLenientParsing)
         {

src/UglyToad.PdfPig/Parser/FileStructure/FileTrailerParser.cs

@@ -7,12 +7,16 @@
     using Tokenization.Scanner;
     using Tokens;
 
-    internal class XrefOffsetValidator
+    internal sealed class XrefOffsetValidator
     {
-        private static readonly long MinimumSearchOffset = 6;
+        private const long MinimumSearchOffset = 6;
+
+        private static ReadOnlySpan<byte> XRefBytes => "xref"u8;
+        private static ReadOnlySpan<byte> SpaceObjBytes => " obj"u8;
 
         private readonly ILog log;
 
+        private List<long>? bfSearchStartXRefTablesOffsets;
         private List<long>? bfSearchXRefTablesOffsets;
         private List<long>? bfSearchXRefStreamsOffsets;
 
@@ -90,16 +94,18 @@ private long BruteForceSearchForXref(long xrefOffset, ISeekableTokenScanner scan
 
             BfSearchForXRefStreams(reader);
 
-            if (bfSearchXRefTablesOffsets != null)
+            if (bfSearchXRefTablesOffsets != null && bfSearchXRefTablesOffsets.Count > 0)
             {
                 // TODO to be optimized, this won't work in every case
                 newOffsetTable = SearchNearestValue(bfSearchXRefTablesOffsets, xrefOffset);
             }
-            if (bfSearchXRefStreamsOffsets != null)
+            
+            if (bfSearchXRefStreamsOffsets != null && bfSearchXRefStreamsOffsets.Count > 0)
             {
                 // TODO to be optimized, this won't work in every case
                 newOffsetStream = SearchNearestValue(bfSearchXRefStreamsOffsets, xrefOffset);
             }
+            
             // choose the nearest value
             if (newOffsetTable > -1 && newOffsetStream > -1)
             {
@@ -126,9 +132,91 @@ private long BruteForceSearchForXref(long xrefOffset, ISeekableTokenScanner scan
                 newOffset = newOffsetStream;
                 bfSearchXRefStreamsOffsets!.Remove(newOffsetStream);
             }
+            else
+            {
+                log.Warn("Trying to repair xref offset by looking for all startxref.");
+                if (TryBruteForceSearchForXrefFromStartxref(xrefOffset, scanner, reader, out long newOffsetFromStartxref))
+                {
+                    newOffset = newOffsetFromStartxref;
+                }
+            }
+            
             return newOffset;
         }
 
+        private bool TryBruteForceSearchForXrefFromStartxref(long xrefOffset, ISeekableTokenScanner scanner, IInputBytes reader, out long newOffset)
+        {
+            newOffset = -1;
+            BruteForceSearchForStartxref(reader);
+            long newStartXRefOffset = SearchNearestValue(bfSearchStartXRefTablesOffsets, xrefOffset);
+            if (newStartXRefOffset < reader.Length)
+            {
+                long tempNewOffset = -1;
+                var startOffset = scanner.CurrentPosition;
+                scanner.Seek(newStartXRefOffset + 9);
+
+                if (scanner.MoveNext() && scanner.CurrentToken is NumericToken token)
+                {
+                    tempNewOffset = token.Long;
+                }
+
+                if (tempNewOffset > -1)
+                {
+                    scanner.Seek(tempNewOffset);
+                    scanner.MoveNext();
+                    if (ReferenceEquals(scanner.CurrentToken, OperatorToken.Xref))
+                    {
+                        newOffset = tempNewOffset;
+                    }
+
+                    if (CheckXRefStreamOffset(tempNewOffset, scanner, true))
+                    {
+                        newOffset = tempNewOffset;
+                    }
+                }
+
+                scanner.Seek(startOffset);
+            }
+
+            return newOffset != -1;
+        }
+
+        private void BruteForceSearchForStartxref(IInputBytes bytes)
+        {
+            if (bfSearchStartXRefTablesOffsets != null)
+            {
+                return;
+            }
+
+            // a pdf may contain more than one startxref entry
+            bfSearchStartXRefTablesOffsets = new List<long>();
+
+            var startOffset = bytes.CurrentOffset;
+
+            bytes.Seek(MinimumSearchOffset);
+
+            // search for startxref
+            while (bytes.MoveNext() && !bytes.IsAtEnd())
+            {
+                if (ReadHelper.IsString(bytes, FileTrailerParser.StartXRefBytes))
+                {
+                    var newOffset = bytes.CurrentOffset;
+
+                    bytes.Seek(newOffset - 1);
+
+                    if (ReadHelper.IsWhitespace(bytes.CurrentByte))
+                    {
+                        bfSearchStartXRefTablesOffsets.Add(newOffset);
+                    }
+
+                    bytes.Seek(newOffset + 9);
+                }
+
+            }
+
+            bytes.Seek(startOffset);
+        }
+
         private void BruteForceSearchForTables(IInputBytes bytes)
         {
             if (bfSearchXRefTablesOffsets != null)
@@ -146,7 +234,7 @@ private void BruteForceSearchForTables(IInputBytes bytes)
             // search for xref tables
             while (bytes.MoveNext() && !bytes.IsAtEnd())
             {
-                if (ReadHelper.IsString(bytes, "xref"))
+                if (ReadHelper.IsString(bytes, XRefBytes))
                 {
                     var newOffset = bytes.CurrentOffset;
 
@@ -180,11 +268,9 @@ private void BfSearchForXRefStreams(IInputBytes bytes)
             bytes.Seek(MinimumSearchOffset);
 
             // search for XRef streams
-            var objString = " obj";
-
             while (bytes.MoveNext() && !bytes.IsAtEnd())
             {
-                if (!ReadHelper.IsString(bytes, "xref"))
+                if (!ReadHelper.IsString(bytes, XRefBytes))
                 {
                     continue;
                 }
@@ -209,7 +295,7 @@ private void BfSearchForXRefStreams(IInputBytes bytes)
 
                         for (int j = 0; j < 10; j++)
                         {
-                            if (ReadHelper.IsString(bytes, objString))
+                            if (ReadHelper.IsString(bytes, SpaceObjBytes))
                             {
                                 long tempOffset = currentOffset - 1;
 
@@ -224,7 +310,7 @@ private void BfSearchForXRefStreams(IInputBytes bytes)
                                     bytes.Seek(tempOffset);
 
                                     // is the digit preceded by a space?
-                                    if (ReadHelper.IsSpace(bytes.CurrentByte))
+                                    if (ReadHelper.IsWhitespace(bytes.CurrentByte))
                                     {
                                         int length = 0;
                                         bytes.Seek(--tempOffset);