fix issue with newlines in object start tokens #88

where we brute force the file and it contains newlines between object tokens we fix the parsing to prevent pseudo-infinite loops.

src/UglyToad.PdfPig.Core/ReadHelper.cs

@@ -123,7 +123,7 @@
         }
         
         /// <summary>
-        /// Determines if a character is whitespace or not.
+        /// Determines if a character is whitespace or not, this includes newlines.
         /// </summary>
         /// <remarks>
         /// These values are specified in table 1 (page 12) of ISO 32000-1:2008.

src/UglyToad.PdfPig.Tests/Parser/Parts/BruteForceSearcherTests.cs

@@ -36,7 +36,7 @@ startxref
 216
 %%EOF";
 
-        private static readonly long[] TestDataOffsets = 
+        private static readonly long[] TestDataOffsets =
         {
             TestData.IndexOf("2 17 obj", StringComparison.OrdinalIgnoreCase),
             TestData.IndexOf("3 0 obj", StringComparison.OrdinalIgnoreCase),
@@ -57,7 +57,7 @@ startxref
         public void SearcherFindsCorrectObjects()
         {
             var input = new ByteArrayInputBytes(OtherEncodings.StringAsLatin1Bytes(TestData));
-            
+
             var locations = BruteForceSearcher.GetObjectLocations(input);
 
             Assert.Equal(4, locations.Count);
@@ -69,11 +69,11 @@ startxref
         public void ReaderOnlyCallsOnce()
         {
             var reader = StringBytesTestConverter.Convert(TestData, false);
-            
+
             var locations = BruteForceSearcher.GetObjectLocations(reader.Bytes);
 
             Assert.Equal(4, locations.Count);
-            
+
             var newLocations = BruteForceSearcher.GetObjectLocations(reader.Bytes);
 
             Assert.Equal(4, locations.Count);
@@ -132,7 +132,7 @@ endobj
 5 0 obj
 << /IsEmpty false >>
 endobj";
-            
+
             var bytes = new ByteArrayInputBytes(OtherEncodings.StringAsLatin1Bytes(s));
 
             var locations = BruteForceSearcher.GetObjectLocations(bytes);
@@ -168,12 +168,35 @@ endobj";
                 Assert.Equal(581, locations[new IndirectReference(7, 0)]);
                 Assert.Equal(5068, locations[new IndirectReference(8, 0)]);
                 Assert.Equal(5091, locations[new IndirectReference(9, 0)]);
-                
+
                 var s = GetStringAt(bytes, locations[new IndirectReference(3, 0)]);
                 Assert.StartsWith("3 0 obj", s);
             }
         }
 
+        [Fact]
+        public void BruteForceSearcherBytesFileOffsetsCorrect()
+        {
+            var bytes = new ByteArrayInputBytes(File.ReadAllBytes(IntegrationHelpers.GetDocumentPath("Single Page Simple - from inkscape.pdf")));
+
+            var locations = BruteForceSearcher.GetObjectLocations(bytes);
+
+            Assert.Equal(13, locations.Count);
+
+            Assert.Equal(6183, locations[new IndirectReference(1, 0)]);
+            Assert.Equal(244, locations[new IndirectReference(2, 0)]);
+            Assert.Equal(15, locations[new IndirectReference(3, 0)]);
+            Assert.Equal(222, locations[new IndirectReference(4, 0)]);
+            Assert.Equal(5766, locations[new IndirectReference(5, 0)]);
+            Assert.Equal(353, locations[new IndirectReference(6, 0)]);
+            Assert.Equal(581, locations[new IndirectReference(7, 0)]);
+            Assert.Equal(5068, locations[new IndirectReference(8, 0)]);
+            Assert.Equal(5091, locations[new IndirectReference(9, 0)]);
+
+            var s = GetStringAt(bytes, locations[new IndirectReference(3, 0)]);
+            Assert.StartsWith("3 0 obj", s);
+        }
+
         [Fact]
         public void BruteForceSearcherFileOffsetsCorrectOpenOffice()
         {

src/UglyToad.PdfPig/Parser/Parts/BruteForceSearcher.cs

@@ -42,6 +42,8 @@
 
             var currentlyInObject = false;
 
+            var objBuffer = new byte[4];
+
             do
             {
                 if (loopProtection > 1_000_000)
@@ -94,17 +96,24 @@
 
                 bytes.Seek(currentOffset);
 
-                if (!ReadHelper.IsString(bytes, " obj"))
+                bytes.Read(objBuffer);
+
+                if (!IsStartObjMarker(objBuffer))
                 {
                     currentOffset++;
                     continue;
                 }
 
                 // Current byte is ' '[obj]
-                var offset = currentOffset - 1;
+                var offset = currentOffset + 1;
 
                 bytes.Seek(offset);
 
+                while (ReadHelper.IsWhitespace(bytes.CurrentByte) && offset >= MinimumSearchOffset)
+                {
+                    bytes.Seek(--offset);
+                }
+
                 while (ReadHelper.IsDigit(bytes.CurrentByte) && offset >= MinimumSearchOffset)
                 {
                     generationBytes.Insert(0, (char)bytes.CurrentByte);
@@ -113,13 +122,16 @@
                 }
 
                 // We should now be at the space between object and generation number.
-                if (!ReadHelper.IsSpace(bytes.CurrentByte))
+                if (!ReadHelper.IsWhitespace(bytes.CurrentByte))
                 {
                     currentOffset++;
                     continue;
                 }
 
-                bytes.Seek(--offset);
+                while (ReadHelper.IsWhitespace(bytes.CurrentByte))
+                {
+                    bytes.Seek(--offset);
+                }
 
                 while (ReadHelper.IsDigit(bytes.CurrentByte) && offset >= MinimumSearchOffset)
                 {
@@ -185,5 +197,17 @@
             bytes.Seek(originalOffset);
             return long.MaxValue;
         }
+
+        private static bool IsStartObjMarker(byte[] data)
+        {
+            if (!ReadHelper.IsWhitespace(data[0]))
+            {
+                return false;
+            }
+
+            return (data[1] == 'o' || data[1] == 'O')
+                   && (data[2] == 'b' || data[2] == 'B')
+                   && (data[3] == 'j' || data[3] == 'J');
+        }
     }
 }