Fixing invalid lengths beyond hardcoded threshold #347

src/foundation/src/PDFsharp/src/PdfSharp/Pdf.IO/Parser.cs

@@ -553,16 +553,30 @@ bool TryReadEndStream(PdfDictionary dict, SizeType streamStart, ref int streamLe
             var oldLength = streamLength;
             //_lexer.DetermineStreamLength(dict.Reference!.Position, streamLength - length, ref streamLength);
 
-            // Try to read 20 extra bytes in case reported stream length is too small.
-            int scanWindow = streamLength + 20;
+            // If this mismatch is larger than a few bytes, search up to the next object (or EOF).
+            // This keeps valid files fast and improves robustness for malformed /Length values.
+            var behindPosition = _document.IrefTable.GetPositionOfObjectBehind(dict, streamStart);
+            int scanWindow;
+            if (behindPosition != -1)
+            {
+                // Read up to next object.
+                scanWindow = (int)(behindPosition - streamStart);
+            }
+            else
+            {
+                // Object is the last one in the file, so read up to EOF.
+                scanWindow = (int)(_lexer.PdfLength - streamStart);
+            }
+
             // Make sure we do not try to read beyond EOF.
             if (streamStart + scanWindow > _lexer.PdfLength)
             {
                 // We're close to the EOF, so casting to int is OK here.
                 scanWindow = (int)(_lexer.PdfLength - streamStart);
-                Debug.Assert(scanWindow >= oldLength);
             }
 
+            Debug.Assert(scanWindow >= oldLength);
+
             streamLength = _lexer.DetermineStreamLength(streamStart, scanWindow, suppressObjectOrderExceptions);
             if (SuppressExceptions.HasError(suppressObjectOrderExceptions))
                 return false;

src/foundation/src/PDFsharp/tests/PdfSharp.Tests/IO/ReaderTests.cs

@@ -17,6 +17,9 @@
 using Xunit;
 using FluentAssertions;
 using PdfSharp.Diagnostics;
+using System.Runtime.CompilerServices;
+using static System.Net.Mime.MediaTypeNames;
+using System.Text;
 
 #if PDFSHARP_DEBUG
 using static PdfSharp.Diagnostics.DebugBreakHelper;
@@ -567,5 +570,29 @@ static void DrawFooter(PdfPage page, XBrush color, XRect box, string footer)
                 gfx.DrawString(footer, font, color, box, format);
             }
         }
+
+        [Fact]
+        public async Task Read_FileWithInvalidStreamLengths_Success()
+        {
+            var doc = new PdfDocument();
+            var page = doc.AddPage();
+            using var gfx = XGraphics.FromPdfPage(page);
+
+            var font = new XFont(UnitTestFontResolver.ArialFont, 20, XFontStyleEx.Regular);
+
+            gfx.DrawString(string.Concat(Enumerable.Repeat("Hello World!", 30)), font, XBrushes.Black, new XPoint(100, 100));
+
+            using var stream = new MemoryStream();
+            await doc.SaveAsync(stream, false);
+
+            using StreamReader reader = new StreamReader(stream, PdfSharp.Pdf.Internal.PdfEncoders.RawEncoding);
+            var fullFile = await reader.ReadToEndAsync();
+            var invalidLengthFile = fullFile.Replace("/Length 417", "/Length 317");
+            fullFile.Should().NotBe(invalidLengthFile);
+            MemoryStream newStream = new MemoryStream(PdfSharp.Pdf.Internal.PdfEncoders.RawEncoding.GetBytes(invalidLengthFile));
+
+            var doc2 = PdfReader.Open(newStream, PdfDocumentOpenMode.Import);
+            doc2.Should().NotBeNull();
+        }
     }
 }