Fixed GzipInputStream to complete reading gracefully if at least one block has been successfully processed. This allows for a compressed stream that may contain trailing garbage.

Author: silijon

ICSharpCode.SharpZipLib.Tests/GZip/GZipTests.cs

@@ -220,6 +220,58 @@ public void WriteAfterClose()
 			}
 		}
 
+        /// <summary>
+        /// Verify that if a decompression was successful for at least one block we're exiting gracefully.
+        /// </summary>
+        [Test]
+        public void TrailingGarbage()
+        {
+            /* ARRANGE */
+			var ms = new MemoryStream();
+			var outStream = new GZipOutputStream(ms);
+
+            // input buffer to be compressed
+			byte[] buf = new byte[100000];
+			var rnd = new Random();
+			rnd.NextBytes(buf);
+
+            // compress input buffer
+			outStream.Write(buf, 0, buf.Length);
+			outStream.Flush();
+			outStream.Finish();
+
+            // generate random trailing garbage and add to the compressed stream
+            byte[] garbage = new byte[4096];
+            rnd.NextBytes(garbage);
+            ms.Write(garbage, 0, garbage.Length);
+
+            // rewind the concatenated stream
+			ms.Seek(0, SeekOrigin.Begin);
+
+
+            /* ACT */
+            // decompress concatenated stream
+			var inStream = new GZipInputStream(ms);
+			byte[] buf2 = new byte[buf.Length];
+			int currentIndex = 0;
+			int count = buf2.Length;
+            while (true) {
+                int numRead = inStream.Read(buf2, currentIndex, count);
+                if (numRead <= 0) {
+                    break;
+                }
+                currentIndex += numRead;
+                count -= numRead;
+            }
+
+
+            /* ASSERT */
+			Assert.AreEqual(0, count);
+			for (int i = 0; i < buf.Length; ++i) {
+				Assert.AreEqual(buf2[i], buf[i]);
+			}
+        }
+
 		// TODO: Fix This
 		//[Test]
 		//[Category("GZip")]

ICSharpCode.SharpZipLib/GZip/GzipInputStream.cs

@@ -47,6 +47,12 @@ public class GZipInputStream : InflaterInputStream
 		/// This is tracked per-block as the file is parsed.
 		/// </summary>
 		bool readGZIPHeader;
+
+        /// <summary>
+        /// Flag to indicate if at least one block in a stream with concatenated blocks was read successfully.
+        /// This allows us to exit gracefully if downstream data is not in gzip format.
+        /// </summary>
+        bool completedLastBlock;
 		#endregion
 
 		#region Constructors
@@ -100,12 +106,23 @@ public override int Read(byte[] buffer, int offset, int count)
 				// If we haven't read the header for this block, read it
 				if (!readGZIPHeader) {
 
-					// Try to read header. If there is no header (0 bytes available), this is EOF. If there is
-					// an incomplete header, this will throw an exception.
-					if (!ReadHeader()) {
-						return 0;
-					}
-				}
+                    // Try to read header. If there is no header (0 bytes available), this is EOF. If there is
+                    // an incomplete header, this will throw an exception.
+                    try
+                    {
+                        if (!ReadHeader())
+                        {
+                            return 0;
+                        }
+                    }
+                    catch (Exception ex) when (completedLastBlock && (ex is GZipException || ex is EndOfStreamException))
+                    {
+                        // if we completed the last block (i.e. we're in a stream that has multiple blocks concatenated
+                        // we want to return gracefully from any header parsing exceptions since sometimes there may
+                        // be trailing garbage on a stream
+                        return 0;
+                    }
+                }
 
 				// Try to read compressed data
 				int bytesRead = base.Read(buffer, offset, count);
@@ -151,14 +168,14 @@ bool ReadHeader()
 
 			headCRC.Update(magic);
 			if (magic != (GZipConstants.GZIP_MAGIC >> 8)) {
-				throw new GZipException("Error GZIP header, first magic byte doesn't match");
+                throw new GZipException("Error GZIP header, first magic byte doesn't match");
 			}
 
 			//magic = baseInputStream.ReadByte();
 			magic = inputBuffer.ReadLeByte();
 
 			if (magic < 0) {
-				throw new EndOfStreamException("EOS reading GZIP header");
+                throw new EndOfStreamException("EOS reading GZIP header");
 			}
 
 			if (magic != (GZipConstants.GZIP_MAGIC & 0xFF)) {
@@ -324,6 +341,9 @@ void ReadFooter()
 
 			// Mark header read as false so if another header exists, we'll continue reading through the file
 			readGZIPHeader = false;
+
+            // Indicate that we succeeded on at least one block so we can exit gracefully if there is trailing garbage downstream
+            completedLastBlock = true;
 		}
 		#endregion
 	}