Apply patch from Jukka from bug #43670 to improve HDGF v11 Separator detection, and handle short strings better, hopefully solving the Negative length of ChunkHeader issue

git-svn-id: https://svn.apache.org/repos/asf/poi/trunk@904052 13f79535-47bb-0310-9956-ffa450edef68

src/documentation/content/xdocs/status.xml

@@ -34,6 +34,7 @@
 
     <changes>
         <release version="3.7-SNAPSHOT" date="2010-??-??">
+           <action dev="POI-DEVELOPERS" type="fix">43670 - Improve HDGF ChunkV11 separator detection, and short string detection, to solve the "Negative length of ChunkHeader" problem</action>
            <action dev="POI-DEVELOPERS" type="add">48617 - Optionally allow the overriding of the Locale used by DataFormatter to control how the default number and date formats should look</action>
            <action dev="POI-DEVELOPERS" type="add">New event based xssf text extractor (XSSFEventBasedExcelExtractor)</action>
            <action dev="POI-DEVELOPERS" type="add">ExtractorFactory can now be told to prefer Event Based extractors (current Excel only) on a per-thread or overall basis</action>

src/scratchpad/src/org/apache/poi/hdgf/chunks/Chunk.java

@@ -180,6 +180,14 @@ public final class Chunk {
 				// A Little Endian String
 				// Starts 8 bytes into the data segment
 				// Ends at end of data, or 00 00
+			   
+				// Ensure we have enough data
+				if(contents.length < 8) {
+					command.value = "";
+					break;
+				}
+			   
+				// Find the end point
 				int startsAt = 8;
 				int endsAt = startsAt;
 				for(int j=startsAt; j<contents.length-1 && endsAt == startsAt; j++) {
@@ -190,7 +198,7 @@ public final class Chunk {
 				if(endsAt == startsAt) {
 					endsAt = contents.length;
 				}
-
+				
 				int strLen = (endsAt-startsAt) / 2;
 				command.value = StringUtil.getFromUnicodeLE(contents, startsAt, strLen);
 				break;

src/scratchpad/src/org/apache/poi/hdgf/chunks/ChunkHeaderV11.java

@@ -33,7 +33,10 @@ public final class ChunkHeaderV11 extends ChunkHeaderV6 {
 		if(hasTrailer()) { return true; }
 
 		if(unknown2 == 2 && unknown3 == 0x55) { return true; }
+		if(unknown2 == 2 && unknown3 == 0x54 && type == 0xa9) { return true; }
 		if(unknown2 == 2 && unknown3 == 0x54 && type == 0xaa) { return true; }
+		if(unknown2 == 2 && unknown3 == 0x54 && type == 0xb4) { return true; }
+		if(unknown2 == 2 && unknown3 == 0x54 && type == 0xb6) { return true; }
 		if(unknown2 == 3 && unknown3 != 0x50) { return true; }
 		if(type == 0x69) { return true; }
 

src/scratchpad/testcases/org/apache/poi/hdgf/TestHDGFCore.java

@@ -63,13 +63,18 @@ public final class TestHDGFCore extends TestCase {
 	}
 
 	/**
-	 * Tests that we can open a problematic file, that initially
+	 * Tests that we can open a problematic file, that used to
-	 *  appears to have a negative chunk length
+	 *  break with a negative chunk length
 	 */
-	public void DISABLEDtestNegativeChunkLength() throws Exception {
+	public void testNegativeChunkLength() throws Exception {
 		fs = new POIFSFileSystem(_dgTests.openResourceAsStream("NegativeChunkLength.vsd"));
 
 		HDGFDiagram hdgf = new HDGFDiagram(fs);
 		assertNotNull(hdgf);
+		
+		// And another file
+		fs = new POIFSFileSystem(_dgTests.openResourceAsStream("NegativeChunkLength2.vsd"));
+		hdgf = new HDGFDiagram(fs);
+		assertNotNull(hdgf);
 	}
 }