better recovery for appended files with damaged cross-reference tables

git-svn-id: svn+q:///qpdf/trunk@649 71b93d88-0707-0410-a8cf-f5a4172ac649

better recovery for appended files with damaged cross-reference tables
git-svn-id: svn+q:///qpdf/trunk@649 71b93d88-0707-0410-a8cf-f5a4172ac649
Jay Berkenbilt
1 parent 35d72c82
Showing 4 changed files with 37 additions and 6 deletions
ChangeLog
TODO
include/qpdf/QPDF.hh
libqpdf/QPDF.cc
+2009-02-20  Jay Berkenbilt  <ejb@ql.org>
+
+	* libqpdf/QPDF.cc (QPDF::insertXrefEntry): when recovering the
+	cross-reference table, have objects we encounter later in the file
+	supersede those we found earlier.  This improves the chances of
+	being able to recover appended files with damaged cross-reference
+	tables.
+
 2009-02-19  Jay Berkenbilt  <ejb@ql.org>
 	* libqpdf/Pl_LZWDecoder.cc: correct logic error for previously
+2.0.4
+=====
+
+ * Spell check to fix types in messages and comments.  Known type in
+   "damanged".
+
+ * Exit with a different exit code if warning are found during check
+   mode.
+
+
 General
 =======
@@ -333,7 +333,8 @@ class QPDF
     int read_xrefTable(off_t offset);
     int read_xrefStream(off_t offset);
     int processXRefStream(off_t offset, QPDFObjectHandle& xref_stream);
-    void insertXrefEntry(int obj, int f0, int f1, int f2);
+    void insertXrefEntry(int obj, int f0, int f1, int f2,
+			 bool overwrite = false);
     QPDFObjectHandle readObject(
 	InputSource*, int objid, int generation,
 	bool in_object_stream);
@@ -435,7 +435,7 @@ QPDF::reconstruct_xref(QPDFExc&amp; e)
 		int obj = atoi(m.getMatch(1).c_str());
 		int gen = atoi(m.getMatch(2).c_str());
 		int offset = this->file.getLastOffset();
-		insertXrefEntry(obj, 1, offset, gen);
+		insertXrefEntry(obj, 1, offset, gen, true);
 	    }
 	    else if ((! this->trailer.isInitialized()) &&
 		     trailer_re.match(line.c_str()))
@@ -865,11 +865,15 @@ QPDF::processXRefStream(off_t xref_offset, QPDFObjectHandle&amp; xref_obj)
 }
 void
-QPDF::insertXrefEntry(int obj, int f0, int f1, int f2)
+QPDF::insertXrefEntry(int obj, int f0, int f1, int f2, bool overwrite)
 {
     // Populate the xref table in such a way that the first reference
     // to an object that we see, which is the one in the latest xref
-    // table in which it appears, is the one that gets stored.
+    // table in which it appears, is the one that gets stored.  This
+    // works because we are reading more recent appends before older
+    // ones.  Exception: if overwrite is true, then replace any
+    // existing object.  This is used in xref recovery mode, which
+    // reads the file from beginning to end.
     // If there is already an entry for this object and generation in
     // the table, it means that a later xref table has registered this
@@ -879,8 +883,16 @@ QPDF::insertXrefEntry(int obj, int f0, int f1, int f2)
 	ObjGen og(obj, gen);
 	if (this->xref_table.count(og))
 	{
-	    QTC::TC("qpdf", "QPDF xref reused object");
-	    return;
+	    if (overwrite)
+	    {
+		QTC::TC("qpdf", "QPDF xref overwrite object");
+		this->xref_table.erase(og);
+	    }
+	    else
+	    {
+		QTC::TC("qpdf", "QPDF xref reused object");
+		return;
+	    }
 	}
 	if (this->deleted_objects.count(obj))
 	{