pdfviewer: load files with missing xref (we need in order to help with testing, as most good pdfx i…

experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.cpp

 
 #include "SkPdfNativeTokenizer.h"
 #include "SkPdfObject.h"
 #include "SkPdfConfig.h"
 
 #include "SkPdfStreamCommonDictionary_autogen.h"
 #include "SkPdfImageDictionary_autogen.h"
 
 // TODO(edisonn): perf!!!
 // there could be 0s between start and end! but not in the needle.
@@ skipping 65 matching lines @@
 
 #else
 #define TRACE_INDENT(level,type)
 #define TRACE_COMMENT(ch)
 #define TRACE_TK(ch)
 #define TRACE_NAME(start,end)
 #define TRACE_STRING(start,end)
 #define TRACE_HEXSTRING(start,end)
 #endif
 
-static const unsigned char* skipPdfWhiteSpaces(int level, const unsigned char* start, const unsigned char* end) {
+const unsigned char* skipPdfWhiteSpaces(int level, const unsigned char* start, const unsigned char* end) {
     TRACE_INDENT(level, "White Space");
-    while (start < end && isPdfWhiteSpace(*start)) {
+    while (start < end && (isPdfWhiteSpace(*start) || *start == kComment_PdfDelimiter)) {
         TRACE_COMMENT(*start);
         if (*start == kComment_PdfDelimiter) {
             // skip the comment until end of line
             while (start < end && !isPdfEOL(*start)) {
                 //*start = '\0';
                 start++;
                 TRACE_COMMENT(*start);
             }
         } else {
             //*start = '\0';
             start++;
         }
     }
     return start;
 }
 
 // TODO(edisonn) '(' can be used, will it break the string a delimiter or space inside () ?
-static const unsigned char* endOfPdfToken(int level, const unsigned char* start, const unsigned char* end) {
+const unsigned char* endOfPdfToken(int level, const unsigned char* start, const unsigned char* end) {
     //int opened brackets
     //TODO(edisonn): what out for special chars, like \n, \032
     TRACE_INDENT(level, "Token");
 
     SkASSERT(!isPdfWhiteSpace(*start));
 
     if (start < end && isPdfDelimiter(*start)) {
         TRACE_TK(*start);
         start++;
         return start;
@@ skipping 512 matching lines @@
     int64_t length = -1;
 
     // TODO(edisonn): very basic implementation
     if (stream->has_Length() && stream->Length(doc) > 0) {
         length = stream->Length(doc);
     }
 
     // TODO(edisonn): laod external streams
     // TODO(edisonn): look at the last filter, to determione how to deal with possible issue
 
+
+    if (length >= 0) {
+        const unsigned char* endstream = start + length;
+
+        if (endstream[0] == kCR_PdfWhiteSpace && endstream[1] == kLF_PdfWhiteSpace) {
+            endstream += 2;
+        } else if (endstream[0] == kLF_PdfWhiteSpace) {
+            endstream += 1;
+        }
+
+        if (strncmp((const char*)endstream, "endstream", strlen("endstream")) != 0) {
+            length = -1;
+        }
+    }
+
     if (length < 0) {
         // scan the buffer, until we find first endstream
         // TODO(edisonn): all buffers must have a 0 at the end now,
         const unsigned char* endstream = (const unsigned char*)strrstrk((char*)start, (char*)end, "endstream");
 
         if (endstream) {
             length = endstream - start;
             if (*(endstream-1) == kLF_PdfWhiteSpace) length--;
             if (*(endstream-2) == kCR_PdfWhiteSpace) length--;
         }
@@ skipping 420 matching lines @@
             SkPdfObject* obj = fAllocator->allocObject();
             fUncompressedStream = nextObject(0, fUncompressedStream, fUncompressedStreamEnd, obj, fAllocator, fDoc);
             // TODO(edisonn): perf maybe we should not expand abreviation like this
             inlineImage->set(inlineImageKeyAbbreviationExpand(key),
                              inlineImageValueAbbreviationExpand(obj));
         }
     }
     // TODO(edisonn): report end of data with inline image without an EI
     return inlineImage;
 }