Properly handle language markers in decoded text

In text like document title 0x001B is used as a marker for the
beginning/end of a language metadata section. Currently PDFium does
nothing with this data, but when returning the 'decoded' text it needs
to be stripped out.

The existing code assumed that the two bytes following a marker would
be the data to be removed and did nothing to track if it was in/out of
one of these regions. This led to a situation where it would always
strip the two bytes following the region, since it assumed the end
marker was the beginning of a new region.

This CL corrects the detection and handling of these regions, and adds
a regression test for the reported bug.

BUG=pdfium:182

Change-Id: I92ddba5666274a8986fed03f502a0331f150f7ac
Reviewed-on: https://pdfium-review.googlesource.com/41070
Reviewed-by: Henrique Nakashima <hnakashima@chromium.org>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>

1 files changed, 26 insertions, 13 deletions

diff --git a/core/fpdfapi/parser/fpdf_parser_decode.cpp b/core/fpdfapi/parser/fpdf_parser_decode.cpp
index 928d650103..063d21ae5f 100644
--- a/core/fpdfapi/parser/fpdf_parser_decode.cpp
+++ b/core/fpdfapi/parser/fpdf_parser_decode.cpp
@@ -29,8 +29,12 @@ namespace {
 
 const uint32_t kMaxStreamSize = 20 * 1024 * 1024;
 
-uint16_t GetUnicodeFromBytes(const uint8_t* bytes, bool bBE) {
-  return bBE ? (bytes[0] << 8 | bytes[1]) : (bytes[1] << 8 | bytes[0]);
+uint16_t GetUnicodeFromBigEndianBytes(const uint8_t* bytes) {
+  return bytes[0] << 8 | bytes[1];
+}
+
+uint16_t GetUnicodeFromLittleEndianBytes(const uint8_t* bytes) {
+  return bytes[1] << 8 | bytes[0];
 }
 
 bool CheckFlateDecodeParams(int Colors, int BitsPerComponent, int Columns) {
@@ -436,21 +440,30 @@ WideString PDF_DecodeText(const uint8_t* src_data, uint32_t src_len) {
       return result;
 
     pdfium::span<wchar_t> dest_buf = result.GetBuffer(max_chars);
-    bool bBE = src_data[0] == 0xfe || (src_data[0] == 0xff && !src_data[2]);
-    const uint8_t* uni_str = src_data + 2;
+    uint16_t (*GetUnicodeFromBytes)(const uint8_t*) =
+        src_data[0] == 0xfe ? GetUnicodeFromBigEndianBytes
+                            : GetUnicodeFromLittleEndianBytes;
+    const uint8_t* unicode_str = src_data + 2;
     for (uint32_t i = 0; i < max_chars * 2; i += 2) {
-      uint16_t unicode = GetUnicodeFromBytes(uni_str + i, bBE);
-      if (unicode != 0x1b) {
-        dest_buf[dest_pos++] = unicode;
-        continue;
-      }
-      i += 2;
-      while (i < max_chars * 2) {
-        uint16_t unicode2 = GetUnicodeFromBytes(uni_str + i, bBE);
+      uint16_t unicode = GetUnicodeFromBytes(unicode_str + i);
+
+      // 0x001B is a begin/end marker for language metadata region that
+      // should not be in the decoded text.
+      if (unicode == 0x001B) {
         i += 2;
-        if (unicode2 == 0x1b)
+        for (; i < max_chars * 2; i += 2) {
+          unicode = GetUnicodeFromBytes(unicode_str + i);
+          if (unicode == 0x001B) {
+            i += 2;
+            unicode = GetUnicodeFromBytes(unicode_str + i);
+            break;
+          }
+        }
+        if (i >= max_chars * 2)
           break;
       }
+
+      dest_buf[dest_pos++] = unicode;
     }
   } else {
     pdfium::span<wchar_t> dest_buf = result.GetBuffer(src_len);