Use CFX_WideString in CPDF_NameTree functions to strip BOMchromium/3162

PDFium doesn't strip BOMs during parsing, but we should strip BOMs when
retrieving parsed strings in CPDF_NameTree to ensure consistency and
appropriate function behavior. See the bug for more info.

As outlined in Bug=pdfium:593, the solution is to call GetUnicodeText()
instead of GetString(). I added a GetUnicodeTextAt() function in
CPDF_Array, which is symmetrical to GetUnicodeTextFor() in
CPDF_Dictionary.

I then changed the input variable types to CPDF_NameTree functions to
be CFX_WideString instead of CFX_ByteString, and modified all the
calls to them.

I also added a unit test for nametree, which would fail prior to this
change. Nametrees with non-unicode names are already tested by embedder
tests.

Bug=pdfium:820

Change-Id: Id69d7343632f83d1f5180348c0eea290f478183f
Reviewed-on: https://pdfium-review.googlesource.com/8091
Reviewed-by: dsinclair <dsinclair@chromium.org>
Commit-Queue: Jane Liu <janeliulwq@google.com>

1 files changed, 4 insertions, 5 deletions

diff --git a/xfa/fxfa/cxfa_ffdoc.cpp b/xfa/fxfa/cxfa_ffdoc.cpp
index 01c570adbe..326fc7f228 100644
--- a/xfa/fxfa/cxfa_ffdoc.cpp
+++ b/xfa/fxfa/cxfa_ffdoc.cpp
@@ -365,13 +365,12 @@ CFX_RetainPtr<CFX_DIBitmap> CXFA_FFDoc::GetPDFNamedImage(
     return nullptr;
 
   CPDF_NameTree nametree(pXFAImages);
-  CFX_ByteString bsName = PDF_EncodeText(wsName.c_str(), wsName.GetLength());
-  CPDF_Object* pObject = nametree.LookupValue(bsName);
+  CPDF_Object* pObject = nametree.LookupValue(CFX_WideString(wsName));
   if (!pObject) {
     for (size_t i = 0; i < nametree.GetCount(); i++) {
-      CFX_ByteString bsTemp;
-      CPDF_Object* pTempObject = nametree.LookupValueAndName(i, &bsTemp);
-      if (bsTemp == bsName) {
+      CFX_WideString wsTemp;
+      CPDF_Object* pTempObject = nametree.LookupValueAndName(i, &wsTemp);
+      if (wsTemp == wsName) {
         pObject = pTempObject;
         break;
       }