Support linearized loading

core/src/fpdfapi/fpdf_parser/fpdf_parser_parser_embeddertest.cpp

Index: core/src/fpdfapi/fpdf_parser/fpdf_parser_parser_embeddertest.cpp
diff --git a/core/src/fpdfapi/fpdf_parser/fpdf_parser_parser_embeddertest.cpp b/core/src/fpdfapi/fpdf_parser/fpdf_parser_parser_embeddertest.cpp
index 96ea766d4dfc437bff2a5fac01a6eb71c7715ca6..4a88e07201191f6ccb1721cc9ada7a2a43c34325 100644
--- a/core/src/fpdfapi/fpdf_parser/fpdf_parser_parser_embeddertest.cpp
+++ b/core/src/fpdfapi/fpdf_parser/fpdf_parser_parser_embeddertest.cpp
@@ -8,8 +8,8 @@
 class FPDFParserEmbeddertest : public EmbedderTest {};
 
 TEST_F(FPDFParserEmbeddertest, LoadError_454695) {
-  // Test trailer dictionary with $$ze instead of Size.
-  EXPECT_TRUE(OpenDocument("testing/resources/bug_454695.pdf"));
+  // Test a dictionary with Hex string rather than expected.
+  EXPECT_FALSE(OpenDocument("testing/resources/bug_454695.pdf"));

[Tom Sepez, 2015/10/14 16:57:45]

Are we sure its OK to fail these documents?  

[jun_fang, 2015/10/15 10:22:28]

Before this CL, we didn't check whether document_ is nullptr. 
For bug_454695.pdf, the returned document_ is nullptr. I added 
a checking in this CL. if document_ is nullptr, it returns false.

 if (!FPDFAvail_IsLinearized(avail_)) {
   document_ = FPDF_LoadCustomDocument(&file_access_, nullptr);
 } else {
   document_ = FPDFAvail_GetDocument(avail_, nullptr);
 }

 }
 
 TEST_F(FPDFParserEmbeddertest, Bug_481363) {
@@ -19,3 +19,7 @@ TEST_F(FPDFParserEmbeddertest, Bug_481363) {
   EXPECT_NE(nullptr, page);
   UnloadPage(page);
 }
+
+TEST_F(FPDFParserEmbeddertest, Feature_Linearized_Loading) {
+  EXPECT_TRUE(OpenDocument("testing/resources/feature_linearized_loading.pdf"));

[Tom Sepez, 2015/10/14 16:57:45]

How do we know that the linearized loading path was taken? If I accidentally
changed feature_linearized_loading.pdf to not do so, this would still pass.

[jun_fang, 2015/10/15 10:22:28]

I am not clear about this question.

[Tom Sepez, 2015/10/15 16:23:30]

Ah.  Suppose you landed this test, but did not land the corresponding code that
implements linearized loading.  In that case, this test would still pass. Is
there a way that we can make it so the test fails unless linearized loading
support is present?  If someone were to break the code such that the linearized
loading path was never taken, instead always falling back to the non-linearized
path, we'd like to detect that in the test.

[jun_fang, 2015/10/19 13:20:30]

You are right. Currently, we don't have a perfect solution to detect.
Pdfium always falls back to non-linearized path when it detects that
it can't handle a pdf file in the linearized way.  

[Tom Sepez, 2015/10/19 16:36:06]

Ok, looking at EmbedderTest::OpenDocument(), we could add an optional argument
called must_linearize, which would then return false instead of failing back at
line
https://code.google.com/p/chromium/codesearch#chromium/src/third_party/pdfium...

+}