Rework of Fixing metadata not read from linearized file.

Move receiving "Info" dictionary form API implementation to CPDF_Document.
Also added test.

Bug: pdfium:664

Change-Id: I273980750fbdd4d20711f651245780fc9ba02789
Reviewed-on: https://pdfium-review.googlesource.com/35490
Commit-Queue: Art Snake <art-snake@yandex-team.ru>
Reviewed-by: Henrique Nakashima <hnakashima@chromium.org>

2 files changed, 108 insertions, 1 deletions

diff --git a/fpdfsdk/fpdf_dataavail_embeddertest.cpp b/fpdfsdk/fpdf_dataavail_embeddertest.cpp
index 2084153852..ba7726613e 100644
--- a/fpdfsdk/fpdf_dataavail_embeddertest.cpp
+++ b/fpdfsdk/fpdf_dataavail_embeddertest.cpp
@@ -9,6 +9,8 @@
 #include <utility>
 #include <vector>
 
+#include "core/fxcrt/bytestring.h"
+#include "core/fxcrt/widestring.h"
 #include "public/fpdfview.h"
 #include "testing/embedder_test.h"
 #include "testing/gtest/include/gtest/gtest.h"
@@ -87,6 +89,9 @@ class TestAsyncLoader : public FX_DOWNLOADHINTS, FX_FILEAVAIL {
     ClearRequestedSegments();
   }
 
+  char* file_contents() { return file_contents_.get(); }
+  size_t file_length() const { return file_length_; }
+
  private:
   void SetDataAvailable(size_t start, size_t size) {
     available_ranges_.Union(RangeSet::Range(start, start + size));
@@ -271,3 +276,105 @@ TEST_F(FPDFDataAvailEmbeddertest, LoadSecondPageIfLinearizedWithHints) {
   EXPECT_TRUE(page);
   FPDF_ClosePage(page);
 }
+
+TEST_F(FPDFDataAvailEmbeddertest, LoadInfoAfterReceivingWholeDocument) {
+  TestAsyncLoader loader("linearized.pdf");
+  loader.set_is_new_data_available(false);
+  avail_ = FPDFAvail_Create(loader.file_avail(), loader.file_access());
+  while (PDF_DATA_AVAIL != FPDFAvail_IsDocAvail(avail_, loader.hints())) {
+    loader.FlushRequestedData();
+  }
+
+  document_ = FPDFAvail_GetDocument(avail_, nullptr);
+  ASSERT_TRUE(document_);
+
+  // The "info" dictionary should still be unavailable.
+  EXPECT_FALSE(FPDF_GetMetaText(document_, "CreationDate", nullptr, 0));
+
+  // Simulate receiving whole file.
+  loader.set_is_new_data_available(true);
+  // Load second page, to parse additional crossref sections.
+  EXPECT_EQ(PDF_DATA_AVAIL, FPDFAvail_IsPageAvail(avail_, 1, loader.hints()));
+
+  EXPECT_TRUE(FPDF_GetMetaText(document_, "CreationDate", nullptr, 0));
+}
+
+TEST_F(FPDFDataAvailEmbeddertest, LoadInfoAfterReceivingFirstPage) {
+  TestAsyncLoader loader("linearized.pdf");
+  // Map "Info" to an object within the first section without breaking
+  // linearization.
+  ByteString data(loader.file_contents(), loader.file_length());
+  Optional<size_t> index = data.Find("/Info 27 0 R");
+  ASSERT_TRUE(index);
+  memcpy(loader.file_contents() + *index, "/Info 29 0 R", 12);
+
+  loader.set_is_new_data_available(false);
+  avail_ = FPDFAvail_Create(loader.file_avail(), loader.file_access());
+  while (PDF_DATA_AVAIL != FPDFAvail_IsDocAvail(avail_, loader.hints())) {
+    loader.FlushRequestedData();
+  }
+
+  document_ = FPDFAvail_GetDocument(avail_, nullptr);
+  ASSERT_TRUE(document_);
+
+  // The "Info" dictionary should be available for the linearized document, if
+  // it is located in the first page section.
+  // Info was remapped to a dictionary with Type "Catalog"
+  unsigned short buffer[100] = {0};
+  EXPECT_TRUE(FPDF_GetMetaText(document_, "Type", buffer, sizeof(buffer)));
+  constexpr wchar_t kExpectedValue[] = L"Catalog";
+  EXPECT_EQ(WideString(kExpectedValue),
+            WideString::FromUTF16LE(buffer, FXSYS_len(kExpectedValue)));
+}
+
+TEST_F(FPDFDataAvailEmbeddertest, TryLoadInvalidInfo) {
+  TestAsyncLoader loader("linearized.pdf");
+  // Map "Info" to an invalid object without breaking linearization.
+  ByteString data(loader.file_contents(), loader.file_length());
+  Optional<size_t> index = data.Find("/Info 27 0 R");
+  ASSERT_TRUE(index);
+  memcpy(loader.file_contents() + *index, "/Info 99 0 R", 12);
+
+  loader.set_is_new_data_available(false);
+  avail_ = FPDFAvail_Create(loader.file_avail(), loader.file_access());
+  while (PDF_DATA_AVAIL != FPDFAvail_IsDocAvail(avail_, loader.hints())) {
+    loader.FlushRequestedData();
+  }
+
+  document_ = FPDFAvail_GetDocument(avail_, nullptr);
+  ASSERT_TRUE(document_);
+
+  // Set all data available.
+  loader.set_is_new_data_available(true);
+  // Check second page, to load additional crossrefs.
+  ASSERT_EQ(PDF_DATA_AVAIL, FPDFAvail_IsPageAvail(avail_, 0, loader.hints()));
+
+  // Test that api is robust enough to handle the bad case.
+  EXPECT_FALSE(FPDF_GetMetaText(document_, "Type", nullptr, 0));
+}
+
+TEST_F(FPDFDataAvailEmbeddertest, TryLoadNonExistsInfo) {
+  TestAsyncLoader loader("linearized.pdf");
+  // Break the "Info" parameter without breaking linearization.
+  ByteString data(loader.file_contents(), loader.file_length());
+  Optional<size_t> index = data.Find("/Info 27 0 R");
+  ASSERT_TRUE(index);
+  memcpy(loader.file_contents() + *index, "/I_fo 27 0 R", 12);
+
+  loader.set_is_new_data_available(false);
+  avail_ = FPDFAvail_Create(loader.file_avail(), loader.file_access());
+  while (PDF_DATA_AVAIL != FPDFAvail_IsDocAvail(avail_, loader.hints())) {
+    loader.FlushRequestedData();
+  }
+
+  document_ = FPDFAvail_GetDocument(avail_, nullptr);
+  ASSERT_TRUE(document_);
+
+  // Set all data available.
+  loader.set_is_new_data_available(true);
+  // Check second page, to load additional crossrefs.
+  ASSERT_EQ(PDF_DATA_AVAIL, FPDFAvail_IsPageAvail(avail_, 0, loader.hints()));
+
+  // Test that api is robust enough to handle the bad case.
+  EXPECT_FALSE(FPDF_GetMetaText(document_, "Type", nullptr, 0));
+}
diff --git a/fpdfsdk/fpdf_doc.cpp b/fpdfsdk/fpdf_doc.cpp
index 465b11302b..7a2bbc4c2e 100644
--- a/fpdfsdk/fpdf_doc.cpp
+++ b/fpdfsdk/fpdf_doc.cpp
@@ -400,7 +400,7 @@ FPDF_EXPORT unsigned long FPDF_CALLCONV FPDF_GetMetaText(FPDF_DOCUMENT document,
   CPDF_Document* pDoc = CPDFDocumentFromFPDFDocument(document);
   if (!pDoc)
     return 0;
-  pDoc->LoadDocumentInfo();
+
   const CPDF_Dictionary* pInfo = pDoc->GetInfo();
   if (!pInfo)
     return 0;