Make CPDF_Document cache CPDF_Pages

We cache pages not by page number, which can bounce around as pages
are inserted or removed, but by page dictionary's object number.

Since the page may be created under one function and used under
another, we can't take the shortcut of not instantiating a render
cache nor not parsing the page.

Change-Id: I9a325cda8b3141153544ac53e78a51a44e6b411a
Reviewed-on: https://pdfium-review.googlesource.com/32830
Commit-Queue: Tom Sepez <tsepez@chromium.org>
Reviewed-by: dsinclair <dsinclair@chromium.org>

1 files changed, 1 insertions, 3 deletions

diff --git a/fpdfsdk/fpdf_flatten.cpp b/fpdfsdk/fpdf_flatten.cpp
index 720fa9ff55..2d81f1230e 100644
--- a/fpdfsdk/fpdf_flatten.cpp
+++ b/fpdfsdk/fpdf_flatten.cpp
@@ -47,9 +47,7 @@ bool IsValidRect(const CFX_FloatRect& rect, const CFX_FloatRect& rcPage) {
 void GetContentsRect(CPDF_Document* pDoc,
                      CPDF_Dictionary* pDict,
                      std::vector<CFX_FloatRect>* pRectArray) {
-  auto pPDFPage = pdfium::MakeRetain<CPDF_Page>(pDoc, pDict, false);
-  pPDFPage->ParseContent();
-
+  RetainPtr<CPDF_Page> pPDFPage = pDoc->GetOrCreatePDFPage(pDict);
   for (const auto& pPageObject : *pPDFPage->GetPageObjectList()) {
     CFX_FloatRect rc;
     rc.left = pPageObject->m_Left;