From dd1083b47c029d2540aceb246bf80f549781f62f Mon Sep 17 00:00:00 2001
From: Henrique Nakashima <hnakashima@chromium.org>
Date: Fri, 6 Jul 2018 17:28:46 +0000
Subject: Revert "Avoid duplicate data buffering in
 CPDF_SyntaxParser::ReadStream()."

This reverts commit 77f15f7883638a4ced131d74c053af10a5970ce9.

Reason for revert: Causes crbug.com/860210
Bug: chromium:860210

Original change's description:
> Avoid duplicate data buffering in CPDF_SyntaxParser::ReadStream().
>
> Allow sub-streams created from an IFX_SeekableReadStream to provide
> stream data without copying memory.
> The data will only reside in the top-level stream.
>
> For example:
> For file
> http://www.major-landrover.ru/upload/attachments/f/9/f96aab07dab04ae89c8a509ec1ef2b31.pdf
> (18 Mb)
>
> The memory usage is reduced by ~13 Mb.
>
> Change-Id: I2595c014d0fbe1fdd181cc04965cfd7d901c2d88
> Reviewed-on: https://pdfium-review.googlesource.com/35930
> Commit-Queue: Art Snake <art-snake@yandex-team.ru>
> Reviewed-by: dsinclair <dsinclair@chromium.org>

TBR=tsepez@chromium.org,dsinclair@chromium.org,art-snake@yandex-team.ru

# Not skipping CQ checks because original CL landed > 1 day ago.

Change-Id: I947fca17052765935a952a4f25ca48f6599c4af9
Reviewed-on: https://pdfium-review.googlesource.com/37210
Reviewed-by: Henrique Nakashima <hnakashima@chromium.org>
Commit-Queue: Henrique Nakashima <hnakashima@chromium.org>
---
 core/fpdfapi/parser/cpdf_syntax_parser.cpp | 67 ++++--------------------------
 1 file changed, 9 insertions(+), 58 deletions(-)

(limited to 'core/fpdfapi/parser')

diff --git a/core/fpdfapi/parser/cpdf_syntax_parser.cpp b/core/fpdfapi/parser/cpdf_syntax_parser.cpp
index f8f36aee11..ec5ac70a29 100644
--- a/core/fpdfapi/parser/cpdf_syntax_parser.cpp
+++ b/core/fpdfapi/parser/cpdf_syntax_parser.cpp
@@ -34,37 +34,6 @@ namespace {
 
 enum class ReadStatus { Normal, Backslash, Octal, FinishOctal, CarriageReturn };
 
-class ReadableSubStream : public IFX_SeekableReadStream {
- public:
-  ReadableSubStream(const RetainPtr<IFX_SeekableReadStream>& pFileRead,
-                    FX_FILESIZE part_offset,
-                    FX_FILESIZE part_size)
-      : m_pFileRead(pFileRead),
-        m_PartOffset(part_offset),
-        m_PartSize(part_size) {}
-
-  ~ReadableSubStream() override = default;
-
-  // IFX_SeekableReadStream overrides:
-  bool ReadBlock(void* buffer, FX_FILESIZE offset, size_t size) override {
-    FX_SAFE_FILESIZE safe_end = offset;
-    safe_end += size;
-    // Check that requested range is valid, to prevent calling of ReadBlock
-    // of original m_pFileRead with incorrect params.
-    if (!safe_end.IsValid() || safe_end.ValueOrDie() > m_PartSize)
-      return false;
-
-    return m_pFileRead->ReadBlock(buffer, m_PartOffset + offset, size);
-  }
-
-  FX_FILESIZE GetSize() override { return m_PartSize; }
-
- private:
-  RetainPtr<IFX_SeekableReadStream> m_pFileRead;
-  FX_FILESIZE m_PartOffset;
-  FX_FILESIZE m_PartSize;
-};
-
 }  // namespace
 
 // static
@@ -608,6 +577,7 @@ std::unique_ptr<CPDF_Stream> CPDF_SyntaxParser::ReadStream(
   ToNextLine();
   const FX_FILESIZE streamStartPos = GetPos();
 
+  std::unique_ptr<uint8_t, FxFreeDeleter> pData;
   if (len > 0) {
     FX_SAFE_FILESIZE pos = GetPos();
     pos += len;
@@ -615,18 +585,12 @@ std::unique_ptr<CPDF_Stream> CPDF_SyntaxParser::ReadStream(
       len = -1;
   }
 
-  RetainPtr<IFX_SeekableReadStream> data;
   if (len > 0) {
-    // Check data availability first to allow the Validator to request data
+    pData.reset(FX_Alloc(uint8_t, len));
+    // We should try read data first to allow the Validator to request data
     // smoothly, without jumps.
-    if (!GetValidator()->CheckDataRangeAndRequestIfUnavailable(
-            m_HeaderOffset + GetPos(), len)) {
+    if (!ReadBlock(pData.get(), len))
       return nullptr;
-    }
-
-    data = pdfium::MakeRetain<ReadableSubStream>(
-        GetValidator(), m_HeaderOffset + GetPos(), len);
-    SetPos(GetPos() + len);
   }
 
   const ByteStringView kEndStreamStr("endstream");
@@ -647,7 +611,7 @@ std::unique_ptr<CPDF_Stream> CPDF_SyntaxParser::ReadStream(
     // specified length, it signals the end of stream.
     if (memcmp(m_WordBuffer, kEndStreamStr.raw_str(),
                kEndStreamStr.GetLength()) != 0) {
-      data.Reset();
+      pData.reset();
       len = -1;
       SetPos(streamStartPos);
     }
@@ -664,27 +628,14 @@ std::unique_ptr<CPDF_Stream> CPDF_SyntaxParser::ReadStream(
     ASSERT(len >= 0);
     if (len > 0) {
       SetPos(streamStartPos);
-      // Check data availability first to allow the Validator to request data
-      // smoothly, without jumps.
-      if (!GetValidator()->CheckDataRangeAndRequestIfUnavailable(
-              m_HeaderOffset + GetPos(), len)) {
+      pData.reset(FX_Alloc(uint8_t, len));
+      if (!ReadBlock(pData.get(), len))
         return nullptr;
-      }
-
-      data = pdfium::MakeRetain<ReadableSubStream>(
-          GetValidator(), m_HeaderOffset + GetPos(), len);
-      SetPos(GetPos() + len);
     }
   }
 
-  auto pStream = pdfium::MakeUnique<CPDF_Stream>();
-  if (data) {
-    pStream->InitStreamFromFile(data, std::move(pDict));
-  } else {
-    DCHECK(!len);
-    // Empty stream
-    pStream->InitStream(nullptr, 0, std::move(pDict));
-  }
+  auto pStream =
+      pdfium::MakeUnique<CPDF_Stream>(std::move(pData), len, std::move(pDict));
   const FX_FILESIZE end_stream_offset = GetPos();
   memset(m_WordBuffer, 0, kEndObjStr.GetLength() + 1);
   GetNextWordInternal(nullptr);
-- 
cgit v1.2.3