tree 6f51f252f7f11f2cf7c450c238a713dd64806f7e
parent 93797296015034f67cbb4464b947544ce9e8cd25
author Lei Zhang <thestig@chromium.org> 1663613233 +0000
committer Pdfium LUCI CQ <pdfium-scoped@luci-project-accounts.iam.gserviceaccount.com> 1663613233 +0000

M107: Copy data in CPDF_SyntaxParser::ReadStream().

Previously, https://pdfium-review.googlesource.com/97491 changed
CPDF_ObjectStream to avoid copying. CPDF_ObjectStream now holds on to
both the data object, and a data stream that references the data object.
The data stream is expected to be destroyed before the data object, to
avoid dangling pointers.

CPDF_ObjectStream then allows CPDF_SyntaxParser an only
CPDF_SyntaxParser to use the data stream. Sometimes CPDF_SyntaxParser
creates new objects that references the data stream, Thereby violating
the object lifetime requirement that CPDF_ObjectStream was expecting.

Fix this issue by doing a data copy in CPDF_SyntaxParser::ReadStream()
only when needed. This copies less data than the previous strategy where
CPDF_ObjectStream does the copying because:
1) CPDF_SyntaxParser only copies in certain cases.
2) CPDF_SyntaxParser copies a substream, which is oftentimes smaller
   than the stream data that CPDF_ObjectStream copies.

This also prevents similar issues from happening when other objects
instantiates CPDF_SyntaxParser.

For the M107 merge, switch to using DataVector, as FixedUninitDataVector
did not make the M107 cut.

Bug: chromium:1364735
Change-Id: Iacf7c3c1ae1a3f97882e52fbacfc68a7ed9d146a
Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/98091
Reviewed-by: Tom Sepez <tsepez@chromium.org>
Commit-Queue: Lei Zhang <thestig@chromium.org>
(cherry picked from commit a52ea7f7b8a6ad60e553ccea8c01ec17bde5825b)
Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/98152