vcl/source/filter/ipdf/pdfdocument.cxx | 6 ++++++
1 file changed, 6 insertions(+)
New commits:
commit 656bf70a0ce5476628292cca45c1cd24bf3450cf
Author: Justin Luth <jl...@mail.com>
AuthorDate: Mon Oct 7 19:57:37 2024 -0400
Commit: Miklos Vajna <vmik...@collabora.com>
CommitDate: Thu Oct 17 08:40:15 2024 +0200
PDFDocument::ReadWithPossibleFixup: clear before reading fixup
Ensure that a read of the pdfdocument
does not use info from failed previous reads.
The result in the problematic documents
was that GetCatalog failed because it was looking
for an m_aTrailerOffset[0] from the failed read,
and thus only empty pages were exported.
I failed to create a clean-room document,
and the single existing unit test that needs a fixup
has the trailer at the same GetPosition,
which replaced the m_aTrailerOffset
instead of creating a second entry.
Change-Id: I470b922323a2b2ebb6d4bbaa58c98f5cc492b750
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/174652
Reviewed-by: Miklos Vajna <vmik...@collabora.com>
Tested-by: Jenkins
Reviewed-by: Justin Luth <jl...@mail.com>
(cherry picked from commit cb7a4d814644bf225127a93bbd6a114b6f19ac52)
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/174955
Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoff...@gmail.com>
diff --git a/vcl/source/filter/ipdf/pdfdocument.cxx
b/vcl/source/filter/ipdf/pdfdocument.cxx
index b5fbc757da57..f63853f86835 100644
--- a/vcl/source/filter/ipdf/pdfdocument.cxx
+++ b/vcl/source/filter/ipdf/pdfdocument.cxx
@@ -1382,6 +1382,12 @@ bool PDFDocument::Read(SvStream& rStream)
rStream.Seek(0);
m_aEditBuffer.WriteStream(rStream);
+ // clear out key items that may have been filled with info from any
previous read attempt
+ m_aOffsetTrailers.clear();
+ m_aTrailerOffsets.clear();
+ m_pTrailer = nullptr;
+ m_pXRefStream = nullptr;
+
// Look up the offset of the xref table.
size_t nStartXRef = FindStartXRef(rStream);
SAL_INFO("vcl.filter", "PDFDocument::Read: nStartXRef is " << nStartXRef);
