Backport fix for XXE vulnerability.

0001-Relax-some-test-requirements.patch

@@ -1,7 +1,7 @@
 From 1ac056bdd30475566dcf630a55aae7cdd9eb81eb Mon Sep 17 00:00:00 2001
 From: Elliott Sales de Andrade <quantum.analyst@gmail.com>
 Date: Wed, 19 Aug 2020 22:44:11 -0400
-Subject: [PATCH] Relax some test requirements.
+Subject: [PATCH 1/5] Relax some test requirements.
 
 Signed-off-by: Elliott Sales de Andrade <quantum.analyst@gmail.com>
 ---

0002-Fix-externalize_inline_images-for-qpdf-10.1.0.patch

@@ -0,0 +1,29 @@
+From e0de53084fe005fafad423a66b261cc45cfaaeb0 Mon Sep 17 00:00:00 2001
+From: "James R. Barlow" <james@purplerock.ca>
+Date: Mon, 4 Jan 2021 20:21:51 -0800
+Subject: [PATCH 2/5] Fix externalize_inline_images for qpdf 10.1.0
+
+Signed-off-by: Elliott Sales de Andrade <quantum.analyst@gmail.com>
+---
+ src/qpdf/page.cpp | 5 ++++-
+ 1 file changed, 4 insertions(+), 1 deletion(-)
+
+diff --git a/src/qpdf/page.cpp b/src/qpdf/page.cpp
+index 72748a9..3f6e26e 100644
+--- a/src/qpdf/page.cpp
++++ b/src/qpdf/page.cpp
+@@ -79,7 +79,10 @@ void init_page(py::module& m)
+         .def("_get_mediabox", &QPDFPageObjectHelper::getMediaBox)
+         .def("_get_cropbox", &QPDFPageObjectHelper::getCropBox)
+         .def("_get_trimbox", &QPDFPageObjectHelper::getTrimBox)
+-        .def("externalize_inline_images", &QPDFPageObjectHelper::externalizeInlineImages,
++        .def("externalize_inline_images",
++            [](QPDFPageObjectHelper &poh, size_t min_size = 0) {
++                return poh.externalizeInlineImages(min_size);
++            },
+             py::arg("min_size") = 0,
+             R"~~~(
+                 Convert inlines image to normal (external) images.
+-- 
+2.29.2
+

0003-libqpdf-10.1.0-raises-different-exception.patch

@@ -0,0 +1,46 @@
+From d9554019340177a2fb0d033fc6ffcf45f025cb2f Mon Sep 17 00:00:00 2001
+From: "James R. Barlow" <james@purplerock.ca>
+Date: Wed, 6 Jan 2021 00:22:16 -0800
+Subject: [PATCH 3/5] libqpdf 10.1.0 raises different exception
+
+The different errors are acceptable to us; actually they are more
+correct than the original behavior.
+
+Signed-off-by: Elliott Sales de Andrade <quantum.analyst@gmail.com>
+---
+ tests/test_filters.py | 13 ++++++-------
+ 1 file changed, 6 insertions(+), 7 deletions(-)
+
+diff --git a/tests/test_filters.py b/tests/test_filters.py
+index b394795..33124bf 100644
+--- a/tests/test_filters.py
++++ b/tests/test_filters.py
+@@ -74,19 +74,18 @@ class FilterInvalid(pikepdf.TokenFilter):
+ 
+ def test_invalid_handle_token(pal):
+     page = pikepdf.Page(pal.pages[0])
+-    with pytest.raises(pikepdf.PdfError):
+-        result = page.get_filtered_contents(FilterInvalid())
++    with pytest.raises((TypeError, pikepdf.PdfError)):
++        page.get_filtered_contents(FilterInvalid())
+ 
+ 
+ def test_invalid_tokenfilter(pal):
+     page = pikepdf.Page(pal.pages[0])
+     with pytest.raises(TypeError):
+-        result = page.get_filtered_contents(list())
++        page.get_filtered_contents(list())
+ 
+ 
+ def test_tokenfilter_is_abstract(pal):
+     page = pikepdf.Page(pal.pages[0])
+-    try:
+-        result = page.get_filtered_contents(pikepdf.TokenFilter())
+-    except pikepdf.PdfError:
+-        assert 'Tried to call pure virtual' in pal.get_warnings()[0]
++    with pytest.raises((RuntimeError, pikepdf.PdfError)):
++        page.get_filtered_contents(pikepdf.TokenFilter())
++    assert 'Tried to call pure virtual' in pal.get_warnings()[0]
+-- 
+2.29.2
+

0004-Fix-test_tokenfilter_is_abstract.patch

@@ -0,0 +1,22 @@
+From 4521a84f5892a38e380ed53f9b4cbcfd7647db45 Mon Sep 17 00:00:00 2001
+From: "James R. Barlow" <james@purplerock.ca>
+Date: Wed, 6 Jan 2021 03:39:50 -0800
+Subject: [PATCH 4/5] Fix test_tokenfilter_is_abstract
+
+Signed-off-by: Elliott Sales de Andrade <quantum.analyst@gmail.com>
+---
+ tests/test_filters.py | 1 -
+ 1 file changed, 1 deletion(-)
+
+diff --git a/tests/test_filters.py b/tests/test_filters.py
+index 33124bf..5604961 100644
+--- a/tests/test_filters.py
++++ b/tests/test_filters.py
+@@ -88,4 +88,3 @@ def test_tokenfilter_is_abstract(pal):
+     page = pikepdf.Page(pal.pages[0])
+     with pytest.raises((RuntimeError, pikepdf.PdfError)):
+         page.get_filtered_contents(pikepdf.TokenFilter())
+-    assert 'Tried to call pure virtual' in pal.get_warnings()[0]
+-- 
+2.29.2
+

0005-Fix-XXE-vulnerability-in-XMP-metadata-parsing.patch

@@ -0,0 +1,124 @@
+From a1babed37c2ffc20104d5cbd840816d890f5bcfc Mon Sep 17 00:00:00 2001
+From: "James R. Barlow" <james@purplerock.ca>
+Date: Sat, 27 Mar 2021 00:43:21 -0700
+Subject: [PATCH 5/5] Fix XXE vulnerability in XMP metadata parsing
+
+For details:
+https://portswigger.net/web-security/xxe
+
+Reported by: Eric Therond eric.therond@sonarsource.com) of Sonarsource (https://www.sonarsource.com/)
+
+Signed-off-by: Elliott Sales de Andrade <quantum.analyst@gmail.com>
+---
+ src/pikepdf/_xml.py            | 30 ++++++++++++++++++++++++++++++
+ src/pikepdf/models/metadata.py | 10 +++++-----
+ tests/test_metadata.py         | 24 ++++++++++++++++++++++++
+ 3 files changed, 59 insertions(+), 5 deletions(-)
+ create mode 100644 src/pikepdf/_xml.py
+
+diff --git a/src/pikepdf/_xml.py b/src/pikepdf/_xml.py
+new file mode 100644
+index 0000000..f0e1c38
+--- /dev/null
++++ b/src/pikepdf/_xml.py
+@@ -0,0 +1,30 @@
++# This Source Code Form is subject to the terms of the Mozilla Public
++# License, v. 2.0. If a copy of the MPL was not distributed with this
++# file, You can obtain one at http://mozilla.org/MPL/2.0/.
++#
++# Copyright (C) 2021, James R. Barlow (https://github.com/jbarlow83/)
++
++
++from typing import IO, Any, AnyStr, Union
++
++from lxml.etree import XMLParser as _UnsafeXMLParser
++from lxml.etree import parse as _parse
++
++
++class _XMLParser(_UnsafeXMLParser):
++    def __init__(self, *args, **kwargs):
++        # Prevent XXE attacks
++        # https://rules.sonarsource.com/python/type/Vulnerability/RSPEC-2755
++        kwargs['resolve_entities'] = False
++        kwargs['no_network'] = True
++        super().__init__(*args, **kwargs)
++
++
++def parse_xml(source: Union[AnyStr, IO[Any]], recover: bool = False):
++    """Wrapper around lxml's parse to provide protection against XXE attacks."""
++
++    parser = _XMLParser(recover=recover, remove_pis=False)
++    return _parse(source, parser=parser)
++
++
++__all__ = ['parse_xml']
+diff --git a/src/pikepdf/models/metadata.py b/src/pikepdf/models/metadata.py
+index f4f3860..8b44b60 100644
+--- a/src/pikepdf/models/metadata.py
++++ b/src/pikepdf/models/metadata.py
+@@ -15,10 +15,11 @@ from io import BytesIO
+ from warnings import warn
+ 
+ from lxml import etree
+-from lxml.etree import QName, XMLParser, XMLSyntaxError, parse
++from lxml.etree import QName, XMLSyntaxError
+ 
+ from .. import Name, Stream, String
+ from .. import __version__ as pikepdf_version
++from .. import _xml
+ 
+ XMP_NS_DC = "http://purl.org/dc/elements/1.1/"
+ XMP_NS_PDF = "http://ns.adobe.com/pdf/1.3/"
+@@ -350,14 +351,13 @@ class PdfMetadata(MutableMapping):
+             data = XMP_EMPTY  # on some platforms lxml chokes on empty documents
+ 
+         def basic_parser(xml):
+-            return parse(BytesIO(xml))
++            return _xml.parse_xml(BytesIO(xml))
+ 
+         def strip_illegal_bytes_parser(xml):
+-            return parse(BytesIO(re_xml_illegal_bytes.sub(b'', xml)))
++            return _xml.parse_xml(BytesIO(re_xml_illegal_bytes.sub(b'', xml)))
+ 
+         def recovery_parser(xml):
+-            parser = XMLParser(recover=True)
+-            return parse(BytesIO(xml), parser)
++            return _xml.parse_xml(BytesIO(xml), recover=True)
+ 
+         def replace_with_empty_xmp(_xml=None):
+             log.warning("Error occurred parsing XMP, replacing with empty XMP.")
+diff --git a/tests/test_metadata.py b/tests/test_metadata.py
+index cf14b21..d968385 100644
+--- a/tests/test_metadata.py
++++ b/tests/test_metadata.py
+@@ -596,3 +596,27 @@ def test_issue_135_title_rdf_bag(trivial):
+         xmp['dc:title'] = {'Title 1', 'Title 2'}
+     with trivial.open_metadata(update_docinfo=False) as xmp:
+         assert b'Title 1; Title 2</rdf:li></rdf:Alt></dc:title>' in xmp._get_xml_bytes()
++
++
++def test_xxe(trivial, outdir):
++    secret = outdir / 'secret.txt'
++    secret.write_text("This is a secret")
++    trivial.Root.Metadata = Stream(
++        trivial,
++        b"""\
++<?xpacket begin='\xef\xbb\xbf' id='W5M0MpCehiHzreSzNTczkc9d'?>
++<!DOCTYPE rdf:RDF [<!ENTITY xxe SYSTEM "file://%s">]>
++<x:xmpmeta xmlns:x='adobe:ns:meta/' x:xmptk='Image'>
++<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
++<note>
++<to>&xxe;</to>
++<from>xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx</from>
++</note>
++</rdf:RDF>
++</x:xmpmeta>
++<?xpacket end='w'?>
++    """
++        % os.fsencode(secret),
++    )
++    with trivial.open_metadata() as m:
++        assert 'This is a secret' not in str(m)
+-- 
+2.29.2
+

python-pikepdf.spec

@@ -2,13 +2,19 @@
 
 Name:           python-%{srcname}
 Version:        1.19.4
-Release:        1%{?dist}
+Release:        2%{?dist}
 Summary:        Read and write PDFs with Python, powered by qpdf
 
 License:        MPLv2.0
 URL:            https://github.com/pikepdf/pikepdf
 Source0:        %pypi_source
 Patch0001:      0001-Relax-some-test-requirements.patch
+# Backport patches to work with qpdf 10.1
+Patch0002:      0002-Fix-externalize_inline_images-for-qpdf-10.1.0.patch
+Patch0003:      0003-libqpdf-10.1.0-raises-different-exception.patch
+Patch0004:      0004-Fix-test_tokenfilter_is_abstract.patch
+# Backport XXE security fix.
+Patch0005:      0005-Fix-XXE-vulnerability-in-XMP-metadata-parsing.patch
 
 BuildRequires:  gcc-c++
 BuildRequires:  qpdf-devel >= 8.4.2
@@ -101,6 +107,10 @@ rm -rf html/.{doctrees,buildinfo}
 
 
 %changelog
+* Thu Apr 01 2021 Elliott Sales de Andrade <quantum.analyst@gmail.com> - 1.19.4-2
+- Backport fix for qpdf 10.1.0
+- Backport fix for XXE vulnerability (#1945365)
+
 * Wed Dec 23 2020 Elliott Sales de Andrade <quantum.analyst@gmail.com> - 1.19.4-1
 - Update to latest version