tags 980426 +patch
thanks
Here's a patch that applies cleanly against upstream v1.17.3 after which the
resulting pikepdf builds and passes its test with qpdf 10.1. Please let me know
if this is sufficient. I put some comments at the top of the patch citing
original commits in case you want to do the DEP-3 thing.
--Jay
Patch created as follows:
cp /dev/null /tmp/a.patch
git show 7ac9b058104219b26747f3fc9761ac6b3c037402 >> /tmp/a.patch
git show 5394b7855eb87f637316432b2b597294adcb974e >> /tmp/a.patch
git show fe4b568ac88ba551458578613b013ac17edf5c4e -- tests/test_filters.py >> /tmp/a.patch
git show 7ca375cb6ff82011d716da1614ef5721a97f91b0 >> /tmp/a.patch
commit 7ac9b058104219b26747f3fc9761ac6b3c037402
Author: James R. Barlow <ja...@purplerock.ca>
Date: Mon Jan 4 20:21:51 2021 -0800
Fix externalize_inline_images for qpdf 10.1.0
diff --git a/src/qpdf/page.cpp b/src/qpdf/page.cpp
index 64b641d..6f70d20 100644
--- a/src/qpdf/page.cpp
+++ b/src/qpdf/page.cpp
@@ -106,7 +106,10 @@ 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.
commit 5394b7855eb87f637316432b2b597294adcb974e
Author: James R. Barlow <ja...@purplerock.ca>
Date: Tue Aug 18 00:36:46 2020 -0700
Fix potential undefined behavior in memoryview
Apply patch based on pybind11 PR2223 to change how Python
memory views on C++ buffers are declared - avoids a possible use-
after-free in Python's code.
This change is not in a released pybind11 version yet so we are
pulling it forward.
diff --git a/src/qpdf/pipeline.cpp b/src/qpdf/pipeline.cpp
index 1856987..8a4fe79 100644
--- a/src/qpdf/pipeline.cpp
+++ b/src/qpdf/pipeline.cpp
@@ -22,6 +22,7 @@
#include "pikepdf.h"
#include "pipeline.h"
+#include "utils.h"
void Pl_PythonOutput::write(unsigned char *buf, size_t len)
@@ -29,8 +30,7 @@ void Pl_PythonOutput::write(unsigned char *buf, size_t len)
py::gil_scoped_acquire gil;
ssize_t so_far = 0;
while (len > 0) {
- py::buffer_info buffer(buf, len);
- py::memoryview view_buffer(buffer);
+ py::memoryview view_buffer = memoryview_from_memory(buf, len);
py::object result = this->stream.attr("write")(view_buffer);
try {
so_far = result.cast<ssize_t>();
diff --git a/src/qpdf/qpdf_inputsource.h b/src/qpdf/qpdf_inputsource.h
index 4a96f6b..8ec6a55 100644
--- a/src/qpdf/qpdf_inputsource.h
+++ b/src/qpdf/qpdf_inputsource.h
@@ -82,8 +82,7 @@ public:
{
py::gil_scoped_acquire gil;
- py::buffer_info buffer_info(buffer, length);
- py::memoryview view_buffer_info(buffer_info);
+ py::memoryview view_buffer_info = memoryview_from_memory(buffer, length);
this->last_offset = this->tell();
py::object result = this->stream.attr("readinto")(view_buffer_info);
diff --git a/src/qpdf/utils.cpp b/src/qpdf/utils.cpp
index bf189f6..36769ca 100644
--- a/src/qpdf/utils.cpp
+++ b/src/qpdf/utils.cpp
@@ -46,3 +46,20 @@
}
#endif
+
+// Copied from pybind11 master branch (pre-2.6), can remove when we require
+// pybind11 v2.6 and replace with py::memoryview::from_memory
+py::memoryview memoryview_from_memory(void *mem, ssize_t size, bool readonly)
+{
+ PyObject* ptr = PyMemoryView_FromMemory(
+ reinterpret_cast<char*>(mem), size,
+ (readonly) ? PyBUF_READ : PyBUF_WRITE);
+ if (!ptr)
+ py::pybind11_fail("Could not allocate memoryview object!");
+ return py::reinterpret_steal<py::memoryview>(ptr);
+}
+
+py::memoryview memoryview_from_memory(const void *mem, ssize_t size)
+{
+ return memoryview_from_memory(const_cast<void*>(mem), size, true);
+}
diff --git a/src/qpdf/utils.h b/src/qpdf/utils.h
index 9f6a411..f26c5bd 100644
--- a/src/qpdf/utils.h
+++ b/src/qpdf/utils.h
@@ -17,3 +17,6 @@ inline bool str_startswith(T haystack, S needle)
{
return std::string(haystack).rfind(needle, 0) == 0;
}
+
+py::memoryview memoryview_from_memory(void *mem, ssize_t size, bool readonly = false);
+py::memoryview memoryview_from_memory(const void *mem, ssize_t size);
commit fe4b568ac88ba551458578613b013ac17edf5c4e
Author: James R. Barlow <ja...@purplerock.ca>
Date: Wed Jan 6 00:22:16 2021 -0800
libqpdf 10.1.0 raises different exception
The different errors are acceptable to us; actually they are more
correct than the original behavior.
diff --git a/tests/test_filters.py b/tests/test_filters.py
index 90c0ca5..36d2a4c 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]
commit 7ca375cb6ff82011d716da1614ef5721a97f91b0
Author: James R. Barlow <ja...@purplerock.ca>
Date: Wed Jan 6 03:39:50 2021 -0800
Fix test_tokenfilter_is_abstract
diff --git a/tests/test_filters.py b/tests/test_filters.py
index 36d2a4c..f6421c9 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]
