The textClipPath member is set in CairoOutputDev::endString() and freed
in CairoOutputDev::endTextObject(). However, if endTextObject() is not
called for whatever reason, the path will just be leaked.

This adds code to the destructor to free this.

This fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=32326

Testing done:

$ wget -O testcase 'https://oss-fuzz.com/download?testcase_id=6659952325296128'
[...]
$ cmake .. -G Ninja -DENABLE_DCTDECODER=unmaintained -DENABLE_BOOST=OFF -DENABLE_LIBOPENJPEG=unmaintained && ninja
[...]
$ git describe
poppler-21.06.1-5-gb7c40059
$ valgrind --leak-check=full ./utils/pdftocairo testcase -png foo
[...]
==104075==
==104075== HEAP SUMMARY:
==104075==     in use at exit: 28,292 bytes in 55 blocks
==104075==   total heap usage: 6,114 allocs, 6,059 frees, 1,617,444 bytes allocated
==104075==
==104075== 24 bytes in 1 blocks are definitely lost in loss record 4 of 37
==104075==    at 0x483877F: malloc (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
==104075==    by 0x48AE748: ??? (in /usr/lib/x86_64-linux-gnu/libcairo.so.2.11600.0)
==104075==    by 0x118995: endString (CairoOutputDev.cc:1474)
==104075==    by 0x118995: CairoOutputDev::endString(GfxState*) (CairoOutputDev.cc:1412)
==104075==    by 0x4B97295: Gfx::doShowText(GooString const*) (Gfx.cc:4010)
==104075==    by 0x4B97CB4: Gfx::opShowSpaceText(Object*, int) (Gfx.cc:3793)
==104075==    by 0x4B8D866: Gfx::go(bool) (Gfx.cc:681)
==104075==    by 0x4B8DCFA: display (Gfx.cc:642)
==104075==    by 0x4B8DCFA: Gfx::display(Object*, bool) (Gfx.cc:622)
==104075==    by 0x4BE1A83: Page::displaySlice(OutputDev*, double, double, int, bool, bool, int, int, int, int, bool, bool (*)(void*), void*, bool (*)(Annot*, void*), void*, bool) (Page.cc:576)
==104075==    by 0x11317C: renderPage (pdftocairo.cc:669)
==104075==    by 0x11317C: main (pdftocairo.cc:1183)
==104075==
==104075== LEAK SUMMARY:
==104075==    definitely lost: 24 bytes in 1 blocks
==104075==    indirectly lost: 0 bytes in 0 blocks
==104075==      possibly lost: 0 bytes in 0 blocks
==104075==    still reachable: 28,268 bytes in 54 blocks
==104075==         suppressed: 0 bytes in 0 blocks
==104075== Reachable blocks (those to which a pointer was found) are not shown.
==104075== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==104075==
==104075== For lists of detected and suppressed errors, rerun with: -s
==104075== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
$ git checkout cairo-leak-textClipPath && git describe && ninja
Zu Branch 'cairo-leak-textClipPath' gewechselt
poppler-21.06.1-6-g8df6f8d2


$ valgrind --leak-check=full ./utils/pdftocairo testcase -png foo
[...]
==104263==
==104263== HEAP SUMMARY:
==104263==     in use at exit: 28,268 bytes in 54 blocks
==104263==   total heap usage: 6,114 allocs, 6,060 frees, 1,617,444 bytes allocated
==104263==
==104263== LEAK SUMMARY:
==104263==    definitely lost: 0 bytes in 0 blocks
==104263==    indirectly lost: 0 bytes in 0 blocks
==104263==      possibly lost: 0 bytes in 0 blocks
==104263==    still reachable: 28,268 bytes in 54 blocks
==104263==         suppressed: 0 bytes in 0 blocks
==104263== Reachable blocks (those to which a pointer was found) are not shown.
==104263== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==104263==
==104263== For lists of detected and suppressed errors, rerun with: -s
==104263== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

As you (might) see, before this commit, there is a "definitely lost"
leak of 24 bytes with this test case. After this commit, this leak is
gone.

Signed-off-by: Uli Schlachter <psychon@znc.in>

instead of lots of fields from it

Makes the signature simpler :)

find . \( -name "*.cpp" -or -name "*.h"  -or -name "*.c"  -or -name "*.cc" \) -exec clang-format -i {} \;

If you reached this file doing a git blame, please see README.contributors (instructions added 2 commits in the future to this one)

Marek Kasik authored Apr 06, 2018 and Albert Astals Cid committed May 12, 2020

Rendering of tiling pattern which has pattern matrix moving pattern cell
far away can fail on allocation of memory. This commit solves the issue by
modifying of cairo pattern matrix so that its offset is closer to the path
filled by the pattern.

Fixes #190

Fixes: poppler/poppler#178

No copyright, this is a mechanical change

Suggested by clang-tidy readability-non-const-parameter check

Check scaledWidth and scaledHeight for 0 at RescaleDrawImage::getSourceImage()
as is done on other places. Set the dimension to 1 if it is 0.

Fixes issue #737

Albert Astals Cid authored Jun 21, 2019 and Albert Astals Cid committed Jun 21, 2019

Fixes shadow warnings and in some cases makes for better C++ code not using this->

Albert Astals Cid authored Jun 21, 2019 and Albert Astals Cid committed Jun 21, 2019

reflects better what it is and fixes several shadow warnings

In ImageOutputDev it comes directly from malloc, and malloc guarantees
alignment for basic types, so we're good

In ArthurOutputDev it comes from QImage::bits that uses malloc
internally, so we're good

In cairo* it comes from cairo_image_surface_get_data that comes from
pixman_image_get_data that returns a uint32_t * so we're only going to
the original type alignment

Oliver Sander authored Nov 09, 2018 and Albert Astals Cid committed Nov 11, 2018

This is the name used by std::string.

to avoid using PS memory on cairo >= 1.5.10

Oliver Sander authored Oct 21, 2018 and Albert Astals Cid committed Oct 23, 2018

These are just non-standard names for bool, true, false, respectively.
Getting rid of these names saves on layer of mental redirection,
and enables proper syntax highlighting in editors.

GCC recommends not using them for a long time and its documentation says:
> These #pragmas have been superceded as of GCC 2.7.2 by COMDAT support
> and the “key method” heuristic mentioned in Vague Linkage.
> Using them can actually cause your program to grow due to
> unnecessary out-of-line copies of inline functions.

Also nobody seems to set USE_GCC_PRAGMAS and sometimes they were
guarded by just __GNUC__ which upsets Clang.

Marked some caches as mutable

Fixed an actual bug in page::text that was swapping the page cropbox in
each call

fixes oss-fuzz/8929

Albert Astals Cid authored Apr 06, 2018 and Albert Astals Cid committed Apr 06, 2018

People that use poppler core, beware of the signature change!

The quality is good enough and the performance is much better.

https://bugs.freedesktop.org/show_bug.cgi?id=103136

More to come, -Wshadow is interesting but if we enable it we still get too many warnings for it to be useful