GitLab

337585e3 unconditionnaly
overrides any potential user-defined CMAKE_C[XX]_FLAGS.
This for example breaks the GDAL oss-fuzz builds which build Poppler
from source (see https://github.com/OSGeo/gdal/blob/master/gdal/fuzzers/build.sh#L54)

So do the same as the GCC path where we save input CMAKE_C[XX]_FLAGS
and reinject them in custom CMAKE_C{XX}_FLAGS_{build_configuration}
The values are identical to GCC, execpt for the _DEBUG configuration
where we remove '-O2 -fno-reorder-blocks -fno-schedule-insns -fno-inline'
since clang does not support -fno-reorder-blocks and -fno-schedule-insns,
so it is likely better to disable any optimization for proper debugging.

This code needs a bit of refactoring to share most of it between gcc and
clang

And fixes for it in the code

This allows creating a PopplerDocument from data stored in a
GBytes.

This is better than poppler_document_new_from_data() since the
document will keep a reference to the data, instead of having
to make sure manually that the data stays around as long as the
document exists. Also poppler_document_new_from_data() is buggy
in that it uses int for the data size instead of gsize.

poppler/poppler#771

   for (const TypeMapEntry &entry : typeMap) {
is much easier to read than
   for (unsigned i = 0; i < sizeof(typeMap) / sizeof(typeMap[0]); i++) {

Only with the performance- checks enabled for now

This way we only const_cast in free()

The problem is that some of the times the pointers hold values
to const tables and some others hold values to dynamic memory.

Instead of const_casting the const tables when needed and holding a non
const pointer, we hold a const pointer and only const cast on free if
needed

This moves 14.4kB from .data to .rodata, and another 11.6kB from .data to
.data.rel.ro.

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

By initializing width and charCount earlier in Annot::layoutText,
there's two early return that need the values to be initialized and
we were initializing them later in the function anyway so just move the
block to the beginning of the function

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

I was tired of having two constructors for GfxDeviceNColorSpace, both
taking GooString ** but one copying those strings and the other keeping
them for itself.

Moved to std::vector<std::string> without much fallout in the rest of
the code

Fixes crash with bug-poppler85281.pdf

Doesn't seem to cause any speed regression and one ifdef less is code easier
to maintain

Otherwise if we construct a DecryptStream and next we call lookupChar it
would use un-initialized memory

Also, added Since: strings and completed missing API doc comments.

Clarify documentation about returned values
of recently added viewer preferences.

poppler/poppler#290