Previously when updating the header fields, SFML assumed the number of
samples written would be the number of samples per channel, which wasn't
the case. Therefore for stereo files the written file length was actually
twice the correct value. This fix uses the file size written as a basis,
no longer counting the samples written alltogether.
This fixes issue #1280.
Add default minimum target to 10.7.
The CMAKE_OSX_* variables need to be defined before the project or the
CMAKE_CXX_STANDARD gets ignored. They also need to be set with FORCE.
This grew out of my work creating an sfml port for macports, but should
be helpful for package maintainers of various distributions:
* add an SFML_USE_SYSTEM_DEPS option to ignore everything in extlibs/
except for headers/stb_image, and use the system versions
* install pkg-config files if a pkg-config program is found
and either lib/pkgconfig or libdata/pkgconfig exists under the
INSTALL_PREFIX, or the SFML_INSTALL_PKGCONFIG_FILES flag is set
explicitly
* install pkg-config files for static libs too, add the necessary
Requires.private and Libs.private entries to the .pc files to support
static linking
* on OS X, honor all INSTALL_NAME and RPATH related cmake variables and
only set the INSTALL_NAME_DIR to "@rpath" if none of them is set, this
preserves the default behavior of using @rpath but also allows
overriding by the usual cmake mechanisms
This should improve the signing process of Mac Applications.
This improves the frameworks of external dependencies used by SFML. To
patch sfml-*.framework, one would need to customised the
`MACOSX_FRAMEWORK_INFO_PLIST` cmake property and provide a custom
Info.plist file with CFBundleSupportedPlatforms property set. See
https://cmake.org/cmake/help/latest/prop_tgt/FRAMEWORK.html
This is however not required (probably) if one used dylibs instead.
Related to #1020 and #1036. Credits go to @kipbits.
Discussion thread: http://en.sfml-dev.org/forums/index.php?topic=20875.0
Basically, the issue with sf::Image::create was, that it would either
occupy space, it doesn't need, because std::vector::resize doesn't
actually shrink the vector, or reallocate in an inefficient way by
needlessly copying the whole old image over. Neither did it grant strong
exception safety guarantee because it changed the non-throwing members
(m_size) prior to doing the critical stuff (reallocating) which may
throw. Changing the order and using a temporary
(create-temporary-and-swap idiom; see http://www.gotw.ca/gotw/059.htm)
fixes both of these problems.
