| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The GdkPixdata API is built as part of the GdkPixbuf shared library,
but:
- it has its own namespace
- it has its own header file
- it's mostly meant for C applications
- it is deprecated
Consumers of the introspection data cannot access this API by using the
GdkPixbuf-2.0 GIR, as the namespace and included headers are different.
Instead of kludging the API, let's move the GdkPixdata introspection
out of the GdkPixbuf-2.0 GIR and into its own module.
This is an introspection ABI break, but there aren't many consumers of
GdkPixdata out there, and it's preferable to straight including the
gdk-pixdata.h header into gdk-pixbuf.h.
Fixes #72
|
|
|
|
|
| |
Makes it easier to see what was enabled, what's going to be built, and
where it's going to be installed.
|
|
|
|
|
| |
Build inside a temporary directory, and ensure that the local
run-docker.sh script imports the repo inside the Docker container.
|
|
|
|
|
| |
The man page for gdk-pixbuf-csource is generated via xsltproc, so
there's no need to keep the troff version in tree.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Existing GdkPixbuf API like gdk_pixbuf_new_from_file_at_size and
gdk_pixbuf_get_file_info have grown smarter since this code was
written. They are now better at selecting the best GdkPixbufLoader for
a given file.
As a side-effect, this also prevents leaking the GFile and its URI,
but those are minor issues since the thumbnailer has a very short life
span
https://bugzilla.gnome.org/show_bug.cgi?id=778517
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently, it's not possible to get the dimensions of the original
image from a scaled GdkPixbuf. This is problematic for thumbnailers
because they want to embed the original dimensions into the resulting
thumbnail file. The only way to get this information is to either
decode the full-resolution image and then downscale it, even when the
decoder supports downscaling on-the-fly; or to re-implement large
chunks of code to avoid decoding the full-resolution image and
retain the original dimensions.
Exposing the original dimensions as options simplifies the thumbnailer
without affecting performance.
https://bugzilla.gnome.org/show_bug.cgi?id=778517
|
|
|
|
|
|
|
|
|
|
| |
All the options are stored under the "gdk_pixbuf_options" key, and only
those were being copied after applying the orientation. Anything else
stored under any other key gets lost with the unoriented GdkPixbuf.
Fallout from 06cf4c78067203b78acbfb29862350cdb8200b73
https://bugzilla.gnome.org/show_bug.cgi?id=778517
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=795705
|
|
|
|
| |
Otherwise the commit hook on git.gnome.org will complain.
|
|
|
|
|
|
|
|
| |
We haven't built them on anything that isn't a 32bit IA platform, and
we could probably get better mileage out of the currently implemented
pixops just by rearranging the C code and letting compilers do the
optimizations for us. We should definitely consider either using pixman
directly, or replacing slow pixops with SSE builtins, instead.
|
|
|
|
|
|
| |
The `docs` directory is empty, and gdk-pixbuf is not part of GTK any
more, so there's no point to have the API reference under a separate
directory.
|
|
|
|
|
|
| |
We use Meson, now. The Meson build has been tested for almost a year in
our continuous integration pipeline, and it's functionally equivalent to
the Autotools one.
|
|
|
|
|
| |
Once we migrate GdkPixbuf over to GitLab, we should already have all the
required bits in place for running a continuous integration pipeline.
|
|
|
|
| |
Outline a contribution workflow, and how GdkPixbuf is logically split.
|
|
|
|
|
| |
It's desolate, at the moment, and free software projects without a
decent README are a bit sad.
|
|
|
|
| |
We cannot expand those, as Xlib does not come with a gtk-doc index.
|
| |
|
|
|
|
|
|
|
| |
The "handling an expose event" example is badly outdated, referencing
concepts and practices that were deprecated halfway through the GTK 2.x
development. These days, it's going to confuse people more than help
them.
|
| |
|
|
|
|
|
| |
Instead of using a template, use Meson's pkgconfig module to generate
the pkg-config file for us from the build targets.
|
|
|
|
|
| |
We really only care about it as a library in order to build a test. We
can simply include the sources into the main shared library.
|
|
|
|
|
|
|
| |
Avoid a bunch of array iterations.
The get_supported_link_arguments() call is gated on a Meson version
check, to avoid requiring a bleeding edge stable version.
|
|
|
|
| |
Use a less ancient baseline.
|
|
|
|
| |
Add myself as a maintainer.
|
|
|
|
| |
Like we did for Meson, do this for the Autotools build.
|
| |
|
|
|
|
|
|
|
| |
If we're installing gdk-pixbuf ourselves, we should generate the loaders
cache file — unless DESTDIR is set or we're cross-compiling, in which
case we assume that there's a packaging system hook in place to do this
at the end of a full build.
|
|
|
|
|
|
| |
Like the Autotools build allows us to decide whether the tests should be
installed in a known location, we should have this option for the Meson
build.
|
|
|
|
|
| |
Remove long since gone intra-library links, and point to the API
reference website instead.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
We have two "set but unused variable" warnings coming from
gdk-pixbuf-xlib. The first is caused by an unnecessary variable used to
hold the return value of XGetWindowAttributes(); we don't do anything
with it, and it's way too late for changing the behaviour of the
function to bail out on error. The second is caused by a variable that
is only used when a debugging printf() is compiled in; to avoid the
warning without creating even more complex macro hell, we tell the
compiler that the variable is unused.
|
|
|
|
|
|
|
|
| |
If there are no built-in modules, the `builtin_module` variable is not
used, and the compiler will warn about it.
Since we're already using the variable inside a macro, we can just
declare it inside the macro block and save us the trouble altogether.
|
|
|
|
| |
A text file should not have the executable bit set.
|
|
|
|
|
| |
We dropped it with the Meson build port, and I'm not entirely sure it's
still tested, but it doesn't cost us anything to re-enable the check.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Use two loops to traverse the array of arrays `DM`, i.e. use `DM[y][x]`
instead of `DM[0][i]`.
This resolves a warning about undefined behavior when compiling with GCC
with aggressive loop optimizations enabled.
While we are at it: Move the variable definitions into the body of the
outer if-statement.
https://bugzilla.gnome.org/show_bug.cgi?id=748211
|
|
|
|
|
| |
Same as commit fc2d02ed908c524cc2831e5404d5f06b02e3f9fd, but for the
Meson build.
|
|
|
|
|
|
|
| |
The generated marshalers are public API, and the header needs to be
installed, even though nobody should be using them.
https://bugzilla.gnome.org/show_bug.cgi?id=795213
|
|
|
|
|
| |
We need to re-generate the gdk-pixbuf.types in order to include all
public types; this requires fixing the list of ignored headers.
|
|
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=794872
Signed-off-by: Quentin Glidic <sardemff7+git@sardemff7.net>
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=795210
|
|
|
|
|
|
| |
gdk_pixbuf_composite_color_simple
https://bugzilla.gnome.org/show_bug.cgi?id=789935
|
|
|
|
|
|
|
| |
Calling g_bytes_unref_to_data() will copy the contents of the bytes
buffer, so we need to free the data.
https://bugzilla.gnome.org/show_bug.cgi?id=787626
|
|
|
|
|
|
|
|
| |
We're currently installing the test launchers and parts of the test
data, but we're not installing the binaries and the whole suite of
image data we have.
https://bugzilla.gnome.org/show_bug.cgi?id=795527
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Relocation works by recognizing paths in the loaders cache
which start with the built time prefix and extract the relative
path from that.
This leads to the following problem when updating the cache:
In case the package is build on another machine one has to
either match the build directory layout or adjust the
cache by hand for the resulting cache to stay relocatable.
This commonly occurs with msys2 where mostly pre-build packages
are used which are built on another machine and the cache gets
generated at install time. Another case is updating the cache
in a separate deployment environment.
This patch takes the package installation directory as a base
and writes relative paths into the cache when relocation
is enabled. When loading the cache a relative path is made
absolute by prepending the package base again.
https://bugzilla.gnome.org/show_bug.cgi?id=776081
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|