| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|
|
|
|
|
| |
We are now using glib-tap.mk for the tests, and all we gain
from Makefile.decl is a predefined EXTRA_DIST, which doesn't
seem worth it.
|
|
|
|
| |
Use --enable-coverage and make coverage to get it.
|
| |
|
|
|
|
|
|
| |
The keys for tEXt chunks are documented as 1-79 character
ASCII strings. Testing this shows that we reject single-
character keys as too short. Fix that.
|
| |
|
|
|
|
|
| |
config.h has no inclusion guards, so including it in headers
is bound to cause redefinition warnings from cpp.
|
|
|
|
| |
Some of the loaders were not doing this.
|
|
|
|
| |
awk is needed for the automake tap driver.
|
| |
|
|
|
|
|
| |
Instead of using gtester, switch to using the TAP support
and the automake test harness.
|
|
|
|
| |
Make this test use g_test_run, so it is compatible with using TAP.
|
|
|
|
|
| |
Make this test use g_test_run, so it is compatible
with using TAP.
|
| |
|
| |
|
|
|
|
|
| |
For update have used values from https://git.gnome.org/browse/pango/log/tools/rgb.txt
that have already updated version of SVG/CSS pallete
|
|
|
|
|
| |
This tests that we correctly load a file with the .xbm extension
using both gdk_pixbuf_new_from_file and gdk_pixbuf_new_from_file_at_scale.
|
|
|
|
|
|
|
|
|
|
|
|
| |
The mime database has no magic at all for xbm, and it
gets sniffed as text/plain. This is a regression from
when we were using our own magic, so be a little more
forgiving here - treat a mimetype of 'text/plain' like
uncertain results, and consult the filename. This lets
us at least recognize files with the extension .xbm
correctly again.
https://bugzilla.gnome.org/show_bug.cgi?id=573726
|
|
|
|
|
|
|
|
|
|
| |
We are using a loader to implement some of the APIs that
are taking a filename, such as gdk_pixbuf_new_from_file_at_scale.
But the loader does not have the filename, so it can't use it
when determining the image type. This makes a difference for
types such as xbm, which have no good magic in the mime database.
Fix this by introducing a private API that lets us pass the
filename on to the loader.
|
|
|
|
| |
Harmless NULL / FALSE confusion.
|
| |
|
|
|
|
| |
This lets us avoid some manual overflow checks.
|
|
|
|
|
| |
The right macro to use in the fill function is N_(),
but some loaders were using _(). Correct this.
|
|
|
|
|
| |
We're now marking the tiff loader as threadsafe, so
remove the comment that says it isn't.
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=144042
|
| |
|
|
|
|
|
| |
Notably, this test catches the regression fixed
in the previous commit.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Commit c62676a284 had an unintended side-effect for loaders
which omit to call size_func, the new size set by
gdk_pixbuf_loader_set_size would overwrite the original
pixbuf size, and in effect force a scale factor of 1.0.
The xpm loader is one of the few which omit the size_func
call, thus the regression that calling
gdk_pixbuf_new_from_file_at_scale does not scale xpms anymore.
The fix is to use separate variables to pass the dimensions
when calling the size_func on behalf of the loader.
https://bugzilla.gnome.org/show_bug.cgi?id=686514
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=712704
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=712704
|
|
|
|
|
| |
We have been requiring a new-enough GLib for a while now,
so drop this unnecessary clutter.
|
|
|
|
|
|
| |
To use gdk-pixbuf-pixdata uninstalled, we need to use the
uninstalled modules as well, so set GDK_PIXBUF_MODULE_FILE
as well.
|
|
|
|
|
|
|
| |
We are using to-pixdata processing for some of the resources
in tests/, but gdk-pixbuf-pixdata has not been installed on
the system yet. So pass the location of the just-built one
via the GDK_PIXBUF_PIXDATA env var.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
Only use the sniff buffer while we need it.
|
|
|
|
|
|
| |
Move the generic incremental loading into its own function
so we don't end up taking two 64k buffers on the stack when
one is enough.
|
|
|
|
|
| |
Turns out gdk_pixbuf_animation_new_from_file() was also
using a different size.
|
|
|
|
|
|
|
| |
GdkPixbufLoader was using 1k as a buffer size for sniffing file types,
and the gdk_pixbuf_new_from_file() family of functions was using 4k,
for no good reason. Unify this so we use the same buffer size
everywhere.
|
|
|
|
|
|
| |
... so that it would be more consistent with the rest of the source files,
and application of patches would be easier and cleaner. The README.txt and
.sln files under build/win32/vs* still has to have DOS/Windows line feeds.
|
| |
|
| |
|
|
|
|
|
| |
If we longjump out of libjpeg, we may have already parsed
the exif data, so free that as well.
|
|
|
|
|
| |
This test verifies that the preceding commit indeed
makes icc-profile loading from jpeg files work.
|
|
|
|
| |
This was apparently never tested, and thus didn't work.
|
|
|
|
|
|
| |
Make all error paths in gdk_pixbuf__jpeg_image_load
go through the same path, so we don't forget to free
the exif data in half of them.
|
| |
|
|
|
|
|
| |
This makes it easier to upgrade to Visual Studio 2012/2013 formats,
like the other *.vcxproj files in GDK-Pixbuf.
|
|
|
|
|
|
|
|
| |
Due to an oversight, combining the --update-cache option with
explicitly specified modules was leading to an attempt to
open the --update-cache.so module - with predictably disappointing
results. Make this work as intended, and as described
in the man page.
|