| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
| |
Use library() to make building a static or dynamic library
optional depending on the builtin "default_library" meson
option. Fixes failures when using the library statically:
/home/buildroot/autobuild/instance-0/output-1/host/opt/ext-toolchain/bin/../lib/gcc/arm-buildroot-linux-uclibcgnueabi/9.3.0/../../../../arm-buildroot-linux-uclibcgnueabi/bin/ld: /home/buildroot/autobuild/instance-0/output-1/host/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libc.a(__uClibc_main.os): in function `__uClibc_fini':
__uClibc_main.c:(.text+0x15c): undefined reference to `__fini_array_start'
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
|
| |
|
|
|
|
|
|
|
|
| |
Add introspection option
Fix #3
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
|
|
|
|
| |
It's pointless to document them.
|
|
|
|
| |
The description is missing.
|
|
|
|
|
| |
There is no point in keeping two build systems in parallel, especially
when one of them is Autotools.
|
|
|
|
| |
Qt 4 was end-of-lifed in December 2015.
|
|\
| |
| |
| |
| | |
Don't use a field before setting it
See merge request GNOME/libmediaart!10
|
| |
| |
| |
| |
| |
| | |
The g_memdup() call was deprecated in GLib 2.68 because it used an
unsigned integer for the buffer size argument. The g_memdup2() function,
which uses a gsize, was introduced in the same version.
|
| |
| |
| |
| |
| |
| | |
We are using the data->len field in our g_memdup() call before we're
actually setting it; since the data structure is zero-filled, it means
we're not copying the buffer at all.
|
|/
|
|
| |
Tools that parse the GIR need the exported package.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This restriction was already in place, since we passed the inputs
directly to GLib functions that expect valid UTF8.
If this library were widely used then we might proactively validate
inputs to protect against exploits. At time of writing, this module is
only used in 3 places (grilo-plugins, gnome-music and rygel), so just
check they are doing the right thing when calling the API.
See https://gitlab.gnome.org/GNOME/libmediaart/-/merge_requests/5 for
background.
|
|
|
|
|
|
|
|
|
|
|
| |
We return NULL if input is NULL and a newly allocated empty string if
input is "". Some comments disagreed with this, and it's possible the
change causes a memory leak in some app, but the alternative of returning
NULL when passed "" is dangerous as some code may free the return value
in this case. (In fact, libmediaart itself does so).
Also, make behaviour occur independently of whether
`G_ENABLE_CONSISTENCY_CHECKS` was defined at build time.
|
|\
| |
| |
| |
| | |
Fix Qt5 backend crashing when running in Qt application
See merge request GNOME/libmediaart!2
|
| |
| |
| |
| |
| | |
Creating a dummy QCoreApplication instance is bad in itself,
but even worse when running in an app that already has one.
|
|\ \
| | |
| | |
| | |
| | | |
meson.build: don't require C++
See merge request GNOME/libmediaart!3
|
| |/
| |
| |
| |
| |
| | |
Does not mandate C++ as it is only required for extractqt.cpp
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
|
| |
| |
| |
| |
| |
| | |
After the removal of mediaartlocal, this is useless now.
It can also cause some lock-ups if gvfs is missing
|
|/ |
|
|
|
|
|
|
|
| |
The meson build did not add the extractdummy.c to the sources, which
contains introspection annotations.
https://bugzilla.gnome.org/show_bug.cgi?id=792272
|
|
|
|
|
| |
This was making the functions defined in extractgeneric.h not really
exported.
|
|
|
|
| |
This must be installed by the meson install as well...
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=786067
|
|
|
|
|
| |
The header is referenced by publicly installed headers, which means
installations of libmediaart without it are now broken.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
These are hopefully complete already. I have compared an Autotools-built
and a Meson-built install of libmediaart and found only the following
differences:
* libmediaart-2.0.la isn't generated by Meson
* External references in the gtk-doc documentation are relative with
Meson and absolute with Autotools
* Some changes in generated .vapi file and .pc file
https://bugzilla.gnome.org/show_bug.cgi?id=783562
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously we relied on libtool's -export-symbols-regex feature, but we
are hoping to drop GNU Autotools and GNU Libtool in soon in favour of
Meson.
Meson doesn't have an equivalent feature, instead the advice is to
control symbol visibility at compile time.
The approach taken in this patch is based on Pango's build system. Pango
tells the compiler to hide symbols by default (if possible), and then defines
a _PANGO_EXTERN variable at compile time which marks a single symbol as
public. In Pango's case there is then further machinary to hide symbols
based on deprecation policies but I have not copied that here, instead I
used _LIBMEDIAART_EXTERN directly.
If a compiler doesn't support hiding symbols then the library we build
makes all symbols available, which is exactly what would happen before
on platforms where the libtool didn't have an implementation for
-export-symbols-regex.
See also:
http://mesonbuild.com/FAQ.html#how-do-i-do-the-equivalent-of-libtools-exportsymbol-and-exportregex
https://git.gnome.org/browse/pango/
https://git.gnome.org/browse/pango/tree/pango/pango-version-macros.h
https://gcc.gnu.org/wiki/Visibility
https://gcc.gnu.org/onlinedocs/gcc/Code-Gen-Options.html#Code-Gen-Options (-fvisibility)
https://bugzilla.gnome.org/show_bug.cgi?id=783562
|
|
|
|
|
|
|
|
|
|
| |
This flag was just enabling a codepath that sets QT_QPA_PLATFORM=minimal
in the environment before constructing a QCoreApplication instance.
This code path has no effect anyway: QT_QPA_PLATFORM only has an effect
if you are using QGuiApplication.
https://bugzilla.gnome.org/show_bug.cgi?id=783562
|
|
|
|
|
|
| |
Fixes this warning:
libmediaart/extract.h:93: Warning: MediaArt: multiple comment blocks documenting 'media_art_error_quark:' identifier (already seen at extract.c:1279).
|
|
|
|
|
| |
We need those files to be generated before the library is built, but it
seems Automake is failing miserably in conditions of massive parallelism.
|
|
|
|
|
|
|
|
|
| |
Under massive parallelism — like the Continuous build — we may end up in
a situation where the sources get built before the marshaller sources
are generated.
Let's try to increase chances that the dependency tracking inside
Automake will do the right thing.
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=767959
|
|
|
|
| |
Otherwise the function is unusable through gobject-introspection.
|
|
|
|
|
| |
There are paths where the stop variable is not set, so might
end up with an uninitialized value when checked.
|
| |
|
|
|
|
|
| |
Use -export-symbols-regex to hide private symbols which should not be
exported.
|
|
|
|
|
|
|
| |
Documentation says "or NULL" and the code also handles it, but the g-i
for this was missing.
https://bugzilla.gnome.org/show_bug.cgi?id=739943
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=739942
|
|
|
|
|
|
|
| |
debug was printing g_strerror(errno) even in success case, leaving
weird confusing messages, indicating an error even if symlinking succeeded.
https://bugzilla.gnome.org/show_bug.cgi?id=739873
|
|
|
|
|
|
|
| |
This is deemed quite useless now and we should not be littering the file
system with .mediaartlocal directories everywhere.
https://bugzilla.gnome.org/show_bug.cgi?id=722795
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=724879
|
| |
|
|
|
|
|
|
|
|
| |
Part of this API changes makes the media_art_get_{path|file}() APIs not do
i/o operations like creating the cache directory. This is now done in
media_art_process_new().
https://bugzilla.gnome.org/show_bug.cgi?id=724879
|
| |
|
|
|
|
|
|
|
|
|
| |
The function media_art_get_file checks for null values in artist, title and
prefix arguments, and puts special values on them if they do.
Do the same on media_art_get_path which calls media_art_get_file internally.
https://bugzilla.gnome.org/show_bug.cgi?id=727478
|
|
|
|
|
|
|
| |
Also added some checks into the API and a return value to know if the function
was successful.
https://bugzilla.gnome.org/show_bug.cgi?id=734837
|
|
|
|
| |
Also updated the unit tests so they don't fail with the new tap system
|
| |
|
| |
|