Age | Commit message (Collapse) | Author | Files | Lines |
|
This allows creating 'light' binaries (i.e. just one arch) when we
are not targetting universal builds
|
|
If which is not in the $PATH, why would whereis be? Fixes the build
on Windows where "whereis" does not exist.
|
|
By default the compilation system used by gobject-introspection (which
is the python distutils compiler system) will create "universal" builds.
We don't need that for g-ir-scanner
|
|
|
|
|
|
Windows
|
|
After polling for file descriptors, the Cocoa select thread did wake up
the main run loop instance that has been stored in a static variable.
This variable might have already been set to NULL by
g_main_context_release, resulting in a segfault of CFRunLoopWakeUp. This
race is fixed by this commit by simply not storing the main run loop in
a variable, but instead calling CFRunLoopGetMain locally where needed.
Within a single process, the main run loop is always the same, and
always accessible. It is therefore not necessary anyway to remember the
run loop instance when acquiring the context.
https://bugzilla.gnome.org/show_bug.cgi?id=758285
|
|
|
|
|
|
downloading
|
|
|
|
|
|
|
|
|
|
Android >= 6.0
Due to the latest changes in Android 6.0 runtime (http://developer.android.com/intl/ko/about/versions/marshmallow/android-6.0-changes.html#behavior-runtime)
Each module should explicitly define it's SONAME. since the building scripts for building android_gstreamer.so doesn't define soname the library is built without one.
This will cause a failure in loading any library that uses libgstreamer_android.so.
https://bugzilla.gnome.org/show_bug.cgi?id=757732
|
|
launched
To avoid rescanning the registry at each start we symlink the plugins
to some user directory and thus we can not have several bundle launched
at once
|
|
|
|
Otherwise we have to re create the cache each time, and that takes time.
|
|
|
|
Allows nesting certain configuration operations. e.g. modifying home_dir in one
file, the configured target/arch in another and setting the build-tools sources
directory in another.
When multiple options are specified, the last one one wins.
https://bugzilla.gnome.org/show_bug.cgi?id=757511
|
|
|
|
gmp.h contains the sizeof(mp_limb_t), which is 32 bits on 32 bit architectures
and 64 bit on 64 bit architectures. nettle will fail with assertions when
mixing these, and other things probably won't work well either.
|
|
|
|
They waste time and space, and because of the missing symbolic link support
they have lots of duplicated files. People on Windows can also just use an
application to extract tar.bz2 files.
|
|
By default, we are now always using the 'latest' SDK, as returned by the
xcodebuild command line tool, i.e. no hard-coded SDK paths are used by
cerbero anymore. It is still possible to ask for a specific SDK version,
by setting 'osx_target_sdk_version' in config (e.g. setting it to
'10.10' to force compilation against the 10.10 SDK). Requesting an SDK
version via osx_target_sdk_version that doesn't exist on the system will
error out.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=757357
|
|
Update to latest version of xz (xz-5.2.2).
Closes https://github.com/EricssonResearch/cerbero/pull/37
|
|
recent autotools
|
|
LIBSPANDSP_EXPORTS must only be defined while building libspandsp,
otherwise code that uses spandsp symbols will be using them with
the dllexport attribute and the binaries will probably fail at runtime.
The change in config_sh is required because we patch the Makefile.am,
so automake needs to run again, and while "make" will detect that,
the code in the tarball is configured to run "automake-1.13", which
will fail on most systems nowadays. Calling autogen.sh again ensures
that we call the correct automake.
|
|
|
|
Otherwise we might initialize gobject before glib, causing assertions
and crashes.
https://bugzilla.gnome.org/show_bug.cgi?id=756139
|
|
able to ship it
|
|
symbols
Because they conflict with libmp3lame symbols when we link
everything statically for mobile targets
https://bugzilla.gnome.org/show_bug.cgi?id=756882
|
|
|
|
|
|
|
|
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=755173
|
|
|
|
|
|
https://github.com/EricssonResearch/cerbero/pull/32
|
|
|
|
|
|
|
|
|
|
See https://openradar.appspot.com/22671534 for more details
|
|
|
|
XCode 7
|
|
There's a compatible element register in audiomixer
|
|
This should fix a non-cross compiling case, Android x86_54 where
this check was failing.
https://bugzilla.gnome.org/show_bug.cgi?id=754997
|
|
By not setting an empty patches array, we endup with the patches
being randomly applied to other recipies.
https://bugzilla.gnome.org/show_bug.cgi?id=754997
|