This reverts commit 8c53b716.

See 26471ad4 for explanation.

Previously, ACLOCAL_AMFLAGS was set to ${ACLOCAL_FLAGS}. This caused
problems when running libtoolize --force --copy as part of autoreconf.
On JHBuild when ACLOCAL_FLAGS was set to the build tree's share/aclocal
directory, libtoolize would actually remove the libtool macros from
there.

When linking with libtool, you are supposed to use -R, not -rpath. This
caused a build failure on OSX due to two -rpath options being given in
one invocation of ld, which libtool fixes when you use -R.

When building on OSX, we would be using Cairo without XLib support, so
instead of erroring out if pkg-config can't find cairo-xlib, print a
warning. The X-based code will not be used anyway if Cairo hasn't
detected XLib.

In C code, the reference passed to the method_call() virtual function
would be passed to the return_value/error function that is eventually
called. But since these functions are called from JS, and the calling
convention is normalized, we need to unref reference passed to
method_call ourselves. See GLib bug 738259 for a more complete discussion.

https://bugzilla.gnome.org/show_bug.cgi?id=738122

In a type fixup, out-return value from JS_ValueToId() was unintentionally
lost.

https://bugzilla.gnome.org/show_bug.cgi?id=736929

Some third-party libraries rely on the ability to install the overrides,
so let's let them.

Michel Dänzer authored Aug 23, 2014 and Jasper St. Pierre committed Aug 25, 2014

Fixes the following JSUnit tests on PPC:

test0040mainloop.js
testByteArray.js
testCairo.js
testEverythingBasic.js
testFundamental.js
testGDBus.js
testGObjectClass.js
testMainloop.js
testParamSpec.js
testTweener.js

Signed-off-by: Michel Dänzer <michel@daenzer.net>

https://bugzilla.gnome.org/show_bug.cgi?id=729554

https://bugzilla.gnome.org/show_bug.cgi?id=735358

Michel Dänzer authored May 05, 2014 and Jasper St. Pierre committed Aug 22, 2014

Fixes crashes in the testsuite and in gnome-shell on at least PPC, due
to treating things like pointers which really aren't.

Signed-off-by: Michel Dänzer <michel@daenzer.net>

https://bugzilla.gnome.org/show_bug.cgi?id=734569

With regression tests

Introduce a new module, called package, which includes convenience
API and a set of simple conventions for packaged applications.
The module takes care of setting up private search paths for JS
modules (as resources, if possible) and private typelibs, and
it optionally does gettext too.

https://bugzilla.gnome.org/show_bug.cgi?id=690136

Needed for g_callable_info_get_instance_ownership_transfer

When calling a method that is (transfer full) on the instance
parameter we need to make an extra ref/copy.

https://bugzilla.gnome.org/show_bug.cgi?id=729545

These need to return void (undefined), and not themselves (which
is what happens if SET_RVAL() is not called, due to spidermonkey
reusing the slot for arguments.callee)

As the comment says, there may not be any interface info,
but g_base_info_unref() crashes if passed NULL.

While Andrea says my alias is configured, I only get emails
from @src.gnome.org...

For GNOME 3.13.3

We want to be able to free cairo objects from the background thread,
and that means we must be able to make xlib calls from multiple
threads. Which is ok, if we initialize xlib beforehand.

We don't have a good place to initialize gjs, where we're sure to
be called before the first XOpenDisplay(), so just stuff it into
a constructor.

https://bugzilla.gnome.org/show_bug.cgi?id=730030

The original Mainloop API had an optional priority argument at the
end of idle_add(), timeout_add() and timeout_add_seconds(). Restore
support for it.

https://bugzilla.gnome.org/show_bug.cgi?id=726333

We need to define GI_DISABLE_CAIRO when scanning to create the GIR,
not just when compiling the library, otherwise we'll fail because
cairo-gobject.h is not found.

When loading a lot of data in memory (for example in the shell
opening the overview, which loads all the desktop files and icons)
the RSS can increase a lot, so we would trigger GCs continously
without any hope of freeing memory, so ratelimit full GCs to at
most one every 5 frames.

https://bugzilla.gnome.org/show_bug.cgi?id=728048

When converting a native object to it's JavaScript proxy, we want to
be as precise as possible to map that native object.
Right now we handle different kinds of proxy for native types :
      * GObject proxies
      * GBoxed proxies
      * GParam proxies
      * GError proxies
      * Function proxies
      * Unions proxies
      * GI proxies
      * Fundamental proxies
      * basic types

The introspection data can provide information that helps us map a
GType to one of the items listed above. But there is a case that is
rather painful to deal with : interfaces.

In GObject interfaces can be attached to different kind of GTypes
(that includes GObjects and custom fundamental types and maybe more, I
haven't checked all the twisted cases). Although there is a way to add
a prerequisite for an interface, not all interface properly use this
feature.

In this particular case (native -> JS) we can use the native object's
gtype to figure out the most appropriate proxy, which what this patch
implements.

https://bugzilla.gnome.org/show_bug.cgi?id=727824

I believe when the installed test suite got created, the way JS files
are loaded changed. That resulted in the window object being different
from the scope of the test files, leaving that particular test suite
disabled.

https://bugzilla.gnome.org/show_bug.cgi?id=728004

We can share the static method code with GObject

We can share the code with object, and get them for free.

https://bugzilla.gnome.org/show_bug.cgi?id=725282

The GParamSpec API provides binding friendly accessors for
name, nick, blurb and default, so we don't need to recurse the
hierarchy looking for a field.

https://bugzilla.gnome.org/show_bug.cgi?id=725282

We must define as methods on prototypes only the functions
that have the IS_METHOD flag, or we get static functions there.

https://bugzilla.gnome.org/show_bug.cgi?id=725282

maxmium instead of maximum

Methods are definitely expected for GParamSpec classes, even
more than fields, and are easy to implement.

https://bugzilla.gnome.org/show_bug.cgi?id=725282

With proper argument marshalling, we don't need the custom code
anymore, the regular GObject functions works just as well.
We still need a lot of JS override code, because the arguments
are swapped...

https://bugzilla.gnome.org/show_bug.cgi?id=725282