Libraries in the GNOME platform are currently being moved
from Docbook to Markdown. This produces HTML at present instead
of Mallard, but in coming commits Mallard will be removed so this
won't be an issue.

This introduces a dependency on python-markdown.

Copy the .gir files that are bundled with g-i to build/ before trying to
compile the .gir's that were generated during the build process, so that
the compilation of these .gir's may succeed.  Also, create the
share/gir-1.0 and lib/girepository-1.0 directories, if needed, before
trying to copy the .gir/.typelib files.

The implementation as a method returning itself as its implemented
interface type.

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

This should fix race conditions when multiple processes attempt to
access the cache concurrently.

https://bugzilla.gnome.org/show_bug.cgi?id=724886

The -I flags for generating GIRepository-2.0.gir should use $(BASEDIR), not
%BASEDIR%, as we are not using a Windows .bat to generate the introspection
files.

Press the secret code too unlock the hidden bonus: GVariantDict!

'ok' != 'return_value', and we never restore 'Returns:'.

At the same time, do the same fix for (out) parameters if the
return value is void, so this

void my_object_get_foo (MyObject *object, Foo *)

is turned into

function get_foo() : Foo
Returns: ....

GClosures are represented in gjs as JS functions for in
parameters, so it makes sense to use Function in place of
the boxed structure (which is also not part of the documentation)

Note that GClosures returned from C code are marshalled
as boxed types, but this is a gjs bug really, and JS code
should never touch them.

They are enums after all.

GValues are transparently marshalled by GJS, so you can
never use them as boxed (and they are not part of the
documentation as such).

This was causing a "warning: function declaration isn't a
prototype" with strict-prototypes enabled.

Add missing page needed for tests to run after commit 7027bb25

https://bugzilla.gnome.org/show_bug.cgi?id=725198

Use ParseState enum instead of a boolean for the ParseContexts embedded_type
flag. This allows specific tracking of the embedded type currently being
parsed which can now either be STATE_STRUCT_FIELD or STATE_CLASS_FIELD (or
allow for future expansion). Add ParseState::STATE_NONE as the default for
this field.

Fix GObject FieldBlob validation to take into account the sizeof
CallbackBlobs (copied from the struct validator).

Add static g_object_info_get_field_offset which parallels
g_struct_info_get_field_offset which is needed since callback fields may
vary in size.

https://bugzilla.gnome.org/show_bug.cgi?id=725198

This shows up in the headers in FreeBSD's stdlib.  Now I didn't add
anything to regress.h because I think that'd introduce an effective
hard dependency on a C11-capable compiler, unless we used #ifdef
trickery.  But the problem with #ifdef is that the -expected.gir isn't
conditional.

https://bugzilla.gnome.org/show_bug.cgi?id=724851

LRN authored Feb 21, 2014 and Colin Walters committed Feb 27, 2014

https://bugzilla.gnome.org/show_bug.cgi?id=724882

Conform shlibs.py to the PEP8 standard. New additions were causing make
check to fail:
 Fixed E211 whitespace before '('
 Fixed E501 line too long (104 > 99 characters)

If a xref points to a node that was not visited, because it
was not introspectable, don't assume it has no parent. Instead,
try to fetch the parent from the node directly.
This is still wrong (the xref points nowhere because no page is
generated), but it looks less wrong.
If OTOH a xref points to a different namespace, generate the
link directly, because yelp-build is not capable of translating
it.

Make sure to set the parent of the function

Switch enum from a topic to a guide page, and include links to
all static methods.

As the comment says, it's an AST bug, the structure there cannot
be reasonably represented in gobject-introspection, and it makes
the tests for C and Python crash, so whatever.

In cases like g_resources_register() and gdk_events_get_angle(),
the c_symbol_prefix of the first parameter (resp. g_resource and
gdk_event) is a prefix of the symbol, but the next character is
not _, so that should not be considered a method.
For backward compatibility reasons, we still generate one, but
then it's not included in the documentation (because of moved_to)

Unless they have no methods, static methods and constructors
(because then they are useless)
Fixes GLib.BookmarkFile, GLib.PatternSpec and probably others.

We should only look at introspectable constructors when resolving
the default and zero args. Also, a disguised structure or one
with zero fields can't be allocated directly.

In addition, restore availability of all constructors, because
we need a page to document the default one, and because some classes
have problems with GObject construction.

Support for fundamental types was just merged in gjs, so just
filter out GParamSpec subclasses (because there is only one
class in gjs)

If a function is shadowed, omit it from the documentation, and
if a function shadows, uses the new name.

The ByteArray class is special in gjs, it's not a normal Array,
so make sure it's flagged as such.

g-ir-compiler ignores them too, and they cause a crash, so whatever.

The introspection build on Windows recently received support to deduce the
DLL name from the library that is passed into g-ir-scanner, so the library
name (without the file name extensions, etc) ought to be passed in for the
build of introspection files on Windows, just like the *nix builds of
g-i.

Commit f3fcdf97 introduced shlib resolution for Windows, but it only works
with MinGW and would therefore break builds of introspection files for
Visual C++.

Improve on the situation by adding capabilities to do the same with Visual
C++, as the introspection dumper program is still compiled and linked with
the Microsoft tool set (only the proprocessor to process the various
headers is using GCC/MinGW).

This makes use of the dumpbin utility that is shipped with Visual C++,
which will query a .lib file, which leads to the DLL that the .lib file
will link to.  The 'gcc -print-search-dirs' can be replaced on Visual C++
builds by querying the LIB environment variable, as this is what the Visual
C++ compiler/linker really looks at during compilation and linkage.

https://bugzilla.gnome.org/show_bug.cgi?id=724890

LRN authored Feb 21, 2014 and Colin Walters committed Feb 22, 2014

GI just adds '.dll' to library name and calls it a day.  There's a
comment about how this function might work, i've used it to implement
something better.  This requires a compiler that supports
-print-search-dirs argument (i.e.  gcc) and a dlltool.

https://bugzilla.gnome.org/show_bug.cgi?id=724890

LRN authored Feb 21, 2014 and Colin Walters committed Feb 22, 2014

https://bugzilla.gnome.org/show_bug.cgi?id=724888

LRN authored Feb 21, 2014 and Colin Walters committed Feb 22, 2014

https://bugzilla.gnome.org/show_bug.cgi?id=724887

LRN authored Feb 21, 2014 and Colin Walters committed Feb 22, 2014

https://bugzilla.gnome.org/show_bug.cgi?id=724887