platform/upstream/gobject-introspection.git
12 years agogirffi: Add new g_function_invoker_new_for_address
Jasper St. Pierre [Mon, 28 Nov 2011 19:51:04 +0000 (14:51 -0500)]
girffi: Add new g_function_invoker_new_for_address

This is a new method designed to make a GIFunctionInvoker for
any GICallableInfo*, for bindings to use.

12 years agoUpdate annotations for GLib 2.31.14
Rico Tzschichholz [Tue, 31 Jan 2012 07:09:21 +0000 (08:09 +0100)]
Update annotations for GLib 2.31.14

12 years agotests: Add more interesting marshalling tests for GTypes
Jasper St. Pierre [Sun, 22 Jan 2012 08:32:44 +0000 (03:32 -0500)]
tests: Add more interesting marshalling tests for GTypes

https://bugzilla.gnome.org/show_bug.cgi?id=668426

12 years agoAdd regress test methods for callbacks taking GError and GHashTable
Martin Pitt [Mon, 23 Jan 2012 07:53:44 +0000 (08:53 +0100)]
Add regress test methods for callbacks taking GError and GHashTable

This covers both "transfer none" and "transfer full" cases.

This is a requisite for writing a pygobject test case for
https://bugzilla.gnome.org/show_bug.cgi?id=666270

12 years agotests: add test functions which return GErrors
Will Thompson [Fri, 20 Jan 2012 14:48:20 +0000 (14:48 +0000)]
tests: add test functions which return GErrors

GStreamer has the following method:

  void gst_message_parse_error (
      GstMessage *message,
      GError **error,
      gchar **debug_message);

This patch adds a number of test functions with similar signatures which
do not follow the standard "throws GError" pattern.

https://bugzilla.gnome.org/show_bug.cgi?id=666098

12 years agoRelease 1.31.10
Colin Walters [Thu, 19 Jan 2012 22:14:19 +0000 (17:14 -0500)]
Release 1.31.10

12 years agoUpdate annotations from glib git GOBJECT_INTROSPECTION_1_31_10
Rico Tzschichholz [Thu, 19 Jan 2012 10:34:32 +0000 (11:34 +0100)]
Update annotations from glib git

12 years agoUpdate annotations from glib git
Colin Walters [Wed, 11 Jan 2012 20:58:08 +0000 (15:58 -0500)]
Update annotations from glib git

12 years agoAllow using GLib.List(Foo) instead of GLib.List<Foo>
Olivier Crête [Tue, 20 Dec 2011 22:45:11 +0000 (17:45 -0500)]
Allow using GLib.List(Foo) instead of GLib.List<Foo>

Using < and > for delimiters is problematic because gtk-doc is Docbook
which in turn is SGML, and those are reserved characters.

https://bugzilla.gnome.org/show_bug.cgi?id=663190

12 years agoUpdate annotations from GLib git
Rico Tzschichholz [Sun, 8 Jan 2012 14:10:04 +0000 (15:10 +0100)]
Update annotations from GLib git

12 years agotests: Add tests for flat GValue arrays
Jasper St. Pierre [Tue, 30 Aug 2011 20:43:18 +0000 (16:43 -0400)]
tests: Add tests for flat GValue arrays

https://bugzilla.gnome.org/show_bug.cgi?id=657766

12 years agoscanner: Remove duplicate declaration of ErrorQuarkFunction
Jasper St. Pierre [Mon, 12 Dec 2011 20:24:20 +0000 (15:24 -0500)]
scanner: Remove duplicate declaration of ErrorQuarkFunction

12 years agogirffi: Fix compiler warnings
Jasper St. Pierre [Fri, 6 Jan 2012 19:55:45 +0000 (14:55 -0500)]
girffi: Fix compiler warnings

12 years agorepository: Squash memory leak in _get_func and fix up style
Jasper St. Pierre [Mon, 12 Dec 2011 21:14:57 +0000 (16:14 -0500)]
repository: Squash memory leak in _get_func and fix up style

12 years agoAdd a marshalling test for returning a NULL array
Paolo Borelli [Tue, 3 Jan 2012 16:00:02 +0000 (17:00 +0100)]
Add a marshalling test for returning a NULL array

12 years ago.dir-locals: Fix unbalanced parenthesis
Colin Walters [Mon, 2 Jan 2012 18:25:38 +0000 (13:25 -0500)]
.dir-locals: Fix unbalanced parenthesis

12 years agoffi: Treat enums as 32 bit signed values to fix PPC64
Ray Strode [Wed, 21 Dec 2011 20:55:18 +0000 (15:55 -0500)]
ffi: Treat enums as 32 bit signed values to fix PPC64

To call a function dynamically using ffi, the caller
first has to tell ffi the size of all the input arguments
of the function. On little endian architectures (like x86_64)
specifying a size that's too large will happen to work because
of how the bits are laid out in memory.  On big endian architectures,
however, specifying the wrong size can lead to reading the wrong
bits.

The function g_type_info_get_ffi_type maps input giargument types to
specific sizes. It was assuming enums were word (pointer) sized; in
fact they can be in theory any size (1,2,4,8 bytes), but in practice
in introspection (via GIArgument) as well as GValue we're limited to 4
byte enums.

This commit fixes PPC64 (big endian, 64 bit).

Signed-off-by: Colin Walters <walters@verbum.org>
https://bugzilla.gnome.org/show_bug.cgi?id=665150

12 years agotests: Add a long string constant test
Colin Walters [Tue, 20 Dec 2011 18:24:03 +0000 (13:24 -0500)]
tests: Add a long string constant test

https://bugzilla.gnome.org/show_bug.cgi?id=662888

12 years agoRelease 1.31.6 GOBJECT_INTROSPECTION_1_31_6
Colin Walters [Tue, 20 Dec 2011 16:14:51 +0000 (11:14 -0500)]
Release 1.31.6

12 years agoUpdate annotations for GLib 2.31.6
Rico Tzschichholz [Tue, 20 Dec 2011 13:43:11 +0000 (14:43 +0100)]
Update annotations for GLib 2.31.6

12 years agoUpdate gitignore
Rico Tzschichholz [Tue, 20 Dec 2011 13:42:25 +0000 (14:42 +0100)]
Update gitignore

12 years agoDocs: fix typo
Claudio Saavedra [Mon, 19 Dec 2011 12:50:40 +0000 (14:50 +0200)]
Docs: fix typo

12 years agoUpdate annotations from GLib git
Colin Walters [Thu, 15 Dec 2011 14:50:04 +0000 (09:50 -0500)]
Update annotations from GLib git

12 years agogiscanner: fix use after decref
Andreas Schwab [Sun, 27 Nov 2011 19:48:24 +0000 (20:48 +0100)]
giscanner: fix use after decref

Decrement reference to temporary string object only after last use of its
value.

12 years agoRelease 1.31.1 GOBJECT_INTROSPECTION_1_31_1
Johan Dahlin [Mon, 5 Dec 2011 18:08:04 +0000 (16:08 -0200)]
Release 1.31.1

12 years agoMakefile.introspection: allow buildir girs in subdirs
Marc-Antoine Perennou [Thu, 1 Dec 2011 10:16:38 +0000 (11:16 +0100)]
Makefile.introspection: allow buildir girs in subdirs

Formerly, trying to build gi/Foo-1.0.gir resulted in namespace being set to "gi/Foo"
Trying to build g-i/Foo-1.0.gir was even setting it to "g"
This fixes this behaviour by only considerating the filename without dirs.
Btw, ensure that the directory exists

Signed-off-by: Marc-Antoine Perennou <Marc-Antoine@Perennou.com>
https://bugzilla.gnome.org/show_bug.cgi?id=665276

12 years agoSave .make-check-passed in builddir, so make distcheck passes
Johan Dahlin [Mon, 5 Dec 2011 18:04:53 +0000 (16:04 -0200)]
Save .make-check-passed in builddir, so make distcheck passes

12 years ago- Fix a substitution:
Jasper Lievisse Adriaanse [Mon, 5 Dec 2011 10:20:57 +0000 (11:20 +0100)]
- Fix a substitution:
use TOOL_SUBSTITUTIONS like the others, to fix an unsubstitued @datarootdir@ in g-ir-doc-tool

https://bugzilla.gnome.org/show_bug.cgi?id=659649

12 years agogitypelib: Fix compiler warnings
Jasper St. Pierre [Mon, 28 Nov 2011 09:59:47 +0000 (04:59 -0500)]
gitypelib: Fix compiler warnings

These variables are unused.

https://bugzilla.gnome.org/show_bug.cgi?id=665249

12 years agogibaseinfo: Fix compiler warnings
Jasper St. Pierre [Mon, 28 Nov 2011 09:56:58 +0000 (04:56 -0500)]
gibaseinfo: Fix compiler warnings

g_slice_free complains when you pass it a pointer of a different type
than it was expecting, like it should.

https://bugzilla.gnome.org/show_bug.cgi?id=665249

12 years agoscanner: Put the -l library names after the .o
Albert Astals Cid [Tue, 29 Nov 2011 18:45:55 +0000 (18:45 +0000)]
scanner: Put the -l library names after the .o

That is how gcc expects them.

From the gcc man page:
-llibrary
   Search the library named library when linking.

   It makes a difference where in the command you write this option; the linker
searches and processes libraries and object files in the order they are
specified.  Thus, foo.o -lz bar.o searches library z after file foo.o but
before bar.o.  If bar.o refers to functions in z, those functions may not be
loaded.

https://bugzilla.gnome.org/show_bug.cgi?id=664616

12 years agoAdd a floating alias for none
Johan Dahlin [Fri, 25 Nov 2011 14:09:51 +0000 (12:09 -0200)]
Add a floating alias for none

https://bugzilla.gnome.org/show_bug.cgi?id=657202

12 years agogiscanner: Add gmodule-2.0 to compiler packages
Florian Müllner [Wed, 23 Nov 2011 21:48:26 +0000 (22:48 +0100)]
giscanner: Add gmodule-2.0 to compiler packages

Gio no longer draws in gmodule, so we need to add the dependency
explicitly.

https://bugzilla.gnome.org/show_bug.cgi?id=664681

12 years agoFix build after GIO removed gmodule dependency
Florian Müllner [Wed, 23 Nov 2011 22:05:22 +0000 (23:05 +0100)]
Fix build after GIO removed gmodule dependency

https://bugzilla.gnome.org/show_bug.cgi?id=664681

12 years agoUpdate glib annotations
Jasper St. Pierre [Tue, 1 Nov 2011 19:45:02 +0000 (15:45 -0400)]
Update glib annotations

12 years agoOpenBSD uses it's own libtool implementation which breaks giscanner in certain
Jasper Lievisse Adriaanse [Fri, 18 Nov 2011 11:01:22 +0000 (12:01 +0100)]
OpenBSD uses it's own libtool implementation which breaks giscanner in certain
situations. So adjust resolve_non_libtool() in this case.

https://bugzilla.gnome.org/show_bug.cgi?id=664282

12 years agolibxml2: add exported package name
Daniel Drake [Sun, 30 Oct 2011 14:57:58 +0000 (14:57 +0000)]
libxml2: add exported package name

This is needed so that g-ir-scanner can find the libxml2 include
path when producing bindings for libraries that interact with libxml2.

12 years agoscanner: Only add rpaths for absolute directories
Colin Walters [Fri, 4 Nov 2011 00:33:53 +0000 (20:33 -0400)]
scanner: Only add rpaths for absolute directories

We shouldn't do it for module-internal paths, since it can't work
and is unnecessary even if it did.

12 years agoscanner: Also add an RPATH for library paths specified
Colin Walters [Thu, 3 Nov 2011 22:23:53 +0000 (18:23 -0400)]
scanner: Also add an RPATH for library paths specified

Introspection scanning is a bit special in that we both compile *and*
run a temporary binary.  In some cases like jhbuild we will pick up
the runtime location of libraries via LD_LIBRARY_PATH.  But this falls
over when we've been given a private subdirectory, since it won't be
in that path.

Let's just "do the right thing" here by injecting an rpath for the
directories as well.

12 years agoAdd tests for chaining up from vfuncs to ancestors
Tomeu Vizoso [Sat, 29 Oct 2011 13:00:26 +0000 (15:00 +0200)]
Add tests for chaining up from vfuncs to ancestors

12 years agoconfigure: Update to 1.31.0
Colin Walters [Wed, 26 Oct 2011 15:36:32 +0000 (11:36 -0400)]
configure: Update to 1.31.0

12 years agofixes and more tests for marshalling of arrays of gvariants
Mikkel Kamstrup Erlandsen [Tue, 4 Oct 2011 10:01:42 +0000 (12:01 +0200)]
fixes and more tests for marshalling of arrays of gvariants

Fix return type annotation for gi_marshalling_tests_array_gvariant_in()

Add function variations of gi_marshalling_tests_array_gvariant_i() with
different transfer modes.

Make gi_marshalling_tests_array_gvariant_none_in() actually respect
the transfer none  annotation on the return value.

https://bugzilla.gnome.org/show_bug.cgi?id=638915

Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
12 years agoAdd regress test method for a (skip) return value without out arguments
Martin Pitt [Fri, 21 Oct 2011 14:13:02 +0000 (16:13 +0200)]
Add regress test method for a (skip) return value without out arguments

This is a requisite for writing a pygobject test case for
https://bugzilla.gnome.org/show_bug.cgi?id=662383

12 years agotests: Also dist headeronly.h
Colin Walters [Thu, 20 Oct 2011 21:15:32 +0000 (17:15 -0400)]
tests: Also dist headeronly.h

Spotted by Rico Tzschichholz

12 years agoWindows port: fix _resolve_non_libtool
Dieter Verfaillie [Mon, 5 Sep 2011 15:46:20 +0000 (17:46 +0200)]
Windows port: fix _resolve_non_libtool

By simply appending '.dll' to the library names.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agotests: Actually add headeronly.h
Colin Walters [Thu, 20 Oct 2011 17:47:37 +0000 (13:47 -0400)]
tests: Actually add headeronly.h

12 years agoUse the correct size when freeing unused info
Sjoerd Simons [Wed, 19 Oct 2011 22:26:35 +0000 (23:26 +0100)]
Use the correct size when freeing unused info

A GIBaseInfo struct can underneath either be GIRealInfo *or*
GIUnresolvedInfo if the type is GI_INFO_TYPE_UNRESOLVED. So when we
eventually free the structures slice use the correct struct type
otherwise things get unhappy.

12 years agoMakefile.introspection: Allow usage with --header-only
Colin Walters [Wed, 19 Oct 2011 19:46:29 +0000 (15:46 -0400)]
Makefile.introspection: Allow usage with --header-only

12 years agoscanner: Support --header-only flag
Colin Walters [Wed, 19 Oct 2011 18:44:48 +0000 (14:44 -0400)]
scanner: Support --header-only flag

This is useful for someone who just wants a set of constants from a .h
file accessible by introspection.

12 years agobarapp: Don't call g_thread_init()
Colin Walters [Tue, 18 Oct 2011 19:25:27 +0000 (15:25 -0400)]
barapp: Don't call g_thread_init()

We don't need to, and if we wanted to, we have to link against
-lgthread.

Spotted by Rico Tzschichholz

12 years agoparser: prevents a segfault when _g_ir_parser_parse_string returns NULL error was...
Alberto Ruiz [Tue, 18 Oct 2011 15:45:38 +0000 (16:45 +0100)]
parser: prevents a segfault when _g_ir_parser_parse_string returns NULL error was not set.

Noticed the segmentation fault while using Vala to generate a .gir, a bug has
been filed tomake sure Vala doesn't export gir symbols outside of a namespace
(see https://bugzilla.gnome.org/show_bug.cgi?id=661952)

https://bugzilla.gnome.org/show_bug.cgi?id=661951

12 years agomessage: Show the file/line even when processing FATAL
Colin Walters [Mon, 17 Oct 2011 18:55:56 +0000 (14:55 -0400)]
message: Show the file/line even when processing FATAL

This helps debug.  Also, add a 'fatal' debug break.

12 years agoadd test for checking refcounting of signals with transfer none object params
John (J5) Palmieri [Fri, 14 Oct 2011 21:31:28 +0000 (17:31 -0400)]
add test for checking refcounting of signals with transfer none object params

12 years agoscanner: allow GObject.Object as a superclass return type
Dan Winship [Thu, 13 Oct 2011 16:21:22 +0000 (12:21 -0400)]
scanner: allow GObject.Object as a superclass return type

6172c268 made the scanner think GObject* wasn't a superclass of its
subclasses. Fix that.

https://bugzilla.gnome.org/show_bug.cgi?id=661677

12 years agodumper: Don't call g_thread_init(), g_type_init() is enough
Colin Walters [Wed, 12 Oct 2011 13:49:06 +0000 (09:49 -0400)]
dumper: Don't call g_thread_init(), g_type_init() is enough

g_thread_init() is deprecated, and unnecessary since GLib 2.24.

12 years agoscanner: split CC environment variable
Natanael Copa [Mon, 10 Oct 2011 13:37:55 +0000 (13:37 +0000)]
scanner: split CC environment variable

This fixes compilation where CC="ccache gcc" and similar.

https://bugzilla.gnome.org/show_bug.cgi?id=660160

Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
12 years agogiconstantinfo: Add API to fix memory leak
Jasper St. Pierre [Wed, 5 Oct 2011 20:31:43 +0000 (16:31 -0400)]
giconstantinfo: Add API to fix memory leak

https://bugzilla.gnome.org/show_bug.cgi?id=654069

12 years agoRevert "gitypelibtest: Remove test on TestStructE size"
Colin Walters [Mon, 3 Oct 2011 14:03:30 +0000 (10:03 -0400)]
Revert "gitypelibtest: Remove test on TestStructE size"

This reverts commit e17c09749c23edbd928db19a8162484ddb6299b8.

We actually want the test for sizeof(GValue).

https://bugzilla.gnome.org/show_bug.cgi?id=660338

12 years agoAdjust dlopened library name on OpenBSD so we always pick the right one.
Jasper Lievisse Adriaanse [Fri, 30 Sep 2011 06:59:12 +0000 (08:59 +0200)]
Adjust dlopened library name on OpenBSD so we always pick the right one.

https://bugzilla.gnome.org/show_bug.cgi?id=660523

12 years agogitypelibtest: Remove test on TestStructE size
Colin Walters [Fri, 30 Sep 2011 14:41:58 +0000 (10:41 -0400)]
gitypelibtest: Remove test on TestStructE size

It's redundant since we're already checking that it's equal to GValue,
and a problem is that padding may break exact assertions on the size.

https://bugzilla.gnome.org/show_bug.cgi?id=660338

12 years agoBug 660338: Account for padding in struct size check
Martin Pitt [Wed, 28 Sep 2011 08:25:26 +0000 (08:25 +0000)]
Bug 660338: Account for padding in struct size check

On some platform like armel or powerpc the compiler adds extra padding to
structs. Make the test_size_of_struct_with_array_of_anon_unions() check more
liberal by only requiring that the size of the struct is at least as big as
expected. Also use g_assert_cmpuint() to make it easier to see the actual diff.

12 years agoSkip analysis of params that have been (skip)'d
Stef Walter [Wed, 28 Sep 2011 12:01:26 +0000 (14:01 +0200)]
Skip analysis of params that have been (skip)'d

Otherwise we get warnings for params that have been skipped

https://bugzilla.gnome.org/show_bug.cgi?id=660352

12 years agorepository: Fix g_irepository_get_c_prefix()
Colin Walters [Wed, 21 Sep 2011 17:13:45 +0000 (13:13 -0400)]
repository: Fix g_irepository_get_c_prefix()

It was returning the wrong data.

https://bugzilla.gnome.org/show_bug.cgi?id=659749

12 years agoconfigure: Release 1.30 GOBJECT_INTROSPECTION_1_30_0
Colin Walters [Tue, 20 Sep 2011 21:00:31 +0000 (17:00 -0400)]
configure: Release 1.30

12 years agoFix incorrect extension for tempfile
Dieter Verfaillie [Mon, 19 Sep 2011 18:20:35 +0000 (20:20 +0200)]
Fix incorrect extension for tempfile

I set ext to None in commit f2a70843905839a1b11795310d2aa3b85a474e77
without realizing _generate_tempfile does a simple string formatting.
A simple test proves this does not generate an empty string:

$ python -c "f = '%s' % None; print f, type(f)"
None <type 'str'>

So pass an empty string instead of None for the suffix parameter.

https://bugzilla.gnome.org/show_bug.cgi?id=659502

12 years agoFix linking against libpython
Dieter Verfaillie [Sat, 17 Sep 2011 11:30:28 +0000 (13:30 +0200)]
Fix linking against libpython

- AM_CHECK_PYTHON_LIBS doesn't work for 64bit sytems (lib64)
- Python extension modules do not need to be linked against
  libpython on linux anyway, but it is needed on Windows

So only run AM_CHECK_PYTHON_LIBS in configure.ac for the windows
case and make it clear in Makefile-giscanner.am linking
against libpython is a windows only thing.

https://bugzilla.gnome.org/show_bug.cgi?id=658914

12 years agoadd tests for arrays of simple structs and arrays of gvalue
John (J5) Palmieri [Thu, 15 Sep 2011 18:49:13 +0000 (14:49 -0400)]
add tests for arrays of simple structs and arrays of gvalue

 * for now we assume simple struct arrays and gvalue arrays are flat
   until gi gets support for annotating the level of indirection
   of an array.

12 years agoAdd constant value annotation
Johan Dahlin [Tue, 13 Sep 2011 14:57:28 +0000 (11:57 -0300)]
Add constant value annotation

Add an annotation tag "Value:" which can be used on
constants to override the value.

12 years agotests: Add gir/ to toplevel typelib path
Colin Walters [Tue, 13 Sep 2011 14:10:24 +0000 (10:10 -0400)]
tests: Add gir/ to toplevel typelib path

Fixes the test when running uninstalled.

12 years agoscanner: Don't add redundant notify signal
Colin Walters [Mon, 12 Sep 2011 18:29:17 +0000 (14:29 -0400)]
scanner: Don't add redundant notify signal

It was actually introduced by e0fea819f0da1ca60cce3280e7e21f3d2955be3a

12 years agoFix g_type_info_is_pointer() for overriden types of arguments.
Pavel Holejsovsky [Mon, 12 Sep 2011 17:54:39 +0000 (19:54 +0200)]
Fix g_type_info_is_pointer() for overriden types of arguments.

Algorithm which detects whether argument type is pointer checks for
trailing '*' characters in c:type .gir elements.  This failed if ctype
is either 'gpointer' or 'gconstpointer'.  Add specific check for
gpointer/gconstpointer types when deducing pointerness of the type.

https://bugzilla.gnome.org/show_bug.cgi?id=658848

12 years agoscanner: Only add notify signal to Object, not ParamSpec etc.
Colin Walters [Mon, 12 Sep 2011 18:10:54 +0000 (14:10 -0400)]
scanner: Only add notify signal to Object, not ParamSpec etc.

12 years agoscanner: correctly handle structs with arrays of anon unions
Torsten Schönfeld [Sat, 10 Sep 2011 14:52:51 +0000 (16:52 +0200)]
scanner: correctly handle structs with arrays of anon unions

This applies mainly to GValue, which is defined as:

  struct _GValue
  {
    /*< private >*/
    GType g_type;

    /* public for GTypeValueTable methods */
    union {
      gint v_int;
      guint v_uint;
      glong v_long;
      gulong v_ulong;
      gint64      v_int64;
      guint64     v_uint64;
      gfloat v_float;
      gdouble v_double;
      gpointer v_pointer;
    } data[2];
  };

Previously, the scanner did not understand the array of unions.  This
resulted in g_struct_info_get_size returning an incorrect size for
GValue (at least on 32bit systems).

Fix this by making up a separate union declaration for the GIR that can
be referenced by the array.

https://bugzilla.gnome.org/show_bug.cgi?id=657040

12 years agoscanner: Add "notify" signal to GObject
Colin Walters [Fri, 9 Sep 2011 22:07:19 +0000 (18:07 -0400)]
scanner: Add "notify" signal to GObject

For gjs we want to switch to using introspection data for signals, and
the "notify" signal being missing from GObject was a problem.

12 years agoWindows port: additionally set PATH in common.mk
Dieter Verfaillie [Mon, 5 Sep 2011 14:51:30 +0000 (16:51 +0200)]
Windows port: additionally set PATH in common.mk

Windows doesn't know about LPATH, but PATH is part
of it's .dll search order.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoRevert "Windows port: Make g-ir-annotiotion-tool, g-ir-doc-tool and g-ir-scanner...
Dieter Verfaillie [Wed, 7 Sep 2011 20:34:38 +0000 (22:34 +0200)]
Revert "Windows port: Make g-ir-annotiotion-tool, g-ir-doc-tool and g-ir-scanner 'relocatable' at runtime."

It's a bit too soon for this one, misunderstood review on irc.
Apologies for the mess!

This reverts commit 0102c517c44d3e8fc3baf2394cb92281511941e3.

12 years agoWindows port: Make g-ir-annotiotion-tool, g-ir-doc-tool and g-ir-scanner 'relocatable...
Dieter Verfaillie [Mon, 5 Sep 2011 19:49:05 +0000 (21:49 +0200)]
Windows port: Make g-ir-annotiotion-tool, g-ir-doc-tool and g-ir-scanner 'relocatable' at runtime.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: g-ir-scanner: Accept -m flags
Dieter Verfaillie [Mon, 5 Sep 2011 19:54:18 +0000 (21:54 +0200)]
Windows port: g-ir-scanner: Accept -m flags

Some pkgconfig files contain these flags on Windows, for example
gtk+-3.0.pc has -mms-bitfields in it's Cflags. Nothing is done yet
with these though, we only accept these flags for now...

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: disable barapp test program on Windows
Dieter Verfaillie [Mon, 5 Sep 2011 19:47:35 +0000 (21:47 +0200)]
Windows port: disable barapp test program on Windows

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: set CAIRO_LIBS in LIBADD, not LDFLAGS
Dieter Verfaillie [Mon, 5 Sep 2011 19:47:08 +0000 (21:47 +0200)]
Windows port: set CAIRO_LIBS in LIBADD, not LDFLAGS

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: Export all symbols on Windows.
Dieter Verfaillie [Mon, 5 Sep 2011 19:34:57 +0000 (21:34 +0200)]
Windows port: Export all symbols on Windows.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: the tmpfile is going to be an exe...
Dieter Verfaillie [Mon, 5 Sep 2011 19:31:50 +0000 (21:31 +0200)]
Windows port: the tmpfile is going to be an exe...

... on Windows, so take care of the extension.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: dont't build gi-dump-types...
Dieter Verfaillie [Mon, 5 Sep 2011 16:48:10 +0000 (18:48 +0200)]
Windows port: dont't build gi-dump-types...

... on Windows, as it uses gio-unix.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: Do not append '/usr/share' to _xdg_data_dirs...
Dieter Verfaillie [Mon, 5 Sep 2011 15:49:45 +0000 (17:49 +0200)]
Windows port: Do not append '/usr/share' to _xdg_data_dirs...

... on Windows as it points to the MinGW installation directory,
which doesn't have any .gir files to start with anyway.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: compute GOBJECT_INTROSPECTION_LIBDIR at runtime.
Dieter Verfaillie [Mon, 5 Sep 2011 15:19:39 +0000 (17:19 +0200)]
Windows port: compute GOBJECT_INTROSPECTION_LIBDIR at runtime.

Otherwise, we fail to properly locate the typelibs, because on Windows
the value of GOBJECT_INTROSPECTION_LIBDIR depends on where Glib has been
installed. Due to the nature of how we handle software that depends on
Glib on Windows (it is recommended that each program bundles it's private
copy), we're working in a "multi-prefixed" environment. Hence the value
computed at build time will most likely not even exist at runtime.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: cmph_time.h includes sys/resource.h which is not available on Windows.
Dieter Verfaillie [Mon, 5 Sep 2011 15:00:37 +0000 (17:00 +0200)]
Windows port: cmph_time.h includes sys/resource.h which is not available on Windows.

So only include it when WIN32 is not defined.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: Work arount MSYS weirdness where it changes --libtool= command line...
Dieter Verfaillie [Mon, 5 Sep 2011 15:39:40 +0000 (17:39 +0200)]
Windows port: Work arount MSYS weirdness where it changes --libtool= command line arguments.

This continues to reuse the LIBTOOL variable from automake if it's set,
but works around some MSYS weirdness: When running g-ir-scanner, MSYS
changes a command-line argument --libtool="/bin/sh ../../libtool" into
--libtool=c:/opt/msys/1.0/bin/libtool. So just use sh.exe without path
because we already "know" where the libtool configure produced is.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: Simplify _giscanner's .pyd file extension handling.
Dieter Verfaillie [Mon, 5 Sep 2011 13:54:14 +0000 (15:54 +0200)]
Windows port: Simplify _giscanner's .pyd file extension handling.

By simply using -shrext ".pyd".

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: set CAIRO_SHARED_LIBRARY in configure.ac
Dieter Verfaillie [Mon, 5 Sep 2011 13:53:15 +0000 (15:53 +0200)]
Windows port: set CAIRO_SHARED_LIBRARY in configure.ac

Both when x$have_cairo_gobject = xyes and x$have_cairo = xyes.

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: check for OS_WIN32 in configure.ac
Dieter Verfaillie [Mon, 5 Sep 2011 13:52:21 +0000 (15:52 +0200)]
Windows port: check for OS_WIN32 in configure.ac

Without repeated output created by AC_MSG_CHECKING([for Win32])

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agoWindows port: Use an improved python.m4
Dieter Verfaillie [Mon, 5 Sep 2011 13:48:26 +0000 (15:48 +0200)]
Windows port: Use an improved python.m4

- AM_CHECK_PYTHON_HEADERS macro now allows PYTHON_INCLUDES to be overridden
  from an environment variable and
- the new AM_CHECK_PYTHON_LIBS macro to check for ability to link against
  libpython. This also allows PYTHON_LIBS and PYTHON_LIB_LOC to be overridden
  from their respective environment variables.

This allows gobject-introspection to be built with MinGW/MSYS by doing:

PYTHON_DIR="/c/Python27"
SRC_DIR="/d/dev/gnome.org/checkout/gobject-introspection/src"
...
PYTHON_INCLUDES="-I${PYTHON_DIR}/include/" \
PYTHON_LIBS="-L${PYTHON_DIR}/libs/ -lpython${PYTHON_VERSION}" \
PYTHON_LIB_LOC="${PYTHON_DIR}/libs/" \
"${SRC_DIR}/configure" \

https://bugzilla.gnome.org/show_bug.cgi?id=620566

12 years agocommon.mk: Drop incorrect $(EXEEXT) extension on scripts
Colin Walters [Sat, 3 Sep 2011 18:46:51 +0000 (14:46 -0400)]
common.mk: Drop incorrect $(EXEEXT) extension on scripts

It only applies to native binaries.

12 years agoLet the user override the default compiler with the CC env variable
Andoni Morales Alastruey [Sat, 3 Sep 2011 17:53:49 +0000 (19:53 +0200)]
Let the user override the default compiler with the CC env variable

12 years agoFix compilation with mingw
Andoni Morales Alastruey [Sat, 3 Sep 2011 10:49:22 +0000 (12:49 +0200)]
Fix compilation with mingw

grealpath.h defines GetFullPathNameA() as windows.h is
not imported, but for gitscanner.c, windows.h is imported and the compiler
throws an error.

12 years agowindows: Use the real python version instead of an hardcoded one
Andoni Morales Alastruey [Sat, 3 Sep 2011 10:44:41 +0000 (12:44 +0200)]
windows: Use the real python version instead of an hardcoded one

12 years agowindows: Fix extension of g-ir-compiler to include EXE
Andoni Morales Alastruey [Sat, 3 Sep 2011 14:23:20 +0000 (16:23 +0200)]
windows: Fix extension of g-ir-compiler to include EXE

12 years agogi-dump-types: New uninstalled debugging program
Colin Walters [Sat, 3 Sep 2011 16:34:29 +0000 (12:34 -0400)]
gi-dump-types: New uninstalled debugging program

Usage: ./_build/gi-dump-types g_object_get_type

12 years agogirffi: Remove unnecessary sys/mman.h include
Colin Walters [Sat, 3 Sep 2011 16:02:54 +0000 (12:02 -0400)]
girffi: Remove unnecessary sys/mman.h include

While we're here move config.h to the top for consistency.

12 years agoregress: Fix wrong type of variable in regression test
Vincent Untz [Fri, 2 Sep 2011 16:40:54 +0000 (18:40 +0200)]
regress: Fix wrong type of variable in regression test

https://bugzilla.gnome.org/show_bug.cgi?id=658075

12 years agoRegenerated GLib annotation cache
Pavel Holejsovsky [Wed, 31 Aug 2011 11:05:22 +0000 (13:05 +0200)]
Regenerated GLib annotation cache

Regenerated from glib fe4fc3e8b5a5ad8d4113c4df1fe8e0e9f295035e