* bootstrap-for-the-insane : restore the message of doom that was
removed during the "bootstrap-generic" reorganization.
svn path=/trunk/gtk-sharp/; revision=48744
makes it safe to create tree nodes in a worker thread as long as
you reparent them into the tree with another thread. Thanks to mk
for allowing a bit of threadedness in :-).
svn path=/trunk/gtk-sharp/; revision=48683
they or their accessors are marked deprecated. (Affects
Gtk.Entry.Editable, Gtk.FontSelection.Font,
Gtk.Notebook.*TabBorder, Gtk.Object.UserData, and a bunch of old
Gtk.ProgressBar properties).
* gtk/Gtk.metadata: Hide Entry.Editable. Mark Notebook.Homogeneous
deprecated.
* gtk/Entry.custom: Implement Editable property with an Obsolete
pointing to IsEditable.
svn path=/trunk/gtk-sharp/; revision=48680
fails, set MarshalType to "" to propagate that failure into
methods that have args of this type. [Fixes#75851]
svn path=/trunk/gtk-sharp/; revision=48679
* bootstrap-generic: New script, handles bootstrapping stuff generically
and saves bootstrap settings so we can regenerate configure.in from
configure.in.in.
* bootstrap
* bootstrap-2.4
* bootstrap-for-the-insane: Modify to use bootstrap-generic
* Makefile.am (configure.in): Add a rule to run bootstrap.status to
regenerate configure.in if configure.in.in changes.
svn path=/trunk/gtk-sharp/; revision=48412
* glib/Object.cs : hold strong refs for managed subclasses and
weakrefs for wrappers.
* gtk/Object.custom : don't hold managed refs here, they are now held
in GLib.Object.
svn path=/trunk/gtk-sharp/; revision=48389
a constructor or a ref, unref, or destroy method, then it must be
a reference type, so mark it "opaque" but then also mark all of
its fields public and writable.
* */*-api*.raw: Regen
* generator/Parser.cs (ParseNamespace): make the opaque attribute
check actually look at the value of the attribute rather than just
checking if it's there, so that you can change a struct's opaque
attribute from "true" to "false" via metadata and have that work.
* generator/BoxedGen.cs (Generate): do not generate the boxed's
"Free" method (since it's guaranteed to crash when we pass it a
stack pointer). If "Copy" is marked deprecated, create a
deprecated no-op for it, otherwise just skip it (since otherwise
it will just leak memory when we copy its result onto the stack).
* pango/Pango.metadata: deprecate Pango.Color.Copy and
Pango.Matrix.Copy. Hide some array fields in Pango.GlyphString
that we've never generated correctly. Tweak Pango.LayoutLine
fields to be the same as they used to be.
* pango/GlyphItem.custom (glyphs, item):
* pango/GlyphString.custom (Zero, New):
* pango/Item.custom (Zero, New):
* pango/LayoutRun.custom (glyphs, item): add deprecated API compat
* gdk/Gdk.metadata: undo the parser's new opaquification of
Gdk.Font; it's been deprecated since pre-gtk# times, and no one
should be using it, so there's no point in fixing it now. Fix up a
few other things to match how they used to be. Fix RgbCmap's
constructor args.
* gdk/RgbCmap.custom (Zero, New): deprecated API compat
* gdk/PangoAttrEmbossed.custom:
* gdk/PangoAttrStipple.custom (Zero, New, Attr): deprecated API
compat
(explicit operator ...): allow casting back and forth between
Pango.Attribute. (We can't usefully make them real subclasses of
Pango.Attribute, because there's no way for
Pango.Attribute.GetAttribute() to be able to dtrt with them.)
* gtk/Gtk.metadata: deprecate Gtk.Requisition.Copy,
Gtk.TextIter.Copy, and Gtk.TreeIter.Copy. Mark the return value of
TextView.DefaultAttributes as "owned". Mark TargetList's fields
private so it stays how it used to be.
* gtk/TextAttributes.custom (Zero, New): deprecated API compat
* gnomevfs/Gnomevfs.metadata: remove a bunch of opaque
declarations that the parser figures out on its own now.
* art/Art.metadata:
* glade/Glade.metadata:
* rsvg/Rsvg.metadata: un-mark everything the parser marked opaque
in these libraries, because all of the structs in question would
still be unusably broken, so the API churn would be pointless.
svn path=/trunk/gtk-sharp/; revision=48387
(Henceforth gda# and gnomedb# will be part of gda and gnomedb. Or
maybe separate modules in Mono SVN. Not part of gtk-sharp though.)
* Makefile.am (SUBDIRS): remove gda and gnomedb
* gda/, gnomedb/: buh-bye
svn path=/trunk/gtk-sharp/; revision=48200
in the context of "Gtk.TreeModel.GetNColumns"
* generator/CallbackGen.cs (GenWrapper): treat InterfaceGen return
values the same as ObjectGen
svn path=/trunk/gtk-sharp/; revision=48192
parent class and interfaces (because we don't want to see the
warnings about certain GtkWidget methods in every single library
that defines a widget, etc). Instead, use the new
ValidateForSubclass() method.
(ValidateForSubclass): only validate the signals
* generator/InterfaceGen.cs (ValidateForSubclass): for interfaces
we need to validate the methods too.
* generator/ObjectGen.cs (Generate): Check for interface method
collisions against the class's own methods too, not just its other
interfaces. Also, it's only a collision if the methods' signatures
have the same types.
* generator/Signal.cs (GenDefaultHandlerDelegate): Use
"{0}_managed" rather than "obj" for the internal variable name, to
avoid compile problems with signals that have a parameter named
"obj".
* generator/SymbolTable.cs (MangleName): mangle "internal" to
"_internal".
svn path=/trunk/gtk-sharp/; revision=48184
parent class and interfaces (because we don't want to see the
warnings about certain GtkWidget methods in every single library
that defines a widget, etc). Instead, use the new
ValidateForSubclass() method.
(ValidateForSubclass): only validate the signals
* generator/InterfaceGen.cs (ValidateForSubclass): for interfaces
we need to validate the methods too.
svn path=/trunk/gtk-sharp/; revision=48177
a bit; only Ref the pointer if Owned is false (and then set it to
true), and vice versa for Unref.
* glib/Opaque.cs (Opaque): set owned before setting Raw, so
that Raw will be reffed properly.
(GetOpaque): Fix this up to dtrt in all cases with refcounted
opaques.
* gtk/TreeView.custom (GetPathAtPos): Use "GetOpaque(...)" rather
than "new TreePath()"
* sample/opaquetest/*: regression test for opaque free/ref/unref
handling
* sample/Makefile.am (SUBDIRS): add opaquetest
* configure.in.in (AC_OUTPUT): add opaquetest files
svn path=/trunk/gtk-sharp/; revision=48168
properties can't be set until Validate-time (eg, Method.IsGetter),
but it's annoying for every potential user of those properties to
have to make sure it has Validated the generatable first. So now
we add an explicit Validate() step after everything is loaded but
before anything is Generated, so that at Generation time,
everything can be assumed to have been Validated.
* generator/IGeneratable.cs: add "bool Validate()"
* generator/CodeGenerator.cs (Main): after loading all of the
generatables, DeAlias the SymbolTable, Validate() all the
generatables, and discard any invalid ones.
* generator/*.cs: Implement Validate() trivially in generatables
that didn't implement it before. Move Validate() calls from
Generate() to Validate(). Remove non-hierarchical Validate()
calls.
* generator/SymbolTable.cs: GPtrArray is IntPtr, not IntPtr[]
svn path=/trunk/gtk-sharp/; revision=48046
TargetEntry[], so you can use methods like
TargetList.AddTextTargets() in situations where you need a
TargetEntry[] rather than a TargetList.
* gtk/glue/targetlist.c: glue for that
svn path=/trunk/gtk-sharp/; revision=48006
* glib/Opaque.cs (Owned): new property saying whether or not gtk#
owns the memory.
(Opaque): Set Owned to true in the void ctor and false in the
IntPtr one.
(GetOpaque): add a new overload that can also create opaques, a la
GLib.Object.GetObject.
(Ref, Unref, Free): empty virtual methods to be overridden by
subclasses.
(set_Raw): Unref() and possibly Free() the old value, Ref() the
new one.
(~Opaque, Dispose): set Raw to IntPtr.Zero (triggering Free/Unref
if needed)
* parser/gapi2xml.pl (addReturnElem): if the method is named Copy
and returns a pointer, set the "owned" attribute on the
return-type.
* */*-api.raw: Regen
* generator/HandleBase.cs (FromNative): Add new
FromNative/FromNativeReturn overloads that takes a "bool owned"
param. Implement the 1-arg FromNative and FromNativeReturn in
terms of that.
* generator/ObjectBase.cs (FromNative): Implement HandleBase's new
overload. Use the two-arg version of GLib.Object.GetObject when
"owned" is true.
* generator/OpaqueGen.cs (Generate): Pull out Ref, Unref, and
Free/Destroy/Dispose methods and handle them specially by
overriding Opaque.Ref, .Unref, and .Free appropriately. (If any
of the methods are marked deprecated, output a deprecated
do-nothing method as well, to save us from having to write all
those deprecated methods by hand.)
(FromNative): use GetOpaque, passing "owned".
* generator/ReturnValue.cs (FromNative): if the value is a
HandleBase, pass Owned to its FromNative().
* generator/Parameters.cs (Owned): new property (for use on out
params)
(FromNative): Call FromNative() on the generatable, handling Owned
in the case of HandleBase.
* generator/ManagedCallString.cs:
* generator/MethodBody.cs:
* generator/Signal.cs: use param.FromNative() rather than
param.Generatable.FromNative(), to get ownership right.
* */*.metadata: Mark opaque ref/unref/free methods deprecated
(except where we were hiding them before). Add "owned" attributes
to return values and out params as needed.
* pango/AttrIterator.custom (GetFont): work around a
memory-management oddity of the underlying method.
* pango/AttrFontDesc.cs (AttrFontDesc): copy the passed-in
FontDescriptor, since the attribute will assume ownership of it.
* gtk/TreeView.custom (GetPathAtPos): set the "owned" flag on the
returned TreePaths.
* gtk/TargetList.custom: Remove refcounting stuff, which is
now handled automatically
* gtk/NodeStore.cs (GetPath): clear the Owned flag on the created
TreePath so that the underlying structure doesn't get freed when
the function returns
* gtkhtml/HTMLStream.custom (Destroy): hide this and then
reimplement it by hand to keep OpaqueGen from using it in
Dispose(), since calling it after an HTMLStream.Close() will
result in a crash.
svn path=/trunk/gtk-sharp/; revision=47928
* gtk/Gtk.metadata: Change the Gtk.Style.Paint* methods to use a
Gdk.Drawable instead of a Gdk.Window
* doc/en/Gtk/Style.xml: Update the documentation for the above
change.
svn path=/trunk/gtk-sharp/; revision=47895
* sources/gtk-sharp-2.6-sources.xml: exclude a bunch of
gnome-print stuff that is not part of gnome-print's public API.
* gnome/gnome-api.raw: regen
* gnome/Gnome.metadata: remove some no-longer-needed metadata
* gnome/GPFontEntry.custom:
* gnome/GPPath.custom: no longer needed
* gnome/FontFamily.cs: moved from FontFamily.custom, since there's
no longer any non-custom portion of this.
svn path=/trunk/gtk-sharp/; revision=47802
params... some functions use that to mark const "out" params. In
fact, let's use it as a hint to mark them pass_as="out" too...
* pango/pango-api-2.4.raw:
* pango/pango-api-2.6.raw:
* gtk/gtk-api-2.6.raw: Regen, fixing pango_script_iter_get_range
and gtk_image_get_icon_name.
* pango/Pango.metadata:
* pango/ScriptIter.cs: Alas, exposing GetRange makes it clear that
PangoScriptIter is really weird and we weren't wrapping it
correctly before anyway, so mark the whole thing hidden and wrap
it by hand.
svn path=/trunk/gtk-sharp/; revision=47748
anonymous function pointer types in method signatures. Before, we
added a <callback> child to the <parameters> node, but the
generator just ignored it. Now we add the callback (with a made-up
name) to the toplevel node, and add an ordinary <param> node
referencing it to the <parameters> node. Also, if the last param
of the callback is a gpointer, rename it from "arg#" to "data" so
it will be treated correctly (as the user data passed from the
calling method). [Fixes#66241]
* art/art-api.raw:
* gdk/gdk-api-2.4.raw:
* gdk/gdk-api-2.6.raw: Regen
* generator/Parameters.cs (IsHidden): loosen the definition of
hideable user_data; it doesn't have to occur at the end of the
parameter list, as long as there's a callback arg before it.
* generator/MethodBody.cs (GetCallString): Use Parameters.IsHidden
to decide whether or not to squash user_data params, rather than
trying to duplicate its logic. As a side effect, this also causes
a handful of methods that take non-hidden IntPtr arguments to
start actually passing those arguments to C rather than always
passing IntPtr.Zero.
* generator/Method.cs (Equals, GetHashCode): Remove unnecessary
and possibly erroneous hashing overrides.
* gtk/Gtk.metadata: Hide Gtk.Container.ForeachFull, since it's
useless and wasn't in gtk# 1.0
* gtk/Menu.custom (Popup):
* gtk/TextIter.custom (ForwardFindChar, BackwardFindChar):
* gnome/App.custom (CreateMenusInterp, InsertMenusInterp,
CreateToolbarInterp):
* gnome/Client.custom (RequestInteractionInterp):
* gnome/Popup.custom (MenuDoPopupModal, MenuDoPopup): Add
[Obsolete] compat overloads for methods that have now lost a
useless IntPtr.
svn path=/trunk/gtk-sharp/; revision=47566
