gtk4.git
9 years agogdk: Drop support for native (and thus foreign) subwindows
Alexander Larsson [Thu, 15 Dec 2016 16:53:08 +0000 (17:53 +0100)]
gdk: Drop support for native (and thus foreign) subwindows

We can't really support these on e.g. wayland anyway, and we're trying
to get rid of subwindow at totally in the long term, so lets drop this.
It allows us to drop a lot of complexity.

9 years agowayland: apply empty input shape on parent commit
Olivier Fourdan [Wed, 16 Nov 2016 14:05:43 +0000 (15:05 +0100)]
wayland: apply empty input shape on parent commit

For subsurfaces, the new state which includes the input shape is not
applied by the compositor if the subsurface is in effective synchronous
mode.

So we need to apply the input shape once parent surface is in effective
desynchronized mode, which is when it's committed, otherwise the input
shape may never be applied if the widget is not using being_paint() /
end_paint() to draw on its subsurface, like clutter does.

We do that only for empty input shape as those won't need update when
the subsurface is resized, for all other non-empty input shape, the
client still has to use begin_paint()/end_paint() for the input shape to
be applied.

https://bugzilla.gnome.org/show_bug.cgi?id=774534

9 years agoUpdate Kazakh translation
Baurzhan Muftakhidinov [Wed, 14 Dec 2016 11:13:24 +0000 (11:13 +0000)]
Update Kazakh translation

9 years agoAdd a css parser test for border-spacing
Matthias Clasen [Sun, 11 Dec 2016 01:41:07 +0000 (20:41 -0500)]
Add a css parser test for border-spacing

9 years agoRemove -gtk-gradient from css parser tests
Matthias Clasen [Sun, 11 Dec 2016 01:40:48 +0000 (20:40 -0500)]
Remove -gtk-gradient from css parser tests

It is not supported anymore.

9 years agoDocument the border-spacing CSS property
Matthias Clasen [Sun, 11 Dec 2016 01:34:24 +0000 (20:34 -0500)]
Document the border-spacing CSS property

9 years agogrid: Implement support for CSS border-spacing
Benjamin Otte [Sat, 10 Dec 2016 03:32:19 +0000 (04:32 +0100)]
grid: Implement support for CSS border-spacing

9 years agoboxgadget: Implement CSS border-spacing
Benjamin Otte [Sat, 10 Dec 2016 03:19:04 +0000 (04:19 +0100)]
boxgadget: Implement CSS border-spacing

9 years agobox: Implement support for border-spacing
Benjamin Otte [Sat, 10 Dec 2016 02:47:53 +0000 (03:47 +0100)]
box: Implement support for border-spacing

9 years agobox: Remove unused member variable
Benjamin Otte [Sat, 10 Dec 2016 02:28:05 +0000 (03:28 +0100)]
box: Remove unused member variable

9 years agocss: Add "border-spacing" CSS property
Benjamin Otte [Sat, 10 Dec 2016 02:26:24 +0000 (03:26 +0100)]
css: Add "border-spacing" CSS property

It's using a GtkCssPositionValue, even though that name is wrong. But
the functionality of managing 2 lengths is exactly what we want.

Nobody is using this yet.

9 years agoUse the new defines
Matthias Clasen [Fri, 9 Dec 2016 20:51:25 +0000 (15:51 -0500)]
Use the new defines

It is no longer GDK_WINDOWING_VULKAN.

9 years agogtk: Remove unneeded gtk_snapshot_push()/pop() calls
Benjamin Otte [Fri, 9 Dec 2016 20:38:35 +0000 (21:38 +0100)]
gtk: Remove unneeded gtk_snapshot_push()/pop() calls

9 years agosnapshot: Always create an initial node
Benjamin Otte [Fri, 9 Dec 2016 20:33:38 +0000 (21:33 +0100)]
snapshot: Always create an initial node

This way we ensure that there is only one root node and everybody can
append() with impunity.

9 years agogsk: Improve GSK_RENDERER env var handling
Benjamin Otte [Fri, 9 Dec 2016 20:18:18 +0000 (21:18 +0100)]
gsk: Improve GSK_RENDERER env var handling

- Recognize "gl" as well as "opengl" for the GL renderer
- GSK_RENDERER=help now works
- g_warning() for an unrecognized renderer (typo detection!)
- g_print() the actual renderer that is used (and error messages when
  selecting) when a GSK_RENDERER is given, so you'll notice if your
  renderer isn't taken.

9 years agobuild: Don't try to detect vulkan library if the header wasn't found.
Benjamin Otte [Fri, 9 Dec 2016 20:17:27 +0000 (21:17 +0100)]
build: Don't try to detect vulkan library if the header wasn't found.

We've already failed. No need to fail again.

9 years agoTrivial doc fixups
Matthias Clasen [Fri, 9 Dec 2016 19:59:46 +0000 (14:59 -0500)]
Trivial doc fixups

9 years agobuild: Make sure GTK compiles without Vulkan headers installed
Benjamin Otte [Fri, 9 Dec 2016 19:21:18 +0000 (20:21 +0100)]
build: Make sure GTK compiles without Vulkan headers installed

Previously, code would work fine with --disable-vulkan if the Vulkan
headers were installed - code would happily just use them as they're
installed in /usr/include.

9 years agoForgotten file
Matthias Clasen [Fri, 9 Dec 2016 19:22:14 +0000 (14:22 -0500)]
Forgotten file

9 years agoRename GDK_WINDOWING_VULKAN
Matthias Clasen [Fri, 9 Dec 2016 19:11:37 +0000 (14:11 -0500)]
Rename GDK_WINDOWING_VULKAN

The WINDOWING defines are only for GDK backends, really.
Define GDK_RENDERING_VULKAN instead.

9 years agoOnly include vulkan.h if we have it
Matthias Clasen [Fri, 9 Dec 2016 19:05:26 +0000 (14:05 -0500)]
Only include vulkan.h if we have it

9 years agoFix the configure check for Vulkan
Matthias Clasen [Fri, 9 Dec 2016 19:00:48 +0000 (14:00 -0500)]
Fix the configure check for Vulkan

GDK_WINDOWING_VULKAN was getting defined regardless of the tests.

9 years agovulkan: Implement texture caching
Benjamin Otte [Fri, 9 Dec 2016 06:23:04 +0000 (07:23 +0100)]
vulkan: Implement texture caching

And with this change, the GPU looks bored again.

9 years agovulkan: Don't create more than one render object
Benjamin Otte [Fri, 9 Dec 2016 05:13:14 +0000 (06:13 +0100)]
vulkan: Don't create more than one render object

By creating unlimited render objects, we would never wait on the GPU.

This would mean that if the GPU was the bottleneck, we would fill its
queue with render commands faster than it could process them.

And because the nvidia binary driver and my code work surprisingly well
and bugfree, this lead to exhaustion of RAM. I had 50GB of swap
configured and my hard disk was quicker as swap storage than my GPU was
at processing the commands, so stuff still filled up.

At that point my computer became rather unresponsive and I decided to
reboot it, so I that could write this patch.

9 years agovulkan: Don't limit number of descriptor sets
Benjamin Otte [Fri, 9 Dec 2016 05:09:52 +0000 (06:09 +0100)]
vulkan: Don't limit number of descriptor sets

If we need more than we have reserved as maximum, recreate the
descriptor pool with a higher maximum.

9 years agovulkan: Add more node operations
Benjamin Otte [Fri, 9 Dec 2016 04:59:19 +0000 (05:59 +0100)]
vulkan: Add more node operations

Add SURFACE and TEXTURE operations. This way, we actually render more
than one node every frame because not everything is a fallback node
anymore that gets composited with its children into a cairo surface.

9 years agovulkan: Push the correct matrix when drawing
Benjamin Otte [Fri, 9 Dec 2016 04:53:04 +0000 (05:53 +0100)]
vulkan: Push the correct matrix when drawing

Instead of pushing the root matrix, push the world matrix for the
current node. That way, the bounds we emit as vertices are actually
properly transformed.

9 years agovulkan: Enable alpha blending
Benjamin Otte [Fri, 9 Dec 2016 04:51:51 +0000 (05:51 +0100)]
vulkan: Enable alpha blending

We will need that once we actually composite stuff.

9 years agovulkan: Redo descriptor set handling
Benjamin Otte [Fri, 9 Dec 2016 01:55:47 +0000 (02:55 +0100)]
vulkan: Redo descriptor set handling

First, we collect all the info about descriptor sets into a hash table,
then we use its size to determine the amount of sets and allocate those
before we finally go ahead and use the hash table's contents to
initialize the descriptor sets.

And then we're ready to render.

9 years agovulkan: Move remaining structs to VulkanRender object
Benjamin Otte [Thu, 8 Dec 2016 23:06:59 +0000 (00:06 +0100)]
vulkan: Move remaining structs to VulkanRender object

It now gets to maintain things.

9 years agovulkan: Create framebuffers from GskVulkanRender object
Benjamin Otte [Thu, 8 Dec 2016 21:35:16 +0000 (22:35 +0100)]
vulkan: Create framebuffers from GskVulkanRender object

Also create them on-demand, as they need to be created per-image and
per-framebuffer, so we don't want to create loads of them.

9 years agovulkan: Turn swapchain image targets into GskVulkanImages
Benjamin Otte [Thu, 8 Dec 2016 20:14:41 +0000 (21:14 +0100)]
vulkan: Turn swapchain image targets into GskVulkanImages

This gives us proper refcounting here, too. And there's no longer a
difference between the different types of images.

9 years agovulkan: Store width/height of GskVulkanImage
Benjamin Otte [Thu, 8 Dec 2016 19:46:02 +0000 (20:46 +0100)]
vulkan: Store width/height of GskVulkanImage

This is to enable lazy framebuffer creation later.

9 years agovulkan: Turn GskVulkanImage into a GObject
Benjamin Otte [Thu, 8 Dec 2016 19:19:56 +0000 (20:19 +0100)]
vulkan: Turn GskVulkanImage into a GObject

This way, we can do real refcounting on them.

9 years agovulkan: Hook up image upload mechanism to env variables
Benjamin Otte [Thu, 8 Dec 2016 19:05:53 +0000 (20:05 +0100)]
vulkan: Hook up image upload mechanism to env variables

Instead of having to decie at compile-time, check if
GSK_RENDERING=staging-image was given, and if so, select the staging
image upload mode.

9 years agovulkan: Don't wait until graphics are done computing
Benjamin Otte [Thu, 8 Dec 2016 18:56:56 +0000 (19:56 +0100)]
vulkan: Don't wait until graphics are done computing

We can let the GPU do its stuff without waiting. The GPU knows what it's
doing.

Which means we now get a lot of time to spend on doing CPU things (read:
we're way better in benchmarks).

The old behavior is safer, so we want to keep it around for debugging.
It can be reenabled with GSK_RENDERING_MODE=sync.

9 years agovulkan: Keep render objects around
Benjamin Otte [Thu, 8 Dec 2016 17:40:35 +0000 (18:40 +0100)]
vulkan: Keep render objects around

That way we can reuse them. We only create a new one if the last render
operation hasn't finished executing.

9 years agovulkan: Move command pool and fence into render object
Benjamin Otte [Thu, 8 Dec 2016 17:03:05 +0000 (18:03 +0100)]
vulkan: Move command pool and fence into render object

9 years agovulkan: Make the GskVulkanRender struct private
Benjamin Otte [Thu, 8 Dec 2016 16:55:32 +0000 (17:55 +0100)]
vulkan: Make the GskVulkanRender struct private

9 years agovulkan: Make the VulkanRender object allocated
Benjamin Otte [Thu, 8 Dec 2016 16:46:08 +0000 (17:46 +0100)]
vulkan: Make the VulkanRender object allocated

That way, we can have multiple of those and keep them around for reuse.

9 years agovulkan: Also wait on the draw semaphore when present()ing
Benjamin Otte [Thu, 8 Dec 2016 16:40:07 +0000 (17:40 +0100)]
vulkan: Also wait on the draw semaphore when present()ing

That way we have a synchronization primitive at the start and at the
end.

9 years agovulkan: Add GskVulkanRenderPass
Benjamin Otte [Thu, 8 Dec 2016 12:26:36 +0000 (13:26 +0100)]
vulkan: Add GskVulkanRenderPass

And move the actual rendering code there.

A RenderPass is a collection of operations on the same target that
get executed one after another. It roughly targets VkRenderPass or
rather the subpasses of a VkRenderPass.

For now, only the infrastructure is there. No real stuff is happening.

9 years agovulkan: Add GskVulkanRender object
Benjamin Otte [Wed, 7 Dec 2016 13:50:52 +0000 (14:50 +0100)]
vulkan: Add GskVulkanRender object

This is refactoring work.

GskVulkanRender is supposed to be the global object for a render
operation, ie GskVulkanRenderer.render() will create this object for
what it does.

The object will be split into stages that perform the operations
necessary to create a drawing.

9 years agovulkan: Make MVP matrix available to vertex shader
Benjamin Otte [Wed, 7 Dec 2016 00:48:31 +0000 (01:48 +0100)]
vulkan: Make MVP matrix available to vertex shader

We use push constants for this.

9 years agovulkan: Upload image data directly
Benjamin Otte [Tue, 6 Dec 2016 22:29:38 +0000 (23:29 +0100)]
vulkan: Upload image data directly

Instead of using a staging iamge, we require the final image to be
linearly allocated and have host-visible memory.

This improves performance quite a bit.

The old code is still there and can be enabled with a simple change
to a #define in gskvulkanimage.h

9 years agovulkan: Make GskVulkanRenderer work
Benjamin Otte [Tue, 6 Dec 2016 13:42:46 +0000 (14:42 +0100)]
vulkan: Make GskVulkanRenderer work

We now create a Cairo renderer, render to an image surface and upload
and retnder it with Vulkan.

9 years agoxim: Use the native parent window for the X connection
Benjamin Otte [Mon, 5 Dec 2016 23:09:59 +0000 (00:09 +0100)]
xim: Use the native parent window for the X connection

Don't try to gdk_window_get_xid() on the current window.

9 years agox11: Don't make random windows native
Benjamin Otte [Mon, 5 Dec 2016 22:40:16 +0000 (23:40 +0100)]
x11: Don't make random windows native

Instead, complain if somebody calls gdk_x11_window_get_xid() on a
non-native window.

We cannot make random windows native anymore because there's no GSK
renderer associated with them, so we cannot draw them.

9 years agogsk: Pass texture coordinates to Vulkan renderer
Benjamin Otte [Mon, 5 Dec 2016 22:07:56 +0000 (23:07 +0100)]
gsk: Pass texture coordinates to Vulkan renderer

We just render the coordinates as color, we don't do anything with them
yet.

9 years agogsk: Loads of work on Vulkan backend
Benjamin Otte [Mon, 5 Dec 2016 01:50:06 +0000 (02:50 +0100)]
gsk: Loads of work on Vulkan backend

We can now upload vertices.

And we use this to draw a yellow background. Which is clearly superior
to not drawing anything.

Also, we have shaders now. If you modify them, you need glslc installed
so they can be recompiled into Spir-V bytecode.

9 years agoconfigure: Improve Vulkan detection
Benjamin Otte [Sun, 4 Dec 2016 17:10:18 +0000 (18:10 +0100)]
configure: Improve Vulkan detection

1. Output Vulkan status in summary
2. Add missing "test" call
3. Check for glslc

The glslc check will be necessary later for the code that automatically
compiles the Vulkan glsl source to Spir-V.
Nothing happens if glslc is not available - unless you modify the glsl.

9 years agovulkan: acquire/present images when drawing
Benjamin Otte [Thu, 1 Dec 2016 03:07:20 +0000 (04:07 +0100)]
vulkan: acquire/present images when drawing

Another step towards the final goal of actually showing something.

9 years agogskvulkan: Create render pass and command pool
Benjamin Otte [Wed, 30 Nov 2016 02:59:28 +0000 (03:59 +0100)]
gskvulkan: Create render pass and command pool

9 years agogskvulkanrenderer: Create imagevies and framebuffers
Benjamin Otte [Wed, 30 Nov 2016 01:08:17 +0000 (02:08 +0100)]
gskvulkanrenderer: Create imagevies and framebuffers

9 years agogsk: Add a vulkan debug category
Benjamin Otte [Wed, 30 Nov 2016 01:07:57 +0000 (02:07 +0100)]
gsk: Add a vulkan debug category

9 years agovulkan: Expose the swapchain images in public API
Benjamin Otte [Tue, 29 Nov 2016 23:54:48 +0000 (00:54 +0100)]
vulkan: Expose the swapchain images in public API

9 years agogsk: Add skeleton for Vulkan renderer
Benjamin Otte [Tue, 29 Nov 2016 15:33:14 +0000 (16:33 +0100)]
gsk: Add skeleton for Vulkan renderer

The renderer itself obviously doesn't do anything.

9 years agovulkan: Register debug handler
Benjamin Otte [Tue, 29 Nov 2016 15:29:19 +0000 (16:29 +0100)]
vulkan: Register debug handler

We now hook Vulkan error reporting into glib's error reporting.

I'm not sure on the severity handling yet, but we can fix that as we go
along.

9 years agogdk: Add GDK_VULKAN env variable
Benjamin Otte [Tue, 29 Nov 2016 14:35:04 +0000 (15:35 +0100)]
gdk: Add GDK_VULKAN env variable

Also add 2 flags:

GDK_VULKAN="disable" will disable Vulkan usage.

GDK_VULKAN="validate" will enable VK_LAYER_LUNARG_standard_validation.

9 years agox11: Call gdk_display_sync() when creating Vulkan surfaces
Benjamin Otte [Tue, 29 Nov 2016 13:47:59 +0000 (14:47 +0100)]
x11: Call gdk_display_sync() when creating Vulkan surfaces

The Vulkan driver wants to operate on an X Window that it is sure does
exist.

9 years agovulkan: More work on GdkVulkanContext
Benjamin Otte [Tue, 29 Nov 2016 02:20:31 +0000 (03:20 +0100)]
vulkan: More work on GdkVulkanContext

Code has now arrived at creation of swapchains.

9 years agovulkan: Turn GdkVulkanContext into a GdkDrawContext
Benjamin Otte [Mon, 28 Nov 2016 16:55:46 +0000 (17:55 +0100)]
vulkan: Turn GdkVulkanContext into a GdkDrawContext

9 years agovulkan: Add more infrastructure
Benjamin Otte [Mon, 28 Nov 2016 15:34:01 +0000 (16:34 +0100)]
vulkan: Add more infrastructure

gdk_window_create_vulkan_context() now exists and will return a Vulkan
context for the given window. It even initializes the surface. But it
doesn't do anything useful yet.

9 years agovulkan: Initial support
Benjamin Otte [Mon, 21 Nov 2016 13:18:43 +0000 (14:18 +0100)]
vulkan: Initial support

Adds the gdk_display_ref_vulkan() and gdk_display_unref_vulkan()
functions which setup/tear down VUlkan support for the display.

Nothing is using those functions yet.

9 years agobuild: Add Vulkan detection code
Benjamin Otte [Mon, 21 Nov 2016 00:25:17 +0000 (01:25 +0100)]
build: Add Vulkan detection code

Also define GDK_WINDOWING_VULKAN if Vulkan is enabled.

9 years agoUpdate Italian translation
Gianvito Cavasoli [Fri, 9 Dec 2016 10:34:45 +0000 (10:34 +0000)]
Update Italian translation
(cherry picked from commit 44fb5c9c704c39fad392ca45c6c2d74d6633a4e4)

9 years agomir: swap buffers on paint
William Hua [Thu, 8 Dec 2016 20:28:31 +0000 (15:28 -0500)]
mir: swap buffers on paint

9 years agowidget: Remove unused adjust_* vfuncs
Timm Bäder [Wed, 23 Nov 2016 16:46:44 +0000 (17:46 +0100)]
widget: Remove unused adjust_* vfuncs

Expose them privately to gtksizerequest.c using normal functions
instead.

9 years agomenu: Replace button-press-handler with gesture
Timm Bäder [Fri, 2 Dec 2016 12:20:47 +0000 (13:20 +0100)]
menu: Replace button-press-handler with gesture

9 years agotoolbar: Replace button-press-event handler with gesture
Timm Bäder [Fri, 2 Dec 2016 11:39:46 +0000 (12:39 +0100)]
toolbar: Replace button-press-event handler with gesture

9 years agolinkbutton: Replace button-press-event handler with gesture
Timm Bäder [Fri, 2 Dec 2016 11:22:41 +0000 (12:22 +0100)]
linkbutton: Replace button-press-event handler with gesture

9 years agoseparatortoolitem: Remove button-event/motion-event handlers
Timm Bäder [Fri, 2 Dec 2016 10:34:56 +0000 (11:34 +0100)]
separatortoolitem: Remove button-event/motion-event handlers

These only exist for the window dragging which does not exist anymore
currently. It will be reintroduced later in a form that does not require
these handlers.

9 years agosnapshot: Add annotations to parameters/return values
Timm Bäder [Thu, 1 Dec 2016 09:19:24 +0000 (10:19 +0100)]
snapshot: Add annotations to parameters/return values

9 years agowidget: Use gtk_widget_measure to measure widget sizes
Timm Bäder [Mon, 5 Dec 2016 17:19:04 +0000 (18:19 +0100)]
widget: Use gtk_widget_measure to measure widget sizes

9 years agowidget: Remove _gtk_widget_supports_clip
Timm Bäder [Tue, 22 Nov 2016 11:08:07 +0000 (12:08 +0100)]
widget: Remove _gtk_widget_supports_clip

9 years agowidget: Remove gtk_widget_invalidate_style_context
Timm Bäder [Tue, 22 Nov 2016 11:05:48 +0000 (12:05 +0100)]
widget: Remove gtk_widget_invalidate_style_context

9 years agowayland: Sync attributes also when drawing with GL
Benjamin Otte [Tue, 6 Dec 2016 17:01:34 +0000 (18:01 +0100)]
wayland: Sync attributes also when drawing with GL

Probably syncing attributes shouldn't happen when drawing in the first
place, but what do I know about Wayland. ¯\_(ツ)_/¯

9 years agowayland: Uncomment erroneously commented out code
Benjamin Otte [Tue, 6 Dec 2016 16:51:35 +0000 (17:51 +0100)]
wayland: Uncomment erroneously commented out code

I read the code as if (use_gl) instead of if (!use_gl) and commented it
out in bddfd7bb41cebd9b4e8ea418ce4d8e1f11822e81. That broke drawing on
Wayland without OpenGL completely.

Whoops.

Now it's back.

9 years agoGDK/Win32: Fix build after GDKGL refactoring
Chun-wei Fan [Tue, 6 Dec 2016 06:37:42 +0000 (14:37 +0800)]
GDK/Win32: Fix build after GDKGL refactoring

There were some parts that need some updates after the refactoring in
GDKGL, so that the code will continue to build and run.

For gdkwindow-win32.c, comment out the parts where we check for use_gl
(which was removed), since we are going to move all drawing to OpenGL,
but don't remove/disable the whole portion as that transition is not
complete at this point.

There a is new GDKGL function that checks for the damaged area of the back
buffer, but since the notion of "damage" is for *NIX (GLX/EGL for
Wayland/mir), meaning that there is no such extension for Windows in this
regard, so we can't support this on Windows as-is, at least for now.

https://bugzilla.gnome.org/show_bug.cgi?id=773299

9 years agoUpdate POTFILES.in
Piotr Drąg [Mon, 5 Dec 2016 18:46:59 +0000 (19:46 +0100)]
Update POTFILES.in

9 years agobuild: add gir depends for GdkX11, GdkWin32, and Gsk on Gdk
William Hua [Mon, 5 Dec 2016 16:37:20 +0000 (11:37 -0500)]
build: add gir depends for GdkX11, GdkWin32, and Gsk on Gdk

https://bugzilla.gnome.org/show_bug.cgi?id=775651

9 years agomir: fix type warning
William Hua [Mon, 5 Dec 2016 16:06:32 +0000 (11:06 -0500)]
mir: fix type warning

9 years agomir: fix build failures
William Hua [Mon, 5 Dec 2016 15:28:34 +0000 (10:28 -0500)]
mir: fix build failures

9 years agomir: implement display monitor vfuncs
William Hua [Tue, 29 Nov 2016 17:38:47 +0000 (12:38 -0500)]
mir: implement display monitor vfuncs

9 years agogdkgl: Add gdk_gl_context_get_damage()
Benjamin Otte [Sun, 4 Dec 2016 15:33:13 +0000 (16:33 +0100)]
gdkgl: Add gdk_gl_context_get_damage()

This is a way to query the damaged area of the backbuffer.

The GL renderer uses this to compute the extents of that damage region
(computed via buffer age) and use them to minimize the area to redraw.

This changes the semantics of GL rendering to "When calling
gdk_window_begin_frame() with a GL context, the area by
gdk_gl_context_get_damage() needs to be redrawn and every other pixel of
the backbuffer is guaranteed to be correct.
After gdk_window_end_frame() on a GL-drawn window, the whole backbuffer
must be correct.

We can always glXBufferSwap() now because of this.

9 years agoinspector: Add support for displayin the render region
Benjamin Otte [Sun, 4 Dec 2016 15:32:38 +0000 (16:32 +0100)]
inspector: Add support for displayin the render region

Now I just need a UI designer that makes this look nice...

9 years agoglrenderer: Always redraw the whole screen
Benjamin Otte [Sun, 27 Nov 2016 15:37:31 +0000 (16:37 +0100)]
glrenderer: Always redraw the whole screen

9 years agogdk: Make gdk_window_begin_draw_frame() take a draw context
Benjamin Otte [Thu, 1 Dec 2016 00:38:20 +0000 (01:38 +0100)]
gdk: Make gdk_window_begin_draw_frame() take a draw context

... instead of a gl context.

This requires some refactoring in the way we mark the shared context as
drawing: We now call begin_frame/end_frame() on it and ignore the call
on the main context.
Unfortunately we need to do this check in all vfuncs, which sucks. But I
haven't found a better way.

9 years agogskrenderer: Add gsk_renderer_begin_draw_frame()
Benjamin Otte [Sun, 27 Nov 2016 15:21:02 +0000 (16:21 +0100)]
gskrenderer: Add gsk_renderer_begin_draw_frame()

This way, we can hijack the begin/end draw process and do out own
processing before passing it on to GDK.

9 years agoinspector: Pass the drawing context into the recorder
Benjamin Otte [Wed, 30 Nov 2016 17:55:45 +0000 (18:55 +0100)]
inspector: Pass the drawing context into the recorder

That way we can capture both the actual changes (clip region) and the
area that was redrawn (render region), which in OpenGL might not be
identical.

Nothing shows the render region yet though...

9 years agoglrenderer: Always redraw the whole screen
Benjamin Otte [Sun, 27 Nov 2016 15:37:31 +0000 (16:37 +0100)]
glrenderer: Always redraw the whole screen

9 years agogsk: Change GSK_USE_SOFTWARE env var to GSK_RENDERER
Benjamin Otte [Tue, 29 Nov 2016 12:36:31 +0000 (13:36 +0100)]
gsk: Change GSK_USE_SOFTWARE env var to GSK_RENDERER

This way, we can use the variable as more than a boolean flag.

This will be useful for Vulkan or D3D renderers.

9 years agogdk: Turn GdkGLContext into a GdkDrawContext
Benjamin Otte [Mon, 28 Nov 2016 16:36:37 +0000 (17:36 +0100)]
gdk: Turn GdkGLContext into a GdkDrawContext

9 years agogdk: Add GdkDrawContext
Benjamin Otte [Mon, 28 Nov 2016 16:22:35 +0000 (17:22 +0100)]
gdk: Add GdkDrawContext

This will be the base class for GdkVulkanContext and GdkGLContext.

9 years agogdk: Redo GL drawing
Benjamin Otte [Wed, 23 Nov 2016 04:54:50 +0000 (05:54 +0100)]
gdk: Redo GL drawing

Reenable GL drawing, but do it without Cairo.

Now, the context passed to gdk_window_begin_draw_frame() decides how
drawing is going to happen. If it is NULL, Cairo is used like before.

If a context is passed, Cairo may not be used for drawing and
gdk_drawing_context_get_cairo_context() is going to return NULL.
Instead, the GL renderer must draw to the GL backbuffer and
end_draw_frame() is then swapping that to the front.

The GskGLRenderer has lost the texture it used to render to and adapted
to render directly to the backbuffer instead.

The only thing missing is for GtkGLArea to gain back a performant way to
render. But it didn't have one since the introduction of GSK, this
patchset doesn't change anything about it.

The new rendering avoids two indirections (the GSK renderer's texture
and the GDK double buffering surface).
It improves icon count in the fishbowl demo by 30%.

9 years agoglcontext: Make begin/end_draw() paired
Benjamin Otte [Wed, 23 Nov 2016 04:18:43 +0000 (05:18 +0100)]
glcontext: Make begin/end_draw() paired

This way, we can query the GL context's state via
gdk_gl_context_is_drawing().

Use this function to make GL contexts as attached and grant them access
to the front/backbuffer for rendering.

All of this is still unused because GL drawing is still disabled.

9 years agogdk: Large GL refactoring
Benjamin Otte [Wed, 23 Nov 2016 02:32:22 +0000 (03:32 +0100)]
gdk: Large GL refactoring

No visible changes as GL rendering is disabled at the moment.

What was done:

1. Move window->invalidate_for_new_frame to glcontext->begin_frame
This moves the code to where it is used (the GLContext) and prepares it
for being called where it is used when actually beginning to draw the
frame.

2. Get rid of buffer-age usage
We want to let the application render directly to the backbuffer.
Because of that, we cannot make any assumptions about the contents the
application renders outside the clip area.
In particular GskGLRenderer renders random stuff there but not actual
contents.

3. Pass the actual GL context
Previously, we passed the shared context to end_frame, now we pass the
actual GL context that the application uses for rendering. This is so
that the vfuncs could prepare the actual contexts for rendering (they
don't currently).

4. Simplify the code
The previous code set up the final drawing method in begin_frame.
Instead, we now just ensure the clip area is something we can render
and decide on the actual method in end_frame.
This is both more robust (we can change the clip area in between if we
want to) and less code.

9 years agogsk: Allow 0 for gsk_gl_driver_begin_frame()
Benjamin Otte [Wed, 23 Nov 2016 00:03:59 +0000 (01:03 +0100)]
gsk: Allow 0 for gsk_gl_driver_begin_frame()

Special-case this situation to unset the current framebuffer and render
to the backbuffer again.

9 years agogdk: Don't recurse when processing updates
Benjamin Otte [Tue, 22 Nov 2016 20:23:21 +0000 (21:23 +0100)]
gdk: Don't recurse when processing updates

We don't need to send expose events for backwards compatibility anymore.

9 years agogdk: Remove all code that only existed because of use_gl
Benjamin Otte [Tue, 22 Nov 2016 19:03:14 +0000 (20:03 +0100)]
gdk: Remove all code that only existed because of use_gl

Now that we don't use GL anymore, this code is unnecessary.