If _gdk_macos_surface_move_resize() was called with various -1 parameters
we really want to avoid changing anything even if we think we know what
the value might be. Otherwise, we risk messing up in-flight operations that
we have not yet been notified of yet.
This improves the chances we place windows in an appropriate location as
they don't et screwed up before window-manager placement.
Previously, a single CVDisplayLink was used to drive updates for all
surfaces across all monitors. It used a 'best guess' rate which would
allow for updates across monitors of mixed rates. This is undesirable for
situations where you might have a 144hz monitor as it does not allow for
reaching up to that frame rate.
Instead, we want to use a per-monitor CVDisplayLink which will fire at the
rate of the monitor down to the level of updates we require. This commit
does just that.
When a surface crosses onto a new monitor, that monitor is used to drive
the GdkFrameClock.
Fixes#4732
Windows can end up on different monitors despite having a parent or
transient-for ancestor. We want them to be driven by the CVDisplayLink
for the best-monitor, and so this needs to be unshared.
This broke with the previous fixes for initial window positioning. We need
the initial positioning so that tails will be displayed correctly when the
popover surface is displayed.
It looks like, particularly on the M1, we might need to double buffer the
contents of the IOSurface<->OpenGL texture bindings. This doesn't appear
to show up on the Intel macbooks I've tried, but I've seen it in the wild
on an M1.
If we have a 2x scale laptop with a 1x scale external display, we would
need to create a new IOSurface for the external display once it crosses
a boundary, otherwise we won't have something capable of displaying
correctly on the second monitor.
This provides a major shift in how we draw both when accelerated OpenGL
as well as software rendering with Cairo. In short, it uses tiles of Core
Animation's CALayer to display contents from an OpenGL or Cairo rendering
so that the window can provide partial damage updates. Partial damage is
not generally available when using OpenGL as the whole buffer is flipped
even if you only submitted a small change using a scissor rect.
Thankfully, this speeds up Cairo rendering a bit too by using IOSurface to
upload contents to the display server. We use the tiling system we do for
OpenGL which reduces overall complexity and differences between them.
A New Buffer
============
GdkMacosBuffer is a wrapper around an IOSurfaceRef. The term buffer was
used because 1) surface is already used and 2) it loosely maps to a
front/back buffer semantic.
However, it appears that IOSurfaceRef contents are being retained in
some fashion (likely in the compositor result) so we can update the same
IOSurfaceRef without flipping as long as we're fast. This appears to be
what Chromium does as well, but Firefox uses two IOSurfaceRef and flips
between them. We would like to avoid two surfaces because it doubles the
GPU VRAM requirements of the application.
Changes to Windows
==================
Previously, the NSWindow would dynamically change between different
types of NSView based on the renderer being used. This is no longer
necessary as we just have a single NSView type, GdkMacosView, which
inherits from GdkMacosBaseView just to keep the tedius stuff separate
from the machinery of GdkMacosView. We can merge those someday if we
are okay with that.
Changes to Views
================
GdkMacosCairoView, GdkMacosCairoSubView, GdkMacosGLView have all been
removed and replaced with GdkMacosView. This new view has a single
CALayer (GdkMacosLayer) attached to it which itself has sublayers.
The contents of the CALayer is populated with an IOSurfaceRef which
we allocated with the GdkMacosSurface. The surface is replaced when
the NSWindow resizes.
Changes to Layers
=================
We now have a dedicated GdkMacosLayer which contains sublayers of
GdkMacosTile. The tile has a maximum size of 128x128 pixels in device
units.
The GdkMacosTile is partitioned by splitting both the transparent
region (window bounds minus opaque area) and then by splitting the
opaque area.
A tile has either translucent contents (and therefore is not opaque) or
has opaque contents (and therefore is opaque). An opaque tile never
contains transparent contents. As such, the opaque tiles contain a black
background so that Core Animation will consider the tile's bounds as
opaque. This can be verified with "Quartz Debug -> Show opaque regions".
Changes to Cairo
================
GTK 4 cannot currently use cairo-quartz because of how CSS borders are
rendered. It simply causes errors in the cairo_quartz_surface_t backend.
Since we are restricted to using cairo_image_surface_t (which happens to
be faster anyway) we can use the IOSurfaceBaseAddress() to obtain a
mapping of the IOSurfaceRef in user-space. It always uses BGRA 32-bit
with alpha channel even if we will discard the alpha channel as that is
necessary to hit the fast paths in other parts of the platform. Note
that while Cairo says CAIRO_FORMAT_ARGB32, it is really 32-bit BGRA on
little-endian as we expect.
OpenGL will render flipped (Quartz Native Co-ordinates) while Cairo
renders with 0,O in the top-left. We could use cairo_translate() and
cairo_scale() to reverse this, but it looks like some cairo things may
not look quite as right if we do so. To reduce the chances of one-off
bugs this continues to draw as Cairo would normally, but instead uses
an CGAffineTransform in the tiles and some CGRect translation when
swapping buffers to get the same effect.
Changes to OpenGL
=================
To simplify things, removal of all NSOpenGL* related components have
been removed and we strictly use the Core GL (CGL*) API. This probably
should have been done long ago anyay.
Most examples found in the browsers to use IOSurfaceRef with OpenGL are
using Legacy GL and there is still work underway to make this fit in
with the rest of how the GSK GL renderer works.
Since IOSurfaceRef bound to a texture/framebuffer will not have a
default framebuffer ID of 0, we needed to add a default framebuffer id
to the GdkGLContext. GskGLRenderer can use this to setup the command
queue in such a way that our IOSurface destination has been
glBindFramebuffer() as if it were the default drawable.
This stuff is pretty slight-of-hand, so where things are and what needs
flushing when and where has been a bit of an experiment to see what
actually works to get synchronization across subsystems.
Efficient Damages
=================
After we draw with Cairo, we unlock the IOSurfaceRef and the contents
are uploaded to the GPU. To make the contents visible to the app,
we must clear the tiles contents with `layer.contents=nil;` and then
re-apply the IOSurfaceRef. Since the buffer has likely not changed, we
only do this if the tile overlaps the damage region.
This gives the effect of having more tightly controlled damage regions
even though updating the layer would damage be the whole window (as it
is with OpenGL/Metal today with the exception of scissor-rect).
This too can be verified usign "Quartz Debug -> Flash screen udpates".
Frame Synchronized Resize
=========================
In GTK 4, we have the ability to perform sizing changes from compute-size
during the layout phase. Since the macOS backend already tracks window
resizes manually, we can avoid doing the setFrame: immediately and instead
do it within the frame clock's layout phase.
Doing so gives us vastly better resize experience as we're more likely to
get the size-change and updated-contents in the same frame on screen. It
makes things feel "connected" in a way they weren't before.
Some additional effort to tweak gravity during the process is also
necessary but we were already doing that in the GTK 4 backend.
Backporting
===========
The design here has made an attempt to make it possible to backport by
keeping GdkMacosBuffer, GdkMacosLayer, and GdkMacosTile fairly
independent. There may be an opportunity to integrate this into GTK 3's
quartz backend with a fair bit of work. Doing so could improve the
situation for applications which are damage-rich such as The GIMP.
This significantly cleans up how we handle various move-resize, compute-
size, and configure (notification of changes) in the macOS GDK backend.
Originally when prototyping this backend, there were some bits that came
over from the quartz backend and some bits which did not. It got confusing
and so this makes an attempt to knock down all that technical debt.
It is much simpler now in that the GdkMacosSurface makes requests of the
GdkMacosWindow, and the GdkMacosWindow notifies the GdkMacosSurface of
changes that happen.
User resizes are delayed until the next compute-size so that we are much
closer to the layout phase, reducing chances for in-between frames.
This also improves the situation of leaving maximized state so that a
grab and drag feels like you'd expect on other platforms.
I removed the opacity hack we had in before, because that is all coming
out anyway and it's a bit obnoxious to maintain through the async flows
here.
We want our tracking area to be limited to the input region so that we
don't pass along events outside of them for the window. This improves the
chances we click-out of a popover with a large shadow.
This still doesn't let us pass-through clicks for large shadows on top-
level windows though.
We need to avoid conflating the managing of frame callbacks from
the freeze/thaw mechanics and ensure we don't perform extra thaw
requests at the wrong time.
This is a port of the fix in the quartz backend to the new macOS backend.
From the original commit:
In macOS-12.sdk CGContextConverSizeToDeviceSpace returns a negative
height and passing that to CGContextScaleCTM in turn causes the cairo
surface to draw outside the window where it can't be seen. Passing the
absolute values of the scale factors fixes the display on macOS 12 without
affecting earlier macOS versions.
Now that we have the display's context to hook into, we can use it to
construct other GL contexts and don't need a GdkSurface vfunc anymore.
This has the added benefit that backends can have different GdkGLContext
classes on the display and get new GLContexts generated from them, so
we get multiple GL backend support per GDK backend for free.
I originally wanted to make this a vfunc on GdkGLContextClass, but
it turns out all the abckends would just call g_object_new() anyway.
The vfunc is called to initialize GL and it returns a "base" context
that GDK then uses as the context all others are shared with. So the GL
context share tree now looks like:
+ context from init_gl
- context1
- context2
...
So this is a flat tree now, the complexity is gone.
The only caveat is that backends now need to create a GL context when
initializing GL so some refactoring was needed.
Two new functions have been added:
* gdk_display_prepare_gl()
This is public API and can be used to ensure that GL has been
initialized or if not, retrieve an error to display (or debug-print).
* gdk_display_get_gl_context()
This is a private function to retrieve the base context from
init_gl(). It replaces gdk_surface_get_shared_data_context().
If our opaque region is the entire surface, then we can make the OpenGL
context opaque like we do for decorated windows. This improves performance
as the compositor does not need to blend the surface with the contents
behind the window.
It was used by all surfaces to track 'is-mapped', but still part of the
GdkToplevelState, and is now replaced with a separate boolean in the
GdkSurface structure.
It also caused issues when a widget was unmapped, and due to that
unmapped a popover which hid its corresponding surface. When this
surface was hidden, it emitted a state change event, which would then go
back into GTK and queue a resize on popover widget, which would travel
back down to the widget that was originally unmapped, causing confusino
when doing future allocations.
To summarize, one should not hide widgets during allocation, and to
avoid this, make this new is-mapped boolean asynchronous when hiding a
surface, meaning the notification event for the changed mapped state
will be emitted in an idle callback. This avoids the above described
reentry issue.
This removes the gdk_surface_set_shadow_width() function and related
vfuncs. The point here is that the shadow width and surface size can now
be communicated to GDK atomically, meaning it's possible to avoid
intermediate stages where the surface size includes the shadow, but
without the shadow width set, or the other way around.
When querying a device, we need to ensure we are providing coordinates
in the coordinate system of the surface. Further, we need to actually
provide the button and keyboard state.
This fixes some issues related to dragging scrollbars and selecting list
box rows more reliably.