Up until now, we have only dealt with games that pass Q = 1 for their
texture coordinates. PrBoom+, however, relies on proper homogenous
texture coordinates for its relatively complex sky rendering, which
means that we should perform this per-fragment division.
When using vertex attribute pointers, we default the Q coordinate for
textures to 0 causing issues if the 4th coordinate is not passed in the
vertex data.
Clean up these defaults and make sure that Q is always set to `1.f`.
Each texture unit now has its own texture transformation matrix stack.
Introduce a new texture unit configuration that is synced when changed.
Because we're no longer passing a silly `Vector` when drawing each
primitive, this results in a slightly improved frames per second :^)
Looking at how Khronos defines layers:
https://www.khronos.org/opengl/wiki/Array_Texture
We both have 3D textures and layers of 2D textures, which can both be
encoded in our existing `Typed3DBuffer` as depth. Since we support
depth already in the GPU API, remove layer everywhere.
Also pass in `Texture2D::LOG2_MAX_TEXTURE_SIZE` as the maximum number
of mipmap levels, so we do not allocate 999 levels on each Image
instantiation.
This makes it consistent with our other `blit_from_color_buffer` and
paves the way for a third method that will be introduced in one of the
next commits.
`GL_COMBINE` is basically a fixed function calculator to perform simple
arithmetics on configurable fragment sources. This patch implements a
number of texture env parameters with support for the RGBA internal
format.
This patch implements rubber band selection in table view while clamping
the rubber band rect to the widget inner rect, matching the behavior of
IconView and ColumnsView.
Previously when selecting a column that was partially scrolled out of
view the rubber band rect would extend outside the widget inner rect.
This patch rewrites the implementation to be more readable and clamps
the rubber band rect to the widget inner rect to match the behavior of
IconView.
Previously the rubber band rect of IconView was not properly constrained
to the widget inner rect, leaving a one pixel gap on the bottom and
right side. This patch removes the gap by inflating the constraint rect
by one pixel on each axis.
Currently, LibGUI modifies the Ctrl+Alt+Space key event to instead
represent the emoji that was selected through EmojiInputDialog. This is
limited to a single code point.
For multiple code point emoji support, individual widgets now set a hook
to be notified of the emoji selection with a UTF-8 encoded string. This
replaces the previous set_accepts_emoji_input() method.
Most of the emoji are 7x10px (or close to that). But some are larger, on
the order of 128x128px. The icon used for the SerenityOS category is one
such large emoji, and must be scaled down to an appropriate size for
rendering.
Currently, we use code point values as a tie break when sorting emoji by
display order. When multiple code point emoji are supported, this will
become a bit awkward. Rather than dealing with varying code point length
while sorting, just set a maximum display order to ensure these are
placed at the end.
Parse emoji from emoji-serenity.txt to allow displaying their names and
grouping them together in the EmojiInputDialog.
This also adds an "Unknown" value to the EmojiGroup enum. This will be
useful for emoji that aren't found in the UCD, or for when UCD downloads
are disabled.
We were consuming all whitespace from the format, but not the input
lexer - that was left to the actual format parsing code. It so happened
that we did not account for whitespace with the conversion specifier
'[', causing whitespace to end up in the output variables.
Fix this by always consuming all whitespace and removing the whitespace
logic from the conversion code.
This patch makes use of helpers implemented for window.length to resolve
two FIXMEs in WindowProxy previously simply assuming no child browsing
contexts :^)
This has two advantages: First the picker no longer changes the active
window state of its parent. Visually this is an additional hint that the
dialog is "fragile" and will close on loss of focus. Second, because
it contains a search box, its own input won't be preempted by global
application shortcuts when typing (pending #15129). This is a problem
in apps like PixelPaint which use shortcuts without modifiers.
Instead of having to negate every focusable widget or textbox, let
windows override all their widgets. These two Dialogs now block
themselves and each other.
If the containing block has indefinite height, we can't resolve
percentage heights against it. Instead of treating it as 0 by accident,
let's treat it as "auto" on purpose.
This brings back the cheek borders on Acid2.
When querying the HTML element (in strict mode) or the BODY element
(in quirks mode), we return the viewport dimensions.
Layout doesn't change the size of the viewport, so we can actually
reorder the steps here and avoid performing layout in some cases.
This removes a bunch of synchronous layouts on pages with reCAPTCHA.
This patch fixes an issue for applications that contain actions without
a modifier (e.g. PixelPaint). Previously when pressing any key bound to
an action while the CommandPalette was visible the action was forwarded
to the parent instead of the CommandPalette.
Previously Menus set themselves as active input solely to make
sure CaptureInput modals would close, but this is a functional
half-truth. Menus don't actually use the active input role; they
preempt normal Windows during event handling instead.
Now the active input window is notified on preemption and Menus
can remain outside the active input concept. This lets us make
more granular choices about modal behavior. For now, the only
thing clients care about is menu preemption on popup.
Fixes windows which close on changes to active input closing
on their own context menus.
According to TR #51, the "best definition of the full set [of emojis] is
in the emoji-test.txt file". This defines not only the emoji themselves,
but the order in which they should be displayed, and what "group" of
emojis they belong to.
OpenGL allows GPUs to approximate a triangle's maximum depth slope
which prevents a number computationally expensive instructions. On my
machine, this gives me +6% FPS in Quake III.
We are able to reuse `render_bounds` here since it is the containing
rect of the (X, Y) window coordinates of the triangle, thus its width
and height are the maximum delta X and delta Y, respectively.
Let's make 16px the default font size instead of 10px. This makes our
layout results match those of other engines in many more cases.
Also make the h1-h6 element styles use relative (em) font sizes, also
matching other browsers.
Also make the font discovery algorithm search subdirectories as well.
This will be used by Ladybird to discover more fonts on non-Serenity
systems. :^)
Instead of hard-coding the names of system fonts to use for the CSS
generic fonts (like "sans-serif", "monospace", etc.) we now call out
to a Platform::FontPlugin and ask for the generic names.
We didn't set their display at all before, and since CSS display is not
inherited, anonymous block wrappers were actually "display: inline", but
it kinda worked anyway because we positioned blocks based on their C++
class (BlockContainer) rather than their CSS display value.
Now that we position based on CSS display value, this broke positioning
of anonymous wrapper blocks, and this fixes that.
Instead of letting buttons determine the relative position
of their menus, a workaround only used by Statusbar segments,
open them all uniformly for a nice, consistent UI.
Passing a rect to popup() now routes to open_button_menu(), an
analog to open_menubar_menu(), which adjusts the menu's popup
position in the same way. Fixes button menus obscuring the buttons
which spawn them and jutting out at odd corners depending on screen
position.
in exclusive, checkable groups. Instead of merely setting the
button checked, call click() so buttons with registered actions
can activate. Fixes ActionGroups like FileManager's view type
checkables not activating when cycled with the keyboard.
Unlike regular buttons, unchecked checkables don't need to repaint
on MouseUp to feel responsive when clicking rapidly. In fact, this
can lead to a flickering effect when a bogus unchecked state gets
painted again before the button's checked one.
Before this change, block-level boxes were laid out vertically by
placing them after the nearest previous BlockContainer sibling. This
only worked if the preceding block-level box happened to be a
BlockContainer.
This fixes an issue where the screenshot on netsurf-browser.org was not
being laid out properly (it was `img { display: block }` which creates
a block-level ImageBox, and ImageBox is not a BlockContainer but a
ReplacedBox, so the following block-level box was skipping over the
ImageBox and being placed next to whatever was before the ImageBox..)
Instead of calling Core::EventLoop directly, LibJS now has a virtual
function on VM::CustomData for customizing this behavior.
We use this in LibWeb to plumb the spin request through to the
PlatformEventPlugin.
Instead of using Core::EventLoop and Core::Timer directly, LibWeb now
goes through a Web::Platform abstraction layer instead.
This will allow us to plug in Qt's event loop (and QTimer) over in
Ladybird, to avoid having to deal with multiple event loops.
We were constantly measuring and re-measuring the "alt" attribute text
of ImageBox layout nodes, even when the alt text didn't change. By
caching this, we avoid a *lot* of repeated text measurement work.
- Use the border box of the floated element when testing if something
needs to flow around it.
- Take the floated element's containing block size into account (instead
of the BFC root) when calculating available space on a line where a
right-side float intrudes.
To prevent lag when the displayed code points are redrawn in support of
a search box, only create the GUI::Button objects for the emoji a single
time. Re-use those buttons when adding them to the dialog.
This will allow easily adding components such as a search box. Also,
increase the number of emoji per row. This does not fix the issue where
too many emoji will cause the dialog to grow limitlessly, but it looks a
bit more reasonable now with the number of emoji that we have.
Our bilinear scaling logic worked well for upscaling, but during
downscaling the bitmap was often shifted one pixel to the bottom right.
This is a common problem, described here in more detail:
https://bartwronski.com/2021/02/15/bilinear-down-upsampling-pixel-grids-and-that-half-pixel-offset/
Fix it by calculating coordinate shift values that align the pixel's
boundaries between the source and target pixels before selecting the
source pixels to interpolate.
This adds icons for folder paths ending in ".git". This adds 16x16 and
32x32 versions of a folder with a Git logo for both the closed and open
versions of the folder icon.
This allows JS::Script to mark its corresponding HTML::Script, even if
it's a little roundabout looking. Fixes an issue where the JS::Script
was kept alive by the execution stack, but the HTML::Script was gone.
This was originally part of 8f9ed415a0
but got lost in a merging accident.
Instead, use the window object from the context element. This fixes an
issue where activating event handlers during fragment parsing would try
to set up callbacks using the internal window object's ESO.
This caused a verify_cast crash on Google Maps, since the internal realm
doesn't have an associated ESO. Perhaps it should, but in this specific
case, it makes more sense for fragment parsing to fully adopt the
context provided.
This allows the garbage collector to keep HTML::Script objects alive and
fixes a bug where a HTMLScriptElement could get GC'd while its code was
executing.
Workers don't have a document, so we can't ask them if scripting is
enabled or not. This is not the right long-term fix, but it fixes an
assertion when trying to query the missing responsible document of a
web worker.
This is *not* according to spec, however we currently store prototypes
and constructors on Window, so the only way for objects in a worker
context to become fully formed is to make a Window.
Long-term we should clean this up and remove the worker window object,
but for now it allows workers to exist without asserting.
wrap() is now basically a no-op so we should stop using it everywhere
and eventually remove it. This patch removes uses of wrap() in
non-generated code.
Unlike ensure_web_prototype<T>(), the cached version doesn't require the
prototype type to be fully formed, so we can use it without including
the FooPrototype.h header. It's also a bit less verbose. :^)
