Only the hand-picked sensible things, not all of it ;-)
Also fix a few typos that cspell complained about when I
tried to commit and some formatting changes that cargo fmt
insisted on when commiting.
For tiling, we will need to know the actual source size in addition to
the scaling factor that can be different. So store the scaling factors
in the scene command, as well as an offset where to start.
This is more accurate in case of clipping and rotation.
For rotation that doesn't matter (appart from the fact that the testing
can now be more strict)
But for clipping this prevent glitches with partial rendering where it
would seem like the image are moving a bit by a pixel when it is redrawn
with a different clip
* Extend the cspell word list
* Remove those extensions from individual source files
* white-list licenses and such as we should not meddle with those
* Fix spelling
TODO:
- Polish the API
- Most screenshot test are failling because the rotation don't draw the
rounded rectangle exactly the same (that's because the border
rectangle algo is not perfect), and also scaled image are not pixel
perfect
The updated screenshot is because of a small change in the image
rendering algorithm that changes the rounding slightly.
Implement the partial rendering with winit and our software renderer.
When the background is not opaque, we must still initialize all the
pixel with 0 otherwise we blend over the previous frame.
(That wasn't visible before because the buffer was always empty)
Implement rgb::Pod/rgb::Zeroable for PremultipliedRgbaColor and Rgb565
Pixel. This allows for bytemuck::cast_slice(), making it easier to copy
software renderer output into another buffer that's the same format but
Vec<u8> for example.
In the screenshot tests this works by setting SLINT_DEFAULT_FONT to
pointer to a directory, instead of a file. We then load all fonts in
that directory and consider their families the default unless a family
is specified. This way for "Noto Sans" a regular as well as an italic
version is registered in fontdb and returned in the list of font
fallback ids. embed_glyphs in the compiler then embeds those variants
and we find them at run-time.
Despite the documentation of softbuffer claiming that the high bits
should be 0, it seems to be interpreted as an alpha value when the
window is actually transparent.
And the values are considered as premultiplied alpha
Also use a Premultiplied aplha for the buffer so we can even have
transparent window with the software renderer
Change the font selection in the software renderer to always prefer
embedded bitmap fonts. This way when running `cargo test -p
test-driver-screenshots` the pixel fonts are used (as before this patch)
as well as in the CI, when all features are enabled (including
"systemfonts").
This originates from #2141
Remove platform glyph abstraction, just use non-zero u16 as glyph index
like everyone else. As a bonus, this reduces the memory consumption of
the glyph buffers per glyph from a size of a pointer to the size of the
u16.
Permit the style metrics to provide a `default-font-size` and bind that
to the `Window` if set. If not provided, then the backend can set a
`default-font-size`. By ensuring that the value is non-zero at run-time,
we can later introduce a rem unit that can act as factor relative to
this non-zero font size.
