And deprecate WheelDown and WheelUp. Funnily enough a similar change was
done for *all other* applications including now-long-gone implementations
like NaClApplication back in 2a77856df2,
which was 2016!!
Frankly, I was first thinking that I'd just deprecate this thing and not
update it anymore, but it seems there's still a use case for a
lightweight wrapper sitting right on top of system APIs. SDL / GLFW is
too heavy for when one just needs to display stuff or debug GPU issues
for which it isn't clear whether they're the driver's fault or the
tookit's. And WindowlessApplication implementations proved very useful
for this, so I guess a similar windowed application still makes sense,
even if not very featureful. Additionally, for Vulkan I might take a stab
at implementing a WaylandApplication if even just an attempt to
understand how the swapchain works internally, and having an XApplication
available would be handy to compare the behavior.
For the most part is the same as in Sdl2Application, except that here
GLFW already returned floating-point coordinates, which this finally
makes use of.
Pointer events are an unified abstraction over mouse, touch, pen and
potential other yet-to-be-invented pointer-like input methods. Their goal
is to expose all such input methods under a single interface so the
application side doesn't need to explicitly make sure that it's
touch-aware or pen-aware. This abstraction is already present in HTML5,
in Qt6 and in WINAPI as well, and is also what I adopted for the new UI
library because it *just makes sense*.
Unfortunately not even SDL3 took the opportunity to introduce that and
instead added a *third* separate event type for pen input in SDL3. At
first I thought that I wouldn't introduce any extra abstractions in the
Application classes (because that's what they are designed to be, as
lightweight as possible), but midway through introducing TouchEvent
classes and fighting SDL's touch->mouse and mouse->touch compatibility
translation (yes, it's both ways, depending on the platform) I realized
that a much simpler solution that doesn't require any event translation
or the users duplicating their event handling logic for several possible
input types is to introduce a single new event type that covers all.
Which is what this commit does -- it doesn't introduce anything
touch-related so far, just creates a new PointerEvent and
PointerMoveEvent class and corresponding virtual functions. Additionally,
I took this as an opportunity to make the position floating-point, since
that's what SDL3 does now as well, and GLFW did so since ever.
Plus, the Pointer and Pointers enums are directly on the Sdl2Application
class, to allow me to *finally* introduce pointer state queries. Which
weren't possible until now, because there were mutually incompatible
MouseEvent::Button and MouseMoveEvent::Button enums and putting them on
the base class would mean one would have to be translated and the other
not. With Pointer it's translated always, because there isn't any similar
enumeration in SDL that would cover mouse, touch and pen at the same
time.
The distance reported by it is useless for any practical purpose because
it doesn't report a ratio of the current and previous radius between all
points, but rather the distance. Which, well, have fun using for any sort
of zooming.
(And yeah, given that the MultiGestureEvent is gone in SDL3, I spent
quite some time looking at what it actually did in order to reimplement
that functionality on my end, and it felt *extremely weird* to me that it
always considered just that single point, never the others in order to
calculate any sort of radius. This is why, because it never considered
any sort of radius between the points, so the "Multi" in there is highly
questionable!)
So this test is just to make the uselessness easy to verify, nothing
more.
Heh, somehow every time I run the full battery of tests I discover
another failure.
NVidia, with Vulkan version forced to 1.0 and when VK_KHR_maintenance1
isn't enabled, returns VK_ERROR_OUT_OF_DEVICE_MEMORY. So whitelist that
error as well and treat it as allocation failure and not a fatal error.
Not sure what I did in 3e4e1bde69 but that
updated XFAIL is now an XPASS on NVidia. Because, apparently, the clear
clears the whole memory, not just the image area, so even though the row
pitch is different, the comparison of the initial N bytes passes.
So I'm ditching the silly XFAILs and doing a proper image comparison that
includes the actual driver-dependent row pitch for the images. Finally,
the image-to-image copy was flat out wrong because it didn't take the
*input* row pitch into account, so it copied garbage and then compared to
a different kind of garbage.
Otherwise it may happen that the clock gets adjusted mid-frame, leading
to an underflow and an assertion like this:
Assertion _measurements[i]._movingSum + data >= _measurements[i]._movingSum failed at src/Magnum/DebugTools/FrameProfiler.cpp:233
The clock adjustment is known to be happening rather frequently under
WSL2.
The check for memory size was enough for llvmpipe, but not for
SwiftShader. And now it wasn't enough for NVidia either, so let's just do
it properly.
Ideally of course this would compare always. Don't feel like doing that
right now tho, so it's just a TODO.
This caused MeshToolsCompileGLTest to fail in a strange way, and
PhongGLTest::renderLowLightAngle() as well. Which looked rare enough that
I first suspected some random driver bug, but apparently it was all
caused by these using the default infinity range instead of explicitly
calling setLightRanges() on the shader.
The test is now updated to explicitly verify the default value when a
setter isn't called, to catch this problem better in case it reappears in
a different form elsewhere.
Neither of those are really critically important. Failure not being a
failure is fine (just don't make the shader broken in the first place),
multiple color output bindings should be in the shader source anyway.
This was quite a mess, the whole array name copypasted to each and every
access. I mean, yeah, originally I thought this would be *the* usage
pattern, but oh god this resulted in SO MANY copypaste errors.
Which ultimately means the two annoying NVidia failures in the
TextureGLTest related to pixel storage in compressed 3D images are no
longer failing as a result of cleanup.
A regression was introduced in 2a8e550b57
that affects only Windows builds. The version downloaded on the Windows
CI unfortunately doesn't contain CMake configs, so this went unnoticed.
I wanted to add this for GlfwApplication, only to realize the timer there
is a double, in seconds, so accepting *integral* milliseconds there felt
very weird. Let's use the fancy new time types instead.
Also updated the docs to (hopefully) clarify that setSwapInterval() has
to be called in order for the interaction between the two to work
properly.
As usual, the old variant taking untyped milliseconds is a deprecated
alias to this one.
Vladimír Vondruš