Along with ReadPixels in the next change, a Framebuffer with depth is enough to
implement screenshot functionality on top of any Backend.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
"handling" means panicing, but at least the panicing is moved up
a layer, leaving future changes to do it properly in GPU.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
OpenGL doesn't care if the texture to a framebuffer changes, but
Direct3D does. Change Backend to better match both APIs.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Direct3D needs to know the texture bind usage up front, in particular
whether the texture is going to be used as a render target.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
It serves no purpose other than paranoia. Perhaps buggy drivers exists that
require unused texture slots cleared before drawing to a texture, but if so the
workaround belongs in the GL backend.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The offset argument to DrawElements is in 16-bit shorts, while the
underlying DrawElements use byte offsets.
No users of DrawElements use a non-zero offset, so nothing changed.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Emulate them for the OpenGL ES backend because 2.0 doesn't support uniform
buffers. The future d3d backend only supports uniform (constant) buffers.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
InputLayout is the abstraction for the mapping between vertex data and
shader inputs. The mapping is implicit in OpenGL but explicit in
Direct3D.
Infer the attribute name location index using shader reflection data,
and get rid of a parameter to NewProgram.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
We're about to add Direct3D support, where shaders are written in
HLSL. Rather than write shaders twice (or more), convert them to
a GLSL variant understood by the glslcc cross-compiler and generate
the OpenGL ES 2.0 and HLSL variants. The HLSL is used by a future
change.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The error message is not appropriate where there are multiple backends,
and there's a much better chance Gio will run with the direct3d backend
where no external DLLs are required.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The srgb package was recently created to contain just the sRGB
emulation, but the names weren't shortened accordingly.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
To prepare package headless for multiple backends, refactor the common headless
driver to no longer assume an OpenGL context. Instead, introduce a headless
backend type and the OpenGL implementation, glBackend.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
This is a refactoring change to prepare for another gpu.Backend
implementations.
Notably, app/loop.go no longer imports gpu/gl.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
The GPU implementation only uses immutable buffers so far, so let's
make it easy and performant for the backends.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Before this change, the index buffer would start empty and grow up to the
maximum size (128kb). It would never shrink. We're about to tighten the GPU
buffer API to be immutable for performance and to better match Direct3D, so
allocate the index buffer once at startup, and limit it to a reasonable size.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
It was left over from a previous approach to enable the program
to decide the ordering between calls to Layout vs Clicked.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Now that pkg.go.dev supports the UNLICENSE, the COPYING file confuses
the automatic license detection. Move the license text back to README.md.
This is merely a cosmetic change, there are no licensing changes to
the project.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
It passes now, thanks to the recent work to add retries to screenshots.
If it turns out to still not be reliable enough, we can look at this
again.
Signed-off-by: Daniel Martí <mvdan@mvdan.cc>
The new field ImageOp.Rect is initialized to cover the entire source
image, but can be modified to draw only a section of it.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
Right now, this was badly needed for the wayland subtest, as it seems
like waiting for a frame to be ready wasn't enough for the screenshot to
show what we want. In practice, even if the machine was idle, it could
sometimes take a few extra milliseconds for the app to first appear on
the display.
This was worse when the machine is under stress, which is often the case
with CI. For example, the command below showed a ~20% failure rate on my
laptop with four cores:
go test -c -o test && stress ./test -test.run EndToEnd/Wayland
Add a generic withRetries helper function, which allows us to keep
trying some action up to a timeout, with sleeps in between that start at
100ms and keep doubling until 2s. The function also logs before each
sleep, in case the user is confused why their test is stuck for
potentially may seconds at once.
Refactor the wantColors function into a separate function that returns
an error, as we can no longer directly report errors via *testing.T. It
still reports all the mismatches at once, which is useful. It can now be
used on to pof withRetries with a thin wrapper.
While at it, make the X11 subtest use withRetries to wait for the X
server to be ready. It was using a simpler method with a fixed number of
static sleeps. It's now more consistent, and a bit better overall.
With the changes above, the 'stress' command from earlier can get past
100 runs on my laptop with no failures at all.
Finally, fix a rogue log.Fatal call I had somehow missed.
Signed-off-by: Daniel Martí <mvdan@mvdan.cc>
A recent change made the OpenGL functions an interface of the functions
required for the implementation of GPU, a renderer for Gio operations.
That allowed for running Gio on external systems where OpenGL is
available.
However, to allow for non-OpenGL flavored backends such as Vulkan,
Metal and Direct3D, this change introduces Backend for the high-level
operations required by GPU. This change also adds a concrete backend
to package gl.
Type Backend is a first cut heavily based on OpenGL. Future changes will add
more backends, where the Backend interface quite possibly will need refinement.
Signed-off-by: Elias Naur <mail@eliasnaur.com>
On Linux and FreeBSD, this means we no longer need to install Go twice,
since we don't need to bootstrap a Go build. This reduces the "install
go" step by over half a minute, and avoids installing a number of distro
packages.
Debian requires the pkg-config package to be explicit, as it's not part
of the build-essential metapackage, and we need it. It was being pulled
as a dependency from the golang package before.
OpenBSD is nearly the same change, modulo the fact that Go doesn't have
a binary release for it. So instead, we keep installing Go via the
system's package, and we download and build a source archive.
Finally, switch from Debian testing+experimental to simply unstable.
Sway 1.4 finally hit unstable a few days ago, and experimental is
extremely unstable, so this is a step in the right direction. Add a TODO
about going to just testing once sway 1.4 arrives there.
Signed-off-by: Daniel Martí <mvdan@mvdan.cc>
Elias Naur