Commit a4018708 authored by Jakob Bornecrantz's avatar Jakob Bornecrantz

doc: Remove old changelog fragments

parent e7323e98
u/vk: Remove unused vk_image struct, this is later recreated for the image
allocator code.
u/vk: Add a new image allocate helper, this is used by the main compositor to
create, export and import swapchain images.
u/vk: Rename `vk_create_semaphore_from_fd` to `vk_create_semaphore_from_native`
aux/ogl: Add a function to compute the texture target and binding enum for a given swapchain image creation info.
- mr.547
- mr.581
aux/android: New Android utility library added.
util: Tidy hand tracking header.
math: Fix doxygen warnings in vector headers.
compositor: Add support for alpha blending with premultiplied alpha.
compositor: Implement subimage rectangle rendering for quad layers.
compositor: Enable subimage rectangle rendering for projection layers.
compositor: Fix printing of current connected displays on nvidia when no whitelisted display is found.
compositor: Add env var to temporarily add display string to nvidia whitelist.
compositor and clients: Use a generic typedef to represent the platform-specific graphics buffer, allowing use of `AHardwareBuffer` on recent Android.
compositor: Check the protected content bit, and return a non-success code if it's set. Supporting this is optional in OpenXR, but lack of support must be reported to the application.
compositor: Implement cylinder layers.
main: Set the maximum layers supported to 16, we technically support more than
16, but things get out of hand if multiple clients are running and all are using
max layers.
main: Add code to check that a format is supported by the GPU before exposing.
compositor: Remove panotools and vive shaders from compositor.
Initial work on a port of the compositor to Android.
render: Implement equirect layer rendering.
main: Fix leaks of sampler objects that was introduced in !566.
psvr: We were sending in the wrong type of time to the 3DOF fusion code,
switch to nanoseconds instead of fractions of seconds.
rs: Make the pose getting from the T265 be threaded. Before we where getting the
pose from the update input function, this would cause some the main thread to
block and would therefore cause jitter in the rendering.
survive: Add lighthouse tracking system type
hydra: Add lighthouse tracking system type
rs: Add slam tracking system type
northstar: Use tracking system from tracker (e.g. rs) if available.
psmv: Introduce proper grip and aim poses, correctly rotate the grip pose to
follow the spec more closely. The aim poses replaces the previous ball tip pose
that was used before for aim.
survive: Implement haptic feedback.
-- mr.573
dummy: Tidy the code a bit and switch over to the new logging API.
psvr: Switch to the new logging API.
Add initial "Cardboard" phone-holder driver for Android.
- mr.427
Generalize handling of native-platform handles in IPC code, allow bi-directional handle transfer, and de-duplicate code between server and client.
generation: Fix handling 'in_handle' by adding a extra sync round-trip, this
might be solvable by using `SOCK_SEQPACKET`.
Implement the `xrt_compositor::import_swapchain` function, uses the earlier
`in_handle` work.
proto: Transport the `xrt_compositor_info` over the wire so that the client can
get the needed information.
client: Implement the usage of the `xrt_image_native_allocator`, currently not
used. But it is needed for platforms where for various reasons the allocation
must happen on the client side.
client: Add a "loopback" image allocator, this code allocates a swapchain from
the service then imports that back to the service as if it was imported. This
tests both the import code and the image allocator code.
ipc: Allow sending zero handles as a reply, at least the Linux fd handling code
allows this.
Use a native AHardwareBuffer allocator on the client side when building for recent-enough Android.
ipc: Add functionality to disable a device input via the `monado-ctl` utility,
this allows us to pass the conformance tests that requires the runtime to turn
off a device.
Support building in-process Monado with meson.
Allow building some components without Vulkan. Vulkan is still required for the compositor and therefore the OpenXR runtime target.
- mr.581
Add an OpenXR Android target: an APK which provides an "About" activity and eventually, an OpenXR runtime.
OpenXR: Add support for attaching Quad layers to action sapces.
OpenXR: Use initial head pose as origin for local space.
OpenXR: Minor fixes for various bits of code: copy-typo in device assignment
code; better stub for the unimplemented function
`xrEnumerateBoundSourcesForAction`; better error message on internal error in
OpenXR: Make the `xrGetCurrentInteractionProfile` conformance tests pass, needed
to implement better error checking as well as generating
`XrEventDataInteractionProfileChanged` events to the client.
- mr.456
OpenXR: Centralize all sub-action path iteration in some x-macros.
OpenXR: Improve the validation in the API function for
OpenXR: Implement the function `xrEnumerateBoundSourcesForAction`, currently we
only bind one input per top level user path and it's easy to track this.
OpenXR: Properly handle more than one input source being bound to the same action
according to the combination rules of the specification.
OpenXR: Fix multiplicity of bounds paths per action - there's one per input/output.
OpenXR: Implement the MND_swapchain_usage_input_attachment_bit extension.
OpenXR: Refactor the native compositor handling a bit, this creates the
compositor earlier then before. This allows us to get the viewport information
from it.
OpenXR: Implement action set priorities and fix remaining action conformance tests.
st/oxr: Fix crash when calling `xrPollEvents` when headless mode is selected.
OpenXR: Add stub functions and support plumbing for a lot of layer extensions.
OpenXR: Be sure to return `XR_ERROR_FEATURE_UNSUPPORTED` if the protected content bit is set and the compositor does not support it.
OpenXR: Update to 1.0.11 and start returning the new `XR_ERROR_GRAPHICS_REQUIREMENTS_CALL_MISSING` code added in this release.
OpenXR: Enable the `XR_KHR_android_create_instance` extension.
OpenXR: Add support for creating swapchains with depth formats and submitting depth layers. The depth layers are passed through to the compositor, but are not used yet.
OpenXR: For pose actions the any path (`XR_NULL_PATH`) needs to be special
cased, essentially turning into a separate action sub path, that is assigned
at binding time.
OpenXR: More correctly implement `xrGetInputSourceLocalizedName` allowing apps
to more accurently tell the user which input to use.
OpenXR: Pass through equirect layer data to the compositor.
add `xrt_device_type` to `xrt_device` to differentiate handed controllers
from controllers that can be held in either hand.
- mr.428
Rename functions and types that assumed the native graphics buffer handle type was an FD: in `auxiliary/vk/vk_helpers.{h,c}` `vk_create_image_from_fd` -> `vk_create_image_from_native`, in the XRT headers `struct xrt_compositor_fd` -> `xrt_compositor_native` (and method name changes from `xrt_comp_fd_...` -> `xrt_comp_native_...`), `struct xrt_swapchain_fd` -> `struct xrt_swapchain_native`, `struct xrt_image_fd` -> `struct xrt_image_native`, and corresponding parameter/member/variable name changes (e.g. `struct xrt_swapchain_fd *xscfd` becomes `struct xrt_swapchain_native *xscn`).
Make some fields on `xrt_gl_swapchain` and `xrt_vk_swapchain` private moving
them into the client compositor code instead of exposing them.
Make `xrt_compositor::create_swapchain` return xrt_result_t instead of the
swapchain, this makes the methods on `xrt_compositor` more uniform.
Add the method `xrt_compositor::import_swapchain` allowing a state tracker to
create a swapchain from a set of pre-allocated images. Uses the same
`xrt_swapchain_create_info` as `xrt_compositor::create_swapchain`.
Make `xrt_swapchain_create_flags` swapchain static image bit match OpenXR.
Add `XRT_SWAPCHAIN_USAGE_INPUT_ATTACHMENT` flag to `xrt_swapchain_usage_bits`
so that a client can create a Vulkan swapchain that can be used as input
Remove the `flip_y` parameter to the creation of the native compositor, this
is now a per layer thing.
Add `xrt_compositor_info` struct that allows the compositor carry information
to about it's capbilities and it's recommended values. Not everything is hooked
up at the moment.
Add a native handle type for graphics sync primitives (currently file descriptors on all platforms).
Add defines for underlying handle types.
Add a whole bunch of structs and functions for all of the different layers
in OpenXR. The depth layer information only applies to the stereo projection
so make a special stereo projection with depth layer.
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment