Affects per-pixel transparency
The current method renders to the screen by copying the GLES output to a
DIB for transparency using the CPU instead of rendering directly to the
window via the GPU. This is slower and also forces the window to be borderless
as WS_EX_LAYERED affects the non-client region as well.
This change uses DWMEnableBlurBehindWindow which allows using the standard
glClearColor() background alpha and is also performed through the GPU,
eliminating CPU bottlenecks
Now the stack saved in a `GDScriptFunctionState` is cleared as soon as the `yield()` operation is known not to be resumed because either the script, the instance or both are deleted.
This clears problems like leaked objects by eliminating cases of circular references between `GDScriptFunctionState`s preventing them and the objects they refer to in their saved stacks from being released. As an example, this makes using `SceneTreeTimer` safer.
Furthermore, with this change it's now possible to print early warnings about `yield()`s to released script/instances, as now we know they won't be successfully resumed as the condition for that happens. However, this PR doesn't add such messages, to keep the observed behavior the same for the time being.
Also, now a backup of the function name in `GDScriptFunctionState` is used, since the script may not be valid by the time the function name is needed for the resume-after-yield error messages.
Skew is x-axis only, because it must be bidirectionally convertible to a 2x3 matrix, but you can subtract it to the rotation to get the effect on y-axis
(cherry picked from commit efb1f7d76b)
`OS.shell_open()` will pass on the path directly to the OS' shell
handler (which can handle file paths or URLs). It can't handle
Godot-specific paths, so these need to be converted with
`ProjectSettings.globalize_path()` first.
(cherry picked from commit d46e411b44)
This fixes numerous false positives coming out of the culling system.
AABB checks are now a full separating-axis check against the frustum, with the points of the frustum being compared to the planes of the box just as the points of the box were being compared to the planes of the frustum. This fixes large objects behind the camera not being culled correctly.
Some systems that used frustums that were (sometimes mistakenly?) unbounded on one or more side have been modified to be fully enclosed.
This moves the instance id member from Variant to the ObjectRC so that Variant is still the same size as before the fix (and also regardless if debug or release build).
This commit addresses multiple issues with `Variant`s that point to an `Object`
which is later released, when it's tried to be accessed again.
Formerly, **while running on the debugger the system would check if the instance id was
still valid** to print warnings or return special values. Some cases weren't being
warned about whatsoever.
Also, a newly allocated `Object` could happen to be allocated at the same memory
address of an old one, making cases of use hard to find and having **`Variant`s pointing
to the old one magically reassigned to the new**.
This commit makes the engine realize all these situations **under debugging**
so you can detect and fix them. Running without a debugger attached will still
behave as it always did.
Also the warning messages have been extended and made clearer.
All that said, in the name of performance there's still one possible case of undefined
behavior: in multithreaded scripts there would be a race condition between a thread freeing
an `Object` and another one trying to operate on it. The latter may not realize the
`Object` has been freed soon enough. But that's a case of bad scripting that was never
supported anyway.
The `TextEdit` one was indeed a potential bug.
The `PCKPacker` one seems to be a false positive, it's already in a
`for` loop that depends on `files.size()`.
(cherry picked from commit ca4e4506db)
The minimum slider value no longer allows decreasing the value below
the default, as this can cause things to break in the editor.
The maximum slider value was also increased to 4096 since it can safely
be increased to that value (some add-ons may require it).
This closes#37052.
(cherry picked from commit 8d8c7a9383)
Namely, move the drive dropdown to just the left of the path text box and don't include the former
in the latter.
This improves the UX on Windows.
In the UNIX case, since its concept of drives is (ab)used to provide shortcuts to useful paths, its
dropdown is kept at the original location.
UDP sockets can be "connected" to filter packets from a specific source.
In case of a bound socket (e.g. server), a new socket can be created on
the same address/port that will receive all packets that are not
filtered by a more specific socket (e.g. the previously connect socket).
This way, a UDPServer can listen to new packets, and return a new
PacketPeerUDP when receiving one, knowing that is a "new client".
Pith bend message now has correct size (was 2 bytes instead of 3).
Recognized (but not implemented) 0xF? messages. SysEx messages will be reocognized as such, but their contents will be ignored.
Fixes#26637.
Fixes#19900.
The viewport_size returned by get_viewport_size was previously incorrect, being half the correct value. The function is renamed to get_viewport_half_extents, and now returns a Vector2.
Code which called this function has also been modified accordingly.
This PR also fixes shadow culling when using ortho cameras, because the correct input for CameraMatrix::set_orthogonal should be the full HEIGHT from get_viewport_half_extents, and not half the width.
It also fixes state.ubo_data.viewport_size in rasterizer_scene_gles3.cpp to be the width and the height of the viewport in pixels as stated in the documentation, rather than the current value which is half the viewport extents in worldspace, presumed to be a bug.
We used a lock signals in the signal_map while emitting, because it was
not allowed to disconnect them while being emitted.
We used that lock to check if we where deleting an object during signal
emission.
Now that we allow to disconnect signals while they are being emitted, if
an object first disconnects, then gets deleted we can't know that a
signal was being emitted during the destructor.
This commit adds a new `_emitting` boolean member to Object to be set
while emitting and checked in the destructor, while removing the old
signal lock which is now unused.
Used to allocate in stack (via alloca) which causes crashes when trying
to encode big variables.
The buffer grows as needed up to `encode_buffer_max_size` (which is
8MiB by default) and always in power of 2.
According to 22637beb2e (commitcomment-36651823)
and as confirmed by @reduz, this seems not to be necessary now that we
copy-on-write.
This triggered freeze scenarios in cases where a node would be deleted
while being used as a target in a signal emission.
Fixes#34650.
Fixes#34769.
Now those two errors go back to reporting:
```
ERROR: emit_signal: Condition ' !target ' is true. Continuing..:
At: core/object.cpp:1191.
```
Avoids crashes on debug mode. Instead it now breaks the execution and
show the error in-editor. Will still crash on release.
Also add a similar check to Marshalls to ensure the debugger doesn't
crash when trying to serialize the invalid instance.
Unify pack file version and magic to avoid hardcoded literals.
`version.py` now always includes `patch` even for the first release in
a new stable branch (e.g. 3.2). The public name stays without the patch
number, but `Engine.get_version_info()` already included `patch == 0`,
and we can remove some extra handling of undefined `VERSION_PATCH` this
way.
Co-authored-by: Rémi Verschelde <rverschelde@gmail.com>
