Some platforms (*cough* web *cough*) have hard limits on the number of
threads that can be spawned.
Currently, ThreadPoolWork (mostly used in rendering/physics servers)
will spawn as many threads as CPUs available causing exception on
machines with high CPU count.
This commit adds a new overridable method to OS that returns the default
thread pool size (still the CPU count by default), and overrides it for
the JavaScript platform so it always allocate only one thread.
We can likely improve the whole ThreadPoolWork in the future to always
allocate X amount of threads, and assign jobs to them on the fly, but
that will require some more architectural changes.
- Rename OpenGL to GLES3 in the source code per community feedback.
- The renderer is still exposed as "OpenGL 3" to the user.
- Hide renderer selection dropdown until OpenGL support is more mature.
- The renderer can still be changed in the Project Settings or using
the `--rendering-driver opengl` command line argument.
- Remove commented out exporter code.
- Remove some OpenGL/DisplayServer-related debugging prints.
First implementation with Linux display manager.
- Add single-threaded mode for EditorResourcePreview (needed for OpenGL).
Co-authored-by: clayjohn <claynjohn@gmail.com>
Co-authored-by: Fabio Alessandrelli <fabio.alessandrelli@gmail.com>
Sets `AlignOperands` to `DontAlign`.
`clang-format` developers seem to mostly care about space-based indentation and
every other version of clang-format breaks the bad mismatch of tabs and spaces
that it seems to use for operand alignment. So it's better without, so that it
respects our two-tabs `ContinuationIndentWidth`.
* Made the Basis euler orders indexed via enum.
* Node3D has a new rotation_order property to choose Euler rotation order.
* Node3D has also a rotation_mode property to choose between Euler, Quaternion and Basis
Exposing these modes as well as the order makes Godot a lot friendlier for animators, which can choose the best way to interpolate rotations.
The new *Basis* mode makes the (exposed) transform property obsolete, so it was removed (can still be accessed by code of course).
Roughly based on https://github.com/godotengine/godot-proposals/issues/3375 (used format is slightly different).
* Implement bitwidth based animation compression (see animation.h for format).
* Can compress imported animations up to 10 times.
* Compression format opens the door to streaming.
* Works transparently (happens all inside animation.h)
Whenever we change the name (or remove) generated cpp files with the `.gen.cpp`
extension, users run into build issues when switching between branches (i.e.
switching before and after the name change/removal). This is because we glob
`*.cpp` so if a now-obsolete file from a previous build is present, we'll
include it too, potentially leading to bugs or compilation failure (due to
missing headers or invalid code).
So globbing patterns in `add_source_files` will now skip files ending with
`.gen.cpp`, which should instead be passed explicitly where they're used.
Extension class properties that have been registered with
ClassDB(`GDNativeInterface::classdb_register_extension_class_property`)
were not fetched previously. (Only directly providing a property
list using `GDNativeExtensionClassCreationInfo::get_property_list_func`
would work.) This especially caused problems with the C++
bindings since they exclusively rely on ClassDB to register properties.
Allow using String or StringName types as operand in any position of the
`in` operator, which is more convenient in scripting when interacting
with data in the engine (such as a Node name).
`core` and `scene` shouldn't depend on `editor`, so they can't query this style
setting in `get_argument_options`. But we can handle it after the fact in
GDScript's completion code.
Also cleans up a couple extra unused invalid includes in `core`.
Adding the copy constructor is needed to solve a `-Wdeprecated-copy` warning
from GCC and Clang, which is raised when upgrading doctest from 2.4.4 to 2.4.6.
The GDVIRTUAL_NATIVE_PTR did not declare the correct GDNativeConstPtr
template, resulting in "void*" being used as it's type info in both the
documentation and the extension API dump.
* Fixed LODs for shadow meshes.
* Added a merging step before simplification. This helps with tesselated
meshes that were previously left untouched. The angle difference at
wich edges ar considered "hard" can be tweaked as an import setting.
* LODs will now start with the highest decimation possible and keep
doubling (approximately) the number of triangles from there. This
makes sure that very low triangle counts are included when possible.
* Given more weight to normal preservation.
* Modified MeshOptimizer to report distance-based error instead of
including attributes in the reported metrics.
* Added attribute transference between the original mesh and the
various LODs. Right now only normals are taken into account,
but it could be expanded to other attributes in the future.
`#pragma once` was used in a few files, yet we settled on using
traditional include guards instead.
The PooledList template comment was also moved to allow editors
such as Visual Studio Code to display the comment when hovering
PooledList.
`app.h` was renamed to `app_uwp.h` to be less generic for the
include guard.
Calling the constructor alone is not enough if the class to be
instantiated is not a base class.
This commit adds two functions, one for retrieving the the extension
class reference, the other to construct an instance using the
constructor and the extension class reference.
This can be used as a shorthand for:
if OS.is_stdout_verbose():
print("...")
Unlike `print_debug()`, this works in release builds too and can
be toggled off in debug builds.
The `Engine.set_editor_hint()` setter method is no longer exposed
to scripting, which makes the property effectively read-only from
an user perspective.
This method extracts the 2 or 3-letter language code from `OS::get_locale()`,
making it easier for users to identify the "main" language code for users
that might have different OS locales due to different OS or region, but
should be matched to the same translation (e.g. "generic" Spanish).
Fixes#40703.
On latest (11.1 as of this commit) GCC, the following warning is
continuously issued during build:
warning: placement new constructing an object of type
'SafeNumeric<unsigned int>' and size '4' in a region of type
'uint32_t*' {aka 'unsigned int*'} and size '0' [-Wplacement-new=]
This happens because on 98ceb60eb4 the new operator override used
was dropped and replaced with standard placement new. GCC sees the
subtraction from the pointer and complains as it thinks that the
SafeNumeric is placed outside an allocation, not knowing that the
address requested is already inside one.
After suggestions, the false positive is silenced, with no other
changes.
Removes _networking_ prefix from some methods and members, now that multiplayer has been largely moved out of Node and SceneTree and is seperated into its own set of classes.
Move multiplayer classes to "core/multiplayer" subdir.
Move the RPCConfig and enums (TransferMode, RPCMode) to a separate
file (multiplayer.h), and bind them to the global namespace.
Move the RPC handling code to its own class (RPCManager).
Renames "get_rpc_sender_id" to "get_remote_sender_id".
There are two ways a class can be added to ClassDB:
- `A`: When an instance of the class is created for the first time. When
this happends the class is not registered/exposed to scripts.
- `B`: When calling `GDREGISTER_CLASS(ClassName)` or similar. When this
happens the class is registered/exposed to scripts.
ClassDB has an API type property to differentiate between the core
and editor APIs. Up until now the API type was determined whenever
the class is added to ClassDB (either `A` or `B`).
The problem comes when a class is instantiated (`A`) before
being registered (`B`).
If at this point the current defined API is not the same as when the
class is later registered, this will result in a mismatch between
`--editor` and non-editor apps.
This is specially bad for C# as it makes the editor player abort.
This was happening with `EditorPaths` which, while being registered
during the editor API classes registrations, it was also being
instantiated earlier when running the editor or the project manager,
via a call to `EditorPaths::create()`. This regression was introduced
in 1074017f04.
This commit fixes this simply by re-assigning the class API type when
the class is registered (`B`).
This is correct because API type describes registered/exposed classes.
It shouldn't cause any regressions as the API type should not be
accessed of classes that are not (or not yet) registered/exposed.
Code locations for reference:
- Method to add a class to ClassDB: `ClassDB::_add_class2`
- Code that adds classes to ClassDB post first initialization (`A`):
`memnew` macros -> `Object::_postinitialize` ->
`Object::initialize_class` -> `ClassDB::_add_class2`.
- Code adds class to ClassDB and registers/exposes it to scripts:
`GDREGISTER_CLASS` macros -> `ClassDB::register_class<T>` ->
`Object::initialize_class` -> `ClassDB::_add_class2`.
The previous error message incorrectly suggested that any Basis could be fixed by calling get_rotation_quation() or orthonormalize(). This PR points out that only a valid rotation Basis can be fixed in this way.
This commit completely removes the RPC_MODE_MASTER ("master" keyword),
and renames the RPC_MODE_PUPPET to RPC_MODE_AUTHORITY ("auth" keyword).
This commit also renames the "Node.[get|set]_network_master" methods to
"Node.[get|set]_network_authority".
This commit also renames the RPC_MODE_REMOTE constant to RPC_MODE_ANY.
RPC_MODE_MASTER in Godot 3.x meant that a given RPC would be callable by
any puppet peer on the master, while RPC_MODE_PUPPET meant that it would
be callable by the master on any puppet.
Beside proving to be very confusing to the user (referring to where it
could be called instead of who can call it) the RPC_MODE_MASTER is quite
useless. It is almost the same as RPC_MODE_REMOTE (anyone can call) with
the exception that the network master cannot. While this could be useful
to check in some case, in such a function you would anyway need to check
in code who is the caller via get_rpc_sender_id(), so adding the check
there for those rare cases does not warrants a dedicated mode.
Segment collision results could be different depending on the direction
when they exactly touch (order of the points in segments). This was due
to the way parallelism was checked, using different logic based on
positive or negative sign of cross products.
Now the results are the same whatever the direction, without changing
the current design, which is that parallel or colinear segments are
not considered colinear.
Fixes inconsistencies with raycasts exactly on edges of convex shapes
depending on the direction.
* Exposed functions in Engine to register and unregister singletons.
* Added the concept of user singletons, which can be removed (the system ones can't).
Adds ability to add error return documetation to the binder and class reference.
Usage example:
```C++
void MyClass::_bind_method() {
[..]
BIND_METHOD_ERR_RETURN_DOC("load", ERR_FILE_CANT_OPEN, ERR_FILE_UNRECOGNIZED);
}
```
One function of ConfigFile was changed as example.
* Allows calling into native extensions directly with a pointer
* Makes it easier to implement some APIs more efficiently
* Appears with a "*" in the documentation for the argument.
* Implementing the pointer handling is entirely up to the implementation, although the extension API provides some hint.
* AudioStream has been implemented as an example, allowing to create NativeExtension based AudioStreams.
* `_gui_input`, `_input`, `_unhandled_input` and `_unhandled_key_input` are now regular C++ virutal functions.
* Everything else converted to GDVIRTUAL
* BIND_VMETHOD is gone, always use the new syntax from now on.
Creating `_gui_input` method and using the binder to register events will no longer work, simply override the virtual function now.
The constructor was expecting a mutable pointer, while the passed pointer was immutable ( returns a const pointer), so the compilation was failing when iterating a constant .
* New syntax is type safe.
* New syntax allows for type safe virtuals in native extensions.
* New syntax permits extremely fast calling.
Note: Everything was replaced where possible except for `_gui_input` `_input` and `_unhandled_input`.
These will require API rework on a separate PR as they work different than the rest of the functions.
Added a new method flag METHOD_FLAG_OBJECT_CORE, used internally. Allows to not dump the core virtuals like `_notification` to the json API, since each language will implement those as it is best fits.
* Extensions are now scanned and loaded on demand.
* Extensions found are cached into a file that is used to load them (which is also exported).
* Editor will ask to restart when an extension requires core functionality.
* Editor will attempt to load extensions always before importing or loading scenes. This ensures extensions can register the relevant types.
Capsule height and radius setters can modify each other, rather than
using clamping, to avoid cases where values are not set correctly when
loading a scene (depending on the order of properties).
Inspector undo/redo:
Added the possibility to link properties together in the editor, so
they can be undone together, for cases where a property can modify
another one.
Gizmo undo/redo:
Capsule handles pass both radius and height values so they can be undone
together.
Like the spawn/despawn feature, it can be completely overridden with 2
custom callables.
The callables will be called with the list of tracked objects.
In SERVER mode, objects are automatically tracked, while in CUSTOM mode
you can manually track them via `track`/`untrack` (but that's optional).
The default sync only happens from server to client, with batch updates,
over unreliable channel (but with custom ordering).
The default sync will warn you, if your state representation gets too
big.
Move the former "spawnables" functions to a dedicated
MultiplayerReplicator class.
Support custom overrides in replicator.
Spawn/despawn messages can now contain a state.
The state can be automatically encoded/decoded by passing the desired
object properties to `spawnable_config`.
You can use script properties to optimize the state representation.
2 Callables can be also specified to completely override the default
implementation for sending and receiving the spawn/despawn event.
(9 bytes overhead, and there's room for improvement here).
When using a custom implementation `spawn` and `despawn` can be called
with any Object, `send_spawn`/`send_despawn` can receive any Variant as
a state, and the path is not required.
Two new functions, `spawn` and `despawn`, convey the implementation
independent method for requesting a spawn/despawn of an Object, while
`send_spawn` and `send_despawn` represent the more low-level send event
for a Variant to be used by the custom implementations.
Fixes undefined behavior, and fixes the logic for negative powers of ten.
Fixes#51764.
Adds tests to validate the changes and prevent regressions.
Adds docs for `String.num`.
- Add MethodBind call (besides ptrcall), since vararg methods don't work
with ptrcall.
- Fix argument name in register constant function to the way it
actually is used in the engine.
- Change the integer constant type to GDNativeInt to keep it consistent.
