This allows the EMotionFX runtime to compile with `/we4267` enabled, which
emits a warning when converting from `size_t` to a smaller type. All tests
for the runtime have been updated accordingly, and they pass.
In instances where a range-for loop could be used, or a std algorithm, that
was used instead of using `size_t numItems = vec.size()` and a for loop.
Casts to `uint32` were removed where possible. Some places remain, like in
the file formats.
Signed-off-by: Chris Burel <burelc@amazon.com>
This translates all usages of MCore::Array to AZStd::vector. It is
designed to be as minimal of a change as possible (no changing to
range-for loops or other C++11 stuff).
We can decide to submit this wholesale, or submit it to a separate
branch that we can then integrate individual files from once we're ready
to do a specific class's transition.
It does not completely solve the `uint32`->`size_t` transition.
One important finding from doing this: `MCore::Array` uses a `memcpy`
when it reallocates. `AZStd::vector` will use the contained type's copy
or move constructor, per element. This is a significant change in
behavior. If you have type, `SomeStruct` that defines a destructor, that
type is copyable and not movable. So if you have a
`MCore::Array<SomeStruct>`, and you call `Add(); Add(); Add()`, that
reallocates 3 times, copying the contents using `memcpy`, and never
invokes `SomeStruct`'s copy constructor or destructor. Translating that
to `AZStd::vector<SomeStruct>` and calling `push_back(); push_back();
push_back();` will still reallocate 3 times, but it sees that
`SomeStruct` is non-movable, and uses the copy constructor to make the
copies, and then the destructor on the previous values. This call to the
destructor wasn't there before, and can cause things to be deleted that
weren't before. The solution to this is to make that struct be a
move-only type. Where possible, this was done by changing that type to
use `AZStd::unique_ptr` instead of a raw pointer, to get the proper move
behavior. Where that is not possible (types that inherit from
`MCore::MemoryObject`), a hand-written move constructor was created.
In general:
GetLength() becomes size()
GetMaxLength() becomes capacity()
GetIsEmpty() becomes empty()
Reserve() becomes reserve()
ReserveExact() becomes reserve()
Resize() becomes resize()
ResizeFast() becomes resize_no_construct()
Add() becomes emplace_back()
AddExact() becomes emplace_back()
AddEmpty() becomes emplace_back()
AddEmptyExact() becomes emplace_back()
GetPtr() becomes data()
GetItem() becomes at()
Shrink() becomes shrink_to_fit()
GetFirst() becomes front()
GetLast() becomes back()
Remove() becomes erase()
RemoveFirst() becomes erase()
RemoveLast() becomes pop_back()
RemoveByValue() becomes if (const auto it = AZStd::find(...); it != end(container)) container.erase(it);
Insert() becomes emplace()
Swap() becomes swap()
Clear(true) becomes clear(); shrink_to_fit()
Clear() becomes clear(); shrink_to_fit()
Clear(false) becomes clear()
Swap() becomes swap()
Find() becomes AZStd::find
MoveElements() becomes AZStd::move
SetMemoryCategory() is removed
Signed-off-by: Chris Burel <burelc@amazon.com>
* Final update copyright headers to reference license files at the repo root
Signed-off-by: spham <spham@amazon.com>
* Fix copyright validator unit tests to support the stale O3DE header scenario
Signed-off-by: spham <spham@amazon.com>
* Fixes saving motions from within the Animation Editor
* Fixes saving actors from within the Animation Editor
* The motion event chunk of the .motion file format now also stores the event data as json (rather than XML) reducing motion file sizes (Example: 60KB motion went down to 49KB, containing only 4 motion events from 2 tracks).
* Fully backward compatible
* New motion meta data rule stores the event data directly rather than command strings or objects. This is the way that aligns with the Json paradigm and as side-effect bypasses the optionals that we use for the commands which fixes the issue.
* [LYN-4574] Adding new motion event meta data rule that stores the event data directly rather than via commands to align with the Json paradigm
* [LYN-4574] Preparing motion, event table and event track for Json serialization
* [LYN-4574] New chunk to store motion event data in Json format (fully backward compatible to XML)
* [LYN-4669] Json: Empty AZStd::vector<AZStd::shared_ptr<T>> serializes into 1x element with nullptr as data
* [LYN-4603] EMotion FX: Cannot save actors with physics or simulated object setup in Json format
- Invalidate 'Physics Materials From Mesh' boolean from collider component
- Removed material library from material selector. Default material library will always be used instead.
- Marking failing automated test as xfail
- Added default material to physics configuration.
- Moved material library asset from physx configuration to physics configuration, as it doesn't need to be physx specific.
- Refactor physics material system having into account that there is only one material library in the project.
- Renaming code from DefaultMaterialLibrary to MaterialLibrary.
- All queries about physics materials unified under PhysicsMaterialRequests bus.
- PhysXSystem only manages the material library asset.
- Saving and reloading the same physics material asset with different content didn't trigger a events that the material library has changed.
- Changing Physics Material Request interface to use shared_ptr instead of weak_ptr to be simpler to handle the returned materials and having a more consistent code.
- Refactored Material Manager to improve its implementation. Still following the same approach of "creating materials on the fly as they are requested", but now it's doing it consistently across the interface, with private helpers functions FindOrCreateMaterial that simplify vastly the implementation.
- Material Manager now listens to change event of material library asset and default material configuration so it updates its materials accordingly.
- Complete Material move constructor and operator.
Delete:
- I3DEngine
- Missions
- Time of day
- GameEffectsSystem Gem
- ImageProcessing Gem
- SVOGI Gem
- Various other things that don't do anything now that the legacy renderer has been removed.
EMotionFX has user-customizable hotkeys. These hotkeys are registered by
individual plugins, and then the user can set what they want the hotkey to
be. The way this was implemented was by reimplementing `keyPressEvent` and
`keyReleaseEvent` for each widget that used customizable hotkeys, and in
there call `KeyboardShortcutManager::Check` to see if key press matched any
existing hotkey mapping.
However, the main Editor has behavior that prevents events from reaching
EMotionFX's `keyPressEvent` method, if a keypress matches a hotkey that is
also used by the main Editor. This is due to the global event filter
defined in `ShortcutDispatcher::eventFilter`. This event filter takes a Qt
`Shortcut` event, and will re-dispatch that event to all parent widgets of
a given receiver. So if a parent widget, like the main Editor, *does* have
a QAction that matches a given key sequence, that widget will receive the
event, the event is marked as processed, and no `KeyPress` event is ever
sent to the original widget where the event occurred.
All this means that processing hotkeys in a `keyPressEvent` won't work
reliably. The main editor defines a hotkey for the `delete` key, so that
can never be received in a `keyPressEvent` by any child widget of the
Editor.
This change removes all the hotkey logic from the `keyPressEvent` methods,
and replaces them with `QAction` instances. Hotkeys are defined with
`QAction::setShortcut`, and added to each widget that they apply to.
In addition, the `KeyboardShortcutManager` class had to be adjusted to suit
this new way of defining the hotkeys. It now has a pointer to each
`QAction*` that can have a customizable hotkey. It has also been greatly
simplified, since it can use a `QKeySequence` instead of separate variables
for `int key; bool hasCtrlModifier; bool hasAltModifier`.
Applying user-defined hotkeys now has to be done after the hotkeys are
registered from a plugin. It is the plugin's responsibility to reload the
user-defined hotkeys after registering all of its actions.
- Fixing cloth working with Actors by reading directly from ModelAsset instead from EMFX Mesh.
- Actor caches map from joint indices in skin metadata to skeleton indices so they can be query later.
- Actor cloth skinning reads indices and weights from Model Asset instead from EMFX Mesh.
- Actor cloth skinning with unlimited skinning bones.
- Sort out cloth unit tests by disabling them until there is a way to create an Atom mesh
- Addressing feedback.
Esteban Papp
pappeste
Chris Burel
hultonha
Benjamin Jillich
Benjamin Jillich
Yuriy Toporovskyy
Steve Pham
Chris Galvan
AMZN-stankowi
Gene Walters
amzn-sean
Aaron Ruiz Mora
lumberyard-employee-dm
Terry Michaels
bosnichd