The MovieClip variable oldest_parent_swf_version and its getter method
have been removed because they were unnecessary, as the UpdateContext
already provides the root movie.
The call of the getter method has been replaced by
activation.context.swf.version().
The suggested changes in the feedback to the pull request have been
implemented.
Therefore, this commit consists of multiple smaller changes:
- The unload_movie method has been renamed to avm1_unload_movie.
- The movieclip_default_state test now doesn't test
getTextSnapshot().getCount() because the underlying methods haven't
been implemented in AVM1 yet.
This should be reverted when they have been implemented or stubbed in
AVM1.
An unload_movie method has been added to the MovieClip. It unloads the
MovieClip, which means that one frame after it has been called,
avm1_unload and the new transform_to_unloaded_state method get called
and the MovieClip enters the unloaded state in which some attributes
have certain values.
To do this, it creates and spawns a future which calls avm1_unload and
transform_to_unloaded_state (which actually makes the MovieClip enter
the unloaded state).
It uses the new MovieUnloader (which has been added to the Loader enum)
to pass the MovieClip to the asynchronous code. The MovieUnloader saves
a handle and the target clip, and the asynchronous code retrieves it
from the LoadManager.
unload_movie is now called when a MovieClip should be unloaded.
TODOs have been added, describing to check whether other avm1_unload
calls also need to be delayed by one frame and whether unload_movie also
needs to be called in other places.
Additionally, a spelling mistake has been corrected.
Previously, the MovieClip default state contained wrong values for
opaqueBackground, transform.pixelBounds and getRect & getBounds with
another default MovieClip as the parameter.
This has been corrected; the calculation of these values has been fixed.
To do that, a use_new_invalid_bounds_value variable has been added to
the UpdateContext and the Player (from which the UpdateContext is
created).
Additionally, an oldest_parent_swf_version variable has been added to
the MovieClip with a getter method for it. It is calculated and saved on
demand in the getter.
These variables are retrieved in the get_bounds function and used to set
use_new_invalid_bounds_value and to calculate the correct bounds.
Ruffle now uses the correct MovieClip default state.
The set_cur_preload_frame and set_current_frame methods have been added
to the MovieClip.
If the Loader now loads an image, these methods are called to set the
MovieClip image state correctly.
Additionally, load_error_swf now uses set_cur_preload_frame instead of
the removed movie_not_available method.
A load_initial_loading_swf function has been added to the Loader. It
makes the MovieClip enter the initial loading state in which some
attributes have certain initial loading values to signal that the file
is currently being loaded and neither an error has occurred nor the
first frame has been successfully loaded yet.
The Loader calls the function when initialising the loading before it
gets an error or success.
Additionally, load_error_swf has been improved: If a local URL is
fetched on a WASM target, the _url property will now be correctly set as
in the flash player.
The documentation has been improved.
All NavigatorBackend implementations have been refactored, resulting in
improved code quality, less duplicated code, more consistent and easier
to understand procedures, additional error handling and better error
messages.
A resolve_url method has been added to the NavigatorBackend trait. It
takes a URL and and resolves it to the actual URL from which a file can
be fetched (including handling of relative links and pre-processing). It
has been implemented in each NavigatorBackend implementation.
Duplicated code has been put into new public functions in
core/src/backend/navigator.rs which are called by all NavigatorBackend
implementations.
ExternalNavigatorBackend:
- The navigate_to_url and fetch methods have been adapted to use
resolve_url, removing redundant code.
- Error handling has been added in case that the URL can't be converted
to a PathBuf.
- A TODO about differences between the flash player fetch and the
Ruffle fetch implementation has been added.
WebNavigatorBackend:
- The previous resolve_url method exclusively to the WebNavigatorBackend
has been replaced by the new resolve_url method. It is used by
navigate_to_url and fetch.
- resolve_url now always pre-processes the URL if it's valid (even if no
base_url exists) and explicitly returns whether the URL can be parsed.
- navigate_to_url now traces an explanatory error each if the URL can't
be parsed or is local.
- fetch now returns an explanatory error each if the URL can't be parsed
or is local (previously, a vague "Got JS error" has been returned).
TestNavigatorBackend & NullNavigatorBackend:
- fetch pre-processes the URL now (using the resolve_url implementation).
- If the URL isn't local, an explanatory error is returned (previously,
it was just an "Invalid URL" error).
- If the URL can't be parsed, an explanatory error with the reason is
returned (previously, it was just an "Invalid URL" error).
Additionally, error messages in all NavigatorBackend implementations
have been improved and made more consistent, e.g. if a local file can't
be read.
Previously, the MovieClip error state contained one TODO in order to be
completely implemented: The HeaderExt variable uncompressed_len needs to
be -1 in the error state, but its type has been u32.
The type has now been changed to i32, and the variable is set to -1 in
the default_error_header function. Therefore, the MovieClip error state
is now fully implemented.
Other connected functions and variables like SwfMovie::uncompressed_len
or MovieClip::total_bytes have been adjusted to this type change. Code
has been adjusted if needed where this value is used.
A load_error_swf function has been added to the Loader. It makes the
MovieClip enter the error state in which some attributes have certain
error values to signal that no valid file could be loaded. This happens
if no file could be loaded or if the loaded content is no valid
supported content.
The function creates an error state movie stub using the new
SwfMovie::error_movie function (which uses a new default_error_header
function) and configures remaining variables with the
movie_not_available method.
One TODO in order for the error state to be completely implemented has
been added.
Since the error state of the MovieClip includes the final URL of the SWF
file obtained after any redirects, the load_error_swf and
movie_loader_error functions (now) take an swf_url attribute.
To get this URL in case no file could be loaded, the
NavigatorBackend::fetch method has been changed to return an
ErrorResponse struct (including the url and the actual error) in the
error case. The Response struct returned in the success case has been
renamed to SuccessResponse.
All fetch implementations have been adapted accordingly. Code has been
adjusted to return the actual error where that's needed.
Documentation has been added and improved.
If an SWF tries to load a movie that can't be loaded, the Flash Player
sets _framesloaded of the MovieClip to -1. This has previously not been
implemented in Ruffle.
A new movie_not_available function has been added to the MovieClip
class. It sets the cur_preload_frame variable to 0.
This function is now called in the movie_loader_error function of the
Loader class (if a movie can't be loaded).
The frames_loaded function (which returns _framesloaded of the
MovieClip, which is given to the SWF) now returns an i32 and not an u16
value. This allows it to return a negative value. Code has been adjusted
where an u16 value is expected (clamping negative values to 0).
frames_loaded now returns cur_preload_frame - 1. Previously, it has
returned cur_preload_frame - 1 and clamped it to zero, but since the -1,
in case cur_preload_frame is zero, is explicitly wanted, this has been
corrected.
* avm2: Do not remove an EditText's selection on unfocus (fix#9006)
There are significant differences between how selection and caret info
work between AVM1 and AVM2.
In AVM1, there is only a single, global selection, which applies to
whichever element is currently focused. Therefore changing the focus
necessarily erases any information of what was selected before. There
can also be no active selection at all.
In AVM2, every text field has its own independent selection info. It is
NOT optional, the default selection is just a caret at position 0. If a
field loses focus, the selection is not rendered, but it is still
present. Movies such as #9006 rely on the same selection still being
there once you give back focus to the field.
Ruffle's model of selections (an Option per text field) is different
from both the AVM1 model (an optional singleton) and the AVM2 model
(mandatory per text field). This fix does not change that, it is only a
narrow fix targeted at 9006.
* avm2: implement selectionBegin/selectionEnd/caretIndex and add a test.
The selection test validates a few situations, including the behaviour
when unfocusing that was fixed in the previous commit.
The test does not validate how the selection changes after replacing
text, as there are still some inaccuracies there.
* avm2: Additional selection fixes needed after merging with master.
1. The default caret for an AVM2 textbox is at the end of the textbox,
not the beginning.
2. Selection should not be changed when focusing on a textbox in AVM2.
3. Fixed a test whose output.txt didn't actually match the flash player
output.
* avm2: Make the selection AVM checks compatible with mixed AVM, and revert the on_focus_changed parameters
---------
Co-authored-by: Nathan Adams <dinnerbone@dinnerbone.com>
The 'gc_arena' dependency was only used to manipulate the `GcCell`s
containing the vertex and fragment shaders; replacing these by a
reference to a plain old `Cell` means tha the Context3D traits and
types do not need to interact with GC'd object anymore.
As a knock-on effect, we can also remove the `Activation` parameter
from most of the `Context3DObject` methods.
* core: add temporary, ruffle-internal copy of `gc-arena` crate
This will allow bumping the upstream `gc-arena` version while
reexporting our own version of the old `GcCell` API, so that
Ruffle's code can be gradually migrated.
Once the migration is done, this crate should be removed.
* core: bump `gc-arena` to kyren/gc-arena#56
Add back the removed `GcCell` to our internal facade crate
* core: bump `gc-arena` to current master
This bump renames `Gc::allocate` to `Gc::new`
* core: rename `GcCell::allocate` to `GcCell::new`, to match `Gc`
* core: bump gc-arena to (slighly after) v0.3.1
Add typedefs for old `*Context` names in the gc-arena facade crate
* core: replace uses of `CollectionContext<'_>` by `&Collection`
* core: Add `gc()` convenience method for `*Context` and `Activation` types
This allows shortening most instances of `[activation.]context.gc_context`
to `activation.gc()` or `context.gc()` (but not all instances, because of
borrowck) Note that this doesn't actually do these shortenings to avoid
major code churn.
The bind group layout only depends on the texture registers
(and 2D/cubemap type) accessed by the fragment shader, not on
the runtime texture bound with Context3D. This means that we can
build and cache it when we compile the AGAL program to a Naga
module.
Since the bind group layout is used for the overall pipeline, I've
refactored the shader caching code into `ShaderPairAgal`, which
holds both the vertex and fragment shader bytecode, and compiles
both in the `compile` function.
When using a 'Loader', properties on the 'contentLoaderInfo' become
set during specific events in the load sequence. In particular,
'LoaderInfo.bytesTotal' becomes available during the first 'progress'
event.
Also, 'LoaderInfo.parameters' is now properly set from the URL query
parameters. In Flash player, this work even with filesystem urls
(e.g. 'file:///some/path/to/file.txt?paramOne=valOne' will load
a file named 'file.txt', setting and expose the parameter 'paramOne'
with value 'valOne' in `LoaderInfo.parameters`). This required some
cleanup to the desktop and test NavigatorBackend impls to strip
out query parameters when loading a parameter from disk.
Previously, we would set `SwfMovie.parameters` manually from the url.
Now, the various `SwfMovie` constructors automatically extract
query parameters from the provided url. Outside of `SwfMovie`,
we only append *extra* parameters (e.g. those set from `flashvars`).
This makes CPMStar ads work, since the loaded SWF needs to access
`LoaderInfo.parameters`
The 'winding' argument and filling behavior described in the docs
are not yet implemented. However, this implementation is good enough
for Scratch to render its default cat image.
We were previously only ever checking children,
and not attributes.
In order to avoid matching both attributes and elements
with a given name in 'descendants', `E4xNode::matches_name`
now checks `is_attribute` on the provided `Multiname`. This
requries changing several other parts of the codebase to
properly set this flag on `Multinames` provided by ActionScript.
The only generic class is `Vector`, which only has a single parameter.
However, we currently store a list of type parameters in several places,
which requires redundant checks that the list only has one entry.
Instead, we now store an `Option` everwhere except for
`swf::Multiname::TypeName` (since the actual SWF format supports
multiple type parameters). We still perform the same check when loading
bytecode (the swf Multiname should only have at most one type
parameter), but the rest of the codebase can deal with an `Option`
instead.
In the special case where y_min==y_max==(height-1),
we would create a bad 'encompassing' region - subtracing
one from the max would make it smaller than the min,
causing `PixelRegion::encompassing_pixels` to treat it
as the minimum, and add one to `height - 1`
There's no work to do when x_min==x_max or y_min==y_max,
so we can also skip the sync and GC write in this case.
This works a bit differently from SWF videos, we do not retain decoded video frames or valid seek points indefinitely. This assumes code interacting with `NetStream` is at least mildly aware of how a video codec is supposed to work and what the performance implications of that are.
We don't actually process tags out of order, we're just tracking what tags have already been processed. FLV embeds critical video metadata in a ScriptData tag that we need to pull data out of and process.
`XML(someXMLObj)` and `XMLList(someXMLList)` perform the
normal cast behavior (returning the same object), instead
of creating a new object like other arguments do.
This also applies to `new XMLList(someXMLObj)`
and `XML(singleElemXMLList)`
* wpgu: Initial implementation of PixelBender shader execution
The implementation is split across four crates:
* `ruffle_render` now holds the main PixelBender bytecode parsing
implementation (previously, this was in `ruffle_core`).
* `ruffle_core` holds some helper functions for converting between
AVM2 `Value`s and the PixelBender vector types.
* `naga-pixelbender` (newly created) constructs a Naga `Module`
from parsed PixelBender bytecode
* `ruffle_render_wgpu` sets up the render pipeline for the shader
constructed by `naga-pixelbender`, and actually executes the shader.
The Actionscript-side shader parameters are passed in through uniforms.
This allows us to cache the compiled `naga::Module` and associated
wgpu types inside `ShaderData`, when it's first created. Each invocation
of a `ShaderJob` only needs to create a bind group and render pass.
Limitations:
* Only a few of the PixelBender opcodes are implemented - however, this is
enough to get Stemlands cannon rotation working, as well as a cool
"donut" shader that I found and included as a test.
* PixelBender matrix types are not supported.
* Only BitmapData is supported as an input/output type - Flash Player
also supports using Vector and ByteArray
* ShaderJob execution is always synchronous.
* Adjust comments
* Address review comments
We create a separate child domain, which is accessible
from the Stage and the root movieclip.
This prevents ActionScript from loading classes into the
special playerglobals domain (Domain.parentDomain is modified
to return null instead of the playerglobals domain when applicable),
so the native method lookup logic will never run for user code.
This lets us print the full error message and stack stacke
in contexts where an Activation is not available, including
the `Debug` impl.
The 'name' and 'error' fields are accessed using hardcodd slot
ids. This is pretty hacky, but will work until we have better
handling of slot properties.
We now support deleting named children, as well as attributes.
Additionally, I've fixed our handling of `XML.parent()` - we now
properly set the parent when a child is created, and clear the parent
when `delete` is used.
If a child named 'foo' is removed by the timeline (without
having been previous added/removed from the timeline by ActionScript),
then the 'foo' field in the parent will be set to null. This occurs
even if the 'foo' field in the parent is not currently set to
the child 'foo' (e.g. 'this.foo = someOtherObject' was executed by
ActionScript).
- we can now get a `&RawTable` from a `&HashSet`, meaning we can store a
`HashSet` directly in the interner;
- `RawTable::iter_hash` now allows the removal of already-yielded
elements during iteration, which simplifies `WeakSet::entry`
Previously `last_mouse_position` was updated irrespectively of
whether the dragged object was inside or outside the constraint
box. Change it follow mouse deltas, after clamping is performed.
Fixes#11254.
When we run a 'goto', a weird "nested frame" gets triggered.
Previously, we were only calling `construct_frame` on the target
MovieClip as part of this "nested frame". However, Flash Player
seems to treat this (in some ways) like a normal frame - *all*
objects on the Stage (and orphans) have `construct_frame` called.
In particular, `gotoAndStop`/`gotoAndPlay` is called during
an "enterFrame" event handler, then unrelated objects on the Stage
will have their children constructed during the execution of
`gotoAndStop`/`gotoAndPlay`. The same logic holds for frame scripts.
This fixes a bug in Steamlands, which relies on children on the main
timeline being constructed immediately following a call to `gotoAndStop`
on an orphan (originally triggered from an "enterFrame" handler).
When we iterate over the orphan list to run frame lifecyle methods,
the orphan list may be modified (e.g. an event listener creates a new
orphan object). To ensure that we iterate over all of the orphans
originally present, this commit wraps the orphan list in an `Rc`
(just like we do for the render list). Modifications to the list use
`Rc::make_mut`, and iteration operates on a clone of the `Rc`.
QName can contain characters like '.' and '<' as part of the
package or class name (though this requires editing the SWF
or using a nonstandard compiler). This broke our attempt to parse
generic type paramters by looking for '.<'
Instead, our 'Vector' special casing now operates on the unparsed
'AVMString' name, instead of attempting to construct a 'QName'.
This means that we don't need to handle generic paramters at
all for obfuscated names (which will never start with '__AS3__'
or 'Vector.<')
This fixes a bug in Red Ball 4v3, which has an obfuscated class
'!D.<H'
If we have two PlaceObject tags in the same frame with the same depth,
only the first one actually places an object. The second one is ignored
(Flash Player logs a warning).
There are only a few places where we actually need to use the
`caller_domain`, so we don't actually need it available for most native
method calls. This means that `Activation::from_nothing` can be used
in the vast majority of cases without causing a panic later on.
We now parse PixelBender bytecode, and populate the parameters
from the bytecode on `ShaderData`. This is enough to progress
Steamlands, which needs to access dynamically set properties
on `ShaderData`
Bytecode execution is not implemented yet.
When we create a DisplayObject from ActionScript, we should always
run `construct_frame` on it, regardless of what frame phase we
are currently in.
This fixes a regression in Fancy Pants World 4 Part 1, where entering
the first door produced an error.
Flash Player performs `x + width` and `y + height` as floating
point operations before `round_to_even`. This affects the extent
covered by a `Rectangle` in various BitmapData methods, as the sum
of two values might be large enough to be rounded up to a larger
value (when rounding `x` and `width` individually would have
produced a smaller overall extent).
These are directly set on the underlying navigator's HTTP
request type, and get printed out in our test navigator backend.
No validation of the header names is performed - on web, this will be
enforced by the browser.
When we iterate over a render list (in order to call
`enter_frame`, `construct_frame, etc.) we need to be sure
that we iterate over all of the original `DisplayObjects`
in the list, even if the list is modified during iteration
(e.g. some ActionScript code calls `parent.addChildAt`).
Previously, `RenderIter` would repeatedly call `child_by_index`
on the original `DisplayObjectContainer`, up to the original
child count. If new DisplayObjects were inserted into the list
during iteration, we could miss some of the original DisplayObjects
in the list (as they are now at a greater position in the list).
To solve this, we now store the render list as an `Rc<Vec>`,
and use `Rc::make_mut` to modify it. See the comments
for more details.
With SWF version < 30, Context3D.configureBackBuffer throws
an error with a less informative message when the width/height
are out of range. Additionally, it seems to special case
the case width=0, height=0, antiAlias=0. enableDepthAndStencil=false,
and does *not* throw an error. This is relied on by Sniper Team.
Kornelius Rohrschneider