Currently, if you were to call (for instance) write_value as part of a custom write_host_object implementation, you would be forced to create aliased exclusive references to the serializer (which is instant UB).
The underlying v8 de/serializers are re-entrant, so for the most part we simply need to pass around shared references instead of exclusive references.
Also adds some bindings to some missing functions, and makes ValueSerializer (and deserializer) garbage collectable.
* Rolling to V8 12.9.202.1
* initial changes for 12.9
* CallbackScope for fast fns
* disable broken thing by default
* use windows 2022 runner
---------
Co-authored-by: snek <the@snek.dev>
A subtle unsoundness / undefined behaviour made its way into the fairly recently added ExternalOneByteStringResource object: The as_str API is not sound as the data inside may be be Latin-1, not ASCII.
As the API was not used anywhere in deno or deno_core, I opted to simply remove it and replace it with an as_bytes API. I also modified the test to showcase the Latin-1 string case and added copious notes and explanations around the code to make sure this doesn't accidentally happen again. The likely reason why the API originally slipped in is because the OneByteConst has this API where it is safe because the OneByteConst creation checks the data for ASCII-ness.
I also tried to add an API to extract an Option<&'static OneByteConst> from an &ExternalOneByteStringResource but run into rust-lang/rust#119618 ie. OneByteConst is actually duplicating the vtables... which is not great.
This commit updates APIs related to named and indexed property
handlers to not use deprecated V8 APIs. Main change is the change
of return value in callbacks that now requires to return v8::Intercepted enum.
Towards #1478
1. [[exceptions] Unify pending and scheduled exceptions](https://chromium-review.googlesource.com/c/v8/v8/+/5050065)
Reset no longer clears exception if it was rethrown. The test had to be adjusted for the same.
2. [[api] Allow passing CppHeap on Isolate creation](https://chromium-review.googlesource.com/c/v8/v8/+/4989254)
`AttachCppHeap` was deprecated but the alternative of passing `CppHeap` via Isolate CreateParams hard crashes (SIGSEGV). There are no tests for this in V8 and it seems the [Chromium CL](https://chromium-review.googlesource.com/c/chromium/src/+/4992764) is also crashing. For now I've just suppressed the deprecation warning until the crash is fixed in V8.
3. v8::Serializer impl must not throw more than one exception.
I changed `get_shared_buffer_id()` to not throw and return `None`. V8 internally calls data clone error when it's the SAB is not clonable.
Other changes:
- `v8::ScriptCompiler` size increased by 3 words with `v8::ScriptCompiler::CompilationDetails`.
- `v8::ObjectTemplate::SetAccessor` & `v8::ObjectTemplate::SetAccessorProperty` signature changed and also deprecated.
- `v8::Context::SetContinuationPreservedEmbedderData` deprecated. Use `v8::Isolate::GetContinuationPreservedEmbedderData` instead.
- `GetStalledTopLevelAwaitMessage` deprecated. Use `GetStalledTopLevelAwaitMessages` instead.
- `v8::Isolate::AttachCppHeap` deprecated. Set the heap on Isolate creation using CreateParams instead.
- `v8::ScriptOrigin` deprecated. Use constructor without the isolate.
- `v8::SnapshotCreator` is deprecated. Use the version that passes CreateParams instead.
- `v8::Isolate` assertion failures.
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
Due to the automatic entry and exit behavior of Isolate upon creation and drop, it is crucial to ensure that v8::OwnedIsolate instances are dropped in the reverse order of their creation. Dropping them in the incorrect order can result in the corruption of the thread-local stack managed by v8, leading to memory corruption and potential segfaults. This introduces a check to verify the `this == Isolate::GetCurrent()` requirement before invoking the exit function. If the requirement is not met, a clean panic is triggered to provide explicit error handling instead of allowing silent memory corruption.
