It is possible for a PaintTask to start exiting soon after sending new
buffers to the compositor. In that case, the compositor should return
the now unnecessary buffers to the PaintTask so that it can properly
free them.
To accomplish this, the compositor now keeps a hash map of paint task
channels per pipeline id. When a PaintTask exists, the constellation
informs the compositor that it can forget about it. Additionally, the
PaintTask should not wait for any buffers when the engine is doing a
complete shutdown. In that case, the compositor is already halted and
has simply let all buffers leak. We pipe through the shutdown type when
destroying the pipeline to make this decision.
Fixes#2641.
This changes those calls whose unsoundness was not picked up by the type system
because of a lifetime constraint that cannot be expressed at this time.
This replaces the specialized TrustedXHRAddress and TrustedWorkerAddress code that was used for the same purpose. A non-zero refcount pins the given DOM object's reflector and prevents it from being GCed even when there are no other outstanding references visible to SpiderMonkey. This will enable us to implement asynchronous operations that refer to particular DOM objects (such as "queue a task to fire a simple event named load at the iframe element" from the spec) safely and conveniently, and paves the way for things like asynchronous network responses.
Some concerns about the resulting size of XHR progress messages have been expressed, but I believe optimizations to reduce that can be implemented in subsequent PRs.
r? @Ms2ger - note in particular the changes to the worker lifetime code. I couldn't figure out how to achieve an identical lifetime to the previous addref/release pairing, and I also was having trouble figuring out why the existing setup was safe. The new implementation now holds the main script task Worker object alive via the TrustedWorkerAddress field in the dedicated worker global scope, which is a significant difference.
Now `#[dom_struct]` also generates Reflectable impls, and there's another lint to ensure that a DOM struct only contains one bare DOM field (as the first field) or a Reflector.
A lot of this was generated by sed -- each autogenerated change has its own commit for easy review; these will be squashed later.
