The idea is that frequently accessed properties, e.g. window.document, can be
cached directly as data properties on the underlying v8::Object, removing the
need for the access call into Zig. This is only used on a handful of properties,
almost all of which are on the Window. It is important that the property be
read-only. For example, window.location cannot be cached this way because
window.location is writable (e.g. window.location.hash = '#blah').
This existed briefly before Zigdom, but was removed as part of the migration.
The implementation has changed. This previously relied on a "postAttach" feature
which no longer exists. It is not integrated in the bridge/callback directly and
lazily applied after the first access.
Previously, we were doing some global setup in the Snapshot, but this was
always being overwritten when creating a context. This meaningless setup in
the snapshot was removed.
The global setup is now done once per isolate, rather than once per context.
Deciding what should be an lp.assert, vs an std.debug.assert, vs a debug-only
assert is a little arbitrary.
debug-only asserts, guarded with an `if (comptime IS_DEBUG)` obviously avoid the
check in release and thus have a performance advantage. We also use them at
library boundaries. If libcurl says it will always emit a header line with a
trailing \r\n, is that really a check we need to do in production? I don't think
so. First, that code path is checked _a lot_ in debug. Second, it feels a bit
like we're testing libcurl (in production!)..why? A debug-only assertion should
be good enough to catch any changes in libcurl.
This adds a crash handler which reports a crash (if telemetry is enabled). On a
crash, this looks for `curl` (using the PATH env), and forks the process to then
call execve. This relies on a new endpoint to be setup to accept the "report".
Also, we include very little data..I figured just knowing about crashes would
be a good place to start.
A panic handler is provided, which override's Zig default handler and hooks
into the crash handler.
An `assert` function is added and hooks into the crash handler. This is
currently only used in one place (Session.zig) to demonstrate its use. In
addition to reporting a failed assert, the assert aborts execution in
ReleaseFast (as opposed to an undefined behavior with std.debug.assert).
I want to hook this into the v8 global error handler, but only after direct_v8
is merged.
Much of this is inspired by bun's code. They have their own assert (1) and
a [more sophisticated] crashHandler (2).
:
(1) beccd01647/src/bun.zig (L2987)
(2) beccd01647/src/crash_handler.zig (L198)
When we create a js.Context, we create the underlying v8.Context and store it
for the duration of the page lifetime. This works because we have a global
HandleScope - the v8.Context (which is really a v8::Local<v8::Context>) is that
to the global HandleScope, effectively making it a global.
If we want to remove our global HandleScope, then we can no longer pin the
v8.Context in our js.Context. Our js.Context now only holds a v8.Global of the
v8.Context (v8::Global<v8::Context).
This PR introduces a new type, js.Local, which takes over a lot of the
functionality previously found in either js.Caller or js.Context. The simplest
way to think about it is:
1 - For v8 -> zig calls, we create a js.Caller (as always)
2 - For zig -> v8 calls, we go through the js.Context (as always)
3 - The shared functionality, which works on a v8.Context, now belongs to js.Local
For #1 (v8 -> zig), creating a js.Local for a js.Caller is really simple and
centralized. v8 largely gives us everything we need from the
FunctionCallbackInfo or PropertyCallbackInfo. For #2, it's messier, because we
can only create a local v8::Context if we have a HandleScope, which we may or
may not.
Unfortunately, in many cases, what to do becomes the responsibility of the caller
and much of the code has to become aware of this local-ness. What does it means
for our code? The impact is on WebAPIs that store .Global. Because the global
can't do anything. You always need to convert that .Global to a local
(e.g. js.Function.Global -> js.Function).
If you're 100% sure the WebAPI is only being invoked by a v8 callback, you can
use `page.js.local.?.toLocal(some_global).call(...)` to get the local value.
If you're 100% sure the WebAPI is only being invoked by Zig, you need to create
`js.Local.Scope` to get access to a local:
```zig
var ls: js.Local.Scope = undefined;
page.js.localScope(&ls);
defer ls.deinit();
ls.toLocal(some_global).call(...)
// can also access `&ls.local` for APIs that require a *const js.Local
```
For functions that can be invoked by either V8 or Zig, you should generally push
the responsibility to the caller by accepting a `local: *const js.Local`. If the
caller is a v8 callback, it can pass `page.js.local.?`. If the caller is a Zig
callback, it can create a `Local.Scope`.
As an alternative, it is possible to simply pass the *Page, and check
`if page.js.local == null` and, if so, create a Local.Scope. But this should only
be done for performance reasons. We currently only do this in 1 place, and it's
because the Zig caller doesn't know whether a Local will actually be needed and
it's potentially called on every element creating from the parser.
This is the first in a series of changes to make globals explicit. The ultimate
goal of having explicit Globals is to move away from the global HandleScope and
to explicit HandleScopes.
Currently, we treat globals and locals interchangeably. In fact, for Global ->
Local, we just ptrCast. This works because we have 1 global HandleScope, which
effectively disables V8's GC and thus nothing ever gets moved.
If we're going to introduce explicit HandleScopes, then we need to first have
correct Globals. Specifically, when we want to act on the global, we need to
get the local value, and that will eventually mean making sure there's a
HandleScope.
While adding explicit globals, we're keeping the global HandleScope so that we
can minimize the change. So, given that we still have the global HandleScope
the change is largely two things:
1 - js.Function.persit() returns a js.Function.Global. Types that persist global
functions must be updated to js.Function.Global.
2 - To turn js.Function.Global -> js.Function, we need to call .local() on it.
The bridge has been updated to support js.Function.Global for both input and
output parameters. Thus, window.setOnLoad can now directly take a
js.Function.Global, and window.getOnLoad can directly return that
js.Function.Global.
This is the first in a series of changes to make globals explicit. The ultimate
goal of having explicit Globals is to move away from the global HandleScope and
to explicit HandleScopes.
Currently, we treat globals and locals interchangeably. In fact, for Global ->
Local, we just ptrCast. This works because we have 1 global HandleScope, which
effectively disables V8's GC and thus nothing ever gets moved.
If we're going to introduce explicit HandleScopes, then we need to first have
correct Globals. Specifically, when we want to act on the global, we need to
get the local value, and that will eventually mean making sure there's a
HandleScope.
While adding explicit globals, we're keeping the global HandleScope so that we
can minimize the change. So, given that we still have the global HandleScope
the change is largely two things:
1 - js.Function.persit() returns a js.Function.Global. Types that persist global
functions must be updated to js.Function.Global.
2 - To turn js.Function.Global -> js.Function, we need to call .local() on it.
The bridge has been updated to support js.Function.Global for both input and
output parameters. Thus, window.setOnLoad can now directly take a
js.Function.Global, and window.getOnLoad can directly return that
js.Function.Global.
TL;DR - use page.arena instead of page.call_arena
This probably comes from copying the implementation of MutationObserver and/or
IntersectionObserver. But those dispatches are different in that they directly
dispatch a slice (e.g. of MutationRecords) which gets mapped to a v8::Array when
doing the callback. The MutationRecords exist on the heap, not in
_pending_records, so the call_arena is fine.
PerformanceObserver returns an Zig object, not a slice. Therefore it gets mapped
to a v8::Object which references the Zig object. The state of that object, the
_entries list, has to exist for the lifetime of that object, not the call_arena.
