| Auto-Submit | +1 |
| Commit-Queue | +1 |
Samuel: as discussed, PTAL.
Thibaud: FYI, this could be useful for retired `WasmSuspenderObject`s.
Tagged<WasmSuspenderObject> suspender =Drive-by fasterification: avoiding repeated handle derefs by working on the raw pointer at first.
old_table->set_protected_uses(
*isolate->factory()->empty_protected_weak_fixed_array());
for (uint32_t i = 0; i < old_length; ++i) {
// Note: We pass `kNewEntry` here since the offheap data was already moved
// to the new table and we do not want to update anything there.
DispatchTableClear(*old_table, i, WasmDispatchTable::kNewEntry);
}We _could_ keep all of this, but we don't need it any more if the entire object is inaccessible anyway. Efficiency++ 😊
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Nice, thanks! One nit and a question but overall looks good!
// Undo earlier publishing, making the handle inaccessible.Nit: maybe "entry" instead of "handle" to be consistent with the documentation for Publish.
old_table->set_protected_uses(
*isolate->factory()->empty_protected_weak_fixed_array());
for (uint32_t i = 0; i < old_length; ++i) {
// Note: We pass `kNewEntry` here since the offheap data was already moved
// to the new table and we do not want to update anything there.
DispatchTableClear(*old_table, i, WasmDispatchTable::kNewEntry);
}We _could_ keep all of this, but we don't need it any more if the entire object is inaccessible anyway. Efficiency++ 😊
SG! Just to double-check, do we have a regression test for this scenario already?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Auto-Submit | +1 |
| Commit-Queue | +1 |
// Undo earlier publishing, making the handle inaccessible.Nit: maybe "entry" instead of "handle" to be consistent with the documentation for Publish.
Done
old_table->set_protected_uses(
*isolate->factory()->empty_protected_weak_fixed_array());
for (uint32_t i = 0; i < old_length; ++i) {
// Note: We pass `kNewEntry` here since the offheap data was already moved
// to the new table and we do not want to update anything there.
DispatchTableClear(*old_table, i, WasmDispatchTable::kNewEntry);
}Samuel GroßWe _could_ keep all of this, but we don't need it any more if the entire object is inaccessible anyway. Efficiency++ 😊
SG! Just to double-check, do we have a regression test for this scenario already?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | +1 |
| Commit-Queue | +2 |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
[sandbox] Introduce ExposedTrustedObject::Unpublish
which can be used to make objects inaccessible that are not supposed
to be used any more.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |