Update wgpu to upstream e084b66aa (2026-10-18)
Categories
(Core :: Graphics: WebGPU, task, P1)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox160 | --- | fixed |
People
(Reporter: jimb, Assigned: jimb)
References
(Blocks 1 open bug)
Details
Attachments
(1 external file)
What it says on the tin.
| Assignee | ||
Updated•1 month ago
|
| Assignee | ||
Updated•1 month ago
|
Updated•29 days ago
|
| Assignee | ||
Comment 1•29 days ago
|
||
Import a new version of wgpu_core, naga and related crates, and
adapt gfx/wgpu_bindings and dom/webgpu/ipc to API changes.
In the newly vendored version of the wgpu crates,
wgpu_core_remote::Global::buffer_get_mapped_range returns a
wgpu_core::resource::BufferMapping value that ensures the pointer to the
mapped memory remains valid until it is dropped. In the C++ bindings, introduce
a new MappedRangeGuard type for use by WebGPUParent that owns the mapped
range and releases the BufferMapping when it is dropped. Change all callers of
wgpu_server_buffer_get_mapped_range to use MappedRangeGuard instead.
Since MappedRangeGuard causes buffer unmapping and destruction to block, take
care not to hold a MappedRangeGuard across any such operation, or any call
that might perform one, to avoid deadlock.
| Assignee | ||
Comment 2•29 days ago
|
||
| Assignee | ||
Comment 3•29 days ago
|
||
Added missing explicit keyword on MappedRangeGuard constructor.
try push
| Assignee | ||
Comment 4•28 days ago
•
|
||
It seems like we have new instances of bug 1840858: mac debug now behaves like mac release.
| Assignee | ||
Comment 6•28 days ago
|
||
queued for landing
Comment 8•28 days ago
|
||
| bugherder | ||
Updated•21 days ago
|
Description
•