Martin Robinson
Martin Robinson
This means that bytes need to be serialized when sent across the IPC channels, which I think isn't what we want. We are also using the same `IpcSharedMemory` to send...
Thinking about this a bit more, I think understand your line of thinking here. I'll take a closer look tomorrow. I would like to attempt wrapping IpcSharedMemory in an Arc...
You might see a change if you use canvas though...
> But I feel like I still don't understand. Canvas gets spawned from a html element so it needs to live in the same ScriptThread thread. You can verify that...
Could you also verify that the bytes of the image cache are not being serialized when sent over an IPC channel?
Perhaps we can do a quick turnaround here and release the 0.11 series in order to not block the pending PRs that should be merged?
Another alternative here would be to ensure that every platform has Ahem loaded into its system font cache before running the tests. It seems like this could be done by...
> 2\. make webfonts critical subresources that block load events and remove our Ahem.css user stylesheet injection ([f363bc2](https://github.com/servo/servo/commit/f363bc2bc8f2d58c188f7de7f42ffa854f66bf12)) About this, I wonder why it is necessary? We are currently waiting...
Oh, nice detective work!
Thanks for filing this. I'm happy to make a PR for this later today. Probably we should have damage consistently update both `need_new_display_list` and `need_new_stacking_context_tree` immediately after it is calculated.