Do not attempt to grab pointer on web if unsupported.

[rectalogic]

Objective

Fixes #21497

Solution

Mobile web does not support pointer APIs and panics if they are used. Check if they exist before using them.

Testing

Tested on iOS Safari using https://github.com/rectalogic/pointerlock

[alice-i-cecile]

This was previously working, and broke in #19668.

[janhohenheim]

I personally want this, but when I argued for better locking support via calling web_sys stuff, @mockersf answer was to instead upstream those things into winit, which is fair.
I tried, but got ghosted by the winit folks. So idk if that "ruling" is still accurate.
But yeah if an SME-Input was against this sorta thing in the past, I cannot in good conscience approve it

[alice-i-cecile]

@janhohenheim can you link the winit discussion here for us?

[janhohenheim]

• https://discord.com/channels/691052431525675048/768253008416342076/1370133196053676144
• https://discord.com/channels/691052431525675048/768253008416342076/1383919145736868000
• Add event for CursorGrabMode changes rust-windowing/winit#4222
• https://matrix.to/#/!DGpLzJTRzBDTwZiogk:matrix.org/$Km7lgNAmxyOtmyR8YCI7InGiq1g8Wqkz0phkCxMGKno?via=matrix.org&via=bwinf.de&via=synod.im

That said, given that this is less about "does winit send correct events" and more about "does winit not crash", this seems more pressing to merge here downstream.

[alice-i-cecile]

That said, given that this is less about "does winit send correct events" and more about "does winit not crash", this seems more pressing to merge here downstream.

I'm strongly in favor of this. I think that this is an elegant way to fix this downstream for now, and I don't think we should block this on an upstream fix.

Upstream fixes would be great, but we can migrate this to that later.

[janhohenheim]

Fully agreed :)

[janhohenheim]

Ah, code looks correct to me!