Handle global mouse state for Emscripten

[Temdog007]

Fixes #12667

[vittorioromeo]

Thanks for giving it a shot! Unfortunately, this doesn't seem to solve the problem.

It appears to work at first, but either resizing the browser window or resizing the canvas by changing the window size via SDL_SetWindowSize seems to screw up everything again.

In contrast, with my crappy workaround, the position is always consistent even after resizing the browser or the window:

Vector2i getPosition(const WindowBase& relativeTo)
{
#ifdef SFML_SYSTEM_EMSCRIPTEN
    return getPosition(); 
#else
    return getPosition() - relativeTo.getPosition();
//                         ^~~~~~~~~~~~~~~~~~~~~~~~
//                  uses SDL_GetWindowPosition() internally
#endif
}

[vittorioromeo]

Thanks for giving it a shot! Unfortunately, this doesn't seem to solve the problem.

It appears to work at first, but either resizing the browser window or resizing the canvas by changing the window size via SDL_SetWindowSize seems to screw up everything again.

In contrast, with my crappy workaround, the position is always consistent even after resizing the browser or the window:

Vector2i getPosition(const WindowBase& relativeTo)
{
#ifdef SFML_SYSTEM_EMSCRIPTEN
    return getPosition(); 
#else
    return getPosition() - relativeTo.getPosition();
//                         ^~~~~~~~~~~~~~~~~~~~~~~~
//                  uses SDL_GetWindowPosition() internally
#endif
}

Thinking about it a bit more, the issue might now be SDL_GetWindowPosition()...

[Temdog007]

These changes will be merged into #12575. So, closing this PR.

[slouken]

Please keep them as a separate PR unless there's an actual code dependency between them.

[Temdog007]

Please keep them as a separate PR unless there's an actual code dependency between them.

A test was added to #12575 for verifying multiple windows for Emscripten. This involves spawning multiple windows. To prevent putting them in the same position, I would spawn them at the global position of the mouse. Thus, that branch needed this change. Since this was marked for 3.4.0, I figured it would be fine to merge this branch into that one to have just one PR.

[slouken]

Merged, thanks!