Cherry-picks for the 3.2 branch (future 3.2.2) - 6th batch

[akien-mga]

No description provided.

[madmiraal]

I think the C4724 warning is due to these, because edges.size() may be 0:

godot/scene/3d/navigation.cpp

Line 255 in 8e91d27

Vector3 b = _get_vertex(from_poly->edges[(pe + 1) % from_poly->edges.size()].point);

godot/scene/3d/navigation.cpp

Line 338 in 8e91d27

_get_vertex(begin_poly->edges[(i + 1) % begin_poly->edges.size()].point)

godot/scene/3d/navigation.cpp

Line 397 in 8e91d27

_get_vertex(p->edges[(i + 1) % p->edges.size()].point)

godot/scene/3d/navigation.cpp

Line 460 in 8e91d27

int prev_n = (p->prev_edge + 1) % p->edges.size();

godot/scene/3d/navigation.cpp

Line 532 in 8e91d27

int prev_n = (p->prev_edge + 1) % p->edges.size();

godot/scene/3d/navigation.cpp

Line 589 in 8e91d27

Geometry::get_closest_points_between_segments(p_from, p_to, _get_vertex(p.edges[i].point), _get_vertex(p.edges[(i + 1) % p.edges.size()].point), a, b);

An alternative is to use:
(i+1) == edges.size() ? 0 : i+1

[akien-mga]

@madmiraal But I'm preventing those to run with ERR_FAIL checks in f390749 already, so it should not be possible for those mods to be called with 0. I guess MSVC is not clever enough to see that the branch has an early return in our macros?

Note that I only guarded 4 of them as the last 2 are already guarded by the bound of their respective for loops: the loop shouldn't be entered at all if the size is 0.

[akien-mga]

I changed the AppVeyor commit a520e03 to use werror=no for now. We should be able to see all remaining warnings in the CI log and see how to solve them. Apparently f390749 doesn't suffice to fix C4724.

[madmiraal]

It's definitely going to take more work. I've managed to extend f390749 to fix all the warnings in scene\3d\navigation.cpp, but now I get the same warnings in scene\2d\collision_polygon_2d.cpp:
https://ci.appveyor.com/project/madmiraal/godot/builds/33002769

Something else has changed to cause these warnings, but I can't work out what it is. The only thing that has changed in cowdata.h is 94f451e, but I don't see how that causes it, especially if it's not happening in master.