Reduce errors in src_c/surface.c when compiling with SDL3

[gresm]

Not a full port.

Things left to do:

• Logic for palettes.
• RLEACCEL
• alphablit
• surface_fill
• simd_blitters_avx2
• simd_blitters_sse2
• simd_surface_fill_avx2
• simd_surface_fill_sse2

Unchecked points will be done in a different PR / different PRs.

[ankith26]

I think this approach is fine for now, but eventually we may have to look into how this can be improved. For now we have no idea how cheap or how expensive these functions are, and they are called in tight loops in performance critical sections.
Many of those places probably need to be refactored to call this method once outside any tight loops and then use that result wherever possible.

[gresm]

Definitely something to remember in the future ports, but in the src_c/surface.c it seems like only surf_get_bounding_rect falls in this category.

[Starbuck5]

Hello.

Looking through this I see some parts that won't work for us. SDL_GetPixelFormatDetails and SDL_GetSurfaceClipRect need to be error checked, for one thing. Also when mapping to or from the same surface over and over again getting the format every time is going to be a large hit to performance, which doesn't come up a bunch in surface.c but does apply at least to surf_get_bounding_rect.

[gresm]

SDL_GetPixelFormatDetails and SDL_GetSurfaceClipRect need to be error checked, for one thing.

Oh right... SDL_GetSurfaceClipRect errors only when invalid surface is passes, so I assumed that it would be checked, but in many cases it doesn't seem like it. And for SDL_GetPixelFormatDetails, didn't notice that.

which doesn't come up a bunch in surface.c but does apply at least to surf_get_bounding_rect.

You're right, but currently seems like it's only for surf_get_bounding_rect.