Repository navigation
Coercion to pointer to vector segfaults at runtime rather than catching the alignment issue at comptime #16997
Description
Activity
The array to vector coercion was removed in #25186 and the other two cases now require
@alignCast:a.zig:35:35: error: expected type '*[1]@Vector(16, f32)', found '*[16]f32' var y: *[1]@Vector(16, f32) = &x; ^~ a.zig:35:35: note: pointer type child '[16]f32' cannot cast into pointer type child '[1]@Vector(16, f32)' a.zig:35:35: note: array of length 16 cannot cast into an array of length 1 a.zig:44:35: error: @ptrCast increases pointer alignment var y: *[1]@Vector(16, f32) = @ptrCast(&x); ^~~~~~~~~~~~ a.zig:44:44: note: '*[16]f32' has alignment '4' var y: *[1]@Vector(16, f32) = @ptrCast(&x); ^~ a.zig:44:35: note: '*[1]@Vector(16, f32)' has alignment '32' a.zig:44:35: note: use @alignCast to assert pointer alignment a.zig:50:30: error: expected type '*align(64) u128', found '*u128' var y: *align(64) u128 = &x; ^~ a.zig:50:30: note: pointer alignment '16' cannot cast into pointer alignment '64'
Zig Version
0.12.0-dev.86+197d9a9eb
Steps to Reproduce and Observed Behavior
It looks like coercion involving pointer-of-vector types doesn't have the correct alignment in its calculations. Zig normally catches potentially unsafe coercion at comptime, so it's weird that this sort of code could segfault.
Expected Behavior
All three of those tests should fail to compile because there's a potentially unsafe coercion from a less aligned type to a more aligned type.