Bug 6286 – Regression(2.054): Static arrays can not be assigned from const(T)[N] to T[N]

The following snipped worked up to DMD 2.053 but fails on 2.054: --- void test() { const(int)[4] src = [1, 2, 3, 4]; int[4] dst; dst = src; // Error: cannot implicitly convert expression (src) of type const(int[4u]) to int[] dst[] = src[]; // still works } --- The assignment "T[N] = const(T)[N]" should work as long as "T = const(T)" works as there is no aliasing goind on but just a plain copy.

I don't think that this is a bug but rather than dmd used to be buggy with regards to this case. dst = src; is assigning src to dst. And assigning a dynamic array to a static one isn't legal. If dst = src; were allowed, then it would be inconsistent with the case where both dst and src are dynamic arrays. dst[] = src[]; on the other hand is specifically copying the elements of src to the elements of dst. So, it's possible that this is a regression and that dst = src; is supposed to just translate to dst[] = src[]; when dst is a static array, but I think that it's far more likely that this is a case where a long-standing bug was fixed.

Note that in the test case both, the source and the destination are static arrays of the same size. There is no reason why a direct assignment should not work here. I think the compiler mistakes the left side here as a dynamic array. In the case of really assigning a dynamic array to a static one, the decision is definitely not so obvious as it could lead to a hidden runtime error.

I fear this one was my fault! https://github.com/D-Programming-Language/dmd/pull/232

Ah yes. I didn't read carefully enough. If both src and dst are static arrays, then dst = src; should work. It's when one is a dynamic array that it shouldn't.