Bug 1768 – CTFE: cant ~= an array literal to an unitinitialized array

The bug is illustrated in case (c). Case (d) shows a workaround: explicitly initialize the array with []. --------- char [] foo() { char [] a; a ~="abc"; // ok char [][] b; b ~= "abc"; // ok char [][][] c; c ~= ["abc", "def"]; // Error: cannot eval at compile time char [][][] d = []; d ~= ["abc", "def"]; // ok return "abc"; } const xx = foo();

*** Bug 2009 has been marked as a duplicate of this bug. ***

Works in dmd 1.030 and 2.014 (after replaceing all char[] with string).

Please recheck. Both compilers fail for me. Other people was having the issue too. dmd 1.030: char [] foo() { char [][][] c; c ~= ["abc", "def"]; // Error: cannot eval at compile time return null; } const xx = foo(); dmd 2.014: char [] foo() { string[][] c; c ~= ["abc", "def"]; // Error: cannot eval at compile time return null; } const xx = foo();

While you are at it, please ensure that this works too: (For dmd 1.030) char[][] foo(char[][] a = null) { a ~= ["oops"]; return a; } static assert(foo()[0] == "oops");

Cannot reproduce on 1.026 or 2.019. Is anyone else still seeing this problem?

*** Bug 2009 has been marked as a duplicate of this bug. ***

These seem to have been fixed except for the one with null as default argument but it's a different bug. And there are still bugs related to uninitialized arrays in structs.

For what it's worth, I can also confirm that this is fixed.