https://issues.dlang.org/show_bug.cgi?id=24570
--- Comment #4 from Steven Schveighoffer <[email protected]> --- (In reply to Salih Dincer from comment #2) > If you do the following instead of `auto save() => return this;`, the > problem is solved: > > ``` > struct R > { > wchar* ptr; > size_t len; > > this(T)(T[] range) > { > ptr = cast(wchar*)range.ptr; > len = range.length; > } > > auto empty() => len == 0; > auto front() => *ptr++; this is an invalid implementation. Using `front` more than once is supported. > auto popFront() => len--; > auto save() > { > auto r = R([]); > r.len = len; > r.ptr = ptr; > return r; > } > } > > void main() > { > auto c = ['€', '₺', '₽']; > auto r = R(c); > auto arr = [r, r, r]; > > assert(!arr.empty); > > import std.conv : text; > auto str = arr.text; // "€₺₽" > > assert(!arr.empty); But it was never arr that was empty. It's just an array of empty ranges at this point. But the difference here is that you are using wchar and not short. format treats character data differently (as evidenced by the printout, it's not an array of items, but a string). Yet another reason why this bug has gone undetected for so long. --
