Bug 9559 – Range of Nullable doesn't work with std.array.array

Using std.array.array on a range of Nullable!T causes a runtime error if at least one of the elements is set to null. Test case: ---------------------- import std.algorithm; import std.typecons; void main() { alias Nullable!int I; auto ints = [0, 1, 2].map!(i => i & 1 ? I.init : I(i)); auto asArray = std.array.array(ints); } ---------------------- Produces the following error: object.Exception@/usr/include/dmd/phobos/std/typecons.d(1181): Enforcement failed Full backtrace here: http://pastebin.com/wMFRKk6L The enforcement exception occurs in Nullable's get() function (I'm using 2.062, I noticed that this has been changed to an assert in HEAD) and Nullable has an alias this to get. Best guess is that this is actually a problem in emplace or array's usage of it.

I don't have a good testing environment at the moment, but it looks like the bug might be here: https://github.com/D-Programming-Language/phobos/blob/master/std/conv.d#L3581 If I understand this correctly, this static if will succeed because the Nullable!int constructor will accept another Nullable!int due to the alias this. When the constructor is called, the alias this causes the get() function to be called in order to resolve the Nullable being copied to an int. Possible fix: add a constructor to Nullable that takes typeof(this) and correctly checks isNull.

(In reply to comment #1) > I don't have a good testing environment at the moment, but it looks like the > bug might be here: > https://github.com/D-Programming-Language/phobos/blob/master/std/conv.d#L3581 > > If I understand this correctly, this static if will succeed because the > Nullable!int constructor will accept another Nullable!int due to the alias > this. When the constructor is called, the alias this causes the get() function > to be called in order to resolve the Nullable being copied to an int. > > Possible fix: add a constructor to Nullable that takes typeof(this) and > correctly checks isNull. This gets fixed by my existing fix for emplace. The pull is here: https://github.com/D-Programming-Language/phobos/pull/1082 And I just added a unittest for this specific bug: https://github.com/monarchdodra/phobos/commit/10025b835e564d17fd960d461e965e9bb80d115d

(In reply to comment #2) > (In reply to comment #1) > > I don't have a good testing environment at the moment, but it looks like the > > bug might be here: > > https://github.com/D-Programming-Language/phobos/blob/master/std/conv.d#L3581 > > > > If I understand this correctly, this static if will succeed because the > > Nullable!int constructor will accept another Nullable!int due to the alias > > this. When the constructor is called, the alias this causes the get() function > > to be called in order to resolve the Nullable being copied to an int. > > > > Possible fix: add a constructor to Nullable that takes typeof(this) and > > correctly checks isNull. > > This gets fixed by my existing fix for emplace. > > The pull is here: > https://github.com/D-Programming-Language/phobos/pull/1082 > > And I just added a unittest for this specific bug: > https://github.com/monarchdodra/phobos/commit/10025b835e564d17fd960d461e965e9bb80d115d Excellent. I'd love to see this in the next release. Unfortunately, it looks like the use of UFCS in my test case is causing the auto-tester to choke.

Commit pushed to master at https://github.com/D-Programming-Language/phobos https://github.com/D-Programming-Language/phobos/commit/e951b2cf465614e87379c4f0b501ee0f1d57a9ad Issue 9559 - Range of Nullable doesn't work with std.array.array

Fixed in Git master.