• via Digitalmars-d-bugs (54/54) Mar 08 2015 https://issues.dlang.org/show_bug.cgi?id=14261

via Digitalmars-d-bugs <digitalmars-d-bugs puremagic.com> writes:

https://issues.dlang.org/show_bug.cgi?id=14261

          Issue ID: 14261
           Summary: Struct destructors shouldn't be called when in a
                    closure
           Product: D
           Version: D2
          Hardware: All
                OS: All
            Status: NEW
          Severity: major
          Priority: P1
         Component: DMD
          Assignee: nobody puremagic.com
          Reporter: thecybershadow gmail.com

Illustrative example:

////////////// example.d /////////////
import std.stdio;

auto getLogger()
{
    auto f = File("log.txt", "w");
    return (string s) => f.writeln(s);
}

void main()
{
    auto logger = getLogger();
    logger("Hello, world!");
}
//////////////////////////////////////

This will crash because the file is already destroyed by the time the lambda
runs.

Test case:

///////////////// test.d ////////////////
struct S
{
    int i;
    ~this() { i--; }
}

auto outerFun()
{
    auto s = S(1);
    void innerFun() { assert(s.i == 1); }
    return &innerFun;
}

void main()
{
    auto f = outerFun();
    f();
}
/////////////////////////////////////////

Now that we have finalization of structs on the heap, maybe it's time to move
closure destruction from end of scope to GC-collection time.

I'm not sure if accessing destroyed objects can be considered a safety problem,
but if it is, all the more reason to fix.

--