[Issue 9438] New: Strange RefCounted stack overflow

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=9438

           Summary: Strange RefCounted stack overflow
           Product: D
           Version: D2
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: major
          Priority: P2
         Component: DMD
        AssignedTo: nobody at puremagic.com
        ReportedBy: monarchdodra at gmail.com


--- Comment #0 from monarchdodra at gmail.com 2013-02-01 07:26:12 PST ---
I've been chasing this on and off for a couple months now. Basically, trying to
access the RefCounted.refCountedStore.isInitialized of a non-initialized
RefCounted in a field of a temporary will create a stack overflow. I know
that's not clear, but here is the reduced usecase:

//----
import std.container, std.stdio, std.typecons, std.exception;

struct S
{
  RefCounted!int _data;

  this(int)
  {_data.refCountedStore.ensureInitialized();}

  int get() @property
  {
      writeln("here");
      enforce(_data.refCountedStore.isInitialized); //OH NOES!!!
      writeln("there");
      return _data.refCountedPayload;
  }
}

void main()
{
    // 1)
    writeln(S(1).get);

    // 2)
    S s;
    writeln(s.get).collectException();

    // 3)
    writeln(S().get);
}
//----

1) This will create a temporary S, and intialize the ref counted. the writeln
works.

2) The creates a non-temporary S. Trying to access the ref counted will
(correctly) throw an exception.

3) This will stack overflow at the "//OH NOES!!!" line: It will first call:

ref inout(RefCountedStore) refCountedStore() inout

To get the store, and then will recursively call "isInitialized" until the
program stack overflows. I have no idea why:

//----
        @property nothrow @safe
        bool isInitialized() const
        {
            return _store !is null;
        }
//----

This seems to me like the tip of a more serious bug somewhere. I would be very
pleased if someone with more knowledge than me could try to look into it?

I think it might also create problems with things such as arrays of arrays:
Every time I've tried to fix http://d.puremagic.com/issues/show_bug.cgi?id=6153
I've had crashes (NOT asserts/enforeces), and I think this might be the reason.

Originally found with this code:
//----
void main()
{
    writeln(Array!int()[0]);
}
//----

Yes, the code is wrong, but it should *assert*. Currently, it just dies.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------