Thread overview
[Issue 10272] New: opAssign() not invoked during variable declaration and initialization
Jun 05, 2013
Peter Williams
Jun 05, 2013
w0rp
Jun 05, 2013
Jonathan M Davis
Jun 05, 2013
Peter Williams
Jun 06, 2013
Jonathan M Davis
June 05, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10272

           Summary: opAssign() not invoked during variable declaration and
                    initialization
           Product: D
           Version: D2
          Platform: x86_64
        OS/Version: Linux
            Status: NEW
          Severity: major
          Priority: P2
         Component: DMD
        AssignedTo: nobody@puremagic.com
        ReportedBy: pwil3058@bigpond.net.au


--- Comment #0 from Peter Williams <pwil3058@bigpond.net.au> 2013-06-04 23:50:09 PDT ---
Created an attachment (id=1219)
Demonstrate the opAssign() problem for struct template

The opAssign() operator is not called when the assignment is part of a variable declaration for both the case where auto used and where the type is declared explicitly. e.g.:

TestStruct t1 = TestStruct(arg);
TestStruct t2 = t1;

the opAssign() operator is not called in either of these cases.  But:

TestStruct t3;
t3 = TestStruct(arg);
t3 = t1;

the opAssign() operator is called in both cases.

This can have serious consequences if (for instance) the purpose of the
opAssign() method being defined was to ensured that the assignee and the
assignor did not share the same internal array.

The attached code demonstrates the problem.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
June 05, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10272


w0rp <devw0rp@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |devw0rp@gmail.com


--- Comment #1 from w0rp <devw0rp@gmail.com> 2013-06-05 00:03:13 PDT ---
Isn't this exactly the same as not calling operator= in C++? I expect this behaviour, because assignment and initialisation are two separate concepts. If you have a case to handle where initialisation copies or destroys another object's memory, do it in this().

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
June 05, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10272


Jonathan M Davis <jmdavisProg@gmx.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |jmdavisProg@gmx.com
         Resolution|                            |INVALID


--- Comment #2 from Jonathan M Davis <jmdavisProg@gmx.com> 2013-06-05 00:08:35 PDT ---
This is completely expected. Variable declarations do not use the assignment operator, and

int a = 1;

does not use the assignment operator at all. It's only assignments which use the assignment operator. e.g.

a = 5;

When the variable is declared, it is initialized, not assigned to. While they may seem similar and have similar syntax, they are fundamentally different. And as w0rp points out, this behavior is exactly the same as C++'s behavior.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
June 05, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10272


Peter Williams <pwil3058@bigpond.net.au> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|INVALID                     |


--- Comment #3 from Peter Williams <pwil3058@bigpond.net.au> 2013-06-05 16:14:39 PDT ---
I should have been clearer.  I agree that it is the expected and desirable behaviour when the initializer is an rValue but disagree when it is an lValue.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
June 06, 2013
http://d.puremagic.com/issues/show_bug.cgi?id=10272


Jonathan M Davis <jmdavisProg@gmx.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
         Resolution|                            |INVALID


--- Comment #4 from Jonathan M Davis <jmdavisProg@gmx.com> 2013-06-05 17:13:09 PDT ---
Sorry, but it's still invalid.

auto foo = bar;

will _never_ use opAssign. opAssign is for overloading the assignment operator, and there is no assignment operator in that statement. It's a variable declaration, not an assignment. Whether it's an lvalue or rvalue is irrelevant for that. The difference is that with an rvalue, it's a move operation, whereas with an lvalue, the postblit operator will be called. So, if you want to overload the behavior of

auto foo = bar;

then you need to declare a postblit constructor:

http://dlang.org/struct.html#StructPostblit

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