http://d.puremagic.com/issues/show_bug.cgi?id=2196

           Summary: Link errors with an abstract class's method implementing
                    an interface
           Product: D
           Version: 1.031
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Keywords: link-failure, spec
          Severity: minor
          Priority: P3
         Component: DMD
        AssignedTo: bugzilla@digitalmars.com
        ReportedBy: diggory.hardy@gmail.com


The following code generates an error at link time:

interface A
{
    void foo ();
}

abstract class B : A
{
    void foo ();
}

void main() {}


The error:
abstract.o: In function `_TMP0':
abstract.d:(.text+0x4): undefined reference to `_D8abstract1B3fooMFZv'
collect2: ld returned 1 exit status


Marking foo in B as abstract avoids the problem. The documentation isn't at all clear on whether this should be the case.

Tried with dmd 1.031 and 1.027, and gcc (Debian 4.3.1-2) 4.3.1.


-- 

http://d.puremagic.com/issues/show_bug.cgi?id=2196





------- Comment #1 from 2korden@gmail.com  2008-07-06 05:09 -------
I'm not sure if it is a bug, this is a correct behaviour to me.

In this code:
abstract class B : A {
    void foo ();
}

you define a method foo() and should provide a function body somewhere. Merely marking class abstract says that you can't instantiate it, but you can any other class that subclasses from it, unless it is abstract, too. If you mark you method as abstract it means that there is no body, child class should provide an implementation.

Interface methods are implicitly abstract.


-- 

http://d.puremagic.com/issues/show_bug.cgi?id=2196


gide@nwawudu.com changed:

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




------- Comment #2 from gide@nwawudu.com  2009-04-23 05:12 -------
(In reply to comment #1)
> I'm not sure if it is a bug, this is a correct behaviour to me.
> If you mark your method as abstract it means that there is no body,
> child class should provide an implementation.

The following compiles correctly, the original code should not link.

interface A {
    void foo ();
}

abstract class B : A
{
    abstract void foo ();
}

void main() {}


-- 

http://d.puremagic.com/issues/show_bug.cgi?id=2196


diggory.hardy@gmail.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED




------- Comment #3 from diggory.hardy@gmail.com  2009-04-23 08:22 -------
Agreed.

I think what I was unsure on is why the linker wanted the definitions of an abstract class that wasn't (attempted to be) used. But of course having to explicitly mark such functions as abstract does make it clear whether or not the function was intended to be defined in that class.


-- 

http://d.puremagic.com/issues/show_bug.cgi?id=2196


Andrei Alexandrescu <andrei@erdani.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|CLOSED                      |RESOLVED
                 CC|                            |andrei@erdani.com


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