std.boxer.Box.opEqualsInternal's last line is:

return cast(bit)type.equals(data, other.data);

but when you've boxed a null Object reference, 'data' is all zeros, 'type' is a TypeInfo_Class, and TypeInfo_Class.equals is:

:     int equals(void *p1, void *p2)
:     {
:         Object o1 = *cast(Object*)p1;
:         Object o2 = *cast(Object*)p2;
:
:         return o1 == o2 || (o1 && o1.opCmp(o2) == 0);
:     }

and o1 and o2 are made into null references, causing a seg-fault.

Which one should be fixed?

<derick_eddington@nospam.yashmoo.com> wrote in message news:d7rd9q$20hk$1@digitaldaemon.com...
> std.boxer.Box.opEqualsInternal's last line is:
>
> return cast(bit)type.equals(data, other.data);
>
> but when you've boxed a null Object reference, 'data' is all zeros, 'type'
> is a
> TypeInfo_Class, and TypeInfo_Class.equals is:
>
> :     int equals(void *p1, void *p2)
> :     {
> :         Object o1 = *cast(Object*)p1;
> :         Object o2 = *cast(Object*)p2;
> :
> :         return o1 == o2 || (o1 && o1.opCmp(o2) == 0);
> :     }
>
> and o1 and o2 are made into null references, causing a seg-fault.
>
> Which one should be fixed?

I think equals should be fixed.
see http://www.digitalmars.com/d/archives/digitalmars/D/bugs/1306.html
It looks like compare was changed but equals wasn't.

'compare' also seg-faults if its 'o1' is null:

:     int compare(void *p1, void *p2)
:     {
:         Object o1 = *cast(Object*)p1;
:         Object o2 = *cast(Object*)p2;
:         int c = 0;
:
:         // Regard null references as always being "less than"
:         if (o1 != o2)  // <--- tries o1.opEquals(o2)
:         {
:             if (o1)
:             {   if (!o2)
:                     c = 1;
:                 else
:                     c = o1.opCmp(o2);
:             }
:             else
:                 c = -1;
:         }
:         return c;
:     }


In article <d7smh2$30pc$1@digitaldaemon.com>, Ben Hinkle says...
>
>
><derick_eddington@nospam.yashmoo.com> wrote in message news:d7rd9q$20hk$1@digitaldaemon.com...
>> std.boxer.Box.opEqualsInternal's last line is:
>>
>> return cast(bit)type.equals(data, other.data);
>>
>> but when you've boxed a null Object reference, 'data' is all zeros, 'type'
>> is a
>> TypeInfo_Class, and TypeInfo_Class.equals is:
>>
>> :     int equals(void *p1, void *p2)
>> :     {
>> :         Object o1 = *cast(Object*)p1;
>> :         Object o2 = *cast(Object*)p2;
>> :
>> :         return o1 == o2 || (o1 && o1.opCmp(o2) == 0);
>> :     }
>>
>> and o1 and o2 are made into null references, causing a seg-fault.
>>
>> Which one should be fixed?
>
>I think equals should be fixed.
>see http://www.digitalmars.com/d/archives/digitalmars/D/bugs/1306.html
>It looks like compare was changed but equals wasn't.
>
>

<derick_eddington@nospam.yashmoo.com> wrote in message news:d7u5eq$ufd$1@digitaldaemon.com...
> 'compare' also seg-faults if its 'o1' is null:

Which version of dmd are you using? In dmd.125 compare has been fixed to
have
if (!(o1 is o2))

TypeInfo_C in std/typeinfo/ti_C.d has the fixed 'compare' but TypeInfo_Class in internal/object.d does not.

In article <d7urnf$1dtc$1@digitaldaemon.com>, Ben Hinkle says...
>
><derick_eddington@nospam.yashmoo.com> wrote in message news:d7u5eq$ufd$1@digitaldaemon.com...
>> 'compare' also seg-faults if its 'o1' is null:
>
>Which version of dmd are you using? In dmd.125 compare has been fixed to
>have
>if (!(o1 is o2))
>
>