[Proposal] Weak reference implementation for D

[Denis Shelomovskij]

14.10.2013 13:04, robert пишет:
>
> Why would I be angry with a stranger who insults me in public? I
> don't understand your concerns.

No insults assumed! Just ugly truth about all of us. )

> If you are more
> experienced in this area I am glad if you share your insights and

Walter and Andrei often do silly mistakes. Can we suppose we are more 
experienced than they? Of course no, they just don't have enough time to 
think and check. Here the situation can be the same. I probably just 
have enough time to investigate the problem and solve it.

> if you think something is wrong, I am glad to discuss it and fix
> it if you are right, but just saying your implementation is
> wrong, does not really help. It implies that you are obviously
> right and everyone who does not see this is obviously a moron, if
> someone has a bit of a problem with his ego, I don't think it is
> me.

Easy, man. I have never met morons here, except, probably, myself.
Concurrent programming is fun so I just don't want to spoil the pleasure 
of investigation. And yes, I'm also lazy. )

Now about your code. First, I was completely incorrect about it, sorry 
for that. My mistake. I didn't even think the code containing such loop 
can be "so much correct". But, and this is the second, the code can't be 
"more or less" correct. It is either correct or not correct. It 
remembers me our (Russian) recent Moscow mayoral elections when we tried 
to change something in our country (we failed, yes) and after government 
won officials said: "It was the most honest elections of all preceding." )

So you code is incorrect and lets show it. When you give your code for 
eating to the compiler, it can does whatever it want but guarantee your 
program will work as you have written it (except special cases like copy 
construction elimination) and it doesn't assume every variable can be 
accessed from other threads. E.g. here is you code with unwinded loop in 
SSA (static single assignment) form:
```
auto tmp1 = atomicLoad(_obj);
void* o1 = tmp1.address;
if(o1 is null) return null;
void* o2 = GC.addrOf(tmp1.address);

auto tmp2 = atomicLoad(_obj);
void* o3 = tmp2.address;
if(o3 is null) return null;
void* o4 = GC.addrOf(tmp2.address);

if(o4) return cast(Object) o4;
return null;
```

`o1` is only used once and `o2` is never used so the compiler is free to 
discard the former and ignore the latter. So your code equals to this code:
```
auto tmp1 = atomicLoad(_obj);
if(tmp1.address is null) return null;
GC.addrOf(tmp1.address);

auto tmp2 = atomicLoad(_obj);
if(tmp2.address is null) return null;
void* o4 = GC.addrOf(tmp2.address);

if(o4) return cast(Object) o4;
return null;
```

So the second iteration gives nothing except decreasing (squaring to be 
precise) a variety of undesired threads execution order.

-- 
Денис В. Шеломовский
Denis V. Shelomovskij