example:
-----------

struct A{
	int* arr;
	
	~this() {
		writeln("A destruct");
	}
}

static ~this() {
	writeln("module destruct");
}

A a;

void main() {
}

-----------

only prints "module destruct"

On Tuesday, 25 December 2018 at 19:32:47 UTC, Dru wrote:
> example:
> -----------
>
> struct A{
> 	int* arr;
> 	
> 	~this() {
> 		writeln("A destruct");
> 	}
> }
>
> static ~this() {
> 	writeln("module destruct");
> }
>
> A a;
>
> void main() {
> }
>
> -----------
>
> only prints "module destruct"

`a`, since declared in the global scope, is constructed at compile-time so its data are in executable dedicated segments (althoufh here it's just equal to A.init) and don't have to be destructed. Add `a = A(null); ` in main and the destructor gets called.

On 12/25/18 3:46 PM, Basilez B. wrote:
> On Tuesday, 25 December 2018 at 19:32:47 UTC, Dru wrote:
>> example:
>> -----------
>>
>> struct A{
>>     int* arr;
>>
>>     ~this() {
>>         writeln("A destruct");
>>     }
>> }
>>
>> static ~this() {
>>     writeln("module destruct");
>> }
>>
>> A a;
>>
>> void main() {
>> }
>>
>> -----------
>>
>> only prints "module destruct"
> 
> `a`, since declared in the global scope, is constructed at compile-time so its data are in executable dedicated segments (althoufh here it's just equal to A.init) and don't have to be destructed. Add `a = A(null); ` in main and the destructor gets called.

That's the destructor of the temporary, not the `a`. Add a writeln("end of main") at the end of main and you will see the destructor is called before main exits (while `a` is still alive).

My expectation is that thread-local-storage variables are not destroyed. It's definitely an oversight, as things like refcounting would be screwed up if a reference is stored there.

However, I think this will take some help from the compiler, and possibly we would have to put those destructors in a constructed static destructor, since there is no "type" that holds those variables.

This seems like something for which a bug likely already exists, but feel free to file one if you don't find it.

-Steve

yeah it was reported
https://issues.dlang.org/show_bug.cgi?id=14650

Status: 	ASSIGNED
Importance: 	P1 normal
Assignee: 	Seb

On 12/26/18 1:45 PM, Dru wrote:
> yeah it was reported
> https://issues.dlang.org/show_bug.cgi?id=14650
> 
> Status:     ASSIGNED
> Importance:     P1 normal
> Assignee:     Seb

Reassigned to Razvan.