Hello,


I was just looking at the source of std.typecons.RefCounted and it seems like it is not thread-safe at all, the count variable may be written concurrently from several threads which could cause data corruption. Is this correct or am I overlooking something (I’m still new to D). If it is correct then this should either be fixed or if it is intentional it should really be documented.


rgds, Kira

12-Jan-2014 23:59, Kira Backes пишет:
> Hello,
>
>
> I was just looking at the source of std.typecons.RefCounted and it seems
> like it is not thread-safe at all, the count variable may be written
> concurrently from several threads which could cause data corruption.

It cannot. In D everything is thread-local by default.
And  shared(RefCounted!T) would most likely fail to compile.

> Is
> this correct or am I overlooking something (I’m still new to D). If it
> is correct then this should either be fixed or if it is intentional it
> should really be documented.



-- 
Dmitry Olshansky

On 2014-01-12 20:08:17 +0000, Dmitry Olshansky <dmitry.olsh@gmail.com> said:

> 12-Jan-2014 23:59, Kira Backes пишет:
>> Hello,
>> 
>> 
>> I was just looking at the source of std.typecons.RefCounted and it seems
>> like it is not thread-safe at all, the count variable may be written
>> concurrently from several threads which could cause data corruption.
> 
> It cannot. In D everything is thread-local by default.
> And  shared(RefCounted!T) would most likely fail to compile.

There's still a race when RefCounted is located in a GC-allocated memory block, as the destructor might get called from any thread.
https://d.puremagic.com/issues/show_bug.cgi?id=4624

-- 
Michel Fortin
michel.fortin@michelf.ca
http://michelf.ca

On Sunday, 12 January 2014 at 20:08:25 UTC, Dmitry Olshansky wrote:
> It cannot. In D everything is thread-local by default.
> And shared(RefCounted!T) would most likely fail to compile.

Hello, thanks for the info, I did not test compiling with shared. You’re right about this, I was just too much used to C++11’s shared_ptr I think ;-)

13-Jan-2014 00:12, Michel Fortin пишет:
> On 2014-01-12 20:08:17 +0000, Dmitry Olshansky <dmitry.olsh@gmail.com>
> said:
>
>> 12-Jan-2014 23:59, Kira Backes пишет:
>>> Hello,
>>>
>>>
>>> I was just looking at the source of std.typecons.RefCounted and it seems
>>> like it is not thread-safe at all, the count variable may be written
>>> concurrently from several threads which could cause data corruption.
>>
>> It cannot. In D everything is thread-local by default.
>> And  shared(RefCounted!T) would most likely fail to compile.
>
> There's still a race when RefCounted is located in a GC-allocated memory
> block, as the destructor might get called from any thread.
> https://d.puremagic.com/issues/show_bug.cgi?id=4624
>
Awful.
The only thing worse is that arrays of structs don't have their destructors called at all.

-- 
Dmitry Olshansky