I'm registering a callback with some C code.  The simplified story is here, but the actual code is on GitHub [1] at the end if you care.

The call looks something like this.

void register(void(*fp)(void*), void* context);

I have a class that holds state for the callback and registers itself:

final class Klass
{
    void method()
    {
        register(callback_function, &this);
    }
}

As of dmd 2.067, doing "&this" is deprecated.  Is there an idiomatic way to do this?

[0] Actual code is at https://github.com/todayman/dubik/blob/master/source/vibe/core/drivers/rx.d#L177 .  The msg object eventually gets passed to the registration function.

Thanks,
-- 
Paul O'Neil
Github / IRC: todayman

On Monday, 30 March 2015 at 02:53:36 UTC, Paul O'Neil wrote:

> As of dmd 2.067, doing "&this" is deprecated.

where is this documented? I don't see it in the release notes.

On 03/29/2015 10:57 PM, weaselcat wrote:
> On Monday, 30 March 2015 at 02:53:36 UTC, Paul O'Neil wrote:
> 
>> As of dmd 2.067, doing "&this" is deprecated.
> 
> where is this documented? I don't see it in the release notes.

I don't see it in the release notes either, but it's happening.  Maybe it's an instance of a more general thing?

-- 
Paul O'Neil
Github / IRC: todayman

On Monday, 30 March 2015 at 03:02:07 UTC, Paul O'Neil wrote:
> On 03/29/2015 10:57 PM, weaselcat wrote:
>> On Monday, 30 March 2015 at 02:53:36 UTC, Paul O'Neil wrote:
>> 
>>> As of dmd 2.067, doing "&this" is deprecated.
>> 
>> where is this documented? I don't see it in the release notes.
>
> I don't see it in the release notes either, but it's happening.
>  Maybe
> it's an instance of a more general thing?

if it works in 2.066 you should submit a bug report, either this shouldn't be a deprecation or it should be in the release notes.

On Monday, 30 March 2015 at 02:53:36 UTC, Paul O'Neil wrote:
> I'm registering a callback with some C code.  The simplified story is
> here, but the actual code is on GitHub [1] at the end if you care.
>
> The call looks something like this.
>
> void register(void(*fp)(void*), void* context);
>
> I have a class that holds state for the callback and registers itself:
>
> final class Klass
> {
>     void method()
>     {
>         register(callback_function, &this);
>     }
> }
>
> As of dmd 2.067, doing "&this" is deprecated.  Is there an idiomatic way
> to do this?
>
> [0] Actual code is at
> https://github.com/todayman/dubik/blob/master/source/vibe/core/drivers/rx.d#L177
> .  The msg object eventually gets passed to the registration function.
>
> Thanks,

This is only deprecated for class not struct. This code below works fine:

---
import std.stdio;

extern(C) void f2(void* ins) {
    auto s = cast(S*)(ins);
    writefln("f2():%s", s);
    writefln("f2():%s", *s);
}

void f1(void* ins) {
    auto s = cast(S*)(ins);
    writefln("f1():%s", s);
    writefln("f1():%s", *s);
}

struct S { // <<-- change to "class" to get deprecated message
    int value = 10;
    void f() {
        f1(&this);
        f2(&this);
    }

}

void main()
{
    auto s = S();
    s.f();
}
---

bye,
lobo

On Sun, 29 Mar 2015 22:53:35 -0400, Paul O'Neil wrote:

> I'm registering a callback with some C code.  The simplified story is here, but the actual code is on GitHub [1] at the end if you care.
> 
> The call looks something like this.
> 
> void register(void(*fp)(void*), void* context);
> 
> I have a class that holds state for the callback and registers itself:
> 
> final class Klass {
>     void method()
>     {
>         register(callback_function, &this);
>     }
> }
> 
> As of dmd 2.067, doing "&this" is deprecated.  Is there an idiomatic way to do this?
> 
> [0] Actual code is at https://github.com/todayman/dubik/blob/master/source/vibe/core/drivers/
rx.d#L177
> .  The msg object eventually gets passed to the registration function.
> 
> Thanks,

you still can cast `this` to void pointer: `cast(void*)this`

but beware of possible `opCast` overloads for `void*` (there are no in stdlib, but...)

On Monday, 30 March 2015 at 02:53:36 UTC, Paul O'Neil wrote:
> I'm registering a callback with some C code.  The simplified story is
> here, but the actual code is on GitHub [1] at the end if you care.
>
> The call looks something like this.
>
> void register(void(*fp)(void*), void* context);
>
> I have a class that holds state for the callback and registers itself:
>
> final class Klass
> {
>     void method()
>     {
>         register(callback_function, &this);
>     }
> }

`this` is already a reference. You're taking the address of that reference. A  simple cast should work: `cast(void*) this`.

On 3/30/15 5:12 AM, "Marc =?UTF-8?B?U2Now7x0eiI=?= <schuetzm@gmx.net>" wrote:
> On Monday, 30 March 2015 at 02:53:36 UTC, Paul O'Neil wrote:
>> I'm registering a callback with some C code.  The simplified story is
>> here, but the actual code is on GitHub [1] at the end if you care.
>>
>> The call looks something like this.
>>
>> void register(void(*fp)(void*), void* context);
>>
>> I have a class that holds state for the callback and registers itself:
>>
>> final class Klass
>> {
>>     void method()
>>     {
>>         register(callback_function, &this);
>>     }
>> }
>
> `this` is already a reference. You're taking the address of that
> reference. A  simple cast should work: `cast(void*) this`.

To build on this further, &this for a class is actually taking a local stack reference, this is why it's not allowed.

And technically, cast(void*) this is dangerous in the general case because opCast can be overridden. If you absolutely need to get a pointer to a class reference, you would need to do this:

auto x = this;
auto p = &x;

For example, for a foolproof implementation of converting a class reference to void *, you would need to do:

auto x = this;
auto p = *(cast(void **)&x);

I wonder if those who made this change thought of this problem?

-Steve

On 03/30/2015 11:32 AM, Steven Schveighoffer wrote:
> On 3/30/15 5:12 AM, "Marc =?UTF-8?B?U2Now7x0eiI=?= <schuetzm@gmx.net>" wrote:
>>
>> `this` is already a reference. You're taking the address of that reference. A  simple cast should work: `cast(void*) this`.
> 
> To build on this further, &this for a class is actually taking a local stack reference, this is why it's not allowed.
> 
> And technically, cast(void*) this is dangerous in the general case because opCast can be overridden. If you absolutely need to get a pointer to a class reference, you would need to do this:
> 
> auto x = this;
> auto p = &x;
> 
> For example, for a foolproof implementation of converting a class reference to void *, you would need to do:
> 
> auto x = this;
> auto p = *(cast(void **)&x);
> 
> I wonder if those who made this change thought of this problem?
> 
> -Steve

Thanks for the explanation.  This makes a lot of sense - I forgot that "ref" actually means "special pointer."

Luckily, I don't plan on overriding opCast!

-- 
Paul O'Neil
Github / IRC: todayman