C++ Logo

STD-PROPOSALS

Advanced search

Subject: Re: [std-proposals] Allowing access to object representations
From: language.lawyer_at_[hidden]
Date: 2019-08-19 14:44:55


1. You propose to add into [intro.object]/1:
> The object representation of an object of object o is guaranteed to appear in the object representation of o.

Looks like you some words.

2. Consider this:

        struct S { int i; } s;
        auto p = std::launder(static_cast<unsigned char*>(static_cast<void*>(&s)));
        // reinterpret_cast was intentionally not used!

In your proposal, the object representation of `s.i` lays ontop of the object representation of `s` and the `unsigned char` element of both of them are within their lifetime.
Pointer to which of two `unsigned char` objects the invocation of `std::launder` should produce?

On 19/08/2019 03:52, Krystian Stasiowski via Std-Proposals wrote:
> Ok, I've gone through the issues that have brought up, here's a link to the
> draft for R1 (unfortunately, I can only provide the raw text for it, on
> vacation right now so I am unable to make the pdf):
>
> In short, I changed [intro.object] p9 to specify that elements of an object
> representation can be overlap (the nested within relationship, in my
> opinion doesn't make sense, since the object representation "overlays" the
> object, and several elements of object representations can overlay each
> other), changed [expr.add] p6 to allow for pointer addition and subtraction
> for pointers of type char, unsigned char, and std::byte (unsigned char
> included since char and byte are their own object representations) pointing
> to an element of an object representation, extended the nested within
> guarantee to objects of nonzero size, and extended to definition of
> contiguous layout objects to multi-dimensional arrays (since array elements
> are contiguously allocated http://eel.is/c++draft/dcl.array#6)
>
>
> On Sun, Aug 18, 2019 at 9:41 AM sdkrystian <sdkrystian_at_[hidden]> wrote:
>
>> Although, it would still be contiguous; http://eel.is/c++draft/dcl.array#6
>>
>> Perhaps extending the definition to cover multidimensional arrays would be
>> worth doing.
>>
>>
>>
>> Sent from my Samsung Galaxy smartphone.
>>
>> -------- Original message --------
>> From: Language Lawyer via Std-Proposals <std-proposals_at_[hidden]>
>> Date: 8/18/19 08:34 (GMT-05:00)
>> To: std-proposals_at_[hidden]
>> Cc: Language Lawyer <language.lawyer_at_[hidden]>
>> Subject: Re: [std-proposals] Allowing access to object representations
>>
>> And multidimensional array of, say, `unsigned char` is its own object
>> representation?
>> If so, this does not match the expectation that the object representation
>> is contiguous.
>>
>> On 18/08/2019 03:33, sdkrystian via Std-Proposals wrote:
>>> Yes, in which case it is its own object representation
>>>
>>>
>>>
>>> Sent from my Samsung Galaxy smartphone.
>>>
>>> -------- Original message --------
>>> From: Language Lawyer via Std-Proposals <std-proposals_at_[hidden]>
>>> Date: 8/17/19 20:30 (GMT-05:00)
>>> To: std-proposals_at_[hidden]
>>> Cc: Language Lawyer <language.lawyer_at_[hidden]>
>>> Subject: Re: [std-proposals] Allowing access to object representations
>>>
>>> Ah, so the object representation of an object of type T is a sequence of
>> `unsigned char` objects unless T is (cv-qualified) `char` or `std::byte`
>> (or array thereof)?
>>>
>>> On 18/08/2019 03:24, sdkrystian via Std-Proposals wrote:
>>> > Same thing applies. I'll have to change [expr.add]/6 to make it work.
>>> >
>>> >
>>> >
>>> > Sent from my Samsung Galaxy smartphone.
>>> >
>>> > -------- Original message --------
>>> > From: Language Lawyer via Std-Proposals <
>> std-proposals_at_[hidden]>
>>> > Date: 8/17/19 20:21 (GMT-05:00)
>>> > To: std-proposals_at_[hidden]
>>> > Cc: Language Lawyer <language.lawyer_at_[hidden]>
>>> > Subject: Re: [std-proposals] Allowing access to object representations
>>> >
>>> > And for an object `a` of type `char`?
>>> >
>>> > On 18/08/2019 03:18, sdkrystian via Std-Proposals wrote:
>>> > > The object representation of the object is itself... For an object
>> `a` of type unsigned char, it's object representation is `a`
>>> > >
>>> > >
>>> > >
>>> > > Sent from my Samsung Galaxy smartphone.
>>> > >
>>> > > -------- Original message --------
>>> > > From: Language Lawyer via Std-Proposals <
>> std-proposals_at_[hidden]>
>>> > > Date: 8/17/19 16:49 (GMT-05:00)
>>> > > To: std-proposals_at_[hidden]
>>> > > Cc: Language Lawyer <language.lawyer_at_[hidden]>
>>> > > Subject: Re: [std-proposals] Allowing access to object
>> representations
>>> > >
>>> > > > The object representation of an object of type unsigned char,
>> char, std::byte, or an array of such types (ignoring cv-qualification), is
>> itself.
>>> > >
>>> > > What does this cryptic "is itself" mean?
>>> > > --
>>> > > Std-Proposals mailing list
>>> > > Std-Proposals_at_[hidden]
>>> > > https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>> > >
>>> > --
>>> > Std-Proposals mailing list
>>> > Std-Proposals_at_[hidden]
>>> > https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>> >
>>> --
>>> Std-Proposals mailing list
>>> Std-Proposals_at_[hidden]
>>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>>
>> --
>> Std-Proposals mailing list
>> Std-Proposals_at_[hidden]
>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>
>
>


STD-PROPOSALS list run by herb.sutter at gmail.com

Standard Proposals Archives on Google Groups