C++ Logo


Advanced search

Re: [wg14/wg21 liaison] C trap representations and unspecified values versus C++ indeterminate values

From: Uecker, Martin <Martin.Uecker_at_[hidden]>
Date: Sat, 10 Oct 2020 19:32:11 +0000
Am Samstag, den 10.10.2020, 17:19 +0200 schrieb Freek Wiedijk:
> Dear Martin,
> > I agree, although I do not see it as a contradiction to
> > the proposed provenance model.
> In the proposed model subobjects are problematic, because
> the "provenance" tagging is on the object level.

Yes, so far this is not addressed.

> In the
> "dual model" that I think is a better approach, subobject
> "provenance" becomes trivial, because there is no provenance
> anymore in the first place.

I do not see how you would achieve the same things
without having essentially something similar as

> The main problem (as always) is how to synthesize an abstract
> object out of an object representation consisting of bytes.

Conceptually, there is a function that maps
representation bytes to the valid values of a
type. It is used during lvalue conversion.
If the representation is not in the domain
of this function, it is a trap representation.
It seems that nowadays this mapping can not be
a fixed property of the type but may depend
on the context.


Received on 2020-10-10 14:32:19