Re: [libstdc++ PATCH] libstdc++ debug mode (fourth try)

[Doug Gregor <dgregor at apple dot com>]

On Wednesday, August 6, 2003, at 6:58AM, Pétur Runólfsson wrote:
> Hi,
>
> Just a small nit about the documentation:
>
> Index: docs/html/debug.html
> [...]
> ! <p>libstdc++ includes many extensions to the C++ standard library. In
> !   some cases the extensions are obvious, such as the hashed
> !   associative containers, whereas other extensions give predictable
> !   results to behavior that would otherwise be undefined, such as
> !   throwing an exception when a <code>std::basic_string</code> is
> !   constructed from a NULL character pointer. This latter category 
> also
> !   includes implementation-defined and unspecified semantics,
>
> The last sentence seems somewhat misleading. Implementation-defined
> behaviour (as defined in 1.3.5) is quite distinct from undefined
> behaviour, and is (or should be) predictable (since it is required to 
> be
> documented). Unspecified behaviour is also different from undefined
> behaviour, and is not an extension.

But in all cases you are relying upon behavior that the C++ standard 
does not guarantee, so your code may not be portable to other 
implementations of the standard library.

> Also, I don't see any place where the debug mode warns about
> implementation-defined or unspecified behaviour (but I may be missing
> something).

There aren't many places where this occurs, but vector capacity is one 
of them: in pedantic mode, the vector tracks the capacity that is 
guaranteed by the standard, e.g., the greatest size the container has 
attained or the largest value that has been passed to reserve(). 
Strings provide examples of libstdc++ extensions that we'll complain 
about, because construction with a NULL string is undefined behavior 
but libstdc++ throws an exception.

	Doug