Re: Large file iostreams

[Martin Sebor <sebor at roguewave dot com>]

Pétur Runólfsson wrote:

> I wrote:
>
> > I came up with this list last night (this includes already reported
> > defects):
>
>
> While trying to come up with a conforming streamoff and fpos<> I
> found some more:
>
> 27.4.3.1 p2: The initial value of state() is not specified.

That's not necessarily a defect, is it? The default ctor is
unspecified, so if stateT is a POD, the initial value of a
default-constructed fpos object is also unspecified. There
isn't much one can do with such an object, anyway (just
like there isn't much one can do with any other
uninitialized POD).

> 27.4.3.2 p1: The sentence "Operations specified in Table 88 are permitted."
> is ambiguous. It's not clear if this means "Users are permitted to use
> the operations in table 88" or "Implementations are permitted to provide
> the operations in table 88".

It could probably be worded better but the intent is clear from
the note below the table:

    [Note: Every implementation is required to supply overloaded
    operators on fpos objects to satisfy the requirements of 27.4.3.2.
    It is unspecified whether these operators are members of fpos,
    global operators, or provided in some other way. --end note]

> 27.4.3.2: The operations q = p + o, p += o, q = p - o, p -= o, o = p - q
> are specified on objects of type fpos and streamoff. Objects of type
> const fpos and const streamoff should also be allowed on the right hand
> side.

What about volatile? Or const volatile? ;-)
> 27.4.3.2: The return value of state() on the objects returned by q = p + o,
> p += o, q = p - o and p -= o is not specified.

But what should the value be? It's perfectly fine to increment
a position into a stream operating on a subsequence encoded in
a fixed-width encoding but the effects are undefined otherwise.
Since the effects don't depend on the value of fpos but on the
subsequence I don't see how they could be specified here.

Martin