FAQ issue: Guaranteed value ranges of fundamental types?

[Alf P. Steinbach]

The C++ FAQ item 29.5 (this seems to be strongly related to C), at
<url: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.5>
mentions that

<quote>
C++ guarantees a char is exactly one byte which is at least 8 bits, short
is at least 16 bits, int is at least 16 bits, and long is at least 32
bits.
</quote>


Questions:

(1) This guarantee seems to come from the C standard. Which I don't
have. Does the C++ standard really guarantee this?

(2) Is this guarantee originally formulated in terms of number of bits,
or in terms of e.g. decimal value ranges?

(3) Concerning (2), if formulated in terms of number of bits, are the number
of bits mentioned simply sizeof(T)*CHAR_BIT, which doesn't say much about
value ranges, or are they stated to be the value representation bits?


(Intentionally cross-posted [comp.lang.c++] and [comp.lang.c]).

 

[pete]

The C++ FAQ item 29.5 (this seems to be strongly related to C), at
<url: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.5>
mentions that

<quote>
C++ guarantees a char is exactly
one byte which is at least 8 bits, short
is at least 16 bits, int is at least 16 bits,
and long is at least 32 bits.
</quote>

Questions:

(1) This guarantee seems to come from the C standard. Which I don't
have. Does the C++ standard really guarantee this?

(2) Is this guarantee originally
formulated in terms of number of bits,
or in terms of e.g. decimal value ranges?

The part of the C standard which is called
"Sizes of integral types" in C89 and
"Sizes of integer types" in C99,
is specified in terms of ranges.

C99 describes padding bits for integer types.

 

[CBFalconer]

The C++ FAQ item 29.5 (this seems to be strongly related to C), at
<url: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.5>
mentions that

<quote>
C++ guarantees a char is exactly one byte which is at least 8
bits, short is at least 16 bits, int is at least 16 bits, and
long is at least 32 bits.
</quote>

Questions:

(1) This guarantee seems to come from the C standard. Which I
don't have. Does the C++ standard really guarantee this?

I dunno. This is c.l.c. Google for N869 to get the last draft of
the C standard. f'ups set, which you should have done in the
original posting.

(2) Is this guarantee originally formulated in terms of number of
bits, or in terms of e.g. decimal value ranges?

By values. Other specifications translate that into viable bits.

(3) Concerning (2), if formulated in terms of number of bits, are
the number of bits mentioned simply sizeof(T)*CHAR_BIT, which
doesn't say much about value ranges, or are they stated to be
the value representation bits?

No, because there may be bits for other purposes, such as trap
values.

 

[Alf P. Steinbach]

* CBFalconer:

I dunno. This is c.l.c. Google for N869 to get the last draft of
the C standard. f'ups set, which you should have done in the
original posting.

It was intentionally cross-posted; F.U.T. overridden.

Thanks for the "N869" reference.

Yes, the C standard specifies those minimum ranges in the documentation
of [limits.h], which the C++ standard refers to the C standard for.

 

[Ioannis Vranos]

The C++ FAQ item 29.5 (this seems to be strongly related to C), at
<url: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.5>
mentions that

<quote>
C++ guarantees a char is exactly one byte which is at least 8 bits, short
is at least 16 bits, int is at least 16 bits, and long is at least 32
bits.
</quote>


Questions:

(1) This guarantee seems to come from the C standard. Which I don't
have. Does the C++ standard really guarantee this?

Yes. Also, except where otherwise is stated, C90 is a subset of C++98
standard.

(2) Is this guarantee originally formulated in terms of number of bits,
or in terms of e.g. decimal value ranges?

Decimal ranges. C90 defines the following (from the last C90 draft):


"Sizes of integral types <limits.h>

The values given below shall be replaced by constant expressions
suitable for use in #if preprocessing directives. Their
implementation-defined values shall be equal or greater in magnitude
(absolute value) to those shown, with the same sign.

* maximum number of bits for smallest object that is not a bit-field
(byte)
CHAR_BIT 8

* minimum value for an object of type signed char
SCHAR_MIN -127

* maximum value for an object of type signed char
SCHAR_MAX +127

* maximum value for an object of type unsigned char
UCHAR_MAX 255

* minimum value for an object of type char
CHAR_MIN see below

* maximum value for an object of type char
CHAR_MAX see below

* maximum number of bytes in a multibyte character, for any
supported locale
MB_LEN_MAX 1

* minimum value for an object of type short int
SHRT_MIN -32767

* maximum value for an object of type short int
SHRT_MAX +32767

* maximum value for an object of type unsigned short int
USHRT_MAX 65535

* minimum value for an object of type int
INT_MIN -32767

* maximum value for an object of type int
INT_MAX +32767

* maximum value for an object of type unsigned int
UINT_MAX 65535

* minimum value for an object of type long int
LONG_MIN -2147483647

* maximum value for an object of type long int
LONG_MAX +2147483647

* maximum value for an object of type unsigned long int
ULONG_MAX 4294967295


If the value of an object of type char sign-extends when used in
an expression, the value of CHAR_MIN shall be the same as that of
SCHAR_MIN and the value of CHAR_MAX shall be the same as that of
SCHAR_MAX . If the value of an object of type char does not sign-extend
when used in an expression, the value of CHAR_MIN shall be 0 and the
value of CHAR_MAX shall be the same as that of UCHAR_MAX./7/"

(3) Concerning (2), if formulated in terms of number of bits, are the number
of bits mentioned simply sizeof(T)*CHAR_BIT, which doesn't say much about
value ranges, or are they stated to be the value representation bits?

Apart from char, and unsigned char that are guaranteed to not have
padding bits (anyone may tell about signed char?), the number of bits
of a type may contain padding bits etc.

 

[Andreas Huber]

Alf P. Steinbach wrote:
[snip]

short is at least 16 bits, int is at least 16 bits, and long
is at least 32 bits.
</quote>

That is contrary to what my copy of the standard says (see 3.9.1/1). It
basically says that char <= short <= int <= long (i.e. there is no
guarantee that a long is any bigger than a char, although there probably
is no platform where sizeof(long) == sizeof(char)). For more information
see 5.3.3 and 1.7..

Regards,

 

[Victor Bazarov]

Alf P. Steinbach wrote:
[snip]

short is at least 16 bits, int is at least 16 bits, and long
is at least 32 bits.
</quote>

That is contrary to what my copy of the standard says (see 3.9.1/1). It
basically says that char <= short <= int <= long (i.e. there is no

I wonder what _exactly_ does your copy say in 3.9.1/1 that makes you
derive that 'char <= short <= int'. But that doesn't matter. You are
actually correct, the C standard defined those relationships between
type sizes. It is true that char <= short <= int. It does not, however,
mean that there is no guarantee that 'int' is larger than 'char'. The
same C90 Standard when describing <limits.h> (and see 18.2.2/2 to know
that C++ mandates the same values for all xx_MAX and xx_MIN values),
does require *at least* ranges -32767..32767 for 'int' and -2^31..2^31
for 'long'.

guarantee that a long is any bigger than a char, although there probably
is no platform where sizeof(long) == sizeof(char)). For more information
see 5.3.3 and 1.7..

For more information see 18.2.2

V

 

[infobahn]

I wonder what _exactly_ does your copy say in 3.9.1/1 that makes you
derive that 'char <= short <= int'. But that doesn't matter. You are
actually correct, the C standard defined those relationships between
type sizes.

C&V please.

 

[Victor Bazarov]

C&V please.

I don't have a copy of C90. I used to have a printed copy at work,
but I don't work there any longer. I can be mistaken, therefore.

V

 

[Keith Thompson]

Alf P. Steinbach wrote:
[snip]

short is at least 16 bits, int is at least 16 bits, and long
is at least 32 bits.
</quote>

That is contrary to what my copy of the standard says (see
3.9.1/1). It basically says that char <= short <= int <= long
(i.e. there is no guarantee that a long is any bigger than a char,
although there probably is no platform where sizeof(long) ==
sizeof(char)). For more information see 5.3.3 and 1.7..

There's no contradiction.

The C standard guarantees that short and int are at least 16 bits, and
long is at least 32 bits (it states these in terms of minimum ranges,
but the requirement for a binary representation implies the sizes
given). It also guarantees that int as at least as wide as short, and
long is at least as wide as int. (Padding bits mean that this applies
to the ranges, not the sizes.)

I think the C++ standard has the same guarantees.

 

[Andreas Huber]

Alf P. Steinbach wrote:
[snip]

short is at least 16 bits, int is at least 16 bits, and long
is at least 32 bits.
</quote>

That is contrary to what my copy of the standard says (see 3.9.1/1).
It basically says that char <= short <= int <= long (i.e. there is no

I wonder what _exactly_ does your copy say in 3.9.1/1 that makes you
derive that 'char <= short <= int'.

I was wrong, it's actually 3.9.1/2...

But that doesn't matter. You are
actually correct, the C standard defined those relationships between
type sizes. It is true that char <= short <= int. It does not,
however, mean that there is no guarantee that 'int' is larger than
'char'. The same C90 Standard when describing <limits.h> (and see
18.2.2/2 to know that C++ mandates the same values for all xx_MAX and
xx_MIN values), does require *at least* ranges -32767..32767 for
'int' and -2^31..2^31 for 'long'.


For more information see 18.2.2

<blush> Ok, you got me on the left foot there. Although the C++ standard
does not itself guarantee anything like that, it *refers* to C standard
of which I don't even have a copy available... oh well.

Thanks for clarifying.

Regards,

 

[CBFalconer]

CBFalconer:

It was intentionally cross-posted; F.U.T. overridden.

You got your inputs from c.l.c, now it should go away. That's why
you should have set fups in the first place, and not have
overridden mine. Notice I didn't object to the initial cross-post.

 

[CBFalconer]

Alf P. Steinbach wrote:
.... snip ...

Yes. Also, except where otherwise is stated, C90 is a subset of
C++98 standard.

See what happens when you don't set followups in the initial
query. Misinformation like this works its way into c.l.c and is
likely to produce an interminable stupid cross group argument.
It's bad enough when the misinformation is native to the group.

 

[Jack Klein]

The C++ FAQ item 29.5 (this seems to be strongly related to C), at
<url: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.5>
mentions that

<quote>
C++ guarantees a char is exactly one byte which is at least 8 bits, short
is at least 16 bits, int is at least 16 bits, and long is at least 32
bits.
</quote>


Questions:

(1) This guarantee seems to come from the C standard. Which I don't
have. Does the C++ standard really guarantee this?

[snip]

The fact that the C++ standard decided to cite the C standard as a
normative reference is irrelevant here in comp.lang.c. The POSIX
standard is also based on the C standard, but that doesn't make POSIX
topical in comp.lang.c either.

What the C++ language guarantees is 100% off-topic in c.l.c, period.
The fact that the C++ standard does or does not refer to any or all
parts of the C standard has no meaning at all in terms of C.

 

[Alf P. Steinbach]

* Jack Klein:

The C++ FAQ item 29.5 (this seems to be strongly related to C), at
<url: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.5>
mentions that

<quote>
C++ guarantees a char is exactly one byte which is at least 8 bits, short
is at least 16 bits, int is at least 16 bits, and long is at least 32
bits.
</quote>


Questions:

(1) This guarantee seems to come from the C standard. Which I don't
have. Does the C++ standard really guarantee this?

[snip]

The fact that the C++ standard decided to cite the C standard as a
normative reference is irrelevant here in comp.lang.c. The POSIX
standard is also based on the C standard, but that doesn't make POSIX
topical in comp.lang.c either.

What the C++ language guarantees is 100% off-topic in c.l.c, period.
The fact that the C++ standard does or does not refer to any or all
parts of the C standard has no meaning at all in terms of C.

You mean, you don't have the slightest clue about the C aspects (although
you could easily check up on that, since you do have that standard), and
therefore choose to answer the 1/3 C++ part -- with some sour grapes.

Have a nice weekend.

 

[Jack Klein]

* Jack Klein:

The C++ FAQ item 29.5 (this seems to be strongly related to C), at
<url: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.5>
mentions that

<quote>
C++ guarantees a char is exactly one byte which is at least 8 bits, short
is at least 16 bits, int is at least 16 bits, and long is at least 32
bits.
</quote>


Questions:

(1) This guarantee seems to come from the C standard. Which I don't
have. Does the C++ standard really guarantee this?

[snip]

The fact that the C++ standard decided to cite the C standard as a
normative reference is irrelevant here in comp.lang.c. The POSIX
standard is also based on the C standard, but that doesn't make POSIX
topical in comp.lang.c either.

What the C++ language guarantees is 100% off-topic in c.l.c, period.
The fact that the C++ standard does or does not refer to any or all
parts of the C standard has no meaning at all in terms of C.

You mean, you don't have the slightest clue about the C aspects (although
you could easily check up on that, since you do have that standard), and
therefore choose to answer the 1/3 C++ part -- with some sour grapes.

Yes, I do, and had you posted this to comp.lang.c++ ONLY, I would have
been happy to answer based on what the C++ standard inherits from the
C standard.

But whether the C++ standard decided to incorporate the C standard in
its entirety, which it does not, or decided to only incorporate the
third comma on page 47, which it also does not, it not a C language
issue or topical for comp.lang.c at all.

There is no mention of the C++ language at all in the (now superceded
version) of the C standard that C++ chose to incorporate.

 

[Mark McIntyre]

C&V please.

5.2.4.2.1 Sizes of Integral types.

 

[Mark McIntyre]

* Jack Klein:

You mean, you don't have the slightest clue about the C aspects

Before posting such remarks, you might want to check up on Jack's
credentials, posting history etc. You just made a fool of yourself.

 

[Keith Thompson]

5.2.4.2.1 Sizes of Integral types.

Which actually discusses the ranges of integer types, which, in the
presence of padding bits, may not be directly related to their sizes.
A sufficiently perverse implementation could have
sizeof(long) < sizeof(int).

 

[Alf P. Steinbach]

* Mark McIntyre:

Before posting such remarks, you might want to check up on Jack's
credentials, posting history etc.

I know Jack's posting history for several years.

You just made a fool of yourself.

Before posting such remarks, you might want to check up on my
credentials, posting history etc.