Re: string is const char [] ?

Re: string is const char [] ?

[Oleg Zabluda]

: On 23 Jun 1998, Alexandre Oliva wrote:
: > >> >   char a1[] = "test not const";
: >
: > >> This line should cause egcs to flag an error.

: Kamil Iskra <kamil@dwd.interkom.pl> writes:
: > > Excuse me?! I missed the original mail, but do you really suggest that al
: > > the programmers should suddenly switch to:
: >
: > > char a1[]={'t', 'e', 's', 't', ' ', 'n', 'o', 't', ' ', 'c', 'o', 'n',
: > > 's', 't', '\0'};

: Alexandre Oliva writes:
: > How about:
: > const char a1[] = "test not const";

: Kamil is trying to initialize a char[], so your suggestion is irrelevant.
: Programmers must sometimes initialize a writable string, and the standard
: way to do this in C is exactly the way you claim is forbidden.  This would
: need to be documented as a language incompatibility.

: I don't believe that you are correct, Alexandre.  I think you are
: confusing assignment (where there is a conversion) with initialization.
: If the standard is unclear, this is a case where the committee should be
: asked to produce a clarification.

I think this is the rare case when Alexandre is confusing issues,
and the standard is very clear (8.5.2)

char a1[] = "test not const"; is definitely legal. What
is also legal, but deprecated is

char *a1 = "test not const";

This generates a warning when compiled with -Wwrite-strings.

Here is a real bug in egcs:

void f(      char *) {}
voif f(const char *) {}

f("test not const"); // error in egcs. f(const char *) must be called,
                     // because the type of string literals now is
                     // array of n const char. However egcs calls
                     // f(char*). -Wwrite-strings has no effect.
                     // And even warning is surpressed.

Oleg.
--
Life is a sexually transmitted, 100% lethal disease.

Re: string is const char [] ?

[Carlo Wood]

| Here is a real bug in egcs:
| 
| void f(      char *) {}
| voif f(const char *) {}
| 
| f("test not const"); // error in egcs. f(const char *) must be called,
|                      // because the type of string literals now is
|                      // array of n const char. However egcs calls
|                      // f(char*). -Wwrite-strings has no effect.
|                      // And even warning is surpressed.
| 
| Oleg.

Shouldn't this be added as a test in the testsuite?

Carlo

----

PS

~>/usr/src/egcs/egcs-cvs-objdir/gcc/g++ -B/usr/src/egcs/egcs-cvs-objdir/gcc/ -I/usr/src/egcs/egcs-cvs/libstdc++ -I/usr/src/egcs/egcs-cvs/libio -static -L/usr/src/egcs/egcs-cvs-objdir/libraries/libstdc++ -Wwrite-strings tst.cc
~>a.out
      char *
~>cat tst.cc
#include <iostream>

void f(      char *) { cout << "      char *" << endl; }
void f(const char *) { cout << "const char *" << endl; }
int main(void) { f("test not const"); }

Re: string is const char [] ?

[Oleg Zabluda]

In article < 199806251626.SAA05600@jolan.ppro > you wrote:
: | Here is a real bug in egcs:
: |
: | void f(      char *) {}
: | voif f(const char *) {}
: |
: | f("test not const"); // error in egcs. f(const char *) must be called,
: |                      // because the type of string literals now is
: |                      // array of n const char. However egcs calls
: |                      // f(char*). -Wwrite-strings has no effect.
: |                      // And even warning is surpressed.

: Shouldn't this be added as a test in the testsuite?

I think it should. I am sorry for not submitting a patch at this 
time. I'll setup everything needed for generating good patches
Real Soon Now. I've posted this hoping somebody else would.

Oleg.
--
Life is a sexually transmitted, 100% lethal disease.

Re: string is const char [] ?

[Kaz Kylheku]

On Wed, 24 Jun 1998, Richard Henderson wrote:

> On Wed, Jun 24, 1998 at 03:07:30PM +0200, Kamil Iskra wrote:
> > Don't get me wrong, I'm sure you know the standard far better than I do,
> > but are you SURE that it disallows this initialisation syntax?
> 
> For non-constant static or file-scope variables, the "" form is
> legal.  For automatic variables, it is not.
> 
> > I find it
> > hard to believe, since first, I think this syntax is very common (at least
> > I use it quite often), and second, this syntax is fine in C,
> 
> Well, gcc provides the "" form for automatic variables as an extension.
> I wish it did not, for I often find folks initializing things at runtime
> taking up hordes of .text space, when they weren't intending to.

To clarify, you mean that:

	{
		char x[] = "abcd";

	}

is wrong? That is perfectly valid C. But it does potentially take up
.text space, since the compiler will likely store the "abcd" string
literal there and generate a copy operation.

Re: string is const char [] ?

[Kaz Kylheku]

On Wed, 24 Jun 1998, Richard Henderson wrote:

> On Wed, Jun 24, 1998 at 03:07:30PM +0200, Kamil Iskra wrote:
> > Don't get me wrong, I'm sure you know the standard far better than I do,
> > but are you SURE that it disallows this initialisation syntax?
> 
> For non-constant static or file-scope variables, the "" form is
> legal.  For automatic variables, it is not.

Could you cite the relevant sections from either the C or C++ standards?
(Or both).

Re: string is const char [] ?

[Ross Smith]

From: Kamil Iskra <kamil@dwd.interkom.pl>
>
>  char a1[] = "test not const";
>
>I didn't want a const array, I wanted a plain one, probably for a good
>reason. Maybe I wanted to change some characters or append something to
>it? Now, how should I do that to satisfy the standard and avoid the
much
>more error-prone standard array-initialisation syntax?
>
>Don't get me wrong, I'm sure you know the standard far better than I
do,
>but are you SURE that it disallows this initialisation syntax? I find
it
>hard to believe, since first, I think this syntax is very common (at
least
>I use it quite often), and second, this syntax is fine in C, and C++ is
>generally supposed to accept C code (I know, there are certain
>incompatibilities, and, IIRC, there is even a list of them. Is this
>particular incompatibility listed there?).
>
>If the standard indeed does deliberately disallow this initialisation
>syntax, I would vote that G++ allows it, preferably without any
warnings
>(unless compiling with -ansi, -pedantic or some such).

The standard *does* allow the above syntax. (Or at least CD2 did; I
haven't seen the FDIS but I haven't heard about any change in this
area.) The relevant section is 8.5.2 [dcl.init.string]. It makes a
special exception to the normal initialisation rules, allowing the
initialisation of a non-const char (or wchar_t) array with a string
literal.

(There is one change from the old C rules, though: if you specify the
array size, you can't leave out the terminating null. E.g. char foo[3] =
"bar"; is legal in C (giving an unterminated string) but illegal in
C++.)

--
Ross Smith ................................... mailto:ross.s@ihug.co.nz
.............. The Internet Group, Auckland, New Zealand ..............
  "Remember when we told you there was no future? Well, this is it."
                                                        -- Blank Reg

Re: string is const char [] ?

[Richard Henderson]

On Wed, Jun 24, 1998 at 03:07:30PM +0200, Kamil Iskra wrote:
> Don't get me wrong, I'm sure you know the standard far better than I do,
> but are you SURE that it disallows this initialisation syntax?

For non-constant static or file-scope variables, the "" form is
legal.  For automatic variables, it is not.

> I find it
> hard to believe, since first, I think this syntax is very common (at least
> I use it quite often), and second, this syntax is fine in C,

Well, gcc provides the "" form for automatic variables as an extension.
I wish it did not, for I often find folks initializing things at runtime
taking up hordes of .text space, when they weren't intending to.

The correct solution to init auto variables is this little-known
function called strcpy.


r~

Re: string is const char [] ?

[Alexandre Oliva]

Joe Buck <jbuck@Synopsys.COM> writes:

> On 23 Jun 1998, Alexandre Oliva wrote:
>> >> >   char a1[] = "test not const";

> I don't believe that you are correct, Alexandre.

Thank you all for pointing out my mistake.  When I was looking for the
requirement in the FDIS that stated that strings are char const[], I
came across the following statement, that surprised me:

  Example:
  char array[4] = "abcd";         // valid C, invalid C++

which I took in a more general sense than it should have been taken,
because I didn't read the text just before it [diff.decl]:

4 Change: In C++, when initializing an array of character with a string,
  the  number  of  characters  in  the string (including the terminating
  '\0') must not exceed the number of elements in the array. [...]


What I was really looking for was [diff.lex]:

4 Change: String literals made const
  The type of a string literal is changed from "array of char" to "array
  of const char."  The type of a wide string  literal  is  changed  from
  "array of wchar_t" to "array of const wchar_t."
[...]
  char* p = "abc";                // valid in C, deprecated in C++
  char* q = expr ? "abc" : "de";  // valid in C, invalid in C++

[conv.array]:

2 A string literal (_lex.string_) that is not a wide string literal  can
  be  converted  to  an  rvalue of type "pointer to char"; [...]
  Note:  this  conversion  is  deprecated.


Sorry for spreading misinformation.

-- 
Alexandre Oliva
mailto:oliva@dcc.unicamp.br mailto:aoliva@acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil

Re: string is const char [] ?

[Horst von Brand]

Alexandre Oliva <oliva@dcc.unicamp.br> said:
> Kamil Iskra <kamil@dwd.interkom.pl> writes:
> > On 23 Jun 1998, Alexandre Oliva wrote:
> >> >   char a1[] = "test not const";

> >> This line should cause egcs to flag an error.

I think this is ridiculous. Why should it matter that the value from which
a particular variable is _initialized_ is or not a constant? With this
reasoning, we'd have:

  int i = 5;   // illegal?!
  int j = k;   // OK, as long as k isn't 'const int'?!

> > Excuse me?! I missed the original mail, but do you really suggest that all
> > the programmers should suddenly switch to:

> > char a1[]={'t', 'e', 's', 't', ' ', 'n', 'o', 't', ' ', 'c', 'o', 'n',
> > 's', 't', '\0'};

By the above reasoning, this should be an error too.

Besides, in C doing what you state above is quite common, is compatibility
with C being thrown out the window just like that for no reason at all?

> How about:

> const char a1[] = "test not const";

That means something totally different.

Something is very broken here. Maybe my understanding of the issue ;)
-- 
Dr. Horst H. von Brand                       mailto:vonbrand@inf.utfsm.cl
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: string is const char [] ?

[jeroen dobbelaere xe44 7682]

Alexandre Oliva writes:
 > Kamil Iskra <kamil@dwd.interkom.pl> writes:
 > 
 > > On 23 Jun 1998, Alexandre Oliva wrote:
 > >> >   char a1[] = "test not const";
 > 
 > >> This line should cause egcs to flag an error.
 > 
 > > Excuse me?! I missed the original mail, but do you really suggest that all
 > > the programmers should suddenly switch to:
 > 
 > > char a1[]={'t', 'e', 's', 't', ' ', 'n', 'o', 't', ' ', 'c', 'o', 'n',
 > > 's', 't', '\0'};
 > 
 > How about:
 > 
 > const char a1[] = "test not const";


What if you want to initialise a changeable char array with a default string ?


// stupid example
void f()
{
  char counter[]="999";  // should work (normal initialization, size =4) 

  // char *counter="000";   // should at least give a warning, and
                         //should not work with this example

  for(;;)
  {
    if(counter[2]-- == '0')
    {
      counter[2]='9';
      if(counter[1]-- == '0')
      {
        counter[1]='9;
        if(counter[0]-- == '0')
        {
          counter[0]='9';

          break; // exit the for(;;)
	}
       }
     }

    // do something with the current  counter
    cout << counter << endl;
   }
}

Re: string is const char [] ?

[Joe Buck]

 

 
On 23 Jun 1998, Alexandre Oliva wrote:
> >> >   char a1[] = "test not const";
> 
> >> This line should cause egcs to flag an error.

Kamil Iskra <kamil@dwd.interkom.pl> writes:
> > Excuse me?! I missed the original mail, but do you really suggest that all
> > the programmers should suddenly switch to:
> 
> > char a1[]={'t', 'e', 's', 't', ' ', 'n', 'o', 't', ' ', 'c', 'o', 'n',
> > 's', 't', '\0'};

Alexandre Oliva writes:
> How about:
> const char a1[] = "test not const";

Kamil is trying to initialize a char[], so your suggestion is irrelevant.
Programmers must sometimes initialize a writable string, and the standard
way to do this in C is exactly the way you claim is forbidden.  This would
need to be documented as a language incompatibility.

I don't believe that you are correct, Alexandre.  I think you are
confusing assignment (where there is a conversion) with initialization.
If the standard is unclear, this is a case where the committee should be
asked to produce a clarification.

Re: string is const char [] ?

[Kamil Iskra]

On 24 Jun 1998, Alexandre Oliva wrote:

> >> >   char a1[] = "test not const";
> >> This line should cause egcs to flag an error.
> > Excuse me?! I missed the original mail, but do you really suggest that all
> > the programmers should suddenly switch to:
> > char a1[]={'t', 'e', 's', 't', ' ', 'n', 'o', 't', ' ', 'c', 'o', 'n',
> > 's', 't', '\0'};
> How about:
> 
> const char a1[] = "test not const";

I didn't want a const array, I wanted a plain one, probably for a good
reason. Maybe I wanted to change some characters or append something to
it? Now, how should I do that to satisfy the standard and avoid the much
more error-prone standard array-initialisation syntax?

Don't get me wrong, I'm sure you know the standard far better than I do,
but are you SURE that it disallows this initialisation syntax? I find it
hard to believe, since first, I think this syntax is very common (at least
I use it quite often), and second, this syntax is fine in C, and C++ is
generally supposed to accept C code (I know, there are certain
incompatibilities, and, IIRC, there is even a list of them. Is this
particular incompatibility listed there?).

If the standard indeed does deliberately disallow this initialisation
syntax, I would vote that G++ allows it, preferably without any warnings
(unless compiling with -ansi, -pedantic or some such).

/ Kamil Iskra    AmigaOS  Linux/i386  Linux/m68k               \
| GeekGadgets GCC maintainer   UNIX system administrator       |
| iskra@student.uci.agh.edu.pl  kiskra@ernie.icslab.agh.edu.pl |
\ kamil@dwd.interkom.pl   http://student.uci.agh.edu.pl/~iskra /

Re: string is const char [] ?

[Alexandre Oliva]

Kamil Iskra <kamil@dwd.interkom.pl> writes:

> On 23 Jun 1998, Alexandre Oliva wrote:
>> >   char a1[] = "test not const";

>> This line should cause egcs to flag an error.

> Excuse me?! I missed the original mail, but do you really suggest that all
> the programmers should suddenly switch to:

> char a1[]={'t', 'e', 's', 't', ' ', 'n', 'o', 't', ' ', 'c', 'o', 'n',
> 's', 't', '\0'};

How about:

const char a1[] = "test not const";

-- 
Alexandre Oliva
mailto:oliva@dcc.unicamp.br mailto:aoliva@acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil

Re: string is const char [] ?

[Kamil Iskra]

On 23 Jun 1998, Alexandre Oliva wrote:

> >   char a1[] = "test not const";
> This line should cause egcs to flag an error.  The only
> backward-compatibility provided in the standard is [conv.array]/2,
> that allows string literals to be converted to non-const `char*'.
> Hence, string literals cannot be used to initialize a char[].  At
> least a warning should be issued in this case.

Excuse me?! I missed the original mail, but do you really suggest that all
the programmers should suddenly switch to:

char a1[]={'t', 'e', 's', 't', ' ', 'n', 'o', 't', ' ', 'c', 'o', 'n',
's', 't', '\0'};

??? IMO, this is INSANE.

/ Kamil Iskra    AmigaOS  Linux/i386  Linux/m68k               \
| GeekGadgets GCC maintainer   UNIX system administrator       |
| iskra@student.uci.agh.edu.pl  kiskra@ernie.icslab.agh.edu.pl |
\ kamil@dwd.interkom.pl   http://student.uci.agh.edu.pl/~iskra /