Subtracting 0 from one-past-end pointer

[Old Wolf]

For the following code:

int main(void)
{
short a[100], *ptr;

ptr = a + 100;
ptr -= 0;
}

a runtime bounds-checker I use gives an error "Pointer
underrun" on the ptr -= 0 line. Is there something wrong
with that statement?

[Ben Pfaff]

Re: Subtracting 0 from one-past-end pointer

oldwolf@inspire .net.nz (Old Wolf) writes:
[color=blue]
> int main(void)
> {
> short a[100], *ptr;
>
> ptr = a + 100;
> ptr -= 0;
> }
>
> a runtime bounds-checker I use gives an error "Pointer
> underrun" on the ptr -= 0 line. Is there something wrong
> with that statement?[/color]

No, not as far as I can tell. (I'm surprised that the compiler
emits any code for that statement though.)
--
"Large amounts of money tend to quench any scruples I might be having."
-- Stephan Wilms

[E. Robert Tisdale]

Re: Subtracting 0 from one-past-end pointer

Old Wolf wrote:
[color=blue]
> For the following code:
>
> int main(void)
> {
> short a[100], *ptr;
>
> ptr = a + 100;
> ptr -= 0;
> }
>
> a runtime bounds-checker I use
> gives an error "Pointer underrun" on the ptr -= 0 line.
> Is there something wrong with that statement?[/color]
[color=blue]
> cat main.c[/color]
int main(int argc, char* argv[]) {
short a[100], *ptr = a + 100;
ptr -= 0;
return 0;
}
[color=blue]
> gcc -Wall -std=c99 -pedantic -o main main.c
> ./main[/color]

It works fine for me.

Your run time bounds checker is probably buggy.

[Robert Gamble]

Re: Subtracting 0 from one-past-end pointer

On Tue, 16 Nov 2004 18:05:11 -0800, Old Wolf wrote:
[color=blue]
> For the following code:
>
> int main(void)
> {
> short a[100], *ptr;
>
> ptr = a + 100;
> ptr -= 0;
> }
>
> a runtime bounds-checker I use gives an error "Pointer
> underrun" on the ptr -= 0 line. Is there something wrong
> with that statement?[/color]

I don't see anything wrong with any part of the code.

Perhaps for some reason your bounds-checker does not like the statement
before the one in question where you assign ptr to point to one past the
last element of a. Do you experience the same error if you change it to
"ptr = a + 99"?

Rob Gamble

[Robert Gamble]

Re: Subtracting 0 from one-past-end pointer

On Tue, 16 Nov 2004 18:27:25 -0800, E. Robert Tisdale wrote:
[color=blue]
> Old Wolf wrote:
>[color=green]
>> For the following code:
>>
>> int main(void)
>> {
>> short a[100], *ptr;
>>
>> ptr = a + 100;
>> ptr -= 0;
>> }
>>
>> a runtime bounds-checker I use
>> gives an error "Pointer underrun" on the ptr -= 0 line.
>> Is there something wrong with that statement?[/color]
>[color=green]
> > cat main.c[/color]
> int main(int argc, char* argv[]) {
> short a[100], *ptr = a + 100;
> ptr -= 0;
> return 0;
> }
>[color=green]
> > gcc -Wall -std=c99 -pedantic -o main main.c
> > ./main[/color]
>
> It works fine for me.
>
> Your run time bounds checker is probably buggy.[/color]

$ cat test7.c
int main(void)
{
short a[100], *ptr;

ptr = a + 1000;
ptr -= 0;
}

$ gcc -Wall --std=c99 -pedantic test7.c
$ ./a.out

Works fine for me.

The point here of course being that just because your compiler didn't
produce any warnings, or even that the resulting program does not crash,
does not imply that the code exhibits no undefined behavior.

Rob Gamble

[E. Robert Tisdale]

Re: Subtracting 0 from one-past-end pointer

Ben Pfaff wrote:
[color=blue]
> Old Wolf writes:
>[color=green]
>> int main(void)
>> {
>> short a[100], *ptr;
>>
>> ptr = a + 100;
>> ptr -= 0;
>> }
>>
>>a runtime bounds-checker I use
>>gives an error "Pointer underrun" on the ptr -= 0 line.
>>Is there something wrong with that statement?[/color]
>
> No, not as far as I can tell.
> (I'm surprised that
> the compiler emits any code for that statement though.)[/color]
[color=blue]
> cat main.c[/color]
int main(int argc, char* argv[]) {
short a[100], *ptr = a + 100;
ptr -= 0;
return 0;
}
[color=blue]
> gcc -Wall -std=c99 -pedantic -S main.c
> cat main.sold[/color]
.file "main.c"
.text
.globl main
.type main, @function
main:
pushl %ebp
movl %esp, %ebp
subl $232, %esp
andl $-16, %esp
movl $0, %eax
addl $15, %eax
addl $15, %eax
shrl $4, %eax
sall $4, %eax
subl %eax, %esp
leal -216(%ebp), %eax// a
addl $200, %eax // a + 100;
movl %eax, -220(%ebp)// ptr = a + 100
movl $0, %eax // return 0;
leave
ret
.size main, .-main
.section .note.GNU-stack,"",@progb its
.ident "GCC: (GNU) 3.4.1"

My compiler doesn't.

[Dave Vandervies]

Re: Subtracting 0 from one-past-end pointer

In article <pan.2004.11.17 .03.19.52.59798 2@gmail.com>,
Robert Gamble <rgamble99@gmai l.com> wrote:[color=blue]
>On Tue, 16 Nov 2004 18:05:11 -0800, Old Wolf wrote:
>[color=green]
>> For the following code:
>>
>> int main(void)
>> {
>> short a[100], *ptr;
>>
>> ptr = a + 100;
>> ptr -= 0;
>> }
>>
>> a runtime bounds-checker I use gives an error "Pointer
>> underrun" on the ptr -= 0 line. Is there something wrong
>> with that statement?[/color]
>
>I don't see anything wrong with any part of the code.[/color]

Nor do I. One past the end of an array is a valid pointer for any use
other than dereferencing it, and adding or subtracting 0 to/from a valid
pointer should be a no-op (and, in particular, give you back a (the same)
valid pointer).

[color=blue]
>Perhaps for some reason your bounds-checker does not like the statement
>before the one in question where you assign ptr to point to one past the
>last element of a. Do you experience the same error if you change it to
>"ptr = a + 99"?[/color]

If that's the case, then the bounds-checker is wrong. It should only
trap if the OP's code tries to dereference ptr, not if it subtracts a
nonnegative value from it. (Well, nonnegative and not greater than 100.)


dave

--
Dave Vandervies dj3vande@csclub .uwaterloo.ca

Silly, yes. But have we shown that it is unhelpful?
--Daniel Fox in comp.lang.c

[Dan Pop]

Re: Subtracting 0 from one-past-end pointer

In <843a4f78.04111 61805.27326514@ posting.google. com> oldwolf@inspire .net.nz (Old Wolf) writes:
[color=blue]
>For the following code:
>
> int main(void)
> {
> short a[100], *ptr;
>
> ptr = a + 100;
> ptr -= 0;
> }
>
>a runtime bounds-checker I use gives an error "Pointer
>underrun" on the ptr -= 0 line. Is there something wrong
>with that statement?[/color]

Nope, the standard explicitly allows it:

If both the pointer operand and the result point to elements
of the same array object, or one past the last element of the
array object, the evaluation shall not produce an overflow;

The definition of C makes *reliable* bound checking at run time incredibly
difficult, but your case can be trivially handled by a run time bound
checker.

Dan
--
Dan Pop
DESY Zeuthen, RZ group
Email: Dan.Pop@ifh.de
Currently looking for a job in the European Union