Two-Dimensional Char Array

[miketigerwoods]

I'm having problems where I need to parse a command line, and place
each token from strtok() into an array. I've read here:
http://www.phim.unibe.ch/comp_doc/c_...PT/arrays.html (at
bottom) that you can access a char[][] by doing char[i] and getting the
string. However when I try the code below I get

error: invalid use of array with unspecified bounds

whenever argv is accessed. It is declared in main() as:

char argv[ARGV_LENGTH][ARGV_LENGTH]; /* ARGV_LENGTH is a const int of
64 */

This is pretty much my first time with C, I'm usually using Java and a
little C++. Any help is appreciated.

Thanks,
Mike
void parse_cmdline(c har cmdline [], char argv [][])
{
char * command;
int i = 1;
command = strtok(cmdline, DELIMITERS);
argv[0] = malloc(sizeof(c har)*(strlen(co mmand)+1));
strcpy(argv[0],command);
while (command != NULL) {
printf("%s\n", command);
command = strtok(NULL, DELIMITERS);
argv[i] = malloc(sizeof(c har)*(strlen(co mmand)+1));
strcpy(argv[i],command);
i++;
}
}

[Vimal Aravindashan]

Netocrat wrote:

On Wed, 19 Oct 2005 10:47:46 +0530, Vimal Aravindashan wrote:

Netocrat wrote:

The original while loop had the correct semantics. This do loop fails
to check for the above function returning NULL. There is nothing
wasted - it is a necessary check.

Check the OP's code again. He doesn't check for the first one himself.

Here is the relevant portion of the OP's code:

command = strtok(cmdline, DELIMITERS); ...
while (command != NULL) {
...
command = strtok(NULL, DELIMITERS);
...
}

This has correct semantics; the do loop does not.

Perhaps you misread my comments as applying to the test for i <
ARGV_LENGTH that I added.

I didn't misread it. Forget about what maybe relevant or not. The
following is this OP's original post:

command = strtok(cmdline, DELIMITERS);
argv[0] = malloc(sizeof(c har)*(strlen(co mmand)+1));
strcpy(argv[0],command);
while (command != NULL) {

As per the above code, even if strtok() returns NULL for the first call,
the OP makes it a point to call strcpy() for it.
Netocrat, I don't think either of us can decide what the "correct
semantics" are for the OP solution. I am surprised that he hasn't
commented on our replies, because apparently he wants a NULL at the end
of the arg list, which neither of our codes do.

Given the OP's code, I think its safe to assume further that his set of
DELIMITERS combined with strtok() will yield the necessary results for
him.

It will likely generate correct results without ill-effect for the OP's
situation, but a general warning re strtok is warranted.

Mike, you've been warned.

Cheers,
Vimal.
--
"If you would be a real seeker after truth, it is necessary that at
least once in your life you doubt, as far as possible, all things."
-- Rene Descartes

[Netocrat]

On Wed, 19 Oct 2005 17:19:58 +0530, Vimal Aravindashan wrote:

Netocrat wrote:

On Wed, 19 Oct 2005 10:47:46 +0530, Vimal Aravindashan wrote:

Netocrat wrote:

The original while loop had the correct semantics. This do loop fails
to check for the above function returning NULL. There is nothing
wasted - it is a necessary check.

Check the OP's code again. He doesn't check for the first one himself.
Here is the relevant portion of the OP's code:

command = strtok(cmdline, DELIMITERS); ...
while (command != NULL) {
...
command = strtok(NULL, DELIMITERS);
...
}

This has correct semantics; the do loop does not.

Perhaps you misread my comments as applying to the test for i <
ARGV_LENGTH that I added.

I didn't misread it. Forget about what maybe relevant or not. The
following is this OP's original post:

command = strtok(cmdline, DELIMITERS);
argv[0] = malloc(sizeof(c har)*(strlen(co mmand)+1));
strcpy(argv[0],command);
while (command != NULL) {

As per the above code, even if strtok() returns NULL for the first call,
the OP makes it a point to call strcpy() for it.

He does and that is unsafe - strlen and strcpy do not take NULL pointers.
The original code's looping is not entirely correct but with those two
lines removed it is - perhaps my omitting them was misleading; the point
is that the while loop is more naturally suited to correctly dealing with
errors.

For the do loop to behave correctly one would have to either:
a) only enter the loop if cmdline != NULL - but why write the code for
that additional test when the while loop code already provides it?
b) eliminate the call to strtok prior to the loop, reverse the order of
the functions within the loop and break if strok returns NULL - but then
to remove unnecessary tests as you seemed to want to do, we should remove
the loop conditional - so a do loop is no more preferable than a
while(1){} or for(;;){} loop and the code is considerably different to
that originally proposed.
Netocrat, I don't think either of us can decide what the "correct
semantics" are for the OP solution.
True, but we can decide when a solution fails to prevent NULL from being
passed to functions to which it should not be passed.
I am surprised that he hasn't commented on our replies, because
apparently he wants a NULL at the end of the arg list, which neither of
our codes do.

Well that can be added immediately prior to the return if he does want it:

if (i == ARGV_LENGTH)
i--;
argv[i++] = NULL;

[...]
--
http://members.dodo.com.au/~netocrat