Character Literals

student04 · 06-20-2008, 09:42 AM

While trying to understand some code for work I came across something I was initially puzzled over, but now have a slight understanding as to what the original author intended to do:

Code:

#include <string.h>

     ...
     char *ptr;
     ptr = strrchr( some_string, '.' );
     if( ptr ) *ptr = '\000';
     ...

Now, what I think the intention is, is to artificially end the specified string at the last location of a period, if found.

However, what can possibly be going wrong with the '\000' literal? The code compiles without warnings (even with -Wall). When I run this it segfaults assigning that literal (run on a Solaris 10 box with gcc 3.4.3). Is this a NUL-byte (value 0) followed by two zero characters (value 30)? From what I can tell the code is pumping 3 bytes into a space allocated to hold only one byte. The literal seems to be okay with the compiler, but not when assigning it to the address pointed to by 'ptr'.

I think it should be rewritten as

Code:

#include <string.h>

     ...
     char *ptr;
     ptr = strrchr( some_string, '.' );
     if( ptr ) *ptr = '\0';
     ...

This code probably never got executed in the original file as the strings given to it probably never had periods in them...

-AM

netdog · 06-20-2008, 10:02 AM

Ok, somebody check me on this, it's been a while since C...

Your assumption is good, the \000 is an octal literal for a one byte 0 which will end the string. I think the problem is the quote marks. By sitting in the quotes it translates to three character bytes instead of one. Try assigning the value as some sort of int. One of the following...

- if(ptr) *ptr = \000; // octal
- if(ptr) *ptr = \0x00; // hex
- if(ptr) *ptr = 0; // dec

That will assign the character at *ptr to 0 (null) and end the string.

Hope that helps

Ciao

student04 · 06-20-2008, 10:39 AM

Oh that's interesting. I hadn't thought of octals for this... See I thought '\000' would translate to 0x003030 (NUL-byte, zero, zero). My task wasn't to clean up the code but to band-aid it.. And it is 8 years old with the original author gone, so I didn't want to touch the rest of it.

So,

Code:

*ptr = '\0'; // and
*ptr = 0;

are the same, correct? Since the decimal 0 is typecast into a single byte 0 before assigned to the location at ptr?

I can't really ask the person who wrote this what the heck they intended to do.. either use an octal value like you mentioned, or write two more zeros, or...

My confusion is only

Code:

'\000';

should not be valid syntax, disregarding semantics. Am I wrong?

ntubski · 06-20-2008, 12:22 PM

See http://www.cppreference.com/escape_sequences.html

Code:

\nnn  	Octal number (nnn)
\0 	Null character (really just the octal number zero)

student04 · 06-20-2008, 12:59 PM

Quote:

Originally Posted by ntubski

See http://www.cppreference.com/escape_sequences.html

Code:

\nnn  	Octal number (nnn)
\0 	Null character (really just the octal number zero)

Okay; just thinking too hard about this.

Thanks!

osor · 06-20-2008, 06:26 PM

Quote:

Originally Posted by student04

Okay; just thinking too hard about this.

Yes. Whenever you have single quote-marks, you will only get one character, period (with width of int in C and width of char in C++).

Quote:

Originally Posted by student04

When I run this it segfaults assigning that literal (run on a Solaris 10 box with gcc 3.4.3).

If the segfault does indeed occur on the assignment (*ptr = '\000';), it is most likely because the original string (some_string) was in read-only memory. This is where you should direct your debugging efforts. In particular, if some_string is a (pointer to a) string literal (and not an array initialized to a string literal), many older compilers (wrongly) allowed this to work, whereas most modern compilers put string literals in read-only memory. For example, the following program would have worked on an older compiler, but should segfault with a newer compiler:

Code:

#include <stdio.h>

int main()
{
        char *string = "safe";
        string[2] = 'v';

        return 0;
}