How to delete first n characters from a binary file

Here is a program which will lop off the first N bytes of ./temp, leaving the original file as ./temp-old. It assumes the file will fit in memory. You specify N on the command line.

/*
 * THIS PROGRAM IS PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND,
 * EITHER EXPRESSED OR IMPLIED.
 */
#include <errno.h>
#include <fcntl.h>
#include <string.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/stat.h>
#include <sys/types.h>

#define ERR (-1)

int main (
        int     argc,
        char ** argv
) {
        int             fdin;
        int             fdout;
        unsigned long   n;
        struct stat     st;
        size_t          sz;
        ssize_t         ssz;
        char *          data;

        if (argc != 2) {
                fprintf(stderr, "usage: %s nbytes\n", argv[0]);
                return 1;
        }
        n = strtoul(argv[1], NULL, 10);
        if (errno) {
                fprintf(stderr, "nbytes (%s) is suspect\n", argv[1]);
                return 1;
        }
        fdin = open("./temp", O_RDONLY, 0);
        if (fdin == ERR) {
                fprintf(stderr, "open input: %s\n", strerror(errno));
                return 1;
        }
        if (fstat(fdin, &st) == ERR) {
                fprintf(stderr, "stat input: %s\n", strerror(errno));
                return 1;
        }
        sz = st.st_size;
        if (sz < n) {
                fprintf(stderr, "file is not that big\n");
                return 1;
        }
        data = malloc(sz);
        if (data == NULL) {
                fprintf(stderr, "insufficient memory\n");
                return 1;
        }
        ssz = read(fdin, data, sz);
        if (ssz < 0) {
                fprintf(stderr, "read input: %s\n", strerror(errno));
                return 1;
        }
        if ((size_t)ssz != sz) {
                fprintf(stderr, "read was short\n");
                return 1;
        }
        (void)close(fdin);
        fdout = open("./temp-new", O_CREAT|O_EXCL|O_WRONLY, st.st_mode);
        if (fdout == ERR) {
                fprintf(stderr, "open output: %s\n", strerror(errno));
                return 1;
        }
        sz -= n;
        ssz = write(fdout, data + n, sz);
        if (ssz < 0) {
                fprintf(stderr, "write output: %s\n", strerror(errno));
                return 1;
        }
        if ((size_t)ssz != sz) {
                fprintf(stderr, "write was short\n");
                return 1;
        }
        if (close(fdout) == ERR) {
                fprintf(stderr, "write close: %s\n", strerror(errno));
                return 1;
        }
        if (link("./temp", "./temp-old") == ERR) {
                fprintf(stderr, "link input: %s\n", strerror(errno));
                return 1;
        }
        if (rename("./temp-new", "./temp") == ERR) {
                fprintf(stderr, "rename output: %s\n", strerror(errno));
                return 1;
        }
        return 0;
}

I wrote it carefully, but, of course you should make a backup copy of your file before using it, just in case...

Use the dd utility. One way is to set the block size to be equal to the number of bytes you want to delete, then tell dd to seek one block ahead in the input file:

# N is number of bytes to delete: wherever N is seen
# substitute this value.

dd if=infile of=outfile bs=N seek=1

Now outfile is a copy of infile, but with the first N bytes removed.

The following command will also work; unfortunately, it will perform 1 byte read and write system calls, since the block size is 1 byte:

# N is number of bytes to delete
dd if=infile of=outfile bs=1 seek=N

We can use separate input and output block sizes (ibs and obs) if N is too small to be a good buffering factor, so that at least we write in larger chunks:

# read N at a time, skipping first N; write 64 kB at a time:
dd if=infile of=outfile ibs=N obs=65536 seek=1

We can simulate an in-place deletion by then renaming outfile to infile.

It should be possible to actually do this in place. The following almost works:

dd if=inoutfile of=inoutfile bs=N conv=notrunc skip=1

The only problem is that without the conv=notrunc option, the file is immediately truncated to zero length. The content is gone! And with the option, the file isn't truncated at all after being written.

GNU Coreutils has a truncate command which can be used:

dd if=inoutfile of=inoutfile bs=N conv=notrunc skip=1
truncate -s -N inoutfile

The - prefix on the size argument in truncate means "reduce the size by that many bytes".