🚀 go-pugleaf

In article <1992Sep12.235241.23213@eng.ufl.edu> zzang@stat.ufl.edu (Zhuo Zang) writes:

> Hello, the following program using lseek will give different results
> in BC++2.0 than that of in UNIX.  the BC++2.0 will print the last
> character twice.

I tried this program with BCC++ for OS/2, BC++ for MS-DOS, and GCC on a
unix box.  All produced the same result.  This assumes that the first line
of the file foo.c is the following:

/* use bcc -DMSDOS foo.c to compile in BC++i

The program seeks to file offset 3, then reads 20 bytes from the file
into a buffer.  These 20 bytes, which the program prints, should be
the following:

use bcc -DMSDOS foo.

There are several dangers and harmless mistakes in this program, though.

First, the buffer is declared as follows:

>char buf[BUFSIZ];

The assumption here is that the buffer is greater than 20 bytes long.
If BUFSIZ is <= 20, bad things could happen.  BUFSIZ is usually 512,
but you can't count on it.

Secondly, the attempt to zero-terminate the string didn't work; it
just converted '\0' to a (char *), and assigned it to sp:

>  sp = '\0';

This should have been:

   buf[20] = '\0';

or

   sp[20] = '\0';

since sp was set to point to buf[0].

However, the program worked without this, because buf[] was declared
as a global variable, and hence was initially zero-filled when the
program was started.  If you had used the buffer again, or had declared
the buffer as a local variable, the program would have failed, because
the buffer would probably not be zero-filled.