Re: [PATCH] read command

[Yoshinori K. Okuji]

On Friday 01 February 2008 23:48, Robert Millan wrote:
> On Fri, Feb 01, 2008 at 11:45:50PM +0100, Robert Millan wrote:
> > +static void *
> > +grub_xrealloc (void *ptr, grub_size_t size)
> > +{
> > +  void *value = grub_realloc (ptr, size);
> > +  if (value == 0)
> > +    grub_fatal ("Virtual memory exhausted");
> > +  return value;
> > +}
> > +
> > +static char *
> > +grub_getline (void)
> > +{
> > +  int i;
> > +  char *line;
> > +
> > +  i = 0;
> > +  line = grub_malloc (1 + i + sizeof('\0'));
> > +
> > +  while ((line[i - 1] != '\n') && (line[i - 1] != '\r'))
> > +    {
> > +      line[i] = grub_getkey ();
> > +      if (grub_isprint (line[i]))
> > +   grub_putchar (line[i]);
> > +      i++;
> > +      line = grub_xrealloc (line, 1 + i + sizeof('\0'));
> > +    }
> > +  line[i] = '\0';
> > +
> > +  return line;
> > +}
>
> Does it make sense to move any of these two to kernel?  Or to normal.mod ? 
> Or maybe just to un-static-ize them and leave them here?

What would be other use cases?

BTW it is a bad idea to use xrealloc. xmalloc and xrealloc are sometimes used 
in the GNU Project, but they must not be used in GRUB. If a program runs in 
user space, when it panics, the control is back to the OS. But GRUB is a 
standalone program. When it panics, the user loses any kind of control. Thus 
grub_fatal must be used only if there is nothing else you can do.

In this case, grub_getline should simply return NULL. How the user deals with 
this situation is up to the user.

Okuji