Re: RFC: Fix various problems with "printf" and warnings

[Daniel Jacobowitz <drow@false.org>]

On Sun, Jan 22, 2006 at 06:28:05AM +0200, Eli Zaretskii wrote:
> > Date: Sat, 21 Jan 2006 19:31:17 -0500
> > From: Daniel Jacobowitz <drow@false.org>
> > 
> > +	      else if (*f == ' ')
> > +		seen_space = 1;
> > +	      else if (*f == ' ')
> > +		seen_plus = 1;
> 
> A typo in the second "else if" clause.

Oops.  Thank you.

> > +	  case long_double_arg:
> > +#ifdef HAVE_LONG_DOUBLE
> > +	    {
> > +	      long double val = value_as_double (val_args[i]);
> > +	      printf_filtered (current_substring, val);
> 
> Won't `value_as_double' lose accuracy here, possibly catastrophically
> (i.e., a non-zero number coming out as zero, etc.)?

Not usually.  value_as_double is somewhat misnamed; it returns a
DOUBLEST, which will be long double if the host supports long double.
Now, if the target has a long double with a greater precision (or just
different precision, due to different format) than the host long
double, this will lose out a bit.  But this isn't the only place in GDB
that will lose in that case.

-- 
Daniel Jacobowitz
CodeSourcery