Re: [avr-libc-dev] Re: [avr-gcc-list] I fixed dtostre()

[E. Weddington]

On 10 Sep 2002 at 22:53, Joerg Wunsch wrote:

> As E. Weddington wrote:
> 
> > Good idea. I was just looking up how IAR does it for their toolset.
> > They have 3 versions of (s)printf according to their docs: 1.
> > Full-blown version. 2. Medium version: no floating point. %f, %g,
> > %G, %e, %E specifiers produce errors. 3. Small version: same as
> > medium except supports only the %%, %d, %o, %c, %s and %x specifiers
> > for int objects, and does not support "field width" and "precision"
> > arguments.
> 
> I wonder how much difference in code size the last two options would
> cause?
> 
The only thing the IAR docs say is that the small version is 10-15% 
of the size of the full version. Also, they default to the small 
version.


> > IAR uses linker switches to switch between versions.
> 
> Do people feel using linker switches would be more practical than
> selecting using a #define?
> 

Well, in the fact that the programmer is selecting which function to 
link into their application, this would seem to be a linking 
function, hence a linker switch. Ultimately though, either way can be 
just as easy, as long as it's documented somewhere.

Has something like this been done on other ports of gcc / libc? How 
did they do it? 

Eric