Re: [avr-libc-dev] Re: [avr-gcc-list] c library with stdio?

[Joerg Wunsch]

As E. Weddington wrote:

> I agree that printf() is too impractical to do.

It's not completely out of the question, but it needs a good design
so it could be useful, and applications could stuff their actual
IO functions (for UART, LCD, whatever) into it.

> Any ideas about implementing at least sprintf() / snprintf(), with 
> caveats about the amount of space it will take up (i.e. use it only 
> on the bigger AVRs)? 

See my mail to the avr-gcc list, there used to be some sample
implementation shipping with Marek's early avr-libc releases.
Maybe this could be a starting point (although the copyright looks
fairly messy ;-), it has both a BSD and a GPL copyright header).

> If not sprintf(), then how about an ftoa() to fill in the gaps (since 
> there's already itoa(), utoa(), ltoa(), etc.)?

There's a dtostre() and dtostrf() already.  However, i've been using
dtostre() in one of my applications, and it sometimes emits `funny'
results (like colons in places where a digit should normally appear),
and i even accuse it of totally crashing my program for some
out-of-bound input (like a NaN maybe).  So i wouldn't consider this
to be really stable yet...  Haven't found the time so far to isolate
the cases with the funny results so i could debug them.  That's
probably something that could be better done with the simulator
though.

> If not both than how about a FAQ as to why it's not implemented since 
> this question keeps popping up. I've seen this question (about having 
> printf()) also on the forums at avrfreaks.net.

Would you mind writing up a paragraph or two?

-- 
J"org Wunsch                                           Unix support engineer
address@hidden        http://www.interface-systems.de/~j/