Re: [avr-libc-dev] <ot> porting a generic libc to an arbitrary uC

[Joerg Wunsch]

As Steve Franks wrote:

> The ARM/lpc folks have a "newlib" implementation for their i/o, but
> no one appears to have implemented anything with stdio, ...

I'm a bit surprised about that.  I always thought newlib contained a
reasonable stdio implementation.  But then, newlib is basically a
half-heartedly assembled conglomerate of libraries that are not as
restricted as GPL is (as in an embedded world, true GPL is probably
not what people are looking for), picked from everywhere around, but
not really maintained as a whole.

> Do I need to make sweeping changes to the stdio system, or does the
> avr-libc implementation use the 'standard' bsd/sysV stdio.h code,
> except fdevopen?

fdevopen() uses malloc(), so you have to have a working malloc()
implementation.  Other than this, well, it has really been tailored
down for minimal resource consumption on avr-libc, but I hope I got it
well enough to be as portable as possible.

For a completely other route to go, I confess I don't know very much
about ARMs.  However, if I were you, I'd at the very least explore the
option to run either NetBSD or maybe even FreeBSD natively on them.
Failing that, have a look whether you could maybe use their libraries.
Unlike the AVR, the ARM is not as resource-constrained, so picking a
well-maintained library that has been written for a larger processor
might not be a too bad option at all.

-- 
cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL

http://www.sax.de/~joerg/                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)