Re: [MiNT] Gcc 3.3.6 v 4.0.1

[Alan Hourihane <alanh@fairlite.co.uk>]

On Mon, 2009-03-02 at 11:35 +0100, Vincent Rivière wrote:
> Petr Stehlik wrote:
> > Sounds like MiNTLib should become a patch to glibc - then tracking the
> > changes by GNU would be easier, perhaps :)
> 
> Yes, that is the definitive solution.
> But the glibc is huge, I heard it can be difficult to port it to 
> something other than Linux. The situation may be better now.

If we want to run this on 68000 machines with 4MB of RAM, then I think
glibc is not an option.

> However, there are alternatives to glibc, for example Newlib, uclibc, 
> dietlibc...

Yes. newlib I think is the best choice here.

> Newlib is a very good choice: it has been designed to be portable. It is 
> required to implement less than 10 "system calls" in order to get it 
> work on a new system. I even managed to let it work on plain TOS as a 
> replacement for the MiNTLib ;-) (be quiet: it was only a very incomplete 
> test).
> 
> However, Newlib has less functionality than glibc (for example, that 
> damn obstream stuff is not present). For information, Cygwin uses 
> Newlib, not glibc. The good news is that most makefiles of GNU software 
> are aware of Newlib, it can ease porting them.
> 
> Unfortunately, I guess the MiNTLib has been heavily patched from the 
> original sources, so it could be extremely difficult to port the added 
> functionalities to another libc. But I think it is the only way to get 
> the new functionalities and bugfixes from the community. I don't think 
> the MiNTLib could survive without this.

I agree we need to look to some integration with an existing project to
reap the benefits.

Alan.