Re: [MiNT] binutils-2.13.2.1-2.m68kmint.rpm

[evan@coolrunningconcepts.com]

Quoting Patrice Mandin <mandin.patrice@wanadoo.fr>:

> mmap() is on kernel level, new binutils would not change this.

The discussion was about RPM.  I know how mmap() works.  Mmap() would 
be needed
to efficiently deal with very large RPMs.  Had nothing to do with binutils.

> However, more and more recent C++ software requires gcc/g++ 3 to be
> compiled, and currently (with my patches) it does not work.

I've got GCC 3.3.5 - using your patches.   Seems to work.

> If anyone is interested, the preliminary patches for binutils and gcc
> are mon my site, but considered non-working. I would like some help to
> fix that. binutils 2.15 for example does not create the translation
> table in the binary, so all relocated addresses still point to NULL.

Yes, binutils 2.15 does not work.  2.13.x seems to be last that does.  Libs
don't link right with newer ones - found that out via trial-n-error :(

> I know gcc 3 maybe more memory hungry and slower than current 2.95, but
> having recent tools is needed to port recent software, specially when
> cross-compiling.

It may also produce more efficient code, and while this isn't true or much
earlier GCC versions, 3.3.x should be doing well.  An extra 10 seconds 
during a
compile to save 1 second every time the program runs adds up if you run the
program often.

If you want fast compiles (like for development and testing), you might 
be able
to port the TCC compiler to generate 68K code.  TCC is blindingly fast 
- enough
to build a linux kernel from source on boot (using a modified boot-loader with
libtcc) or compile on demand (using libtcc).   Its an interesting project,
although for production code, I'd stick to GCC as it optimizes better and can
take advantage newer CPUs, while TCC is fairly generic in its optimization.

-- Evan