RTEMS vs. egcs-1.1a

[ccj at acm.org (Chris Johns)]

Ralf Corsepius wrote:
> 
> Eric Norum wrote:
> 
> > My attempts at building an m68k-rtems egcs-1.1 haven't been very
> > successful.
> 
> Well, I have been able to build egcs-1.1 for sh-rtems and h8300hms under linux
> (using a native egcs-1.1 on the host) more or less successfully.
> 
> Some problems I noticed still to be present:
> * The limits.h problem (I still couldn't figure out how to fix it. The way Joel
> suggests doesn't work for me)
> * Joel's bit scripts don't work.

It worked for me. I got the lastest from the beta3/c_tools directory,
modified the user.cfg and off they went.

> * Symlinking one-tree source-trees apparently doesn't seem work.
> 

The bit script uses links and it seemed to build. I have not built
anything with the m68k compiler yet.

> FYI: The only reliable way to build the gnu-toolchain seems to be  the way the
> egcs' docs mention (Don't ask me where :-):
> * Build a one-tree source-tree by physically copying newlib/* and [binutils|gas]/*
> into the egcs source tree (The sequence of copying is important: If I recall
> correctly, I used : 1. newlib, 2. egcs, 3. gas ([gas|binutils] seem to contain the
> latest libiberty, configure scripts etc.)).
> * configure
> * make
> 
> Just for curiosity I tried to build the tool-chain for m68k-rtems (including objc
> and f):
> 

This is what I am building.

> Unified source tree consisting of:
> * newlib-1.8.0 + rtems-newlib-patches
> * gas-980825
> * egcs-1.1a
> 

I used egcs-1.1b (not that a to b is a real change), binutils-2.9.1,
newlib-1.8.0 and Joel's patch.

> 
> Now, my guess is that your host's binutils and gcc might be too old/broken and/or
> that your target's binutils are too old (At least on ix86 egcs is known to require
> very recent versions of binutils).
> 

I used egcs-1.1b and binutils-2.9.1 on Linux to build the tools with.

Regards