Re: Building a mips toolchain for an AMD Alchemy processor and problems

[Dan Kegel <dank at kegel dot com>]

Ted Schroeder wrote:
> I'm trying to build a toolchain using the crossgcc scripts for an AMD 
> Alchmey processor. I'm configuring with gcc-3.4.2, glibc-2.3.3, 
> binutils-2.15 and a kernel based on 2.4.20 (some modifications to 
> support our specific hardware platform).
>
> When I begin running the script things go pretty well until I start 
> building the glibc-headers. At that point I get an error caused by 
> picking up the wordsize.h file from sysdeps/mips/bits directory. The 
> problem I get here is that instead of specifying a specific wordsize 
> __WORDSIZE is defined to _MIPS_SZPTR. Now, I presume, that this should 
> get #defined someplace, but it's not obvious how this is supposed to be 
> done.

gcc-3.4.2/gcc/config/mips/linux.h defines the macro
TARGET_OS_CPP_BUILTINS which, among other things. defines _MIPS_SZPTR.
http://gcc.gnu.org/onlinedocs/gcc-3.4.1/gccint/Run-time-Target.html

It wsn't too long ago that this macro was introduced; see e.g.
http://gcc.gnu.org/ml/gcc-patches/2002-07/msg00154.html
Maybe somebody botched it?

BTW I haven't been able to build a similr combination; see
http://kegel.com/crosstool/current/buildlogs/mips-gcc-3.4.2-glibc-2.3.3.sh.log.txt
for my failure log.  The build dies very early with
  cc1: error: invalid option `abi=32'
because I'm trying to build with the linux-2.6.8 kernel headers;
I should try the older kernel and/or the sanitized kernel headers;
then I'd be able to see the problem you ran into, probably.

> I "patched" this to just #define _MIPS_SZPTR to 32 in that same 
> file, but then I run into the next problem.
>
> When I'm building the main glibc I get errors in the memcpy.S file 
> assembly. The registers t4-t7 are not defined. This appears to be 
> because the _MIPS_SIM preprocessor symbol is not being defined to 
> MIPS_SIM_ABI32. 

That's starting to look like the problem described in
http://gcc.gnu.org/ml/java-patches/2003-q4/msg00042.html
Can you try his patch and see if it even still applies
(let alone fixes your problem :-)?

Good luck!

- Dan


--
Trying to get a job as a c++ developer?  See http://kegel.com/academy/getting-hired.html

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com