Hi,

since we have the "--enable-targets=all" flag now I wrote a little script 
(based on gdb_mbuild.sh) to make use of it for testing GDBs build.

In comparison to gdb_mbuild this script builds a single GDB with --enable-targets=all
flag (and --enable-64-bit-bfd if wanted). After a successfull build it tries
to recognize the architectures supported by GDB. For this it invokes "set architecture"
without a parameter and analyzes the output. The output looks like that:

<snip>
 powerpc:rs64iii
 powerpc:7400
 powerpc:e500
 powerpc:MPC8XX
 powerpc:750
 s390:31-bit
 s390:64-bit
 score
 sh
 sh2
 sh2e
 sh-dsp
 sh3
</snip>

It then sets the current architecure to each of thoses and stores the result from
"maint print architecture" command to appropriate log files (<target>.log).

In the first run it found some errors (like the one I mentioned in
http://sourceware.org/ml/gdb-patches/2008-01/msg00642.html).

The script is kept as simple as possible. There is no flag to change the verbosity
for example.

This is an example of the scripts output:

<snip>
... h1 OK
... mips OK
... mips:3000 OK
... mips:3900 ERR: gdb panic
... mips:4000 OK
... mips:4010 OK
... mips:4100 ERR: gdb panic
... mips:4111 OK
</snip>

The disadvantage of gdb_mbuild.h is the fact that it builds a complete GDB from source 
for each and every target which took quite a lot of time for a full build. For my opinion 
this shouldn't be needed anymore since we have "--enable-targets=all" flag now.

Would this script be ok to commit it to mainline?

I neither attached a ChangeLog nor a patch for documentation as this is just an RFC and I'd appreciate
your opinion on this very much. If this comes out to be useful for mainline I'll add ChangeLog and 
documentation.


Regards,
Markus


-- 
  Markus Deuling
  GNU Toolchain for Linux on Cell BE
  deuling@de.ibm.com