> On Aug 28, 12:44pm, Andrew Cagney wrote:
> 
> 
>> > I agree, but we're talking about a simple bug-fix for an existing target, where the multi-arch-correct solution is not all that clear, because...
>> > 
>> > - re-arange tm-rs6000.h so that the define wasn't needed
>> > - move the define to nm-rs6000.h (sick but effective :-)
>> > 
>> > ...it's an MI part of the debugger that is testing the define.  How would you suggest modifying tm-rs6000.h to deal with this?
> 
>> 
>> Create an aix specific tm-rs6000.h, move the define to there and then 
>> kill the #undef's.  Given that this macro will be deleted rather than 
>> multi-arched, moving [burrying] it to nm-rs6000.h (and zapping the 
>> #undefine is easier).
> 
> 
> I agree that moving it to nm-rs6000.h is the way to go.  I'm not terribly
> worried about cross-debugging scenarios with AIX as the target since
> I'm pretty sure that it never worked anyway.
> 
> I just took a look at the places where IBM6000_TARGET are used.  This
> code is _really_ ugly.  E.g, anything vmap related in exec.c ought to
> be moved somewhere else, maybe xcoffsolib.c.

How's the attached?  It deletes the #undef's, and moves a renamed 
#define (DEPRECATED_IBM6000_TARGET) to the nm-rs6000.h file.  (If you've 
a better new name ... :-)

Eli, note that I tweaked the doco.

Andrew