This is the mail archive of the
mailing list for the GDB project.
Question about rs6000- targets
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: gdb at sourceware dot org
- Date: Wed, 29 Nov 2006 01:56:36 +0100 (CET)
- Subject: Question about rs6000- targets
in continuing to clean up the TM files, I've gotten stuck in config/rs6000.
This directory currently supports three configurations (as host and target):
rs6000-*-lynxos* rs6000lynx.mh / rs6000lynx.mt
rs6000-*-aix4* aix4.mh / aix4.mt
rs6000-*-* rs6000.mh / rs6000.mt
Now in looking at these you will notice:
- The generic rs6000 configuration is 100% identical to the rs6000-*-aix4*
configuration (both as host and as target) -- the .mh and .mt files are
identical except for comments.
- Both of those are mostly identical to the powerpc/aix configuration.
There are a few changes, but they all appear to be recent bugfixes to
the powerpc/aix version that simply didn't get applied to the rs6000
This would appear to indicate that the rs6000 AIX configurations are
simply obsolete, and we should just support only powerpc/aix. (I guess
we could continue to support the target triples, but map them to the
same .mh / .mt files as powerpc/aix.)
Now, the RS/6000 LynxOS configuration also appears to be quite out of
date. In fact, I noticed just when testing my recent REALTIME_ change
that the configuration wouldn't even build.
Also, it does a number of annoyingly weird things in its TM file, in
particular it is the only configuration in the tree that uses coff-solib.h.
(B.t.w. it would appear that LynxOS has in fact switched from COFF to
ELF a long time ago ...)
How should such cases be handled? Can we declare rs6000-lynxos obsolete
in GDB 6.6 and remove it from mainline? Or do we need to continue to
attempt to support the configuration?
I'd appreciate suggestions how to proceed with this ...
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE