This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [rfa, v3] Fix inferior calls during interrupted system calls on PowerPC
On Sat, May 03, 2008 at 05:31:35PM +0200, Ulrich Weigand wrote:
> - powerpc-32.xml and powerpc-64.xml used to provide AltiVec registers,
> and now they don't. To compensate for that, I've switched the generic
> "powerpc" and "powerpc64" targets in rs6000.c:variants back to use
> the new AltiVec target descriptions.
>
> However, I'm wondering why those were using AltiVec in the first place;
> shouldn't the generic description be compatible with old gdbserver
> implementations (that do not use AltiVec)?
This is just how GDB worked before my XML conversion, so I left it
alone. The "powerpc" variant in rs6000-tdep.c included AltiVec and I
wanted to leave it that way. It's quite possible we could clean it up
now.
> - In the SPE register set for gdbserver, register numbers 71 and 72 were
> already used for "acc" and "spefscr". As it is much simpler to always
> use the same numbers for "orig_r3" and "trap", I've moved those SPE
> registers to 73 and 74. I *think* this shouldn't break anything as
> those numbers were only ever used with XML target descriptions ...
Yes, this is OK. Giving them 73 and 74 should work too - the script
will generate "0:" as appropriate to make things line up.
> - Should we attempt to detect SPE registers in core files?
I don't know how SPE registers are dumped, offhand. We can fix this
up later.
The patch looks OK to me.
--
Daniel Jacobowitz
CodeSourcery