Re: [patch] Fix i386 memory-by-register access on amd64

["Ulrich Weigand" <uweigand@de.ibm.com>]

Jan Kratochvil wrote:

> Therefore fixed linux_xfer_partial.  Therefore it no longer needs the fix to
> be present also in inf_ptrace_xfer_partial.  I do not know how non-Linux OSes
> handle the debugging of 32bit inferior on 64bit GDB so I have no opinion
> whether the inf_ptrace_xfer_partial patch makes sense, i can drop it.

I think at this point it would probably be better to just to the Linux fix.
I've looked at a couple of ptrace implementations in the Linux kernel, and
it seems this fix should be correct on all bi-arch Linux platforms.

I simply do not know enough how ptrace behaves on other OSes -- until we
get a specific bug report on some other OS, I think we should leave the
behaviour unchanged for now.

> 2009-07-08  Jan Kratochvil  <jan.kratochvil@redhat.com>
> 
> 	Fix memory access from signed 32bit inferior registers on 64bit GDB.
> 	* inf-ptrace.c (inf_ptrace_xfer_partial <TARGET_OBJECT_MEMORY>): New
> 	variable addr_bit.  Mask OFFSET by the ADDR_BIT width.
> 	* linux-nat.c (linux_xfer_partial <TARGET_OBJECT_MEMORY>): Likewise.
> 
> 2009-07-08  Jan Kratochvil  <jan.kratochvil@redhat.com>
> 
> 	* gdb.arch/amd64-i386-address.exp, gdb.arch/amd64-i386-address.S: New.

The linux-nat.c change and the test case are OK.

Thanks,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com