Re: mipsbe solib event breakpoint endian?

[Daniel Jacobowitz <drow@mvista.com>]

On Thu, Jul 10, 2003 at 11:06:01AM -0400, Kris Warkentin wrote:
> Finally got around to chasing this again and I've found the problem.  I'm
> hoping someone might be able to explain why I'm seeing what I'm seeing.
> 
> GDB-5.2.1 targetting QNX MipsBE:
> Breakpoint 4, enable_break () at ../../gdb/solib-svr4.c:939
> 939           if (!load_addr_found)
> (top) n
> 940             load_addr = read_pc () - tmp_bfd->start_address;
> (top) p read_pc()
> $1 = 1882417036
> (top) p /x read_pc()
> $2 = 0x7033678c
> (top)
> 
> 
> GDB-HEAD targetting QNX MipsBE:
> Breakpoint 3, enable_break () at ../../gdb/solib-svr4.c:1039
> 1039            load_addr = read_pc () - tmp_bfd->start_address;
> (top) p read_pc()
> $1 = 8084919611063009279
> (top) p /x read_pc()
> $2 = 0x7033678ceeffffff
> (top)
> 
> Looks like someone thinks the program counter is a 64 bit value....  Any
> hints?

The CORE_ADDR type _is_ a 64-bit value, and MIPS addresses do sign
extend; this has caused all sorts of interesting problems, if you check
the list archives.  Something has read it in endian-swapped, however. 
I'd investigate what; maybe REGISTER_BYTE fallout?

I haven't seen this happen on my MIPS targets either.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer