On Wed, Jun 04, 2003 at 04:16:48PM +0200, Maciej W. Rozycki wrote:
> On Wed, 4 Jun 2003, Ralf Baechle wrote:
>
> > > But the need is to raise an exception after instr1 (at addr1) is executed.
> > > One solution is using a break at instr2 (at addr2).
> > > But suppose instr1 is a jmp then there is no point
> > > in keeping a break at addr2.
> > > (inorder to raise an exception after instr1 is executed).
> >
> > You understood correctly. Now jumps and even more so the conditional
> > branches are sort of the ugly part of the whole thing. The easiest
> > method is probably inserting a branch at the jump's destination address
> > or in case of a branch at the branch target and the instruction following
> > it's delay slot. So that's a lot of inserting and removing of
> > breakpoints ...
>
> In a more finegrained but also more complicated example, you probably
> want to insert a breakpoint in the delay slot first and at the second step
> evaluate the branch's condition and put a breakpoint at the next
> instruction to be executed. I'm not sure if the current version of gdb
> does the first step, but it inserts a single breakpoint in the second one
> only. For branch likely instructions adjust the two steps as necessary.
Does that actually work reliably across MIPS processors? I don't
believe that it will. I suppose you could re-execute the branch to get
the delay slot executed...
GDB simply executes the branch and its delay slot as a unit.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer
|