Re: [commit] Split d10v-tdep.c into trad-frame.[hc]

[Daniel Jacobowitz <drow@mvista.com>]

On Sun, Jun 08, 2003 at 12:15:31PM -0400, Andrew Cagney wrote:
> Hello,
> 
> This moves a chunk of d10v's frame unwind code into trad-frame.[hc]. It 
> creates a simple framework that traditional unwinders can use - a 
> successor to get_frame_saved_regs().
> 
> Handling the problem of the previous frame's PC being in this frame's LR 
> is remarkably simple - just copy the saved location of LR into PC!

I like it.  Just one complaint:

> +struct trad_frame
> +{
> +  /* If non-zero (and regnum >= 0), the stack address at which the
> +     register is saved.  By default, it is assumed that the register
> +     was not saved (addr == 0).  Remember, a LONGEST can always fit a
> +     CORE_ADDR.  */
> +  LONGEST addr;
> +  /* else, if regnum >=0 (and addr == 0), the REGNUM that contains
> +     this registers value.  By default, it is assumed that the
> +     registers are not moved (the register's value is still in that
> +     register and regnum == the index).  */
> +  int regnum;
> +  /* else, if regnum < 0, ADDR is the registers value.  */
> +};

Gyuh?  I looked at this for a couple of minutes and couldn't make heads
or tails of it until I went to look at trad-frame.c: this structure
isn't a frame at all, it's a single saved register.  Could you rename
it, and update the comments?

I'm also not sure I understand all the combinations:
  regnum < 0, addr is the register value
  regnum >= 0 and addr 0, regnum holds the new register number (in the
    next frame?)
  regnum >= 0 and addr non-zero, addr is the save address - but what's
    regnum mean?  Nothing?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer