Re: Better MI memory commands

[Daniel Jacobowitz <dan@codesourcery.com>]

On Fri, Jul 09, 2010 at 10:54:07PM +0400, Vladimir Prus wrote:
> On Wednesday 07 July 2010 20:29:55 Daniel Jacobowitz wrote:
> 
> > > -           if (xfer == 0)
> > > -             return xfered;
> > > -           if (xfer < 0)
> > > -             {
> > > -               remaining = half;               
> > > -             }
> > > -           else
> > > -             {
> > > -               /* We have successfully read the first half.  So, the
> > > -                  error must be in the second half.  Adjust start and
> > > -                  remaining to point at the second half.  */
> > > -               xfered += xfer;
> > > -               start += xfer;
> > > -               remaining -= xfer;
> > > -             }
> > > -           half = remaining/2;
> > > +           xfree (buffer);
> > > +           xfered += to_read;
> > >           }
> > 
> > Why do we skip to_read bytes if we succeed at reading zero bytes?
> > For that matter, what does a return value of zero mean?  It seems like
> > this would mean the same as -1.
> 
> I am not really sure. Per documentation of target_read:
> 
>    Return the number of bytes actually transfered, or -1 if the
>    transfer is not supported or otherwise fails.  Return of a positive
>    value less than LEN indicates that no further transfer is possible.
> 
> So, value of 0 seems to mean 'there are no more bytes that that, honest', 
> and that we probably don't need to try further. Documentation for
> to_xfer_partial seem to give such meaning to return of 0.

I'd suggest treating 0 and -1 the same, for memory.

> On the
> other hand, it's not clear what return value of <LEN might mean,
> and whether we should try to read remaining chunk. What would you suggest?

A return of less than LEN from xfer_partial doesn't mean anything; you
just retry.  A return of less than LEN from target_read, though, is
supposed to mean that there is no point in retrying; the next byte is
inaccessible or does not exist.

It doesn't look like memory reads (unlike other partial transfers)
implement that; usually they just fail.  But we can treat it that way
anyway.  So <LEN means we got some number of bytes, and then we should
see what happens after those successful bytes.

-- 
Daniel Jacobowitz
CodeSourcery