Re: [Xc3sprog-users] bscan 59a6 bug

[joosteto <joo...@gm...>]

    joosteto> So my question is: * am I right in thinking that the magic
    joosteto> should be two different words, not twice the same

> I only think that this will shift the problem to some other sequence.

Actually, no, it will not.

I just verified  by using the magic x"59a659a3", and by inserting that
magic in my 59a6.bit file at 4 different offsets (mod 4).

The reason is that bscan only compares the top 48 bits of its header signal
with 59a659a3 when have_header is 0, thus after
  CAPTURE || RESET || UPDATE || !SEL1
becomes true. This only happens when the SPI sender is ready sending
its data (I presume, as mentioned earlier, I cannot find where the .cpp code
does this).

The problem with using twice the same word as magic (59a659a6)
is that have_header is already '1' when the first word of magic is read,
so that word can be combined with the last word of data.

When using something like 59a659a3 (two different words of magic),
this condition is not reached.


On 6 June 2013 17:51, Uwe Bonnes <bo...@el...>wrote:

> >>>>> "joosteto" == joosteto  <joo...@gm...> writes:
>
>
>     joosteto> So my question is: * am I right in thinking that the magic
>     joosteto> should be two different words, not twice the same
>
> I only think that this will shift the problem to some other sequence.
>
> Benedikt Heinz, thr original author of of bscan_spi admits the problem and
> proposed a better high-level SPI-Over-Jtag protocoll. Neither he nor me
> have
> time to do so at the moment.
>
> Bye
>
> --
> Uwe Bonnes                bo...@el...
>
> Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
> --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------
>