On Tuesday 22 September 2009, Øyvind Harboe wrote:
> I'm not keen on adding dependency on some wonderful
> bit vector utilities...

As opposed depending on some not-so-wonderful ones,
like we do today?  :)

Regardless of how things change, change is needed.
That's at the level of some basic *models* ... and I
think it's only sensible to notice systems that do
not have the flaws of the current interface.


> The first thing that I'd like to see fixed is for the jtag API
> to use 32 bit words instead of byte words. That would
> make the code less awkward.

I'd say "unsigned long" since that wouldn't require us
to pessimize 64-bit systems; and on 32-bit ones it'd be
the same as explicit spec of 32-bits.

Either way, that seems to be the necessary initial change.

Being able to change later to "one field of N bits" rather
than today's "M fields of ((N+31)/32) bits" will simplify
a bunch of code too ... but that can wait a bit.


> On the other hand, I just 
> can't work up the enthusiasm to throw that much testing
> overboard by making a sweeping change :-)
> 
> Probably better to do this earlier rather than later though...

Seems like not-for-0.3 ...

- Dave


_______________________________________________
Openocd-development mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/openocd-development

Reply via email to