Re: [ECOS] Still confused by virtual vectors

[Grant Edwards <grante at visi dot com>]

On Thu, Oct 04, 2001 at 09:37:24PM +0100, Jonathan Larmour wrote:
> Grant Edwards wrote:
> > 
> > Some boards won't have RedBoot installed, and I'm assuming that
> > setting USE_ROM_MONITOR to false is the right way to run on
> > those boards.
> 
> Not needed. For a start GDB stub ROMs have had VV support since
> before RedBoot existed.

On two models of the boards I have neither RedBoot nor GDB stub
ROMs, and are running 1.3.1 (no VV stuff at all). There are
four or five models that are running a January 2001 snapshot
and do have RedBoot, but they're currently not using VV
features from RedBoot because their HAL was a quick port of the
1.3.1 HAL.

I want to transition both the older and newer products over to
a current snapshot using VV (services provided by RedBoot on
newer boards), but for the moment, I need my HAL to support all
three configurations.

> What this means is that you should implement
> CYGINT_HAL_VIRTUAL_VECTOR_SUPPORT_NOT_GUARANTEED instead if you
> want to run on both old and new boards.
>
> Look at hal.cdl for the effects.

I see. Here's my plan:

My HAL has a cdl option that controls implimentation of
CYGSEM_HAL_VIRTUAL_VECTOR_SUPPORT.  That controls whether VV
table is used at all.  That seems to work.

My HAL needs a second cdl option that controls implimentation
of CYGINT_HAL_VIRTUAL_VECTOR_SUPPORT_NOT_GUARANTEED, which (if
the above is enabled) controls whether the app uses the ROM
monitor (RedBoot) services or the app initializes the VV table
and uses it's own service routines.

Having to support three different configurations sounds a bit
odd, but the "C" in eCos stands for "configurable"!

Thanks again.

-- 
Grant Edwards
grante@visi.com