[PicForth] Re: bug+workaround: code page switching

[David McNab]

Samuel Tardieu wrote:
> I have applied your fix (using no-cbank in begin, actually in backref)
> with the following comment:
> 
> \ In a backward reference, we cannot assume that the bank will be properly
> \ restored to its correct value. We could do that by remembering the
> \ value it should have, but this may be an underoptimization as we may
> \ need to explicitely call code in another bank just after the backward
> \ reference.
> 
> I'll be releasing a new PicForth version today with this fix and
> others, as I regard this one as being critical.

Great one Sam. Good also to see your presence again on this list.

A couple of points regarding this and future releases:

1) To make life easier for 16F8xxA users, you might like to consider 
applying the change to serial-monitor.fs I suggested earlier - having 
two vector blocks of 4 words at start, instead of 2 blocks of 3 words.
Otherwise, anyone implementing a host bootloading client has to handle 
granularity issues (to address the requirement of flashing program words 
in blocks of 4).

2) I installed 'tla', and had a look at the GNU Arch archive for 
PicForth (as per details on PicForth website), but couldn't see 
picforth-dev. All I found there was picforth-release, picforth-0.31.
Can you please tell me where 'picforth-dev' lives? If you're willing to 
grant me checkin access, I'd be able to contribute code snippets - my 
intention would be to contribute new files only, and not touch the 
existing codebase.

> Thanks again.

And thanks of course to you too, Sam. The more time I spend using 
PicForth, the more I'm appreciating its elegance, power, practicality 
and sanity.

-- 
Cheers
David