Then there is the problem that compilation often involves rewriting memory (resolving BEGIN, DO .. LOOP references, etc). The MSP430 main memory is divided into 512 byte segments. And, each segment must be erased before writing (technically, you can write twice safely without an erase). What if the word definition spans two segments? I would need at least 512 bytes of RAM for buffering. Will I need more?
Another approach is to use a "tethered" development environment. With this, I would attach the PC uForth to the MSP430 uForth, compile the word locally and send the byte codes to the MSP430 to be flashed. This simplifies Flashing since there doesn't need to be a concern for rewriting bytes. However, this means that the PC and MSP430 uForths must be tightly bound.
SwiftX Forth takes a more formal tethered approach, but I like the more standalone nature of uForth right now.
The batch approach I take right now is to develop and test the word on the PC uForth, dump the dictionary and recompile the MSP430 Forth with this dictionary (flash based). The big problem with that can be summarized by the following example -->
: led-on bit0 port1 mem! ;
Ugh. There is no analogue for this on the PC uForth. My only recourse is to use a RAM based MSP430. But, after defining "port1", "mem!", etc. a 1KB RAM MSP430 part won't be big enough to support the dictionary. (And, obviously, using a Flash based MSP430 uForth, I won't be able to define led-on !)
Solution to follow... hopefully...