Hi,
I'd agree that in the code Rohit sent, it is perhaps not reasonable for
teh compiler to know that byte accesses are disallowed.
The casting-away-volatile issue is a red herring I think: notice that the
struct elements are themselves volatile.
What I'd like to call your attention to is that face that the msp430-gcc
currently used by TinyOS does the right thing with the test code,
producing this output at -Os:
getCtl0:
mov &0x01A0, r15
ret
Pehaps this is a symptom of a more serious regression?
John Regehr
