In the TI MSPM0 SVDs, some registers use the non-standard `read-writeonce` access. For obvious reasons this is going to result in a parse error. Would there be any objections to an addition to just map this access type to `read-write` (whether to guard it behind a flag is a different question)?
Activity
burrbull commentedon Mar 15, 2025
read-writeonce=>read-writeOnceburrbull commentedon Mar 15, 2025
#277