I'm happy to say that I have incorporated etracer's mods to [u]NewSoftSerial[/u] and after running the Windows-compiled object through my baud rate test suite, I find that the slightly longer RX interrupt handlers do not significantly affect the error rate.
ERROR RATES
| 300 | 1200 | 2400 | 4800 | 9600 | 14.4K | 19.2K | 28.8K | 31.25K | 38.4K | 57.6K | 115.2K
TX | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0%
RX | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0% | 0.1% | 0.35% | N/A
@mellis: I would love to be able to make etracer's fix conditional on whether it is needed. This seems an excellent example of a use case that would benefit from the
#define ARDUINO_VERSION 14
proposal I made.
OSX users: Please test. And would someone please compile NSS 5 on OSX for Diecimila/Duemilanove and send me the binary? I would like to run it through my test bench.
Mikal