this makes the core MAVLink functions into library calls, instead of
being inlined. The resulting code size doesn't change, but it means we
can safely call MAVLink functions from multiple places without causing
undue code bloat
this makes mavlink_message_t much smaller, but setting
MAVLINK_MAX_PAYLOAD_LEN to the maximum message size in this
dialect. This saves us 320 bytes of memory
git-svn-id: https://arducopter.googlecode.com/svn/trunk@3239 f9c3cf11-9bcb-44bc-f272-b75c42450872
this new implementation reduces code size, and also reduces stack
usage, while avoiding the gcc union stack bug
Note that we will gain even more when we move to the new protocol
version, especially in terms of code size
git-svn-id: https://arducopter.googlecode.com/svn/trunk@3200 f9c3cf11-9bcb-44bc-f272-b75c42450872