I still wish there was a patch that combined both dtrace and systemtap support, und used as much code sharing between them as feasible. I'd be +1 on such a patch, and -0 on two separate patches that achieve the same functionality, but on different code paths.

For example, the systemtap version has a helper function get_frame_marker_info that covers more cases than the dtrace version; OTOH, the dtrace version has more trace points.

If consensus on functionality is not easily achieved, I propose to have the intersection on functionality first, i.e. only use the function-entry/exit trace points even in the dtrace version. Or else you agree on what trace points both systems ought to provide.