shielding #789 behind a new #define to effectively disable it now

[eddelbuettel]

No description provided.

[kevinushey]

Can't we just have Rcpp_eval_impl() always use Rf_eval() for now, rather than sprinkling the #ifdefs around the associated usage? I think that would imply just ensuring that RCPP_USE_PROTECT_UNWIND is never defined by us?

[kevinushey]

(I think this looks fine but I think we can accomplish the same thing with a bit less code)

[eddelbuettel]

Sure, we could. So far I figured @lionel- 's PR was somewhat limited so I could just leave it in place, "protect" it (or us ;-) ) from it and give him a chance to revisit in time.

These ifdef will not stay forever. Either we can make this work, or I'll remove them in, say, six month. At least now I know where to find the code as the #ifdef provide a clear demarcation.

[lionel-]

What I don't understand is that Rcpp_eval_impl() was defined as a simple wrapper to Rf_eval() unless RCPP_PROTECTED_EVAL was defined..

[eddelbuettel]

See the diff. Your PR affected other areas, and things generally turned sour ... on systems other than Linux. I spent two days cleaning this and am not too interested in discussions of the "in theory this should have worked" type because I have ample evidence that "in practice" this didn't. Part of it may be R. We don't quite know.

But now is neither the time to take chance, nor for experimentation. Let's get a working 0,12,15 out the door.