[Rcpp-devel] Rcpp::exception + threads = disaster

[Joshua N Pritikin]

On Sat, Feb 08, 2020 at 03:28:22PM -0500, JJ Allaire wrote:
>    Agreed that it would be good to have more clear docs here. Note 
>    that as Dirk pointed out both Writing R Extensions and 
>    RcppParallel docs are pretty clear about the fact that you 
>    shouldn't call any R APIs when in a background thread.

Rcpp::stop superficially looks like a nice way to throw C++ exceptions 
that also does printf style message formatting. I didn't realize that 
it calls R APIs until I looked at the implementation.

>    I think the point is that the *only* reason Rcpp::stop exists is 
>    to do forwarding to Rf_error. If that isn't your intention it's 
>    strictly worse than a normal C++ exception.

Sure, but does it *have* to be that way? It seems to me that
exceptions.h line 40,

  rcpp_set_stack_trace(Shield<SEXP>(stack_trace()))

could just as easily be done around macros/macros.h line 51.