small fix to make sure an exception is raised in '_RcppCCTZ_getOffset…

[lsilvest]

…' when the specified timezone is not found

Otherwise is just falls back silently on UTC and in nanotime the user cannot know if for example there was a typo in the timezone he supplied.

[eddelbuettel]

Nice. My preferred idiom these days is Rcpp::stop() (which thows) over thow.

Shall we change that to

Rcpp::stop("Cannot retrieve timezone '%s'", tzstr);

which also lists the bad timezone thanks to the tinyfmt library we have in there.

[eddelbuettel]

Ok, I noticed that your master checkout was also listed as 'one commit ahead, one behind' (because my last commit was merging from you, one then needs a rebase in the repo the PR originated from).

So I rebased, and then applied one cleanup commit with Rcpp::stop() and a changelog entry. Have a look. Probably ready as is :) but I wait to hear from you.

(And yes, I first accidentally pushed it back to my repo in a new branch rather than yours. My bad. Also taken care of.)

[lsilvest]

Thanks! Will make sure to rebase next time!

The stop is fine in our case because that's what we end up doing in nanotime (actually to be precise we call forward_exception_to_r), but maybe the throw is a bit more flexible because this function is intended to be called at C++ level in another package and the other package might want to do something else than stop if the conversion fails, or might simply want a custom error message. I don't think it matters much though. We have a similar throw in the other two functions (_RcppCCTZ_convertToCivilSecond and _RcppCCTZ_convertToTimePoint) so we should probably align all three to either throw or stop...

[eddelbuettel]

Yes, I use them interchangeably too. There is, as I recall, a bit of refined work in the stop() to unwind the stack and whatnot. So if and when we have Rcpp.h included, it may be preferable. But it is, as you say, a small point. Merging now...