Garbage collection issues

[Enchufa2]

Initially, I reported this in tidyverse/magrittr#146 (see the reprex there, please). But after many attempts, I'm definitely unable to take Rcpp out of the equation, and I'm starting to think that the issue has something to do with it. I cross-post here so that the Rcpp gurus can take a look.

[eddelbuettel]

I will close this as there is absolutely nothing reproducible here.

And an reprex with a pipe is not minimal.

[Enchufa2]

I know, I know... but this is the best I managed to do so far. Do you know if it is possible that a reference to an environment is still visible to the gc but not to ls, and where?

[eddelbuettel]

You need to talk to the folks who deal in pipes, and standard vs non-standard evaluation. We "just" have C++ code here that very much guarantees you that an destructor is executed (but in our context maybe only once gc() happens). What we do is more straightforward. The other one is ... harder.

I would imagine you might be able to construct something without Rcpp exhibiting the same behaviour.