Fix examples that crash or hang up R session on CRAN machines

[IndrajeetPatil]

cf. https://cran.r-project.org/web/checks/check_results_lintr.html

[AshesITR]

I don't understand the logs, to me all output seems to be commented code (##D ...)
Do they --run-dontrun on r-devel under windows only?

[IndrajeetPatil]

Do they --run-dontrun on r-devel under windows only?

I am not sure. I reverse-inferred that they must be doing so based on the output. But maybe that's not the case?

---

That said, as I mention in #1776, \donttest examples are run in an additional check by BDR, and probably \dontrun examples are also run in some other checks.

Therefore, making sure all examples run successfully (or are never run if they are just for illustrative purposes) is a good preventive practice in general. This is currently not true for lintr-main branch, and is remedied in #1776.

[MichaelChirico]

I would say it's not likely the case. It seems that's just the particular example that's running when the machine times out -- this looks like a similar issue to that befalling {data.table}, see

Rdatatable/data.table#5507

[IndrajeetPatil]

Thanks. That was helpful to read, and I guess there isn't much we can do.

But I still think #1776 is worth considering for the reasons mentioned in the top-level comment in that PR.

[AshesITR]

So essentially, some test is too heavy for CRAN-Windows to digest?
Maybe we can gain some insight by profiling our tests to look for slow code...

[mattdowle]

No, not really. Here's a snapshot of lintr's results. It shows the same pattern as data.table; i.e. very large Tinstall times on Windows. In fact, yours is even more severe. Tinstall is just installing. It doesn't include tests at all. The trouble is Uwe is incommunicado so all I can do it guess that the CRAN Windows server is severely over laden with tasks, it could even be swapping. It would be helpful if you could contact CRAN -- maybe they or Uwe will reply to you. Perhaps they think it is a data.table-only problem despite my emails explaining that it isn't. Contact from other package maintainers might help.

[IndrajeetPatil]

It would be helpful if you could contact CRAN -- maybe they or Uwe will reply to you. Perhaps they think it is a data.table-only problem despite my emails explaining that it isn't. Contact from other package maintainers might help.

@MichaelChirico, @jimhester Would either of you have time and be willing to write to CRAN about this issue?