Help with CRAN 'Additional Issues': No Long Double

Package development
1

Been trying to figure out the solution to this today without much success so I figured I'd ask here.

Got an email today from Brian Ripley about an "Additional Issue" on my lakemorpho package. In the email he provided some details from Writiing R Extensions, specifically the last bullet of Section 1.6. I've read that many times and am having trouble interpreting what the issue is exactly.

Also, I checked the CRAN Package Results page which lists the additional issues. All OS flavors have an OK status except for the additional info section which lists the noLD issue. The output for this indicates one of my examples is throwing an error.

So it appears I could solve this for CRAN by wrapping my examples is a \dontrun{}. But I am more interested in knowing what the underlying issue is that caused the problem. Also why did it just get flagged after nearly a year on CRAN?

Thanks!
Jeff

2

recipes was included on that email too so I'm looking into it.

In my case, there are tests that fail on on x86_64 Linux with R-devel configured --without-long-double that pass otherwise. I assume that this is related to the default tolerances used in all.equal but, I won't know until I test...

I'm working on (my first) docker build that will use --without-long-double to test but that probably won't happen until after the holiday.

1 Like
3

Thanks for the reply, @Max. I'll keep an eye out here for what you find out. The testing issues make sense. Odd thing on my package is that it doesn't have tests. I may just wrap my examples in \dontrun{} for now and hope for the best!

4

Yours is odd in that it seems like an issue with classes. It would be pretty weird for that to be affected by how doubles are stored (but maybe)...

It looks like maptools was updated since yours went to CRAN. Maybe a newer version of that package broke yours?

5

Possibly. Bare minimum I can remove maptools. I had a built in check for polygon holes, but that seems like something I can offload to the user to make sure they are passing clean polygons.

Thanks for your suggestions!

6

I also received this email for a build error in my package drtmle. Were you two able to resolve this issue?

I can't tell from the error why things are failing and I have no way to configure R devel on the relevant Linux system to debug. Not sure what my best option is going forward.