Skip to content

[R-pkg-devel] CRAN submission with warnings

10 messages · Michael Dewey, Duncan Murdoch, Murray Efford +1 more

Original
Kent Johnson
# 
Hi,

What is the usual process to submit a new package to CRAN which has false-positive warnings?

I submitted rtree_0.2.0 last week (7/21). This package compiles with warnings from Boost headers. The warnings are expected, unavoidable, and unimportant, which I documented in cran-comments.md. After submitting the package, I got an email that the package failed the pre-test due to the warnings. I replied-all, pointing to the explanation in cran-comments.md. I haven't heard anything since, and the package is still in the archive folder at ftp://cran.r-project.org/incoming.

I'm new to this process. Is there something else I should do, or should I just sit tight and wait for a response?

win-builder results are here: https://win-builder.r-project.org/incoming_pretest/rtree_0.2.0_20210721_153225/Windows/00check.log
cran-comments.md here: https://github.com/akoyabio/rtree/blob/master/cran-comments.md

Thank you for any guidance,
Kent Johnson
Akoya Biosciences
Michael Dewey
# 
Dear Kent

I think the canonical way to report these is in the space on the 
submission form, unless I have not correctly grasped what you did.

Michael
On 26/07/2021 16:02, Kent Johnson wrote:

  
    
  


  


      

    

    

      
      

        


  

        

      Kent Johnson
    

    

      
      #
    

  


  

      
Thanks Michael! So I should re-submit the package with the documentation on WARNINGS in the comments section of the submission?

Kent

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Kent Johnson
    

    

      
      #
    

  


  

      
Actually I think the contents of cran-comments.md were included in the comments on the submission form. I used devtools::submit_cran() for the submission. It automates the form submission and IIUC copies cran-comments into the form. My comments are in the confirmation email I received.

Kent

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Duncan Murdoch
    

    

      
      #
    

  


  

      
On 26/07/2021 11:02 a.m., Kent Johnson wrote:

            

      
      
      
    

        
Generally I'd sit for a week or so, then send a gentle reminder.  It's 
summer, and people are on holiday, so you might make that 10 days.

Regarding the issues:

Since one of them comes from the BH package rather than yours, you might 
want to talk to Dirk for advice about it.

The second issue you mention in your comment looks more serious:
"Found 'abort', possibly from 'abort' (C), 'runtime' (Fortran) Found 
'exit', possibly from 'exit' (C), 'stop' (Fortran) Found 'printf', 
possibly from 'printf' (C)".

Your response that you don't call  'abort', 'exit' or 'printf' isn't 
sufficient here:  those functions appear to have been linked in to your 
DLL, so somebody is calling them.  I'd try to track down those calls and 
see if there's a way to compile without them.

Duncan Murdoch

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Kent Johnson
    

    

      
      #
    

  


  

      
Thanks, Duncan,

I did consult with Dirk about the warnings. The #pragma in Boost headers which disables those warnings is not allowed in a CRAN submission, so he removed the #pragma, enabling the warnings.

I see that Boost does call std::abort in its error handling. I don't think I can do anything about that...

Kent

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Murray Efford
    

    

      
      #
    

  


  

      
FWIW my package 'secr' has for some time had the same 'abort', 'exit' or 'printf' issue when checked locally on Windows. It is happily accepted by CRAN, and I have learned to ignore the local message. The only common element I can see between secr 4.4.5 and rtree 0.2.0 is that both import and link to Rcpp.
Murray

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Duncan Murdoch
    

    

      
      #
    

  


  

      
On 26/07/2021 3:34 p.m., Kent Johnson wrote:

            

      
      
      
    

        
If you look at the BH page on CRAN, there are a ton of "reverse linking 
to" packages, i.e. packages that are doing something like what you do. 
It would be worth looking at a few of those that don't generate those 
warnings (e.g. abcADM, ACEt, etc.) to see if they're using the same 
parts of Boost as you're using, and have found a way to avoid the warnings.

Duncan Murdoch

            

      
      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Kent Johnson
    

    

      
      #
    

  


  

      
The {anytime} package by Dirk Eddelbuettel uses {BH} and generates the same NOTE when I CMD Check it locally on Windows. It seems safe to ignore this NOTE.

Kent

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Duncan Murdoch
    

    

      
      #
    

  


  

      
On 26/07/2021 4:43 p.m., Kent Johnson wrote:

            

      
      
      
    

        
But it generates no messages on CRAN:  OK all the way.

Duncan Murdoch