Skip to content

[R-pkg-devel] rhub, docker and Bioconductor

6 messages · Gábor Csárdi, Christian Hennig, brian knaus +1 more

Original
Christian Hennig
# 
Hi there,

I'm trying to track down some errors and issues of my fpc package. As I can't reproduce any of these on my machine, I'm using rhub and docker.

I'm getting Bioconductor related issues with both

  local_check_linux("fpc_2.2-3.tar.gz", image = "rhub/debian-clang-devel")

and

rhub::check("fpc_2.2-3.tar.gz", platform = "debian-clang-devel")

namely locally:
source('https://bioconductor.org/biocLite.R')
Error: With R version 3.5 or greater, install Bioconductor packages using BiocManager; see https://bioconductor.org/install
Execution halted
Error in run(bash, c(file.path(wd, "rhub-linux.sh"), args), echo = TRUE,  :
  System command error

...and on rhub:

91<https://builder.r-hub.io/status/fpc_2.2-3.tar.gz-2492f95fc1c94fcbbff4d6f47cb423f8#L91>#> Error: package or namespace load failed for 'BiocManager':

92<https://builder.r-hub.io/status/fpc_2.2-3.tar.gz-2492f95fc1c94fcbbff4d6f47cb423f8#L92>#> .onAttach failed in attachNamespace() for 'BiocManager', details:

93<https://builder.r-hub.io/status/fpc_2.2-3.tar.gz-2492f95fc1c94fcbbff4d6f47cb423f8#L93>#> call: if (version[, 1:2] != version) return(sprintf("version '%s' must have two components, e.g., '3.7'",

94<https://builder.r-hub.io/status/fpc_2.2-3.tar.gz-2492f95fc1c94fcbbff4d6f47cb423f8#L94>#> error: argument is of length zero

95<https://builder.r-hub.io/status/fpc_2.2-3.tar.gz-2492f95fc1c94fcbbff4d6f47cb423f8#L95>#> Error: loading failed

96<https://builder.r-hub.io/status/fpc_2.2-3.tar.gz-2492f95fc1c94fcbbff4d6f47cb423f8#L96>#> Execution halted

97<https://builder.r-hub.io/status/fpc_2.2-3.tar.gz-2492f95fc1c94fcbbff4d6f47cb423f8#L97>#> ERROR: loading failed

(To be totally honest, I'm not even sure whether the latter rhub issue doesn't refer to the local docker thing as well, because this was linked from an R-hub builder PREPERROR email, it also mentions docker, and all the time indications don't make it unambiguously clear to what build they refer.)

Not sure what to do about this. Are there any other options apart from rhub/docker to get wiser about my errors and issues (error on debian-clang in a routine that shows nothing suspicious on my machine, additional issues  ATLAS<https://www.stats.ox.ac.uk/pub/bdr/Rblas/ATLAS/fpc.out> MKL<https://www.stats.ox.ac.uk/pub/bdr/Rblas/MKL/fpc.out> noLD<https://www.stats.ox.ac.uk/pub/bdr/noLD/fpc.out> OpenBLAS)?<https://www.stats.ox.ac.uk/pub/bdr/Rblas/OpenBLAS/fpc.out>

Yesterday I was able to check a prabclus noLD error on rhub, but no luck yet with fpc.

Thanks a lot,

Christian

--
Christian Hennig
Dipartimento di Scienze Statistiche "Paolo Fortunati",
Universita di Bologna, phone +39 05120 98163
christian.hennig at unibo.it<mailto:christian.hennig at unibo.it>
Christian Hennig
# 
I should maybe add that I'm not even sure why I need Bioconductor, the 
package has:

Imports: MASS, cluster, mclust, flexmix, prabclus, class, diptest, 
robustbase, kernlab, grDevices, graphics, methods, stats, utils, parallel
Suggests: tclust, pdfCluster, EMMIXskew, mvtnorm

Christian
On 1/7/20 2:54 PM, Christian Martin Hennig wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Gábor Csárdi
    

    

      
      #
    

  


  

      
It is an R-hub bug: https://github.com/r-hub/rhub/issues/318

Gabor

On Tue, Jan 7, 2020 at 2:05 PM Christian Martin Hennig

        
<christian.hennig at unibo.it> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Christian Hennig
    

    

      
      #
    

  


  

      
That's good to know. I'm still at a loss what to do though. How are you 
checking your packages on systems that are not available to you, other 
than using rhub?

Chances are if something from Bioconductor is required by one of the 
packages that I require, I can't get rid of it, right?

Thanks,

Christian

        
On 1/7/20 3:06 PM, G?bor Cs?rdi wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      brian knaus
    

    

      
      #
    

  


  

      
I was tracking down problems with my package yesterday that were specific
to r-devel and I had trouble reproducing the errors. I believe the issue is
that the rhub and the docker/rocker r-devel were not using the most recent
version of r-devel. My solution was to add r-devel to my travis ci.

https://towardsdatascience.com/travis-ci-for-r-advanced-guide-719cb2d9e0e5

This allowed me to reproduce the errors.

Brian

On Tue, Jan 7, 2020 at 6:35 AM Christian Martin Hennig <

        
christian.hennig at unibo.it> wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Ivan Krylov
    

    

      
      #
    

  


  

      
On Tue, 7 Jan 2020 13:54:38 +0000

        
Christian Martin Hennig <christian.hennig at unibo.it> wrote:

        

            

      
      
      
    

        
I cannot answer your original questions, but I can offer some advice on
the problems uncovered by package checks:

            

      
      
      
    

        
In R-devel, matrices are now also arrays [*]:

            

      
      
      
    

        
Use !inherits(convinv, 'try-error') instead.

            
        
These mostly seem to be concerned with the package startup message from
the mclust package. Try wrapping the calls to library(mclust) into
suppressPackageStartupMessages()?

As for the numbers being different between .Rout.save and actual test
results in ATLAS/MKL/noLD/OpenBLAS tests despite set.seed() being used,
I am not sure what could be done. RNGversion() might help, but this
could also be caused by differences in mclust version, in which case
the result might just be impossible to reproduce.