Skip to content

[R-pkg-devel] How to debug CRAN errors?

11 messages · David Hugh-Jones, William Dunlap, Klaus Schliep +2 more

#
Hi all,

My package has errors on CRAN's Linux and Solaris:

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

which I can't reproduce on my local OSX Machine, nor on Linux on Travis.

Does anyone have any general hints on how to reproduce and/or debug such
errors?

Specifically the error relates to a call to openxlsx::saveWorkbook
producing a message. openxlsx hasn't changed recently, though.

Cheers,
David
#
The complaint
-- 1. Failure: Data written in appropriate format
(@test-openxlsx.R#101) ------
     `openxlsx::saveWorkbook(wb, file = "test-xlsx.xlsx", overwrite =
TRUE)` produced messages.

comes from your call to testthat::expect_silent()

test_that("Data written in appropriate format", {
  hx <- huxtable(a = 1:2 + 0.5, b = -1:-2 + 0.5, d = letters[1:2],
add_colnames = TRUE)
  wb <- as_Workbook(hx)
  expect_silent(openxlsx::saveWorkbook(wb, file = "test-xlsx.xlsx",
overwrite = TRUE))


Perhaps you should suggest to the authors of testthat that it would be nice
if expect_silent() showed some of the text of the messages, etc., instead
of just saying that messages were produced.

By the way, I think your test test should write to a file in [a
subdirectory of]] tempdir(), not to a file in the current directory.

comes from
Bill Dunlap
TIBCO Software
wdunlap tibco.com


On Wed, Mar 13, 2019 at 8:50 AM David Hugh-Jones <davidhughjones at gmail.com>
wrote:

  
  
#
Hi Bill,

I take your point. Meanwhile, though, is there any way to debug the
problem? I?ll assume that making repeated uploads to CRAN is not a viable
approach.... it would be great if there were a Docker image of their setup
available, for example.
On Wed, 13 Mar 2019 at 17:00, William Dunlap <wdunlap at tibco.com> wrote:

            
#
On Wed, Mar 13, 2019, 1:13 PM David Hugh-Jones <davidhughjones at gmail.com>
wrote:
Have you tested your package on https://builder.r-hub.io/ or
https://win-builder.r-project.org/ ?

  
  
#
You can stick your own version of expect_silent into the test file.  Have
it print
the messages, output, and warnings that it finds.

Bill Dunlap
TIBCO Software
wdunlap tibco.com


On Wed, Mar 13, 2019 at 10:13 AM David Hugh-Jones <davidhughjones at gmail.com>
wrote:

  
  
#
On 13/03/2019 12:59 p.m., William Dunlap wrote:
Yes.
That could be the source of the error:  it's a violation of CRAN policy 
to write to the current directory.  An obvious way to test this would be 
to make it unwriteable.

Duncan Murdoch
#
Hi guys

That might be it. But there was no error until the latest release - whereas
that test has been there for 9 months. The file is written within
testthat... is there a new check for this in CRAN?

Meanwhile, I am still figuring out how to debug this problem. I uploaded to
r-hub... does anyone know how I can access the artifacts created after the
build? The URL build was

https://builder.r-hub.io/status/huxtable_4.4.0.tar.gz-b91485ad24b7a2f9c73bad2d51ebdb9d

David

  
  
#
On 13/03/2019 8:34 p.m., David Hugh-Jones wrote:
You should have got an email giving the link.  In the ones I've seen, 
you replace the prefix

https://builder.r-hub.io/status/

with

https://artifacts.r-hub.io/

and you'll see them.  It works for yours.

Duncan Murdoch
#
Thank you! Got it. This should be close enough to CRAN, and there?s a
Docker image available too.

On Thu, 14 Mar 2019 at 00:40, Duncan Murdoch <murdoch.duncan at gmail.com>
wrote:
Sent from Gmail Mobile
#
To follow up on the great answers regarding R-hub,- to use R-hub Linux Docker images you can use this brand-new rhub function?https://r-hub.github.io/rhub/reference/local_check_linux.html (untested on Windows)- here's WIP R-hub docs entry about choosing a platform to reproduce an error?https://docs.r-hub.io/#rhub-cran-platforms
Ma?lle.




    Den torsdag 14 mars 2019 08:48:34 CET, David Hugh-Jones <davidhughjones at gmail.com> skrev:  
 
 Thank you! Got it. This should be close enough to CRAN, and there?s a
Docker image available too.

On Thu, 14 Mar 2019 at 00:40, Duncan Murdoch <murdoch.duncan at gmail.com>
wrote:
Sent from Gmail Mobile

??? [[alternative HTML version deleted]]

______________________________________________
R-package-devel at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
1 day later
#
Thanks to Ma?lle for getting me fixed up with r-hub, which is looking quite
slick.

I debugged the problem and it comes down to the R tinytex package not
playing nicely with Debian?s texlive installation.

I  think the answer is just to turn off all relevant tests on CRAN. It?s
reasonable for them to use texlive, and reasonable for Yihui to not want to
support it; meanwhile Travis etc still run the tests.

Thanks for your help, everyone!

David
On Thu, 14 Mar 2019 at 12:50, Ma?lle SALMON <maelle.salmon at yahoo.se> wrote: