[R-pkg-devel] Need advice CRAN package check results

[Spencer Graves]

On 4/30/22 12:58 PM, Duncan Murdoch wrote:
> On 30/04/2022 12:57 p.m., Matthias Gondan wrote:
>> Dear developers,
>>
>> I have recently submitted my first r package, and bear with me, I am 
>> still learning things. Most importantly, there is a page with the CRAN 
>> package checks,
>>
>> https://cran.rstudio.com/web/checks/check_results_rolog.html
>>
>> The package check for r-release-macos-arm64 fails, in two places.
>>
>> • The one in the table refers to Version 0.9, that one is less 
>> interesting because I have already fixed it. Well, minor question, 
>> will the version numbers for these tests increase automatically?
> 
> You manually change the version number of your package, and the version 
> number in the check table shows which version was tested.  New 
> submissions won't show up there until they are accepted.
> 
> Occasionally a new version could be accepted but fail to build on some 
> platforms; in that case the version number for that platform won't be 
> updated.  I think it's also possible for CRAN maintainers to manually 
> skip checks if they want.
> 
>> • More importantly, there are two links called M1mac below the table, 
>> „Additional issues“. They refer to the current version of the package, 
>> 0.9.2. The left link points to 
>> https://www.stats.ox.ac.uk/pub/bdr/M1mac/rolog.log, and there is an 
>> error message at 69% that needs to be fixed somehow. Note that the 
>> message is not really helpful (clang: error: unable to execute 
>> command: Abort trap: 6, clang: error: clang frontend command failed 
>> due to signal)
> 
> That looks like clang crashed.  If you can't duplicate it locally, it 
> will be hard to fix.
> 
> At the start of your log, there were these messages:
> 
> CMake Warning at cmake/port/Darwin.cmake:39 (message):
>    Could not find Macport or Homebrew to provide dependencies          
> trying to
>    configure with default search paths
> 
> I would try to test on a system that doesn't have Macport or Homebrew 
> installed.
> 
>> To investigate this in more detail, I tried the mac-builder service, 
>> but the build doesn’t fail there: 
>> https://mac.r-project.org/macbuilder/results/1651148567-352c2175cfbba55d/. 
>> I think this will be deleted soon, so this one is more recent, with 
>> only little changes: 
>> https://mac.r-project.org/macbuilder/results/1651332515-22f4e1b1c9492557/
>>
>> Does anyone have a good idea how to move on? I don’t have a Mac M1 at 
>> hand…
> 
> I think you'll need to find one to test on.  I'm not sure if Github 
> offers that platform for running Github actions, but I imagine there's 
> one somewhere or other.


	  One of my "R-CMD-check.yaml" files includes the following:


{os: macOS-latest, r: 'release'}


	  If you have this and it does not replicate your error, I don't know 
what other Mac options are available or if they would help.


TWO OTHER THOUGHTS:


1.  HOW DID YOU SUBMIT TO CRAN?  I know two ways.  You can submit 
directly via:


https://cran.r-project.org/submit.html


	  Alternatively, you can use "devtools::release()" as described in 
Chapter 20 of Wickham and Bryan, R Packages, available at:


https://r-pkgs.org


	  IF YOU SUBMITTED ONLY ONE OF THESE TWO WAYS, try the other ;-)


2.  DOCUMENT THIS PROBLEM IN a "cran-comments.md" file, as described in 
Wickham and Bryan.  Explain you can NOT replicate this problem.  With 
luck, if the package fails, a CRAN maintainer may look at that file and 
provide a more helpful suggestion -- or even fix the problem, if it's 
actually a problem with their platform.


	  Hope this helps.


	  Spencer Graves


p.s.  Jim Ramsay and I just went through some of this with the "fda" 
package.  We fixed that problem with help from Duncan Murdoch (THANKS, 
DUNCAN!!!).  We still have other problems, but ...

> 
> Duncan Murdoch
> 
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel