[R-pkg-devel] Recommended procedure for solving "Additional issues"?

[Uwe Ligges]

On 18.06.2018 16:38, Marius Hofert wrote:
> On Sun, Jun 17, 2018 at 2:13 AM Uwe Ligges
> <ligges using statistik.tu-dortmund.de> wrote:
>>
>>
>>
>> On 16.06.2018 21:46, Marius Hofert wrote:
>>> Dear DevelopeRs,
>>>
>>> Is there a recommended procedure for addressing "Additional issues" if
>>> they appear on CRAN checks?
>>>
>>> Say you have some C code in a package which checks fine (Status OK for
>>> all flavors) but shows "Additional issues" due to a potential memory
>>> problems. One can compile an R version configured with valgrind
>>> support and run a corresponding check on the package, but that
>>> a) ... sometimes shows false positives (on Mac, for example, as some
>>> have reported, see
>>> https://stackoverflow.com/questions/29312335/valgrind-on-os-x-yosemite-giving-bogus-errors)
>>> b) ... sometimes doesn't provide useful information (at least not in a
>>> case I recently ran into).
>>> c) Also, one can virtually not always check for all the "Additional
>>> issues" that might appear *before* submitting a package to CRAN -- or
>>> can one?
>>>
>>> Concerning c), it would be great if there was a way to 'see' the
>>> "Additional issues" *before* submitting to CRAN (like a winbuilder or
>>> so). Is there? (I couldn't find anything in WRE). Otherwise, one can
>>> do the usual (like --as-cran), hope everything is fine, wait until the
>>> CRAN checks are reported, and then, if there are problems, carefully
>>> look at the code again, try and find the problem and resubmit in the
>>> hope that the problem is fixed => but that's not a good 'procedure' as
>>> it creates more work for CRAN etc.
>>
>>
>> Well, you can try to run with valgrind and the sanitizers before you
>> submit to CRAN yourself.
> 
> Hi Uwe,
> 
> thanks for helping.
> 
> valgrind reported false positives, and 'ASAN' I still struggle with
> (setting the flags in ~/.R/Makevars as in WRE led me to some
> 'Interceptors not working' errors -- might post a question on this
> one, not sure if I understood that part correctly).
> 
>> Or use R-hub which offers to use  builds with
>> enabled sanitizers.
> 
> I tried that, thanks for this hint. I got some weird bioconductor
> errors (which seem to have nothing to do with my package). I contacted
> admin using r-hub.io but didn't get a response yet (don't think that's the
> right way to contact them, but found no other way...).

Not sure, Gabor Czardi is a main maintainer of that service, I am sure 
he will respond quickly.




>>
>> But you are right, running with valgrind + sanitizers + rchk (protection
>> stack imbalance checks) + different BLAS installations is not feasible
>> in case you do not suspect problems.
>>
>> That is also the reason why CRAN does not always detect these issues in
>> the incoming checks, as these checks run only on different specific
>> machines in batches.
> 
> Would of course be great if we could use exactly the same machines, so
> uploading the .tar.gz would reproduce the 'Additional issues' and one
> could try that until all issues are resolved (without the need for
> local tinkering) before officially submitting a package to CRAN.

Yes, but unfortunately that is not possible for the machines we are 
talking about.

Best,
Uwe


> 
> Cheers,
> M
> 
>>
>> Best,
>> Uwe Ligges
>>
>>
>>
>>>
>>> Am I missing anything?
>>>
>>> Thanks & cheers,
>>> Marius
>>>
>>> ______________________________________________
>>> R-package-devel using r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-package-devel
>>>