[R-pkg-devel] NOTE checking for detritus in the temp directory on windows

[Henrik Bengtsson]

Hi.

> * checking for detritus in the temp directory ... NOTE
> Found the following files/directories:
> 'Rscript306c1f1adf1' 'Rscript3fb41f1adf1'

Whenever seeing detritus files with name Rscript<hexcode> like these,
it's a strong suggestion that parallel workers on MS Windows are
involved.  More precisely, each of them is from a separate PSOCK
cluster node worker typically launched by parallel::makeCluster().
When such workers are launched in example code, unit tests, or
vignettes those files are created by R itself when running on MS
Windows.  If the workers are not properly shutdown (e.g. forgetting to
call parallel::stopCluster()), those files are left behind, and 'R CMD
check' will detect them.

Now, from experience, but neither fully investigated not understood
yet, it *might* be that even if one calls parallel::stopCluster() at
the end of examples, unit tests, or vignettes, it might be that there
is still a race condition where the parallel workers are still in the
process of shutting down when 'R CMD check' is checking for detritus
files.  If that is the case, then it might be sporadic and tricky to
reproduce this problem.  This has happened to me exactly once. I'm
saying this, only in case you're struggle to reproduce it, but my
guess is that this is *not* the case here. Please see below for
options how to reproduce.

Now, I see you're using 'future' for parallelization.  Because of
this, I suspect you use plan(multisession), which launches parallel
PSOCK workers like parallel::makeCluster().  To shut down those
workers at the end, call plan(sequential).  My guess is that this is
your problem.  Before you do that, I would make sure you can reproduce
the problem first - see below.

> I did not manage to reproduce it with Rhub and github action.

1. Looking at your
https://github.com/comeetie/greed/blob/master/.github/workflows/R-CMD-check.yaml,
I see you're not checking with **R-devel** on MS Windows. I suggest
you add that and too see if you can reproduce it there.  You could
also add an explicit `env: _R_CHECK_THINGS_IN_TEMP_DIR_: true` just in
case (but I'd guess --as-cran does this).

2. Similarly, did you make sure to test with R-devel on MS Windows on R-hub?

3. Did you try with R-devel on the win-builder service?

For your future needs (pun not intended), I would make sure you can
reproduce the problem before trying to fix it.

Hope this helps,

Henrik

On Fri, Sep 30, 2022 at 12:14 AM <etienne.come using univ-eiffel.fr> wrote:
>
> Dear all,
>
> I'm getting the following Note for CRAN pre-tests (for package greed https://github.com/comeetie/greed) :
>
> * checking for detritus in the temp directory ... NOTE
> Found the following files/directories:
> 'Rscript306c1f1adf1' 'Rscript3fb41f1adf1'
>
> this note only appears on windows :
>
> * using R Under development (unstable) (2022-09-26 r82921 ucrt)
> * using platform: x86_64-w64-mingw32 (64-bit)
>
> The full pretest logs are located at https://win-builder.r-project.org/incoming_pretest/greed_0.6.1_20220927_163632/Windows/00check.log
>
> I did not manage to reproduce it with Rhub and github action.
>
> I've seen other questions on the mailing list about "detritus" but the root causes seems to be different in my case since i did not open a navigator and i did not create
> files or folders in my tests or vignettes. Any hint on how to solve this note will be welcome.
>
> Thanks a lot
>
> Etienne
>
>
> Etienne Côme, @comeetie
> Chargé de Recherche
> Université Gustave Eiffel
> GRETTIA/COSYS
> Tel : 01 81 66 87 18
> Web : http://www.comeetie.fr
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel