CRAN re-submission

[lorenzwalthert]

Dear maintainer,

Please see the problems shown on
https://cran.r-project.org/web/checks/check_results_styler.html.

Please correct before 2022-11-09 to safely retain your package on CRAN.

The CRAN Team

Seems like only occurring on R devel with fedora flavor. Also, it's not clear if the problem was already present for version 1.7.0, as the below screenshot has no data for 1.7.0 anymore. @IndrajeetPatil did you check the CRAN results page before 1.8.0 wsa released?

[lorenzwalthert]

This seems like a needle in the heystack thing to find -.-
Potential sources:

• bug in R devel in 202 (for debian only, file starts with Rscript and has extension): https://stat.ethz.ch/pipermail/r-package-devel/2020q4/006161.html
• opening web browser (file names seem to not to start with Rscript): https://stackoverflow.com/questions/62456137/r-cran-check-detritus-in-temp-directory
• parallel testing (issue on Windows only it seems and no file extension, but we only started enabling parallel testing with this releaes on CRAN): https://stackoverflow.com/questions/72677239/how-do-i-solve-this-two-notes-from-r-package-auto-check (with links to https://stat.ethz.ch/pipermail/r-package-devel/2022q3/008527.html on how to R cmd check with _R_CHECK_THINGS_IN_TEMP_DIR_).

[lorenzwalthert]

I am trying to reproduce the problem with {rhub}:

• with rhub::check(platforms = 'fedora-gcc-devel') without --as-cran: https://builder.r-hub.io/status/styler_1.8.0.9000.tar.gz-d6a944f58f3940c9a4e8fd2a312837d5
  *with rhub::check(platforms = 'fedora-gcc-devel', check_args = '--as-cran') that does the temp directory check but finds no problems: https://builder.r-hub.io/status/styler_1.8.0.9000.tar.gz-d91cf7c05eee41bc8f4e3b0c91369435
• with rhub::check(platforms = 'fedora-clang-devel', check_args = '--as-cran'): https://builder.r-hub.io/status/styler_1.8.0.9000.tar.gz-eca3f6c096f2423184aae90b8ea10d33

[lorenzwalthert]

@krlmlr I guess here we might need your help.

[krlmlr]

Do I understand correctly that none of these checks have shown the error?

Is there a way to find out when this error started to occur? A hint in the most recent R-devel changes? Unfortunately, the Wayback Machine is of no help here, they don't archive CRAN check results.

[IndrajeetPatil]

@IndrajeetPatil did you check the CRAN results page before 1.8.0 wsa released

IIRC, yes, I did. But there was only the NOTE about is_shebang.Rd file; nothing about detritus in the temp directory.

Is the problem that we are leaving the cache directory behind, or something similar?

Btw, we error on a NOTE in R CMD check run on GHA, and the fact that it's still not showing errors makes me think that this is a change that was very recently introduced in R-devel, and we may start seeing this failure on GHA soon.

styler/.github/workflows/check-full.yaml

Line 59 in 4c1cb51

error-on: '"note"'

[lorenzwalthert]

Do I understand correctly that none of these checks have shown the error?

yes.

Wayback Machine is of no help here, they don't archive CRAN check results.

yes, tried that also.

[lorenzwalthert]

Is the problem that we are leaving the cache directory behind, or something similar?

I don’t think so, these problems looked differently: #836

[krlmlr]

These machines are fairly special:

• https://www.stats.ox.ac.uk/pub/bdr/Rconfig/r-devel-linux-x86_64-fedora-gcc
• https://www.stats.ox.ac.uk/pub/bdr/Rconfig/r-devel-linux-x86_64-fedora-clang

Maybe we can even replicate the problems locally?

[lorenzwalthert]

Mirracles happen. r-devel-linux-x86_64-debian-gcc no longer shows a note. 😮. Maybe it was just a bug in R devel. Let's see if the other note also disappears.

[lorenzwalthert]

@krlmlr do you think we should to write on r devel? I have never done that before. 7 days till removal.

[krlmlr]

Let's wait for five more days, and then push an update to CRAN if the failures persist. This will buy us a bit more time.

[lorenzwalthert]

Ok, and what update should we push if we don't even know what the root cause is?

[krlmlr]

We could disable all tests to see if problems still occur. Then, we at least know if it's the tests or the examples. Another thing to try would be to run R CMD check on Linux devel with the flags described in https://www.stats.ox.ac.uk/pub/bdr/Rconfig/r-devel-linux-x86_64-fedora-gcc . It puzzles me that _R_CHECK_THINGS_IN_TEMP_DIR_ is set to "true" both here and in the other custom Fedora setting, but the other Fedora setting now works. I really hope this is a transient thing.