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.

[krlmlr]

Also, the last checks with this R version ran with

using R Under development (unstable) (2022-10-21 r83155)

Let's give it some more time.

[lorenzwalthert]

For your last point: Agree, all other checks have newer R versions.

[lorenzwalthert]

@krlmlr days passed and the NOTE still appears on one machine, which uses the same old 20222-10-21 R devel. Should we re-submit with all tests skipped? How would I do that? Just comment out the code in tests/test that.R?

[krlmlr]

Commenting out sounds good. The error is on CRAN's end, but we don't want styler to be pulled, which is likely to happen automatically if we don't act. And acting means resubmitting, e-mails are likely to be ignored.

[IndrajeetPatil]

Update is on CRAN! 🎉

[lorenzwalthert]

and all checks green. Let's hope it was not the tests that caused a problem but just a transient flavour problem.