NOTE detritus in the temp

[christophsax]

@eddelbuettel, I have a problem with seasonal on CRAN M1mac, where it complains about detritus in temp directory:

https://www.stats.ox.ac.uk/pub/bdr/M1mac/seasonal.out

I saw that I indeed write to tempdir() without cleaning up, but I cannot recreate the the check result, even on an M1 mac. I tried with _R_CHECK_THINGS_IN_TEMP_DIR_=true but no success.

Now I see that x13binary has the same issue:

https://www.stats.ox.ac.uk/pub/bdr/M1mac/x13binary.out

* checking for detritus in the temp directory ... NOTE
Found the following files/directories:
  ‘Rtmp41Q’ ‘RtmpqmD’

Here, indeed, we write to tempdir() without cleanup:

x13binary/R/checkX13binary.R

Line 25 in ba9858c

file.copy(system.file("testdata", "Testairline.spc", package="x13binary"), tdir)

This would be easy to fix, but would that help? Also, do you have any idea how to recreate the NOTE on any system?

I got a CRAN ultimatum for seasonal, so this is probably something we need to solve. Thanks a lot!

[krlmlr]

Yeah, I guess we need on.exit(unlink(file.path(tdir, x13.bin))) (and not forget to use add = TRUE in the subsequent on.exit() calls).

[eddelbuettel]

Or be even more explicit and assign a tempfile() at the start of example and simply remove at end which can always be hidden in a \dontshow{} segment.

Do you known if it comes from examples or tests? Have you tried r-devel with --as-cran?

[eddelbuettel]

I do not get it in seasonal under r-devel (but I don't have all optional packages installed so maybe not all code paths get opened). The Dept of Commerce website gets opened which is a bit ... weird.

Same in x13binary. Cannot reproduce locally (yet).

[eddelbuettel]

I propose to replace tdir <- tempdir() (which just gets us the temporary directory of the current session, which we cannot nuke) with dir.create(tdir <- tempfile()) which is what I use elsewhere. And then at end we remove that tdir recursively. PR to x13binary in a minute or two.

[eddelbuettel]

@christophsax if you review / approve the PR I can make a new x13binary release.

Which will show if the suggested fix works at CRAN....

[christophsax]

Super! I call x13binary::checkX13binary() on startup of seasonal, so every package that loads seasonal has this problem.

[eddelbuettel]

Yes, from a glance a the seasonal code it is not clear where it would create them -- so makes some sense that the call to x13binary did this by oversighgt. Oops. Should be better soon, hopefully.

[eddelbuettel]

Still waiting for the package to proceed through CRAN. Has been in state 'pretest' for the whole day -- maybe Uwe's machines are ill or something...

edd@rob:~$ cranIncoming.r x13binary        # wrapper for foghorn package function
   package  version cran_folder                time size
 x13binary 1.1.39.3     pretest 2021-03-24 14:23:00 8131
edd@rob:~$