[R-pkg-devel] checking CRAN incoming feasibility

Ivan Krylov |kry|ov @end|ng |rom d|@root@org
Tue Jan 16 08:58:58 CET 2024


В Tue, 16 Jan 2024 05:49:01 +0000
Rolf Turner <rolfturner using posteo.net> пишет:

> The problem is persistent/repeatable.  I don't believe that there is
> any faulty connection.

One of the things done by R CMD check --as-cran at this point is
sending a HEAD request to every Web link mentioned in the package
documentation and DESCRIPTION. One of the hosts may be slow to respond,
either by accident or due to misguided anti-robot countermeasures.
(Most website protection systems would say that R CMD check counts as a
robot because there's no human behind it to look at the ads.)

Here's what you could try. Unpack your built source package. If you
have a fresh .Rcheck directory from an R CMD check, use
YOURPACKAGE.Rcheck/00_pkg_src/YOURPACKAGE. Then profile the check
function, using the subdirectory from the source package archive as the
argument:

Rprof(); tools:::.check_package_CRAN_incoming(dir); Rprof('NULL')

Does any one function stand out in the subsequent summaryRprof()
output? For me, it's readRDS (not very helpful), but by reading
Rprof.out I can see that it's used by CRAN_package_db and
CRAN_archive_db to download web/packages/packages.rds and
src/contrib/Meta/archive.rds from the chosen CRAN mirror, which for me
takes a few seconds for both files.

Do you have a CRAN mirror set up in ~/.Rprofile? It could be having a
slow day.

-- 
Best regards,
Ivan



More information about the R-package-devel mailing list