Now we get an error from downlit which we traced to this (premature?) optimization:
CRAN_urls <- function() {
# Substantially faster to use RStudio mirror: in my testing this reduced
# download time from ~2s to 0.6s
withr::local_envvar(R_CRAN_WEB = "https://cran.rstudio.com")
tools::CRAN_package_db()
}
This overwrites the R_CRAN_WEB variable (thus disabling our internal proxy) and then fails ungracefully.
Maybe as a compromise, CRAN_urls() should only change R_CRAN_WEB if it has the default values? Or maybe there's a way to completely avoid relying on the web when pkgdown.internet = FALSE is set?
Got a very similar error to #177 when trying to build a package website from behind a firewall. Until maybe 3-6 months this worked:
Now we get an error from downlit which we traced to this (premature?) optimization:
This overwrites the R_CRAN_WEB variable (thus disabling our internal proxy) and then fails ungracefully.
Maybe as a compromise,
CRAN_urls()
should only change R_CRAN_WEB if it has the default values? Or maybe there's a way to completely avoid relying on the web whenpkgdown.internet = FALSE
is set?Thanks!