(my reply to reviewer, cran-submissions, cran re: 2019-05-16_m-s
email)

Hello MS,

Writing R Extensions says that \donttest{} is to be "used to mark code
that should be run by example() but not by R CMD check". This package
provides facilities for handling Google service account tokens and
OAuth2 user credentials. If, for example, this token fetching code runs
on CRAN, I will need a way to provide a shared secret to CRAN, so it can
decrypt a token file, which is the approach I take on remote continuous
integration services. There is also the problem that sometimes there are
intermittent 500 errors (server errors) for anything like this, which
must also be tolerated. Can you please advise how I should proceed? If I
change all of these from \dontrun{} to \donttest{}, can you assure me
this will not cause problems?

Currently gargle uses one central function for outputting information to
the user (cat_line()). This is primarily used for print methods and that
is its long term use. I also have some usage that provides more of a
"debugging mode", just in case users need more information about which
credential functions token_fetch() is actually executing. This
centralized function cat_line() is under control of a documented option,
wrapped in a documented function, and it defaults to FALSE. A user will
never see this output unless they actively change the option, presumably
at my request, while we debug something. I believe I will have a richer
set of UI options in the future and, if this debugging mode proves to be
useful, I will refactor it accordingly.

Under which CRAN policy does this repeated directive about cat() fall
and does it apply to all packages? Because I would like to alert my
fellow maintainers to this new rule, so we can all help look for code in
existing CRAN packages and base R that may need modification.

"Please rephrase the first sentence of the description." Can you please
provide more specifics? In an experience with another package, my
attempt to comply with a similar request was even less acceptable to
CRAN and lead to another round of submission and review. I would like to
get it right the first time.

-- Jenny
