Set NOT_CRAN to true for the checks ?

[gaborcsardi]

Maybe optionally?

[hadley]

I don't think we want to do this, just so we're as similar to CRAN as possible

[jimhester]

So callr has
https://github.com/r-lib/callr/blob/2dffbbe502c7aba84fc0785add4a1e588f88b03c/R/rcmd.R#L106

which basically always sets NOT_CRAN=true unless it is already set, which I believe is causing revdepcheck to always have NOT_CRAN=true set when running revdepchecks. I am pretty sure I have purged NOT_CRAN settings in all of my environment configurations and it is still ends up being set when I run revdepcheck::revdep_check()

[gaborcsardi]

callr probably should not touch NOT_CRAN at all, no? @hadley ?

[jimhester]

I think not touching it is probably best.

In the meantime I can successfully work around it with

withr::with_envvar(c(NOT_CRAN="false"), revdepcheck::revdep_check())

[hadley]

Hmmmm, I'm not sure. We are asserting that the code is not running on CRAN, which is true.

Maybe we're conflating too many definitions of "on cran"?

It probably should at least be an option.

[gaborcsardi]

Hmmmm, I'm not sure. We are asserting that the code is not running on CRAN, which is true.

In callr not always, actually, because we do use callr::r for many things now, and some of these do run on CRAN.

[hadley]

Ok, that's true. We should be setting NOT_CRAN at a higher level.