cran.md reporting 0 packages from CRAN

[eliocamp]

I'm running revdepcheck and after seemingly checking all packages correctly, the file cran.md states:

## revdepcheck results

We checked 36 reverse dependencies (0 from CRAN + 36 from Bioconductor), comparing R CMD check results across CRAN and dev versions of this package.

 * We saw 0 new problems
 * We failed to check 0 packages

However I'm pretty certain that most if not all the packages I checked are on CRAN. The package is ggnewscale.

For example, one of the reverse dependencies is cmcR, which is on CRAN and not in Bioconductor.

[gaborcsardi]

I think you are using RSPM, and repdevcheck does not know that that is CRAN. It could try to guess it. What's the output of

getOption("repos")

[eliocamp]

Yes, I'm using RSPM: "https://packagemanager.rstudio.com/all/__linux__/focal/latest"

[gaborcsardi]

Can you please run

getOption("repos")

and show the complete output? Details might matter. Thanks!

[eliocamp]

Yep, no problem.

> getOption("repos")
                                                           RSPM 
"https://packagemanager.rstudio.com/all/__linux__/focal/latest

[gaborcsardi]

Yeah, so if you name that repo CRAN instead of RSPM, then revdepcheck will probably know that they are CRAN packages. (I haven't tested, though.)

[eliocamp]

I've just renamed the repo and ran it again. I still get the same message.

[gavinsimpson]

I'm also using RSPM. I see the same behaviour if the revdep check process times out as per #355; solution there is to make timeout a very large number (it's in seconds). Increasing the timeout allows the revdep checks to complete, but do note the second comment in #355 as this timeout applies to the entire revdep checking process (at the time of writing) and if your package's reverse dependencies themselves have a lot of dependencies, just installing those dependencies can take a very long time indeed

[MichaelChirico]

I think I'm seeing a similar problem on revdep_check() of {bit64}.

What I notice is that my revdep_check() terminated mid-way for an unknown reason; README.md correctly reports ## Failed to check (85), but cran.md incorrectly reports (17 from CRAN + 85 from Bioconductor) (glancing at the CRAN landing page we see ~8 Bioconductor revdeps: https://cran.r-project.org/web/packages/bit64/index.html).

It seems all too convenient that those two numbers match exactly, so I suspect revdep_check() has somehow mixed up "these packages failed" to mean "they must come from Bioconductor".

[MichaelChirico]

Based on this that would make sense:

revdepcheck/R/report.R

Line 294 in 5a0c138

paste0(" (", sum(on_cran), " from CRAN + ", sum(!on_cran), " from Bioconductor)"),

revdepcheck/R/report.R

Line 284 in 5a0c138

on_cran <- map_lgl(results, on_cran)

revdepcheck/R/report.R

Lines 330 to 332 in 5a0c138

	on_cran <- function(x) {
	isTRUE(x$new$cran)
	}

We're categorizing Bioconductor ↔️ !isTRUE(<cran flag set>), but "the check failed" is another reason that flag wouldn't be set.