`use_pkgdown_github_pages()` causes warning in urlchecker

[malcolmbarrett]

And on CRAN checks.

I believe it's because the trailing / is removed, as the corrected URL has the trailing /:

usethis/R/pkgdown.R

Line 77 in 196b85a

site_url <- sub("/$", "", site$html_url)

However, this is a peculiar thing to find in the code, which makes me think it fixes some other problem.

Here's the result from tidysmd

urlchecker::url_check()
#> fetching [ 0 / 8 ]fetching [ 1 / 8 ]fetching [ 2 / 8 ]fetching [ 3 / 8 ]fetching [ 4 / 8 ]fetching [ 5 / 8 ]fetching [ 6 / 8 ]fetching [ 7 / 8 ]                       
#> x Error: README.md:12:56 404: Not Found
#> status](https://www.r-pkg.org/badges/version/tidysmd)](https://CRAN.R-project.org/package=tidysmd)
#>                                                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#> ! Warning: DESCRIPTION:12:5 Moved
#>     https://malcolmbarrett.github.io/tidysmd
#>     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#>     https://malcolmbarrett.github.io/tidysmd/

^{Created on 2021-10-20 by the reprex package (v2.0.1)}

[jennybc]

Maybe whether we want or can have the trailing / varies depending on the form of the URL. E.g., https://usethis.r-lib.org is OK, but https://malcolmbarrett.github.io/tidysmd/ is not?

Also I cannot help but note the ridiculousness of this URL grooming situation we find ourselves in 🙄

[malcolmbarrett]

😅

Maybe this is as simple as moving sub() to tidyverse_url(), then... at least for these two particular types of URLs

[jennybc]

https://searchfacts.com/url-trailing-slash/

The short answer is that the trailing slash does not matter for your root domain or subdomain. Google sees the two as equivalent.

But trailing slashes do matter for everything else because Google sees the two versions (one with a trailing slash and one without) as being different URLs.

Yes, short of really getting into the URL parsing business, I think your suggestion re: moving the sub() is probably the best move.

[malcolmbarrett]

Since this is limited to GH pages URLs, the simple solution seems best. But another option could be to call urlchecker::url_update() on the fly (or maybe silently run urlchecker::url_check() and suggest the user fix via ui_todo())

[jennybc]

I have a strong preference for the simple solution.

[malcolmbarrett]

I think that's reasonable. In theory, the release checklist should have caught this or other URL issues, anyway; I just happened to run use_pkgdown_github_pages() after urlchecker