Fix coord_sf() regression introduced in draw_axis rewrite

[paleolimbot]

In the draw_axis() rewrite (#3349), I neglected a piece of code that allowed list()s of language objects to be passed as labels. This is only used in coord_sf() to draw degree signs next to labels, and because the coord_sf() tests are all skipped, I didn't notice that I had missed this piece of code. This PR fixes the regression so that coord_sf() plots have degree signs in axis labels once more.

[clauswilke]

I've been thinking that the decision to skip the sf tests may not be a good one. I realize the problems created by tests whose outcome depends on the version of other packages, but it may be better to update the reference images to those tests from time to time than to risk introducing major regressions.

[paleolimbot]

I agree! The issue is #2684.

[clauswilke]

It seems like 2 of the 3 items I listed in that issue have been resolved already. Should we turn tests for sf back on and see what happens? @hadley?

[hadley]

Maybe we should just skip them on CRAN? That way we'll still see on travis and locally, but changes to sf or other stream packages won't force a ggplot2 release.

[clauswilke]

All visual tests are already skipped on CRAN. That's what I meant when I said 2 of the 3 items I listed in the issue have been resolved: We skip visual tests on CRAN, and on Travis we only run visual tests for the currently released R version.

[hadley]

Ok got it, then it's definitely worth re-enabling.

[paleolimbot]

I think the instability was because we were reading the North Carolina dataset from a shapefile and/or gpkg. I changed this to be a polygon from one county constructed manually, which results in the dopplegangers passing on the CIs and my computer.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/