Doesn't try to combine labels if labels are NULL

[eliocamp]

I encountered this strange error. It took me a while to get a reprex because it is incredibly elusive.

library(ggplot2)

grid <- expand.grid(x = seq(0, 360, by = 2.5), y = seq(-90, -20, by = 2.5))
grid$z <- with(grid, x^2*y)

ggplot(grid, aes(x, y)) +
  geom_contour(aes(z = z)) +
  coord_polar() +
  scale_x_continuous(breaks = seq(0, 360, by = 30), labels = NULL)
#> Error in labels[[n]] <- combined: attempt to select less than one element in integerOneIndex

The issue was that coord_polar() tries to combine labels even if they are NULL. I added a check. All tests succeed and the bug is fixed.

[clauswilke]

A test seems to be failing on R devel. Could you see if you can figure out why?

@yutannihilation Is there a reason we check R devel only on macOS? Makes it difficult to figure out whether a problem is related to R devel or to macOS. I think we should check at least on ubuntu also.

[eliocamp]

Some visual test unrelated to this change are failing. Might have to do with fonts or some other platform-specific reason?

══ testthat results  ═══════════════════════════════════════════════════════════
[ OK: 1515 | SKIPPED: 19 | WARNINGS: 1 | FAILED: 58 ]
1. Error: aesthetics are drawn correctly (@test-aes.r#160) 
2. Error: alpha is drawn correctly (@test-aes.r#175) 
3. Error: segment annotations transform with scales (@test-annotate.r#30) 
4. Error: cartesian coords draws correctly with limits (@test-coord-cartesian.R#25) 
5. Error: cartesian coords draws correctly with clipping on or off (@test-coord-cartesian.R#56) 
6. Error: USA state map drawn (@test-coord-map.R#7) 
7. Error: coord_map scale position can be switched (@test-coord-map.R#16) 
8. Error: polar coordinates draw correctly (@test-coord-polar.r#96) 
9. Error: basic coord_trans() plot displays both continuous and discrete axes (@test-coord-transform.R#99) 
1. ...

[clauswilke]

Looks like the failing tests are vdiffr tests. Maybe some graphics related detail has changed in R devel?

[yutannihilation]

Because R devel is not yet available on Linux.

c.f. r-lib/actions#49 (comment)

[yutannihilation]

Let me handle the CI issue on #4116. I'm sure the failure is unrelated to this PR.

[clauswilke]

Thanks!