[R] Investigate date handling issue in test-r-rhub-debian-gcc-devel-lto-latest crossbow job

[amoeba]

Describe the bug, including details regarding any error messages, version, and platform.

While verifying the 18.1.0 R CRAN submission in #44914, we noticed the test-r-rhub-debian-gcc-devel-lto-latest crossbow job failing due to a test failure related to datetime round-tripping. We should investigate the cause:

══ Failed tests ════════════════════════════════════════════════════════════════
── Failure ('test-dplyr-funcs-datetime.R:302:5'): timestamp round trip correctly via strftime and strptime ──
`object` (`actual`) not equal to `expected` (`expected`).

actual vs expected
                         %Y-%m-%d %OS
- actual[1, ]                      NA
+ expected[1, ]   1999-03-16 00:00:12
- actual[2, ]                      NA
+ expected[2, ]   1999-10-08 00:00:18
- actual[3, ]                      NA
+ expected[3, ]   1999-04-04 00:00:03
- actual[4, ]                      NA
+ expected[4, ]   1999-05-28 00:00:11
- actual[5, ]                      NA
+ expected[5, ]   1999-03-16 00:00:08
- actual[6, ]                      NA
+ expected[6, ]   1999-09-25 00:00:00
- actual[7, ]                      NA
+ expected[7, ]   1999-10-12 00:00:20
- actual[8, ]                      NA
+ expected[8, ]   1999-04-15 00:00:20
- actual[9, ]                      NA
+ expected[9, ]   1999-05-01 00:00:03
- actual[10, ]                     NA
+ expected[10, ]  1999-12-15 00:00:01
and 90 more ...

     actual$%Y-%m-%d %OS | expected$%Y-%m-%d %OS                
 [1] NA                  - "1999-03-16 00:00:12" [1]            
 [2] NA                  - "1999-10-08 00:00:18" [2]            
 [3] NA                  - "1999-04-04 00:00:03" [3]            
 [4] NA                  - "1999-05-28 00:00:11" [4]            
 [5] NA                  - "1999-03-16 00:00:08" [5]            
 [6] NA                  - "1999-09-25 00:00:00" [6]            
 [7] NA                  - "1999-10-12 00:00:20" [7]            
 [8] NA                  - "1999-04-15 00:00:20" [8]            
 [9] NA                  - "1999-05-01 00:00:03" [9]            
[10] NA                  - "1999-12-15 00:00:01" [10]           
 ... ...                   ...                   and 90 more ...
Backtrace:
    ▆
 1. └─arrow:::expect_equal(...) at test-dplyr-funcs-datetime.R:302:5
 2.   └─testthat::expect_equal(...) at tests/testthat/helper-expectation.R:43:

cc @jonkeane

Component(s)

R

[amoeba]

@jonkeane did some sleuthing and found a commit to r-devel we should investigate, wch/r-source@1e4a70c.

[jonkeane]

It looks like this was actually resolved in R-devel (I believe with wch/r-source@a790cd2)

[amoeba]

Thanks, that looks reasonable to me.