clock-dependent dirSummary test failure

[kyleam]

Depending on the local time zone and current time, the below test failure can be triggered:

  ══ Failed tests ════════════════════════════════════════════════════════════════
  ── Failure ('test-dirSummary.R:37:3'): dirSummary captures the expected QC status of all scripts ──
  as.Date(dirSummaryRes$data$`Latest edit`[1]) (`actual`) not equal to Sys.Date() (`expected`).

  `actual`:   "2023-12-05"
  `expected`: "2023-12-04"

• svn info --xml returns the date in UTC (e.g., "2023-12-05T14:04:27.050339Z"), and svnInfo() parses that to a POSIXct object. (There's some other processing that goes on downstream, but this is the only processing that matters.)

• a test in test-dirSummary.R converts that value with as.Date() and expects it to be equal to Sys.Date()

That expectation is invalid because Sys.Date() is the current day for the current time zone.

One fix would be to replace the Sys.Date() value with a date for UTC (say format(Sys.time(), "%Y-%m-%d", tz = "UTC")). However, that would leave a harder-to-hit failure: if the test is executed right at the boundary of a day change, with the svnInfo call before and the Sys.time() call for the assertion after, the results would not match.

Instead you could avoid converting to a date. Subtract the POSIXct value in dirSummaryRes from Sys.time(), and assert that the time difference is under N seconds. N should be large enough that, if it were to be reached, it'd be an indication that something was going very wrong with the test run (say 120 seconds in this case?).

cc: @andersone1 @michaelmcd18

[andersone1]

@kyleam is this blocking this version of review from getting into the next mpn? (I e how urgent is this)

[kyleam]

is this blocking this version of review from getting into the next mpn?

No.

[andersone1]

is this blocking this version of review from getting into the next mpn?

No.

@kyleam Ok, that's good to know, thanks for bringing this to our attention.

@michaelmcd18 i can review the PR for this issue whenever it's ready

[michaelmcd18]

Great idea @kyleam, we'll implement this: Instead you could avoid converting to a date. Subtract the POSIXct value in dirSummaryRes from Sys.time(), and assert that the time difference is under N seconds. N should be large enough that, if it were to be reached, it'd be an indication that something was going very wrong with the test run (say 120 seconds in this case?).