fixes R-devel foverlaps failures due to c.POSIXct change

[jangorecki]

fixes

  Test 1390.3 produced 1 errors but expected 0
  Expected:
  Observed: 'origin' must be supplied
  Test 1390.4 produced 1 errors but expected 0
  Expected:
  Observed: 'origin' must be supplied
  Test 1390.5 produced 1 errors but expected 0
  Expected:
  Observed: 'origin' must be supplied
  Test 1985.1 produced 1 errors but expected 0
  Expected:
  Observed: 'origin' must be supplied
  Test 1993.1 produced 1 errors but expected 0
  Expected:
  Observed: 'origin' must be supplied

[codecov]

Codecov Report

Merging #4428 into master will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #4428      +/-   ##
==========================================
- Coverage   99.61%   99.60%   -0.01%     
==========================================
  Files          72       72              
  Lines       13917    13918       +1     
==========================================
  Hits        13863    13863              
- Misses         54       55       +1     

Impacted Files	Coverage Δ
R/foverlaps.R	100.00% <100.00%> (ø)
src/assign.c	99.84% <0.00%> (-0.16%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update dd7609e...864d9eb. Read the comment docs.

[mattdowle]

Great. Needs news item. For example, when this change in R-devel is released, and users start getting this error message in data.table <= 1.12.8, then if they search for the error, there's a chance they'll find the news item and find out they need to upgrade data.table.

[jangorecki]

@mattdowle done

[mattdowle]

This missed coverage line is odd, doesn't seem related to this PR. Will let it go. If it's truly now uncovered then master coverage will show it and it won't get forgotten.

[jangorecki]

@mattdowle see #4424

[eddelbuettel]

Are we sure this only affects r-devel? It seems to also affect R-patches if I am reading the CRAN results page for data.table correctly (but not R-release).

It affects us in Debian as the release data.table fails its tests with R 4.0.0, effectively block transition of r-base 4.0.0 (and now r-base 4.0.1) from unstable into testing. See #962497.

[jangorecki]

Quite possibly R-patched as well. It is actually surprising that base R decided to make a breaking change like this without first giving a warning for at least one version.
We are planning to publish to CRAN soon, so hopefully it will not last for long.

[eddelbuettel]

And the actual fix is a one-liner to maybe the Debian maintainer for data.table can carry that over. I probably would if it were my package (though this speculating as there may be other things too betweeb R 4.0.0 and data.table 0.12.8).

I think the change to c.POSIXct was pretty late-breaking as well. Oh well. Spilled milk now.