Is test_possibly_better() in test/test_hoe.rb really accurate?

[dleidert]

We observe a test failure when running the testsuite on Debian with Ruby 2.7:

[..]
> TestHoe#test_possibly_better = 0.00 s = F
> TestHoe#test_nosudo = 0.00 s = .
> TestHoe#test_setting_licenses = 0.00 s = .
> TestHoePublish#test_make_rdoc_cmd = 0.00 s = .
> 
> Finished in 0.084995s, 376.4912 runs/s, 1153.0043 assertions/s.
> 
>   1) Failure:
> TestHoe#test_possibly_better [/<<PKGBUILDDIR>>/test/test_hoe.rb:274]:
> Expected: 2020-02-23 00:00:00 UTC
>   Actual: 2020-02-06 00:00:00 UTC

The date expected is the result of Gem::Specification::TODAY whereas the actual date always is determined by the date of the actual file hoe.gemspec (not the date value inside). Can you please explain what you actually expect to happen here?

Here is the bug report against the Debian package:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=952041

[utkarsh2102]

Heh, I get:

Failure:
TestHoe#test_possibly_better [/home/utkarsh/debian/ruby-team/hoe/ruby-hoe/test/test_hoe.rb:274]:
Expected: 2020-04-02 00:00:00 UTC
  Actual: 2020-03-02 00:00:00 UTC

[dleidert]

We've found the cause. The issue is only visible when packaging hoe for downstream, installing the gem into a directory and then run the tests with Ruby 2.7 and this directory in RUBYLIB/GEM_PATH. The relevant change in Ruby 2.7, which caused this, is (to our research) in rubygems/specification.rb (changing :date from TODAY to nil). dd845b9#diff-715213bedd26ddebc71af81759d8fd75L3 actually fixes the problem. I'm therefor closing this report myself.