locale-dependent test failure for day-of-week

[dilbernd]

Building 707f16daf19e04cf182cc532a2cce8db170ea178 using Leiningen, I get a test failure:

FAIL in (constructors) (java_time_test.clj:61)
enums
expected: (= (j/day-of-week 4) (j/day-of-week :thursday) (j/day-of-week (j/local-date clock)) (j/day-of-week "ee" "05"))
  actual: (not (= #object[java.time.DayOfWeek 0x9d818b5 "THURSDAY"] #object[java.time.DayOfWeek 0x9d818b5 "THURSDAY"] #object[java.time.DayOfWeek 0x9d818b5 "THURSDAY"] #object[java.time.DayOfWeek 0x4f8e6101 "FRIDAY"]))

This is a locale issue. In a REPL with my system locale (English in Austria):

(ns jt-test (:require [java-time :as jt] [java-time.format :as jtf]))
=> nil
(jt/day-of-week (jtf/parse "ee" "05"))
=> #object[java.time.DayOfWeek 0x6ab351fe "FRIDAY"]

In a REPL with -Duser.country=US set on the JVM:

(ns jt-test (:require [java-time :as jt] [java-time.format :as jtf]))
=> nil
(jt/day-of-week (jtf/parse "ee" "05"))
=> #object[java.time.DayOfWeek 0x67860a8c "THURSDAY"]

This is consistent with the expected first DoW being Sunday in the US, but Monday in (at least continental) Europe.

[dm3]

Thanks for the report! What would you think would be the right solution for the issue?

[dilbernd]

No idea, sorry.

(Thought about it a bit when I discovered it; really drove home the value of explicit parameter (or closure/partials) dependencies, this is errno all over.)

[dm3]

The solution for this issue - make the tests locale-independent. Easiest - remove the offender identified in this issue from the tests. Harder - set the default TZ to the known value during test execution and revert to the previous value afterwards.

[dm3]

Removed the locale-dependent test.