Skip to content

Time.at(t, in: nil) returns UTC instead of the local time (or the time zone of t), unlike CRuby #10018

Open

Description

@laurent-martin

Environment Information

  • JRuby: jruby 10.1.3.0 (4.0.9) 2026-11-06 ffffffffff OpenJDK 64-Bit Server VM 27.0.1.0.3 on 27.0.1.0.3 +indy +jit [arm64-darwin]
    (also reproduced with jruby 10.1.2.0 (4.0.0) 2026-09-21)
  • Command line: jruby script.rb, no flags, JRUBY_OPTS empty
  • OS: Darwin 27.0.0 ... arm64 (macOS, Apple silicon)
  • Environment: TZ=Europe/Paris set on the command line (the system time zone is also Europe/Paris)
  • For comparison: CRuby 4.0.9 (ruby 4.0.9 (2026-10-23 revision 3e93b23afc) +PRISM [arm64-darwin27])

Expected Behavior

in: defaults to nil in the documented signature of Time.at
(at(time, subsec = false, unit = :microsecond, in: nil)), so passing in: nil explicitly
should be the same as not passing in:: the result is in the local time zone, or keeps the
time zone of t when t is a Time. CRuby does this. A library that forwards an optional
time zone to Time.at(t, in: zone) works as is, with zone being nil when the caller
wants the default.

Actual Behavior

With only one positional argument, Time.at(t, in: nil) returns a UTC Time (utc? is true,
utc_offset is 0) even though the local time zone is not UTC.
Time.at(t), Time.at(t, 0, in: nil), Time.now(in: nil), Time.new(..., in: nil) and
Time.at(t).getlocal(nil) are correct on JRuby.

Reproduction:

# TZ=Europe/Paris jruby tz_nil.rb
p RUBY_DESCRIPTION
p Time.at(0, in: nil)
p Time.at(0, in: nil).utc?
p Time.at(0, in: nil).utc_offset
p Time.at(0)
p Time.at(0).utc?
p Time.at(0, 0, in: nil)
p Time.at(0).getlocal(nil)
p Time.now(in: nil).utc_offset
p Time.new(2025, 1, 1, in: nil).utc_offset

JRuby 10.1.2.0 and 10.1.3.0 (TZ=Europe/Paris):

1970-01-01 00:00:00 UTC            # Time.at(0, in: nil)             <-- wrong
true                               # .utc?                           <-- wrong
0                                  # .utc_offset                     <-- wrong
1970-01-01 01:00:00 +0100          # Time.at(0)
false
1970-01-01 01:00:00 +0100          # Time.at(0, 0, in: nil)
1970-01-01 01:00:00 +0100          # Time.at(0).getlocal(nil)
3600                               # Time.now(in: nil).utc_offset
3600                               # Time.new(2025, 1, 1, in: nil).utc_offset

CRuby 4.0.9 (TZ=Europe/Paris):

1970-01-01 01:00:00 +0100          # Time.at(0, in: nil)
false                              # .utc?
3600                               # .utc_offset
1970-01-01 01:00:00 +0100          # Time.at(0)
false
1970-01-01 01:00:00 +0100          # Time.at(0, 0, in: nil)
1970-01-01 01:00:00 +0100          # Time.at(0).getlocal(nil)
3600
3600

The result is the same without TZ set (the system time zone is Europe/Paris): on JRuby 10.1.3.0,
Time.at(0, in: nil) still returns 1970-01-01 00:00:00 UTC, while CRuby 4.0.9 returns
1970-01-01 01:00:00 +0100.

It also happens when the first argument is a Time: Time.at(t, in: nil) should keep the time zone
of t (same as Time.at(t)), but JRuby returns UTC.

# TZ=Europe/Paris
l = Time.new(2025, 12, 31, 1, 0, 0)
o = Time.new(2025, 12, 31, 3, 0, 0, "+03:00")
p Time.at(l, in: nil)   # JRuby: 2025-12-31 00:00:00 UTC      CRuby: 2025-12-31 01:00:00 +0100
p Time.at(o, in: nil)   # JRuby: 2025-12-31 00:00:00 UTC      CRuby: 2025-12-31 03:00:00 +0300
p Time.at(l)            # JRuby: 2025-12-31 01:00:00 +0100    (correct)

If t is already UTC (Time.utc(...)), both give UTC, so there is no difference.

Impact: a library that stores an optional time zone and calls Time.at(t, in: @tz).iso8601
prints ...Z instead of the local offset when @tz is nil.

Workaround: Time.at(t).getlocal(tz). It gives the same result as Time.at(t, in: tz) on CRuby
for nil, "UTC", "Z", "+HH:MM" and integer offsets. On JRuby 10.1.3.0 the two forms
are identical for every value of tz except nil, where only getlocal returns local time:

# TZ=Europe/Paris, t = 1_767_139_200
Time.at(t, in: nil)           # JRuby: 2025-12-31 00:00:00 UTC        CRuby: 2025-12-31 01:00:00 +0100
Time.at(t).getlocal(nil)      # JRuby: 2025-12-31 01:00:00 +0100      CRuby: 2025-12-31 01:00:00 +0100

Possible fix

In RubyTime.atOpts (core/src/main/java/org/jruby/RubyTime.java), the one-argument path calls
time.gmtime(context) before checking in:, so a nil zone leaves the result in UTC.
atMulti (used by Time.at(t, 0, in: nil), which is correct) only converts to UTC when a zone is given.
The same can be done in atOpts:

--- a/core/src/main/java/org/jruby/RubyTime.java
+++ b/core/src/main/java/org/jruby/RubyTime.java
@@ -1563,10 +1563,8 @@
 
         if (arg2.isNil() && (arg3 == null || arg3.isNil())) {
             RubyTime time = at1(context, recv, arg1);
-
-            time = time.gmtime(context);
 
-            if (!zone.isNil()) time = timeZoneLocal(context, zone, time);
+            if (!zone.isNil()) time = timeZoneLocal(context, zone, time.gmtime(context));
 
             return time;
         }

With this change Time.at(x, in: nil) takes the same path as Time.at(x), which already gives the CRuby
result on JRuby 10.1.3.0 for numbers and for Time arguments (local, UTC and fixed offset).
I read this on the current master and I have not built JRuby or run any test with this change.

I found no existing issue about this. #6229 (in: with a TZInfo::Timezone) and #9436
(utc? when the local zone is UTC) are different.

Activity

  1. changed the title [-]`Time.at(t, in: nil)` returns UTC instead of local time, unlike CRuby (`Time.at(t, 0, in: nil)`, `Time.now(in: nil)` and `Time.new(..., in: nil)` are correct)[/-] [+]`Time.at(t, in: nil)` returns UTC instead of the local time (or the time zone of `t`), unlike CRuby[/+] on Nov 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions