[keelerm84]

Summary

FlagBuilder.value_for_all (in TestData) and FlagBuilderV2.value_for_all (in TestDataV2) replaced the flag's variations with the single value and called variation_for_all(0), which leaves the off variation unchanged. A new test flag is a boolean flag with off variation 1, so the built flag had variations: [value] and offVariation: 1. That index is out of range, so the flag failed to evaluate (malformed flag) when targeting was off. The method is documented to return the value whether targeting is on or off.

value_for_all now also sets the off variation to 0 in both builders. variation_for_all is unchanged; it still leaves the off variation alone.

The new tests check the built flag with targeting on and off for both builders, and check that the client evaluates a value_for_all flag with targeting off to the value with an OFF reason.

This matches the Ruby fix in launchdarkly/ruby-server-sdk#489.

_{Opened from termag}

---

Note

Overview
Fixes TestData / TestDataV2 so value_for_all behaves as documented when targeting is off.

Previously, value_for_all replaced variations with a single value and set fallthrough via variation_for_all(0), but left off variation at the default boolean flag index (1). With only one variation, that index was invalid and evaluation failed when targeting was off. The builders now also call off_variation(0) so the off path returns the same value.

Tests cover the built flag shape (including targeting off) for both builders, plus a client evaluation test asserting value 'x' with reason OFF when targeting is disabled.

^{Reviewed by Cursor Bugbot for commit d9958e4. Bugbot is set up for automated code reviews on this repo. Configure here.}