test(index): make createContentIndexAndDelete index-count assertion phase-aware (#36501)

[fabrizzio-dotCMS]

Proposed Changes

Fixes one of the phase-induced test failures tracked in #36501 (found while running the MainSuite battery under an ES→OS migration phase — see #36320).

ContentletIndexAPIImplTest.createContentIndexAndDelete asserted an exact single-store delta after creating a working + live index:

assertEquals((long) oldIndices + 2, newIndices);

Under a dual-write migration phase (-Dopensearch.phase=1/2), createContentIndex fans out to both the ES and OS stores, so listDotCMSIndices() reports two physical entries per logical index and the delta is +4 — failing with expected:<158> but was:<160>.

Fix

Assert the two created indices are present (already checked via foundWorking/foundLive) and that the total grew by at least two, instead of a single-store exact delta:

assertTrue("expected at least two new indices, delta was " + (newIndices - oldIndices),
        newIndices >= oldIndices + 2);

Phase-agnostic: passes in phase 0 (ES only) and under dual-write.

Verification

Ran under phase 1 locally (-Dopensearch.phase=1 -Dit.test=ContentletIndexAPIImplTest): Tests run: 16, Failures: 0, Errors: 0 ✅ (was 1 failure before).

Refs #36501 (tracking), #36320 (ContentletIndexAPIImplTest phase-aware decoupling — this assertion is item #4 there).

[claude]

Claude finished @fabrizzio-dotCMS's task in 2m 11s —— View job

---

Claude finished @fabrizzio-dotCMS's task in 1m 4s —— View job

---

PR Review — phase-aware createContentIndexAndDelete assertion

• Read the diff (git diff of the two PR commits)
• Verified IndexTag.strip API and semantics
• Reviewed for bugs, conventions, and test correctness

New Issues

No issues found.

The change is a test-only fix and is correct:

• newIndices >= oldIndices + 2 (line 354–355) — right call. The exact +2 delta was single-store; under dual-write (opensearch.phase=1/2) each logical index produces two physical entries (+4). The >= lower bound is phase-agnostic and holds in phase 0 (ES only, exact +2) and under dual-write (+4). Presence of the two specific indices is still asserted exactly via foundWorking/foundLive, so the weaker count assertion doesn't lose meaningful coverage.
• IndexTag.strip(index) (lines 339, 378) — correct use of the sanctioned helper. IndexTag.java:216 documents it as safe on untagged names (returns unchanged), so phase 0 names still match, and .os-tagged names in phases 1/2/3 are normalized before comparison. This respects the "only IndexTag may manipulate the marker" invariant in the enum's Javadoc — no ad-hoc .endsWith(".os").
• Comments accurately describe the phase-3 OS-only behavior and the dual-write fan-out.

Note (non-blocking)

The final assertion at line 390, assertEquals(oldIndices, newIndices) after the deletes, is unchanged by this PR and remains an exact-equality check. It's phase-safe only if indexAPI.delete(...) fans out and removes both physical entries per logical index under dual-write. That's outside this diff and tracked separately under #36320 — flagging only for awareness, not as a defect in this PR.

Looks good to merge. ✅
· issue-36501-fix-createcontentindex-phase-aware

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request