Issue 628 Step Execution Type Default Fix

Issue 628 — E_WARNING "Undefined array key execution_type" When Creating Test Cases Without Explicit Step Execution Type

Issue #628

Creating test cases whose step arrays omit the optional execution_type key (minimal CLI/import fixtures, REST-API-style payloads) worked functionally but wrote one E_WARNING row per created case into the Event Viewer (events table):

E_WARNING | Undefined array key "execution_type"
  - in lib/functions/testcase.class.php - Line 804

Reproduced on a fresh DB with php tmp/repro_628.php: two cases created → events ids 8 and 9 (log_level=2, source GUI), exactly one per case.

Root Cause

testcase::createVersion() — the private method behind testcase::create() / createFromObject() — loops over the step array and reads an optional key unguarded:

// lib/functions/testcase.class.php:799-805 (pre-fix)
$this->create_step($tcase_version_id,
                   $item->steps[$jdx]['step_number'],
                   $item->steps[$jdx]['actions'],
                   $item->steps[$jdx]['expected_results'],
                   $item->steps[$jdx]['execution_type']);   // ← line 804

Two facts make this warning pure noise:

1. create_step()'s signature already defaults the parameter: function create_step($tcversion_id, $step_number = null, $actions = null, $expected_results = null, $execution_type = TESTCASE_EXECUTION_TYPE_MANUAL) (lines 5774-5775).
2. It even sanitizes out-of-range values back to MANUAL ($dummy = (isset($this->execution_types[$dummy])) ? $dummy : TESTCASE_EXECUTION_TYPE_MANUAL;, line 5796).

So under PHP 8 the missing key raises E_WARNING "Undefined array key" (a silent notice in PHP 7), which TestLink's error handler persists to events, while the stored result is correct anyway. Any caller omitting the key hits it on every creation — and per the in-code comment at line 797 this loop is shared with the REST API, so API clients are first-class victims.

Blast radius: createVersion() has exactly two call sites — createFromObject() (:432, fed by create()) and copy_to() (:2264, which always sets the value from a DB row). Similar raw reads at :2323/:2341/:2602 are fed from DB rows where the column always exists — untouched.

Fix Approach

Chosen: guard the single raw read with the same idiom already used twice in the class (:5786, :6072):

// lib/functions/testcase.class.php:799-807 (post-fix)
$this->create_step($tcase_version_id,
                   $item->steps[$jdx]['step_number'],
                   $item->steps[$jdx]['actions'],
                   $item->steps[$jdx]['expected_results'],
                   isset($item->steps[$jdx]['execution_type'])
                     ? $item->steps[$jdx]['execution_type']
                     : TESTCASE_EXECUTION_TYPE_MANUAL);

Rejected alternatives:

• Defaulting execution_type upstream in every caller — N call sites, larger diff, easy to miss future callers.
• Changing create_step()'s signature — unnecessary; it already defaults and sanitizes.
• ?? null-coalescing — equivalent here, but the codebase style at the two existing guards is isset() ? : , so consistency wins.

Verification

Regression matrix (php tmp/regr_628.php, suite 62 in tmp/TLU_Test_Cases.md):

Case	Input	Stored tcsteps.execution_type	New error/warning events
Missing key	steps without execution_type	1 (MANUAL default)	none
Explicit value	execution_type=2	2 preserved	none
Invalid value	execution_type=99	1 (create_step sanitize, unchanged)	none

Event Viewer diff across the whole matrix: NO_ERROR_WARNING_EVENTS.

Files Changed

• lib/functions/testcase.class.php — the +3/-1 guard at line 804.
• tmp/repro_628.php — repro driver (project REP628 → suite → 2 TCs, events diff).
• tmp/regr_628.php — post-fix regression matrix driver.
• tmp/TLU_Test_Cases.md — Suite 62 documenting the above.