Bugfix Issue 1415 buildCfields EWarning

Bugfix — Issue #1415: print.inc.php — E_WARNING Undefined variable $buildCfields on every report render with unexecuted TCs on a build

Problem

Generating a Test Plan Execution report on a build (printDocument.php?type=testreport_onbuild → DOC_TEST_PLAN_EXECUTION_ON_BUILD) raises one E_WARNING "Undefined variable $buildCfields" per linked Test Case that has NO execution row on the selected build. Each warning is logged as an event (log_level=2, source 'GUI - Test Project ID : 1') in the events table, polluting the Event Viewer used for CI integration checks.

Reproduced at HEAD (fresh-import CI box, PHP 8.x, MySQL), both on the legacy apikey path and the logged-in session path:

E_WARNING\nUndefined variable $buildCfields - in .../lib/functions/print.inc.php - Line 1595

Verified 2026-09-11 on fixture data (plan with 3 TCs, 2 builds, an execution on build 2 for 1 TC only): exactly 2 new events per render (one per unexecuted-on- build-2 TC). Document output is unaffected because the undefined variable short-circuits its own guard.

Root Cause

Chain (each hop backed by file:line):

1. lib/functions/print.inc.php:922 — renderTestCaseForPrinting(&$db,&$node,&$options,$env,$context,$indentLevel) never declares $buildCfields; a grep of the function confirms the single bare reference is line 1595.
2. print.inc.php:1067-1071 — build custom fields are fetched onto the static tree object: $st->buildCfields[$tbuild_id] = $st->build_mgr->html_table_of_custom_field_values(...), gated on the build_cfields print option and only when $exec_info is not null.
3. print.inc.php:1594-1602 — for a TC with no execution on the selected build (is_null($exec_info)) the build-name block runs its build-CF row; its outer guard !is_null($buildCfields) (line 1595) reads the undefined variable → PHP8 E_WARNING → event row. The row body already used the real source, $st->buildCfields[$build_id].
4. Git history pinpoints the regression source:
   • 25619cd009 (2016) used static $buildCfields; populated per build id inside the same function.
   • 90c461a54 ("#0008556 DB Access Error", 2019) refactored the storage to $st->buildCfields, migrating the write, the row body and buildTestExecResults() — but forgot the guard on the then-old line 1563. From that commit on, $buildCfields is a leftover name: the variable is undefined, and the whole CF row is dead code.

Why it breaks now: PHP 8.x surfaces the 2019 refactor leftover as an E_WARNING on every affected render. Not a modernization regression; the legacy printDocument.php path reproduces it identically.

Blast radius: grep buildCfields in lib/functions/print.inc.php → 7 hits; only the guard under review used the bare variable; the other 6 already use $st->buildCfields (lines 1068, 1069, 1595, 1596, 1599, 1629) plus an unrelated $things->buildCfields (2355). No other file references it. Single-line blast radius.

Fix

Committed on branch fix/issue-1415, commit e9fa9f0e9:

if(is_null($exec_info)) {
  if(isset($st->buildCfields[$build_id]) &&
     $st->buildCfields[$build_id] != '') {
    $code .= '<tr><td width="' . $cfg['firstColWidth'] .
             '" valign="top"></td>' . '<td colspan="'  . $tsp . '">' .
             $st->buildCfields[$build_id] . "</td></tr>\n";
  }
}

The stale !is_null($buildCfields) clause is deleted. The remaining isset($st->buildCfields[$build_id]) provides the same "CF data already fetched for this build" semantics the original static-null check was meant to express, against the real data source. The row renders its build-CF table when data exists — the behaviour the 2019 refactor intended but had accidentally dead-branched.

Why this method: the guard should read exactly what the row body reads. isset() is null-safe and covers "initialised" — an extra is_null($st->buildCfields) adds nothing.

Rejected alternatives: (a) $buildCfields = null; initializer — silences the warning but keeps the CF row permanently dead, losing intended output; (b) switching the whole guard to $st->buildCfields with an added null check — redundant given isset().

Files Changed

• lib/functions/print.inc.php — line 1595: removed the undefined !is_null($buildCfields) clause from the build-CF guard (-2/+1).
• tmp/TLU_Test_Cases.md — regression suite "Regression — Issue #1415", 7/7 PASS.
• tmp/fixtures_1415.php — fixture (1 project, plan with 3 linked TCs on platform 1, 2 builds, execution on build 2 for the 3rd TC only).

Verification

All checks on branch fix/issue-1415, PHP 8.x, MySQL fresh-import schema:

• Pre-fix control (file reverted to HEAD): session render of the affected URL → HTTP 200 + 2 new E_WARNING ... Line 1595 events.
• Post-fix R1 (build_id=2, allOptionsOn=1): HTTP 200, 0 new events.
• Post-fix R2 (build_id=1, no executions at all): HTTP 200, 0 new events.
• Post-fix R3 (build_id=2, print options off): HTTP 200, 0 new events.
• Post-fix R4 (anonymous/apikey path): HTTP 200, 0 buildCfields events.
• Post-fix R5 (positive path): a build custom field seeded (cf_build_1415, value BUILDCF-MARKER-VALUE) renders its row in the document exactly once for the executed TC via buildTestExecResults() — the restored isset() branch is now live for no-execution TCs when the static has been populated by an earlier executed TC in the same render.
• Post-fix R6 (restored branch proof): with the executed TC (tcversion 10) reordered to render FIRST, BUILDCF-MARKER-VALUE appears 3x in the HTML — once per TC under its "Build: Build1415-2" line, i.e. the line-1595 row now renders for the two no-execution TCs (isset() true from the earlier executed TC). Pre-fix those rows were impossible (guard always false: is_null(undefined) → true ⇒ !is_null(...) → false).
• php -l lib/functions/print.inc.php clean.
• Events table after all fixed-path renders: 0 rows matching %buildCfields%.

Regression suite: tmp/TLU_Test_Cases.md — "Regression — Issue #1415", 7/7 PASS.

Related Finding

While testing, the anonymous/apikey direct access to printDocument.php also emits 3 E_WARNING "Undefined array key basehref" events (printDocument.php:175/254, print.inc.php:700) because $_SESSION['basehref'] is unset without a session. Pre-existing (reproduced identically against the pre-fix file), filed separately as a bug issue.

Address: Fixes #1415 (verified + pushed on fix/issue-1415).