Modernization of Execute Tests (lib/execute/execNavigator.php + lib/execute/execSetResults.php) — GitHub issue #662.

The legacy tree-navigator + execution-form pair is replaced by a standalone Dashio two-pane page (gui/templates/execute/execTest.html) backed by a plain-PHP REST BFF (api/execute/index.php, extended with four new actions). The ASIDE Execute → Execute Tests entry now points to the new HTML screen when a test plan is selected (switched in getActions() / initUserEnv() in lib/functions/common.php).

Path: Execute → Execute Tests URL: gui/templates/execute/execTest.html?tproject_id=<id>&tplan_id=<id> BFF API: api/execute/index.php (same file also serves the Execution History screen via ?action=history) Rights: testplan_execute to record results, exec_ro_access for the read-only variant — enforced server-side on every BFF route.

---

1. What the screen does
2. REST API Reference
3. Legacy parity notes
4. i18n Keys
5. Security
6. Testing

• Toolbar: build selector (only active+open builds are executable; closed ones are listed with a "(closed)" hint), platform selector (hidden when the plan has no platforms), result filter and live text search over external id, name and suite path.
• Left pane: all test cases linked to the plan (latest linked version per case) with suite path, version, last-execution timestamp on the selected build/platform and a colored status badge.
• Right pane: the execution form — summary, preconditions, importance, manual/automated marker, prior execution recap for this build/platform, result-status buttons (Not Run / Passed / Failed / Blocked), execution notes and step-level results (per-step status select + notes) with the recorded partial-step results pre-filled from the prior execution.
• Save records one execution row through the legacy exec.inc.php:write_execution(), so every side effect keeps working: requirement-link freezing, relation closing, step partial-execution rows.
• Read-only mode: users with only exec_ro_access see everything but cannot save (inputs disabled; API rejects writes with 403).

All endpoints are GET unless stated otherwise; session auth required; tplan_id must be one of the user's accessible test plans.

Action	Params	Returns
init	tplan_id	project/plan context, grants, builds (+executable flag, default build = newest active+open), platforms, localized status vocabulary
tcList	tplan_id, build_id, platform_id, optional q	latest linked version per case: external id, name, suite path, importance, execution type, last execution (status/ts/tester)
tcDetails	tplan_id, tcase_id, tcversion_id, build_id, platform_id	version block, ordered steps, prior execution incl. prior step results, and exec_history (all executions of this tcversion in the plan, DESC, with per-row run-mode/version/build/tester/status + can_edit_notes/can_delete grants)
save (POST JSON)	tplan_id,tcase_id,tcversion_id,version_number,build_id,platform_id,status(code),notes,steps{stepId:{status,notes}}	saved:true + execution id/ts, or reason:'not_run'; validates executable build + plan link; delegates to write_execution()
deleteExecution (POST JSON)	tplan_id, execution_id	deleted:true; gates: testplan_execute + exec_delete (public-plan-aware), execution must belong to the plan, build must be open; delegates to delete_execution() and logs an AUDIT event only on success
updateNotes (POST JSON)	tplan_id, execution_id, notes	updated:true; gates: testplan_execute + exec_edit_notes, execution must belong to the plan, build must be open; delegates to updateExecutionNotes()
linkBug (POST JSON)	tplan_id,execution_id,tcstep_id,bug_id	validates bug syntax + existence via ITS, writes execution_bugs row, returns {ok:true,id}; audit-logged
createBug (POST JSON)	tplan_id,execution_id,tcstep_id,bug_summary,bug_description	creates a real issue via its->addIssue() (GitHub returns new id), links it to the execution, returns {ok:true,id}; audit-logged

Status codes travel as codes (n/p/f/b) exactly as served by init (statuses[].code).

• Only active+open builds can receive results (legacy UI offered the same set; the BFF additionally rejects violations server-side).
• Not Run results are never written as executions (identical to write_execution() behavior since 1.9.20) — the UI explains this via toast.
• Plans without platforms store platform_id = 0 on execution rows (the client sends -1, mapped exactly like legacy $executedInPlatform).
• Step results reuse the legacy execution_tcsteps storage; steps whose status is not_run and whose notes are empty are skipped like legacy.
• The aside entry keeps its menuGrants gating (testplan_execute, exec_ro_access fallback label) in aside.tpl.

Deliberate v1 scope cuts (tracked on #662): step-level attachment upload, execution custom fields in the form, bulk mode. Execution-level attachment upload/list + "Copy attachments from latest execution" are implemented (see Bugfix-Issue-792-Execute-Tests-Attachments).

Issue-tracker integration inside the execution form (legacy bugAdd.php link / create / doCreate flows), enabled when the project has an issue tracker configured and exec_cfg->features->issue_tracker->enabled.

• Link existing bug — from the case toolbar or a specific step row, open the Link bug to execution modal, switch to Link existing bug, type a bug id and submit. BFF postBugAction('linkBug') validates syntax + existence against the tracker (GitHub), then write_execution_bug($db,$execId,$bugId,$tcstepId) stores the execution_bugs row (execution-level or step-level via tcstep_id).
• Create new bug — switch to Create new bug, fill Title (summary, required) + optional Description, submit. BFF postBugAction('createBug') resolves the tracker, requires addIssue() support, calls its->addIssue($summary,$description) to create a real GitHub issue, then links it to the execution and audit-logs Bug {id} created and linked to execution {execId}.
• Static-name calls guarded so CI fix-bug.yml never picks up test issues.
• The modal target line shows Execution <id> (case) or Execution <id> on step N (<id>) (step).

The test-case details panel renders a Linked bugs DataTables card (Bug ID / Title / Labels / Status / Step), fed real-time from the GitHub API via the tcDetails linked_bugs array:

• Bug ID is a clickable GitHub link (bug_url), unless the issue is deleted.
• Deleted / rejected GitHub issues (API returns HTTP 410) set unavailable:true → the row shows a grey Deleted status badge, an "Unavailable" title and no dead link (#805/#806 demonstrate this; live issues show real title/labels/status).
• Labels render as GitHub-colored badges. githubrestInterface::getIssue() and listIssues() expose label_colors (name→hex) alongside labels; execTest.html renderLinkedBugsTable() uses the same map; missing color falls back to grey #6c757d.
• BFF dedupes by bug id and aggregates the steps each bug is bound to (steps) plus the number of executions (exec_count).
• The same unavailable + GitHub label-color handling is mirrored in the Dashboard (mainPage.html bugs + issues tables, dash.bugDeleted / dash.bugUnavailable).

All strings use the exe.* prefix, defined in ALL ten client bundles: en, de, es, fr, it, ja, pt, ro, ru, zh (46 keys each after the Linked Bugs table keys exe.bugDeleted / exe.bugUnavailable), consumed through TLi18n. Server-side labels (statuses) come from lang_get('test_status_*') like legacy localize_tc_status().

• Session auth + same-origin guard (bffSameOriginGuard) on every request.
• testplan_resolveContext: plan must be within the user's accessible plan set for the owning project (else 403/404).
• Write route requires testplan_execute on (project, plan); exec_ro_access alone yields HTTP 403.
• All ids cast to int; notes prepared inside legacy prepare_string() paths.

Suite 63 in tmp/TLU_Test_Cases.md: 14/14 PASS — covers init/list/detail contracts, full UI save flow, filters/search, not-run guard, closed-build guard, read-only path (banner + disabled UI + 403), locale switch and Event Viewer cleanliness after fixes.

Fixes landed during testing:

• options blob read instead of vestigial option_platforms column (E_WARNING in Event Viewer);
• locale/en_GB/custom_strings.txt adding file_upload_step_exec_ok/ko (surfaces #621's missing en_GB keys on every step-level save).

write_execution() (lib/functions/exec.inc.php:257-258) evaluates lang_get('file_upload_step_exec_ok'/'ko') on every step-level save. When the active locale lacks the keys, lang_get() falls back to en_GB and writes a string '...' is not localized for locale '<loc>' - using en_GB event (log_level 32) into the Event Viewer.

Root cause / coverage: the keys were added to 16 locales' custom_strings.txt and lived in strings.txt for pt_BR/pt_PT, but ro_RO had neither (no custom_strings.txt, 0 hits in strings.txt) — a 19-locale scan confirmed ro_RO was the only missing bundle.

Fix (Refs #621): created locale/ro_RO/custom_strings.txt with both keys (UTF-8 Romanian translation, same Fixes #621 / #662 header convention as the 16 existing bundles).

Verification (regression suite 621, 5/5 PASS):

• ro_RO save pre-fix → 2 L18N events (reproduced, events id 9/10);
• ro_RO save post-fix → 0 new file_upload_step_exec events;
• en_GB + pt_BR saves → unchanged (still 0);
• browser UI save on the modernized Execute Tests screen → "Execution saved." with no console errors and no L18N events.

Side-discoveries logged as new issues (out of this fix's scope):

• #773 — Undefined property: stdClass::$requirementsEnabled E_WARNING on every execution save when testprojects.options is empty;
• #774 — other ro_RO server-side keys (manual, automated, the_format_tc_xml_import) still missing.

Symptom: opening a test case in the modernized Execute Tests screen showed no inline execution-history table at all. The legacy 1.9.20 screen had, under the prior-execution note line, a full history table with per-row edit-notes / delete / print actions and run-mode icons.

Root cause: the modernized execTest.html simply never rendered that table — the BFF tcDetails response carried no execution-history payload and there were no BFF actions for edit-notes / delete, so 1.9.20 functionality was silently dropped during the rewrite.

Fix approach (restore 1.9.20 parity, following the legacy code paths):

• BFF (api/execute/index.php): tcDetails now also returns exec_history — the full execution set for this tcversion+plan (all builds, DESC), each row with build_name/build_is_open, tester login/name, localized status, execution_ts/execution_duration/run_type, tcversion_number, plus can_edit_notes + can_delete per-row grants (from exec_edit_notes / exec_delete + open build). Two new write routes:
  • deleteExecution — mirrors legacy execSetResults.php delete flow (exec_delete right, open build, execution belongs to the plan); calls delete_execution() (cleans execution_bugs / cfield_execution_values / execution_tcsteps / attachments) and writes an AUDIT event only on success.
  • updateNotes — mirrors legacy editExecution.php doUpdate (testplan_execute + exec_edit_notes, open build, plan ownership); calls updateExecutionNotes().
• Frontend (gui/templates/execute/execTest.html): new renderHistoryHtml() renders the EXECUTION HISTORY table (Date/time, Build+lock icon, Executed by, Status badge, Run-mode icon, Version, Actions) plus a per-row notes line. Row actions: edit-notes → legacy editExecution.php popup; print → legacy execPrint.php popup; delete → confirm dialog → BFF deleteExecution → refresh form + list.
• i18n: added exe.executionHistory / actions / deleteExecution / confirmDeleteExecution / deletedOk / errDelete and exechist.closedBuild to all ten client bundles; also repaired pre-existing corrupted English/Romanian values for the exechist.* keys the table uses.

Verification (regression suite 796, 7/7 PASS, see tmp/TLU_Test_Cases.md): init/list/detail contracts unchanged; history table renders all runtime rows with per-row action buttons; edit-notes popup preloads the execution notes; print popup renders the full execution; delete removes the execution + child rows and logs an AUDIT delete event; rights/state gates confirmed by code review (testplan_execute + fine-grained right, plan ownership, open build); Event Viewer clean (only AUDIT events, no new Error/Warning).

Symptom: the modernized Execute Tests screen had no way to save partial step results for later resumption. Legacy 1.9.20 rendered a "Save Steps Work In Progress Execution" button under the steps table (inc_exec_test_spec.tpl → TLS_saveStepsForPartialExec) that wrote execution_tcsteps_wip rows; the rewrite dropped the button, the BFF action and the WIP read-back, so mid-run progress could be lost.

Root cause: three missing pieces during the #662 rewrite, all present in legacy:

• no button in the form;
• no BFF route (legacy: execSetResults.php → testcase::saveStepsPartialExec());
• tcDetails did not merge execution_tcsteps_wip into prior_step_results, so saved WIP rows never pre-filled the form on reopen.

Fix approach (restore 1.9.20 parity):

• BFF (api/execute/index.php):
  • new POST ?action=savePartialSteps — gates testplan_execute (403), active+open build (400), valid tcversion_id (400); validates every step id against the version (tcsteps+nodes_hierarchy join, forged ids dropped) and every status via results.code_status (fallback not_run); delegates to testcase::saveStepsPartialExec() with testplan_id, platform_id (normalized >0 ? id : 0), build_id, tester_id = session user. Legacy delete-then-insert semantics preserved (a resave replaces the tester's WIP rows for that plan/build/platform).
  • tcDetails now merges WIP rows over prior executed steps for the current (plan, build, platform, tester) context — the resume pre-fill.
• Frontend (gui/templates/execute/execTest.html): non-read-only forms with steps render the Save Steps Work In Progress Execution button + a hint; savePartialSteps() replicates legacy checkStepsHaveContent client-side ("nothing to save" toast, no request), POSTs the same step payload shape as saveExecution(), then toasts on success and re-opens the case so resumed values show.
• i18n: exe.partialExecInfo / partialExecNothingToSave / partialExecSaved / savePartialSteps added to all ten bundles; button label reuses the legacy English string, translations reuse legacy locale/*/strings.txt where present.

Verification (regression suite 64, 11/11 PASS, see tmp/TLU_Test_Cases.md): partial save writes only execution_tcsteps_wip (no executions row); empty save is blocked client-side (no POST, DB untouched); resume pre-fill in-session and across full reloads; full "Save execution" after a partial save clears the WIP rows (legacy exec.inc.php flow) and writes the execution + step rows; zero-right user → 403; inactive build → 400; forged step id dropped; Event Viewer clean (only INFO-level AUDIT rows).

Before: issue-795-execTest-no-partial-save.png (no button).

Symptom: the modernized Execute Tests screen offered only Save execution and Reset in the save row. The 1.9.20 toolbar buttons btn_save_exec_and_movetonext ("Save and move to next") and btn_next ("Next" — move to the next test case WITHOUT saving) were dropped in the #662 rewrite, and the BFF init never surfaced exec_mode->save_and_move, so the configured navigation mode could not be honored.

Root cause: three pieces missing during the rewrite, all present in legacy (lib/execute/execSetResults.php doExecSave()/move2next window at lines 140-345 + 617-632, gui/templates/tl-classic/execute/inc_exec_controls.tpl lines 89-97):

• no "Save and move to next" button / BFF save-then-navigate flow;
• no "Next" (navigate, no save) button / BFF move2next equivalent;
• no read of $tlCfg->exec_cfg->exec_mode->save_and_move (config.inc.php:1119, default 'unlimited') in the BFF init response.

Fix (restore 1.9.20 parity):

• BFF (api/execute/index.php): init now returns save_and_move from config_get('exec_cfg')->exec_mode->save_and_move (fallback 'unlimited').
• Frontend (gui/templates/execute/execTest.html): extracted visibleItems() (the exact renderList() filter predicate — navigation always matches what the user sees); saveAndMoveMode(); execNextTarget() — cyclical successor in the visible set, 'unlimited' = whole list, 'limited' = members of the SAME test suite (grouped by displayed tsuite_path, the flat-list equivalent of legacy getTestCaseNextSibling — note the BFF tsuite_id field actually carries the test-case node id, not the suite id, and is unused by the frontend); moveToNext() = legacy move2next (no save); saveExecution('next') saves then loadList() + openCase(next), successor computed pre-save like legacy. Save row: Save execution · Save and move to next · Next · Reset, all sharing the existing roMode disabled guard.
• i18n: exe.saveAndMoveNext / exe.moveToNext / exe.noNextCase in all ten bundles (legacy locale/*/strings.txt translations reused).

Note — bulk results mode: the legacy bulk results grid is a testsuite-level feature ($args->level == 'testsuite'); the modern flat-list screen has no testsuite view, so bulk execution is out of scope here.

Verification (regression suite 69, 12/12 PASS, see tmp/TLU_Test_Cases.md): BFF init returns save_and_move:"unlimited"; four-button save row; save-and-move writes the execution and advances to the successor; move-to-next navigates WITHOUT writing; unlimited cycle wraps; limited mode scoped per suite; search-filtered navigation honors the visible subset; plain Save unchanged; i18n bundles valid 10/10; Event Viewer clean (no new Error/Warning rows).

Before: issue-793-save-and-next-before.png (only Save execution + Reset).

The Execute Tests toolbar now shows a collapsible Test Plan Notes, Build Notes and Platform Notes panel fed by the BFF init response (tplan.notes, builds[].notes, platforms[].notes). Each section toggles closed/open on click (mirroring legacy execSetResults.tpl show_hide('tplan_notes'|'build_notes'|'platform_notes')). Build and Platform boxes auto-hide when their note is empty; the whole panel disappears when there is nothing to display. Notes re-render whenever the selected build or platform changes. i18n keys exe.tplanNotes/exe.buildNotes/exe.platformNotes added to all 10 bundles.

Verification (suite 1163, 7/7 PASS): all three panels render; plan notes toggle collapse/expand; empty build/platform boxes hide; empty-all hides the panel; notes follow build selection; console clean; Event Viewer clean.