Modernize Copy Requirement Specification reqSpecCopy

Modernize: Copy Requirement Specification (reqSpecCopy)

• Issue: #1797 (enhancement) — bugs found while testing: #1798, #1799, #1800
• Legacy screen: lib/requirements/reqSpecEdit.php actions copy / doCopy → gui/templates/dashio/requirements/reqSpecCopy.tpl
• Modern screen: gui/templates/requirements/reqSpecCopy.html
• BFF: api/reqspeccopy/index.php
• Entry point: a new Copy Requirement Specification button on the modern Requirement Specification Viewer (gui/templates/requirements/reqSpecView.html) + $actions->reqSpecCopy

Why this screen

The TODO section of docs/MODERNIZATION-STATUS.md is empty (every ASIDE entry maps to a modern screen), so the smallest coherent legacy-only slice was picked: the requirement-specification copy popup, the last un-ported reqSpecEdit.php action.

Two things made it worth doing on its own:

1. The capability had been silently dropped in 2.0.1. The legacy gui/templates/dashio/requirements/reqSpecView.tpl:52 rendered a Copy Req. Spec button pointing at reqSpecEdit.php?doAction=copy. No modern screen replaced it, so after the modernization the modern Requirement Specification Viewer had no copy-specification entry point at all.
2. The legacy write was unsafe. reqSpecEdit.php runs checkRights() against the SESSION context only, while containerID and tproject_id came straight out of the request body — a copy could be aimed at any node of any test project. It also offered the source specification as its own destination, and its target_position (top / bottom) radio pair was dead UI: doCopy() called copy_to() with four arguments and never read the position.

Legacy behaviour carried over

Legacy	Modern
Source card: doc id, title, test project, prefix, author, scope	same, scope rendered as escaped plain text (stripHtml)
Destination = the test project that owns the source	kept as the default, but a project switcher is offered
Container <select> of the whole tree	indented list, source subtree removed
target_position radios (rendered, never used)	actually honoured — node_order renumbered in two collision-free phases
Message A copy of Req. Spec (DOCID:… - …) has been done (DOCID:… [n])	verbatim
Copy / Cancel	Copy + Open the copy shortcut to the new specification

Hardening over legacy

• mgt_view_req and mgt_modify_req on the destination project (the legacy screen only ever checked the session project).
• mgt_view_req on the project that owns the source, so a specification cannot be read (or its metadata leaked) across projects.
• The destination node is proved to be a test project / requirement specification of the destination project (rscResolveContainer), so a foreign container is a 400.
• A destination inside the source subtree is refused (destination_inside_source) — legacy happily copied a spec into its own child.
• bffSameOriginGuard() + bffEnforceSession() on the write, stable machine codes on every error answer (400/401/403/404/405).
• The legacy copy / doCopy URLs are now a non-mutating 302 shim placed before new reqSpecCommands() / initialize_gui(), so the old write is not executed and the request does not even reach the GUI initializers. It carries ?legacy_intent=, and the popup then says "Nothing has been copied yet".

BFF contract

GET  ?action=init&req_spec_id=N[&tproject_id=P]
       → source, destination, default_project_id, destinations[], rights, tproject_fallback_from
GET  ?action=projects          → the test projects the caller may write into
POST ?action=copy&req_spec_id=N&container_id=M[&tproject_id=P]&target_position=top|bottom
       → new{id,doc_id,title}, position

Bugs found and fixed

Issue	What
#1798	?action=projects listed orphan node_type_id=1 nodes (a nodes_hierarchy row with no testprojects row, which testproject::delete() leaves behind) as destinations. rscResolveContainer proves the destination against testprojects, so every pick of one died with 400. Now an INNER JOIN testprojects … AND TP.active = 1.
#1799	Five screen defects: (1) the success box was wiped by the post-copy load(), so a successful copy looked like a no-op; (2) showState() picked "the first <p> without a <code>", which is the access-denied hint — every error card showed the wrong explanation and an empty machine code; (3) the first repair of (2) lower-cased notFound while the markup id is notFoundCode, so it silently did nothing; (4) a stale ?tproject_id= (exactly what the legacy shim hands over — the SESSION id) produced zero containers, an empty selector and a disabled Copy; (5) the ?legacy_intent= notice was wiped twice and threw ReferenceError: p is not defined.
#1800	Event Viewer: E_WARNING Undefined property: tlUser::$id on every denial (tlUser has no ->id; the id is the protected dbID of tlObject), and one Trying to access array offset on null pair per shim hit because the redirect ran after initialize_gui().

Test results

Suite 1797 = 25 / 25 PASS (tmp/TLU_Test_Cases.md), covering the two positions, the project root, a cross-project copy, the empty-destination state, the 400/401/403/404/405 matrix, both legacy shim branches, the toolbar button, the "every other reqSpecEdit.php action still works" regression, and the Romanian bundle (0 missing data-i18n keys). Event Viewer clean — three consecutive shim hits add 0 rows.

Screenshots

• — the screen
• — after a successful copy
• — no destination available
• — the 404 card + machine code
• — the ?legacy_intent= notice
• — Romanian