adacovex 1.34.0¶

Date: 2026-08-29

Version bumped 1.33.0 -> 1.34.0.

Changes¶

C1: Overview paired cards now fill 50-50¶

The Doc Coverage and Dependency Scope cards on the Overview tab sit in a .chart-pair grid, but that pair was itself a single item inside the overview .chart-grid (auto-fit, min 280px per cell), so the two cards were squeezed into one narrow column instead of each taking half the content width. The pair now spans every grid track (grid-column: 1 / -1), so the two cards fill 50-50 of the Overview content container exactly like Source Overview and Quick Stats do, and still stack on narrow screens.

C2: Search finds page sections, not just tabs¶

The header search indexed each tab as one blob of words, so a query only ever jumped to a whole tab and multi-word page text such as “orphan tags” or a section heading rarely matched. The index is now built from the rendered DOM at the granularity of a section:

every .card / .chart-card (labelled by its heading);
every compliance-table row, so “orphan tags” lands on the exact row;
every dependency node and every HLR-tagged element;
plus one catch-all entry per tab.

Queries are tokenised and every token must match (AND semantics), so multi-word page content and section headings resolve. Selecting a hit switches to the tab and scrolls to the matched section, briefly flashing it to show where the match is, instead of only switching tabs. Dependency names, versions, scopes, licences and PURLs stay indexed, so the box still filters the dependency tree.

C3: System-tool probe cache self-invalidates¶

The referenced-tools cache key now folds in every curated tool’s version-probe flag and the probe fallback chain. Editing how a tool’s version is read (for example the go subcommand or the --version, -v, version fallbacks) busts the cache automatically. The 1.33-era control relied on a manually bumped |tools-v3| salt; forgetting the bump silently served stale probe results within a release. No hand-maintained salt is needed now.

C4: gnatdoc and gnatformat documented as Alire dev dependencies¶

gnatdoc_bin and gnatformat_bin were already dev dependencies declared in alire-dev.toml (run via alr exec for the make doc and make fmt targets), but the README’s Requirements list only mentioned gnatpp / gnatdoc as optional tools. The list now states that gnatdoc_bin and gnatformat_bin are managed by Alire as dev dependencies, so the published crate still installs and builds with no toolchain beyond the GNAT compiler.

Fixes¶

H1: Proof Check Types bars now scale with value¶

1.33 rolled out proof-category bars with a per-row track-width style, but the CSS .hbar-track { flex: 1 } overrode that inline width, so every proof bar stretched to the full row and showed only the proved rate (100% green), never the magnitude – unlike the test-category bars, which scale because their fill is measured against the largest category. The proof fills are now also measured against the largest category: the green fill is Proved/Max and the red remainder is (Checks-Proved)/Max inside a full-width track, with the grey track showing the scale gap – so a 407-VC category reads as a longer bar than a 56-VC one, precisely like the test chart.

H2: Links render in the accent colour¶

Links outside the Credits and footer areas – the dependency-detail source link, the /api/deps / /api/metrics references, and the nomnoml note – fell back to the browser’s default blue/purple, while the Credits and footer links used --accent. A global a { color: var(--accent) } rule now styles every link uniformly (the Credits/footer rules become redundant but harmless), so the detail panel, dependency views and search results match the rest of the dashboard in both themes.

Test Suite¶

Native suite grows to 1066 tests (14 categories). The HTML/Markdown renderer suite gains two assertions: the proof-check-type bar scales against the largest category (Flow’s green fill is a quarter of the full-width track when Flow is 2 of 8 runtime checks, and the largest category fills it), and the partial-proof red remainder is sized by the maximum category. The existing donut, chart-pair and hoverable scope-ring checks still pass unchanged.

Proof Results¶

Platinum, 723/723 VCs proved across 50 analysed units (unchanged at 1.34.0 – the renderer and manifest-parser bodies are not in SPARK_Mode On, so the bar scaling and cache-key changes add no new VCs). Invocation: adacovex prove (--steps=10000, gnatprove 16.1.0). 0 unproved, 0 justified.

Traceability¶

No new HLRs. Coverage:

RENDER-HTML / HLR-DASH – C1 paired-card width, C2 section search, H1 proof-bar scaling, H2 uniform link colour.
MANIFEST / HLR-SBOM – C3 self-invalidating system-tool probe cache.
HLR-ARCH – C4 doc changes (gnatdoc_bin / gnatformat_bin declared as Alire dev dependencies in alire-dev.toml and run via alr exec for make doc / make fmt).

See docs/dashboard.md.