docs(benchmarks): record Zlib, MPL-2.0, and EPL/Apache-dual targets

[mstykow]

Summary

Records three verify-benchmark-target rows from this round, each stressing a distinct license family, measured on merged main with cached ScanCode baselines:

Target	Family	Files	Provenant	ScanCode	Speedup
glfw/glfw @ ed6452b	Zlib	177	5.18s	88.79s	17.15×
mozilla/glean @ a4166fa	MPL-2.0	1,079	6.26s	128.06s	20.46×
eclipse-vertx/vert.x @ 78ade62	EPL-2.0/Apache dual	1,752	9.29s	194.51s	20.94×

Corpus is now 252 of 252 runs faster, 19.6× median. Chart (scan-duration-vs-files.svg) and the stats line regenerated via generate-benchmark-chart.

Detector fixes this round surfaced merged separately in #1281 (Javadoc @author HTML-anchor extraction; bare <copyright> XML-tag junk). Rows are written as present-tense Provenant-vs-ScanCode end-state differences.

Deferred follow-ups (own PRs)

• glean: pyproject.toml PEP 621 license = { file = "LICENSE" } table form is detected at file level (mpl-2.0) but not propagated to the package's declared_license_expression — needs parser + file-reference/assembly wiring (like the existing about.rs/podspec.rs/rpm license-file precedent).
• vert.x: 4 files with @author Name <a href=url>@handle</a> (trailing handle-anchor) still miss the author.
• glfw: multi-line duplicate/merged holder from consecutive © copyright lines inside a <copyright> XML block (vendored Wayland deps).

Testing

• markdown-lint + prettier via pre-commit hook: pass.
• Timings are same-host compare-outputs --profile common --build-mode optimized runs on merged main; ScanCode rows are cache hits from the original clean runs.

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR adds three new verify-benchmark-target entries — eclipse-vertx/vert.x, glfw/glfw, and mozilla/glean — to the benchmark documentation, each covering a distinct license family (EPL/Apache dual, Zlib, MPL-2.0). The corpus count and SVG scatter chart are updated to reflect the 252 completed runs.

• Each entry is inserted at the correct alphabetical position within its section, follows the established format, and the SVG coordinate offsets are consistent with the surrounding log-log data points.
• The stated 17.15× speedup for glfw/glfw cannot be reproduced by dividing the recorded timing pair (88.79 s ÷ 5.18 s = 17.14×); it is unclear whether this is a rounding artifact from higher-precision measurements or a transcription error.
• The overall statistics line (252 of 252 runs faster, 19.6× median) is self-consistent with the three added entries.

Confidence Score: 4/5

Safe to merge; all changes are documentation additions with no code or configuration impact.

The three new benchmark entries are well-formed and correctly placed. The only open question is the glfw/glfw speedup figure, which does not reproduce from the documented timing pair and may reflect rounding at an unstated precision level.

docs/BENCHMARKS.md — the glfw/glfw speedup header warrants a quick check against the raw timing data.

Important Files Changed

Filename	Overview
docs/BENCHMARKS.md	Adds three new benchmark entries (eclipse-vertx/vert.x, glfw/glfw, mozilla/glean) with correct alphabetical placement; the stated 17.15× speedup for glfw/glfw is off by 0.01× from the ratio derived by dividing the documented timings (88.79s / 5.18s = 17.14×).
docs/scan-duration-vs-files.svg	Adds six SVG data points (three orange ScanCode rects and three blue Provenant circles) for the new benchmark targets; x/y positions are internally consistent with the log-log scale and the surrounding anchors checked.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[compare-outputs --profile common
--build-mode optimized] --> B[Raw timing pair
Provenant vs ScanCode]
    B --> C[Compute speedup ratio
ScanCode ÷ Provenant]
    C --> D[Write BENCHMARKS.md entry
Files · Run context · Timing · Notes]
    D --> E[Update corpus stats line
252 of 252 runs faster]
    D --> F[generate-benchmark-chart]
    F --> G[scan-duration-vs-files.svg
new rect + circle per target]
    subgraph New entries this PR
        H[eclipse-vertx/vert.x
EPL-2.0 AND Apache-2.0 · 20.94×]
        I[glfw/glfw
Zlib · 17.15×]
        J[mozilla/glean
MPL-2.0 · 20.46×]
    end
    D --> H
    D --> I
    D --> J

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
flowchart TD
    A[compare-outputs --profile common
--build-mode optimized] --> B[Raw timing pair
Provenant vs ScanCode]
    B --> C[Compute speedup ratio
ScanCode ÷ Provenant]
    C --> D[Write BENCHMARKS.md entry
Files · Run context · Timing · Notes]
    D --> E[Update corpus stats line
252 of 252 runs faster]
    D --> F[generate-benchmark-chart]
    F --> G[scan-duration-vs-files.svg
new rect + circle per target]
    subgraph New entries this PR
        H[eclipse-vertx/vert.x
EPL-2.0 AND Apache-2.0 · 20.94×]
        I[glfw/glfw
Zlib · 17.15×]
        J[mozilla/glean
MPL-2.0 · 20.46×]
    end
    D --> H
    D --> I
    D --> J

Loading

_{Reviews (1): Last reviewed commit: "docs(benchmarks): record Zlib, MPL-2.0, ..." | Re-trigger Greptile}

[greptile-apps]

Speedup figure doesn't match the documented timings

The stated 17.15× speedup for glfw/glfw cannot be derived from the timings on the next lines: 88.79 ÷ 5.18 = 17.142…, which rounds to 17.14×, not 17.15×. The same figure appears in the PR-description table. If the actual measurements were taken at higher precision than the two-decimal-place values recorded here (e.g. Provenant 5.175 s → 88.79 ÷ 5.175 = 17.15×), the stored timing value rounds too coarsely to reproduce the advertised ratio; otherwise the header multiplier is off by 0.01×.