fix(copyright): keep a trailing acronym-with-period holder (e.g. CERN.)

[mstykow]

Summary

• Fixes a copyright-detection bug where the holder was dropped from a statement of the shape Copyright © <year> <ACRONYM>. when prose preceded the marker on the same line.
• Root cause: an all-caps acronym with a sentence-final period (CERN.) lexes as a Pn token, whereas its period-less spelling (CERN) lexes as Caps. A bare Caps reduces to a NameCaps node and is absorbed onto a year-only copyright via the strong_first path in should_start_absorbing; a bare trailing Pn matched no absorption branch and was dropped, yielding a holder-less Copyright (c) <year>.
• Isolated lines were rescued only by the line-anchored extend_year_only_copyrights_with_trailing_text repair (regex anchored at ^copyright), which cannot fire when prose precedes the copyright — so any real-world file with the copyright mid-line lost the holder.
• Observed on flask tests/test_cli.py @ 22d924701a6ae2e4cd01e9a15bbaf3946094af65 (... Revised BSD License. Copyright © 2015 CERN.): reported Copyright © 2015 with no holder; now reports Copyright © 2015 CERN. with holder CERN.
• Fix: in should_start_absorbing, absorb a lone trailing Pn leaf on the same line as a holder-less, year-bearing copyright clause, mirroring the existing Caps/NameCaps treatment so the two source-faithful spellings behave consistently. Scoped to a year-only clause so it cannot swallow ordinary prose.

Scope and exclusions

• Included: general absorption fix for a trailing Pn acronym holder; detector regression tests.
• Explicit exclusions: no lexer retagging of CERN. (kept as Pn, which is still correct in name contexts such as Name Pn); no change to the line-anchored repair heuristic.

How to verify

• Scan a file whose copyright is preceded by prose on the same line, e.g. a line containing its Revised BSD License. Copyright © 2015 CERN. with --copyright; observe the holder CERN is captured and the statement keeps CERN..
• Edge cases exercised manually: Copyright © 2015 CERN. All rights reserved. (rights-reserved stripped from holder, holder still CERN); Copyright © 2015 U.S. Government (multi-token name unaffected); a bare year-only Copyright © 2015 after prose (correctly yields no holder — no over-capture of surrounding prose).

Expected-output fixture changes

• Files changed: none. No golden/expected fixtures shifted; the existing copyright golden suite (copyright_golden), scanner_copyright_credits, post_processing_golden, and the copyright:: unit tests all pass unchanged. New coverage is added as unit tests in src/copyright/detector/tests.rs.

[greptile-apps]

Greptile Summary

Fixes copyright extraction for same-line, trailing acronym holders with sentence-final periods.

• Absorbs a standalone Pn token following a holder-less, year-bearing copyright clause.
• Adds regression coverage for CERN., existing holder variants, and genuinely holder-less notices.

Confidence Score: 5/5

The PR appears safe to merge with no actionable defects identified.

The new absorption path is constrained to same-line Pn tokens after holder-less year-bearing clauses, and the added tests verify both the intended acronym recovery and preservation of genuinely holder-less output.

Important Files Changed

Filename	Overview
src/copyright/detector/tree_walk/copyright/absorb.rs	Narrowly extends trailing-token absorption to same-line Pn leaves after year-only copyright clauses.
src/copyright/detector/tests.rs	Adds focused regression and guard tests covering acronym holders and holder-less notices.

_{Reviews (1): Last reviewed commit: "fix(copyright): keep a trailing acronym-..." | Re-trigger Greptile}