contributing: Use SPDX copyright tags in unlabeled file headers

[wenzeslaus]

Converts the 1030 files whose copyright notice does not use the labeled COPYRIGHT field: Doxygen file blocks in the C libraries, Python module docstrings in the GUI and the Python packages, and assorted headers. Together with #7743 this covers the whole tree apart from vendored code and the cases listed below.

• Years and names preserved as is; the collective line and the license identifier follow the same rules as contributing: Use SPDX copyright tags in file headers #7743.
• Vendored code excluded by directory and by the holder named in the notice.
• Translation catalogs under locale excluded as a standing decision.
• Multi-holder notices and notices whose license text is separated from them by an author list are left for reviewed one-off replacements in a follow-up.

A typical replacement:

  *
  * Dynamic balanced k-d tree implementation
  *
- * (C) 2014 by the GRASS Development Team
- *
- * This program is free software under the GNU General Public License
- * (>=v2).  Read the file COPYING that comes with GRASS for details.
+ * SPDX-FileCopyrightText: 2014 Other GRASS authors
+ * SPDX-License-Identifier: GPL-2.0-or-later
  *
  * \author Markus Metz

The policy decisions were made during GRASS Community Meeting 2026 in San Michele all'Adige.

Like #7743, this should technically merge after #7749, which defines "Other GRASS authors", but it does not have to.

The conversion was performed by a deterministic script, not by a language model. The script was extended for this corpus and this description drafted with AI assistance (Claude Code with Opus 4.8 and Fable 5); the decisions are the author's.

Details: scope, rules, decisions, deferred cases, and how it was checked (AI generated)

Decisions beyond #7743

• Files whose notice carries no license text at all get the identifier when the notice is team-held, covered by the repository-wide grant in COPYING, or when all its years fall in the CERL era, in or before 1992. Individual-held notices from 1993 on with no license text are left alone and reported.
• The locale/po exclusion is a standing decision: the gettext headers already state the license, distributed under the same license as the GRASS package, and they are rewritten by the translation platform sync, so an edit there would not survive.
• Some headers spell the license paragraph as Licence:-labeled text (imagery testsuites) or mark it with a Doxygen \copyright tag (lib/rst); both are recognized as the same GPL paragraph and converted.

Scope

	files
converted	1030
ambiguous holders, follow-up	63
license text separated by an author list, follow-up	54
locale/po catalogs, standing exclusion	85
vendored by path or holder	211
individual holder, 1993 or later, no license text	7
not a notice, not UTF-8, no anchor, or lint-conflicted	12
baseline: files with a notice and no COPYRIGHT field	1462

Rules

Same as #7743: one holder per SPDX-FileCopyrightText line, named holders first, the collective line last, the identifier after all copyright lines, years and names verbatim, the collective line carries years only where it is the sole copyright line, and neither AUTHOR(S) nor the git history was consulted. Author tags (\author, @author) and all other header text outside the notice and license paragraph are untouched. Files half-converted earlier, carrying an identifier but a legacy notice, are completed without duplicating the identifier.

Checked

Every added line is one of the two SPDX shapes and every removed line is a known notice or license-paragraph form (2246 added, 4649 removed, none unclassifiable). No byte outside the replaced span changed, trailing bytes are identical, the four-digit years on notice lines are conserved exactly, no emitted C line exceeds 78 columns, and a second run converts nothing. pre-commit over all 1030 files passes with nothing reformatted. All C compiles; all 399 converted Python files parse. 43 tests cover the script, including the wrapped-holder, \copyright-tag, and interposed-author-list cases.

One file, man/build_check.py, is deferred because the pinned ruff hook rewrites it on the unmodified base already, so converting it would pull an unrelated lint fix into the diff.

The script (spdx_headers.py.txt), its tests (spdx_headers_test.py.txt), and the full deferred report (deferred-bare.md) are attached.

deferred-bare.md
spdx_headers_test.py.txt
deferred-cases.md

[cwhite911]

Misses author

[cwhite911]

Should these authors be picked up?

[cwhite911]

Here is an example of the author getting picked up by the script.

[cwhite911]

Missed authors.

[cwhite911]

Another type of missed author

[cwhite911]

Missed authors

[cwhite911]

@wenzeslaus overall, the conversion worked well but I found a few examples that were not properly picked up.

I didn't comment on every instance where I found an issue, but if you look at the comments they should capture the 3-4 variants that need to be addressed.

I'm on my phone otherwise I would have added the patterns myself here.

Some of these may have been picked up in the other PRs.