diff --git a/tests/fixtures/comments/english.tsv b/tests/fixtures/comments/english.tsv
new file mode 100644
index 0000000..8c526db
--- /dev/null
+++ b/tests/fixtures/comments/english.tsv
@@ -0,0 +1,301 @@
+lang	text	origin
+en	"string", "char" (a literal of one character) or "none"	tools/count_loc.py
+en	(ADR-0008). The workflow checks out the calling repository, fetches this file	tools/quality-checks-cpp.sh
+en	(MIL-004 task 6). The repository is public, so no token is needed.	.gitea/scoped_workflows/sync-github-metadata.yml
+en	(compile-commands-dir) are optional.	.gitea/workflows/quality-checks-cpp.yml
+en	(tools/publish_wiki.py), fetched at tools-ref.	.gitea/workflows/publish-wiki-pages.yml
+en	----------------------------------------------------------------- helpers	tests/test_quality_workflows_cpp_csharp.py
+en	--------------------------------------------------------------------- C#	tests/test_quality_workflows_cpp_csharp.py
+en	--------------------------------------------------------------------- C++	tests/test_quality_workflows_cpp_csharp.py
+en	.c files are C; everything else (including headers) is checked as C++.	tools/quality-checks-cpp.sh
+en	.gitmodules uses ssh:// URLs and the runner has no SSH key, so the	.gitea/workflows/sync-milestones.yml
+en	A directory with a `python3` that runs the interpreter of the tests.	tests/test_code_metrics.py
+en	A failure with a category.	tools/publish_wiki.py
+en	A runner must be registered for the repository (or its organisation).	.gitea/workflows/sync-milestones.yml
+en	A script module, readable and writable by item and by attribute.	tests/workflow_source.py
+en	A script run: the finished process and the tool calls the stubs logged.	tests/test_quality_workflows_cpp_csharp.py
+en	A source file or section needed for a page is missing.	tools/build_wiki.py
+en	A translation has the same code blocks, table rows and heading levels.	tests/test_build_wiki.py
+en	A wiki that already has pages named with spaces, as Gitea stores them.	tests/test_publish_wiki.py
+en	ADR-0009, MIL-009: the scoped publish-wiki.yml and the marker are removed.	tests/test_publish_wiki.py
+en	An exact stable version of the .NET SDK, 10 or later (no channel, no range).	tools/quality-checks-csharp.sh
+en	An input is invalid; the message is printed as a `[configuration]` failure.	tools/badges.py
+en	Answers like the Gitea repository and wiki endpoints.	tests/test_publish_wiki.py
+en	Answers the calls of the release script and records the writes.	tests/test_release.py
+en	Black or white, whichever has the higher contrast on the background.	tools/badges.py
+en	Build the English wiki pages in ``docs/wiki/`` from ``README.md`` and ``guides/``.	tools/build_wiki.py
+en	CLANG_TIDY_VERSION, QC_VENV (and QC_BIN for tests).	tools/quality-checks-cpp.sh
+en	COVERAGE_WARN, COVERAGE_GOOD, BADGES_BRANCH, BADGES_OUT (the directory of the badge files; a temporary	tools/code-metrics.sh
+en	Copy the badge files into the worktree and stage them; return 0 when	tools/code-metrics.sh
+en	Copy the description and topics of a Gitea repository to its GitHub push mirror.	tools/sync_github_metadata.py
+en	Count the code files and the lines of code of a directory (ADR-0010).	tools/count_loc.py
+en	Count the code files under directory; see the module docstring for the result.	tools/count_loc.py
+en	Create a release of this repository: version tag, release page, major tag.	tools/release.py
+en	Create and update the pages; delete the others only when prune is set.	tools/publish_wiki.py
+en	Create the release page (and with it the tag), then move the major tag.	tools/release.py
+en	Describe the shape of a token without revealing any of its characters.	tools/publish_wiki.py
+en	Empty on a push (a real run); "true" or "false" on a manual run.	.gitea/workflows/sync-milestones.yml
+en	Env: SHELLCHECK, SHFMT the commands to run (default: from the PATH)	tools/lint-action-scripts.sh
+en	Every failure is printed as ::error::[category] message.	tools/code-metrics.sh
+en	Every failure is printed as ::error::[category] message. A key is never printed.	actions/connect-wireguard/connect-wireguard.sh
+en	Exit: 0 every script is clean (or there is none); 1 at least one finding or	tools/lint-action-scripts.sh
+en	Gitea keeps a hyphen in a title literal (Home-da.-.md, address Home%2Dda).	tests/test_publish_wiki.py
+en	Gitea shows subpages only through _Sidebar, so it must reach every page.	tests/test_build_wiki.py
+en	H2 titles to keep; None keeps the file	tools/build_wiki.py
+en	Helpers that read workflow files and the Python embedded in them.	tests/workflow_source.py
+en	INSTALLING THIS FILE IS THE STANDING AUTHORIZATION TO PUBLISH. This project	.gitea/workflows/sync-milestones.yml
+en	Import a script file without running its `main` (its name is not `__main__`).	tests/workflow_source.py
+en	Import tools/changelog.py, which sits next to this file.	tools/release.py
+en	Import tools/publish_wiki.py.	tests/workflow_source.py
+en	Import tools/sync_github_metadata.py.	tests/workflow_source.py
+en	It checks the C and C++ files of the repository with clang-format and	.gitea/workflows/quality-checks-cpp.yml
+en	It checks the C# solution or project of the repository with `dotnet format`	.gitea/workflows/quality-checks-csharp.yml
+en	It checks the Python and shell files of the repository with pinned tools,	.gitea/workflows/quality-checks.yml
+en	It counts the code files and the lines of code of the repository (comment	.gitea/workflows/code-metrics.yml
+en	It needs read and write on issues and milestones, and read on the framework	.gitea/workflows/sync-milestones.yml
+en	It runs no test, needs no secret and writes only the badges branch.	.gitea/workflows/code-metrics.yml
+en	LANGUAGES, WORKING_DIRECTORY, RUFF_VERSION, MYPY_VERSION,	tools/quality-checks.sh
+en	LF on every platform	tests/test_lint_action_scripts.py
+en	Leading emoji and symbols in README headings; removed so anchors stay stable.	tools/build_wiki.py
+en	Link text of the sidebar: English, then one list per translation code.	tools/build_wiki.py
+en	Lint the shell scripts of the composite actions (MIL-012 task 6): shellcheck,	tools/lint-action-scripts.sh
+en	MIL-004 Go/No-Go criterion 5: every script is a file under tools/.	tests/test_workflows.py
+en	Markdown pages of wiki_dir to the wiki of the calling repository, creating	.gitea/workflows/publish-wiki-pages.yml
+en	Needs: bash, find, shellcheck, shfmt	tools/lint-action-scripts.sh
+en	No translation follows an older version of its English page.	tests/test_build_wiki.py
+en	Offline tests for the composite action actions/connect-tailscale.	tests/test_connect_tailscale.py
+en	Offline tests for the composite action actions/connect-wireguard.	tests/test_connect_wireguard.py
+en	Offline tests for the parts of tools/ that the flow tests do not reach.	tests/test_tool_entrypoints.py
+en	Offline tests for the reusable C++ and C# quality-check workflows (ADR-0008).	tests/test_quality_workflows_cpp_csharp.py
+en	Offline tests for the reusable code-metrics workflow and tools/code-metrics.sh (ADR-0010).	tests/test_code_metrics.py
+en	Offline tests for the reusable quality-check workflow (ADR-0005).	tests/test_quality_workflow.py
+en	Offline tests for tools/badges.py (ADR-0010): the coverage reader and the SVG renderer.	tests/test_badges.py
+en	Offline tests for tools/build_wiki.py.	tests/test_build_wiki.py
+en	Offline tests for tools/changelog.py and the CHANGELOG.md of this repository.	tests/test_changelog.py
+en	Offline tests for tools/count_loc.py (ADR-0010): hand-counted fixtures per language.	tests/test_count_loc.py
+en	Offline tests for tools/lint-action-scripts.sh (MIL-012 task 6).	tests/test_lint_action_scripts.py
+en	Offline tests for tools/publish_wiki.py and the publish-wiki workflows.	tests/test_publish_wiki.py
+en	Offline tests for tools/release.py and .gitea/workflows/release.yml.	tests/test_release.py
+en	Offline tests for tools/sync_github_metadata.py.	tests/test_sync_github_metadata.py
+en	On a push to main that changes docs/milestones/ or docs/sqa/reviews/, this	.gitea/workflows/sync-milestones.yml
+en	One `## [x]` section of the changelog.	tools/changelog.py
+en	One generated wiki page.	tools/build_wiki.py
+en	Publish accepted milestones to the git host.	.gitea/workflows/sync-milestones.yml
+en	Publish the pages of a folder to the Gitea wiki of the repository.	tools/publish_wiki.py
+en	Publishes the wiki of this repository from docs/wiki/ by calling the	.gitea/workflows/publish-own-wiki.yml
+en	Publishing an empty directory would delete the whole wiki.	tools/publish_wiki.py
+en	PyYAML reads `on` as True	tests/test_code_metrics.py
+en	QC_DOTNET_INSTALL_SCRIPT, QC_DOTNET for tests).	tools/quality-checks-csharp.sh
+en	Read a Cobertura coverage report and write the code metrics badges (ADR-0010).	tools/badges.py
+en	Read, check and cut release notes from CHANGELOG.md (ADR-0006).	tools/changelog.py
+en	Records calls and answers like the Gitea and GitHub REST APIs.	tests/test_sync_github_metadata.py
+en	Releases this repository (ADR-0006, ADR-0007, MIL-004 task 5). It is started	.gitea/workflows/release.yml
+en	Remove the leading emoji from a heading line.	tools/build_wiki.py
+en	Replace urllib.request.urlopen; set `.reply` to a body or an exception.	tests/test_tool_entrypoints.py
+en	Return (code, comment, blank) line counts of the text of a file.	tools/count_loc.py
+en	Return (exact tag, major tag, whether the major tag moves).	tools/release.py
+en	Return ``{file name: content}`` for every English page and the sidebar.	tools/build_wiki.py
+en	Return a flat two-part SVG badge; the width follows the number of characters.	tools/badges.py
+en	Return a message for every page that is missing or out of date.	tools/build_wiki.py
+en	Return a message for every translation that is missing or stale.	tools/build_wiki.py
+en	Return every tag name of the repository.	tools/release.py
+en	Return every workflow file in the known workflow directories.	tests/workflow_source.py
+en	Return one message per breach of the convention.	tools/changelog.py
+en	Return red below warn, yellow from warn to below good, green from good.	tools/badges.py
+en	Return the Markdown of the release page for ``version``.	tools/changelog.py
+en	Return the boolean of the WIKI_PRUNE input: true or false, nothing else.	tools/publish_wiki.py
+en	Return the full Markdown of one wiki page.	tools/build_wiki.py
+en	Return the identity of a page: Gitea stores a space as a hyphen.	tools/publish_wiki.py
+en	Return the index after the string that opens at start; the end of line if unclosed.	tools/count_loc.py
+en	Return the keys of the pages that exist in the wiki.	tools/publish_wiki.py
+en	Return the language of a file name by its extension, or None.	tools/count_loc.py
+en	Return the line coverage of a Cobertura XML report as a whole percent, half up.	tools/badges.py
+en	Return the line that links a page to its other languages.	tools/build_wiki.py
+en	Return the lines before the first H2 and the lines of each H2 section.	tools/build_wiki.py
+en	Return the navigation shown beside every wiki page.	tools/build_wiki.py
+en	Return the problems of one workflow's default `tools-ref`, or [].	tests/test_tools_ref.py
+en	Return the release page text for a version, or fail as a validation error.	tools/release.py
+en	Return the releases in file order; text before the first one is ignored.	tools/changelog.py
+en	Return the short content hash a translation records for its source.	tools/build_wiki.py
+en	Return the source of each `python3 - <<'PY'` heredoc in a workflow.	tests/workflow_source.py
+en	Return the title to send to Gitea for a page key.	tools/publish_wiki.py
+en	Return the version as a tuple, or raise a configuration error.	tools/release.py
+en	Return the wiki form of a link target found in ``source``.	tools/build_wiki.py
+en	Return the workflow files of this repository that a caller can use.	tests/test_tools_ref.py
+en	Return {page key: content} for every Markdown file in wiki_dir.	tools/publish_wiki.py
+en	Reusable workflow (ADR-0005). A repository calls it with	.gitea/workflows/quality-checks.yml
+en	Reusable workflow (ADR-0008). A repository calls it with	.gitea/workflows/quality-checks-cpp.yml
+en	Reusable workflow (ADR-0009, MIL-008). A repository calls it with	.gitea/workflows/publish-wiki-pages.yml
+en	Reusable workflow (ADR-0010). A repository calls it with	.gitea/workflows/code-metrics.yml
+en	Run a tool as `python tools/<name>` with only the given environment.	tests/test_tool_entrypoints.py
+en	Run it once by hand from the Actions page with dry_run left on: it prints	.gitea/workflows/sync-milestones.yml
+en	Run the generator; return the process exit code.	tools/build_wiki.py
+en	Run the script with a stub wg-quick that records its arguments.	tests/test_connect_wireguard.py
+en	SHELLCHECK_PY_VERSION, SHFMT_PY_VERSION, QC_VENV (and QC_BIN for tests).	tools/quality-checks.sh
+en	Scan one line; return (has_code, has_comment) and update the state.	tools/count_loc.py
+en	Scoped workflow: it runs in every repository of an organization. The script	.gitea/scoped_workflows/sync-github-metadata.yml
+en	Secret: TOKEN_FOR_GITEA, the bare Gitea token of the organization (no JSON).	.gitea/workflows/sync-milestones.yml
+en	Settings come from the environment, never from interpolated expressions:	tools/code-metrics.sh
+en	Static validation of every workflow file in this repository.	tests/test_workflows.py
+en	Steps of the composite action actions/connect-tailscale/action.yml.	actions/connect-tailscale/connect-tailscale.sh
+en	Steps of the composite action actions/connect-wireguard/action.yml.	actions/connect-wireguard/connect-wireguard.sh
+en	Steps of the reusable workflow .gitea/workflows/code-metrics.yml (ADR-0010).	tools/code-metrics.sh
+en	Steps of the reusable workflow .gitea/workflows/quality-checks-cpp.yml	tools/quality-checks-cpp.sh
+en	Steps of the reusable workflow .gitea/workflows/quality-checks-csharp.yml	tools/quality-checks-csharp.sh
+en	Steps of the reusable workflow .gitea/workflows/quality-checks.yml (ADR-0005).	tools/quality-checks.sh
+en	Strip heading emoji and rewrite links, leaving fenced code untouched.	tools/build_wiki.py
+en	Succeeds when one of the named files exists in the working directory or above	tools/quality-checks-cpp.sh
+en	TOKEN_FOR_GITEA_WIKI: ${{ secrets.TOKEN_FOR_GITEA_WIKI }}	.gitea/workflows/publish-wiki-pages.yml
+en	TS_ARGS is split on spaces on purpose: it holds flags, never a secret	actions/connect-tailscale/connect-tailscale.sh
+en	TS_AUTHKEY, TS_OAUTH_SECRET, TS_TAGS, TS_HOSTNAME, TS_VERSION, TS_USERSPACE,	actions/connect-tailscale/connect-tailscale.sh
+en	TS_SHA256, TS_ARGS, TS_TIMEOUT, TS_HEALTH_CHECK_HOST. TS_TAILSCALE,	actions/connect-tailscale/connect-tailscale.sh
+en	TS_TAILSCALED, TS_SUDO, TS_STATE_DIR, TS_START_WAIT (tries) and	actions/connect-tailscale/connect-tailscale.sh
+en	TS_WAIT_INTERVAL (seconds) name the commands, the work folder and the waits,	actions/connect-tailscale/connect-tailscale.sh
+en	The Home page names each workflow file of the repository and draws it.	tests/test_build_wiki.py
+en	The `if __name__ == "__main__"` block: exit code 1 and ERROR: on stderr.	tests/test_tool_entrypoints.py
+en	The auth key is passed to tailscale in a file of mode 600, never on a command	actions/connect-tailscale/connect-tailscale.sh
+en	The caller file is the opt-in; there is no marker file. It publishes the	.gitea/workflows/publish-wiki-pages.yml
+en	The comment and string rules of one language.	tools/count_loc.py
+en	The commit is the tip of the branch and its validation run is green.	tools/release.py
+en	The default `tools-ref` of every reusable workflow names the current major.	tests/test_tools_ref.py
+en	The exclude patterns of the caller; a pattern without `/` matches a name anywhere.	tools/count_loc.py
+en	The page a link to a source file should point to. The README is the home page.	tools/build_wiki.py
+en	The pages in wiki/ match what README.md and guides/ produce.	tests/test_build_wiki.py
+en	The reusable workflow sets WIKI_PRUNE: the caller is the opt-in (MIL-008).	tests/test_publish_wiki.py
+en	The wiki of this repository equals docs/wiki/ (ADR-0009, MIL-008 task 4).	tests/test_publish_wiki.py
+en	The workflow checks out the calling repository, fetches this file into	tools/code-metrics.sh
+en	Translate an exclude pattern: `*` stops at `/`, `**` crosses it, `?` is one character.	tools/count_loc.py
+en	Translations (ADR-0001): `<Page>.<code>` written by hand in docs/wiki/.	tools/build_wiki.py
+en	True when the tag holds the file; True when the tag is not in this checkout.	tests/test_tools_ref.py
+en	Two merges in a row queue instead of racing.	.gitea/workflows/sync-milestones.yml
+en	Usage: lint-action-scripts.sh [<directory>] (default: actions)	tools/lint-action-scripts.sh
+en	Use the user-space implementation when the kernel has no WireGuard module,	actions/connect-wireguard/connect-wireguard.sh
+en	WG_ALLOWED_IPS, WG_PRESHARED_KEY, WG_DNS, WG_KEEPALIVE, WG_INTERFACE,	actions/connect-wireguard/connect-wireguard.sh
+en	WG_CONFIG, WG_PRIVATE_KEY, WG_ADDRESS, WG_PEER_PUBLIC_KEY, WG_ENDPOINT,	actions/connect-wireguard/connect-wireguard.sh
+en	WG_HEALTH_CHECK_HOST. WG_DIR is the configuration directory (/etc/wireguard	actions/connect-wireguard/connect-wireguard.sh
+en	WORKING_DIRECTORY, EXCLUDE, COVERAGE_REPORT, COVERAGE_ARTIFACT,	tools/code-metrics.sh
+en	WORKING_DIRECTORY, PROJECT, DOTNET_SDK_VERSION, DOTNET_ROOT (and	tools/quality-checks-csharp.sh
+en	WORKING_DIRECTORY, STD, COMPILE_COMMANDS_DIR, CLANG_FORMAT_VERSION,	tools/quality-checks-cpp.sh
+en	Windows has no POSIX file modes	tests/test_connect_wireguard.py
+en	Without a .clang-tidy, clang-tidy enables no check and refuses to run, so	tools/quality-checks-cpp.sh
+en	Write every page; return the names of the files that changed.	tools/build_wiki.py
+en	Write the badge files and the metrics JSON; return the paths in a fixed order.	tools/badges.py
+en	`install` creates the venv and installs `==` versions; checked with a fake pip.	tests/test_quality_workflows_cpp_csharp.py
+en	`ip link add` fails in the test environment, so wireguard-go must exist	tests/test_connect_wireguard.py
+en	a breaking entry needs a new major compared with the release before it	tools/changelog.py
+en	a called workflow has its own runner	tests/test_workflows.py
+en	a daemon that already runs belongs to the host; this action would log it out	actions/connect-tailscale/connect-tailscale.sh
+en	a full configuration: mask the value of every PrivateKey and PresharedKey	actions/connect-wireguard/connect-wireguard.sh
+en	a line break would add directives (PostUp runs as root) to the file	actions/connect-wireguard/connect-wireguard.sh
+en	a missing tool	tools/lint-action-scripts.sh
+en	a sudo that resets the environment, as the real one does	tests/test_connect_wireguard.py
+en	a wrapped item: the lines that continue it belong to the same item	tools/changelog.py
+en	a.py has one code line, run.sh one	tests/test_code_metrics.py
+en	an entry for the version. dry_run defaults to true: it checks and prints and	.gitea/workflows/release.yml
+en	and `dotnet build` (warnings as errors) on an exact .NET SDK 10 or later,	.gitea/workflows/quality-checks-csharp.yml
+en	and blank lines are not counted), reads an optional Cobertura XML coverage	.gitea/workflows/code-metrics.yml
+en	and nothing when none of them changed	tools/code-metrics.sh
+en	and shfmt in the format .editorconfig sets (two spaces, indented case labels).	tools/lint-action-scripts.sh
+en	and updating pages, and deleting pages that are not in the folder only when	.gitea/workflows/publish-wiki-pages.yml
+en	branches: [ main ]	.gitea/workflows/publish-wiki-pages.yml
+en	branches: [main]	.gitea/workflows/code-metrics.yml
+en	by default); WG_QUICK, WG_SUDO and WG_PING name the commands, for the tests.	actions/connect-wireguard/connect-wireguard.sh
+en	by hand with the version to release; it creates the version tag, the release	.gitea/workflows/release.yml
+en	caller a project copies from framework/templates/ci/publish-wiki.yml, with	.gitea/workflows/publish-own-wiki.yml
+en	changes no file and needs no secret.	.gitea/workflows/quality-checks.yml
+en	changes no file and needs no secret. The package restore needs the network.	.gitea/workflows/quality-checks-csharp.yml
+en	clang-tidy at exact versions, changes no file and needs no secret. The	.gitea/workflows/quality-checks-cpp.yml
+en	clang-tidy exits 0 on a warning, so every finding is made an error: a	tools/quality-checks-cpp.sh
+en	code-metrics.sh measure count the code, read the coverage report and	tools/code-metrics.sh
+en	code-metrics.sh publish commit the badge files to the badges branch,	tools/code-metrics.sh
+en	code-metrics.sh validate check the inputs, before anything is counted	tools/code-metrics.sh
+en	command line.	.gitea/workflows/sync-milestones.yml
+en	commit is the tip of main, its validation run is green and the changelog has	.gitea/workflows/release.yml
+en	compile-commands-dir is relative to the repository root, not to the working directory	tools/quality-checks-cpp.sh
+en	configuration the same way.	tools/quality-checks-cpp.sh
+en	connect-tailscale.sh down log the node out and stop tailscaled	actions/connect-tailscale/connect-tailscale.sh
+en	connect-tailscale.sh mask print ::add-mask:: for every secret input	actions/connect-tailscale/connect-tailscale.sh
+en	connect-tailscale.sh up install tailscale if needed, start tailscaled and join the tailnet	actions/connect-tailscale/connect-tailscale.sh
+en	connect-wireguard.sh down take the tunnel down and delete the configuration	actions/connect-wireguard/connect-wireguard.sh
+en	connect-wireguard.sh mask print ::add-mask:: for every secret input	actions/connect-wireguard/connect-wireguard.sh
+en	connect-wireguard.sh up write the configuration and bring the tunnel up	actions/connect-wireguard/connect-wireguard.sh
+en	contents: write	.gitea/workflows/code-metrics.yml
+en	deleted. Each run updates the issues of every accepted milestone (the sync is	.gitea/workflows/sync-milestones.yml
+en	directory under RUNNER_TEMP by default).	tools/code-metrics.sh
+en	env keeps the userspace variable through sudo; the file holds the key, so	actions/connect-wireguard/connect-wireguard.sh
+en	finding must fail the check, as a failing tool does for the other languages.	tools/quality-checks-cpp.sh
+en	for the tests.	actions/connect-tailscale/connect-tailscale.sh
+en	idempotent, but the host records the edit).	.gitea/workflows/sync-milestones.yml
+en	installed it on 2026-10-11 at the request of S01.	.gitea/workflows/sync-milestones.yml
+en	into shared-scripts/ and runs one command per step:	tools/quality-checks-cpp.sh
+en	is a file of TirSystem/github-action (tools/sync_github_metadata.py), fetched	.gitea/scoped_workflows/sync-github-metadata.yml
+en	is created or updated as a Milestone with its tasks as Issues. A Proposed	.gitea/workflows/sync-milestones.yml
+en	it is removed when the interface does not come up	actions/connect-wireguard/connect-wireguard.sh
+en	it, up to the top of the repository: clang-format and clang-tidy find their	tools/quality-checks-cpp.sh
+en	latest Version History row is Accepted and whose latest review record is a Go	.gitea/workflows/sync-milestones.yml
+en	line, and nothing else, so the rest of the log stays readable	actions/connect-wireguard/connect-wireguard.sh
+en	line. Every failure is printed as ::error::[category] message.	actions/connect-tailscale/connect-tailscale.sh
+en	metrics:	.gitea/workflows/code-metrics.yml
+en	milestone, or one without a Go review, is skipped. Nothing is closed or	.gitea/workflows/sync-milestones.yml
+en	name -> (source, code, comment, blank), counted by hand.	tests/test_count_loc.py
+en	names that start with GITEA_ or GITHUB_.	.gitea/workflows/sync-milestones.yml
+en	no final newline	tests/test_count_loc.py
+en	often has no sudo. TS_SUDO overrides it (empty for none).	actions/connect-tailscale/connect-tailscale.sh
+en	often has no sudo. WG_SUDO overrides it (empty for none).	actions/connect-wireguard/connect-wireguard.sh
+en	on failure the captured output is the ::error:: line, which must be shown	actions/connect-wireguard/connect-wireguard.sh
+en	only a daemon this action started is logged out and stopped	actions/connect-tailscale/connect-tailscale.sh
+en	only the key values are masked	tests/test_connect_wireguard.py
+en	only the probe ran: no second daemon, no up	tests/test_connect_tailscale.py
+en	outside this repository, so it gets a copy of the repository's file	tests/test_lint_action_scripts.py
+en	page from the changelog entry and moves the major tag, and only when the	.gitea/workflows/release.yml
+en	parsed YAML has no fixed shape	tests/test_quality_workflows_cpp_csharp.py
+en	paths: [ "docs/wiki/**" ]	.gitea/workflows/publish-wiki-pages.yml
+en	permissions:	.gitea/workflows/code-metrics.yml
+en	pretend the kernel has WireGuard and the tools are present	tests/test_connect_wireguard.py
+en	prune is true. The script is a file of this repository	.gitea/workflows/publish-wiki-pages.yml
+en	quality-checks-cpp.sh install install the exact clang tool versions into a venv	tools/quality-checks-cpp.sh
+en	quality-checks-cpp.sh run run clang-format and clang-tidy on the tracked files	tools/quality-checks-cpp.sh
+en	quality-checks-cpp.sh validate check the inputs, before anything is installed	tools/quality-checks-cpp.sh
+en	quality-checks-csharp.sh install install the exact .NET SDK version	tools/quality-checks-csharp.sh
+en	quality-checks-csharp.sh run run dotnet format and dotnet build on the project	tools/quality-checks-csharp.sh
+en	quality-checks-csharp.sh validate check the inputs, before anything is installed	tools/quality-checks-csharp.sh
+en	quality-checks.sh install install the pinned tools into a virtual environment	tools/quality-checks.sh
+en	quality-checks.sh run run every selected tool on the tracked files	tools/quality-checks.sh
+en	quality-checks.sh validate check the inputs, before anything is installed	tools/quality-checks.sh
+en	quality:	.gitea/workflows/quality-checks-cpp.yml
+en	real/a.py only, not link.py	tests/test_count_loc.py
+en	render_config: the wg-quick configuration on stdout	actions/connect-wireguard/connect-wireguard.sh
+en	report (an artifact of an earlier job of the caller, or a file of the	.gitea/workflows/code-metrics.yml
+en	repository needs a .clang-format; a .clang-tidy and a compile database	.gitea/workflows/quality-checks-cpp.yml
+en	repository) and commits three SVG badges to a branch of the calling	.gitea/workflows/code-metrics.yml
+en	repository, which is checked out as a submodule. Gitea does not accept secret	.gitea/workflows/sync-milestones.yml
+en	repository.	.gitea/workflows/code-metrics.yml
+en	reusable workflow of this repository (ADR-0009, MIL-008 task 4). It is the	.gitea/workflows/publish-own-wiki.yml
+en	secrets:	.gitea/workflows/publish-wiki-pages.yml
+en	shared-scripts/ and runs one command per step:	tools/code-metrics.sh
+en	shellcheck disable=SC2206	actions/connect-tailscale/connect-tailscale.sh
+en	shfmt reads .editorconfig from the folder upwards; a temporary folder is	tests/test_lint_action_scripts.py
+en	something differs from the branch.	tools/code-metrics.sh
+en	stop at the top of the repository, or at the file system root	tools/quality-checks-cpp.sh
+en	submodules (framework, and framework/qc inside it) are fetched over HTTPS	.gitea/workflows/sync-milestones.yml
+en	sudo is used only when the job is not root: a job container runs as root and	actions/connect-tailscale/connect-tailscale.sh
+en	the command is required	tests/test_tool_entrypoints.py
+en	the plan and creates nothing. Turn dry_run off only for a real run.	.gitea/workflows/sync-milestones.yml
+en	the project gets the compiler diagnostics and the static analyzer.	tools/quality-checks-cpp.sh
+en	the stub records its arguments and, for `up`, the content of the key file	tests/test_connect_tailscale.py
+en	the tools are in the virtual environment of the interpreter that runs the tests	tests/test_lint_action_scripts.py
+en	tools-ref is main so that the script that is run is the one of this commit.	.gitea/workflows/publish-own-wiki.yml
+en	two differences: the reusable workflow is called by its local path, and	.gitea/workflows/publish-own-wiki.yml
+en	type: ignore[index]	tools/changelog.py
+en	type: ignore[return-value]	tools/changelog.py
+en	uses: TirSystem/github-action/.gitea/workflows/code-metrics.yml@v2	.gitea/workflows/code-metrics.yml
+en	uses: TirSystem/github-action/.gitea/workflows/publish-wiki-pages.yml@v1	.gitea/workflows/publish-wiki-pages.yml
+en	uses: TirSystem/github-action/.gitea/workflows/quality-checks-cpp.yml@v1	.gitea/workflows/quality-checks-cpp.yml
+en	uses: TirSystem/github-action/.gitea/workflows/quality-checks-csharp.yml@v1	.gitea/workflows/quality-checks-csharp.yml
+en	uses: TirSystem/github-action/.gitea/workflows/quality-checks.yml@v1	.gitea/workflows/quality-checks.yml
+en	validate-workflows.yml measures tools/ and fails below the .coveragerc limit.	tests/test_workflows.py
+en	validate-workflows.yml runs it on every pull request and push to main.	tools/lint-action-scripts.sh
+en	variables that must reach wg-quick through sudo, which resets the environment	actions/connect-wireguard/connect-wireguard.sh
+en	wait_for_daemon: tailscaled needs a moment to create its socket	actions/connect-tailscale/connect-tailscale.sh
+en	which is the usual case in a Gitea runner container.	actions/connect-wireguard/connect-wireguard.sh
+en	with a checkout so that it is readable, lintable and tested as a file	.gitea/scoped_workflows/sync-github-metadata.yml
diff --git a/tests/fixtures/comments/foreign.tsv b/tests/fixtures/comments/foreign.tsv
new file mode 100644
index 0000000..f775cfe
--- /dev/null
+++ b/tests/fixtures/comments/foreign.tsv
@@ -0,0 +1,139 @@
+lang	text	origin
+da	hent brugeren fra databasen	written
+da	Denne funktion beregner totalen for kurven	written
+da	vi skal huske at lukke forbindelsen bagefter	written
+da	TODO: ret fejlen når listen er tom	written
+da	returnerer sandt hvis filen findes	written
+da	Gem resultatet i en midlertidig mappe	written
+da	her tjekker vi om brugeren har adgang	written
+da	læs konfigurationen fra miljøvariablerne	written
+da	hvis der ikke er noget svar, prøver vi igen	written
+da	Opret en ny forbindelse til serveren	written
+da	dette er en midlertidig løsning indtil videre	written
+da	loop gennem alle filerne og tæl linjerne	written
+da	Vi bruger en cache for at spare tid	written
+da	fejlbesked skal vises til brugeren	written
+da	sorter listen efter dato, nyeste først	written
+da	ingen grund til at gentage kaldet	written
+da	Kontroller at tokenet ikke er udløbet	written
+da	slet de gamle filer før vi starter	written
+da	Her håndterer vi alle undtagelser	written
+da	afrund til nærmeste hele tal	written
+da	venter på at serveren bliver klar	written
+da	Parametrene er valgfrie, standardværdien er nul	written
+da	kald funktionen to gange for at være sikker	written
+da	skriv loggen til disken hver time	written
+da	Det her virker kun på Linux	written
+da	vi springer tomme linjer over	written
+da	opdater tælleren når en ny post kommer ind	written
+da	fjern dubletter fra resultatet	written
+da	Brug en kø så arbejdet ikke blokerer	written
+da	den øverste mappe i projektet	written
+da	gennemløb træet fra roden og ned	written
+da	Denne klasse holder styr på sessionerne	written
+da	hvis navnet er tomt, brug standardnavnet	written
+da	tjek om porten er optaget	written
+da	Beregn gennemsnittet af de sidste ti målinger	written
+da	vi skal have en bedre løsning her senere	written
+da	Sæt tiden til midnat	written
+da	læser hele filen ind i hukommelsen	written
+da	Pas på, her kan der opstå en kapløbstilstand	written
+da	flyt filen til arkivmappen	written
+da	Dette er kun til test	written
+da	slå op i tabellen efter navn	written
+da	henter næste side af resultaterne	written
+da	hvorfor gør vi det her? Fordi serveren kræver det	written
+da	gem ændringerne og luk vinduet	written
+ms	dapatkan pengguna daripada pangkalan data	written
+ms	Fungsi ini mengira jumlah bagi troli	written
+ms	kita perlu ingat untuk menutup sambungan selepas itu	written
+ms	TODO: betulkan ralat apabila senarai kosong	written
+ms	mengembalikan benar jika fail wujud	written
+ms	Simpan hasil dalam folder sementara	written
+ms	di sini kita semak sama ada pengguna mempunyai akses	written
+ms	baca konfigurasi daripada pemboleh ubah persekitaran	written
+ms	jika tiada jawapan, kita cuba lagi	written
+ms	Buat sambungan baharu ke pelayan	written
+ms	ini penyelesaian sementara buat masa ini	written
+ms	ulang semua fail dan kira baris	written
+ms	Kita guna cache untuk jimat masa	written
+ms	mesej ralat mesti dipaparkan kepada pengguna	written
+ms	susun senarai mengikut tarikh, yang terbaru dahulu	written
+ms	tidak perlu mengulang panggilan	written
+ms	Semak bahawa token belum tamat tempoh	written
+ms	padam fail lama sebelum kita mulakan	written
+ms	Di sini kita mengendalikan semua pengecualian	written
+ms	bundarkan kepada nombor bulat yang terdekat	written
+ms	menunggu pelayan sedia	written
+ms	Parameter adalah pilihan, nilai lalai ialah sifar	written
+ms	panggil fungsi dua kali untuk pastikan	written
+ms	tulis log ke cakera setiap jam	written
+ms	Ini hanya berfungsi pada Linux	written
+ms	kita langkau baris kosong	written
+ms	kemas kini pembilang apabila rekod baharu masuk	written
+ms	buang pendua daripada hasil	written
+ms	Guna giliran supaya kerja tidak menyekat	written
+ms	folder teratas dalam projek	written
+ms	lalui pokok dari akar ke bawah	written
+ms	Kelas ini menjejaki sesi	written
+ms	jika nama kosong, guna nama lalai	written
+ms	semak sama ada port sedang digunakan	written
+ms	Kira purata sepuluh ukuran terakhir	written
+ms	kita perlukan penyelesaian yang lebih baik nanti	written
+ms	Tetapkan masa kepada tengah malam	written
+ms	membaca seluruh fail ke dalam memori	written
+ms	Awas, di sini boleh berlaku keadaan perlumbaan	written
+ms	alihkan fail ke folder arkib	written
+ms	Ini hanya untuk ujian	written
+ms	cari dalam jadual mengikut nama	written
+ms	mendapatkan halaman seterusnya bagi hasil	written
+ms	kenapa kita buat begini? Kerana pelayan memerlukannya	written
+ms	simpan perubahan dan tutup tetingkap	written
+de	Hole den Benutzer aus der Datenbank	written
+de	Diese Funktion berechnet die Summe des Warenkorbs	written
+de	wir müssen die Verbindung danach schließen	written
+de	Gibt wahr zurück, wenn die Datei existiert	written
+de	Speichere das Ergebnis in einem temporären Ordner	written
+de	hier prüfen wir, ob der Benutzer Zugriff hat	written
+de	Lies die Konfiguration aus den Umgebungsvariablen	written
+de	wenn keine Antwort kommt, versuchen wir es erneut	written
+de	Das ist nur eine vorläufige Lösung	written
+de	Fehlermeldung muss dem Benutzer angezeigt werden	written
+de	lösche die alten Dateien, bevor wir anfangen	written
+de	Achtung, hier kann ein Wettlauf entstehen	written
+sv	hämta användaren från databasen	written
+sv	Den här funktionen räknar ut summan av varukorgen	written
+sv	vi måste stänga anslutningen efteråt	written
+sv	returnerar sant om filen finns	written
+sv	Spara resultatet i en tillfällig mapp	written
+sv	här kontrollerar vi om användaren har åtkomst	written
+sv	läs konfigurationen från miljövariablerna	written
+sv	om det inte kommer något svar försöker vi igen	written
+sv	Det här är bara en tillfällig lösning	written
+sv	felmeddelandet ska visas för användaren	written
+sv	ta bort de gamla filerna innan vi börjar	written
+sv	Se upp, här kan det uppstå en tävlingssituation	written
+no	hent brukeren fra databasen	written
+no	Denne funksjonen beregner summen av handlekurven	written
+no	vi må huske å lukke forbindelsen etterpå	written
+no	returnerer sant hvis filen finnes	written
+no	Lagre resultatet i en midlertidig mappe	written
+no	her sjekker vi om brukeren har tilgang	written
+no	les konfigurasjonen fra miljøvariablene	written
+no	hvis det ikke kommer noe svar, prøver vi igjen	written
+no	Dette er bare en midlertidig løsning	written
+no	feilmeldingen skal vises til brukeren	written
+no	slett de gamle filene før vi begynner	written
+no	Pass på, her kan det oppstå en kappløpstilstand	written
+fr	récupère l'utilisateur depuis la base de données	written
+fr	Cette fonction calcule le total du panier	written
+fr	il faut penser à fermer la connexion ensuite	written
+fr	renvoie vrai si le fichier existe	written
+fr	Enregistre le résultat dans un dossier temporaire	written
+fr	ici on vérifie si l'utilisateur a accès	written
+fr	lit la configuration depuis les variables d'environnement	written
+fr	s'il n'y a pas de réponse, on réessaie	written
+fr	Ce n'est qu'une solution provisoire	written
+fr	le message d'erreur doit être affiché à l'utilisateur	written
+fr	supprime les anciens fichiers avant de commencer	written
+fr	Attention, une condition de concurrence peut survenir ici	written