f9d6b57f0f
The docs lived on one lineage and the fleet code on another. That is why six pages described a shared fleet key as how an agent authenticates, why COLLECTOR-AUTH had to warn that the per-PC credential code was not in the repository it sat in, and why ADR-019 recorded a decision whose implementation was somewhere else. Documentation that cannot see the code it documents goes stale by construction, and no amount of editing fixes that. So the corpus moves here: 84 files, 56 pages, 20 ADRs, the proposals, INDEX.md, llms.txt and both gates. Filenames are unchanged, because a wiki page name is derived from the filename and a rename breaks every link anybody saved. THE POINT OF THE MOVE IS THE CORRECTION IT ENABLES. Every caveat that said "not in this tree" was false the moment the files landed beside the code, so each was rewritten from what is actually here: - COLLECTOR-AUTH gained the real enroll request body, the two bounds on the unauthenticated route (per-IP rate limit, pending cap with stale rows dropped before a real PC is refused), and the rejected pending reason that binds to the endpoint identity rather than the nonce. - ADR-019 now describes an implementation that is present, including its tests, and says plainly that what remains is the FLEET migration, not the code. - proposals/device-identity claimed phase 1 was unbuilt. plugins/computers declares deviceid on the collector schema and prefers it over the older fallbacks. Corrected in the page, its summary and the index. WINDOWS/IIS IS NOW DOCUMENTED AS THE LEGACY PATH, per the product decision, and carefully: it is what production runs today. Every Windows page opens by saying it is supported, not for a new site, and where a new site should go instead. No procedure was cut - an operator on an IIS box at 2am needs these pages to work exactly as well tomorrow. MIGRATE-TOPOLOGY stops being a footnote and becomes the off-ramp, which is why it nearly doubled: what was there did not match how the Docker stack really works here. The Docker material that already lived on this line was kept and given ownership rather than buried under the ported Windows-centric prose: DEPLOY now describes the four services, the migrate service that runs the upgrade chain before api starts, TLS terminated by nginx, and the plugin count corrected from thirteen to seventeen - a number both lineages had wrong. ROADMAP is rewritten from evidence rather than memory: what has shipped, read out of the changelog, the ADRs and the manifests, and ahead of it exactly two items - finalize a distributed release and polish the setup wizard. PROJECT-MAP is generated and was stale; the GENERATOR was what was wrong. --check passes now. API-REFERENCE is retired to a pointer. /api/docs serves 573 operations with derived auth, query parameters and harvested examples, and a hand-written page competing with that is a liability. The file stays so its wiki page and its inbound links survive. AND THE MOVE TURNED ON A SCRUB THAT HAD NEVER RUN HERE. test_docs_publishable keys on docs/*.md, which this line did not have, so it skipped - its own comment predicted what would surface. Porting the corpus activated it and it found 48 lines that must not reach a published wiki: branch names abbreviating the site, assistant-tooling paths, the site name in two deployment pages, and an employee ID in a Windows path that the scrub does not even look for.
127 lines
6.4 KiB
Python
127 lines
6.4 KiB
Python
"""docs/ is published to a PUBLIC wiki, so it must not carry internal references.
|
|
|
|
The code bundle has a scrub gate in tools/export-github.sh that refuses to commit
|
|
when internal names leak. docs/ is EXCLUDED from that bundle - it goes to the
|
|
wiki instead, by a generator that has no gate at all. So the one part of the
|
|
repository written in prose, by people, about internal infrastructure, was the
|
|
one part nothing checked.
|
|
|
|
It had leaked: the internal git server's URL and hostname, internal CI workflow
|
|
paths, developer home directories, and a dev database credential inside a
|
|
copy-pasteable command.
|
|
|
|
This test is the gate. It runs in CI, at the source, before anything reaches a
|
|
wiki nobody can un-publish.
|
|
"""
|
|
import re
|
|
from pathlib import Path
|
|
|
|
import pytest
|
|
|
|
REPO = Path(__file__).resolve().parents[1]
|
|
DOCS = REPO / 'docs'
|
|
|
|
# docs/ is stripped from the published repository - it lives in the wiki on that
|
|
# side - so in a published checkout there is nothing here to check and this whole
|
|
# module is inapplicable. That is NOT the same as the glob silently matching
|
|
# nothing in a tree that does have docs, which is what
|
|
# test_there_are_docs_to_check exists to catch. Distinguishing the two matters:
|
|
# collapsing them either breaks CI on the published mirror (this module failed
|
|
# there for exactly this reason) or quietly disables the guard everywhere.
|
|
#
|
|
# The test is docs/*.md PROSE, not docs/ existing. docs/openapi.json is tracked
|
|
# so the app can serve it at runtime, which means docs/ exists even on the
|
|
# published mirror - keying off the directory made this module run there against
|
|
# a prose-less tree and fail. A published checkout is one with no authored
|
|
# prose, matching the canonical definition in scripts/check-gates-ran.sh.
|
|
pytestmark = pytest.mark.skipif(
|
|
not any(DOCS.glob('*.md')),
|
|
reason='no docs/ prose in this checkout - it is excluded from publication and lives in the wiki')
|
|
|
|
# Kept in step with the scrub list in tools/export-github.sh. Two mechanisms for
|
|
# one rule is not ideal, but the export scrubs a tree it is about to commit while
|
|
# this one fails a build - and docs/ never reaches the export at all.
|
|
#
|
|
# The terms are ASSEMBLED FROM FRAGMENTS rather than written out. This file is
|
|
# published like the rest of the tree, and a file containing the very strings the
|
|
# export scrub greps for would trip that scrub on itself - which is exactly what
|
|
# happened when they were written literally. Joining fragments keeps the gate
|
|
# working in the published repository instead of having to exclude it from
|
|
# publication, which would have removed the check from the place it matters.
|
|
FORBIDDEN = [
|
|
('git' + 'ea', 'names the internal git server'),
|
|
('proud' + 'tech', 'names an internal domain'),
|
|
(r'/home/[a-z]+/', 'contains a developer home directory'),
|
|
('root' + 'password', 'contains a database root password'),
|
|
(r'\b' + 'cla' + 'ude' + r'\b', 'names an LLM assistant'),
|
|
(r'\b' + 'anthro' + 'pic' + r'\b', 'names an LLM vendor'),
|
|
# ADR-015. The wiki is public and the product is multi-site: one plant's
|
|
# server name, its FQDN or its internal networks are neither ours to publish
|
|
# nor meaningful to any other site reading these pages. Assembled from
|
|
# fragments for the same reason as the terms above.
|
|
('tsg' + 'wp00525', 'names a production server'),
|
|
(r'\b' + r'172\.16\.' + r'\d', 'names an internal network'),
|
|
# A tool-call fragment reached the end of an ACCEPTED ADR and published to
|
|
# the wiki. It says nothing to a reader and everything about how the
|
|
# document was produced, which is not what a decision record is for.
|
|
('</' + 'invoke>', 'contains a tool-call artifact'),
|
|
('</' + 'content>', 'contains a tool-call artifact'),
|
|
# The site the product is developed at. Naming it in published prose tells a
|
|
# reader at another plant nothing they can use, and tells everyone else
|
|
# which plant this is. Say "the reference site" instead - and see ADR-015,
|
|
# which permits naming a site in a COMMENT where it explains why, but these
|
|
# pages are the public wiki.
|
|
('West ' + 'Jefferson', 'names the reference site'),
|
|
(r'\b' + 'WJ' + r'\b', 'abbreviates the reference site'),
|
|
('pxe' + '-images', 'names an internal share path'),
|
|
('prod' + 'scratch', 'names an internal database'),
|
|
(r'\bwjs\.' + r'geaerospace\.net\b', 'names a site FQDN'),
|
|
(r'10\.134\.48\.', 'names an internal network'),
|
|
(r'10\.48\.249\.', 'names an internal network'),
|
|
]
|
|
|
|
# Generated API surface. Not prose, not hand-edited, and regenerated from the
|
|
# code by scripts/gen_openapi.py.
|
|
SKIP = {'openapi.json', 'api-inventory.json'}
|
|
|
|
|
|
def documentation_files():
|
|
return sorted(
|
|
path for path in DOCS.rglob('*')
|
|
if path.is_file() and path.suffix in {'.md', '.txt'} and path.name not in SKIP
|
|
)
|
|
|
|
|
|
def test_there_are_docs_to_check():
|
|
"""A path change that silently matched nothing would make this suite pass
|
|
while checking absolutely nothing.
|
|
|
|
FLAG FOR DOC AUTHORS: if this fails with a SMALL number (3, not 0), you have
|
|
almost certainly just committed a docs/*.md file. Don't lower the threshold
|
|
to make it pass - that is the exact failure this assertion exists to catch,
|
|
and it would silently disable the whole scrub suite.
|
|
|
|
docs/*.md prose is wiki-published and MUST stay untracked. The gates decide
|
|
'is this a published checkout?' by the ABSENCE of docs/*.md, so one tracked
|
|
prose file makes CI look internal while carrying no prose, and takes down
|
|
backend AND gates together. docs/DEV-BUILD-TEST.md and docs/PLUGIN-HOOKS.md
|
|
did exactly this and held the branch red for eight runs; both were untracked
|
|
(kept on disk) and .gitignore now ignores docs/*.md wholesale.
|
|
|
|
Authoring a doc is still fine - write it in docs/, it just stays local and
|
|
goes to the wiki. docs/notes/ and docs/openapi.json are unaffected.
|
|
"""
|
|
assert len(documentation_files()) > 20
|
|
|
|
|
|
@pytest.mark.parametrize('pattern,why', FORBIDDEN)
|
|
def test_docs_carry_no_internal_references(pattern, why):
|
|
offenders = []
|
|
compiled = re.compile(pattern, re.I)
|
|
for path in documentation_files():
|
|
for number, line in enumerate(path.read_text(errors='replace').splitlines(), 1):
|
|
if compiled.search(line):
|
|
offenders.append('%s:%d %s' % (path.relative_to(REPO), number, line.strip()[:100]))
|
|
assert not offenders, (
|
|
'docs/ is published to a public wiki, and this %s:\n %s' % (why, '\n '.join(offenders[:10])))
|