Files
Sylpheed/tools/re/check-capture-citations
Fabian Hamm d9ba2c73b1 fix(re): the capture checker counted its own selftest fixture as a citation
`check-capture-citations` exits 1 on a clean checkout of this branch, and would
have kept doing so for ever:

    🔴 cited but NOT committed: 1
        docs/re/captures/no-such-file-anywhere.png

`SEARCH` includes `tools`, so the scan reads this file -- and `selftest()`'s
planted fixture is a literal capture path. The checker cited its own test data,
to a file that will never exist by construction. A gate that is red before
anyone changes anything teaches people to ignore it, which is the opposite of
what Phase 4 built it for.

Excluding only this one file (`SELF`) keeps every other tool in scope. A real
citation belongs in a page, or in a tool that opens the capture -- never in the
checker itself.

Added a selftest case for it, because this is exactly the class of bug the rest
of this file already documents surviving: the ERE that matched nothing while
five ticks stayed green. Verified the guard can fail, rather than assuming it:

    exclusion neutralised -> "own fixtures not counted 🔴 FAILED",
                             selftest rc=2, real check rc=1 reproducing the bug
    exclusion restored    -> selftest ok (7 cases), check rc=0

After the fix, on this branch:

    captures committed          : 212
      cited by a page or a tool : 212
      cited by nothing          : 0
      🔴 cited but NOT committed: 0        rc=0

which is also the first independent confirmation of Phase 4's "212 captures,
all cited" -- the checker now agrees with the page.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-14 21:18:55 +02:00

208 lines
9.1 KiB
Python
Executable File

#!/usr/bin/env python3
"""Do the captures and the pages that cite them agree?
tools/re/check-capture-citations # assert
tools/re/check-capture-citations --selftest # can it fail?
tools/re/check-capture-citations --orphans # list the unreferenced files
`tools/port/check-citations` has checked `docs/port/*.md` since it was written.
**`docs/re/` has never had one**, and `docs/re/captures/` is 118 MB — the
largest thing in the repository, and the one place a file can be added, never
cited, and never noticed.
Two failures, which are opposites and must not be conflated:
* a page cites a capture that **is not committed** — a reader following it
gets nothing. That is an error, exactly as in `check-citations`.
* a capture that **no page cites** — not an error. It may be evidence a page
should have cited, and deleting on that basis would silently ratify the
omission. Reported, counted, never failed on.
⚠️ THE NAIVE VERSION OF THIS CHECK REPORTS 10 DANGLING AND IS WRONG ABOUT 9.
A first pass with a plain path regex claimed ten; the real number is one. Eight
were **directory** references — `docs/re/captures/ui-layout/` — which resolve
perfectly well and are absent only from `git ls-tree`, which lists files. One
was a path at the end of a sentence, with the full stop pulled into the match.
Both are handled below, and both are in the selftest, because a checker that
cries wolf nine times out of ten is worse than none: it teaches you to ignore it.
"""
import os
import re
import subprocess
import sys
ROOT = "docs/re/captures/"
# A capture path, optionally with the `docs/re/` prefix, optionally a directory.
CITE = re.compile(r"(?:docs/re/)?captures/[A-Za-z0-9._/-]+")
# ⚠️ `git grep -E` is POSIX ERE: it has no `(?:`, and a pattern carrying one
# matches NOTHING while exiting 0. That silence read as "no page cites any
# capture" — 257 orphans, 0 citations — which is the same shape as the export
# table that returned empty and let the build succeed.
CITE_ERE = r"(docs/re/)?captures/[A-Za-z0-9._/-]+"
# Where a citation may live. Prose AND code: tests and tools open these files.
SEARCH = ["docs", "crates", "tools", "port", "authored"]
# 🔴 AND NOT THIS FILE. `SEARCH` includes `tools`, so the scan read *itself* --
# and `selftest()`'s planted fixture below is a literal capture path. The check
# therefore reported a dangling citation to a file that will never exist, and
# exited 1 on a clean checkout, for ever. A gate that is red before anyone
# changes anything teaches people to ignore it.
# Excluding only this one file keeps every other tool in scope; a real citation
# belongs in a page or a tool that opens the capture, never in the checker.
SELF = ":!tools/re/check-capture-citations"
def git(*args: str) -> str:
return subprocess.run(["git", *args], capture_output=True, text=True).stdout
def committed() -> tuple[set[str], set[str]]:
"""Every capture in the WORKING TREE, and every directory they imply.
⚠️ This read `git ls-tree HEAD` first, and that made the check useless for
the thing it is for: restoring a missing capture left it still reporting the
citation as dangling, because the fix was in the index and the check was
looking at the last commit. A pre-merge check must see the change being
proposed. `tools/port/check-citations` uses the working tree for exactly
this reason; so does this.
"""
files = {l for l in git("ls-files", ROOT).splitlines() if l}
dirs = set()
for f in files:
parts = f.split("/")
for i in range(4, len(parts)): # docs/re/captures/<dir>/...
dirs.add("/".join(parts[:i]) + "/")
return files, dirs
def cited() -> set[str]:
raw = git("grep", "-rhoE", CITE_ERE, "--", *SEARCH, SELF).splitlines()
out = set()
for line in raw:
c = line.split(":")[-1]
# ⚠️ A path at the end of a sentence takes the punctuation with it.
c = c.rstrip(".,;:)`")
if not c.startswith("docs/"):
c = "docs/re/" + c
out.add(c)
return out
def resolves(c: str, files: set[str], dirs: set[str]) -> bool:
if c in files:
return True
# A directory reference, written with or without its trailing slash.
return c.rstrip("/") + "/" in dirs
def named_bare() -> set[str]:
"""Basenames that appear anywhere outside the capture tree itself.
🔴 A CAPTURE CAN BE USED WITHOUT ITS PATH. `tools/re-capture/screen_match.py`
opens `movie-frame-attract-a.png` and `-b.png` by **filename**, building the
directory separately. A path-only scan calls both orphans, and deleting them
on that basis breaks the tool — which is what nearly happened. Anything
named at all counts as used.
"""
out = set()
for line in git("grep", "-rhoE",
r"[A-Za-z0-9._-]+\.(png|jpg|csv|log|txt|json|jsonl|bin|tsv)",
"--", *SEARCH, SELF).splitlines():
out.add(line.split(":")[-1].strip())
return out
def scan():
files, dirs = committed()
refs = cited()
bare = named_bare()
dangling = sorted(c for c in refs if not resolves(c, files, dirs))
# 🔴 A FILE INSIDE A CITED DIRECTORY IS CITED. A page that says "see
# `docs/re/captures/hud-runtime/`" is citing the whole directory, so every
# file in it is referenced and none is an orphan. Without this rule the
# first prune deleted the entire contents of two such directories and the
# check then went red on the very citations that made them evidence — it
# caught its own damage, which is the only reason this rule exists.
cited_dirs = {c.rstrip("/") + "/" for c in refs if resolves(c, files, dirs) and c not in files}
orphans = sorted(
f for f in files
if f not in refs
and os.path.basename(f) not in bare
and not any(f.startswith(d) for d in cited_dirs)
)
return files, refs, dangling, orphans
def selftest() -> int:
"""🔴 A CHECK THAT CANNOT FAIL IS NOT A CHECK — and this one has three ways
to be wrong, not one. It must catch a real dangling citation, and it must
NOT flag a directory reference or a path ending a sentence, which is how
the naive version produced nine false alarms."""
files, dirs = committed()
real_dir = next(iter(dirs), None)
real_file = next(iter(files), None)
if not real_dir or not real_file:
print("selftest: 🔴 no captures committed — nothing to test against")
return 2
cases = [
("planted dangling caught", "docs/re/captures/no-such-file-anywhere.png", False),
("real file passed", real_file, True),
("directory reference passed", real_dir, True),
("directory without slash passed", real_dir.rstrip("/"), True),
("sentence punctuation stripped", real_file + ".", True),
]
# 🔴 THE SELFTEST USED TO PASS WHILE THE SCAN RETURNED NOTHING. It exercised
# `resolves()` and never the gathering, so an ERE the grep could not compile
# produced "0 cited, 257 orphans" and five green ticks above it. A check
# whose selftest cannot see the failure that actually happened is decoration.
refs = cited()
gathering_ok = len(refs) > 50 and any(r.startswith(ROOT) for r in refs)
print(" %-34s %s (%d citations found)"
% ("citation gathering works", "ok" if gathering_ok else "🔴 FAILED", len(refs)))
# 🔴 REGRESSION GUARD FOR `SELF`. The planted fixture above is a literal
# capture path living in a file `SEARCH` covers. Before the exclusion
# existed the scan counted it as a real citation, so the check reported a
# dangling capture and exited 1 on a clean tree — permanently. If anyone
# drops `SELF`, this goes red here instead of in everybody else's run.
fixture_counted = cases[0][1] in refs
print(" %-34s %s"
% ("own fixtures not counted", "🔴 FAILED" if fixture_counted else "ok"))
ok = gathering_ok and not fixture_counted
for name, path, want in cases:
got = resolves(path.rstrip(".,;:)`"), files, dirs)
mark = "ok" if got == want else "🔴 FAILED"
if got != want:
ok = False
print(" %-34s %s" % (name, mark))
print("selftest: %s" % ("ok" if ok else "🔴 BROKEN"))
return 0 if ok else 2
def main() -> int:
if "--selftest" in sys.argv:
return selftest()
files, refs, dangling, orphans = scan()
if "--orphans" in sys.argv:
for f in orphans:
print(f)
return 0
print("captures committed : %d" % len(files))
print(" cited by a page or a tool : %d" % (len(files) - len(orphans)))
print(" cited by nothing : %d (reported, not failed —" % len(orphans))
print(" an orphan may be evidence a page owes)")
if dangling:
print(" 🔴 cited but NOT committed: %d" % len(dangling))
for d in dangling:
print(" %s" % d)
print("\n🔴 a reader following those gets nothing. Commit the capture, fix the")
print(" path, or drop the citation.")
return 1
print(" 🔴 cited but NOT committed: 0")
return 0
if __name__ == "__main__":
raise SystemExit(main())