#!/usr/bin/env python3 """verify-generated.py — the wall reads back its own generators' output. The escaped-title catch (2026-08-16, the reader's-rule wake) proved the wall can fail to read its OWN generators' files: sitemap.py escapes the desc for XML and the wall's sitemap-title check never unescaped — a derived word the machinery itself wrote sat outside the wall's reading until the coda's '>' exposed it. This gate is the pattern one surface further: every derived file the finish script regenerates is emitted by a generator and consumed by the wall — does each have a reverse gate (generated ⊆ read)? The finish script (finish-vigo-wake.sh) regenerates, every wake: site/feed.xml (tools/feed.py — from wake-log/index.html) site/sitemap.xml (tools/sitemap.py — from the filesystem) site/robots.txt (tools/sitemap.py — same pass) site/tools/index.html (tools/regpage.py — from unillustrate --json) site/art/index.html (tools/artpage.py — from the SVGs' own descs) site/index.html (ops/scripts/sync-vigo-home-now.py — from the wake log's top-5, the home Now block) site/light/index.html (tools/light-quote.py — the register quote) This gate closes the read-back for the files the wall reads by hand-written regex or never reads at all: 1. FEED PARSES: feed.xml must be well-formed XML with the machinery's own parser (ElementTree). Nothing parsed the feed before — a tag-like word or a broken escape in any item would reach subscribers while every wall passed. Same class as verify-svg-xml.py, one surface over in the site's own derived output. 2. FEED ROUND-TRIP (generated ⊆ read): every feed item's title and description must equal the wake-log's own words — the wall re-derives the plain text exactly as feed.py does (strip tags, unescape, collapse) and requires each item's carried words to match. The generator escapes for XML; the wall unescapes; a title the wall cannot read back fails the build, named by its timestamp. And the reverse: every wake entry must have a feed item (the count must match). 3. SITEMAP PARSES: sitemap.xml must be well-formed XML. The wall regex-reads its image entries but nothing ever PARSED the file itself — the file the whole-site img scan reads from was never verified to be readable by the machinery's own parser. 4. ROBOTS CARRIES ITS DERIVED LINE: robots.txt is generated in the same pass as the sitemap; the wall reads back the line it must carry (the Sitemap: URL) — a wake that hand-edits or breaks the derived floor fails. 5. HOME NOW == WAKE LOG'S TOP-5: the home page's Now block is DERIVED by sync-vigo-home-now.py from the wake log. The wall re-derives the top-5 wake entries and requires the home page to carry exactly them — the home page's newest words ARE the wake log's newest words, read back. 6. THE REGISTER'S OWN PAGE (Q101, 2026-08-16): the finish script regenerates site/tools/index.html (regpage.py), site/art/index.html (artpage.py), and site/light/index.html (light-quote.py) every wake — and the wall read only PIECES of them: the art caps' substrings, the count line, the quote's substring. A hand-edited card number on the register, a hand-added piece on the art page, or a duplicated seam in the light page walked past every gate while the page silently disagreed with the machine that made it. The reverse gate, whole-page: - tools/index.html must equal regpage.build() byte-for-byte (the page IS the generator's output — regenerated from unillustrate's live JSON); - art/index.html must equal artpage.build() byte-for-byte (every piece, cap, and count IS the generator's output from the SVGs' own words); - light/index.html's regs-quote seam must carry EXACTLY the derived quote, and the seam must appear exactly once (a duplicated seam is a page the generator cannot read back — the substring check could not see it). Mechanical, verified both ways (verify-generated-both-ways.py): the honest estate exits 0; a hand-edited feed title, a malformed sitemap, a broken robots line, a stale Now block, a hand-edited register card, a hand-added art piece, or a tampered light seam each exit 1 named by their file. """ import html as _html import re import sys import xml.etree.ElementTree as ET from pathlib import Path SITE = Path(__file__).resolve().parent.parent BASE = "https://vigo.trentuna.com" FAILURES: list[str] = [] LINES: list[str] = [] def fail(name: str, why: str) -> None: FAILURES.append(f"{name}: {why}") LINES.append(f" FAIL {name}: {why}") # ---- shared wake-block extraction (the same contract feed.py uses) -------- _START_RE = re.compile('
])') _OPEN_RE = re.compile(r"]") _CLOSE_RE = re.compile(r"
") def extract_wake_blocks(html_text: str) -> list[str]: blocks: list[str] = [] pos = 0 while True: m = _START_RE.search(html_text, pos) if not m: break start = m.start() i = m.end() depth = 1 while depth > 0: o = _OPEN_RE.search(html_text, i) c = _CLOSE_RE.search(html_text, i) if c is None: fail("wake-log", f"unbalanced div at offset {i}") return blocks if o is not None and o.start() < c.start(): depth += 1 i = o.end() else: depth -= 1 i = c.end() blocks.append(html_text[start:i]) pos = i return blocks def entry_plain(block: str) -> tuple[str, str, str]: """The wake block's own words, as feed.py reads them: (timestamp, title, description). Tags stripped, entities unescaped, whitespace collapsed.""" t = re.search(r'
([^<]+)
', block).group(1).strip() b = re.search(r'
(.*?)
', block, re.S).group(1) text = re.sub(r"<[^>]+>", "", b) text = _html.unescape(text).strip() text = " ".join(text.split()) strong = re.search(r"(.*?)", b, re.S) title = _html.unescape(strong.group(1)).strip() if strong else t title = " ".join(title.split()) return t, title, text def check_feed(wake_html: str) -> None: feed_path = SITE / "feed.xml" if not feed_path.exists(): fail("feed.xml", "the generator's output is missing — the finish script must run feed.py") return # 1. FEED PARSES — the machinery's own parser. try: root = ET.parse(feed_path).getroot() except ET.ParseError as e: fail("feed.xml", f"not well-formed XML — {e}") return items = root.findall(".//item") LINES.append(f"feed: parsed {len(items)} items as XML (ElementTree)") # 2. FEED ROUND-TRIP — every item's words must equal the wake log's own. blocks = extract_wake_blocks(wake_html) if not blocks: fail("wake-log", "no wake blocks found") return by_time: dict[str, tuple[str, str]] = {} for block in blocks: t, title, desc = entry_plain(block) by_time[t] = (title, desc) if len(items) != len(blocks): fail("feed.xml", f"carries {len(items)} items but the wake log carries {len(blocks)} blocks " f"— generated {'' if len(items) == len(blocks) else '⊄'} read") for item in items: title = (item.findtext("title") or "").strip() desc = (item.findtext("description") or "").strip() desc = " ".join(desc.split()) guid = item.findtext("guid") or "" # guid is wake-YYYYMMDDHHMM — rebuild the timestamp's clock face m = re.match(r"wake-(\d{4})(\d{2})(\d{2})(\d{2})(\d{2})", guid) t = "" if m: t = f"{m.group(1)}-{m.group(2)}-{m.group(3)} · {m.group(4)}:{m.group(5)}" src = by_time.get(t) if src is None: fail("feed.xml", f"item {t} has no wake-log block — the feed speaks a time the log does not") continue want_title, want_desc = src if title != want_title: fail("feed.xml", f"item {t} title was not read back: '{title[:60]}…' != '{want_title[:60]}…'") if desc != want_desc: fail("feed.xml", f"item {t} description was not read back (generator escaped, wall unescapes — the escaped-title class)") LINES.append(f"feed round-trip: {len(items)}/{len(blocks)} items == wake log words") def check_sitemap() -> None: sitemap_path = SITE / "sitemap.xml" if not sitemap_path.exists(): fail("sitemap.xml", "the generator's output is missing — the finish script must run sitemap.py") return try: ET.parse(sitemap_path) except ET.ParseError as e: fail("sitemap.xml", f"not well-formed XML — {e}") return text = sitemap_path.read_text(encoding="utf-8") urls = re.findall(r"([^<]+)", text) LINES.append(f"sitemap: parsed as XML, {len(urls)} urls") def check_robots() -> None: robots_path = SITE / "robots.txt" if not robots_path.exists(): fail("robots.txt", "the generator's output is missing — the finish script must run sitemap.py") return text = robots_path.read_text(encoding="utf-8") want = f"Sitemap: {BASE}/sitemap.xml" if want not in text: fail("robots.txt", f"does not carry its derived line '{want}'") def check_home_now(wake_html: str) -> None: home_path = SITE / "index.html" if not home_path.exists(): fail("index.html", "the home page is missing") return home = home_path.read_text(encoding="utf-8") blocks = extract_wake_blocks(wake_html) if not blocks: fail("wake-log", "no wake blocks found") return top5 = blocks[:5] h2 = "

Now

" start = home.find(h2) if start < 0: fail("index.html", "Now block not found — sync-vigo-home-now.py's marker is missing") return # The Now block is the raw wake-block source between

Now

and the # "read the full wake log" link (the same marker sync-vigo-home-now.py # uses — the derivation's own seam); compare against the wake log's top-5 # raw blocks. link_marker = '

None: """THE REGISTER'S OWN PAGE (Q101, 2026-08-16): the finish script regenerates site/tools/index.html (regpage.py), site/art/index.html (artpage.py), and site/light/index.html (light-quote.py) every wake. The wall read only PIECES of them — the art caps' substrings, the count line, the quote's substring — so a hand-edited card number, a hand-added piece, or a duplicated seam walked past every gate. The reverse gate, whole-page: each generated page must equal its own generator's build(), byte-for-byte. """ # --- the register: the page IS regpage.build() ------------------------- try: import regpage data = regpage.live_json() fresh = regpage.build(data) except SystemExit as e: fail("tools/index.html", f"regpage.build() itself failed — {e}") return reg_path = SITE / "tools" / "index.html" if not reg_path.exists(): fail("tools/index.html", "the register's own page is missing — the finish script must run regpage.py") else: on_disk = reg_path.read_text(encoding="utf-8") if on_disk != fresh: fail("tools/index.html", "the register's own page drifted from regpage.build() — a hand-edited card, a reordered card, " "or a generator change not re-run (generated ⊄ read)") else: LINES.append(f"register page: tools/index.html == regpage.build() byte-for-byte ({len(data)} cards)") # --- the art page: every piece and cap IS artpage.build() --------------- try: import artpage fresh = artpage.build() except SystemExit as e: fail("art/index.html", f"artpage.build() itself failed — {e}") return art_path = SITE / "art" / "index.html" if not art_path.exists(): fail("art/index.html", "the art page is missing — the finish script must run artpage.py") else: on_disk = art_path.read_text(encoding="utf-8") if on_disk != fresh: fail("art/index.html", "the art page drifted from artpage.build() — a hand-added piece, a hand-edited cap, " "or a generator change not re-run (generated ⊄ read)") else: LINES.append(f"art page: art/index.html == artpage.build() byte-for-byte ({len(artpage.SVG_PIECES)} SVG caps derived)") # --- the light page: the seam carries EXACTLY the derived quote --------- light_path = SITE / "light" / "index.html" if not light_path.exists(): fail("light/index.html", "the light page is missing — the finish script must run light-quote.py") return text = light_path.read_text(encoding="utf-8") begin = "" end = "" if text.count(begin) != 1 or text.count(end) != 1: fail("light/index.html", f"the regs-quote seam must appear exactly once (found begin={text.count(begin)}, end={text.count(end)}) " f"— a duplicated seam is a page the generator cannot read back") return b = text.find(begin) + len(begin) e = text.find(end) inside = text[b:e] try: import regpage as _regpage seq = _regpage.catalog_seq() except SystemExit as ex: fail("light/index.html", f"catalog_seq() itself failed — {ex}") return if inside != seq: fail("light/index.html", f"the seam's quote was not read back: carried '{inside}' != derived '{seq}'") else: LINES.append(f"light page: the seam's quote == catalog_seq() exactly ('{seq}')") def check_all() -> tuple[int, list[str]]: FAILURES.clear() LINES.clear() wake_html = (SITE / "wake-log" / "index.html").read_text(encoding="utf-8") check_feed(wake_html) check_sitemap() check_robots() check_home_now(wake_html) check_pages() return len(FAILURES), LINES if __name__ == "__main__": n, lines = check_all() print("verify-generated: the wall reads its own generators' files") for l in lines: print(l) if n: print(f"verify-generated: {n} failure(s)") sys.exit(1) print("OK: feed parses and round-trips, sitemap parses, robots carries its line, home Now == wake log's top-5, the register's own page == its generator (tools/art/light read back byte-for-byte)") sys.exit(0)