BROWSER USE - Browser Use Agents: give Browser Use a task and receive completed work. API V4 is current for new integrations. - Browser Infrastructure: connect your agent or automation to managed browsers through SDK, REST, or CDP. Starts at $0.02/browser-hour. - Developer tools: Open Source, Browser Harness, SDK, and MCP support the two products above. [Developer Index](https://browser-use.com/index.md) [Product Map](https://browser-use.com/llms.txt) [Full Product Context](https://browser-use.com/llms-full.txt) [Pricing](https://browser-use.com/pricing.md) [Cloud Docs](https://docs.browser-use.com/cloud/quickstart) [Open Source Docs](https://docs.browser-use.com/open-source/introduction) --- # Tripadvisor scraper skill for AI agents Source: https://browser-use.com/skills/tripadvisor. One file from github.com/browser-use/browser-harness, MIT licensed. Written for browser-harness: code calls its helpers (http_get, js, goto_url). Sites change; these notes describe what worked when last updated. --- # Tripadvisor — Scraping hotels and restaurants for a city Field-tested against www.tripadvisor.com on 2026-10-23 using a Browser Use cloud browser (plain HTTP was also tried and is blocked); every code block re-run verbatim in a fresh cloud browser the same day. robots.txt (checked 2026-10-23) allows the paths this skill loads (`/`, `/Tourism-`, `/Hotels-`, `/Restaurants-`, `/Hotel_Review-`, `/Restaurant_Review-`) for `User-agent: *` and disallows `/Search`; it disallows `/` entirely for named AI crawlers such as ClaudeBot, GPTBot and CCBot. ## Quick summary - Plain HTTP is blocked (status 403, 778-byte block page) on the city hotel list and the city restaurant list; this skill uses a normal browser session. - In the browser, logged out, no consent banner or captcha appeared across ~30 page loads (Boston and Lisbon, US-English site, USD). - Every list page embeds a schema.org **`ItemList` JSON-LD** with the 30 organic results of that page (name, rating, review count, URL, address, geo, phone, price range). Rank, the "#N Best Value" text and the hotel "from $X" price come from the cards (`data-automation` attributes). - Pagination is a URL offset: insert `-oa30-`, `-oa60-`, ... right after the `g` segment. 30 ranked results per page, plus sponsored cards to skip. - Hotels page 1 is special: it renders only 5 ranked hotels until you click **"See all N hotels"**; after that it has 30 like every other page. - Detail pages (`Hotel_Review-...`, `Restaurant_Review-...`) embed `LodgingBusiness` / `FoodEstablishment` JSON-LD, and the visible "#N of M ... in City" popularity rank. ## What works ### 1. Find a city's list URLs (site typeahead) The `/Search?q=` page rendered no results in this run. The header search box typeahead does, and its city option links to `Tourism-g--Vacations.html`, from which the hotel and restaurant list URLs follow. ```python import re, time QUERY = "Lisbon" new_tab("https://www.tripadvisor.com/") time.sleep(2) wait_for_load(30) time.sleep(2) xy = js("""(() => { const i = document.querySelector('input[aria-label="Search"], input[type="search"]'); if (!i) return null; i.scrollIntoView({block: 'center'}); const r = i.getBoundingClientRect(); return [r.x + r.width / 2, r.y + r.height / 2]; })()""") if not xy: raise SystemExit("search box not found") click_at_xy(*xy) time.sleep(1) type_text(QUERY) time.sleep(4) # typeahead suggestions render as role="option" elements options = js("""[...document.querySelectorAll('[role="option"]')].map(o => ({ label: o.innerText.replace(/\\n/g, ' | ').trim(), href: (o.querySelector('a') || o.closest('a') || {}).href || null, }))""") or [] cities = [o for o in options if o["href"] and "/Tourism-g" in o["href"]] for o in cities: m = re.search(r"/Tourism-(g\d+)-(.+)-Vacations\.html", o["href"]) o["restaurants_url"] = f"https://www.tripadvisor.com/Restaurants-{m.group(1)}-{m.group(2)}.html" o["hotels_url"] = f"https://www.tripadvisor.com/Hotels-{m.group(1)}-{m.group(2)}-Hotels.html" print(o) press_key("Escape") ``` Real output (trimmed): ``` {'label': 'Lisbon | Central Portugal, Portugal', 'href': 'https://www.tripadvisor.com/Tourism-g189158-Lisbon_Lisbon_District_Central_Portugal-Vacations.html', 'restaurants_url': 'https://www.tripadvisor.com/Restaurants-g189158-Lisbon_Lisbon_District_Central_Portugal.html', 'hotels_url': 'https://www.tripadvisor.com/Hotels-g189158-Lisbon_Lisbon_District_Central_Portugal-Hotels.html'} {'label': 'Lisbon | Ohio, United States', 'href': 'https://www.tripadvisor.com/Tourism-g50559-Lisbon_Ohio-Vacations.html', ...} ``` Same-name places exist (Lisbon, Portugal / Ohio / New Hampshire), so pick by label. Both derived URLs loaded correctly for Lisbon in this run. ### 2. Restaurants: 3 pages (90 ranked restaurants) ```python import json, re, time from urllib.parse import urljoin CITY_URL = "https://www.tripadvisor.com/Restaurants-g60745-Boston_Massachusetts.html" PAGES = 3 # 30 ranked restaurants per page PER_PAGE = 30 def page_url(city_url, page): # page 1 = city URL; page N inserts -oa<30*(N-1)>- right after the g segment if page == 1: return city_url return re.sub(r"-(g\d+)-", lambda m: f"-{m.group(1)}-oa{PER_PAGE * (page - 1)}-", city_url, count=1) def open_page(url, first=False): new_tab(url) if first else goto_url(url) time.sleep(2) wait_for_load(30) wait_for_element('[data-automation="restaurantCard"]', timeout=20) time.sleep(1.5) def ld_itemlist(): for s in js("[...document.querySelectorAll('script[type=\"application/ld+json\"]')].map(s => s.textContent)") or []: try: d = json.loads(s) except ValueError: continue if isinstance(d, dict) and d.get("@type") == "ItemList": return [e.get("item", {}) for e in d.get("itemListElement", [])] return [] CARDS_JS = r""" [...document.querySelectorAll('[data-automation="restaurantCard"]')].map(c => { const a = c.querySelector('a[href*="Restaurant_Review-"]'); const lines = c.innerText.split('\n').map(s => s.trim()).filter(Boolean); const title = lines.find(l => /^\d+\.\s/.test(l)) || null; return { href: a ? a.getAttribute('href') : null, rank: title ? parseInt(title) : null, // null = Sponsored card name: title ? title.replace(/^\d+\.\s*/, '') : lines.find(l => l !== 'MICHELIN'), rating: (c.querySelector('[data-automation="bubbleRatingValue"]') || {}).innerText || null, reviews: (c.querySelector('[data-automation="bubbleReviewCount"]') || {}).innerText || null, price: lines.find(l => /^\$+( - \$+)?$/.test(l)) || null, }; }) """ def loc_id(url): m = re.search(r"-d(\d+)-", url or "") return m.group(1) if m else None rows, seen = [], set() for p in range(1, PAGES + 1): open_page(page_url(CITY_URL, p), first=(p == 1)) ld = {loc_id(i.get("url")): i for i in ld_itemlist()} total = js("(document.querySelector('[data-automation=\"resultsTotal\"]') || {}).innerText || null") for c in js(CARDS_JS) or []: lid = loc_id(c["href"]) if c["rank"] is None or not lid or lid in seen: continue # skip sponsored cards and repeats seen.add(lid) j = ld.get(lid, {}) agg = j.get("aggregateRating") or {} addr = j.get("address") or {} rows.append({ "rank": c["rank"], "name": j.get("name") or c["name"], "rating": float(agg.get("ratingValue") or c["rating"] or 0) or None, "review_count": agg.get("reviewCount") or int(re.sub(r"\D", "", c["reviews"] or "") or 0) or None, "price_level": j.get("priceRange") or c["price"], "cuisine": j.get("servesCuisine"), "address": addr.get("streetAddress"), "telephone": j.get("telephone"), "url": urljoin("https://www.tripadvisor.com", c["href"]).split("?")[0].split("#")[0], "location_id": lid, }) print(f"page {p}: {js('location.href')} total={total} rows so far={len(rows)}") time.sleep(4) # polite pacing between pages print(json.dumps(rows[:2], indent=1)) print([(r["rank"], r["name"]) for r in rows[-3:]]) print("ranks contiguous:", [r["rank"] for r in rows] == list(range(1, len(rows) + 1))) ``` Real output (trimmed): ``` page 1: https://www.tripadvisor.com/Restaurants-g60745-Boston_Massachusetts.html total=2,561 results rows so far=30 page 2: https://www.tripadvisor.com/Restaurants-g60745-oa30-Boston_Massachusetts.html total=2,561 results rows so far=60 page 3: https://www.tripadvisor.com/Restaurants-g60745-oa60-Boston_Massachusetts.html total=2,561 results rows so far=90 [ { "rank": 1, "name": "Cafe Bonjour", "rating": 4.7, "review_count": 312, "price_level": "$$ - $$$", "cuisine": "French, Cafe", "address": "55 Temple Place, Boston, MA 02111", "telephone": "+1 617-779-0062", "url": "https://www.tripadvisor.com/Restaurant_Review-g60745-d19799010-Reviews-Cafe_Bonjour-Boston_Massachusetts.html", "location_id": "19799010" }, ... ] [(88, "Rino's Place"), (89, 'Krasi'), (90, 'Sweet Cheeks Q')] ranks contiguous: True ``` The same script with the Lisbon URL returned 60 rows over 2 pages, `total=6,462 results`, first row `Floresta Das Escadinhas`, rating 4.8, 7365 reviews, `$`. ### 3. Hotels: 2 pages (60 ranked hotels) ```python import json, re, time from urllib.parse import urljoin CITY_URL = "https://www.tripadvisor.com/Hotels-g60745-Boston_Massachusetts-Hotels.html" PAGES = 2 # 30 ranked hotels per page PER_PAGE = 30 TITLES = '[data-automation="hotel-card-title"]' def page_url(city_url, page): if page == 1: return city_url return re.sub(r"-(g\d+)-", lambda m: f"-{m.group(1)}-oa{PER_PAGE * (page - 1)}-", city_url, count=1) def ranked_count(): return js(f"[...document.querySelectorAll('{TITLES}')].filter(e => /^\\d+\\./.test(e.innerText)).length") or 0 def open_page(url, first=False): new_tab(url) if first else goto_url(url) time.sleep(2) wait_for_load(30) wait_for_element(TITLES, timeout=20) time.sleep(2) SEE_ALL_JS = """[...document.querySelectorAll('button')].find(b => /^See all [\\d,]+ hotels/.test(b.innerText.trim()))""" def expand_first_page(): """Page 1 renders only 5 ranked hotels until 'See all N hotels' is clicked. Raises instead of scraping 5.""" m = re.search(r"([\d,]+) properties in ", js("document.body.innerText") or "") if not m: raise RuntimeError("'N properties in City' total not found on page 1") want = min(PER_PAGE, int(m.group(1).replace(",", ""))) # small cities have fewer than 30 if ranked_count() >= want: return if not js(f"(() => {{ const b = {SEE_ALL_JS}; if (!b) return false; b.scrollIntoView({{block: 'center'}}); return true; }})()"): raise RuntimeError(f"'See all N hotels' button not found; page 1 has {ranked_count()} of {want} ranked hotels") time.sleep(1.5) press_key("Escape") # scrolling can pop the sticky date picker over the button time.sleep(1) xy = js(f"(() => {{ const r = ({SEE_ALL_JS}).getBoundingClientRect(); return [r.x + r.width / 2, r.y + r.height / 2]; }})()") click_at_xy(*xy) for _ in range(20): if ranked_count() >= want: return time.sleep(0.5) raise RuntimeError(f"'See all' click did not expand page 1: {ranked_count()} of {want} ranked hotels after 10 s") def ld_itemlist(): for s in js("[...document.querySelectorAll('script[type=\"application/ld+json\"]')].map(s => s.textContent)") or []: try: d = json.loads(s) except ValueError: continue if isinstance(d, dict) and d.get("@type") == "ItemList": return [e.get("item", {}) for e in d.get("itemListElement", [])] return [] CARDS_JS = r""" [...document.querySelectorAll('[data-automation="hotel-card-title"]')].map(t => { const c = t.closest('li') || t.parentElement; const a = c.querySelector('a[href*="Hotel_Review-"]'); const text = c.innerText; const price = text.match(/(?:^|\n)from\n([^\n]*\d[^\n]*)/); // "from\n$165"; descriptions also say "from South Station" return { title: t.innerText.trim(), href: a ? a.getAttribute('href') : null, rating: (c.querySelector('[data-automation="bubbleRatingValue"]') || {}).innerText || null, reviews: (c.querySelector('[data-automation="bubbleReviewCount"]') || {}).innerText || null, rank_text: (text.match(/#\d[\d,]* [^\n]* of [\d,]+ [^\n]*/) || [null])[0], price_from: price ? price[1] : null, }; }) """ def loc_id(url): m = re.search(r"-d(\d+)-", url or "") return m.group(1) if m else None rows, seen = [], set() for p in range(1, PAGES + 1): open_page(page_url(CITY_URL, p), first=(p == 1)) if p == 1: expand_first_page() ld = {loc_id(i.get("url")): i for i in ld_itemlist()} for c in js(CARDS_JS) or []: m = re.match(r"(\d+)\.\s*(.+)", c["title"]) lid = loc_id(c["href"]) if not m or not lid or lid in seen: continue # no "N." prefix = Sponsored card seen.add(lid) j = ld.get(lid, {}) addr = j.get("address") or {} rows.append({ "rank": int(m.group(1)), "name": m.group(2), "rating": float(c["rating"]) if c["rating"] else None, "review_count": int(re.sub(r"\D", "", c["reviews"] or "") or 0) or None, "rank_text": c["rank_text"], "price_from": c["price_from"], "price_tier": j.get("priceRange"), "address": ", ".join(filter(None, [addr.get("streetAddress"), addr.get("addressLocality")])) or None, "url": urljoin("https://www.tripadvisor.com", c["href"]).split("?")[0].split("#")[0], "location_id": lid, }) print(f"page {p}: {js('location.href')} rows so far={len(rows)}") time.sleep(4) # polite pacing between pages print(json.dumps(rows[:2], indent=1)) print([(r["rank"], r["name"], r["price_from"]) for r in rows[-3:]]) print("ranks contiguous:", [r["rank"] for r in rows] == list(range(1, len(rows) + 1))) print("rows joined to JSON-LD:", sum(1 for r in rows if r["address"])) print("rows with a price:", sum(1 for r in rows if r["price_from"])) ``` Real output (trimmed): ``` page 1: https://www.tripadvisor.com/Hotels-g60745-Boston_Massachusetts-Hotels.html rows so far=30 page 2: https://www.tripadvisor.com/Hotels-g60745-oa30-Boston_Massachusetts-Hotels.html rows so far=60 [ { "rank": 1, "name": "Cambria Hotel Boston Downtown", "rating": 4.2, "review_count": 244, "rank_text": "#1 Best Value of 293 places to stay in Boston", "price_from": "$165", "price_tier": "€", "address": "6 W Broadway, Boston", "url": "https://www.tripadvisor.com/Hotel_Review-g60745-d15618152-Reviews-Cambria_Hotel_Boston_Downtown-Boston_Massachusetts.html", "location_id": "15618152" }, ... ] [(58, 'Hyatt Regency Boston', '$237'), (59, 'Boston Harbor Hotel', '$502'), (60, 'Boston Marriott Long Wharf', '$234')] ranks contiguous: True rows joined to JSON-LD: 60 rows with a price: 60 ``` With the Lisbon hotels URL and `PAGES = 3` the same script returned 90 contiguous ranks, `#1 Best Value of 2,494 places to stay in Lisbon` (Hotel Da Baixa, 4.9, 3479 reviews, from $273), all 90 joined to JSON-LD. ### 4. Detail page basics (hotel or restaurant) ```python import json, re, time URLS = [ "https://www.tripadvisor.com/Hotel_Review-g60745-d15618152-Reviews-Cambria_Hotel_Boston_Downtown-Boston_Massachusetts.html", "https://www.tripadvisor.com/Restaurant_Review-g60745-d19799010-Reviews-Cafe_Bonjour-Boston_Massachusetts.html", ] TYPES = {"LodgingBusiness", "Hotel", "FoodEstablishment", "Restaurant"} def detail(url, first=False): new_tab(url) if first else goto_url(url) time.sleep(2) wait_for_load(30) time.sleep(2) j = {} for s in js("[...document.querySelectorAll('script[type=\"application/ld+json\"]')].map(s => s.textContent)") or []: try: d = json.loads(s) except ValueError: continue if isinstance(d, dict) and d.get("@type") in TYPES: j = d break text = js("document.body.innerText") or "" rank = re.search(r"#\d[\d,]* of [\d,]+ [^\n]+", text) stars = re.search(r"(\d(?:\.\d)?) of 5 stars", text) agg = j.get("aggregateRating") or {} addr = j.get("address") or {} cuisine = j.get("servesCuisine") return { "type": j.get("@type"), "name": j.get("name"), "rating": float(agg["ratingValue"]) if agg.get("ratingValue") else None, "review_count": agg.get("reviewCount"), "popularity_rank": rank.group(0) if rank else None, "price_level": j.get("priceRange"), "hotel_class": float(stars.group(1)) if stars and j.get("@type") in ("LodgingBusiness", "Hotel") else None, "cuisine": ", ".join(cuisine) if isinstance(cuisine, list) else cuisine, "address": ", ".join(filter(None, [addr.get("streetAddress"), addr.get("addressLocality"), addr.get("postalCode")])), "telephone": j.get("telephone"), "geo": [j["geo"]["latitude"], j["geo"]["longitude"]] if j.get("geo") else None, "amenities": [a["name"] for a in j.get("amenityFeatures") or [] if a.get("value")][:10], "opening_hours": [(h.get("dayOfWeek"), h.get("opens"), h.get("closes")) for h in j.get("openingHoursSpecification") or []], "url": j.get("url") or url, "location_id": (re.search(r"-d(\d+)-", url) or [None, None])[1], } for i, u in enumerate(URLS): print(json.dumps(detail(u, first=(i == 0)), indent=1)) time.sleep(4) ``` Real output (trimmed): ``` { "type": "LodgingBusiness", "name": "Cambria Hotel Boston Downtown", "rating": 4.2, "review_count": 244, "popularity_rank": "#68 of 96 hotels in Boston", "price_level": "$ (Based on Average Nightly Rates for a Standard Room from our Partners)", "hotel_class": 4.0, "address": "6 W Broadway, Boston, 02127-1016", "telephone": "1 (617) 463-9147", "geo": [42.343216, -71.05699], "amenities": ["Valet parking", "Free High Speed Internet (WiFi)", "Fitness Center with Gym / Workout Room", ...], ... "location_id": "15618152" } { "type": "FoodEstablishment", "name": "Cafe Bonjour", "rating": 4.7, "review_count": 312, "popularity_rank": "#2 of 2,612 Restaurants in Boston", "price_level": "$$ - $$$", "cuisine": "French, Cafe", "opening_hours": [["Sunday", "07:00:00", "15:00:00"], ...], ... "location_id": "19799010" } ``` ## Fields | Field | Where | Source | Example | |---|---|---|---| | `rank` | lists | card title `"N. Name"` (`hotel-card-title` / `restaurantCard`) | `1` | | `name` | lists, detail | JSON-LD `name` or card title | `"Cafe Bonjour"` | | `rating` | lists, detail | JSON-LD `aggregateRating.ratingValue` or `bubbleRatingValue` | `4.7` | | `review_count` | lists, detail | JSON-LD `aggregateRating.reviewCount` or `bubbleReviewCount` `"(312 reviews)"` | `312` | | `price_level` | restaurant list, detail | JSON-LD `priceRange` | `"$$ - $$$"`, `"$$$$"`, `"$"` | | `rank_text` | hotel list | card text | `"#1 Best Value of 293 places to stay in Boston"` | | `price_from` | hotel list | card text after `from` (default dates, USD) | `"$165"` | | `price_tier` | hotel list | ItemList JSON-LD `priceRange` | `"€"`, `"€€"` | | `cuisine` | restaurant list, detail | JSON-LD `servesCuisine` (string in list, array on detail) | `"French, Cafe"` | | `address` | lists, detail | JSON-LD `address` | `"55 Temple Place, Boston, MA 02111"` | | `telephone` | restaurant list, detail | JSON-LD `telephone` | `"+1 617-779-0062"` | | `geo` | detail | JSON-LD `geo` (also in the list ItemList, not extracted above) | `[42.343216, -71.05699]` | | `popularity_rank` | detail | page text `#N of M ... in City` | `"#68 of 96 hotels in Boston"` | | `hotel_class` | hotel detail | page text `N of 5 stars` | `4.0` | | `amenities` | hotel detail | JSON-LD `amenityFeatures[].name` | `"Valet parking"` | | `opening_hours` | restaurant detail | JSON-LD `openingHoursSpecification` | `("Sunday", "07:00:00", "15:00:00")` | | `location_id` | lists, detail | `-d-` in the URL, stable join key | `"19799010"` | | total results | restaurant list | `[data-automation="resultsTotal"]` | `"2,561 results"` | ## Gotchas - **Hotels page 1 shows only 5 ranked hotels.** The 30-item list appears only after clicking "See all N hotels" (URL does not change; ItemList JSON-LD grows from 8 to 30). Scrolling that button into view opened the sticky check-in date picker on top of it in this run, and a blind click picked a date instead. Press Escape, then click. `expand_first_page()` raises if the button is missing or the list does not grow, so a failed click never yields a silent 5-row page. Its target is `min(30, "N properties in City")`: Lisbon, Ohio shows `6 properties in Lisbon` with all 6 ranked and a "See all 34 hotels" button (nearby towns), so small cities need no click. - **Pages 2+ need no click.** `-oa30-`, `-oa60-` load 30 ranked hotels directly. `-oa0-` and `-oa5-` redirect to page 1 (the 5-hotel view). - **JSON-LD `position` restarts at 1 on every page.** Take the global rank from the card title (`"31. The Bostonian Hotel Boston"`), not from `position`. - **Sponsored cards** have no `"N."` prefix (hotels: a `Sponsored` label; the same sponsored hotel appeared twice on one page). Sponsored hotel links are absolute and carry `?spAttributionToken=`. Skip them and dedupe by `location_id`. - **List rank is not the detail rank.** Hotel lists default to "Best Value" sort (`#1 Best Value of 293 places`), restaurant lists to "Featured". The detail page shows a different popularity rank (Cambria: #1 Best Value in the list, `#68 of 96 hotels` on its page; Cafe Bonjour: rank 1 in the list, `#2 of 2,612 Restaurants`). Record which one you store. List counts also differ from detail counts (2,561 list results vs 2,612 on detail). - **Hotel `price_from` depends on the default dates** the site pre-fills (here 1 night, Nov 01-02, 1 room, 2 guests) and on currency (USD). Hotel list JSON-LD `priceRange` is a tier written in euro signs even on the USD site; hotel detail `priceRange` is `"$ (Based on Average Nightly Rates ...)"`. - **Do not regex `from (\S+)` for the hotel price.** AI-written card descriptions contain phrases like "from South Station"; a loose regex returned `"South"` once. The price sits on its own line after a line that is exactly `from`. - **Hotel order on pages 2+ shifted between two loads minutes apart** (rank 58 was Hyatt Regency Boston in one run, HI Boston - Hostel in the next). Scrape all pages in one session and dedupe by `location_id`; do not stitch pages from different runs. - **Hotel list JSON-LD `url` points to `HotelHighlight-...?m=...`**, not `Hotel_Review-...`. Join on `location_id` and use the card link for the canonical URL. Restaurant list JSON-LD uses `Restaurant_Review-...?m=...`. - **Restaurant list `address.streetAddress` holds the full address**; `addressLocality` is empty in the list but filled on the detail page. - **`wait_for_load()` right after `goto_url()` can return for the previous page.** Sleep ~2 s first and wait for a card selector. - **Pacing:** ~30 page loads with 3-5 s between them produced no captcha, block or consent dialog. Keep the delay; Tripadvisor bot-protects the plain HTTP path, so do not hammer the browser path either. - Detail pages also show reviewer names and review text; this skill does not extract them. ## What does not work - **Plain HTTP:** plain HTTP is blocked (status 403) for `Hotels-g60745-Boston_Massachusetts-Hotels.html` and `Restaurants-g60745-Boston_Massachusetts.html`; this skill uses a normal browser session. - **`/Search?q=Lisbon+Portugal`** rendered only the header and footer in the browser (no result links after 10+ s). Use the typeahead (block 1). - **Clicking "See all N hotels" without closing the date picker** selects a calendar date instead and leaves 5 hotels on the page. - **`-oa0-` / `-oa5-` offsets** redirect to the unexpanded page 1; only multiples of 30 from `-oa30-` up behaved as pages in this run.