fix(downloads): a shared or feed-opened Facebook group post downloads its own video

A Facebook share link (/share/v/…) redirects to a group post, but only the
permalink form went through the post-page read, so yt-dlp followed the
share link itself and downloaded an ad from the signed-in group feed
(prod job 471). The link of a post opened from the group's feed
(/groups/<g>/?multi_permalinks=<id>) was not recognised at all, and yt-dlp
answers it with "Unsupported URL" (prod job 472).

With a cookie, a share link is now followed in the same request that reads
the post page, and a feed link is read as the post it names. Without one,
the feed link goes to yt-dlp in the permalink form it knows. A post page's
"no single video" verdict now names the post for a share link too.

backend/app/downloads/adapters.py

@@ -12,7 +12,7 @@ Resolution works both before extraction (from source_kind / URL host) and after
 """
 
 import re
-from urllib.parse import urlparse
+from urllib.parse import parse_qs, urlparse
 
 from yt_dlp.networking.impersonate import ImpersonateTarget
 
@@ -183,10 +183,23 @@ class FacebookAdapter(SourceAdapter):
         # in the page scripts that carry the post id; its watch link downloads (plain first, then the
         # Chrome retry above). Needs the cookie: signed out, the page is the login wall.
         post_id = facebook_group_post_id(url)
-        if post_id is None or not ydl_opts.get("cookiefile"):
+        if not ydl_opts.get("cookiefile"):
+            # yt-dlp knows the permalink form and answers the feed's form with "Unsupported URL".
+            if post_id is not None and not _FB_GROUP_POST_PATH.match(urlparse(url).path):
+                return _group_permalink(url, post_id)
+            return url
+        # A share link is a post only once followed: handed to yt-dlp, it was followed to the group
+        # page and an ad from the signed-in feed came down (prod job 471). One request does both — the
+        # redirect lands on the post page, which is then read as any post link's.
+        if post_id is None and not _is_share_link(url):
             return url
         with egress.open_ydl(ydl_opts) as ydl:
-            raw = ydl.urlopen(url).read(self._POST_PAGE_MAX_BYTES)
+            response = ydl.urlopen(url)
+            if post_id is None:
+                post_id = facebook_group_post_id(response.url)
+                if post_id is None:
+                    return url  # a shared reel or video: yt-dlp follows the link as it always has
+            raw = response.read(self._POST_PAGE_MAX_BYTES)
         page = raw.decode("utf-8", "replace")
         # Measured: a refused or expired cookie gets the login page (~450 KB, a login
         # form, no post id at all), which is not a post without a video — the user's cookie is the cause.
@@ -246,6 +259,7 @@ class FacebookAdapter(SourceAdapter):
 
 
 _FB_GROUP_POST_PATH = re.compile(r"^/groups/[^/]+/(?:permalink|posts)/(\d+)/?$")
+_FB_GROUP_PATH = re.compile(r"^/groups/[^/]+/?$")
 _SCRIPT_BLOCK = re.compile(r"<script[^>]*>(.*?)</script>", re.S)
 _VIDEO_NODE = re.compile(r'"__typename":"Video","id":"(\d+)"')
 
@@ -256,12 +270,37 @@ def _quoted(post_id: str) -> str:
 
 
 def facebook_group_post_id(url: str) -> str | None:
-    """The post id of a Facebook group post link (`/groups/<group>/permalink|posts/<id>/`), else None."""
+    """The post id of a Facebook group post link, else None: `/groups/<group>/permalink|posts/<id>/`,
+    or a post opened from the group's feed, `/groups/<group>/?multi_permalinks=<id>` (one id only —
+    with several, which one was meant is not ours to guess)."""
     parsed = urlparse(url)
     if not _host_is(parsed.netloc.lower(), "facebook.com"):
         return None
     match = _FB_GROUP_POST_PATH.match(parsed.path)
-    return match.group(1) if match else None
+    if match:
+        return match.group(1)
+    if not _FB_GROUP_PATH.match(parsed.path):
+        return None
+    ids = parse_qs(parsed.query).get("multi_permalinks", [])
+    return ids[0] if len(ids) == 1 and ids[0].isdigit() else None
+
+
+def _group_permalink(url: str, post_id: str) -> str:
+    """The permalink form of a group post link — the one form yt-dlp itself accepts."""
+    group = urlparse(url).path.split("/")[2]
+    return f"https://www.facebook.com/groups/{group}/permalink/{post_id}/"
+
+
+def facebook_post_link(url: str) -> bool:
+    """A link that `resolve_url` may read as a group post: a post link, or a share link (which may
+    lead to one) — what the failure catalog accepts for the post page's own verdicts."""
+    return facebook_group_post_id(url) is not None or _is_share_link(url)
+
+
+def _is_share_link(url: str) -> bool:
+    # facebook.com/share/<code>/ and /share/v|r|p/<code>/ — a redirect to the shared object.
+    parsed = urlparse(url)
+    return _host_is(parsed.netloc.lower(), "facebook.com") and parsed.path.startswith("/share/")
 
 
 def facebook_post_video_id(page: str, post_id: str) -> str | None:

backend/app/downloads/errors.py

@@ -32,6 +32,7 @@ from app.downloads.adapters import (
     POST_NOT_SHOWN_MESSAGE,
     POST_WITHOUT_VIDEO_MESSAGE,
     facebook_group_post_id,
+    facebook_post_link,
 )
 from app.downloads.credentials import host_key_for_url
 
@@ -227,13 +228,14 @@ _CATALOG: tuple[_Entry, ...] = (
     # answers "No video formats found" (measured on prod jobs 66 and 145, whose videos are reshared
     # clips). With a cookie yt-dlp only ever gets the post's VIDEO link, so the same text there is about
     # that video, not the post — the category says it. Only for group post links (the adapter's own
-    # test of one): a reel with no formats is a different failure.
+    # test of one): a reel with no formats is a different failure. The adapter's own text also comes
+    # from a share link that led to a post, so its test takes those too.
     _Entry(
         host="facebook",
         markers=(POST_WITHOUT_VIDEO_MESSAGE.lower(),),
         code="format_unavailable",
         detail="facebook.group_post_video",
-        link=facebook_group_post_id,
+        link=facebook_post_link,
     ),
     _Entry(
         host="facebook",

backend/tests/test_download_errors.py

@@ -218,6 +218,12 @@ def test_our_post_without_a_video_verdict_names_the_group_link():
         "format_unavailable",
         "facebook.group_post_video",
     )
+    # The share link a post was reached by (prod job 471's kind) names the post too.
+    share = "https://www.facebook.com/share/v/19bCAHKyrd/"
+    assert dlerrors.diagnose(msg, url=share, cookie_attached=True) == (
+        "format_unavailable",
+        "facebook.group_post_video",
+    )
     assert dlerrors.diagnose(msg, url=_REEL, cookie_attached=True) == ("format_unavailable", None)
 
 

backend/tests/test_facebook_impersonation.py

@@ -27,6 +27,18 @@ _GROUP_LINK = f"https://www.facebook.com/groups/983554479435596/permalink/{_POST
 _POST_VIDEO = "1408015447857753"
 _FEED_VIDEO = "1056963063816966"  # what yt-dlp downloaded from this post's link: a feed video
 _CHROME = ImpersonateTarget.from_str("chrome")
+# The two other ways a group post reaches us (prod jobs 471 and 472, the same post): Facebook's "Share"
+# link, which redirects to the post's permalink (measured), and the link of a post opened from the
+# group's feed, which yt-dlp does not know at all ("Unsupported URL").
+_SHARE_LINK = "https://www.facebook.com/share/v/19bCAHKyrd/"
+_LANDED = (
+    f"https://www.facebook.com/groups/it.humor.and.memes/permalink/{_POST}/"
+    "?rdid=y9HAMntF7aCoAKZE&share_url=https%3A%2F%2Fwww.facebook.com%2Fshare%2Fv%2F19bCAHKyrd%2F"
+)
+_MULTI_LINK = (
+    f"https://www.facebook.com/groups/983554479435596/?multi_permalinks={_POST}"
+    "&hoisted_section_header_type=recently_seen&__cft__[0]=AZgK&__tn__=%2CO%2CP-R"
+)
 
 
 def _video_node(video_id: str) -> str:
@@ -86,6 +98,9 @@ def ydl(monkeypatch):
             body = web["page"].encode()
 
             class Response:
+                # Where the request ended up after its redirects (yt-dlp's `Response.url`).
+                url = web.get("landed") or web["opened"][-1]
+
                 def read(self, amt=None):
                     web["reads"].append(amt)
                     return body
@@ -228,12 +243,79 @@ def test_without_a_cookie_the_group_post_link_goes_to_ytdlp_unchanged(ydl, tmp_p
     assert web["extracted"] == [_GROUP_LINK]
 
 
+def test_a_share_link_to_a_group_post_downloads_the_posts_own_video(ydl, tmp_path):
+    # Prod job 471: handed the share link, yt-dlp followed it to the group page and downloaded an
+    # ad from the signed-in feed. Followed here first, it is the group post it leads to.
+    _, outcomes, web = ydl
+    web["page"], web["landed"] = _page(_POST_VIDEO), _LANDED
+    outcomes += [{"id": _POST_VIDEO}]
+
+    assert _run(tmp_path, "/cookies.txt", _SHARE_LINK) == {"id": _POST_VIDEO}
+
+    assert web["opened"] == [_SHARE_LINK]  # one request: the redirect and the post page at once
+    assert web["reads"] == [adapters.FacebookAdapter._POST_PAGE_MAX_BYTES]
+    assert web["extracted"] == [f"https://www.facebook.com/watch/?v={_POST_VIDEO}"]
+
+
+def test_a_share_link_to_a_reel_goes_to_ytdlp_unchanged(ydl, tmp_path):
+    _, outcomes, web = ydl
+    web["landed"] = _URL
+    outcomes += [{"id": "806379462330494"}]
+
+    _run(tmp_path, "/cookies.txt", _SHARE_LINK)
+
+    assert web["reads"] == []  # not a post: its page is not read
+    assert web["extracted"] == [_SHARE_LINK]  # the path that already downloads shared reels
+
+
+def test_a_share_link_that_fails_as_a_group_post_says_why(ydl, tmp_path):
+    _, _, web = ydl
+    web["page"], web["landed"] = _page(), _LANDED
+
+    with pytest.raises(RuntimeError) as exc:
+        _run(tmp_path, "/cookies.txt", _SHARE_LINK)
+
+    assert web["extracted"] == []
+    # The job's own link is the share link: the verdict still names the post.
+    assert dlerrors.diagnose(str(exc.value), url=_SHARE_LINK, cookie_attached=True) == (
+        "format_unavailable",
+        "facebook.group_post_video",
+    )
+
+
+def test_a_feed_opened_group_post_downloads_the_posts_own_video(ydl, tmp_path):
+    # Prod job 472: yt-dlp answers this link with "Unsupported URL".
+    _, outcomes, web = ydl
+    web["page"] = _page(_POST_VIDEO)
+    outcomes += [{"id": _POST_VIDEO}]
+
+    assert _run(tmp_path, "/cookies.txt", _MULTI_LINK) == {"id": _POST_VIDEO}
+
+    assert web["opened"] == [_MULTI_LINK]
+    assert web["extracted"] == [f"https://www.facebook.com/watch/?v={_POST_VIDEO}"]
+
+
+def test_without_a_cookie_a_feed_opened_group_post_goes_as_its_permalink(ydl, tmp_path):
+    # Signed out there is no page to read, but yt-dlp knows the permalink form and not this one.
+    _, outcomes, web = ydl
+    outcomes += [{"id": _POST_VIDEO}]
+
+    _run(tmp_path, None, _MULTI_LINK)
+
+    assert web["opened"] == []
+    assert web["extracted"] == [
+        f"https://www.facebook.com/groups/983554479435596/permalink/{_POST}/"
+    ]
+
+
 def test_the_group_post_link_forms_are_recognised():
     for url in (
         _GROUP_LINK,
         f"https://www.facebook.com/groups/983554479435596/posts/{_POST}/",
         f"https://m.facebook.com/groups/983554479435596/permalink/{_POST}/?",
         f"https://www.facebook.com/groups/chatgptmagyarul/posts/{_POST}",
+        _MULTI_LINK,
+        f"https://www.facebook.com/groups/983554479435596?multi_permalinks={_POST}",
     ):
         assert adapters.facebook_group_post_id(url) == _POST, url
     for url in (
@@ -241,6 +323,10 @@ def test_the_group_post_link_forms_are_recognised():
         f"https://www.facebook.com/watch/?v={_POST_VIDEO}",
         f"https://facebook.com.evil/groups/x/permalink/{_POST}/",
         "https://www.facebook.com/groups/983554479435596/",
+        # More than one post in the link: which one was meant is not ours to guess.
+        f"https://www.facebook.com/groups/983554479435596/?multi_permalinks={_POST},{_POST_VIDEO}",
+        f"https://www.facebook.com/groups/983554479435596/members/?multi_permalinks={_POST}",
+        _SHARE_LINK,  # a share link is a post only once it is followed (`resolve_url`)
     ):
         assert adapters.facebook_group_post_id(url) is None, url
     assert adapters.SourceAdapter().resolve_url(_GROUP_LINK, {"cookiefile": "/c"}) == _GROUP_LINK