fix: point source refs at paths held at the pin

trident cites the Panda3DS fork 3dsTrident vendors as a submodule, now
its upstream at the submodule pin 69095ee6. pce-emu, uqm and stonesoup
wrote paths from a subdirectory (PCE.emu/, sc2/, crawl-ref/), nesoid
named two Java files without their directory; lines unchanged, each read
at the pin.

The pages no longer take a binary citation (DesktopKinect.dll, an
address continuing a BAM.dll range) for a source path, and a Forgejo
tree is read page by page: mirrors/Ryujinx holds 4704 entries, 1000 a
page.

scripts/profile_sync.py

@@ -46,6 +46,8 @@ from sourceref import (  # noqa: F401  (re-exported for callers and tests)  # no
     _PROSE_LIST_DELIMITERS,
     _REF_RE,
     _SPLIT_RE,
+    _BINARY_SUFFIXES,
+    _HEX_ADDRESS_RE,
     PROSE_CITE_RE,
     PROSE_CONT_RE,
     PROSE_CONT_SPACED_RE,
@@ -58,6 +60,7 @@ from sourceref import (  # noqa: F401  (re-exported for callers and tests)  # no
     _prose_external,
     _prose_runs,
     _trim_prose,
+    is_binary_citation,
     parse_source_ref,
     ref_prose_text,
     ref_role,
@@ -195,33 +198,9 @@ def is_external_citation(path: str) -> bool:
     return "/" not in head and "." not in head
 
 
-_BINARY_SUFFIXES = (
-    ".dll", ".exe", ".so", ".dylib", ".apk", ".elf", ".o", ".a", ".jar", ".dex",
-)
-_HEX_ADDRESS_RE = re.compile(r"\b0x[0-9a-fA-F]{2,}")
 _VERSION_WORD_RE = re.compile(r"v?\d+(\.\d+)+[\w.-]*")
 
 
-def is_binary_citation(path: str) -> bool:
-    """True when a ref locates code inside a shipped binary, not a source file.
-
-    `BAM.dll 1.5-408 .text:0x100c3960-0x100c3a97`, `libmain.so LoadNES
-    0x3bb49` and `.rdata dialog filter` were read by disassembly. No revision
-    of any repository holds those bytes, so the ref can neither anchor nor
-    break: 77 parts of `future-pinball-fploader` alone were reported GONE.
-    """
-    if _HEX_ADDRESS_RE.search(path):
-        return True
-    if " " not in path:
-        # A bare `DesktopKinect.dll` names a shipped library, never a source
-        # file: no directory, no line, only its import table to read.
-        return "/" not in path and path.lower().endswith(_BINARY_SUFFIXES)
-    head = path.split(" ", 1)[0]
-    if head.lower().endswith(_BINARY_SUFFIXES):
-        return True
-    return head.startswith(".") and "/" not in head
-
-
 def strip_repo_word(
     path: str, known: frozenset[str] | set[str] = frozenset()
 ) -> str:

scripts/sitesources.py

@@ -14,7 +14,9 @@ import urllib.request
 
 import upstream
 from siterender import _by_mode
-from sourceref import _prose_runs, ref_role, source_ref_values, split_source_ref
+from sourceref import (
+    _prose_runs, is_binary_citation, ref_role, source_ref_values, split_source_ref,
+)
 
 
 def _repo_pins(profile: dict, field: str, label: str) -> list[tuple[str, str, str]]:
@@ -309,7 +311,10 @@ def _source_ref_markdown(profile: dict, value: str | dict | None) -> str:
             # broke 470 links. Two repositories and no tree to read: no link.
             selected: tuple[upstream.Repo, str] | None = None
             real_path = part.path
-            if path_re.fullmatch(part.path):
+            # A shipped binary (`DesktopKinect.dll`, `0x10037130-0x100371de`
+            # continuing a BAM.dll range) is no path of either repository:
+            # without a cached tree the single declared one took it.
+            if path_re.fullmatch(part.path) and not is_binary_citation(part.path):
                 if current_role:
                     # Only the repository the part names, or no link at all.
                     selected = by_role.get(current_role)

scripts/sourceref.py

@@ -25,6 +25,32 @@ _LOCATED_RE = re.compile(r"^[^\s:]+:\d+(?:-\d+)?$")
 REPO_ROLES = ("source", "upstream")
 
 
+_BINARY_SUFFIXES = (
+    ".dll", ".exe", ".so", ".dylib", ".apk", ".elf", ".o", ".a", ".jar", ".dex",
+)
+_HEX_ADDRESS_RE = re.compile(r"\b0x[0-9a-fA-F]{2,}")
+
+
+def is_binary_citation(path: str) -> bool:
+    """True when a ref locates code inside a shipped binary, not a source file.
+
+    `BAM.dll 1.5-408 .text:0x100c3960-0x100c3a97`, `libmain.so LoadNES
+    0x3bb49` and `.rdata dialog filter` were read by disassembly. No revision
+    of any repository holds those bytes, so the ref can neither anchor nor
+    break: 77 parts of `future-pinball-fploader` alone were reported GONE.
+    """
+    if _HEX_ADDRESS_RE.search(path):
+        return True
+    if " " not in path:
+        # A bare `DesktopKinect.dll` names a shipped library, never a source
+        # file: no directory, no line, only its import table to read.
+        return "/" not in path and path.lower().endswith(_BINARY_SUFFIXES)
+    head = path.split(" ", 1)[0]
+    if head.lower().endswith(_BINARY_SUFFIXES):
+        return True
+    return head.startswith(".") and "/" not in head
+
+
 def ref_role(path: str) -> tuple[str, str]:
     """The role word a cited path opens with, and the path without it."""
     word, _, rest = path.partition(" ")

scripts/upstream.py

@@ -780,12 +780,28 @@ def list_tree(
     payload = _api(url, cache_dir, offline)
     if not isinstance(payload, dict):
         return [], True
+    entries = list(payload.get("tree") or [])
+    truncated = bool(payload.get("truncated"))
+    # Forgejo pages a recursive tree, 1000 entries at a time, and says how
+    # many there are: mirrors/Ryujinx holds 4704, and read as one page every
+    # citation past the first thousand was unverifiable. GitHub reports no
+    # count, its truncation is final.
+    total = payload.get("total_count")
+    if truncated and repo.family == "forgejo" and isinstance(total, int):
+        for page in range(2, MAX_TREE_PAGES + 1):
+            if len(entries) >= total:
+                break
+            more = _api(f"{url}&page={page}", cache_dir, offline)
+            if not isinstance(more, dict) or not more.get("tree"):
+                break
+            entries.extend(more["tree"])
+        truncated = len(entries) < total
     paths = [
         entry["path"]
-        for entry in payload.get("tree") or []
+        for entry in entries
         if isinstance(entry, dict) and entry.get("type") == "blob" and entry.get("path")
     ]
-    return paths, bool(payload.get("truncated"))
+    return paths, truncated
 
 
 def list_gitlinks(