From 739f651a169a6ac0840726774fef675b035a1304 Mon Sep 17 00:00:00 2001
From: Abdessamad Derraz <3028866+Abdess@users.noreply.github.com>
Date: Thu, 5 Nov 2026 11:58:22 +0100
Subject: [PATCH] feat: check refs into a submodule

idtech4a++ cites xash3d-fwgs filesystem/filesystem.c, starship and
ghostship Torch src/Companion.cpp: files of submodules of the declared
repository. No blob of the declared tree answers, so the parts stayed
external and a bump moved the submodule commit with nothing checked.
A project word now also resolves to the one submodule named after it,
and a file under a submodule is read in the repository .gitmodules
names, at the commit the declared revision pins it to.
---
 scripts/profile_sync.py    | 50 +++++++++++++++++++++++++++++++-----
 scripts/upstream.py        | 52 ++++++++++++++++++++++++++++++++++++++
 tests/test_profile_sync.py | 44 ++++++++++++++++++++++++++++++++
 3 files changed, 140 insertions(+), 6 deletions(-)

diff --git a/scripts/profile_sync.py b/scripts/profile_sync.py
index 741ac22b7..9b9448a93 100644
--- a/scripts/profile_sync.py
+++ b/scripts/profile_sync.py
@@ -1463,14 +1463,40 @@ def tag_pinned_views(
     return tagged
 
 
+def resolve_through_gitlinks(path: str, links: Iterable[Iterable[str]]) -> str | None:
+    """Path of `project file` inside the one submodule named after the project.
+
+    idtech4a++ cites `xash3d-fwgs filesystem/filesystem.c` and starship
+    `Torch src/Companion.cpp`: the declared tree holds a gitlink there, not
+    the files, so no blob path answers and the part stayed external.
+    """
+    word, _, rest = path.partition(" ")
+    rest = rest.strip().lstrip("/")
+    if not rest:
+        return None
+    word = word.lower()
+    for paths in links:
+        matches = [
+            link for link in paths or ()
+            if word in link.rsplit("/", 1)[-1].lower()
+        ]
+        if len(matches) == 1:
+            return f"{matches[0]}/{rest}"
+        if matches:
+            return None
+    return None
+
+
 def _resolve_project_words(
-    refs: list, tree_order: Callable[[], Iterable[list[str] | None]]
+    refs: list, tree_order: Callable[[], Iterable[list[str] | None]],
+    link_order: Callable[[], Iterable[Iterable[str]]] = lambda: (),
 ) -> list:
     """Point `project file` parts at the declared tree that vendors them.
 
-    tree_order() yields the trees to search, pin before HEAD. It is only
-    called for a part that needs it, since reading a HEAD tree costs a
-    request. Parts no tree carries stay external.
+    tree_order() yields the trees to search, pin before HEAD, and
+    link_order() the submodule paths of the declared trees at their pins.
+    Each is only called for a part that needs it, since reading a tree costs
+    a request. Parts no tree carries stay external.
     """
     resolved: dict[str, str | None] = {}
     out = []
@@ -1479,7 +1505,9 @@ def _resolve_project_words(
         for part in parts:
             if is_external_citation(part.path):
                 if part.path not in resolved:
-                    resolved[part.path] = resolve_project_word(part.path, tree_order())
+                    resolved[part.path] = resolve_project_word(
+                        part.path, tree_order()
+                    ) or resolve_through_gitlinks(part.path, link_order())
                 if resolved[part.path]:
                     part = replace(part, path=resolved[part.path])
             rewritten.append(part)
@@ -1860,7 +1888,14 @@ def build_report(
         if key in mute:
             return None
         try:
-            return upstream.fetch_file(repo, sha, wanted, cache_dir, offline)
+            found = upstream.fetch_file(repo, sha, wanted, cache_dir, offline)
+            if found is None and "/" in wanted:
+                # A file of a submodule is read at the commit its gitlink
+                # pins in this revision, so a bump moves it with the port.
+                inner = upstream.through_gitlink(repo, sha, wanted, cache_dir, offline)
+                if inner is not None:
+                    found = upstream.fetch_file(*inner, cache_dir, offline)
+            return found
         except (upstream.RateLimitError, upstream.NotCachedError):
             raise
         except upstream.UpstreamError as exc:
@@ -1917,6 +1952,9 @@ def build_report(
         lambda: itertools.chain(
             map(_pin_tree_for, views), map(lambda v: _context_for(v)[1], views)
         ),
+        lambda: (
+            upstream.list_gitlinks(v.repo, v.pin, cache_dir, offline) for v in views
+        ),
     )
 
     lines_cache: dict[tuple[str, str, int | None], list[str] | None] = {}
diff --git a/scripts/upstream.py b/scripts/upstream.py
index bf0b2f7e6..131d0bc83 100644
--- a/scripts/upstream.py
+++ b/scripts/upstream.py
@@ -1115,6 +1115,58 @@ def list_gitlinks(
     }
 
 
+def submodule_urls(text: str, base: Repo) -> dict[str, Repo]:
+    """Submodule path -> repository, read from a .gitmodules text.
+
+    A relative url (`../xash3d-fwgs.git`) names a sibling of the repository
+    that declares it, on the same forge.
+    """
+    repos: dict[str, Repo] = {}
+    path = url = None
+    for line in [*text.splitlines(), "[end]"]:
+        stripped = line.strip()
+        if stripped.startswith("["):
+            if path and url:
+                if url.startswith(("../", "./")):
+                    segments = [base.owner, base.name]
+                    for part in url.split("/"):
+                        if part == "..":
+                            segments.pop()
+                        elif part not in (".", ""):
+                            segments.append(part)
+                    url = f"https://{base.host}/{'/'.join(segments)}"
+                repo = parse_repo(url)
+                if repo is not None:
+                    repos[path] = repo
+            path = url = None
+            continue
+        key, _, value = stripped.partition("=")
+        if key.strip() == "path":
+            path = value.strip()
+        elif key.strip() == "url":
+            url = value.strip()
+    return repos
+
+
+def through_gitlink(
+    repo: Repo, sha: str, path: str, cache_dir: str, offline: bool = False
+) -> tuple[Repo, str, str] | None:
+    """The submodule repository, commit and inner path a path falls under.
+
+    starship cites `Torch src/Companion.cpp`, a file of the Torch submodule:
+    the declared tree holds only the gitlink, which pins the revision.
+    """
+    links = list_gitlinks(repo, sha, cache_dir, offline)
+    link = next((p for p in links if path.startswith(p + "/")), None)
+    if link is None:
+        return None
+    modules = fetch_file(repo, sha, ".gitmodules", cache_dir, offline)
+    target = submodule_urls("\n".join(modules or ()), repo).get(link)
+    if target is None:
+        return None
+    return target, links[link], path[len(link) + 1:]
+
+
 def commit_date(repo: Repo, sha: str, cache_dir: str, offline: bool = False) -> str | None:
     """When a revision was committed, as the forge reports it (ISO 8601)."""
     if repo.family == "gitlab":
diff --git a/tests/test_profile_sync.py b/tests/test_profile_sync.py
index b7f2d424a..025a56fe2 100644
--- a/tests/test_profile_sync.py
+++ b/tests/test_profile_sync.py
@@ -3950,6 +3950,50 @@ class TestCollectCitations(unittest.TestCase):
         self.assertIsNone(resolve("munt ROMInfo.cpp", [tree]))
         self.assertIsNone(resolve("mupen64plus Config.cpp", [tree]))
 
+    def test_project_word_resolves_inside_a_submodule(self):
+        """idtech4a++ and starship cite files of a submodule. The declared
+        tree holds only the gitlink, so the parts stayed external and moved
+        unchecked with every bump."""
+        links = ["Q3E/src/main/jni/xash3d/xash3d-fwgs", "tools/Torch", "libs/zlib"]
+        resolve = profile_sync.resolve_through_gitlinks
+        self.assertEqual(
+            resolve("xash3d-fwgs filesystem/filesystem.c", [links]),
+            "Q3E/src/main/jni/xash3d/xash3d-fwgs/filesystem/filesystem.c",
+        )
+        self.assertEqual(
+            resolve("Torch src/Companion.cpp", [links]), "tools/Torch/src/Companion.cpp"
+        )
+        self.assertIsNone(resolve("munt ROMInfo.cpp", [links]))
+        self.assertIsNone(resolve("lib x.c", [["a/lib1", "b/lib2"]]))
+
+    def test_a_submodule_file_is_read_at_its_pinned_commit(self):
+        port = upstream.make_repo("github.com", "glKarin", "com.n0n3m4.diii4a")
+        files = {
+            (port.slug, "p1", ".gitmodules"): [
+                '[submodule "Q3E/src/main/jni/xash3d/xash3d-fwgs"]',
+                "\tpath = Q3E/src/main/jni/xash3d/xash3d-fwgs",
+                "\turl = ../xash3d-fwgs.git",
+            ],
+            ("glKarin/xash3d-fwgs", "s1", "filesystem/filesystem.c"): ["getenv(x)"],
+        }
+
+        def fetch(repo, sha, path, *_):
+            return files.get((repo.slug, sha, path))
+
+        with mock.patch.object(upstream, "fetch_file", side_effect=fetch), \
+                mock.patch.object(
+                    upstream, "list_gitlinks",
+                    return_value={"Q3E/src/main/jni/xash3d/xash3d-fwgs": "s1"},
+                ):
+            target = upstream.through_gitlink(
+                port, "p1",
+                "Q3E/src/main/jni/xash3d/xash3d-fwgs/filesystem/filesystem.c", "c",
+            )
+        self.assertEqual(
+            (target[0].slug, target[1], target[2]),
+            ("glKarin/xash3d-fwgs", "s1", "filesystem/filesystem.c"),
+        )
+
     def test_run_is_rendered_with_its_written_path(self):
         parts = [RefPart("mupen64plus-video-rice/upstream/src/Config.cpp", 38, 38, "38")]
         self.assertEqual(