todo_hooks: skip unpushed shas, scan upstream

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

todo_hooks/gh.py

@@ -17,6 +17,10 @@ class GhError(Exception):
     """gh missing, offline, failing or answering nonsense."""
 
 
+class GhNotFound(GhError):
+    """GitHub does not know the commit (HTTP 404/422), e.g. an unpushed sha."""
+
+
 @dataclass(frozen=True)
 class PullRequest:
     """A pull request GitHub links to a commit."""
@@ -44,7 +48,10 @@ def run_gh(args: list[str]) -> str:
         msg = "gh timed out"
         raise GhError(msg) from err
     if done.returncode != 0:
-        raise GhError(done.stderr.strip() or f"gh exit {done.returncode}")
+        message = done.stderr.strip() or f"gh exit {done.returncode}"
+        if "(HTTP 404)" in message or "(HTTP 422)" in message:
+            raise GhNotFound(message)
+        raise GhError(message)
     return done.stdout
 
 

todo_hooks/import_pr.py

@@ -21,6 +21,7 @@ if TYPE_CHECKING:
     from todo_hooks.context import Context
 
 MAX_COMMITS = 30
+_PAIR_FIELDS = 2
 USAGE = (
     "usage: python3 -m todo_hooks import-pr "
     "post-merge | post-rewrite [rebase] | --range A..B"
@@ -38,11 +39,15 @@ def commits_post_merge(root: Path) -> list[str]:
 
 
 def commits_post_rewrite(root: Path, command: str, pairs: str) -> list[str]:
-    """New shas from git's ``old new`` stdin lines; only for a rebase."""
+    """Rebase only: ORIG_HEAD..HEAD plus the new shas of stdin ``old new`` lines."""
     if command != "rebase":
         return []
-    shas = [row.split()[1] for row in pairs.splitlines() if len(row.split()) >= 2]  # noqa: PLR2004
-    return shas[:MAX_COMMITS] or commits_post_merge(root)
+    shas = commits_post_merge(root)
+    for row in pairs.splitlines():
+        fields = row.split()
+        if len(fields) >= _PAIR_FIELDS and fields[1] not in shas:
+            shas.append(fields[1])
+    return shas[:MAX_COMMITS]
 
 
 def read_stdin(timeout: float = 1.0) -> str:
@@ -74,6 +79,15 @@ def _add(client: TaskClient, repo: str, number: int, item: TodoItem) -> bool:
     return True
 
 
+def _prs_of(
+    slug: tuple[str, str], sha: str, runner: Callable[[list[str]], str] | None
+) -> list[gh.PullRequest]:
+    try:
+        return gh.merged_prs(slug[0], slug[1], sha, runner)
+    except gh.GhNotFound:
+        return []
+
+
 def import_for_commits(
     ctx: Context, shas: list[str], runner: Callable[[list[str]], str] | None
 ) -> int:
@@ -89,7 +103,7 @@ def import_for_commits(
     client: TaskClient | None = None
     try:
         for sha in shas:
-            for pr in gh.merged_prs(slug[0], slug[1], sha, runner):
+            for pr in _prs_of(slug, sha, runner):
                 items = parse_todo_items(pr.body)
                 if not pr.merged or pr.number in seen or not items:
                     continue

todo_hooks/systemd/todo-drift.service

@@ -6,4 +6,4 @@ Type=oneshot
 ExecStart=%h/src/utils/scripts/todo_hooks.sh sweep
 Nice=19
 IOSchedulingClass=idle
-TimeoutStartSec=30min
+TimeoutStartSec=infinity

todo_hooks/tests/test_import_pr_commits.py

@@ -4,7 +4,7 @@ from __future__ import annotations
 
 from typing import TYPE_CHECKING
 
-from todo_hooks import import_pr
+from todo_hooks import gh, import_pr
 from todo_hooks.tests._fakes import FakeGh, line, make_pr
 from todo_hooks.tests._pr_helpers import pr_repo, search_client
 from todo_hooks.tests.conftest import git
@@ -160,3 +160,45 @@ def test_nothing_new_to_import_is_silent(
     ctx, out = make_ctx(search_client(existing), root)
     assert import_pr.import_for_commits(ctx, [sha], FakeGh({sha: [make_pr(7)]})) == 0
     assert out == []
+
+
+def test_rewrite_includes_upstream_commits_after_pull_rebase(tmp_path: Path) -> None:
+    upstream = tmp_path / "up.git"
+    git(tmp_path, "init", "-q", "--bare", "-b", "main", str(upstream))
+    mine, other = tmp_path / "mine", tmp_path / "other"
+    git(tmp_path, "clone", "-q", str(upstream), str(mine))
+    (mine / "a").write_text("a")
+    git(mine, "add", "a")
+    git(mine, "commit", "-q", "-m", "base")
+    git(mine, "push", "-q", "origin", "HEAD:main")
+    git(tmp_path, "clone", "-q", str(upstream), str(other))
+    (other / "b").write_text("b")
+    git(other, "add", "b")
+    git(other, "commit", "-q", "-m", "arrives via pull")
+    git(other, "push", "-q", "origin", "HEAD:main")
+    arrived = git(other, "rev-parse", "HEAD")
+    (mine / "c").write_text("c")
+    git(mine, "add", "c")
+    git(mine, "commit", "-q", "-m", "local only")
+    git(mine, "pull", "-q", "--rebase", "origin", "main")
+    replayed = git(mine, "rev-parse", "HEAD")
+    shas = import_pr.commits_post_rewrite(mine, "rebase", f"old {replayed}\n")
+    assert arrived in shas
+    assert shas.count(replayed) == 1
+
+
+def test_unpushed_commit_does_not_stop_the_scan(
+    make_repo: Callable[..., Path],
+    make_ctx: Callable[..., tuple[Context, list[str]]],
+) -> None:
+    root = pr_repo(make_repo, 1)
+    fake = FakeGh({"known": [make_pr(7)]})
+
+    def runner(args: list[str]) -> str:
+        if "unpushed" in args[1]:
+            raise gh.GhNotFound("No commit found (HTTP 422)")
+        return fake(args)
+
+    client = search_client()
+    ctx, _ = make_ctx(client, root)
+    assert import_pr.import_for_commits(ctx, ["unpushed", "known"], runner) == 3

todo_hooks/tests/test_pr_parse_gh.py

@@ -71,3 +71,14 @@ def test_merged_prs_parses_fields() -> None:
     prs = gh.merged_prs("kuhyx", "demo", "abc", fake)
     assert [(p.number, p.merged) for p in prs] == [(5, True), (6, False)]
     assert fake.calls[0] == ["api", "repos/kuhyx/demo/commits/abc/pulls"]
+
+
+def test_run_gh_unknown_commit_is_not_found(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    fake = tmp_path / "gh"
+    fake.write_text('#!/bin/sh\necho "gh: No commit found (HTTP 422)" >&2\nexit 1\n')
+    fake.chmod(0o755)
+    monkeypatch.setenv("PATH", f"{tmp_path}:/usr/bin")
+    with pytest.raises(gh.GhNotFound):
+        gh.run_gh(["x"])