videos: hide the transcoder's temporary output from the file refresh (#731)

ffmpeg wrote `<stem>.transcode.<container>` next to the source.  A refresh
that overlapped the transcode discovered that file, probed the half-written
output (errors in the log), and created a FileGroup and Video for a path the
final rename removed moments later.

The temporary file is now `.<stem>.transcode.<container>`: every discovery
path already skips hidden files, and staying in the same directory keeps
the rename over the final path atomic.  `transcode_tmp_path` is the single
place the name is built.

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>

modules/videos/test/test_transcode_job.py

@@ -7,7 +7,7 @@ from unittest import mock
 import pytest
 
 from modules.videos.models import Video
-from modules.videos.transcode import make_ffmpeg_progress_parser, transcode_video_job
+from modules.videos.transcode import make_ffmpeg_progress_parser, transcode_video_job, transcode_tmp_path
 from wrolpi import jobs
 from wrolpi.cmd import CommandResult
 from wrolpi.files.models import FileGroup
@@ -293,3 +293,28 @@ async def test_video_transcode_api_fragmented(async_client, test_session, video_
     request, response = await async_client.post(f'/api/videos/{video.file_group_id}/transcode',
                                                 content=json.dumps({'container': 'mkv', 'fragmented': True}))
     assert response.status == 400
+
+
+@pytest.mark.asyncio
+async def test_transcode_temp_file_is_not_discovered_by_refresh(test_session, test_directory, async_client,
+                                                                 video_factory, refresh_files):
+    """A refresh that runs while ffmpeg is still writing the transcoder's output must not model it.
+
+    The temporary file is renamed over the final path moments later; a FileGroup for it would point
+    at a file that no longer exists, and probing a half-written file logs errors.
+    """
+    video_path = test_directory / 'videos/NO CHANNEL/movie.mp4'
+    video_path.parent.mkdir(parents=True, exist_ok=True)
+    video = video_factory(with_video_file=video_path)
+    test_session.commit()
+
+    tmp_path = transcode_tmp_path(video_path, 'mp4')
+    shutil.copy(PROJECT_DIR / 'test/big_buck_bunny_720p_1mb.mp4', tmp_path)
+    assert tmp_path.is_file()
+
+    await refresh_files()
+
+    primary_paths = {str(i.primary_path) for i in test_session.query(FileGroup)}
+    assert str(video_path) in primary_paths
+    assert str(tmp_path) not in primary_paths, 'the transcoder temporary file was indexed'
+    assert not any(str(tmp_path) in str(p) for p in primary_paths)

modules/videos/transcode.py

@@ -149,6 +149,15 @@ async def _acquire_transcode_lock():
     return lock
 
 
+def transcode_tmp_path(video_path: pathlib.Path, container: str) -> pathlib.Path:
+    """Where ffmpeg writes its output before it is renamed over the final path.
+
+    Hidden (dot-prefixed) in the video's own directory: a refresh skips hidden files, so a refresh that
+    overlaps the transcode neither probes the half-written output nor creates a FileGroup for a file
+    that the rename removes moments later; and same-directory keeps the final rename atomic."""
+    return video_path.with_name(f'.{video_path.stem}.transcode.{container}')
+
+
 async def transcode_video_file(video_path: pathlib.Path,
                                target_vcodec: Optional[str] = None,
                                target_acodec: Optional[str] = None,
@@ -363,7 +372,7 @@ async def _transcode_video_file(video_path: pathlib.Path, target_vcodec: Optiona
         mux_args = ()
 
     final_path = video_path.with_suffix(f'.{container}')
-    tmp_path = video_path.with_suffix(f'.transcode.{container}')
+    tmp_path = transcode_tmp_path(video_path, container)
     cmd = (FFMPEG_BIN, '-y',
            '-i', video_path,
            *map_args,