Fix ffmpeg backend misidentifying metadata as audio stream info

audioread/ffdec.py

@@ -229,7 +229,7 @@ def _get_info(self):
                 raise UnsupportedError()
             elif 'duration:' in line:
                 out_parts.append(line)
-            elif 'audio:' in line:
+            elif 'stream #' in line and 'audio:' in line:
                 out_parts.append(line)
                 self._parse_info(''.join(out_parts))
                 break

test/test_ffdec.py

@@ -0,0 +1,82 @@
+# This file is part of audioread.
+# Tests for ffdec.py parsing logic.
+
+import io
+
+import pytest
+
+from audioread.ffdec import FFmpegAudioFile
+
+
+class _FakeProc:
+    """Minimal subprocess stand-in for testing _get_info without ffmpeg."""
+
+    def __init__(self, stderr_lines):
+        self.stderr = io.BytesIO(b"".join(
+            line if line.endswith(b"\n") else line + b"\n"
+            for line in stderr_lines
+        ))
+        self.stdout = io.BytesIO(b"")
+        self.returncode = 0
+
+    def poll(self):
+        pass
+
+    def kill(self):
+        pass
+
+    def wait(self):
+        pass
+
+
+def _parse_from_lines(stderr_lines):
+    """Call FFmpegAudioFile._get_info with fake process stderr lines.
+
+    Returns an object with `channels`, `samplerate`, and `duration`
+    attributes as parsed by the real _get_info / _parse_info code, without
+    running ffmpeg.
+    """
+    obj = FFmpegAudioFile.__new__(FFmpegAudioFile)
+    obj.proc = _FakeProc(stderr_lines)
+    FFmpegAudioFile._get_info(obj)
+    return obj
+
+
+NORMAL_FFMPEG_OUTPUT = [
+    b"Input #0, mp3, from 'test.mp3':",
+    b"  Duration: 00:00:02.0, start: 0.000000, bitrate: 128 kb/s",
+    b"    Stream #0:0: Audio: mp3, 44100 Hz, stereo, fltp, 128 kb/s",
+]
+
+# File whose metadata contains the substring "audio:" – triggers the bug.
+METADATA_AUDIO_FFMPEG_OUTPUT = [
+    b"Input #0, mp3, from 'out.mp3':",
+    b"  Metadata:",
+    b"    description     : audio: broken",
+    b"    encoder         : Lavf58.29.100",
+    b"  Duration: 00:00:02.04, start: 0.025057, bitrate: 129 kb/s",
+    b"    Stream #0:0: Audio: mp3, 44100 Hz, stereo, fltp, 128 kb/s",
+]
+
+
+def test_normal_parse():
+    """Normal ffmpeg output is parsed correctly."""
+    af = _parse_from_lines(NORMAL_FFMPEG_OUTPUT)
+    assert af.channels == 2
+    assert af.samplerate == 44100
+    assert af.duration == pytest.approx(2.0, abs=0.2)
+
+
+def test_metadata_containing_audio_colon():
+    """Metadata field 'audio: …' must not be mistaken for the stream line.
+
+    Regression test for https://github.com/beetbox/audioread/issues/119.
+    When a file has a metadata tag whose value contains the substring
+    'audio:', the previous code matched that metadata line instead of the
+    real stream descriptor, resulting in channels=0, samplerate=0,
+    and duration=0.
+    """
+    af = _parse_from_lines(METADATA_AUDIO_FFMPEG_OUTPUT)
+    assert af.channels == 2
+    assert af.samplerate == 44100
+    assert af.duration == pytest.approx(2.04, abs=0.2)