MEDIUM: audio download in fingerprint.py has no byte-count cap — pirate server can stream unbounded bytes

[walidboulanouar]

Problem

apps/scanner/fingerprint.py lines 227–250:

async for chunk in resp.aiter_bytes(chunk_size=65536):
    handle.write(chunk)

The download has a 60-second timeout but no byte-count limit. A piracy server returning HTTP 200 with an infinite body will fill disk until the OS kills the process or the timeout fires. At 10 Mbps for 60 seconds that is ~75 MB per fingerprint job.

Also: subprocess.run for fpcalc and ffmpeg at lines 188–197 has no timeout= parameter — a maliciously crafted audio file or codec bug can hang a thread-pool slot indefinitely.

Fix

MAX_DOWNLOAD_BYTES = 50 * 1024 * 1024
total = 0
async for chunk in resp.aiter_bytes(chunk_size=65536):
    total += len(chunk)
    if total > MAX_DOWNLOAD_BYTES:
        raise FingerprintError(f'download exceeded {MAX_DOWNLOAD_BYTES} bytes')
    handle.write(chunk)

And for subprocess: subprocess.run(cmd, timeout=CLIP_SECONDS + 30, ...) with except subprocess.TimeoutExpired: raise FingerprintError('ffmpeg timed out').

[walidboulanouar]

Verified fixed in commit cd7472b — fingerprint downloads capped at MAX_DOWNLOAD_BYTES=50MB with ffmpeg/fpcalc subprocess timeouts (apps/scanner/fingerprint.py).