empty-catch: sdks/python/pmxt/client.py:518 — `except Exception: pass` silently swallows sidecar restart failure in _fetch_with_retry

[realfishsam]

Violation

                # Connection failed -- try to restart sidecar on first failure
                if attempt == 0 and not self.pmxt_api_key:
                    try:
                        self._server_manager.ensure_server_running()
                    except Exception:
                        pass
                time.sleep(delays[attempt])

Location

sdks/python/pmxt/client.py:514-520 (in _fetch_with_retry)

Why It Matters

When the initial API call fails with a connection error, _fetch_with_retry makes a best-effort attempt to restart the sidecar via ensure_server_running() before retrying. If that restart attempt itself raises — for any reason, including bugs in the server-manager code, permission errors, or port conflicts — the exception is discarded with a bare except Exception: pass and no log line. The caller only ever sees the original connection error re-raised after retries are exhausted, with zero trace of why the sidecar restart didn't help. This makes diagnosing "the SDK couldn't recover a dead sidecar" issues much harder than necessary.

Suggested Fix

Log the swallowed exception at DEBUG/WARNING before continuing the retry loop:

                    except Exception as exc:
                        logger.debug(
                            "Sidecar restart attempt failed",
                            {"error": str(exc)},
                        )

---

Found by automated code hygiene audit