b7cdc3b1d1
This works around b/156126072 which introduces app startup latencies that are multiplied with binder txn tracing enabled. Skiping tracing interfaces without binder txn names avoids tracing interfaces like ContentProviders and app-defined interfaces. This is still an improvement on the status quo since we still trace the auto-generated framework interfaces. When we fix sporadic binder calls in the app-startup critical path, we can remove the restriction and enable tracing more broadly. Note that we still keep the old behavior of 'best effort' trace names when binder call stack sampling tracing is enabled Test: Manual Bug: 246650647 Change-Id: Id04b701f97241cbd5d14bbb368fdd8527b6f8c93