results from JNI performance testing

Author: thestinger

static/releases.html

@@ -593,7 +593,7 @@ <h3><a href="#2025040400">2025040400</a></h3>
                         <li>Sandboxed Google Play compatibility layer: remove StatsManager from hidden services and replace that approach with stubbing out all of the methods since Play services recently introduced new code using it that's missing a null check and triggers a null pointer exception which has blocked us from pushing out the newer versions of Play services beyond our Alpha channel</li>
                         <li>Network Location: switch to making at most a single request to the service per position estimation by requesting up to 40 Wi-Fi APs at once</li>
                         <li>Network Location: optimize making requests to the service for Wi-Fi AP data</li>
-                        <li>Network Location: optimize Rust JNI bindings to the point that it no longer causes a noticeable overhead without introducing unsafe code (it could be optimized further with unsafe code to use more of the standard JNI optimizations but does not appear to be needed)</li>
+                        <li>Network Location: optimize Rust JNI bindings to the point that it no longer causes a noticeable overhead without introducing unsafe code (it could be optimized further with unsafe code to cache more JNI binding work but the difference is barely measurable so we don't plan to do it)</li>
                         <li>Network Location: use correct Accept header value to more closely match macOS to avoid future compatibility issues</li>
                         <li>fix upstream system_server crash from null pointer exception in F2fsUtils</li>
                         <li>add infrastructure for more restricted access to global and per-user settings instead of allowing all system apps to read them and all privileged systems apps with the WRITE_SECURE_SETTINGS privileged permission to write them</li>