[Lldb-commits] [lldb] [LLDB][Telemetry]Define DebuggerTelemetryInfo and related methods (PR #127696)

[Jonas Devlieghere via lldb-commits]

================
@@ -761,12 +767,29 @@ void Debugger::InstanceInitialize() {
 
 DebuggerSP Debugger::CreateInstance(lldb::LogOutputCallback log_callback,
                                     void *baton) {
+#ifdef LLVM_BUILD_TELEMETRY
+  lldb_private::telemetry::SteadyTimePoint start_time =
+      std::chrono::steady_clock::now();
+#endif
   DebuggerSP debugger_sp(new Debugger(log_callback, baton));
   if (g_debugger_list_ptr && g_debugger_list_mutex_ptr) {
     std::lock_guard<std::recursive_mutex> guard(*g_debugger_list_mutex_ptr);
     g_debugger_list_ptr->push_back(debugger_sp);
   }
   debugger_sp->InstanceInitialize();
+
+#ifdef LLVM_BUILD_TELEMETRY
+  if (auto *telemetry_manager = telemetry::TelemetryManager::getInstance()) {
+    if (telemetry_manager->getConfig()->EnableTelemetry) {
+      lldb_private::telemetry::DebuggerTelemetryInfo entry;
+      entry.start_time = start_time;
+      entry.end_time = std::chrono::steady_clock::now();
+      entry.debugger = debugger_sp.get();
+      telemetry_manager->atDebuggerStartup(&entry);
+    }
+  }
+#endif
----------------
JDevlieghere wrote:

When I introduced the CMake variable, I had something in mind like what we do for signposts on macOS (`LLVM_SUPPORT_XCODE_SIGNPOSTS`) where we have a wrapper and then compile out support when it's not available. Admittedly, that's not what I did by conditionalizing the library, so that's on me and I'm happy to rectify that. 

That said, as I'm seeing how this is actually used in practice, I'm not so sure of the benefit of compiling things out, so maybe keeping the CMake variable and making it so that it's impossible to turn on at runtime if the flag is set, is good enough. 

https://github.com/llvm/llvm-project/pull/127696