[Lldb-commits] [PATCH] Fix TestThreadSpecificBreakpoint with LLGS

[Tamas Berghammer]

No regression with LLGS LOCAL (I think it uses different code path)


================
Comment at: source/Plugins/Process/Linux/NativeProcessLinux.cpp:2612
@@ +2611,3 @@
+                                                               // Pass this signal number on to the inferior to handle.
+                                                               SetState(eStateRunning, true);
+                                                               return Resume (tid_to_resume, (signo > 0 && !supress_signal) ? signo : LLDB_INVALID_SIGNAL_NUMBER);
----------------
ovyalov wrote:
> Should we update process state only if the following Resume call succeeds?
Fixed

================
Comment at: source/Plugins/Process/Linux/NativeProcessLinux.cpp:2627
@@ +2626,3 @@
+                                                       reinterpret_cast<NativeThreadLinux*> (thread_sp.get ())->SetStepping ();
+                                                       SetState(eStateStepping, true);
+                                                       const auto step_result = SingleStep (tid_to_step,(signo > 0 && !supress_signal) ? signo : LLDB_INVALID_SIGNAL_NUMBER);
----------------
ovyalov wrote:
> Should we update process state only if the following SingleStep call succeeds?
Fixed (if step is unsuccessful then the assert will fire)

================
Comment at: source/Plugins/Process/Linux/NativeProcessLinux.cpp:2727
@@ -2728,1 +2726,3 @@
 
+    Mutex::Locker locker (m_threads_mutex);
+
----------------
ovyalov wrote:
> minor optimization - could you move locker down to "for (auto thread_sp : m_threads)"?
Moved it just before the for loop. It can't be moved inside as the main purpose of this mutex is to make events inside the loop unbreakable by other thread state changing events

http://reviews.llvm.org/D7374

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/