[Lldb-commits] [lldb] r274763 - [LLGS] Work around an adb bug on Android <=M

Pavel Labath via lldb-commits lldb-commits at lists.llvm.org
Thu Jul 7 08:46:00 PDT 2016

Author: labath
Date: Thu Jul  7 10:46:00 2016
New Revision: 274763

URL: http://llvm.org/viewvc/llvm-project?rev=274763&view=rev
[LLGS] Work around an adb bug on Android <=M

On android M it can happen that we get a ETXTBSY, when we try to launch the inferior. Sleeping
and retrying should help us get more stable results.


Modified: lldb/trunk/source/Plugins/Process/Linux/NativeProcessLinux.cpp
URL: http://llvm.org/viewvc/llvm-project/lldb/trunk/source/Plugins/Process/Linux/NativeProcessLinux.cpp?rev=274763&r1=274762&r2=274763&view=diff
--- lldb/trunk/source/Plugins/Process/Linux/NativeProcessLinux.cpp (original)
+++ lldb/trunk/source/Plugins/Process/Linux/NativeProcessLinux.cpp Thu Jul  7 10:46:00 2016
@@ -561,6 +561,18 @@ NativeProcessLinux::ChildFunc(const Laun
     // Execute.  We should never return...
     execve(args.m_argv[0], const_cast<char *const *>(args.m_argv), const_cast<char *const *>(envp));
+    if (errno == ETXTBSY)
+    {
+        // On android M and earlier we can get this error because the adb deamon can hold a write
+        // handle on the executable even after it has finished uploading it. This state lasts
+        // only a short time and happens only when there are many concurrent adb commands being
+        // issued, such as when running the test suite. (The file remains open when someone does
+        // an "adb shell" command in the fork() child before it has had a chance to exec.) Since
+        // this state should clear up quickly, wait a while and then give it one more go.
+        usleep(10000);
+        execve(args.m_argv[0], const_cast<char *const *>(args.m_argv), const_cast<char *const *>(envp));
+    }
     // ...unless exec fails.  In which case we definitely need to end the child here.

More information about the lldb-commits mailing list