|
From: Julian C. <jj...@ya...> - 2006-04-26 17:00:24
|
Hi, I wondered if anyone could help me with using helgrind. I'm running valgrind 2.2.0 (the last known version that works with helgrind it appears) and am getting many spurious 'possible race condition' messages. They are all in cases where the contested variable has been protected by semaphores. Does helgrind work with pthread_mutex's only? Am I doing something wrong when I run it or interpret the output? I run with valgrind --tool=helgrind <my program> The error messages are: "Possible data race writing variable at 0x....." followed by a stack trace. All very helpful but when I check the code the area is always protected by pthread semaphores. Extra text at the end of the error goes thus: "Previous state: shared RO, no locks". I presume the RO means read operation or somesuch - but the no locks claim is completely out of the window, to get to the stack trace printed it has to go through the semaphore. Any help would be greatly appreciated, Julian Churchill |