Comments on: ThreadNinja: Finding Rogue POSIX Threads http://tdistler.com/2010/07/08/threadninja-finding-rogue-posix-threads "To err is human, but to really foul things up you need a computer.” Wed, 30 Nov 2011 07:44:27 +0000 hourly 1 http://wordpress.org/?v=3.2.1 By: Tom http://tdistler.com/2010/07/08/threadninja-finding-rogue-posix-threads/comment-page-1#comment-16101 Tom Wed, 13 Apr 2011 16:33:08 +0000 http://tdistler.com/?p=277#comment-16101 Thanks Jeff! I'll take a look. Thanks Jeff! I’ll take a look.

]]> By: Jeff Frontz http://tdistler.com/2010/07/08/threadninja-finding-rogue-posix-threads/comment-page-1#comment-16100 Jeff Frontz Tue, 12 Apr 2011 19:23:50 +0000 http://tdistler.com/?p=277#comment-16100 Not sure which kernel (or thread implementation) this was written for, but under Linux 2.6.23, the pthread_t ID value can be reused. For processes that create/destroy a lot of threads, this makes the first assert() in DataCollection::AddThread() fail pretty quickly. Adding _threadInfoMap.erase(threadId); at the end of DataCollection::CloseThread() seems to help, except when there is truly a rogue thread and then the assert() fails. Not sure which kernel (or thread implementation) this was written for, but under Linux 2.6.23, the pthread_t ID value can be reused. For processes that create/destroy a lot of threads, this makes the first assert() in DataCollection::AddThread() fail pretty quickly.

Adding

_threadInfoMap.erase(threadId);

at the end of DataCollection::CloseThread() seems to help, except when there is truly a rogue thread and then the assert() fails.

]]>