On 06.08.2014 22:47, Stefan Fuhrmann wrote: > On Fri, Aug 1, 2014 at 7:37 PM, Ivan Zhakov <i...@visualsvn.com > <mailto:i...@visualsvn.com>> wrote: > > > Stefan, > > It seems you broke the code (again): SVN_MUTEX__WITH_LOCK() will > release THREAD_MUTEX immediately after file lock will be obtained. > That means that access to shared memory will not be synchronized on > posix platforms (!). And this also could cause undefined behavior when > unlock() function will try to unlock THREAD_MUTEX without actually > owning it [1]. > > > Thank you for spotting this! It's one of those "too obvious to see" > snafus. > The lock() is obviously supposed to keep the locks ... Fixed in r1615354.
I suggest you add that to the backport proposal. -- Brane -- Branko Čibej | Director of Subversion WANdisco | Realising the impossibilities of Big Data e. br...@wandisco.com
