On Sun, Aug 11, 2019 at 9:53 AM Don Poitras <[email protected]> wrote: > > Hello, > I see that proxy_util.c was updated a few days ago. Would it be considered > a faux-pas to contact the developer directly to request a review of this > patch? It's been almost 2 months and I'm worried it will just get lost in the > noise.
Probably not so effective but following up here is OK. I am a bit curious if what's really happening on your system is a failure in an existing mutex that goes unreported, because it seems that the new use of the thread mutex is redundant due to either the global mutex that's held or the mutex used internally by apr_reslist_*(). IIUC the unlocked access to the worker->cp in e.g. ap_proxy_acquire_connection() is a path only used in single-threaded MPMS. If you still have your testbed, it would be interesting to see which lock is necessary and if any of the existing mutexes are silently failing. -- Eric Covener [email protected]
