On Mon, Oct 26, 2009 at 11:56 PM, Eric Chamberlain <eric at rf.com> wrote: > Hello, > I'm using a worker thread to startup pjsip, > otherwise?pjsua_media_transports_create?blocks the UI for a few seconds, > while ICE is initialized. Yeah. > The thread calls: pjsip_create(),?pjsua_pool_create(), etc. > When the thread is done, it ends. ?The main thread is then registered with > pjsip. That's probably the part where pjlib doesn't like it. Main thread (the thread that calls pj_init()) is treated differently, i.e. it's thread descriptor is kept alive in a global variable and is never destroyed. If the thread itself is terminated, there may be something that goes out of sync. > When a call is made (about one out of three calls) and?pjsua_verify_sip_url > is called (from the main thread), the application crashes in > cpool_create_pool with an EXEC_BAD_ACCESS. > Am I doing something wrong? ?Do I have to initialize pjsip from the main > thread? ?If so, can?pjsua_media_transports_create be called from a worker > thread? You should be able to, and that's probably the best option, i.e. to initialize pjsip from the main thread, and call pjsua_media_transports() from the worker thread, terminating the worker thread if necessary after the function call completes. > Can pjsip be initialized in a worker thread that ends after the > initialization is done? I'm not sure, I never tried this. Probably it can't, based on above. > Do I have to keep the worker thread around Depends. If this is the main thread then probably yes you'd have to. > and call pjsua_verify_sip_url from the worker thread? Definitely no. Once threads are setup you can call the API from any threads. Cheers Benny
