jrandom wrote: > Hi y'all, > > I'm adding support for a high degree transport to i2p and noticed > that kaffe's PlainDatagramSocketImpl synchronizes all the native > methods, in addition to explicit locks on the receive(...) and > send(...) queues. This obviously b0rks anything that tries to > use a single socket to send and receive from (one thread blocking > on .receive, and other threads occationally calling .send), so I've > made the attached patch removing the native object scoped > synchronization, resolving the issue (firing up -vmdebug NATIVENET > showed no problems either). While I personally don't explicitly > need all of the native calls to be unsynchronized (just the send0 > and receive0 calls), I thought it'd make sense to remove it across > the board (unless someone out there knows why it needs to be?) > > (if it matters, yes, sun's jvm works fine with a thread blocking > on receive(...) and other threads calling send(...))
thanks! patch checked in. keep 'em coming! cheers, dalibor topic _______________________________________________ kaffe mailing list [email protected] http://kaffe.org/cgi-bin/mailman/listinfo/kaffe
