| Add function to register at fork handlers to be called in child after each fork. |
|
More...
|
almost 2 years ago
|
| Organize internal_fork() parameters into a struct; |
|
More...
|
almost 2 years ago
|
| Add REUSE_PORT socket option |
|
More...
|
almost 2 years ago
|
| Improve child startup error handling by creating a "startup" mutex, locking it before fork in parent and letting child to unlock it once initialization is complete (or fails). The parent then waits on this mutex to read out error code and proceed accordingly. |
|
More...
|
almost 2 years ago
|
| Pass error code to the caller if ProfilerStart() fails. |
|
More...
|
almost 2 years ago
|
| Implement lock_destroy() for pthread-based locking. |
|
More...
|
almost 2 years ago
|
| Merge pull request #3037 from okhowang/ping_interval |
|
More...
|
almost 2 years ago
|
| Merge pull request #3038 from okhowang/tracer-leak |
|
More...
|
almost 2 years ago
|
| Merge pull request #3012 from purecloudlabs/bug/rport_param_corrupts_via_header |
|
More...
|
almost 2 years ago
|
| Initialize hostname to make compiler happy. |
|
More...
|
almost 2 years ago
|
| Reduce number of sockets created by closing off one half of the socketpair in child after forking. This results in 30% decreate of the total number of sockets allocated, from 8k to some 5.8k on my test config with 10 workers and 10 sockets. |
|
More...
|
almost 2 years ago
|
| [dialog] fix bad examples with invalid profile name |
|
More...
|
almost 2 years ago
|
| Add support for "cunix" variant of the local unix domain communication socket, which is similar to "unix" but does not require re-connecting the socket for every request thus having lower overhead compared to that one. It is supported by rtpproxy for number of years. |
|
More...
|
almost 2 years ago
|
| nat_traversal: remove min interval limit |
|
More...
|
almost 2 years ago
|
| tracer: Don't modify shm msg in trace_onreply_out |
|
More...
|
almost 2 years ago
|
| Set POLLRDHUP on Linux when polling rtpproxy's socket to drain any unclaimed data. This seems to be necessary on TCP sockets otherwise poll() appears to hang indefinitely if rtpproxy has been terminated despite timeout being set to 0 (i.e. don't block). |
|
More...
|
almost 2 years ago
|
| event_route: Fix off-by-one error in some $param warning log |
|
More...
|
almost 2 years ago
|
| Some missing extern var declaration |
|
More...
|
almost 2 years ago
|
| Merge branch 'jes-master' |
|
More...
|
almost 2 years ago
|
| Just a small var rename |
|
More...
|
almost 2 years ago
|
| Merge branch 'master' of github.com:jes/opensips into jes-master |
|
More...
|
almost 2 years ago
|
| PR Comments |
|
More...
|
almost 2 years ago
|
| Rebuild documentation |
|
More...
|
almost 2 years ago
|
| Cap TCP workers at tcp_workers_max_no |
|
More...
|
almost 2 years ago
|
| [async] fix setting restoring type route after running report route for launch() |
|
More...
|
almost 2 years ago
|
| If we did not request qop to start with when challenging then no additional work needs to be done after receiving authentication with no qop. |
|
More...
|
almost 2 years ago
|
| Merge pull request #3020 from nexbridge/temp |
|
More...
|
almost 2 years ago
|
| Update drouting.c |
|
More...
|
almost 2 years ago
|
| event_route: fix crash when raising with no params |
|
More...
|
almost 2 years ago
|
| Expect that protos[proto].name can be NULL. |
|
More...
|
almost 2 years ago
|
