The recent change to the calling sequence for the function 'ws_socket(struct ws_server *ws_srv)' causes problems if the 'ws_server' structure is in the stack of the calling function and the function exits after calling a non-blocking 'ws_socket()'.
Previously, the callers 'ws_events' structure was copied the wsServer's 'ws_server' structure. Now, the caller's pointer to it's 'ws_server' structure is placed in the wsServer's 'ws_accept_params' structure.
When the calling function exits, it's 'ws_server' structure becomes undefined. The solution for the caller is to allocate the 'ws_server' structure in local space.
A note of this should be included in the documentation. And maybe the examples like 'echo' should do this instead of using stack allocation.
My environment:
- wsServer version (commit hash)
- OS version: Debian GNU/Linux 12 (bookworm)
- If Linux:
- kernel version: 6.1.0-15-amd64
- Libc used: Debian GLIBC 2.36-9+deb12u3
- Architecture: x86-64...
The recent change to the calling sequence for the function 'ws_socket(struct ws_server *ws_srv)' causes problems if the 'ws_server' structure is in the stack of the calling function and the function exits after calling a non-blocking 'ws_socket()'.
Previously, the callers 'ws_events' structure was copied the wsServer's 'ws_server' structure. Now, the caller's pointer to it's 'ws_server' structure is placed in the wsServer's 'ws_accept_params' structure.
When the calling function exits, it's 'ws_server' structure becomes undefined. The solution for the caller is to allocate the 'ws_server' structure in local space.
A note of this should be included in the documentation. And maybe the examples like 'echo' should do this instead of using stack allocation.
My environment: