Hi Sergey, > Since this is not a real memory leak (the allocations made in the > init_process method of the ngx_event_core_module module are used up > to this function and already freed on exit automatically, as this > function never returns), I don't think there is something to fix.
Agreed, this is not a "real" leak. But fixing those "false" leaks allows you to run test with sanitizers on the CI, which can protect you and/or external contributors from introducing new bugs. > (Further, the patch misses cycle->files for NGX_USE_FD_EVENT event > methods. Good catch, thanks! > Also, it's probably better to free the memory in the > exit_process method to obey the ownership (this would also fix > missing ngx_free() calls on win32), but this would require moving > code that uses these connections afterwards to catch socket leaks.) Freeing memory in exit_process could result in use-after-free, since cleanups for the cycle->pool might still access those connections. Best regards, Piotr Sikora _______________________________________________ nginx-devel mailing list nginx-devel@nginx.org https://mailman.nginx.org/mailman/listinfo/nginx-devel
