The commit From: "Dr. Stephen Henson" <st...@openssl.org> Date: Fri, 1 Apr 2011 15:46:03 +0000 Subject: [PATCH] Add additional OPENSSL_init() handling add dummy call to (hopefully) ensure OPENSSL_init() is always linked into an application. https://github.com/openssl/openssl/commit/c4acfb1fd049f52fb074b103be01cab5cf5c04f8
seems to have broken CRYPTO_set_mem_functions. After this commit, CRYPTO_set_mem_functions() calls OPENSSL_init(), which calls CRYPTO_malloc(), which sets `allow_customize` to 0, which is then checked by CRYPTO_set_mem_functions(), which then returns without doing anything. See also earlier reports http://openssl.6102.n7.nabble.com/CRYPTO-set-mem-functions-Doesn-t-Work-in-Version-1-0-1b-td46745.html http://bugs.python.org/msg191610 The death test program #include <stdio.h> #include <stdlib.h> #include <openssl/ssl.h> void * my_alloc(size_t n) { abort(); } void my_free(void *p) { abort(); } void * my_realloc(void *p, size_t n) { abort(); } int main(int argc, const char **argv) { const SSL_METHOD *method; SSL_CTX *ctx; CRYPTO_set_mem_functions(my_alloc, my_realloc, my_free); SSL_library_init(); method = SSLv23_client_method(); ctx = SSL_CTX_new(method); printf("Got ctx %p\n", ctx); return 0; } aborts as expected on Ubuntu 10.04 with openssl0.9.8k, but blithely prints a ctx with openssl 1.0.*. The bug is probably not present in openssl-1.1.0-pre5, as the offending code was removed by https://github.com/openssl/openssl/commit/bbd86bf5424a611cb6b77a3a17fc522931c4dcb8 but a fix for 1.0.0 and 1.0.1 would be much appreciated. -- Ticket here: http://rt.openssl.org/Ticket/Display.html?id=4559 Please log in as guest with password guest if prompted -- openssl-dev mailing list To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
