(I'm guessing this belongs on dev.) 1. 0.9.8g (and e and probably earlier) on Solaris 7 with no /dev/*random (and no engine) 'openssl speed rsa' or default which includes rsa fails as not_seeded in RSA_setup_blinding. In speed.c the DSA, ECDSA, and ECDH blocks have logic "if RAND_status() != 1 then RAND_seed(fake) ..." but RSA does not. If I add (copy) there it works.
2. Alternatively, I see no reason not to just do the seed-if-needed once, unconditionally; even if no selected tests/algorithms need PR data, it does no harm. Or even to use app_RAND_loadfile(s?) like other utilities, just to be consistent. (Of course 'speed' doesn't need 'good' randomness, since it isn't securing any real data.) 3. Also (in g only) if aes-*-ige is included in the test set, as it (now) is by default, subsequent tests or shutdown may fail in any of various ways, because that mode uses IV of 2block = 32B but only 1AESblock = 16B is declared and it clobbers memory. 4. This initially exhibited obvious symptoms only on Windows (VC-WIN32 + nasm) but to debug I had to patch the .mak(s) since mk1mf + VC-32.pl no longer puts /Zi /Yd into CFLAG for debug. ______________________________________________________________________ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager [EMAIL PROTECTED]