Hi! > When configurin openssl-0.9.6a on a Sun Blade100 machine running > solaris2.8 (using gcc-2.95.3) ./config decides to use gcc > -mcpu=ultrasparc > > IMO this is a bad idea. No released versions of gcc supports > ultrasparc code generation correctly. Yes, the flag works, but there > are known problems with the sparc 64bit code generation. You're mixing up -mcpu=ultrasparc and -m64 options. It's -m64 which is severely broken and officially unsupported, not -mcpu=ultrasparc. gcc is perfectly capable of generating and optimizing 32-bit code for UltraSPARC and that's what happens when it's invoked with -mcpu=ultrasparc flag alone (i.e. without -m64). Well, "perfectly"... in the sense that they seem to care (bug reports are accepted and bugs are getting fixed), it doesn't cost a dime, it's perfect. > Even the gcc-3.0 CVS branch as of now does not support 64bit sparc, > and it's quite unlikely that the released gcc-3.0 will support it... > So using -mcpu=ultrasparc is a disaster waiting to happen. Well, one can as well say that using gcc is a disaster waiting to happen. Yes, they can screw things up for -mcpu=ultrasparc, but the point is that they can as well screw things up for something else. Linux kernel is probably the best example. The only question is what gets discovered and fixed. So if you care periodically grab gcc snapshots, get 'em through testsuite with -mcpu=ultrasparc (optionally through openssl's 'make test':-) and submit results to GCC team. Andy. ______________________________________________________________________ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
