On Sat, 18 Jan 2003, Andy Polyakov via RT wrote: > > > > > > > Config was adding "386" to the Configure line causing the build > > > > > > to fail on the assembler modules. > > > > > ^^^^ in what way? > > > > UX:as: ERROR: asm/sx86-elf.s:35:invalid register for instruction: %al in xchg > > Would it work if you add b to xchg mnemonic, i.e. make it look "xchgb > %al,%ah"? I'm not saying that we'll stick to 386 code, I just wonder if > xchgb does the trick.
Yes it does. > > An argument in favor of knowing which processor it really has would > > be if at some future date we wanted to automaticly use > > say, crypto/des/asm/des686.pl instead of > > crypto/des/asm/des-586.pl on i686. > > To start with des686.pl is completely out-of-date and is not even > operational, isn't it? Then it's perfectly possible to produce blended > optimized code, and if it will be proven that des686.pl provides > *superior* preformance, I'd rather fix des-586.pl and provide *next* to > superior performance on P6 core. OK. > To summarize. I'm hardcoding i586 to all Caldera/SCO targets. And > according to RT#460 we also should get rid of -lresolv on those > platforms, right? A. Carfefull when you start talking about "all Caldera/SCO targets" SCO has Linux based products and two totally different UNIX based products. Don't interchange OpenServer and UnixWare. (OpenUNIX is UnixWare 7.1.2) But yes, hardcoding the UnixWare/OpenUNIX targets to i586 would be fine. And yes, get rid of -lresolv on the sco5 (OpenServer 5) targets. -- Tim Rice Multitalents (707) 887-1469 [EMAIL PROTECTED] ______________________________________________________________________ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]