> ... the "ucomp runs
> forever with +O4" compiler bug has again reared it's ugly head.
I actually don't think +O4 is an appropriate option for a library. From
what I've heard (I could get it wrong though) +O4 generates some
pseudo-code which in turn is used in some really global optimization
pass during ld phase (code inlining and elimination, some deeper
inter-procedural analysis). As it's not well-documented it wouldn't
surprise me if this pseudo-code appears to be specific to compiler
release or even patch level which might make it absolutely inappropriate
for a library.
> It used
> to be that applying the latest HP compiler patch would fix the problem and
> I've been building okay with +O4 on HPUX since the beginning. But no more.
>
> Can anyone else build with hpux10-cc any more?
>
> BTW I'm trying to build vanilla OpenSSL, without RC4_CHUNK defined. So
> Andy's changes probably have nothing to do with the build problem.
But there are other changes, right? As it sounds like you still can
build 0.9.4 I wonder if you could do following:
- build 0.9.4 with +O4 and send output from 'ls -la apps/openssl' and
'apps/openssl speed';
- build 0.9.4 with +O3 and send same output as above;
- optinally link apps/openssl with +O3 +Oprocelim and send output from
'ls -la apps/openssl';
- see if you can build a snapshot with +O3 [+Oprocelim];
i.e. my suggestion is to replace +O4 with +O3 +Oprocelim or simply +O3
in next release. And of course I appreciate if you could throw RC4_CHUNK
at the snapshot and post output from 'apps/openssl speed rc4'... Of
course provided that +O3 works:-)
Andy.
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
