OBJECT_MODE for AIX.

perfectly safe [compiler driver adds it] and in some situation even
perfectly appropriate [mixing -pthread and -lc on FreeBSD can have
lethal effect on apps/openssl]. I'd say we should get rid of more,
but I remove those I can test myself...

Makefile.shared was a bit overcomplicated.

Make the shell variables LDFLAGS and SHAREDFLAGS in Makefile.shared
get the values of $(CFLAGS) or $(LDFLAGS) as appropriate depending on
the value the shell variables LDCMD and SHAREDCMD get.  That leaves
much less chance of confusion, since those pairs of shell variables
always are defined together.

I change from -notall to -none synonym in irix rules to improve backward
compatibility with IRIX 5.x.
PR: 987

"remaining relocations" in assembler modules. The latter seems to
be new behaviour, elder as/ld managed to resolve this relocations
as internal. It's possible to address this problem differently,
but I settle for -Bsymbolic...
PR: 546

"Proper" means "compiles and passes test." Versioning is broken (I think).

COFF and a.out targets [similar to ELF targets]. You might notice some
rudementary support for shared mingw builds under cygwin. It works (it
produces cryptoeay32.dll and ssleay32.dll with everything exported by
name), but it's primarily for testing/debugging purposes, at least for
now...

libraries. Old .sl extension works just fine, but it .so which is default.

I also used this opportunity to clean up some out-of-date targets and
re-group targets by OS.

think that proper Makefile clean-up is required.

this flag RAND_poll ends up in end-less loop calling RAND_add. But
don't ask me why...

both vendor and GNU compilers. ./engine shared build are still busted.
I mean always were...

after more tests...

For reference. Note that both cc and gcc support -Wl flag, but we can't
use -Wl,-[not]all with both drivers, because cc rearranges options
passed through -Wl. We can't use -Wl,-all,libcrypto.a,-notall with cc
either, because it refuses to start with "no input" error.

It seems that svr3 and svr5 differ, after all.

libraries.  What's worse, the directories given in LD_LIBRARY_PATH are
checked first!  Therefore, we need a hack to prepend all the
directories we give with -L to the current value of LD_LIBRARY_PATH,
thereby temporarly forming a hacked value.

Only copy LIBEXTRAS if they are given.

Svr5 doesn't use -z allextract...

-shared rather than -G.

applications, at least on the platforms where it's known how
to do it.

Note: this has only been tested on GNU-based platforms (Linux), and
needs to be tested on all others.  Additionally, it's not yet
supported on the following platforms, for lack of information:

Darwin (MacOS X)
Cygwin
OSF1/Alpha
SVR3
ReliantUNIX

Please help out with testing and the platforms we don't yet know well
enough.

assignment (modulo those I missed) individual statements.

Makefiles... I suppose it wasn't tested very much...

0.9.7-stable.

Add missing semicolons.
Add a comment explaining a bunch of targets without any action lines.

idea why, but that was definitely wrong...

Make sure test doesn't barf because of missing spaces before the
closing ].
Add -lc to the list of libraries we depend upon.  Not always
necessary, but never hurts.

wrong, and my test engine doesn't show it.  The verbosity will be
temporary for about a day.

shared library, set those flags conditionally.