Call for testing.

[Ben Lindstrom]

On Tue, 26 Feb 2002, Gert Doering wrote:

> Hi,
>
> On Mon, Feb 25, 2002 at 03:48:53PM -0600, Ben Lindstrom wrote:
> > I commited the change so tomorrow please grab the snapshot and try it.
>
> Did so just now.
>
> - cvs update (-d)
> - make clean ; autoreconf ; ./configure (with skey)
> - make
>   -> didn't compile with installed openssl (0.9.5a), bombs out in cipher.c
>      - interesting enough, last week's snapshot (before the reorganization)
>      still built with 0.9.5a.
>
I'm well aware of the 0.9.5a issue.  We did a massive change to cipher.c.
I have not gotten back to seeing if we can still use anything pre 0.9.6,
but my gut feeling (since we are using a struct field that does not
exist in older versions) is that 0.9.6 is our new standard.

> - installed openssl-0.9.6c, ./configure, make -> went fine
>
> - tested: everything related to aix_usrinfo works fine, basic login/logout
>   tests with password and .shosts, protocol -1 and -2, also work.
>
> So...
>
>  + good job!
>
>  + can we add a check to configure for "this version of openssl is too
>    old"? This would be a lot easier for people to understand than
>    "configure works fine, but compiling cipher.c bombs with undefined
>    EVP_CIPHER_CTX_set_key_length".
>

Maybe.. EVP_CIPHER_CTX_set_key_length() is a macro.  I would just
check that the release of OpenSSL is equal or greater than 0.9.6
if we decide to go that way.

- Ben