Call for testing: OpenSSH 8.0 [AIX Fail]

[Kevin Brott]

On 4/2/19 12:36 PM, Darren Tucker wrote:
> On Tue, 2 Apr 2019 at 18:08, Kevin Brott <kevin.brott at gmail.com> wrote:
>> Mixed results - native builds using AIX's licensed xlC compiler compile and pass all tests.
>> Builds using gcc/make from the Linux Toolbox get further - but die at loginrec.o now:
> I think I just fixed that one too:
> http://anongit.mindrot.org/openssh.git/commit/?id=79a87d32783d6c9db40af8f35e091d9d30365ae7
>
> I'm not even sure why xlc didn't choke on that too (my guess is that
> it would have written incorrect login entries).  If you are able,
> could you please login in and out of a newly built sshd (running as
> root) a couple of times and check that the reported "last logged in
> from" data seems sane.
>
> Thanks!
>

Looks like the latest patches did the trick ... using openssh-SNAP-20190405.tar.gz ...


AIX 7.2
=====
OpSys   : AIX 7200-03-02-1846
Compiler: gcc (GCC) 8.1.0
Make    : GNU Make 4.2.1
OpenSSL : OpenSSL 1.0.2p  14 Aug 2018
Results : all tests passed
=====
OpSys   : AIX 7200-03-02-1846
Compiler: xlC (xlC.rte:13.1.3.1)
Make    : make (bos.adt.base:7.2.3.15)
OpenSSL : OpenSSL 1.0.2p  14 Aug 2018 (openssl.base:1.0.2.1601)
Results : all tests passed
=====

AIX 7.1
=====
OpSys   : AIX 7100-05-03-1846
Compiler: gcc (GCC) 6.3.0
Make    : GNU Make 4.2.1
OpenSSL : OpenSSL 1.0.2p  14 Aug 2018
Results : all tests passed
=====
OpSys   : AIX 7100-05-03-1846
Compiler: xlC (xlC.rte:13.1.3.1)
Make    : make (bos.adt.base:7.1.5.31)
OpenSSL : OpenSSL 1.0.2p  14 Aug 2018 (openssl.base:1.0.2.1601)
Results : all tests passed
=====

also - just for regression testing - using the same snapshot ...

=====
OpSys   : Debian GNU/Linux 9.8 (stretch)
Compiler: gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516
Make    : GNU Make 4.1
OpenSSL : OpenSSL 1.1.0j  20 Nov 2018
Results : all tests passed
=====

Testing with with newly built ./sshd ( -Ddef $PWD/regres/sshd_config ) on AIX 7.1 and 7.2 - fired up a working ssh daemon that I could connect to repeatedly. The test logins show up in 'last' output and in the user attribute time_last_login when using lsuser against the account.  So it looks like it's working as expected.


-- 
# include <stddisclaimer.h>
/* Kevin Brott <Kevin.Brott at GMail.com> */