Re: Test failures

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Tue, Mar 06, 2012 at 10:15:12PM -0800, Hisashi T Fujinaka wrote:
> On Tue, 6 Mar 2012, Thor Lancelot Simon wrote:
> 
> >On Tue, Mar 06, 2012 at 07:06:41PM +0100, Martin Husemann wrote:
> >>On Tue, Mar 06, 2012 at 08:01:30PM +0200, Jukka Ruohonen wrote:
> >>>(1) Failures in "fs/psshfs/t_psshfs" and "lib/librumphijack/t_tcpip".
> >>>    The standard error output is:
> >>>
> >>>       rsa_generate_private_key: key generation failed.
> >>>
> >>>    Probably caused by the RNG changes?
> >>
> >>Not that you can ssh into a system running -current (i.e. sshd works),
> >>but you can not ssh out there:
> >
> >Huh.  What happens if you run "openssl rand -out /tmp/foo 100"?  That's
> >how I tested that I hadn't broken RNG output, and it does still work on
> >my systems.  I get 100 bytes of output and no error.
> 
> This seems to work for me, but ssh-ing out gives me the "dh_gen_key"
> error as well.

Yeah, I see it too.  The BN_rand() call seems to fail at a call to
RAND_bytes() but I do not see why that would fail when simple tests
like "openssl rand" and more complex ones like "openssl gendh" and
"openssl gendsa" seem to work fine.  Unfortunately, I have run out of
hours to debug tonight, so it will have to wait, unless someone sees
what I'm missing before I can rebuild openssl with symbols on and
sit for a while with gdb -- likely very late tomorrow.