Trusted HP-UX Patch from Re: PLEASE TEST snapshots

[Darren Cole]

> On Tue, 23 Apr 2002, Darren Cole wrote:
> :> + if (s->ttyfd != -1 && c->istate == CHAN_INPUT_OPEN)
> :> + chan_read_failed(c);
> :> +#endif
> :
> :The problem is that you always hang on logout if I don't I have this
code.
> :If  someone has a better fix I would love to have, I haven't found one
yet.
> :If I login, and imediately type exit without any other commands ssh
hangs.
> :This is really annoying, and since most haven't liked this change I put
it
> :only for TRUSTED_HPUX.
>
> does it happen on 10.20 with uselogin=yes?
> also see:
> http://www.snailbook.com/faq/background-jobs.auto.html

I don't know if this is a problem on 10.20 as I only have 10.26 boxen and
few 11 right now.  Maybe tomorrow or next week (customers in the office
today) I will scare one up and try it.   Thanks for the link, I think I have
an idea on where to look in hpux 10.26 and in the opensshd to figure out why
this problem happens more often on 10.26 hpux.

> :> other than that the only other question is why did you add
> :> disable_ptmx_check?
> :
> :Login wont work correctly if ptmx is used.  From what I have found it
> :appears to have to be one of the little weird things in trusted hpux.
>
> does telnetd use STREAMS ptys?

I honestly don't know off the top of my head. I'll look it up and find out.

> :> for now i have this, which is everything but the above against
> :> -current (the uselogin fix applied to HP-UX in general and has
> :> already been applied):
> :
> :Thanks a lot.  If there is anyway to make the above better, or more
> :acceptable for inclusion let me know.
>
> should i apply that then?  your remaining issue is the hang.

Everything but the hang solution seems to have been passed by everyone.

> did you review the auth-passwd.c changes?
>
> :> --- auth-passwd.c 4 Apr 2002 19:02:28 -0000 1.40
> :> +++ auth-passwd.c 23 Apr 2002 21:53:11 -0000
> :> @@ -55,11 +55,11 @@
> :>  #  include <hpsecurity.h>
> :>  #  include <prot.h>
> :>  # endif
> :> -# ifdef HAVE_SCO_PROTECTED_PW
> :> +# ifdef HAVE_SECUREWARE
> :>  #  include <sys/security.h>
> :>  #  include <sys/audit.h>
> :>  #  include <prot.h>
> :> -# endif /* HAVE_SCO_PROTECTED_PW */
> :> +# endif /* HAVE_SECUREWARE */
>
> is audit.h really required here?

I don't think it is needed for Trusted HPUX (compiled with it included and
without. Each time had the same set of warnings). Someone on SCO will have
to comment for them,  I think that is why it was included.

> _______________________________________________
> openssh-unix-dev at mindrot.org mailing list
> http://www.mindrot.org/mailman/listinfo/openssh-unix-dev
>