Call for testing: OpenSSH-6.2

[Kevin Brott]

On Sun, Mar 3, 2013 at 3:45 PM, Damien Miller <djm at mindrot.org> wrote:

> On Fri, 1 Mar 2013, Kevin Brott wrote:
>
> > # *5 make tests fails immediately - xlc_r
> > # *6 make tests fails immediately - gcc
>
> I suspect that the patch I just sent will avoid these.
>
>
*​Confirmed - solid fix for AIX builds so far.​*


> > # *7 Make tests fails here:  hangs forever
> >      run test forwarding.sh ...
> >      Warning: Could not request remote forwarding.
> >      ssh_exchange_identification: Connection closed by remote host
> >      cmp: EOF on /usr/src/UTILS/SSH/openssh/regress/ls.copy
> >      corrupted copy of /bin/ls
> >      Warning: remote port forwarding failed for listen port 3350
>
> This might be a spurious failure caused by the regress script picking a
> port that is already in use on the host running the tests. Could you
> check whether anything is using ports 3300-3399? You can adjust the
> leading digits of the ports that the test choses using the 'base='
> variable at the start of regress/forwarding.sh
>
>
​*Confirmed - xrdp was using that port - shutting it off during the
build/test process gets 'all tests passed'. Disregard.  Although a test for
xrdp active might not hurt.​*
*
*
*​xrdp-sesm  3558       root    6u  IPv4  45542      0t0  TCP
127.0.0.1:3350(LISTEN)
xrdp-sesm  3558       root    7u  IPv4  97938      0t0  TCP 127.0.0.1:3350->
127.0.0.1:52157 (ESTABLISHED)
sshd      29009       root   32u  IPv4  98454      0t0  TCP 127.0.0.1:52157
->127.0.0.1:3350 (ESTABLISHED)
*



> > # *8 make test failes here:
> >      run test integrity.sh ...
> >      test integrity: hmac-sha1 @2900 Invalid modification spec
> "xor:2900:1"
> > ssh_exchange_identification: Connection closed by remote host.
> >      unexpected error mac hmac-sha1 at 2900
>
> This one is strange - the modpipe tool is not parsing its commandline
> properly. I suspect that HP/UX's sscanf(3) lacks the "hh" conversion.
>
> Could you try this patch?
>
>
> Index: regress/modpipe.c
> ===================================================================
> RCS file: /var/cvs/openssh/regress/modpipe.c,v
> retrieving revision 1.5
> diff -u -p -r1.5 modpipe.c
> --- regress/modpipe.c   20 Feb 2013 10:16:09 -0000      1.5
> +++ regress/modpipe.c   3 Mar 2013 23:44:12 -0000
> @@ -74,20 +76,29 @@ static void
>  parse_modification(const char *s, struct modification *m)
>  {
>         char what[16+1];
> -       int n;
> +       int n, m1, m2;
>
>         bzero(m, sizeof(*m));
> -       if ((n = sscanf(s, "%16[^:]%*[:]%lli%*[:]%hhi%*[:]%hhi",
> -           what, &m->offset, &m->m1, &m->m2)) < 3)
> +       if ((n = sscanf(s, "%16[^:]%*[:]%lli%*[:]%i%*[:]%i",
> +           what, &m->offset, &m1, &m2)) < 3)
>                 errx(1, "Invalid modification spec \"%s\"", s);
>         if (strcasecmp(what, "xor") == 0) {
> -               m->what = MOD_XOR;
>                 if (n > 3)
>                         errx(1, "Invalid modification spec \"%s\"", s);
> +               if (m1 < 0 || m1 > 0xff)
> +                       errx(1, "Invalid XOR modification value");
> +               m->what = MOD_XOR;
> +               m->m1 = m1;
>         } else if (strcasecmp(what, "andor") == 0) {
> -               m->what = MOD_AND_OR;
>                 if (n != 4)
>                         errx(1, "Invalid modification spec \"%s\"", s);
> +               if (m1 < 0 || m1 > 0xff)
> +                       errx(1, "Invalid AND modification value");
> +               if (m2 < 0 || m2 > 0xff)
> +                       errx(1, "Invalid OR modification value");
> +               m->what = MOD_AND_OR;
> +               m->m1 = m1;
> +               m->m2 = m2;
>         } else
>                 errx(1, "Invalid modification type \"%s\"", what);
>  }
>

*
*
*​This seems to have fixed the issue on hp-ux 11.23 - running regression
tests against all builds using both modepipe.c patches applied to
​openssh-SNAP-20130305.tar.gz as openssh-SNAP-20130305-p2.tar.gz.  Will
report status when done.*


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