Peculiar behavior of a 5.0.6 machine:

If one telnets into the system, TZ and HZ are correct, but if one
comes in with ssh, neither HZ nor TZ are in the environment.


--
JP

Jean-Pierre Radley wrote:
>
> Peculiar behavior of a 5.0.6 machine:
>
> If one telnets into the system, TZ and HZ are correct, but if one
> comes in with ssh, neither HZ nor TZ are in the environment.
>
> --
> JP

What version of (I assume) openssh? 3.8p1 was overzealous in clearing
it's environment and would cause the problem you describe. It's fixed
in 3.8.1p1
--
Roger Cornelius racpop@tenzing.org

Roger Cornelius typed (on Thu, Jul 15, 2004 at 10:14:10PM +0000):
| Jean-Pierre Radley wrote:
| >
| > Peculiar behavior of a 5.0.6 machine:
| >
| > If one telnets into the system, TZ and HZ are correct, but if one
| > comes in with ssh, neither HZ nor TZ are in the environment.
|
| What version of (I assume) openssh? 3.8p1 was overzealous in clearing
| it's environment and would cause the problem you describe. It's fixed
| in 3.8.1p1

sshd version OpenSSH_3.8p1, OpenSSL 0.9.7d 17 Mar 2004

--
JP

Jean-Pierre Radley <jpr@jpr.com> wrote in message news:<20040715225334.GD5829@jpradley.jpr.com>...
> Roger Cornelius typed (on Thu, Jul 15, 2004 at 10:14:10PM +0000):
> | Jean-Pierre Radley wrote:
> | >
> | > Peculiar behavior of a 5.0.6 machine:
> | >
> | > If one telnets into the system, TZ and HZ are correct, but if one
> | > comes in with ssh, neither HZ nor TZ are in the environment.
> |
> | What version of (I assume) openssh? 3.8p1 was overzealous in clearing
> | it's environment and would cause the problem you describe. It's fixed
> | in 3.8.1p1
>
> sshd version OpenSSH_3.8p1, OpenSSL 0.9.7d 17 Mar 2004

I see maintenance pack 3 includes 3.8p1. Here's a link describing the problem:

http://marc.theaimsgroup.com/?l=open...6319901648&w=2

3.8.1p1 also has a problem on SCO. Here's a link to that:

http://marc.theaimsgroup.com/?l=open...0516232529&w=2

Roger Cornelius typed (on Mon, Jul 19, 2004 at 08:55:26AM -0700):
| Jean-Pierre Radley <jpr@jpr.com> wrote in message news:<20040715225334.GD5829@jpradley.jpr.com>...
| > Roger Cornelius typed (on Thu, Jul 15, 2004 at 10:14:10PM +0000):
| > | Jean-Pierre Radley wrote:
| > | >
| > | > Peculiar behavior of a 5.0.6 machine:
| > | >
| > | > If one telnets into the system, TZ and HZ are correct, but if one
| > | > comes in with ssh, neither HZ nor TZ are in the environment.
| > |
| > | What version of (I assume) openssh? 3.8p1 was overzealous in clearing
| > | it's environment and would cause the problem you describe. It's fixed
| > | in 3.8.1p1
| >
| > sshd version OpenSSH_3.8p1, OpenSSL 0.9.7d 17 Mar 2004
|
| I see maintenance pack 3 includes 3.8p1. Here's a link describing the problem:
|
| http://marc.theaimsgroup.com/?l=open...6319901648&w=2
|
| 3.8.1p1 also has a problem on SCO. Here's a link to that:
|
| http://marc.theaimsgroup.com/?l=open...0516232529&w=2

I don't find that the hack in that last URL fixes the problem.

--
JP

Jean-Pierre Radley wrote:
>
> Roger Cornelius typed (on Mon, Jul 19, 2004 at 08:55:26AM -0700):
> | Jean-Pierre Radley <jpr@jpr.com> wrote in message news:<20040715225334.GD5829@jpradley.jpr.com>...
> | > Roger Cornelius typed (on Thu, Jul 15, 2004 at 10:14:10PM +0000):
> | > | Jean-Pierre Radley wrote:
> | > | >
> | > | > Peculiar behavior of a 5.0.6 machine:
> | > | >
> | > | > If one telnets into the system, TZ and HZ are correct, but if one
> | > | > comes in with ssh, neither HZ nor TZ are in the environment.
> | > |
> | > | What version of (I assume) openssh? 3.8p1 was overzealous in clearing
> | > | it's environment and would cause the problem you describe. It's fixed
> | > | in 3.8.1p1
> | >
> | > sshd version OpenSSH_3.8p1, OpenSSL 0.9.7d 17 Mar 2004
> |
> | I see maintenance pack 3 includes 3.8p1. Here's a link describing the problem:
> |
> | http://marc.theaimsgroup.com/?l=open...6319901648&w=2
> |
> | 3.8.1p1 also has a problem on SCO. Here's a link to that:
> |
> | http://marc.theaimsgroup.com/?l=open...0516232529&w=2
>
> I don't find that the hack in that last URL fixes the problem.

Here's the simplest patch I could come up with. In wtmpx_write(), you
want to prevent the use of updwtmpx() and use the alternate code
instead. One of the openssh maintainers has checked in a patch for the
problem, but I don't know what it is.

*** loginrec.c.orig 2004-07-20 22:03:24.000000000 -0400
--- loginrec.c 2004-07-20 22:04:15.000000000 -0400
***************
*** 1183,1189 ****
static int
wtmpx_write(struct logininfo *li, struct utmpx *utx)
{
! #ifndef HAVE_UPDWTMPX
struct stat buf;
int fd, ret = 1;

--- 1183,1189 ----
static int
wtmpx_write(struct logininfo *li, struct utmpx *utx)
{
! #ifdef HAVE_UPDWTMPX
struct stat buf;
int fd, ret = 1;
--
Roger Cornelius racpop@tenzing.org

Roger Cornelius typed (on Wed, Jul 21, 2004 at 02:12:52AM +0000):
| Jean-Pierre Radley wrote:
| >
| >
| > I don't find that the hack in that last URL fixes the problem.
|
| Here's the simplest patch I could come up with. In wtmpx_write(), you
| want to prevent the use of updwtmpx() and use the alternate code
| instead. One of the openssh maintainers has checked in a patch for the
| problem, but I don't know what it is.
|

Instead of what you did, Tim Rice defines BROKEN_UPDWTMPX.

This works for me, to a point. I'm working with Kean Johnston and Bela
Lubkin and Boyd Gerber on this issue.

--
JP

Jean-Pierre Radley wrote:
>
> Roger Cornelius typed (on Wed, Jul 21, 2004 at 02:12:52AM +0000):
> | Jean-Pierre Radley wrote:
> | >
> | >
> | > I don't find that the hack in that last URL fixes the problem.
> |
> | Here's the simplest patch I could come up with. In wtmpx_write(), you
> | want to prevent the use of updwtmpx() and use the alternate code
> | instead. One of the openssh maintainers has checked in a patch for the
> | problem, but I don't know what it is.
> |
>
> Instead of what you did, Tim Rice defines BROKEN_UPDWTMPX.

[I'm cc'ing this to Tim Rice]

But is it broken? It seems to work as described in the man page. When
I discovered the problem I did a google search for "updwtmpx man page".
The results showed that some implementations worked like SCO's, others
updated only wtmpx. Here's a couple of examples:

http://www.uwo.ca/cgi-bin/rmanx.pl?t...ent/section=3C
http://www.mcsr.olemiss.edu/cgi-bin/man-cgi?getutx+3

It seems like this is more a problem of different implementations rather
than broken ones.

> This works for me, to a point. I'm working with Kean Johnston and Bela
> Lubkin and Boyd Gerber on this issue.

What do you mean by "to a point"?
--
Roger Cornelius racpop@tenzing.org

Roger Cornelius typed (on Wed, Jul 21, 2004 at 03:14:16PM +0000):
| Jean-Pierre Radley wrote:
| >
| > Roger Cornelius typed (on Wed, Jul 21, 2004 at 02:12:52AM +0000):
| > | Jean-Pierre Radley wrote:
| > | >
| > | > I don't find that the hack in that last URL fixes the problem.
| > |
|
| But is it broken? It seems to work as described in the man page. When
| I discovered the problem I did a google search for "updwtmpx man page".
| The results showed that some implementations worked like SCO's, others
| updated only wtmpx. Here's a couple of examples:
|
| http://www.uwo.ca/cgi-bin/rmanx.pl?t...ent/section=3C
| http://www.mcsr.olemiss.edu/cgi-bin/man-cgi?getutx+3
|
| It seems like this is more a problem of different implementations rather
| than broken ones.
|
| > This works for me, to a point.
|
| What do you mean by "to a point"?

The TZ variable is maintained, but HZ disappears.
There are no duplicate entries in the output of 'last',
but 'brand -t -u' does not show a license being consumed.

--
JP

Jean-Pierre Radley wrote:
>
> Roger Cornelius typed (on Wed, Jul 21, 2004 at 03:14:16PM +0000):
> | Jean-Pierre Radley wrote:
> | >
> | > Roger Cornelius typed (on Wed, Jul 21, 2004 at 02:12:52AM +0000):
> | > | Jean-Pierre Radley wrote:
> | > | >
> | > | > I don't find that the hack in that last URL fixes the problem.
> | > |
> |
> | But is it broken? It seems to work as described in the man page. When
> | I discovered the problem I did a google search for "updwtmpx man page".
> | The results showed that some implementations worked like SCO's, others
> | updated only wtmpx. Here's a couple of examples:
> |
> | http://www.uwo.ca/cgi-bin/rmanx.pl?t...ent/section=3C
> | http://www.mcsr.olemiss.edu/cgi-bin/man-cgi?getutx+3
> |
> | It seems like this is more a problem of different implementations rather
> | than broken ones.
> |
> | > This works for me, to a point.
> |
> | What do you mean by "to a point"?
>
> The TZ variable is maintained, but HZ disappears.
> There are no duplicate entries in the output of 'last',
> but 'brand -t -u' does not show a license being consumed.

I can't explain why your unpatched copy of 3.8.1p1 does not produce
duplicate entries in wtmp/wtmpx. It did here and was confirmed by Tim
Rice on 5.0.4 and 5.0.7. The environment problem is a different issue.
Apparently it wasn't completely fixed between 3.8p1 and 3.8.1p1.
--
Roger Cornelius racpop@tenzing.org