Tung Huynh wrote:
> Subject: X11 forwarding
> Hi,
>
> I have a strange problem with X11 forwarding. I have 2 very similar
> Solaris server. Both running Solaris 7 with SSH 2.1.0 and SSH 1.2.27.
> Both configs have X11forwarding to yes.
>
> I use the same user ID to login. The accounts on both machine are setup
> similarly (PATH, LD_LIBRARY_PATH, etc). On one machine, when I ssh in,
> .Xauthority got created properly and X11 forward is setup correctly, ie
> DISPLAY set.
>
> So obviously I'm missing something or different. Can anyone help me
> track this down further. I've compared everything, environments, ssh
> configs, system patchs, PATH to xauth, etc.
>
> Thanks.
>
> Tung
>-- End of excerpt from Tung Huynh
I've had a similar problem with ssh2 running on several SGI IRIX 6.x
machines. I did a number of tests (with the help of Atro Tossavainen)
and we determined that the problem was that the DISPLAY was not being
set.
In the last few days, I've figured out that out problem has to do with
the compiler. I can not be very specific, but when I recompiled using
gcc, rather than the SGI cc compiler, the problem vanished. The
DISPLAY variable is now properly set. I think the problem is
associated with whether inet_ntoa is broken. WHen using the SGI compiler
inet_ntoa is listed as NOT broken, but the gcc compiler says that it is.
In the latter case (with gcc), ssh2 is configured to use its own internal
version of inet_ntoa, whereas in the former it used the SGI version (I
think). I've forgotten which version of the SGI compulier is installed,
probably 7.2.1.
Using the gcc compiler also automatically enabled SSH1 compatibility.
WHen using the SGI cc compiler, I needed to set this up manually, not so
with gcc. Configuration with the SGI cc compiler produced these remarks:
checking for ssh1... no-ssh1
checking for compatible ssh1.. no new enough ssh1 - COMPATIBILITY DISABLED
However, the gcc compiler configured it correctly
checking for ssh1... /usr/local/bin/ssh1
checking for compatible ssh1... yes
Unfortunately, using gcc did *NOT* fix the utmpx problem I posted
yesterday.
Gary