On Fri, Mar 08, 2024 at 08:53:45AM +1100, Darren Tucker wrote:
> On Fri, 8 Mar 2024 at 06:59, Corinna Vinschen <vinschen@xxxxxxxxxx> wrote:
> [...]
> > After a lot of tinkering I found that the following change in
> > dynamic-forward.sh suddenly made the test succeed.
> >
> > In check_socks():
> >
> > ${REAL_SSH} -q -F $OBJ/ssh_config \
> > - -o "ProxyCommand ${proxycmd}${s} $h $PORT 2>/dev/null" \
> > + -o "ProxyCommand ${proxycmd}${s} $h $PORT" \
> > somehost cat ${DATA} > ${COPY}
> >
> > It occured to me that my login shell is tcsh, not bash. So I changed
> > my login shell to bash and, lo and behold, dynamic-forward.sh succeeded
> > even with the stderr redirection.
>
> Nice find! Wow, tsch, I don't think I've used that in this millenium!
Yes, nice find.
> > Having said that, can this test be changed to be independent of the
> > user's long shell?
>
> Yes we should be able to change to something that invokes
> ${TEST_SHELL} -c "[whatever]", although it might take a couple of
> attempts to get the quoting right. Lemme have a try...
Wouldn't it be simpler to just use '#!/bin/sh' as the shebang line for
the dynamic-forward.sh script?
It seems it is a long-standing shortcoming of the C shell that there is
no simple way to redirect only stderr. The goal of that regression test
isn't specifically to test a multitude of shell implementations, right?
Kind regards,
Job
_______________________________________________
openssh-unix-dev mailing list
openssh-unix-dev@xxxxxxxxxxx
https://lists.mindrot.org/mailman/listinfo/openssh-unix-dev
