Slackware-12 still fails three tests

[Iain Morgan]

On Tue, Aug 17, 2010 at 12:06:38 -0500, Andy Tsouladze wrote:
> On Tue, 17 Aug 2010, Iain Morgan wrote:
> 
> > On Mon, Aug 16, 2010 at 23:18:50 -0500, Andy Tsouladze wrote:
> >> I have re-run builds of openssh-SNAP-20100817.tar.gz on slackware-12.0 and
> >> slackware-13.0.  On 13.0, there both build and all tests run fine.
> >>
> >>
> >> On slackware-12.0, openssh is built successfully, but three tests fail.  I
> >> had to test them individually (by removing successful ones from
> >> .../regress/Makefile).  I also put `set -x' as the first line of every
> >> problematic test file (banner.sh, stderr-data.sh, and stderr-after-eof.sh)
> >> Any chance this can be addressed before the release?
> >>
> >
> > The output in your previous email, particularly the numberous complaints
> > from stty, makes me suspect that this is a user environment issue. Have
> > you tried, for example, renaming your ~/.bashrc prior to running "make
> > tests?"
> 
> I use tcsh, so .bashrc does not exist.  I have removed a few stty settings 
> from my .cshrc, and - you were correct - all tests passed.  So that's a 
> good thing.  As a side note, maybe tests should be made more independent 
> of a user environment.
> 
> Thanks,
> 
> Andy
> 
> Dr Andy Tsouladze
> Sr Unix/Storage SysAdmin
> _______________________________________________
> openssh-unix-dev mailing list
> openssh-unix-dev at mindrot.org
> https://lists.mindrot.org/mailman/listinfo/openssh-unix-dev

That's easier said than done. The issue is that sshd invokes remote
commands via the user's shell. Thus any problems with a .cshrc or
(possibly) a .bashrc can break tthe tests in unexpected ways.

I suppose a test for broken dot-files could be done at the beginning of
the regression tests. For example, if $SHELL -c true yields any output
to stderr or stdout, an appropriate error could be reported.

-- 
Iain Morgan