Regression tests (again)

[Darren Tucker]

There's a couple of other changes I've been using locally which aren't in
the patch (eg grabs HAVE_GETPEEREUID from config.h to decide whether to
run the peereuid tests.

Ben Lindstrom wrote:
> I'm worried about the removals of $OBJ in different places, but the
> keeping of them in others.
> 
> You changed the main ssh_config file with this patch.. so watch that.

Yeah, I saw that.  It wasn't intended.

> FYI I addedd a bunch of stuff to sftp*.sh testing recently that will
> need to be pulled in.

OK.
 
> Can you review your patch for authorized_key usage?  It seems to be
> scattered on how the changes happen.

The authorized_keys will be a problem.  (eg: an out-of-the-box AIX build
will have / owned by bin.bin, and if you're not building in $HOME then
StrictModes will check it and the tests will fail).  I'm thinking of just
setting StrictModes=no and abandoning the other related changes.

> Also any reason why you stripped out $OBJ in some places and not others?

Not that I can think of, I'll have to check.

> You pulled out agent-ptrace from the test list, and modified the test.
> You plan on re-adding it?

If it works, yes.  I forget what the issue was.

> I think other than that.. I think we should get the portable regression
> tests up to par.

Agreed.  Ideally it should be possible to routinely run "make test" as
part of an install.

FWIW, Markus has said he'll take a look at the patch too.

-- 
Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
    Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.