"Eric J. Schwertfeger" <[EMAIL PROTECTED]> writes:
> It seems that the -o option that lsh-authorize passes to sexp_conv has
> been removed, which causes sexp_conv to fail, which causes lsh-authorize
> to fail. For the time being, I'm just removing -o canonical from the
> sexp_conv arguments in lsh-authorize.
Ooops. The -o flag has been renamed to -f (as in "format"). This was
done for consistency reasons, -o is used for specifying an output file
name for those programs for which that makes sense. I'll fix the
lsh-authorize script.
> Also, I've noticed that somewhere between 0.9.5 and 0.9.10, some old bugs
> are back. One was already reported here and fixed by 0.9.10 (can't
> remember which one it was, but it was one that I tripped over trying to
> compile lsh under FreeBSD).
Sorry, I can't do much without the actual compilation errors.
> The one that I didn't notice until yesterday, on the other hand, hasn't
> been fixed, though it worked in 0.9.5. Basically, under FreeBSD, if I
> try to use lsh to a machine that I have not set up public key userauth
> for, I get the following error (which I saw with older versions of lsh,
> but not with 0.9.5):
>
> Unhandled exception of type 0x4001: Public key userauth failed.
> Abort trap (core dumped)
>
> lsh dumps core, but lshd doesn't have a problem.
This (or at least the symptoms) indeed matches an old bug. Can you run
the client with --trace and see what it says? That should tell us
where the exception is raised, and how it is propagated. I've tried to
reproduce it (the --no-publickey flag to the server is handy for
that), but I don't see the bug here.
Regards,
/Niels
