[lug] cvs pserver

[D. Stimits]

Tom Tromey wrote:
> 
> >>>>> ">" == D Stimits <stimits at idcomm.com> writes:
> 
> >> This is on an internal network used only by me, firewalled from ppp
> >> interface to the outside world. In any case, unless I am mistaken,
> >> the pserver set simply changes the CVS_RSH method to use ssh
> >> instead of rsh, so on
> 
> Not really.  In pserver mode the cvs server is usually started from
> inetd.  In `ext' mode it is started directly by the user on the server
> via ssh or rsh (or whatever CVS_RSH is set to).  Try it with rsh and
> see if that helps.  It very well might.

I was eventually able to get ssh to work, though pserver is still being
stupid.

> 
> >> I've been poking around there (as well as in a book), and can't
> >> seem to find any reference as to why /root/.cvsignore must be
> >> readable for this to work. I'm tempted to remove and reinstall cvs,
> >> but I have a feeling that whatever config problem I have it would
> >> still be there.
> 
> This is curious.  I imagine what is going on is that since cvs is
> being started as root on the server, it is trying to read
> ~root/.cvsignore.  However there seems to be a twist, since if it
> isn't readable that would mean that the cvs server has probably
> already changed uid to the invoking user.  But then why would it be
> looking at ~root/.cvsignore?  This is all hypothetical though.

I suspect something like this is correct. It runs as root but then gets
twisted during uid change.

> 
> Maybe it is just a bug in cvs.  You'd have to dig through the source,
> or debug it, to find out.

Probably this is a bug.

> 
> Perhaps some misconfiguration in CVSROOT/passwd could cause the
> problem.  I don't know.  I only ever use pserver for anonymous cvs.

It correctly asks for the user password that I tell it to login with,
and that user does not have root access. Just another config snafu.

> 
> Another choice might be to invoke `cvs -f pserver' from inetd.  I
> really have no idea whether that would even work.  cvs might reject
> it.  It might be worth a try.

I tried this briefly, with no luck. Since ssh is now working, there is
no dire need to solve it now.

> 
> Tom

D. Stimits, stimits at idcomm.com