On Thu, Jul 08, 2004 at 02:46:05PM -0500, Ben Collins-Sussman wrote:
> On Thu, 2004-07-08 at 14:38, Mathias Gumz wrote:
>
> > +----------------------
> > Authentication realm: 4dae73a5-61d1-0310-b756-82608f2b7470
> > Username:
> > +----------------------
>
> Sure enough, if you look in ra_local's open() function, it's using
> session->uuid as the "realmstring" to the auth system. But why the heck
> is it *ever* calling a prompting function? Is this a bug in our auth
> system?
>
> Maybe not. Looking at the username-provider, it looks like we (1) try
> to get --username, (2) try to get a username from the auth cache, (3)
> run getuid().
>
> Somehow, by running in a chroot environment, our auth system is "rolling
> over" into a prompting provider, which is prompting for a username. I
> guess the behavior is explainable, but not the underlying cause of the
> provider rollover. Perhaps getuid() is failing?
ra_local always wanted a username beause it looked for a default or
prompted in the ra open() call. r10194 changes that. It now only looks
up/prompts for the username when it's actually going to use it. Which
is only write operations and operations that require building a
transaction in the db.
--
Ben Reser <ben@reser.org>
http://ben.reser.org
"Conscience is the inner voice which warns us somebody may be looking."
- H.L. Mencken
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Jul 9 08:22:23 2004