On Wed, 2005-02-23 at 09:26 +0100, Paul J Stevens wrote:
>
> Geo Carncross wrote:
>
> > This is not happening; the config_read demands I install dbmail.conf
> > in /etc -- I simply don't know you that well :)
> >
> > I modified the test cases to expect a dbmail-local.conf (attached) - and
> > it's mostly successful.
>
> Ok. I've changed the test-suites to use /etc/dbmail/dbmail-test.conf instead.
> The actual location is now defined in check_dbmail.h and used by all the test
> suites.
Can we use an environment variable to override?
Consider using:
getenv("DBMAIL_CONF") ? getenv("DBMAIL_CONF") : "/etc/dbmail/dbmail-
test.conf"
or something to that effect...
> > check_dbmail_deliver.c:399:F:Auth:test_auth_validate: auth_validate
> > positive failure
> >
> > I _suspect_ this is because I didn't preload my database with a
> > testuser1 account. After doing this, the test succeeds. I think this is
> > a bug with the test-case.
>
> Well, actually it's more like a 'bug' in user.c. Currently, the only way to
> enter a new user record *with* encryption is by using the dbmail-users cli.
> I'm
> working to move mkpassword -- where the actual hashing of the password takes
> place -- to db_user_create (was auth_adduser). That way we can insert new
> test-users in the database with encrypted passwords by simply calling
> db_user_create(), a much cleaner and more consistent approach imo.
That's fine.
> > All other tests succeed!
> >
> > Please correct me if I'm wrong on this.
>
> Actually, apart from myself you're probably the first one to actually use
> those
> test-cases. As I'm sure you've noticed they are far from complete. The only
> api
> that's fully covered atm is auth.h.
No doubt.
> > [[ FYI: I found no mention of "check package" in the mailing list
> > archive. After much digging, I located http://check.sf.net/
>
> Well, that location *is* documented in the test-suites.
Yeah, but figuring out where those were wasn't obvious...
> > The current version of "check" doesn't use "suite_free()" anymore, so I
> > deleted those lines from the check scripts. srunner_free() appears to
> > free the suites now. ]]
>
> I'm using the version that comes with debian/sarge which is way behind
> upstream.
> So I'll stick with suite_free for now. Someone is working on updating the
> check
> package already though.
It's versioned in check.h -- I'm not certain when the change occurred
(someplace between your version and mine). How about:
#if CHECK_MAJOR_VERSION > 0 || CHECK_MINOR_VERSION >= 9
#define suite_free() /* nothing! */
#endif
--
Internet Connection High Quality Web Hosting
http://www.internetconnection.net/
