On Thu, 2006-02-16 at 19:45 +0100, Paul J Stevens wrote:
> >>Are you sure? Sometimes it's cool to make assignment to nonexistent
> >>structure members; good to test that your compiler noticed, you know?
>
> I wasn't sure it was allowed. Or rather, I accept it is allowed under some
> circumstances (--export-dynamic??), but /understand/ it I do not.
Joking, it was totally a bug. {db|auth|sort}module.c aren't compiled at
all for static linking. I've been contemplating adding a little bit more
code and function prefixes that would allow the same dbmail binary to be
statically linked against mysql and pgsql at the same time, though. I'm
not sure this is worth the effort.
-W,lexport-dynamic allows a loaded module to access functions in the
host program. All of our modules use dm_malloc, for example, and so they
need access to the host program's symbols for that.
> Anyway, great to see you've started using unit-testing :-). I did some changes
> in check_dbmail_dsn.c today, to make sure it doesn't collide with
> check_dbmail_user.c (testfailuser).
The check framework is really good stuff! The old dsnuser_resolve was
already too big, and I wanted to get subaddress and username catchalls
working without much pain. Refactoring it was fun, actually.
Aaron
