living in a multi-mech world

[Nicolas Williams]

On Sun, Apr 29, 2007 at 11:00:54PM -0600, Philip Guenther wrote:
> On 4/29/07, Tom Yu <tlyu at mit.edu> wrote:
> > Do we have mechanism plug-in modules export a function which returns a
> > struct full of function pointers?  (This is similar to what our
> > Sun-derived mechglue implementation currently does.)
> 
> I've found the latter to be simpler to implement.  It also lets the
> function perform mech initialization without having to worry about the
> portability or details of dload() invoked .init functions.

You need to add versioning so that adding an extension to the mechglue
and one mechanism doesn't require patching all other mechanisms to
inlude a NULL pointer.  That _was_ a problem with the Solaris mechglue.

But this is useless code -- it does something that the RTLD is supposed
to be good at.