On Wed, Oct 12, 2005 at 09:23:49PM +0100, David Laight wrote:
> If 'a.so has a NEEDED entry for b.so' and 'b.so defines symbol b'
> then dlsym(dlopen("a.so", 0), "b") should load both a.so and b.so and
> return the address of 'b'.

Well, others seem to agree with you and the standard certainly allows this
implementation (there is a very old patch from Charles posted to tech-userlevel
in 2003 that implements it even) - but it's not clear to me that this is what
the standard requires. Of course the point is moot if everyone else implements
the "this module and all it's deppendencies" semantics we probably should not
be different without good reason (and ease of implementation is not a good
one).

Martin