New observer objfile_mapped; was ...

[Andrew Cagney <cagney@gnu.org>]

>>> Think PIE.
> 
> 
> I am, in fact, thinking about pie.  Oh, you mean PIE...
> 
> 
>>> For our purposes, both new executable loaded and new shlib loaded are 
>>> the same event - there's been an objfile_loaded event.  There's no 
>>> reason to differentiate them.
>>> 
>>> The new inferior event is orthogonal, and far more low level.
> 
> 
> I don't see how "new inferior" is any lower level than "new object".

At the time of inferior_created, nothing, zero zip is known about the 
inferior.  Things like the objfile loader and the shlib loader hang off 
of it.

> In any case, thread-db wants to initialize when these two conditions
> are true:
>   (A) The inferior is running
> and
>   (B) The thread library has been loaded
> 
> So using an inferior created hook makes perfect sense to me...

We're comparing:

-> inferior_created
	-> thread-db
-> solib loaded
	-> objfile mapped
		-> thread-db

with

-> inferior_created
	-> objfile mapped
		-> thread-db
-> shlib loaded
	-> objfile mapped
		-> thread-db

thread-db only needs to know when objfiles have been mapped in.

Andrew