[RFA] Add observer notification in normal_stop

[Andrew Cagney]

> [sigh, once more, with the patch this time...]

Welcome to my world.

> Hello,
> 
> I am currently writing a regression test for observer.c, which will live
> in testsuite/gdb.gdb. It's using a mechanism similar to xfullpath.exp.
> But in order for the test to function, observer.o needs to be linked in.
> This wasn't the case up to now, because there is currently no code using
> it.
> 
> This patch is adding a call to observer_notify_normal_stop() inside
> normal_stop. This notification will be useful later, when the latest
> version of ada-task.c (from ACT's repository, accessible from
> libre.act-europe.fr BTW) is contributed. In the meantime, the overhead
> is really minimal and it allows us to test observer.c by debugging gdb.
> 
> 2003-03-17  J. Brobecker  <brobecker@gnat.com>
> 
>         * infrun.c (observer.h): Add #include.
>         (normal_stop): Add call to observer_notify_normal_stop.
>         * Makefile.in (infrun.o): Add dependency on observer.h.
> 
> Ok to apply?

I think down the end, next to the annotate_stop() call, would be better.

I'm assuming that the secret agenda is to replace all annotate_xxx() and 
xxx_hook() calls with a equivalent observer_notify_xxx() call.  Hence, 
adding an observer_xxx() call where ever either of the former appears 
should be pretty obvious and largely mechanical.

An orthogonal task is rationalizing some of those events and/or their 
call site.  For instance, merging registers changed and memory changed 
into a single observer_notify_target_changed().

Andrew