Re: PATCH: Remove unnecessary zero-initializations

[Michael Snyder <msnyder at redhat dot com>]

Andrew Cagney wrote:
> 
> > Currently, thirteen files which provide a target_ops explicitly initialize
> > members they don't support to NULL.  I plan to delete a number of these
> > methods, and rather than making sure I got all the necessary target files
> > each time I just wanted to delete the unnecessary lines up-front.  All of
> > these are called-once functions initializing a statically or globally
> > declared object; C will guarantee zero-initialization for us.  And several
> > of the functions explicitly called memset anyway.
> >
> > Besides, this way grepping for .to_require_attach\ = will only find targets
> > which define it to something useful.
> >
> > I'll commit this tomorrow unless someone sees a problem with it.
> >
> > Note1: remote-st.c hasn't been compilable in a while; m68*-tandem-* is
> > probably a good candidate for the hitlist.  From a glance it looks like it
> > has been broken since the HP merge added the NULL assignments I'm removing,
> > which is about three years now I think.
> >
> > Note2: The DONT_USE member of struct target_ops can go now.
> 
> The fact that 13 files were doing it should suggest that it was
> intentional.  Might want to wait a bit longer while someone dregs up the
> history.

Not necessarily.  At one time it was required.  Then someone
(either Stan Shebs or John  Metzler, I forget) revamped the way
target vectors are created, so that it is no longer required.
These may either be legacy, or someone doing the initialization
from habit.  In any case, it's definitely not required now.