Re: top level: make more dependencies explicit

["H. J. Lu" <hjl@lucon.org>]

On Sun, Sep 29, 2002 at 03:51:18PM -0400, Daniel Jacobowitz wrote:
> On Sun, Sep 29, 2002 at 01:45:40PM -0400, Andrew Cagney wrote:
> > >On Sun, Sep 29, 2002 at 01:09:22PM -0400, Andrew Cagney wrote:
> > >
> > >>Nathanael,
> > >>
> > >>FYI, I'm about to revert this change:
> > >>
> > >>2002-09-25  Nathanael Nerode  <neroden@gcc.gnu.org>
> > >>
> > >>        * Makefile.tpl: Make subsituted variables more autoconfy.
> > >>        * Makefile.in: Regenerate.
> > >>        * configure: Make seds more autoconfy.
> > >>
> > >>It breaks both the GDB and BINUITLS snapshot processes.  Details to 
> > >>follow, however, suggest a short pause.
> > >
> > >Uh... yuck.
> > >
> > >Are there some details on the GDB and BINUTILS snapshot processes so that
> > >I can fix *them*?  This change is going to happen eventually, even if
> > >it's reverted for now; the Makefile changes will be necessary for
> > >autoonfiscation.
> > >
> > >Wait, let me look at Makefile.in...
> > >
> > >Ewwwww.  The taz rules use Makefile.in *as a Makefile*. 
> > >That's the root of the problem, isn't it?  Incidentally, that's
> > >disgusting. :-)  
> > 
> > Well I think ``SHELL = @config_shell@'' is ugly :-)  Why can't the 
> > configury code edit ``s/^SHELL =.*/SHELL = $.../''.
> 
> Because that's disgusting?  Having something which doesn't look like a
> configury-overridden variable be edited in configure is very fragile.
> 
> > The mechanism is very old (it pre-dates me as GDB release engineer). 
> > Changing it is going to involve updates to many things - snapshot 
> > scripts, release process doco, .... so won't happen overnight.
> 
> GCC has had a separate script to make releases for years and years. 
> It's time GDB and Binutils did the same.

FWIW, I like the current GDB release scheme a lot. It is very useful in
my envrionment.


H.J.