Re: ARM Simulator build broken

[Hans-Peter Nilsson <hans-peter.nilsson@axis.com>]

> From: Richard Earnshaw <rearnsha@gcc.gnu.org>
> Date: Wed, 17 Nov 2004 16:06:19 +0000

> On Wed, 2004-11-17 at 15:03, Richard Earnshaw wrote:
> > On Tue, 2004-11-16 at 15:35, Hans-Peter Nilsson wrote:
> > > BTW, you shouldn't have edited out sim-arange.c, a gcc -MM says
> > > it's actually included from sim-basics.h (and from looking at
> > > the code).  For the same reason sim-assert.h should be in
> > > sim_basics_headers.  Maybe all files should use the
> > > $(sim-assert_h) et al macros.
> > 
> > The list of headers I used was the list directly in sim-basics.h.  I
> > only moved existing includes from sim_main_headers and then added
> > further ones at the first level that were obviously missing.  So if it's
> > wrong now it was wrong before.

Yes, it *was* wrong before. :-)  Many more were missing.
Less wrong with your patch.  Anyway, you weren't that far off: I
got the list from "gcc -MM" with the rest of the command-line
from an arm-sim compilation of sim-load.c, but with the file a
single '#include "sim-basics.h"'.  Notable thing missing:
sim-basics.h contains the specific lines
#include "gdb/callback.h"
#include "gdb/remote-sim.h"
Supposedly they were added at some point but dependencies were
for some reason or another not updated.  It kind of makes the
occurrences of $(remote_sim_h) and $(callback_h) in .o rules
redundant...

> > The best way to do all this is to define a series of macros
> > sim-basics_h, sim-assert-h etc, and for each of these to list the macros
> > that form the direct dependencies.

Right, some are already there, but apparently unused in
sim_main_headers for unknown reasons, while the similar hw-*_h
macros are used in hw_main_h.

> > But that's a fair bit of work... :-(  ... and it needs active
> > maintenance to keep it up to date.
> > 
> > R.
> 
> Attached is a revised version of the patch that implements this for
> sim-*.h.

Nice.

>  Really further work should be done so that the C files depend
> on the correct headers rather than all depending on sim_main_headers.

Along the lines of the dep-am rule in bfd/Makefile.am, perhaps.

brgds, H-P