PATCH: eliminate warnings from c-opts.c

PATCH: eliminate warnings from c-opts.c

[Ben Elliston]

c-opts.c invokes C_COMMON_OVERRIDE_OPTIONS if it is defined.  In most
ports that use this target macro, it is defined to be a do/while (0)
expression.  The spu-elf target defines this macro to be the name of a
function (prototyped in spu-protos.h).  

Without this patch, the compiler warns about the lack of a prototype for
the function where it is invoked from c-opts.c.

While I was at it, I found the internals manual contained a FIXME
comment about the tm_p.h header that doesn't seem right to me:

  FIXME: why is such a separate header necessary?

Unless I'm mistaken, it is done this way so that GCC sources can include
an independently named "tm_p.h" without needing to know the name of the
<machine>-protos.h header.  However, this seems too obvious, so I
suspect I'm wrong.  :-)

Ben


2007-07-22  Ben Elliston  <bje@au.ibm.com>

	* c-opts.c: Include tm_p.h.
	* doc/configfiles.texi (Configuration Files): Remove FIXME.

Index: c-opts.c
===================================================================
--- c-opts.c	(revision 126810)
+++ c-opts.c	(working copy)
@@ -40,6 +40,7 @@ Software Foundation, 51 Franklin Street,
 #include "options.h"
 #include "mkdeps.h"
 #include "target.h"
+#include "tm_p.h"
 
 #ifndef DOLLARS_IN_IDENTIFIERS
 # define DOLLARS_IN_IDENTIFIERS true
Index: doc/configfiles.texi
===================================================================
--- doc/configfiles.texi	(revision 126810)
+++ doc/configfiles.texi	(working copy)
@@ -67,5 +67,4 @@ machine.
 @item
 @file{tm_p.h}, which includes the header @file{@var{machine}-protos.h}
 that contains prototypes for functions in the target @file{.c} file.
-FIXME: why is such a separate header necessary?
 @end itemize

Re: PATCH: eliminate warnings from c-opts.c

[Daniel Jacobowitz]

On Sun, Jul 22, 2007 at 07:27:18AM +1000, Ben Elliston wrote:
> While I was at it, I found the internals manual contained a FIXME
> comment about the tm_p.h header that doesn't seem right to me:
> 
>   FIXME: why is such a separate header necessary?
> 
> Unless I'm mistaken, it is done this way so that GCC sources can include
> an independently named "tm_p.h" without needing to know the name of the
> <machine>-protos.h header.  However, this seems too obvious, so I
> suspect I'm wrong.  :-)

I think the question is relative to tm.h (and it has something to do
with which files include one and not the other).

-- 
Daniel Jacobowitz
CodeSourcery

Re: PATCH: eliminate warnings from c-opts.c

[Ben Elliston]

On Wed, 2007-07-25 at 07:17 -0400, Daniel Jacobowitz wrote:

> > Unless I'm mistaken, it is done this way so that GCC sources can include
> > an independently named "tm_p.h" without needing to know the name of the
> > <machine>-protos.h header.  However, this seems too obvious, so I
> > suspect I'm wrong.  :-)
> 
> I think the question is relative to tm.h (and it has something to do
> with which files include one and not the other).

OK, so let's ignore the doc change I proposed.  Is the patch to
eliminate the warning going about it the right way?

Cheers, Ben

Re: PATCH: eliminate warnings from c-opts.c

[Kaveh R. GHAZI]

On Wed, 25 Jul 2007, Daniel Jacobowitz wrote:

> On Sun, Jul 22, 2007 at 07:27:18AM +1000, Ben Elliston wrote:
> > While I was at it, I found the internals manual contained a FIXME
> > comment about the tm_p.h header that doesn't seem right to me:
> >
> >   FIXME: why is such a separate header necessary?
> >
> > Unless I'm mistaken, it is done this way so that GCC sources can include
> > an independently named "tm_p.h" without needing to know the name of the
> > <machine>-protos.h header.  However, this seems too obvious, so I
> > suspect I'm wrong.  :-)
>
> I think the question is relative to tm.h (and it has something to do
> with which files include one and not the other).

I recall something similar (perhaps having to do with cross-compilers?)

Anyway, here is the patch which created the tm_p.h mechanism, but I didn't
see the thread where the reationale was discussed.  Perhaps someone can
dig it up, and for posterity we can insert a comment.

http://gcc.gnu.org/ml/gcc-patches/1999-09n/msg00862.html

		--Kaveh
--
Kaveh R. Ghazi			ghazi@caip.rutgers.edu