[current tinderbox] failure on ...all...

[David O'Brien]

On Fri, Jun 10, 2005 at 10:47:06AM +0300, Ruslan Ermilov wrote:
> On Fri, Jun 10, 2005 at 09:32:14AM +0200, Dag-Erling Sm?rgrav wrote:
> > Ruslan Ermilov <ru at FreeBSD.org> writes:
> > > And if you feel that -fno-strict-aliasing is evil, why not dike it
> > > out from sys.mk?
> > 
> > 'ncvs annotate /usr/src/share/mk/sys/mk | grep aliasing' and you'll
> > realize that any attempt to touch it would result in a huge flamewar.
> > 
> I don't see a flamewar, only the mention that it breaks some notable
> ports.  If it's not suitable for ports, then we should invent a mean
> to compile only src/ *without* -fno-strict-aliasing.

I tried.  But Kris refused to consider the following for committing.
The problem is something like 3 ports will not build with
"-fno-strict-aliasing".  Those are the gcc28, gnat[*] ports.

[*] I really don't understand why we have a GCC 2.8 based Ada compiler
when Ada has been a native part of GCC since version 3.1...

-- 
-- David  (obrien at FreeBSD.org)

Index: bsd.port.mk
===================================================================
RCS file: /home/ncvs/ports/Mk/bsd.port.mk,v
retrieving revision 1.512
diff -u -r1.512 bsd.port.mk
--- bsd.port.mk	9 Jun 2005 20:39:43 -0000	1.512
+++ bsd.port.mk	11 Jun 2005 14:38:58 -0000
@@ -1396,6 +1402,11 @@
 .endif
 .endif
 .endif
+.if ${CFLAGS:M-O[23s]} != ""
+.if !defined(WITHOUT_NO_STRICT_ALIASING)
+CFLAGS+= -fno-strict-aliasing
+.endif
+.endif
 
 .if defined(NOPORTDOCS)
 PLIST_SUB+=	        PORTDOCS="@comment "