cvs commit: src/share/mk sys.mk

[Xin LI]

Andrey Chernov wrote:
> On Wed, Jun 27, 2007 at 11:39:34AM +0200, Harti Brandt wrote:
>> On Wed, 27 Jun 2007, Kostik Belousov wrote:
>>
>> KB>On Tue, Jun 26, 2007 at 07:10:21PM +0000, Andrey A. Chernov wrote:
>> KB>> ache        2007-06-26 19:10:21 UTC
>> KB>> 
>> KB>>   FreeBSD src repository
>> KB>> 
>> KB>>   Modified files:
>> KB>>     share/mk             sys.mk 
>> KB>>   Log:
>> KB>>   This is temp workaround of nasty gcc 4.2.0 -O2 bug which may skip the rest
>> KB>>   of the loop when arrays used inside.
>> KB>>   http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32500
>> KB>>   
>> KB>>   Approved by:    re (kensmith)
>> KB>
>> KB>This seems to break cross-build of HEAD on RELENG_6.
>>
>> And even building on a pre-gcc-4.2.0 CURRENT.
> 
> I just sent this cross-building compatible variant to re@ for approve:
> 
> --- sys.mk.orig	2007-06-27 14:04:15.000000000 +0400
> +++ sys.mk	2007-06-27 14:04:33.000000000 +0400
> @@ -35,7 +35,7 @@
>  CFLAGS		?=	-O
>  .else
>  CC		?=	cc
> -CFLAGS		?=	-O2 -fno-strict-aliasing -fno-tree-vrp -pipe
> +CFLAGS		?=	-O1 -fno-strict-aliasing -pipe
>  .endif

Nitpicking: I think -O1 implies no strict-aliasing.  So -O1 -pipe might
be just Ok.

Well, I'd say that all these changes looks scary to me.

Is there any code in our base system to trigger tree-vrp bug?  Do we
still have some time to have gcc fixed and tested rather than using
band-aid like this?  IMHO fixing gcc sounds better than "fix"ing sys.mk
if time permits us to fix and test a vendor solution.

Cheers,