building on FreeBSD 10

Waitman Gobble uzimac at da3m0n8t3r.com
Tue Feb 14 22:55:02 UTC 2012


Waitman Gobble <uzimac at da3m0n8t3r.com> wrote ..
> JINMEI Tatuya <jinmei at isc.org> wrote ..
> > At Tue, 14 Feb 2012 00:15:11 -0800 (PST),
> > "Waitman Gobble" <uzimac at da3m0n8t3r.com> wrote:
> > 
> > > I have not yet submitted the bug report, I want to feel extremely confident
> that
> > I understand the bug so i'm doing some research.  It seems this strategy "removes"
> > extraneous warnings from boost :
> > > 
> > > 
> > > instead of 
> > > 
> > > #include <boost/numeric/conversion/cast.hpp>
> > > 
> > > change to
> > > 
> > > #include "hushup/cast.hpp"
> > > 
> > > then put cast.hpp in hushup directory:
> > > 
> > > #ifdef __GNUC__
> > > #pragma GCC system_header
> > > #endif
> > 
> > What kind of magic does this pragma do?  Do you have a reference?
> > 
> > At the moment, I'm personally not so sure if we go for it if the
> > problem only happens for a special FreeBSD version of g++.  Adding a
> > per-compiler pragam for a selected issue may not be harmful itself,
> > but it will make the code structure more complicated, and the indirect
> > dependency will probably annoy the current and future developers.
> > 
> > ---
> > JINMEI, Tatuya
> > Internet Systems Consortium, Inc.
> 
> Hi,
> 
> Sure, I totally agree that it should not be changed for a "special build", however
> according to the handbook the version of GCC has not 
> changed since FreeBSD 8.0, and I think the future plans will be to go with clang
> - however it is unclear when this will happen. 
> 
> I have a production server running CentOS 5.7 (Which is obviously dated, but still
> widely used in production use) .. gcc on that 
> machine is what shipped, 4.1, i'll try building on that and see if there are any
> issues.
> 
> 
> -- 
> Waitman Gobble
> San Jose California USA


It is clear that building on CentOS 5.7 is not possible with the distribution, however is a possibility with some customization and source installs outside of the available system packages. Python3.1+ is a big worry on CentOS because it can potentially break and obliterate the package manager if not done carefully.

Anywhow, perhaps the issue is truly a documentation issue, or best left that way, the bind10-guide.txt states GCC 3.4.3+ when I think maybe it is better to say 4.3.(?) + 
Perhaps a typo.. dunno.
Otherwise I don't see how it's going to happen to build boost with -Werror .. The bind10 code does not appear to produce warnings, it's a boost issue. The workaround mutes the warnings but there's the unknown problems that might happen with improper binary.
Thanks!'


-- 
Waitman Gobble
San Jose California USA



More information about the bind10-users mailing list