develooper Front page | perl.perl5.porters | Postings from January 2007

Re: New release ?

Thread Previous | Thread Next
Marcus Holland-Moritz
January 7, 2007 15:49
Re: New release ?
Message ID:
On 2007-01-05, at 20:07:09 +0100, H.Merijn Brand wrote:

>   - initial value of NV_PRESERVES_UV_BITS
> As long as no-one explicitly tells me to alter the behaviour regarding
> the second item, I will take no action.

Currently, the default (/initial) value of NV_PRESERVES_UV_BITS
is literal 'undef', which causes compilation failure of sv.c,
despite Configure finishing successfully. So it's hard to make
it worse in this case by changing the default.

As Nicholas pointed out, zero would be a much better default:

> > b) Let Configure choose a safe minimum for nv_preserves_uv_bits
> >    and warn the user. Probably something like (4*$nvsize) or,
> >    to be perfectly safe, zero like in uconfig.h.  
> I'd go for 0. It generates valid C code, and I think that it's also
> semantically correct.

The test failures that were caused by setting NV_PRESERVES_UV_BITS
to zero should have been fixed by Sadahiro's patch, which has just
been applied (change #29693).

So, I'd vote for fixing this. I had already suggested this change:

--- Configure.orig      2007-01-02 11:11:59.000000000 +0100
+++ Configure   2007-01-02 19:18:48.000000000 +0100
@@ -15322,7 +15322,7 @@
 [1-9]*)        $echo "Your NVs can preserve only $nv_preserves_uv_bits bits of your UVs."  2>&1
        d_nv_preserves_uv="$undef" ;;
 *)     $echo "Can't figure out how many bits your NVs preserve." 2>&1
-       nv_preserves_uv_bits="$undef" ;;
+       nv_preserves_uv_bits="0" ;;
 $rm -f try.* try


Sometimes you have to make products just because they are cool.
            --Larry Augustin

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About