Re: Rebootstrap after adding some files...

[Ed Smith-Rowland <3dw4rd at verizon dot net>]

Paolo Carlini wrote:
> Ed Smith-Rowland wrote:
>
> > I got some warnings.  I don't know if that's normal or not:
> > MacOSX:~/gcc/libstdc++-v3 ed$ autoreconf
> > autoheader: warning: missing template: HAVE____BUILTIN_ABS
> > autoheader: Use AC_DEFINE([HAVE____BUILTIN_ABS], [], [Description])
> > autoheader: warning: missing template: HAVE____BUILTIN_COS
> > autoheader: warning: missing template: HAVE____BUILTIN_COSF
> > autoheader: warning: missing template: HAVE____BUILTIN_COSL
> > autoheader: warning: missing template: HAVE____BUILTIN_FABS
> > autoheader: warning: missing template: HAVE____BUILTIN_FABSF
> > autoheader: warning: missing template: HAVE____BUILTIN_FABSL
> > autoheader: warning: missing template: HAVE____BUILTIN_LABS
> > autoheader: warning: missing template: HAVE____BUILTIN_SIN
> > autoheader: warning: missing template: HAVE____BUILTIN_SINF
> > autoheader: warning: missing template: HAVE____BUILTIN_SINL
> > autoheader: warning: missing template: HAVE____BUILTIN_SQRT
> > autoheader: warning: missing template: HAVE____BUILTIN_SQRTF
> > autoheader: warning: missing template: HAVE____BUILTIN_SQRTL
> > autoreconf: /usr/local/bin/autoheader failed with exit status: 1
> >
> > I started the rebuild anyway so I guess we'll see.
>
> I hope the issue is already solved but I see a plain failure, not just 
> warnings and in that case for sure there are no hopes... Anyway: are 
> you using the right version of the tools, that is autoconf-2.59 and 
> automake-1.9.6 ??
>
> Paolo.
I basically got a whole new setup and started over again.
I'll see about the autotools versions.  Apple seems to get cranky about 
building newer versions of these IIRC.

Anyway, I was busy with work a lot so I didn't have time to work too 
much anyway.

But I *did* find a bug in GSL (that I need to tell them about) that was 
throwing off some of the test generation for the Laguerre polynomials.  
I compared against results from Mathematica and found out that if we 
(and GSL) use recursion for the ordinary laguerre(n, x) for largish n we 
should all be just fine.  The GSL routine punts on recursion for 
laguerre(n, a, x) for a = 0 and does a hypergeometric approximation that 
goes titsup for large n.  (I actually think having hypergeometric 
functions in a standard library is nuts but that is another missive.)

So I'll send out the function comparison and test generation toys next 
like I said I would.

Then I'll send out another patch that should work for most platforms 
Monday.  I'm using i686 cygwin or i686-linux because these seem to be 
the weakest links in terms of accuracy.

I think I need to explore this compile farm.

Ed