Re: [boost] Any sunpro experts out there?

From: Timo Geusch (timo_at_[hidden])
Date: 2006-06-17 13:07:09

On Sat, 2006-06-17 at 01:20 +0100, John Maddock wrote:
> Timo Geusch wrote:
> > Unfortunately it doesn't so it's not quite that easy a fix. Several
> > tests don't compile and a couple of those that do, fail.
> >
> > I've put up the output of a test run at
> > http://www.unixconsult.co.uk/programming/type_traits_test_sunpro.out.
> > Please note that these currently are just the build failures as I
> > thought that it may be better to fix those first and then address the
> > test failures seperately.
> 
> Actually I'm less worried about those that don't compile, than those that 
> compile and do the wrong thing: for example is_empty<function-type>::value 
> being *true*, that could be really really bad, and I'm frankly at a complete 
> loss to understand how that could possibly happen.  Other failures like the 
> is_abstract ones are expected if BOOST_NO_IS_ABSTRACT is set: which it is 
> for compiler versions <= 0x580, so it looks like maybe the compiler version 
> hasn't been updated from previous releases?

I think this is partially my fault - I believe that Simon Atanasyan also
patched the compiler configuration header file and I didn't apply those
changes. That said, the difference between 1.33.1 with his source
patches applied and 1.34 RC with no additional patches is one more
failure against 1.34. I'll rerun the tests with his changes applied and
will report the results.

>From reading Simon's blog I'm also under the impression that the
compiler which succeeds with in type_traits tests is their repository
version of the compiler and not the -02 patch version, but I may be
wrong on that count.

> The remaining failures are all 
> function type related: looks like sunpro doesn't like partial 
> specialisations on pointers to functions, or maybe something has gone wrong 
> before the metacode gets that far it's impossible to say.

There was a comment in his blog somewhere that it still has problems
with 'complex' template expressions. Maybe this one is complex enough.

> Anyway I'm going to contact Simon Atanasyan direct and see if he has any 
> suggestions.

I've seen your email and will add a bit more info for Simon once I'm
done enjoying the decidedly unBritish weather.

Regards,
Timo