[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lmi] rounding unit test under Linux/x86-64

From: Greg Chicares
Subject: Re: [lmi] rounding unit test under Linux/x86-64
Date: Fri, 17 Jun 2011 18:01:07 +0000
User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv: Gecko/20110414 Thunderbird/3.1.10

[oops, sent to VZ rather than to ML]

On 2011-06-16 16:41Z, Vadim Zeitlin wrote:
> On Thu, 16 Jun 2011 14:23:12 +0000 Greg Chicares <address@hidden> wrote:
> GC> On 2010-09-28 00:33Z, Vadim Zeitlin wrote:
> GC> > 
> GC> >  I finally have a sequence of patches which implement the proposed 
> changes.
> GC> > For now I've uploaded them at http://review.bakefile.org/r/231/
> GC> 
> GC> With this change, 'fenv_lmi_x86.hpp' loses the property tested by
> GC> 'make check_physical_closure' (and described in section 3.4 of
> GC> Lakos, _Large-Scale C++ Software Design_), and now this test:
> GC>   g++ -x c++ -w -O0 -fsyntax-only fenv_lmi_x86.hpp
> GC> produces these:
> GC>   /lmi/src/lmi/fenv_lmi_x86.hpp:135: error: `e_ieee754_precision' does 
> not name a type
> GC>   /lmi/src/lmi/fenv_lmi_x86.hpp:136: error: `e_ieee754_rounding' does not 
> name a type
> GC> and other consequent diagnostics.
>  Sorry, I wasn't aware of this test and so didn't run it. Do I understand
> correctly that I would have needed to run "make test_various_build_types"
> to detect this problem?

No, 'make check_physical_closure' suffices. That's one of the tests
I run all the time. A few months ago, I wrote documentation for the
things I habitually do; I should clean that up and post it here, now
that you're close to reproducing my working environment.

'test_various_build_types' is intended to be run as part of the
'nychthemeral_test' target. I'm not yet running that nychthemerally
because I haven't made time to figure out how to use 'cron' under
Cygwin, where no one seems to succeed the first time. But if that's
something you can do in your sleep (under Cygwin), please show me how.

> GC> Is there a good alternative to moving those enumerations from
> GC> 'fenv_lmi_x86.hpp' to 'fenv_lmi.hpp'?
>  The simplest would probably be to simply include fenv_lmi.hpp from
> fenv_lmi_x86.hpp, wouldn't it? While this is contrary to how it used be
> done, I don't see any serious problem with doing it.

I'll have a look; but right now, I've moved this aside to focus on
Vaclav's changes.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]