Re: [lmi] '-pedantic', '-std=c++98', and __STRICT_ANSI__

[Greg Chicares]

On 2005-11-2 17:24 UTC, Vadim Zeitlin wrote:
> On Wed, 02 Nov 2005 17:13:12 +0000 Greg Chicares <address@hidden> wrote:
> 
> GC> '-std=' flags actually do two things. First, they enforce actual
> GC> rules of a particular dialect
> ...
> GC> Second, they define the macro __STRICT_ANSI__; the gcc manual says
> ...
> GC> I wish they'd separate those two purposes,
> 
>  Me too. I totally agree with the first one being useful but the second one
> is not and platform_dependent.hpp workaround for it is IMHO actively ugly.
> 
> GC> because the second one creates some needless problems.
> 
>  So shouldn't we just #undef __STRICT_ANSI__ in config.hpp then instead of
> trying to pre-include all standard headers from platform_dependent.hpp?

That sounds like a great idea.

Here's how I'd like to proceed:

First, do you find any other diagnostics because of __STRICT_ANSI__?
The last one you mentioned led me to a defect elsewhere, which I was
glad to learn about (and fix). It seems possible that you might find
other defects that I never suspected.

Second, I'll try to add a test to the 'check_conformity' target that
will find defects like the one mentioned above, if I can find some
easy way to automate it.

Third, I'll investigate the idea you propose here. It sounds good,
but of course I want to test it carefully.

If no concrete, actual problems remain, then I have to think about
deferring the second and third steps. It's important to improve the
code, yet I'm under pressure to add new features, too. Our customers
don't understand that stabilizing and perfecting unit tests is the
best way for us to serve them.