Re: RFA: Avoid simulator timeouts on more libstdc++ tests

[Mark Mitchell <mark at codesourcery dot com>]

Paolo Carlini wrote:

> I was rebuilding with the last changes, and noticed only today (sorry!)
> a couple of things: the first one, really trivial, is that apparently
> you didn't change the Copyright years, adding 2009 (or 2009 *and* 2008.

I'm sorry; I'll add 2009 today.  I don't know of a rule about adding
years in which no changes were made, but I can certainly do that if
people think it's the standard practice.

> The second thing, a tad less trivial, is that for some of those
> testcases we have the wchar_t counterpart, which you didn't touch.
> Normally we keep those in sync, and I think we should do that here too,
> unless you have strong reasons to believe that simulators never enable
> the wchar_t support (in fact, that would surprise me)

I didn't notice them because the simulator I was using didn't have that
enabled, but I agree that it makes sense.  Is it pre-approved to make
the corresponding changes to the corresponding wchar_t tests?

Y'all should use templates to avoid this code duplication in the tests.
:-p :-)

More generally, I actually think "simulator" is the wrong predicate.  It
should be "slow target" (which doesn't exist as a predicate at present);
"simulator" is just a special case of that.  A little microcontroller
might well have the same problem.  We could honor a setting in the
DejaGNU board file, falling back to "simulator" if the board didn't
explicitly set the flag.

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713