[win32] Fix couple of bugs in signal unwinder

[Dave Korn]

On 31/03/2010 11:23, Eric Botcazou wrote:
>> Everything is obscured on Cygwin target by other problems, I'll have to
>> retest on MinGW. There will now be an interlude...
> 
> Yes, our Windows compiler is MinGW-based.  But it just occured to me that the 
> FSF Ada compiler is still configured for SJLJ exceptions:
> 
>    ZCX_By_Default            : constant Boolean := False;
> 
> in system-mingw.ads so my patch is actually a no-op for Ada. *sigh*

  Heh, and I suffered my usual teeth-gnashing experience trying to run
cross-configured (as compared to the system compiler) gnat tests.

  Well, I'm running the C,C++ and Java tests on the compilers I built anyway;
let's see if we can show it makes a difference somewhere, because it would be
surprising to me if a bug in the fallback unwinder didn't have nasty effects
somewhere.

    cheers,
      DaveK